Handbook of Isolation and Characterization of Impurities in Pharmaceuticals (SST) (Separation Science and Technology)

HANDBOOK OF ISOLATION AND CHARACTERIZATION OF IMPURITIES IN PHARMACEUTICALS

This is Volume 5 of SEPARATION SCIENCE AND TECHNOLOGY A reference series edited by Satinder Ahuja

HANDBOOK OF ISOLATION AND CHARACTERIZATION OF IMPURITIES IN PHARMACEUTICALS Edited by

Satinder Ahuja Ahuja Consulting Calabash, NC

Karen Mills Alsante Pfizer, Inc. Groton, CT

Amsterdam Boston London New York Oxford San Francisco Singapore Sydney Tokyo

Paris

San Diego

Academic Press An imprint of Elsevier Science 525 B Street, Suite 1900 San Diego, California 92101-4495

ß 2003 Elsevier Science (USA) All rights reserved. This work is protected under copyright by Elsevier Science, and the following terms and conditions apply to its use: Photocopying Single photocopies of single chapters may be made for personal use as allowed by national copyright laws. Permission of the Publisher and payment of a fee is required for all other photocopying, including multiple or systematic copying, copying for advertising or promotional purposes, resale, and all forms of document delivery. Special rates are available for educational institutions that wish to make photocopies for non-profit educational classroom use. Permissions may be sought directly from Elsevier’s Science & Technology Rights Department in Oxford, UK: phone: (+44) 1865 843830, fax: (+44) 1865 853333, e-mail: permissions@elsevier. com. You may also complete your request on-line via the Elsevier Science homepage (http://www.elsevier.com), by selecting ‘Customer Support’ and then ‘Obtaining Permissions’. In the USA, users may clear permissions and make payments through the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA; phone: (+1) (978) 7508400, fax: (+1) (978) 7504744, and in the UK through the Copyright Licensing Agency Rapid Clearance Service (CLARCS), 90 Tottenham Court Road, London W1P 0LP, UK; phone: (+44) 207 631 5555; fax: (+44) 207 631 5500. Other countries may have a local reprographic rights agency for payments. Derivative Works Tables of contents may be reproduced for internal circulation, but permission of Elsevier Science is required for external resale or distribution of such material. Permission of the Publisher is required for all other derivative works, including compilations and translations. Electronic Storage or Usage Permission of the Publisher is required to store or use electronically any material contained in this work, including any chapter or part of a chapter. Except as outlined above, no part of this work may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without prior written permission of the Publisher. Address permissions requests to: Elsevier’s Science & Technology Rights Department, at the phone, fax and e-mail addresses noted above. Notice No responsibility is assumed by the Publisher for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. Because of rapid advances in the medical sciences, in particular, independent verification of diagnoses and drug dosages should be made. British Library Cataloguing in Publication Data Handbook of isolation and characterization of impurities in pharmaceuticals. – (Separation science and technology; v. 5) 1. Drugs – Analysis 2. Contamination (Technology) I. Ahuja, Satinder, 1933 – II. Alsante, Karen Mills 615.10 901 ISBN 012044982X Library of Congress Cataloging-in-Publication Data Handbook of isolation and characterization of impurities in pharmaceuticals / edited by Satinder Ahuja, Karen Mills Alsante. p. ; cm – (Separation science and technology; v. 5) Includes index. ISBN 0-12-044982-X (alk. paper) 1. Drugs–Purification–Handbooks, manuals, etc. 2. Drugs–Separation–Handbooks, manuals, etc. I. Ahuja, Satinder, 1933 – II. Alsante, Karen Mills. III. Separation science and technology (San Diego, Calif.); v. 5. [DNLM: 1. Pharmaceutical Preparations–analysis–Handbooks. 2. Pharmaceutical Preparations–analysis–Laboratory Manuals. 3. Drug Contamination–prevention & control–Handbooks. 4. Drug Contamination–prevention & control– Laboratory Manuals. 5. Pharmaceutical Preparations–standards–Handbooks. 6. Pharmaceutical Preparations– standards–Laboratory Manuals. 7. Technology, Pharmaceutical–methods–Handbooks. 8. Technology, Pharmaceutical– methods–Laboratory Manuals. QV 25 H2364 2003] RS404.5 H355 2003 6150 .19–dc21 2003040309 First edition 2003 ISBN: 0-12-044982-X The paper used in this publication meets the requirements of ANSI/NISO Z39.48-1992 (Permanence of Paper). Printed in The United Kingdom.

CONTENTS

PREFACE xi CONTRIBUTORS

1.

xv

Overview: Isolation and Characterization of Impurities SATINDER AHUJA I. II. III. IV. V. VI. VII. VIII. IX.

2.

Introduction 1 Designations of Impurities 4 Regulatory Requirements 7 Sources of Impurities 8 Analytical Method Development Isolation Methods 18 Characterization Methods 20 Case Studies 22 Summary 24 References 24

14

Review of Regulatory Guidance on Impurities RADHIKA RAJAGOPALAN I. II. III. IV.

Introduction 27 Types of Impurities—Drug Substance 28 Role of Compendia 30 Role of Drug Master Files (DMF)—Type II and Impurities Evaluation 31 V. Reference Standards for the Quantitation of Impurities and Analytical Procedures 32

v

vi

CONTENTS

VI. VII. VIII. IX. X. XI.

3.

Qualification of Impurities and New Impurities 32 Impurities in Drug Products 33 Analytical Methodology for Impurities in Drug Product Impurities Quantitation Post-Approval 36 Role of Sponsors 36 Summary 36 References 37

33

Polymorphic and Solvatomorphic Impurities HARRYG. BRITTAIN AND ALES MEDEK I. II. III. IV. V.

4.

Introduction 39 X-ray Diffraction 40 Thermal Methods of Analysis 44 Vibrational Spectroscopy 49 Solid-State Nuclear Magnetic Resonance Spectrometry References 69

57

Impurities in Drug Products KENNETH C.WATERMAN, ROGER C. ADAMI, AND JINYANG HONG I. II. III. IV. V. VI. VII. VIII. IX. X. XI. XII.

5.

Introduction 75 Water 76 Peroxides 78 Aldehydes 79 Metal Impurities 80 Small Molecule Carboxylic Acids Leachables/Extractables 82 Alcohols as Impurities 83 Biological Impurities 83 Additives in Excipients 84 Final Observations 85 Summary 85 References 85

82

Strategies for Investigation and Control of Process- and Degradation-Related Impurities BERNARD A. OLSEN AND STEVEN W. BAERTSCHI I. II. III. IV. V.

Introduction 89 Goals and Strategies 91 Process-Related Impurities 95 Degradation-Related Impurities 102 Summary and Conclusions 115 References 116

vii

CONTENTS

6.

Reference Standards PAUL A. CULBERTAND BRUCE D. JOHNSON I. II. III. IV. V. VI.

7.

Introduction 119 Definitions 120 Life Cycle 121 Governance 125 Qualification Process Summary 139 References 139

127

Sample Selection for Analytical Method Development HUGH J. CLARKE AND KENNETH J. NORRIS I. Introduction 145 II. Components of the Key Predictive Sample Set (KPSS) 147 III. Stereoisomers 147 IV. Matrix Components 150 V. Process-Related Impurities (PRIs) 150 VI. Purposeful Degradation Samples 152 VII. Stability Samples 155 VIII. Phase-Solubility Analysis 156 IX. Sample Selection Strategies 159 X. Summary 162 Glossary 162 References 163

8.

Sample Preparation Methods for the Analysis of Pharmaceutical Materials DAVID T. ROSSI AND KENNETH G. MILLER I. II. III. IV. V. VI. VII. VIII.

Introduction 166 Solid-Phase Extraction (SPE) 166 Liquid–Liquid Extraction (LLE) 174 Supercritical Fluid Extraction (SFE) 181 Accelerated Solvent Extraction (ASE) 189 Centrifugation 194 Filtration 195 Summary 199 References 199

viii

CONTENTS

9.

Isolation Methods I: Thin-Layer Chromatography PAMELA M.GORMAN AND HONG JIANG I. II. III. IV.

Introduction to Thin-Layer Chromatography (TLC) 203 TLC Applications in Pharmaceutical Industry 206 TLC Method Development and Validation 207 Impurity Isolation and Characterization by TLC 221 References 228

10. Isolation Methods II: Column Chromatography MARKGUINN, RONALD BATES, BENJAMIN HRITZKO, TERI SHANKLIN, GLENN WILCOX, AND SAM GUHAN I. II. III. IV. V. VI. VII.

Introduction 231 Background 232 Stationary Phases 233 Equipment 237 Screening 240 Development of Preparative Method 244 Scaleup of Preparative Method 246 Summary 248 References 248

11. Mass Spectral Characterization DAVID J. BURINSKYAND FENG WANG I. Introduction 249 II. Relevance of Impurity Characterization 252 III. The coupling of Liquid-Phase Separations and Mass Spectrometers 259 IV. Ion Formation 264 V. Analyzers 273 VI. Ion Structure Interrogation 277 VII. Data Acquisition and Interpretation 282 VIII. Applications 286 IX. Conclusions 288 X. Summary 289 References 290

ix

CONTENTS

12. NMR Characterization of Impurities LINDA L. LOHR, ANDREW J. JENSEN, AND THOMAS R. SHARP I. II. III. IV. V. VI. VII. VIII. IX. X.

Introduction to Nuclear Magnetic Resonance (NMR) Information Gathering 304 Sample Preparation for NMR 305 Sample Preparation for LC-NMR 307 NMR Instrumentation 309 NMR Experiments 314 Choosing an Experiment Set 324 Data Interpretation 325 Final Steps 334 Summary 336 References 337

301

13. Hyphenated CharacterizationTechniques THOMAS N. FEINBERG I. Introduction 341 II. Experimental Examples III. Conclusions 356 References 357

350

14. Solving Impurity/Degradation Problems: Case Studies KAREN M. ALSANTE, TODD D. HATAJIK, LINDA L. LOHR, DINOS SANTAFIANOS, AND THOMAS R. SHARP I. Introduction and Background 361 II. Case Studies 368 III. Summary and Conclusions 398 Appendix—Lessons Learned 398 References 399

Index 401

This Page Intentionally Left Blank

PREFACE

The pharmaceutical industry is required by the Food, Drug, and Cosmetic Act to establish the identity and purity of all marketed drug products. The United States Food and Drug Administration (FDA) and other regulatory bodies around the world require that impurities in drug substance and drug product when present at threshold levels recommended by the International Conference on Harmonisation (ICH) be isolated and characterized. This book fills the need for a text on the complex process of isolation and characterization of process-related (synthesis and formulation) impurities and degradation products to meet critical regulatory requirements. The identification of process-related impurities and degradation products can provide an understanding on production of impurities and define degradation mechanisms. When this process is performed at an early stage of drug development, there is ample time to address various aspects of drug development to prevent or control the production of impurities and degradation products well before the regulatory filing and thus assure production of a high-quality drug product. The chapters in this book have been organized in a logical sequence to reflect the process used for the isolation and characterization of impurities. Chapter 1 points out that there are ethical, economic, and persuasive regulatory reasons to isolate and characterize impurities and degradation products. It provides an understanding of various sources of impurities and degradation products, the process and methodologies involved in separation, isolation, and characterization of impurities. Chiral impurities are also discussed from the standpoint of their origin, analytical methodology, and regulatory perspective for controlling them. Regulatory guidance is provided in Chapter 2, which includes a significant discussion on ICH guidelines. Thresholds for identification, safely qualification, and reporting impurities have been set in these guidelines with

xi

xii

PREFACE

specific action levels reflective of the proposed daily dose of drugs under development. The rule of thumb for identifying impurities in drug substance and drug product is 0.1% (depending on the daily dose). Identification and control of toxic impurities present even well below the 0.1% level may be required. Chapter 3 discusses polymorphic and solvatomorphic impurities. Since a major pharmaceutical manufacturing goal is to produce drug substance that is phase-pure and stable, the question of small amounts of polymorphic and solvatomorphic impurities in a bulk solid is of great importance. The drug substance should also remain phase-pure during drug product formulation and shelf life. Chapter 4 specifically addresses impurities in drug product originating from formulation ingredients and processing that are likely to cause stability or performance issues in the drug product. The sources of the impurities as well as resulting drug stability issues are outlined in this chapter. Chapter 5 describes strategies for investigating process-related and degradation-related impurities in drug substances and drug products, with emphasis on a ‘‘chemistry-guided approach.’’ A critical component in the analytical quantification of impurity levels is reference standard materials. Chapter 6 provides useful information on the role of reference standards in monitoring impurities. It includes the qualification process and governance of reference standards. Chapter 7 reviews analytical method development for the quantification of impurities and degradation products present in drug substance and drug product. This process involves selecting the key impurities/degradation product sample set, screening of chromatographic conditions, and optimizing method parameters. To assure the high quality of pharmaceutical products, it is of critical importance to carry out elucidation of structure of impurities and degradation products present in the drug substance and drug product throughout the drug development process. For low-level impurities/ degradation products, this quite often involves isolation. The next three chapters detail the isolation of impurities and degradation products. Chapter 8 provides guidance on extraction and isolation techniques for successful sample preparation including specificity for the targeted material, homogeneity, and good recovery. An excellent isolation technique is exemplified by thin-layer chromatography (TLC), discussed in Chapter 9. TLC is particularly useful when highperformance liquid chromatography (HPLC) fails to yield useful information because of retention on the head of column, early elution, or poor detection issues. It can be easily scaled up for preparative work. The resolving power of HPLC is frequently needed for challenging isolation problems. Chapter 10 details the use of HPLC for the isolation of impurities and covers the various options that are available for stationary phases, detectors, and the preparative scaleup process. Structure elucidation of impurities and degradation products at trace levels in complex matrices requires advanced instrumental techniques and collaborative efforts of scientists from various disciplines. Chapter 11 describes the fundamental of mass spectrometry-based techniques for ion

PREFACE

xiii structure analysis, including aspects of ion formation, attributes of various mass analyzers, and scan modes used for collision-induced dissociation experiments and interpretation of mass spectra. It also discusses at some length LC-MS, the powerful combination of a versatile separation technique like HPLC with the universal detectability of MS. Chapter 12 focuses on nuclear magnetic resonance (NMR) spectroscopy, which provides key structural information and intramolecular interactions not readily available from other analytical methods. Vast improvements in NMR sensitivity limits have been made to assist with structural elucidation of impurities at low levels. Nondestructive NMR analysis allows additional characterization experiments to be performed with the same sample. Chapter 13 explains how hyphenated techniques have improved efficiency in structure elucidation of impurities and degradation products. Techniques discussed include HPLCDAD, LC-MS, GC-MS, LC-IR, and LC-NMR. Chapter 14 provides practical guidance with case studies on isolating and characterizing process-related impurities and degradation products for pharmaceutical drug candidates. The case studies utilize isolation or synthesis in conjunction with mass spectral and NMR characterizations. A collaborative multiple disciplinary strategy has been found to be the most efficient way to solve impurity/degradation product problems. We sincerely believe that the detailed information provided by all authors, actively working with the pharmaceutical industry and FDA, will be of great value to various readers who are interested in the isolation and characterization of impurities and degradation products. Satinder Ahuja Karen Mills Alsante

This Page Intentionally Left Blank

CONTRIBUTORS

Numbers in parentheses indicate the page on which the authors’ contributions begin.

Roger C. Adami (75) Pfizer Global Research and Development Division, Pharmaceutical Research and Development Department, Groton, CT 06340 Satinder Ahuja (1) Ahuja Consulting, 1061 Rutledge Court, Calabash, NC 28467 Karen M. Alsante (361) Pfizer Global Research and Development Division, Analytical Research and Development Department, Groton, CT 06340 Steven W. Baertschi (89) Eli Lilly and Company, Indianapolis, IN 46285 Ronald Bates (231) Pfizer Global Research and Development Division, Bioprocess Research and Development Department, Groton, CT 06340 Harry G. Brittain (39) Center for Pharmaceutical Physics, 10 Charles Road, Milford, NJ 08848 David J. Burinsky (249) Chemical Development Division, GlaxoSmithKline, Five Moore Drive, PO Box 13398, Mail stop 2-4075-4A, Research Triangle Park, NC 27709-3398 Hugh J. Clarke (145) Pfizer Global Research and Development Division, Analytical Research and Development Department, Groton, CT 06340 Paul A. Culbert (119) Lilly Analytical Research Laboratories, Eli Lilly Canada Inc., 3650 Danforth Ave., 3rd Floor, Bldg. 600, Scarborough, ON, MIN 2E8 Thomas N. Feinberg (341) Cardinal Health, Structural Chemistry Division, Magellan Pharmaceutical Development, P.O. Box 13341, Research Triangle Park, NC 27709 Pamela M. Gorman (203) Pfizer Global Research and Development Division, Analytical Research and Development Department, Groton, CT 06340

xv

xvi

CONTRIBUTORS

Sam Guhan (231) Pfizer Global Research and Development Division, Bioprocess Research & Development Department, Groton, CT 06340 Mark Guinn (231) Pfizer Global Research and Development Division, Bioprocess Research & Development Department, Groton, CT 06340 Todd D. Hatajik (361) Pfizer Global Research and Development Division, Analytical Research and Development Department, Groton, CT 06340 Jinyang Hong (75) Pfizer Global Research and Development Division, Pharmaceutical Research and Development Department, Groton, CT 06340 Benjamin Hritzko (231) Pfizer Global Research and Development Division, Bioprocess Research and Development Department, Gorton, CT 06340 Andrew J. Jensen (301) Pfizer Global Research and Development Division, Analytical Research and Development Department, Groton, CT 06340 Hong Jiang (203) Pfizer Global Research and Development Division, Analytical Research and Development Department, Groton CT 06340 Bruce D. Johnson (119) Pfizer Global Research and Development Division, Analytical Research and Development Department, Ann Arbor, MI 48105 Linda L. Lohr (301, 361) Pfizer Global Research and Development Division, Analytical Research and Development Department, Groton, CT 06340 Ales Medek (39) Pfizer Global Research and Development Division, Analytical Research and Development Department, Groton, CT 06340 Kenneth G. Miller (165) Pfizer Global Research and Development Division, Analytical Research and Development Department, Ann Arbor, MI 48105 Kenneth J. Norris (145) Pfizer Global Research and Development Division, Analytical Research and Development Department, Groton, CT 06340 Bernard A. Olsen (89) Eli Lilly and Company, 1650 Lilly Road, DC TL12 Lafayette, IN 47909 Radhika Rajagopalan (27) Center for Drug Evaluation and Research, FDA, 7500 Standish Place, Rockville, MD 20855 David T. Rossi (165) Pfizer Global Research and Development Division, Analytical Research and Development Department, Ann Arbor, MI 48105 Dinos Santafianos (361) Pfizer Global Research and Development Division, Analytical Research and Development Department, Groton, CT 06340 Teri Shanklin (231) Pfizer Global Research and Development Division, Bioprocess Research & Development Department, Groton, CT 06340 Thomas R. Sharp (301, 361) Pfizer Global Research and Development Division, Analytical Research and Development Department, Groton, CT 06340 Feng Wang (249) Chemical Development Division, GlaxoSmithKline, Five Moore Drive, PO Box 13398, Mail stop 2-4075-4A, Research Triangle Park, NC 27709-3398 Kenneth C. Waterman (75) Pfizer Global Research and Development Division, Pharmaceutical Research and Development Department, Groton, CT 06340 Glenn Wilcox (231) Pfizer Global Research and Development Division, Bioprocess Research and Development Department, Groton, CT 06340

1 OVERVIEW: ISOLATION AND CHARACTERIZATION OF IMPURITIES SATINDER AHUJA Ahuja Consulting,Calabash, NC 28467

I. INTRODUCTION A. Definitions B. The Need to Isolate and Characterize Impurities II. DESIGNATIONS OF IMPURITIES A. Common Names B. United States PharmacopeiaTerminology C. ICH Terminology III. REGULATORY REQUIREMENTS IV. SOURCES OF IMPURITIES A. Crystallization-Related Impurities B. Stereochemistry-Related Impurities C. Residual Solvents D. Synthetic Intermediates and By-products E. Formulation-Related Impurities F. Impurities Arising During Storage V. ANALYTICAL METHOD DEVELOPMENT A. The Role of Reference Standards B. Spectroscopic Methods C. Separation Methods VI. ISOLATION METHODS VII. CHARACTERIZATION METHODS VIII. CASE STUDIES IX. SUMMARY REFERENCES

I. INTRODUCTION Pharmaceutical analysts play a major role in isolation and characterization of impurities. The success in this endeavor requires a broad knowledge of a variety of fields of chemistry and excellent interactions with experts in various other disciplines.1,2

1

2

S. AHUJA

A. Definitions Let us first briefly review what constitutes an impurity in a drug substance (a new chemical entity of therapeutic interest) or drug product (a drug substance formulated into a suitable product for administration to patients) to enable us to better understand the need to isolate and characterize impurities. As recently as the 1990s, there was no precise definition for ‘‘impurity’’ in the pharmaceutical world, because of the apparent negativity attached to this word.3 The definition for impurity in Webster’s dictionary is ‘‘something that is impure or makes something else impure.’’ In the pharmaceutical industry, it is the latter meaning that is frequently attached to the meaning of an impurity. An impure substance is sometimes described as a substance of interest mixed or impregnated with an extraneous or usually inferior substance. A simple definition of impurity has been offered:2 An impurity is any material that affects the purity of the material of interest, viz., drug substance or drug product. The following definition of impurity is currently under consideration by the regulatory bodies, which is likely to be included in the future guidance:4 Impurity: any entity of the drug substance (bulk material) or drug product (final container product) that is not the chemical entity defined as the drug substance, an excipient, or other additives to the drug product.

This definition of impurity is broad enough to include degradation products as impurities. The term degradation product is defined in International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use (ICH; the American spelling for harmonization will be used from here on in this chapter) as follows:5 Degradation product: a molecule resulting from a change in the drug substance (bulk material) brought about over time. For the purpose of stability testing of the products in this guidance, such changes could occur as a result of processing or storage (e.g., deamidation, oxidation, aggregation, and proteolysis).

These definitions can be used as a guide in the search for a number of inorganic or organic compounds that can be found in a drug substance or a drug product until a finalized definition is available. At present, the impurities are given various names (see Section II); some of the terms, such as related compounds, may tend to soft-pedal them. In the pharmaceutical world, an impurity is generally considered as any other organic material besides the drug substance or active pharmaceutical ingredient (API) that arises out of synthesis. Most of the time, the inorganic contaminants are not given adequate consideration as impurities unless they are toxic, such as heavy metals or arsenic. Organic volatile impurities (OVI, which are generally made up of residual solvents as well as other organic volatile impurities used in the synthesis) are often considered virtual impurities. Interaction products produced during formulation processes and degradation products (frequently referred to colloquially as degradants in the

1 ISOLATION AND CHARACTERIZATION OF IMPURITIES

3

pharmaceutical industry; the terms have been used interchangeably in this text) that can be produced prior to use by the patient are additional sources of impurities. Section IV lists various sources of impurities in API or drug products. It is important to recognize at this stage that any material that leads to a decrease in the purity value of the API should be considered an impurity. Therefore, for all intents and purposes, various contaminants mentioned here can be called impurities and should be labeled as such because they decrease the purity of API.

B. The Need to Isolate and Characterize Impurities Impurities are generally assumed to be inferior to API because they might not have the same level of pharmacologic activity. However, they are not necessarily always inferior. From the standpoint of its usage, the drug substance is compromised in terms of purity even if it contains another material with superior pharmacologic or toxicologic properties. At first pass this may not be readily apparent; however, on further thought it will become clear that if we are to ensure that the accurate amount of the drug substance is being administered to the patient, we must assess its purity independent of the extraneous materials. Therefore, any extraneous material present in the drug substance or active ingredient must be considered an impurity even if it is totally inert or has superior pharmacologic properties, so that an appropriate evaluation of its content in the drug product can be made. The control of low-level impurities is of great importance when a drug is taken in large quantities; for example, the use of methotrexate (10–20 g) to treat neoplasia or the use of vitamins as a fad, notably vitamin C. Penicillins and cephalosporins have been known to sustain facile cleavage of the -lactam bond in aqueous solution. This is particularly interesting since some studies on penicillins have shown that their lack of stability may influence possible reactions involved in penicillin allergy.6 Special attention should be paid to the detection of DNA in all finished biotechnology products because DNA can be incorporated in the human genome and become a potential oncogene. It is expedient to exhibit the absence of DNA at the picogram-per-dose level to warrant the safety of biotechnology products.7 This book addresses isolation and characterization of a variety of materials mentioned in Section II, which can be simply called impurities that can affect the purity of API or can be harmful to patients. It is necessary to isolate and characterize a number of impurities and degradation products mentioned in Section II because it is not always possible to unambiguously characterize them with the widely used hyphenated methods that are frequently the first line of defense. These methods utilize detectors such as diode array UV detector (DAD), nuclear magnetic resonance spectrometer (NMR), and mass spectrometer (MS) with separation methods such as gas chromatography (GC), high-pressure or high-performance liquid chromatography (HPLC), supercritical fluid chromatography (SFC), and capillary electrophoresis (CE). Methods such as gas chromatography-mass spectrometry (GC/MS), high-pressure or high-performance liquid chromatography–mass spectrometry (HPLC/MS, or simply

4

S. AHUJA

LC/MS), GC or LC/MS/MS, LC/DAD/MS, LC/NMR, LC/DAD/MS/NMR, SFC/MS, or CE/MS and various other combinations are extremely useful and can shorten the time needed for characterization of impurities. These methods have been discussed at length (Chapters 11–14 in this book; also see references 1 and 2). It should be apparent from the preceding discussion that complete characterization of the impurities or evaluation of their toxicologic properties generally requires that the impurities be isolated in sufficient quantities or be synthesized.

II. DESIGNATIONS OF IMPURITIES Impurities have been named differently by various groups of scientists who deal with them. Described here are commonly used terms and those terms that are used by official bodies such as compendia or that have been found acceptable by ICH and various regulatory bodies.

A. Common Names Various terms that have been commonly used to describe impurities are listed alphabetically below. . . . . . . .

By-product Degradation product Interaction product Intermediate Penultimate intermediate Related product Transformation product

Some of these terms indicate potential sources of impurities; e.g., intermediates; others tend to downplay the negativity, as exemplified by the use of the term ‘‘related product.’’ By-products: The unplanned compounds generated in the reaction to produce API are generally called by-products. Because it might not be possible to theorize all of them, they present a significant challenge to the analytical chemist. Degradation products: The compounds produced as a result of decomposition of the material of interest or API are often called degradation products (or degradants). It is necessary to be concerned with these products as well as those brought about by degradation of other compounds that may also be present as impurities in the drug substance. Interaction products: This term is slightly more inclusive and more difficult to evaluate than the two previously described, i.e., by-products and degradation products, in that it takes into account interactions that could possibly occur between various involved chemicals—purposely or inadvertently.

1 ISOLATION AND CHARACTERIZATION OF IMPURITIES

5

Intermediates: The planned compounds produced during synthesis of the desired substance are called intermediates, especially if they have been isolated and characterized. The most important requirements are isolation and characterization, i.e., they cannot be just potential reaction products that may be produced theoretically (see by-products). The theorized products are best designated as ‘‘potential intermediates.’’ Penultimate intermediate: As the name implies, this is the last compound in the synthetic chain just preceding the production of the ultimate desired compound. Confusion sometimes occurs when the desired material is a salt of a free base or acid. It is not appropriate to label the free acid or base as the penultimate intermediate if the drug substance is a salt. Related products: As suggested previously, the term ‘‘related products’’ tends to imply that the impurity is similar to the drug substance, and it thus tends to downplay the negativity frequently attached to the term ‘‘impurity.’’ These products may have similar chemical structures and potentially similar biological activities; however, we know that the structure alone does not provide any surety about biological activity. Transformation products: This is a relatively commonplace term that relates to theorized and non-theorized products that may be produced in the reaction. Transformation products are comparable to by-products, except that this term tends to imply that more is known about the reaction products.

B. United States Pharmacopeia Terminology The United States Pharmacopeia (USP)8 discusses impurities in various sections: . . .

Impurities in Official Articles Ordinary Impurities Organic Volatile Impurities

The pharmacopeia states that our notions about purity are likely to change with time and that purity is closely related to current developments in analytical chemistry. Therefore, what we regard as pure today may be considered impure at some future time if methods are found that can determine other components in a particular compound. Inorganic, organic, biochemical, isomeric, or polymeric components may all be considered impurities. The following terms have been used to describe impurities: . . . . . .

Concomitant components Foreign substances Ordinary impurities Organic volatile impurities Signal impurities Toxic impurities

Concomitant components: Bulk pharmaceutical chemicals (frequently referred to as API in the pharmaceutical industry) may have concomitant

6

S. AHUJA

components, e.g., geometric and optical isomers and antibiotics that are mixtures. Foreign substances: These are materials that are introduced by contamination or adulteration, and not as a result of formation or preparation; they are classified as foreign substances, e.g., pesticides in oral analgesics. Ordinary impurities: The types of impurities in bulk pharmaceutical chemicals that are harmless by virtue of having no serious undesirable biological activity in the amounts present are specified as ordinary impurities. Organic volatile impurities: This term describes the residual solvents that may be detected in the drug substance. Signal impurities: These are differentiated from ordinary impurities discussed earlier in that they call for individual identification and quantification by specific tests. These impurities include some process-related impurities or degradation products that provide a significant knowledge about the process. Toxic impurities: These impurities have important undesirable biological activity, even as minor constituents, and demand specific identification and quantification by explicit tests. Compendial articles may contain additional inorganic impurities. These impurities may be as common as table salt or a variety of other compounds that are regulated at low levels, such as heavy metals, arsenic, and so forth.

C. ICH Terminology According to ICH guidelines (on the Internet, see http://www.fda.gov/ cder/guidance), impurities can be broadly classified into the following three categories for the drug substance produced by chemical synthesis: . . .

Organic impurities (starting materials, process-related products, intermediates, and degradation products). Inorganic impurities (salts, catalysts, ligands, and heavy metals or other residual metals). Residual solvents (organic and inorganic liquids used during production and/or recrystallization).

The International Conference on Harmonization addresses the questions relating to impurities as follows: . . . . .

Q1A(R) Stability testing of new drug substances and products Q3A(R) Impurities in drug substances Q3B Impurities in drug products Q3C Impurities: residual solvents Q6A Specifications: test procedures and acceptance criteria for new drug substances and new drug products: chemical substances

It should be noted that none of the terminologies given above adequately highlights polymorphic, solvatomorphic, and chiral impurities (see Section III).

1 ISOLATION AND CHARACTERIZATION OF IMPURITIES

7

III. REGULATORY REQUIREMENTS Ethical, economic, and competitive reasons, as well as those of safety and efficacy, support the need to monitor impurities in drug products.9 However, monitoring impurities and controlling these impurities mean different things to different people or to the same people at different times, even those in the pharmaceutical sciences and industry.2 A unified terminology is necessary to assure that everyone uses the same vocabulary when addressing questions related to impurities. In this context, the leadership provided by ICH is very helpful. A number of requirements have an effect on monitoring impurities (see Chapter 2). For example, a country’s pharmacopeia or the one accepted by that country often provides the primary guidance as to how impurities are to be monitored and regulated. In a majority of countries these pharmacopeias are run under the auspices of the government. The USP is a notable exception to this case. If a product is considered a pharmacopeial item, it must meet the compendial requirements. The United States Food and Drug Administration (FDA) has endorsed the guidance prepared under the auspices of the ICH. The guidance, developed with the joint efforts of regulators and industry representatives from the European Union, Japan, and the United States, has helped ensure that the different regions have consistent requirements for the data that should be submitted to the various regulatory agencies. For the threshold of various impurities allowed in drug substances and drug products, see Chapter 2. The guidelines not only aid the sponsors of New Drug Applications (NDA) or Abbreviated New Drug Applications (ANDA) with the type of information that should be submitted with their applications, but also assist the FDA reviewers and field investigators in their consistent interpretation and implementation of regulations. In the United States, the federal FD&C Act and its amendments require that a manufacturer demonstrate the safety and efficacy of a new drug prior to introducing it into interstate commerce. The requirements are clearly spelled out in the Notice of Claimed Investigational Exemption for a New Drug (IND) and the NDA. The FDA also provides the following guidance on impurities: . .

NDAs: Impurities in drug substances ANDAs: Impurities in drug substances

INDs require ‘‘a statement of the methods, facilities and controls used for the manufacturing, processing, and packing of the new drug to establish and maintain appropriate standards of identity, strength, quality, and purity as needed for safety and give significance to clinical investigation made with the drug.’’ NDAs demand more specific and explicit information, including stability studies to guarantee that the identity, quality, and purity of the drug product is maintained until its expiration date.

8

S. AHUJA

The FDA may initiate action under the Food and Drug Act to bring about removal of a product from the market, or, as is granted in the law, a manufacturer may voluntarily withdraw from the marketplace any batches that do not meet the approved specifications. In each country, regulatory authorities use their own standards for conducting clinical studies on a new drug product, i.e., a drug substance that has been formulated or the product that is approved for commerce. Specific comments may be found in Chapter 2 in Reference 2 as to how regulations vary from country to country and the subsequent effect on international trade. Efforts are being made to unify these approaches, as demonstrated by ICH guidelines; nonetheless, there is still much to be done. ICH Guidelines on Impurities in New Drug Substances have the following requirements: The studies conducted to characterize the structure of actual impurities present in the new drug substance at a level greater than 0.1% (depending on the daily dose, calculated using the response factor of the drug substance) should be described. Similarly, ICH Guidelines on Impurities in New Drug Products require that the degradation products observed in stability studies conducted at recommended storage conditions be identified when present at a level greater than the identification thresholds (1% for a maximum daily dose of < 1 mg to 0.1% for a maximum daily dose of > 2 g).

IV. SOURCES OF IMPURITIES From the preceding discussion, it should be clear that impurities can originate from various sources. The most obvious source of impurities is the synthesis, where intermediates and by-products may be carried into the API as impurities or become a source of other impurities resulting from them. Any impurity that may be present in the starting material has the potential to be carried into the active ingredient of interest. Furthermore, the impurities that relate to inert ingredients (excipients) and solvents used during synthesis must also be considered. Impurities can be produced during various drug product formulation steps. These impurities have the possibility of being present in the final drug product. Potential reaction products relating to these impurities must also be evaluated. The impurities in drug products can be attributed not only to the drug substance or inert ingredients used for formulating a drug product; they can be brought into the drug product through the formulation process or by contact with packaging (see Chapter 4). Of the various impurities that can be found in drug products, there are relatively few that can influence the stability or performance of the dosage form. It is important to give greater consideration to these detrimental impurities. In general, most of these impurities are small molecules. This is especially true in solid dosage forms where the limited mobility restricts the reactivity of larger molecules. For most drugs, the reactive species consists of water (which can hydrolyze some drugs or affect the dosage form performance), small electrophiles (e.g., aldehydes and carboxylic acid derivatives), peroxides (which can oxidize

1 ISOLATION AND CHARACTERIZATION OF IMPURITIES

9

some drugs), and metals (which can catalyze oxidation and other drug degradation pathways). All of these impurities are discussed in Chapter 4, which focuses on how these impurities are likely to alter dosage forms through chemical reactivity and physical changes to the systems. Additionally, some impurities can cause toxicological problems. These impurities may not directly affect the performance or stability of a dosage form, but must be controlled to make a safe drug product.

A. Crystallization-Related Impurities Polymorphism is the term used to denote crystal systems where a substance can exist in different crystal packing arrangements, all of which have the same elemental composition (see Chapter 3). It is also possible to have crystal systems where the substance exists in different crystal packing arrangements, each of which has a different elemental composition; this phenomenon is known as solvatomorphism. Based on the realization that the nature of the structure adopted by a given compound upon crystallization could exert a profound effect on the solid-state properties of that system, the pharmaceutical industry is required by regulatory authorities to take a strong interest in polymorphism and solvatomorphism. The nature of the crystal structure of a given material can influence the following properties: . . . . . . . . . . . . . . . . . . . .

Conductivity Crystal hardness Crystal shape and color Density Diffusivity Dissolution rate Electrolytic conductivity Enthalpy of transitions Heat capacity Heat of solution Hygroscopicity Latent heat of fusion Melting or sublimation properties Phase diagrams Rates of reactions Refractive index Solubility Surface tension Viscosity Volume

It is usually the goal in pharmaceutical manufacturing to produce a drug substance that is phase-pure and remains in that state as long as the bulk material is stored. Another goal is to formulate the drug substance in a manner so that it remains in the same phase-pure state during the manufacture of the drug product and during any subsequent storage. This requires the

10

S. AHUJA

development and validation of assay methodology for the determination of phase composition. In Chapter 3, the question of small amounts of polymorphic and solvatomorphic impurities in a bulk solid is addressed. Methods such as X-ray diffraction, thermal methods, vibrational spectroscopy, and solid-state NMR are useful for performing the required studies. These methods are outlined and illustrated with appropriate examples.

B. Stereochemistry-Related Impurities It is of paramount importance to look for stereochemistry-related compounds, i.e., those compounds that have similar chemical structure but different spatial orientation. These compounds can be considered impurities in the API. Included in this group are various stereoisomers. The simplest case of chirality can be seen in a molecule that has one or more tetrahedral carbons with four different substituents (asymmetric carbon atom) such that its mirror image is not superimposable. Chiral molecules may also occur for a number of other reasons and must be factored into any evaluation of impurities.10–12 Stereoisomerism is possible in molecules that have any of the following characteristics: . . . . . .

One or more center of chirality Helicity (e.g., helical nature of tertiary structure of proteins, polysaccharides, and nucleotides) Planar chirality (e.g., polycyclophanes) Axial chirality (e.g., spiranes with cyclic skeleton) Torsional chirality (e.g., torsion about double or single bonds like cis and trans isomers and rotomers) Topological asymmetry (e.g., catenanes)

Chiral molecules are frequently called enantiomers. Enantiomers are optical isomers that have the same chemical structure but different spatial arrangement, which leads to different optical rotation. It is important not to overlook them because the d-isomer of a compound can have different pharmacologic or toxicologic activity from that of the l-isomer.11 Therefore, the undesired optical isomer is considered a chiral impurity of the API. Furthermore, it is important to remember that the number of chiral impurities increases with the increasing number of asymmetric carbon atoms in a molecule. The first set of guidelines regarding this issue was issued by the FDA in 1987 where the question of stereochemistry was approached directly on the manufacture of drug substances.13 The FD&C Act requires a full description of the methods used in the manufacture of the drug, which includes testing to demonstrate its identity, strength, quality, and purity. For chiral compounds, this includes identification of all chiral centers. The enantiomer ratio, although 50:50 by definition for a racemate, should be defined for any other admixture of stereoisomers. It is expected that the toxicity of impurities, degradation products, and residues from the manufacturing process be investigated as the development of the drug is pursued. The same standards

1 ISOLATION AND CHARACTERIZATION OF IMPURITIES

11

should, therefore, be applied to the racemates. For all practical purposes, optical isomers are considered impurities.

C. Residual Solvents Water is commonly present in drug products. As a result, water is by far the most commonly found volatile impurity in drug products, and most of the time it is not even considered an impurity (see Chapter 4). It is prevalent both in drug substances and excipients (water is present in a number of excipients as supplied by the vendor) and is used in dosage form preparations as well. Moisture content can be important when a dosage form is packaged such that equilibration with the environment does not occur. Under these conditions, the moisture brought into the system through the excipients can be sufficient to bring about hydrolysis. In addition, water from the environment can affect drug products and can very often be detrimental to their chemical stability or dosage form performance. The chemical stability problems relating to water are generally caused by hydrolysis, particularly of carboxylic acid derivatives (esters, amides), acetals, and other susceptible functionalities. Even in nonaqueous liquid formulations, water can be present in amounts sufficient to lead to drug degradation. For some drugs, the moisture level will determine if a hydrate can form. In some cases, hydrate formation or generation of an anhydrous form (by having a sufficiently low humidity level) can lead to a loss in crystallinity, which in turn may bring about greater chemical instability of the drug. In addition, changes in crystal morphology may affect changes in drug dissolution rates (see Chapter 3). A number of solvents that are used for the synthesis of the API or formulation of the drug product can be present in the drug product. The content of these solvents, which are commonly called organic volatile impurities (OVI), is generally determined by the OVI methods specified in the compendia. Their content is controlled by the guidelines offered by various bodies (see Chapter 2). Residual solvents can affect the stability of drug product (see Sections IV. D and E).

D. Synthetic Intermediates and By-products In addition to the residual solvents, polymorphic, solvatomorphic, and chiral impurities mentioned previously, impurities in a pharmaceutical compound or a new chemical entity (NCE) can originate during the synthetic process from raw materials, intermediates, and/or by-products. Raw materials are usually produced to lesser purity requirements than a drug substance. Therefore, it is easy to understand why they could include a number of components that in turn could have an impact on the purity of the drug substance. The solvents used in synthesis are also likely to involve a number of impurities that may extend from trace levels to critical quantities that can react with various chemicals used in the synthesis, to give rise to other

12

S. AHUJA

impurities. Intermediates are also not usually held to the same purity level as that of the drug substance—hence the observations made for the raw materials apply. By-products are often unknown and are very seldom controlled. So they too are a cause of concern in terms of monitoring impurities. ‘‘Pot reactions’’ (reactions where the intermediates are not isolated) are convenient, economical, and time saving; however, they raise havoc in terms of the generation of impurities because a number of reactions can occur simultaneously. The penultimate intermediate in the pharmaceutical synthesis is generally required to meet certain preset specifications. However, the more demanding standards of purity for the drug substance are very rarely exercised at this stage. It is important to bear in mind that this step is the last possible source of impurities from the synthesis. The methods used for analysis at this stage should be rigorous, and the tightest economically and practically feasible specifications should be applied.

E. Formulation-Related Impurities As mentioned before, many impurities in a drug product can originate from excipients used to formulate a drug substance. In addition, a drug substance is subjected to a variety of conditions in the process of formulation that can cause its degradation or have other undesirable reactions. For example, if heat is used for drying purposes or for some other reason, it can hasten degradation. Solutions and suspensions are inherently prone to degradation due to hydrolysis or solvolysis.2 These reactions may also take place in the dosage form in a solid state, such as in the case of capsules and tablets, when water or another solvent has been employed for granulation. The water used in the formulation can contribute not only its own impurities, but can also afford a perfect situation for hydrolysis and metal catalysis. Comparable reactions are conceivable in other solvents that may be used. If no precautions are taken, oxidation is quite possible for freely oxidized materials. In the same way, light-sensitive materials may sustain photochemical reactions. Details concerning the ways in which varied excipients may contribute to degradation and the resulting impurities may be found in Chapter 6 of Reference 2. In a number of cases, chemical stability issues associated with an excipient are due to impurities in the excipient rather than the excipient itself (see Chapter 4). The most reactive impurities tend to be small molecules. In addition to chemical stability, impurities in drug products can cause problems associated with toxicity or dosage form performance. It is often valuable to quantitatively determine the level of important impurities in drug products and to trace the origin of those impurities to their source. If the source is from an excipient, variability from lot to lot may make a marginal product unacceptable for reliability. If the source is related to the dosage form preparation process, it may be worthwhile considering technologies and processes that render less of the undesirable impurity.

1 ISOLATION AND CHARACTERIZATION OF IMPURITIES

13

F. Impurities Arising During Storage A number of impurities can originate during storage (shelf life) or shipment of drug products. It is essential to carry out stability studies to predict, evaluate, and ensure drug product safety.9 Stability, however, can have different meanings to different people, based on their discipline in the pharmaceutical sciences and industry. A variety of terms are currently used to encompass the what, how, and why of stability: kinetic study, compatibility study, stability evaluation, stability-indicating assay, expiration dating, outdating, shelf life, storage legend, preformulation study, failure of a batch to meet specifications, microbiological stability, stability of the active ingredient, stability of the formulation, stability in the marketed package, stability in the sample package, stability in the dispensing package, and stability in the hands of the consumer. All of these are considerations in stability of a drug product, and it is important to ensure that everyone understands the importance of stability studies. Chapter 8 in Reference 2 reviews the many facets of stability studies and describes what a modern stability program includes. Efforts have been made to relate scientific considerations with regulatory requirements. Degradation kinetics is discussed later in this section to highlight their importance in monitoring and evaluating impurities.2 The common goal for investigation of both process- and degradationrelated impurities is to determine which of the many potential impurities are, in fact, produced in the manufacturing process and which occur under a given set of storage conditions (see Chapter 5). During the development phase, this knowledge can be used to modify the process in an effort to eliminate or minimize levels of impurities. Similarly, the knowledge of stability characteristics can be used to store the drug substance and drug product in appropriate packaging and environmental conditions to minimize or eliminate unacceptable degradation. As mentioned previously, a better understanding of the degradation behavior of the drug and its interactions with excipients is also useful in developing the best formulation. The information obtained from impurity investigations can help establish specification tests and acceptance limits at various control points in the manufacturing process. 1. Degradation Kinetics The majority of the degradation reactions of pharmaceuticals take place at finite rates and are chemical in nature. Solvent, concentration of reactants, temperature, pH of the medium, radiation energy, and the presence of catalysts are important factors that affect these reactions. The order of the reaction is characterized by the manner in which the reaction rate depends on the reactant concentration. The degradation of most pharmaceuticals is classified as zero order, first order, or pseudo-first order, although the compounds may degrade by complicated mechanisms, and the true expression may be of higher order or be complex and noninteger. To ensure better stability predictions, an understanding of the limitations of experimentally obtained heat of activation values is critical. For instance,

14

S. AHUJA

the apparent heat of activation of a pH value where two or more mechanisms of degradation are involved is not inexorably constant with temperature. Furthermore, the ion product of water, pKw, is temperature-dependent, and Ha is approximately 12 kcal, an often overlooked factor that must be considered when calculating hydroxide concentration. Therefore, the heat of activation for all bimolecular rate constants involved in a rate–pH profile to predict degradation rates at all pH values for different temperatures must be obtained. Pharmaceutical scientists generally perform the desired kinetic studies to predict stability of a drug substance. It is important, however, to recognize the qualifying factors of such predictions. The significance of kinetic studies and the effect of a variety of additives on the reaction rates have been discussed at some length.2

V. ANALYTICAL METHOD DEVELOPMENT New drug development requires that meaningful and reliable analytical data be produced at various stages of the development.1,2,12,14,15 Assuring the safety of a new pharmaceutical compound or drug substance demands that the new drug substance meet the established purity standards as a chemical entity or when admixed with animal feeds for toxicity studies or when formulated with or without pharmaceutical excipients for human use. Furthermore, it should exhibit excellent stability throughout its shelf life. These requirements mandate that the analytical method(s) employed for this purpose should be sufficiently sensitive to measure low levels of impurities. This has resulted in development of analytical techniques that are appropriate for measurement of trace/ultratrace levels, i.e., sub-microgram quantities of a variety of chemical entities.15 Analytical method development for the quantification of low-level impurities present in pharmaceuticals can be thought of as a three-step process (see Chapter 7): 1. Sample set selection for analytical method development 2. Screening of chromatographic conditions and phases, typically using the linear-solvent-strength model of gradient elution 3. Optimization of the method to fine-tune parameters related to ruggedness and robustness. This can be accomplished using a factorial optimization approach. The key sample set selection for analytical method development has been discussed at length in Chapter 7. There are a great variety of methods used for monitoring impurities.1,2 The primary requirement for such techniques is the capacity to differentiate between the compounds of interest. This requirement frequently necessitates utilization of separation methods (covered in Section V. C) in combination with a variety of detectors (Section V. B). For gas chromatography, flame ionization and electron capture detectors are commonly used. However, these detectors are not suitable for isolation and characterization of impurities, which require

1 ISOLATION AND CHARACTERIZATION OF IMPURITIES

15

nondestructive detectors. Thermal conductivity detector fills that bill, but is generally not sufficiently sensitive. HPLC is the technique of choice for separation and analysis of pharmaceutical compounds because the compounds do not have to be volatile (or made volatile by derivatization), in contrast with gas chromatography, and thus they are not subjected to excessive heat. A number of spectroscopic detectors may be used with HPLC (see Section V. B). The spectroscopic methods may also be used independently for monitoring impurities; however, the combination of separation power of chromatography with spectroscopic detection provides very useful methods that provide high resolution of impurities. Electrochemical and other detectors are used where spectroscopic detection is not possible. Chiral Compounds Spectroscopy or chromatography can be used to monitor chiral compounds. Uniquely selective chromatographic approaches are often needed10–12; therefore, they have been highlighted here. The separations of chiral compounds discussed in various chapters of this book are often necessary because chiral pharmaceutically active moieties are increasingly common. It is imperative to determine the enantiomeric purity of chiral pharmaceutical compounds, including reference standards (see Chapter 6). In general, the minor enantiomer is considered an impurity. A chiral pair can be distinguished spectrally or chromatographically through the interaction with other chiral molecules to form either long-lived or transient diastereomers. The exploitation of the latter mechanism is the basis for the majority of enantiomeric chromatographic separations. The formation of a transient diastereomer is ordinarily performed either by using a chiral stationary phase or by adding a chiral selector to the mobile phase. Numerous chromatographic techniques, including those discussed in this book, have been demonstrated to achieve chiral separations through one of these routes. Chiral separation technology has now matured to the extent that quantification down to 0.1% of the minor enantiomer is commonly attained.

A. The Role of Reference Standards Practical guidance on pharmaceutical reference standards, which are important in all phases of drug discovery, development, and commercialization is provided in Chapter 6. The key objective of the chapter is to provide clarity to the overall life cycle and qualification and governance of reference standards used in the development and control of new drugs. Reference standards serve as the basis of evaluation of both process and product performance and are the benchmarks for assessment of drug safety for patient consumption. These standards are needed not only for the active ingredients in dosage forms but also for impurities, degradation products, starting materials, process intermediates, and excipients. The FDA has provided some guidance on the topic of reference standards (Chapter 2); some information also exists in the form of best practice documents from a variety of sources. However, the availability of reference standards and the degree to which

16

S. AHUJA

they are characterized and governed is often contingent on the stage of the drug development process. At any phase of drug development a reference standard should be assessed versus its intended use, and a balance should be struck between resource commitments, scientific judgment, and regulatory requirements.

B. Spectroscopic Methods The following spectroscopic measurement techniques have been used for characterizing impurities; most of these are very useful as detectors for chromatographic methods: . . . . .

Ultraviolet (UV) Infrared (IR) Raman spectroscopy Mass spectrometry (MS) Nuclear magnetic resonance (NMR)

Ultraviolet spectrophotometry (UV) at a single wavelength furnishes minimum selectivity of analysis; however, with the current accessibility of diode array detectors, it is conceivable to obtain sufficient simultaneous information at various wavelengths to assure greater reliability. Infrared spectrophotometry (IR) affords specific information on some functional groups that offer selectivity and allow quantification. However, low-level detectability is difficult. This requires more complex approaches, which are generally a deterrent to pharmaceutical analysts. Raman spectroscopy is based on the measurement of scattered electromagnetic radiation resulting from the irradiation of matter. Specifically, when a material is irradiated with a strong monochromatic light source (e.g., laser), a small amount of radiation is inelastically scattered at a wavelength different from the original incoming wavelength. It is this difference in vibrational energy between the scattered beam and incident beam that is measured. Raman spectroscopy is considered complementary to IR spectroscopy, as the two techniques provide a complete vibrational picture of a material. Raman spectroscopy is not as widely used for identification purposes as IR spectroscopy because of the relative complexity and the cost of instrumentation. However, it should be noted that Raman spectroscopy is an extremely powerful tool in characterizing the presence of polymorphs (see Chapter 3). Mass spectrometry (MS) provides excellent structural information, and, based on the resolution of the instrument, it may be an effective tool for differentiating molecules with small differences in molecular weight (see discussions in Section V. C and in Chapter 11). However, it has finite uses as a quantitative procedure. Nuclear magnetic resonance spectroscopy (NMR) provides reasonably detailed structural information on a molecule and is an extremely useful method for characterization of impurities (see discussions in Section VII and in Chapter 12); however, its use as a quantitative method is limited.

1 ISOLATION AND CHARACTERIZATION OF IMPURITIES

17

To summarize, UV detectors are commonly used for analyzing most pharmaceutical samples with HPLC. IR, NMR, and MS are excellent techniques for characterization of impurities that have been isolated by any of the techniques discussed here. The combination of MS and NMR with separation methods (discussed in Section VII and Chapters 11 and 12) provide excellent tools for characterization of impurities.

C. Separation Methods The following methods (listed in alphabetic order) can be used for separation of impurities and degradation products: . . . . . .

Capillary electrophoresis (CE) Chiral separations Gas chromatography (GC) High-pressure liquid chromatography (HPLC) Supercritical fluid chromatography (SFC) Thin-layer chromatography (TLC)

The nature and complexity of the separation problem determines which method should be used. The primary goal of a good separation method is resolution of all impurities of interest. A brief account of the abovelisted methods is given here to provide a quick review of their potential use. Except for CE, all these methods are chromatographic methods. CE is an electrophoretic method that is frequently lumped with the chromatographic methods because it shares with chromatography many of the common requirements. However, it is not strictly a two-phase separation system— a primary requisite for chromatography. Capillary electrochromatography meets this requirement. Hyphenated methods such as GC-MS, LC-MS, GC-LC-MS, LC-MS-MS, etc. are all discussed throughout this book. Also, Chapter 13 has been especially planned for this purpose. Capillary electrophoresis is an effective technique in situations where very low quantities of samples are available and high resolution is essential. Its relatively lower reproducibility is the principal difficulty of this procedure. Gas chromatography is an extremely useful technique for quantification. It can afford the desired resolution, selectivity, and ease of quantification. The chief limitation, however, is that the sample must be volatile or must be made volatile by derivatization. This technique is very practical for organic volatile impurities (OVI). High-pressure liquid chromatography is often referred to as highperformance liquid chromatography today. Both terms can be abbreviated as HPLC, and the terms are used interchangeably by chromatographers. The applications of this very effective technique have been significantly expanded for the pharmaceutical chemist by the use of a variety of detectors such as fluorescence, electrometric, MS, and so forth. Supercritical fluid chromatography (SFC) offers some of the advantages of GC in terms of detection and of HPLC in terms of separations, in that volatility of the sample is not of paramount importance. The greatest

18

S. AHUJA

application of this technique has been found in the extraction of samples. SFC is generally performed in the normal phase (NP) mode, and often NP-TLC or NP-HPLC methods can be readily adapted to SFC methods. SFC generally provides an orthogonal separation method to traditional reversed phase HPLC. Because of the similarity to HPLC in the chromatographic measurement process, this technique can be used to accurately quantify nonpolar impurities of the sample of interest. Thin-layer chromatography coupled with densitometric detection is a highly sensitive method for quick assessment of the purity of various compounds including reference standards (see Chapter 6). High-performance TLC (HPTLC) is an improved version of TLC that uses stationary phases of decreased thickness and lower particle size, providing improved resolution over shorter elution distances. TLC can resolve a large range of compounds by employing a variety of different plates and mobile phases. Limited resolution, detection, and ease of quantification are the main problems associated with this method. The foremost advantages are ease of use and low cost.

VI. ISOLATION METHODS It is often necessary to isolate impurities because the instrumental methods that were mentioned earlier for directly characterizing impurities without isolating them are not available or when the authentic material is needed for further confirmation of the structure or its toxicity. Isolation entails removal of the compound of interest from the other compounds present in a mixture. Further purification is achieved based on the compound’s intended use. Chapter 8 provides practical guidance on the use of widely used extraction and isolation techniques from the sample preparation perspective. The first two sections, solid-phase extraction and liquid–liquid extraction deal with liquid samples. The sections on supercritical fluid extraction and accelerated solvent extraction focus mainly on solid samples while the centrifugation and filtration sections handle suspensions. A successful sample preparation protocol accounts for specificity and homogeneity as well as recovery and final physical state of the targeted material. The ultimate aim is to produce a sample that is compatible with the desired analytical technique to assure generation of maximum information. Isolation methods include both chromatographic and nonchromatographic methods. Simple methods should be tried first, as they can lead to considerable savings in time and can produce a larger quantity of materials with greater ease. For isolation of a given compound from a complex mixture, the chromatographic methods utilized for separation of impurities in analytical determinations are the methods of first choice that are suitably modified for the purpose of isolation of impurities where an appropriate fraction is collected. A list of the methods that can be used for isolation of impurities is given below. . .

Solid-phase extraction methods Liquid–liquid extraction methods

1 ISOLATION AND CHARACTERIZATION OF IMPURITIES . . . . . . . . .

19

Accelerated solvent extraction methods Supercritical fluid extraction Column chromatography Flash chromatography Thin-layer chromatography (TLC) Gas chromatography (GC) High-pressure liquid chromatography (HPLC) Capillary electrophoresis (CE) Supercritical fluid chromatography (SFC)

As mentioned earlier, it is desirable to initiate the isolation process on the basis of simple extraction or partition methods as exemplified by the first four methods listed above (see Chapter 8). Frequently, it is possible to extract impurities selectively on the basis of acidity, basicity, or neutrality of the compounds in question.2 The extraction process usually involves liquid– liquid extraction, where one phase is an aqueous solution and the other is a nonpolar organic phase. By appropriate adjustment of the pH of the aqueous solution, one can extract acidic, basic, or neutral impurities. These methods work well when a few impurities are present and their polarity or pKa is sufficiently different from that of API. If necessary, further separations can be achieved by chromatographic methods. It is well known that techniques such as chromatography can provide separation of compounds from each other and from the main compound. Separations are based on properties such as adsorption, partition, ion exchange, or molecular size. The choice of isolation method should be based on all available separation methods (see those listed here and in Section V. C). However, it should be pointed out that scaleup to preparative scale might be necessary. Also, special steps may have to be taken in some cases to circumvent particular problems. For example, it is not possible to utilize a destructive detector, such as a flame ionization detector, to isolate a component in GC. It would be necessary to replace it with a nondestructive detector, such as a thermal conductivity detector, or to use a split stream with accurate timing control that would allow collection of the desired separated component. The selection of an optimum method for isolation of impurities is dependent on a number of considerations, the foremost being the method that was initially used to find the impuritiy. If the method is not too cumbersome to carry out and the projected amount of the isolated impurity can be handled by the methodology in place, that method is obviously the preferred technique. If the amount of impurity needed for characterization, pharmacologic studies, or toxicologic studies is much greater, it may be necessary to rely on simple separation methods such as column chromatography, flash chromatography, or TLC. In this book special chapters have been included on TLC (see Chapter 9) and on column chromatographic methods including HPLC (see Chapter 10), as these methods are more commonly used. Because of its simplicity, flexibility, speed of analysis, and unique detection methods for both qualitative and quantitative analysis, TLC is

20

S. AHUJA

a valuable technique that is commonly used for separations of pharmaceutical compounds (see Chapter 9). Furthermore, it is complementary to commonly used HPLC methods; the majority of TLC separations are carried out in normal phase mode, whereas the method of choice in HPLC is frequently reversed phase. This generally entails selection of a suitable solvent system (mobile phase) for resolution of impurities on a silica gel plate. If the nature of impurities is not known, systems that are acidic, neutral, or basic are investigated. Universal detectors (UV or spray reagents) are preferred. Details on these methods and examples of sophisticated preparative separations are discussed in Chapter 9. Analytical chemists generally run a series of different assays (orthogonal assays) to confirm that they have sufficiently characterized the sample and have identified all impurity peaks found in the sample. Furthermore, they ensure that peaks which may co-elute using one technique can be detected with an orthogonal method (see Chapter 10). Analytical HPLC is the most common technology employed for assay and impurity profiling of pharmaceuticals. When the sample has been analyzed, the impurity may need to be isolated (for structural characterization, for instance). While techniques such as LC-MS may give an indication of the compound identification from the bulk assay, the definitive proof is always obtained from an independent analysis of the isolated compound. Because of the relatively low levels of the impurities (and also in many cases, their close resemblance to the main compound), conventional separation techniques are usually not successful. The resolving power of chromatography (and in many cases, HPLC) is needed for this isolation. Chapter 10 details the use of column chromatography for the isolation of impurities. It also discusses the various options that are available for stationary phases and analytical detectors, as well as the current equipment available for this work. The choice of purification techniques (prep HPLC, low-pressure silica columns, etc.) is also discussed. Finally, a procedure is described for both analytical methods development as well as scale-up to preparative columns. Capillary electrophoresis can provide only limited quantities of substances. However, this technique has been used for micro-preparative applications.16 Supercritical fluid chromatography as an isolation method is useful in those cases where it has been used as an analytical method for resolving impurities. It offers a great advantage in terms of removal of the mobile phase from the isolated fractions, since the mobile phase is mostly gaseous.

VII. CHARACTERIZATION METHODS When an impurity has been detected, it becomes necessary to estimate its content. Adequate detectability frequently means that a given component provides a signal at least twice that of background noise or baseline noise. At times, the multiple is set higher for greater assurance of detectability. Initial estimations are generally done against the parent compound because

1 ISOLATION AND CHARACTERIZATION OF IMPURITIES

21

in most cases the authentic sample of impurity is not available. It is important that the authentic sample be used for estimations when it is available. If the estimations indicate that a given impurity content is greater than 0.1%, it must be characterized according to FDA requirements. The ability of NMR to provide information regarding the specific bonding structure and stereochemistry of molecules of pharmaceutical interest has made it a powerful analytical tool for structure elucidation (see Chapter 12). Unfortunately, NMR has traditionally been sensitivity-limited compared to other analytical techniques. Conventional sample requirements for NMR are on the order of 10 mg, as compared with mass spectroscopy, for example, which requires less than 1 mg. Therefore, NMR spectroscopy historically has not been the first choice for an analytical chemist when identifying an unknown compound. Technological advancements in the field of magnetic resonance have made significant strides in improving sensitivity levels. This is particularly important in the structural characterization of drug impurities and degradation products, which often are available only in extremely limited quantities. The nondestructive, noninvasive nature of NMR spectroscopy makes it a valuable tool for the characterization of low-level impurities and degradation products. In addition, NMR can be considered close to a universal detector for hydrogen and carbon, as well as for other magnetically active nuclei. This is both good and bad because all signals are detected, those arising from the compound of interest and all other components in the sample, such as solvents and starting materials. Quantification is accurate over a dynamic range of nominally four orders of magnitude, although not as precisely as with other analytical tools, particularly at low levels. This poses a challenge for studying mixtures containing low-level components. It is therefore preferred, if possible, to isolate a given impurity prior to study by NMR rather than to analyze a mixture. This technique also will simplify spectral interpretation to a great extent. Mass spectrometry has had an increasingly significant impact on the pharmaceutical development process over the past several decades (see Chapter 11). Advances in the design and efficiency of the interfaces that directly connect separation techniques with mass spectrometers have afforded new opportunities for monitoring, characterizing, and quantifying drugrelated substances in active pharmaceutical ingredients and pharmaceutical formulations. Possessing exceptional analytical specificity and sensitivity, mass spectrometry significantly reduces the cycle time of chromatographic method development, validation, and sample analysis. The popularity of LC-MS-MS systems for complex mixture analysis of thermally labile, biologically relevant molecules is largely attributed to the ‘‘soft’’ nature of atmospheric pressure ionization techniques such as electrospray ionization (ESI), atmospheric pressure chemical ionization (APCI), and atmospheric pressure photoionization (APPI). Chapter 11 describes the fundamentals of LC-MS-based techniques for ion structure analysis, including aspects of ion formation in API sources, attributes of various mass analyzers and scan modes used for collision-induced dissociation experiments, and issues

22

S. AHUJA

surrounding the interpretation of the resulting mass spectra. Although the determination of chemical identity or molecular structure for related substances in pharmaceutical products has continuously benefited from the availability and evolution of modern instrumentation, fundamental knowledge about solution phase chemistry, ionization, and gas phase processes is still vitally important for achieving success in this endeavor. There is economic pressure to devise instrumentation methods that can deliver many data sets at once, thus shortening the time-to-market of drug products (see Chapter 13). Coupling of instruments has become easier with expanded capabilities, e.g., HPLC-DAD-MS (HPLC coupled with a diode array UV detector and a mass spectrometer), is almost routinely used. Nuclear magnetic resonance spectrometry has now been added to this combination to provide HPLC-DAD-NMR-MS capabilities in a commercial instrument. Another pressure on pharmaceutical scientists is the promise of biopharmaceuticals and high-potency active pharmaceutical ingredients. These compounds often have complicated impurity and degradation profiles at low absolute concentrations. Fortunately, instrument manufacturers have been quick to attempt to satisfy both these needs. The challenge now to the pharmaceutical scientist is the organization of the many data sets into presentation-quality formats so that scientists and managers can make correct decisions quickly. Software has been developed to speed the collation, analysis, and presentation of the many spectroscopic characterization techniques necessary. Detection limits for mass spectrometers are now approaching the zeptomole level, and NMR spectrometers have recently seen dramatic increases in sensitivity down to the nanogram level. Hyphenated methods such as gas chromatography-mass spectrometry (GC-MS) or liquid chromatography-mass spectrometry (LC-MS) and a number of other chromatographic and spectroscopic configurations are perfectly suitable for initial characterization of the impurities. Of course, these methods are not always applicable, especially when the authentic material is needed for purposes of structure confirmation, synthesis, or toxicity studies (this subject is discussed in various chapters of this book; also see reference 2).

VIII. CASE STUDIES Diverse applications have been sought in the areas of monitoring quality, stability, and safety of pharmaceutical compounds, whether made synthetically, extracted from natural products, or produced by recombinant methods. Many of these applications are included throughout this book to afford ready reference to various methods. In addition, the reader may want to refer to the scientific literature where various applications of monitoring impurities are generally classified in the following categories:2 . . .

Alkaloids Amines Amino acids

1 ISOLATION AND CHARACTERIZATION OF IMPURITIES . . . . . . . .

23

Analgesics Antibacterials Anticonvulsants, antidepressants, and tranquilizers Antineoplastic agents Local anesthetics Macromolecules Steroids Miscellaneous

This classification scheme has been preferably used over strictly chemistrybased schemes; however, the latter approach may be better suited for method development purposes. A number of case studies are included in this book. A brief review of the approach used for these studies follows (see Chapter 14). The process of identification of impurities and/or degradation products should begin early in drug development. It is imperative to seek participation of all parties familiar with the project. The project group should meet to assess the timelines for completion and to gather all pertinent information. The initial planning and discussions can save significant time in the experimental stage. The following questions should be raised at this early stage: . . . .

Is the compound of interest arising out of synthesis, formulation, or degradation? Is the impurity or degradation product present at significant levels? Is this a process- or degradation-related impurity, and under what conditions is it produced? Are enriched samples with the unknown impurity/degradation products available?

The next step is to determine at what level the unknown impurities are present. Identification of impurities below the 0.1% level is generally not necessary unless the potential impurities are expected to be unusually potent or toxic. Therefore, it is imperative to determine the level of the unknown impurity and/or degradation product early in the process. If the unknown is at or above the 0.1% limit, effort should be made to identify it. However, if the unknown is below the 0.1% threshold, a discussion should be conducted with the project team members to determine if isolation and identification are necessary. After a decision has been made to identify an unknown, the next logical step is to review all known process-related impurities, precursors, intermediates, and degradation products. A critical analysis of all available data can save considerable time and energy. By reviewing the HPLC retention data of all known process-related impurities, precursors, and intermediates (if available), it can quickly be determined whether the impurity of interest is truly unknown. If the retention time of the unknown impurity matches that of a standard, the unknown can be identified by using HPLC-DAD-MS. The identity can be confirmed by correlating the retention time, UV spectra, and mass spectra of the unknown impurity with that of the reference standard. Identifying an unknown by using a reference

24

S. AHUJA

standard as in the preceding description is a quick and easy process. However, other steps must be taken when the retention time of an unknown does not match that of a standard. The next step generally is to obtain molecular mass and fragmentation data via HPLC-MS. It is essential to determine the molecular mass of the unknown. Not only does the molecular mass help in the identification of the unknown, it also allows tracking the correct peak by HPLC when isolation becomes necessary. To run LC-MS, a mass spectrometry-compatible HPLC method is necessary. If such a method is not available, it has to be developed, which adds considerable time to the identification process. The decisions regarding the best approach to characterizing impurities and degradation products should be based on sound reasoning in order to minimize the time and cost of new drug development.

IX. SUMMARY A new drug development program should include a series of physicochemical tests to fully define the impurity profile of a pharmaceutical compound prior to performance of extensive pharmacologic and toxicologic studies. This is necessary to assure that observed toxicologic or pharmacologic profiles are in fact due to the compound of interest and not to impurities. ICH guidelines should be used to control impurities. Most importantly, the degree to which any impurity should be controlled ought to be dictated principally by its pharmacologic and toxicologic effects. Here it is necessary to include all impurities, those originating out of synthesis and those from other origins, such as degradation. The hyphenated methods should be used as the first line of defense for characterization of impurities. To obtain absolute confirmation of the structure of impurities or to conduct pharmacologic or toxicologic studies on impurities, it is frequently necessary to isolate and characterize them.

REFERENCES 1. Ahuja, S. and Scypinski, S., Handbook of Modern Pharmaceutical Analysis, Academic Press, NY, 2001. 2. Ahuja, S., Impurities Evaluation of Pharmaceuticals, Marcel Dekker, NY, 1998. 3. Ahuja, S., Eastern Analytical Symposium, November 16, 1995. 4. Ahuja, S., Personal communication, August 30, 2002. 5. International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use, see on the Internet, http://www.fda.gov/cder/guidance. 6. Yamana, T. and Tsuji, A., J. Pharm. Sci., 65:1563, 1976. 7. Bogdansky, F. M., Pharm. Technol., 72 Sept. 1987. 8. United States Pharmacopeia, Rockville, MD, 2000. 9. Mollica, J. A., Ahuja, S. and Cohen, J. J. Pharm. Sci., 67:443, 1978. 10. Ahuja, S., Chiral Separations by Liquid Chromatography, Am. Chem. Soc., Washington DC, 1991.

1 ISOLATION AND CHARACTERIZATION OF IMPURITIES

25

11. Ahuja, S., Chiral Separations: Applications and Technology, Am. Chem. Soc., Washington, DC, 1997. 12. Ahuja, S., Chiral Separations by Chromatography, Oxford University Press, NY, 2000. 13. Guidelines for Submitting Supporting Documentation in Drug Applications for the Manufacture of Drug Substances: Office of Drug Evaluation and Research (HFD-100), Food and Drug Administration, Rockville, MD, 1987. 14. Ahuja, S., Chromatography of Pharmaceuticals. Natural, Synthetic and Recombinant Products, ACS Symposium Series #512, Am. Chem. Soc., Washington, DC, 1992. 15. Ahuja, S., Trace and Ultratrace Analysis by HPLC, Wiley, New York, NY, 1992. 16. Altria, K. D., Isolation and Purification, 2:113, 1996.

This Page Intentionally Left Blank

2 REVIEW OF REGULATORY GUIDANCE ON IMPURITIES RADHIKA RAJAGOPALAN* Center for Drug Evaluation and Research, FDA, Rockville, MD 20855

I. INTRODUCTION II. TYPES OF IMPURITIESDRUG SUBSTANCE A. Foreign Substances and Other Impurities B. Toxic Impurities C. Ordinary Impurities III. ROLE OF COMPENDIA IV. ROLE OF DRUG MASTER FILES (DMF)TYPE II AND IMPURITIES EVALUATION V. REFERENCE STANDARDS FOR THE QUANTITATION OF IMPURITIES AND ANALYTICAL PROCEDURES VI. QUALIFICATION OF IMPURITIES AND NEW IMPURITIES VII. IMPURITIES IN DRUG PRODUCTS VIII. ANALYTICAL METHODOLOGY FOR IMPURITIES IN DRUG PRODUCT IX. IMPURITIES QUANTITATION POST-APPROVAL X. ROLE OF SPONSORS SUMMARY REFERENCES

I. INTRODUCTION The United States Food and Drug Administration (FDA) has endorsed the guidance prepared under the auspices of the International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use (ICH). The guidance, developed through the joint efforts of regulators and industry representatives from the European Union, Japan, and the United States, helps ensure that the different regions have consistent requirements for the data that should be submitted to the various

*The author is writing this article on her personal time, and the views expressed do not necessarily represent the views of the Agency or the United States.

27

28

R. RAJAGOPALAN

regulatory agencies. They not only aid the sponsors of New Drug Applications (NDA) with what information should be submitted with their applications, but also assist the FDA reviewers, Abbreviated New Drug Applications (ANDA) and field investigators in their consistent interpretation and implementation of regulations. In the area of impurities, both through ICH efforts, and by the efforts of the Agency in publishing Guidance, a tremendous amount of information is available. Technology advances have mandated that the global regulatory community, pharmacopeia, and the multinational sponsors of regulated products find common ground on implementing tighter limits and stringent controls on the production of drug substances and drug products. These trends have created a need to adopt consistent quality standards. This chapter will discuss the existing Guidance on drug substances and drug products, the role of compendia, Drug Master File requirements, and the role of sponsors in detail. The scope of this article is also limited to chemically synthesized materials and does not apply to biological/ biotechnological substances.

II. TYPES OF IMPURITIESDRUG SUBSTANCE Increasing the purity of a drug substance depends on lowering the level of impurities, not only at the time of release, but also maintaining low levels of degradants during the shelf life of the drug substance. Procedures are usually proposed to control synthetic precursors, synthesis-related materials, intermediates, heavy metals, moisture, and volatile solvents. Stabilityindicating methodology that can differentiate between active ingredient and degradation products, process impurities or other potential impurities is necessary to monitor the degradation profile over a long period of time. Starting the synthesis with high purity chemicals certainly adds to the purity profile of the finished product. Table 1 is a summary of the current Guidance that pertain to impurity information for NDA and ANDA products. In addition, a number of other Guidances are also in the draft stage.1 These Guidances are available on the Internet at http://www.fda.gov/cder/ guidance. In addition to the above, ICH Q6B, and Q5C are also available at the Web site, although they are not discussed here. Q6B and Q5C relate to biological and biotechnological products. ICH Guidance addresses impurities in drug substances to accomplish the following: . .

Classify, identify, qualify, set specifications, and discuss analytical procedures for impurities. Discuss the stability evaluation of the drug substance under long-term and accelerated conditions while taking into account the type of packaging materials used for storage and distribution.

It should be noted that the methodology used for identifying and quantifying impurities is covered by another ICH guidance2 for the validation

29

2 REVIEW OF REGULATORYGUIDANCE ON IMPURITIES

TABLE 1 Guidances on Impurities Title

Authorship

Q1A(R) Stability Testing of New Drug Substances and Products

ICH

Q3A(R) Impurities in Drug Substances

ICH

Q3B Impurities in Drug Products

ICH

Q3C Impurities: Residual Solvents

ICH

Q6A Specifications: Test Procedures and Acceptance Criteria for New Drug Substances and New Drug Products: Chemical Substances

ICH

NDAs: Impurities in Drug Substances

FDA

ANDAs: Impurities in Drug Substances

FDA

of chromatographic methods. Impurities can be broadly classified into the following three categories3 when the drug substance is chemically synthesized: . . .

Organic impurities (starting materials, process-related products, intermediates, and degradants) Inorganic impurities (salts, catalysts, ligands, and heavy metals or other residual metals) Residual solvents (organic and inorganic liquids used during production and/or recrystallization)

It is therefore necessary to evaluate impurities utilizing a variety of techniques and instruments, prior to assigning purity values to the drug substance. In addition to the above, the United States Pharmacopeia (USP) addresses impurities using additional terminology:4 foreign substances and other impurities, toxic impurities, and ordinary impurities.

A. Foreign Substances and Other Impurities Many USP monographs contain tests for specific impurities. However, USP recognizes that other impurities may come from a variety of situations, such as a change in processing or extraneous sources, and tests should be employed to detect these other impurities in addition to the specific tests provided in the individual monograph. The USP Committee of Revision recommends that whenever a new impurity is present consistently in quantities >0.1% and the monograph HPLC procedure does not detect the new impurity, a revision be proposed to include this substance in the monograph. USP further recommends that the sum of all impurities including the monograph-specified impurities should not exceed 2.0%.

30

R. RAJAGOPALAN

B. Toxic Impurities Substances known to be toxic can arise from synthesis or degradation of compendial articles and must not be listed under this category. The manufacturer should provide data supporting the classification and the limit for such impurities (refer to Section VI).

C. Ordinary Impurities These arise from the synthesis, preparation, or degradation of compendial articles. In general, USP provides tests to limit these impurities. However, since process-related impurities are different for varying processes, holders of Type II Drug Master File(s) and NDA/ANDA sponsors should address the impurities that are unique to their processes. The identification of impurities present at a level greater than the threshold is usually recommended for organic impurities. Normally 0.1% or 0.05% (calculated using the response factor of the drug substance, Table 3) is considered an acceptable threshold. The inorganic impurities usually can be quantitated using pharmacopeial or other appropriate procedures. ICH Q3C guides in determining, on a safety basis, acceptable residual solvent levels for intake by use of the term ‘‘permitted daily exposure’’ (PDE). This Guidance classifies residual solvents used in the synthesis and processing into four categories. The Guidance recommends that Class I solvents be avoided. These include benzene, carbon tetrachloride, 1,2-dichloromethane, 1,1-dichloroethane, and 1,1,1-trichloroethane. Table 2 is an example from the list of Class II solvents that should be limited because of their inherent toxicity either by calculation of concentration (PPM) or by PDE. This guidance defines class III solvents as having low toxic potential and a PDE of 50 mg or more per day and describes solvents for which no adequate toxicological data are found as class IV solvents.

III. ROLE OF COMPENDIA The USP is recognized as the official compendium by the FDA (21 U.S.C € 321(j)). Under General Notices in the USP 25, both foreign substances and TABLE 2 Common Organic Volatile Impurities (Class II) Organic volatile impurity

Limit (PPM)

PDE (mg/day)

Acetonitrile

410

4.1

Chloroform

60

0.6

600

6.0

80

0.8

1,4-Dioxane

380

3.8

Pyridine

200

2.0

Methylene chloride 1,1,2-Trichloroethene

2 REVIEW OF REGULATORYGUIDANCE ON IMPURITIES

31

other impurities are controlled. However, many USP monographs do not take into account multiple synthetic pathways or sources for drug substance and varying excipients or manufacturing processes for drug products. Nevertheless, for any drug article to meet identity, strength, quality, purity and potency attributes, the evaluation of impurities is a must. Often, there are no USP monographs for drug substances and drug products covered by INDs and NDAs. The sponsors of NDAs or ANDAs initiate USP monographs through articles in the Pharmacopeial Forum (PF), after approval of the application. The draft PF goes through a public comment period where FDA, industry sponsors and any other interested party may offer suggestions before the monograph is finalized for incorporation into the compendium. It should be recognized that despite this mechanism, many older monographs might remain unrevised for a long period despite the presence of test procedures using updated technology. An example of the lack of adequate control (for a drug substance) is an assay test by titration combined with a purity test by thin-layer chromatography (TLC). Updated procedures using, for example, high-performance liquid chromatography (HPLC) or other superior technology should be incorporated in the monograph. Regardless of the age of the USP monograph, the requirements in the ‘‘General Notices’’ section of the USP apply to the possibility of the presence of ordinary, toxic, and foreign substances or impurities. In general, evaluation beyond what is prescribed in the USP is needed by the sponsors of NDA/ANDA, and Type II Drug Master File (DMF) holders. In many cases, drug articles may be covered by a monograph in a foreign compendium (European Pharmacopeia, British Pharmacopeia, Japanese Pharmacopeia, etc.). Until such time as pharmacopeial harmonization across the regions occurs, the FDA does not accept the items covered by foreign pharmacopeia as official. However, the specifications (test procedures and acceptance criteria) cited in these monographs can be used as a secondary source of information.

IV. ROLE OF DRUG MASTER FILES (DMF)TYPE II AND IMPURITIES EVALUATION A Type II DMF may be filed for a drug substance (DS) or for a DS intermediate. For NDAs, drug substance information is usually filed in the NDA itself, instead of a DMF. The Agency reviews Drug Master Files only in connection with the evaluation of a filed IND, NDA, ANDA, or supplemental application when a letter authorizing such review is provided. There is no regulatory requirement that a DMF be submitted to support any of the above applications. The agency does not approve or deny approval to a DMF. The DMF holder is expected to follow the ICH and FDA guidance and ensuring that the different classes of impurities due to synthesis and degradation are addressed and adequately controlled in the drug substance. In addition, the DMF holder should address the stability of the drug substance and document it as per ICH guidance. Type II DMFs are to follow ICH conditions for ambient and accelerated testing and for the use of stability-indicating methods while testing. The amount of information

32

R. RAJAGOPALAN

expected in a DMF is the same as what is usually filed in a NDA for drug substance synthesis and characterization. The regulatory responsibilities of DMF holders are detailed in 21 CFR 314.420.

V. REFERENCE STANDARDS FOR THE QUANTITATION OF IMPURITIES AND ANALYTICAL PROCEDURES While compendial standards are available for some monographed article impurities, it may be difficult at times to obtain pure standards of impurities. Manufacturers of pharmaceuticals function as a potential source for obtaining reference standards of impurities, which may be synthesis precursors, process intermediates, or degradation products. The characterization and evaluation of these impurities reference standards should be constant with their intended use. In many cases, analytical procedures are developed and validated, where the response of an impurity is compared to that of the new drug substance itself. Response factor evaluation of impurities at the chosen detection wavelength is necessary to determine if a correction factor is needed (when the responses differ). Potentiometric detection, fluorescence/ chemiluminescence detection, and refractive index detection are some examples of detection modes available for compounds that may not be suitable for UV detection.

VI. QUALIFICATION OF IMPURITIES AND NEW IMPURITIES The profile of impurities in a new drug substance may change for a variety of reasons, such as process scale-up changes, synthetic route changes, and changes made to key intermediates. ICH decision trees help classify, qualify, and select limits for new molecular entities (NMEs). If an impurity exceeds the qualification threshold listed below in Table 3 (ICH Q3A(R)), studies are needed to qualify that impurity in drug substances. In many cases, studies performed to qualify an impurity will depend not only on the daily dose intake, but also on the patient population, route of administration, and duration of drug administration. In some cases,

TABLE 3 Thresholds Maximum daily dose1

Reporting threshold2,3

 2 g/day

> 2 g/day 1

Identification threshold3

Qualification threshold3

0.05%

0.10% or 1.0 mg per day intake (whichever is lower)

0.15% or 1.0 mg per day intake (whichever is lower)

0.03%

0.05%

0.05%

The amount of drug substance administered per day. Higher reporting thresholds should be scientifically justified. 3 Lower thresholds can be appropriate if the impurity is unusually toxic. 2

2 REVIEW OF REGULATORYGUIDANCE ON IMPURITIES

33

decreasing the level of impurity below the threshold (by optimizing the synthesis or manufacturing conditions), rather than providing additional studies and data, may be the simplest course of action. For ANDAs,5 comparison with innovator products, and evaluation using Quantitative Structure Activity Relationship (QSAR)6 models, may also be an acceptable way of evaluating impurities, in addition to in vitro genotoxicity studies. Adequate data may be available in the scientific literature to qualify an impurity. Additional safety testing becomes necessary when safety data from the literature is not available and decreasing the level of the impurity below the threshold is not possible. Figure 1 is an illustration of the decision tree for consideration of Safety Studies for drug substances in NDAs.

VII. IMPURITIES IN DRUG PRODUCTS ICH Guidance covers the degradation product(s) of drug substances (New Molecular Entities, NMEs) in the finished drug product and reaction products of the drug substance with an excipient and/or immediate container/ closure system. Excluded are extraneous contaminants (these are covered by good manufacturing practices, or GMP), and process impurities present in the drug substance. In general, sponsors who conduct stability studies should propose limits for observed impurities/degradation products based on their experience in analyzing representative pilot batches (at least two) and the third one can be smaller if justified. Figure 2 is a summary of identification and qualification thresholds for impurities in drug products, from the ICH Guidance on how to file information necessary to support the application. During NDA filing, the proposed limits should address . . .

Each specified degradation product Any unspecified degradation product Total degradation products

Adequacy of the proposed limits should be, where possible, evaluated using the principle of mass balance or material balance.

VIII. ANALYTICAL METHODOLOGY FOR IMPURITIES IN DRUG PRODUCT Many NDAs and ANDAs contain one assay method for the quantitation of the drug product and its impurities/degradation products. While consolidating the two tests can be advantageous to a Quality Control Laboratory, it should be noted that regulations do not limit the number of test methods that can be proposed for supporting an application. Indeed, the selection of one stability-indicating method for the drug product analysis with impurity quantitation capabilities can be challenging. The presence of excipients, the interaction of excipients with drug substance, and the drug substance’s degradation tendency make it difficult to devise an optimum

34

R. RAJAGOPALAN

FIGURE 1 Decision tree for identification and qualification. (a) If considered desirable, a minimum screen (e.g., genotoxic potential) should be conducted. A study to detect point mutations and one to detect chromosomal aberrations, both in vitro, are considered an appropriate minimum screen. (b) If general toxicity studies are desirable, one or more studies should be designed to allow comparison of unqualified to qualified material. The study duration should be based on available relevant information and performed in the species most likely to maximize the potential to detect the toxicity of an impurity. On a case-by-case basis, single-dose studies can be appropriate, especially for single-dose drugs. In general, a minimum duration of 14 days and a maximum duration of 90 days would be considered appropriate. (c) Lower thresholds can be appropriate if the impurity is unusually toxic. For example, do known safety data for this impurity or its structural class preclude human exposure at the concentration present?

35

2 REVIEW OF REGULATORYGUIDANCE ON IMPURITIES

THRESHOLDS FOR REPORTING OF DEGRADATION PRODUCTS IN NEW DRUG PRODUCTS Maximum Daily Dose1 1g >1g

Threshold3 0.1 % 0.05 %

THRESHOLDS FOR IDENTIFICATION OF DEGRADATION PRODUCTS IN NEW DRUG PRODUCTS Maximum Daily Dose1 < 1 mg 1 mg–10 mg 10 mg–2 g >2g

Threshold3 1.0% or 5 mg TDI2 whichever is lower 0.5% or 20 mg TDI whichever is lower 0.2% or 2 mg TDI whichever is lower 0.1%

THRESHOLDS FOR QUALIFICATION OF DEGRADATION PRODUCTS IN NEW DRUG PRODUCTS Maximum Daily Dose1 < 10 mg 10 mg–100 mg 100 mg–2 g >2g

Threshold3 1.0% or 5 mg TDI whichever is lower 0.5% or 20 mg TDI whichever is lower 0.2% or 2 mg TDI whichever is lower 0.1%

FIGURE 2 Description of different thresholds. 1The amount of drug substance administered per day. 2Total daily intake. 3Threshold is based on percent of the drug substance.

method, as is evidenced by the sheer number of high-performance liquid chromatography (HPLC) methods. HPLC is still the leading candidate for assay and impurity quantitation. Often, literature examples from other researchers may help in identifying suitable HPLC methods for drug product method development. When compendial methods or otherwise published methods are chosen for regulatory purposes, the method must still be shown to be suitable for the proposed use through validation. ICH Q2A and Q2B documents should be followed while validation is performed for quantitative impurities’ procedures. Often the inactive ingredients in dosage forms such as solutions, suspensions, and emulsions are themselves prone to degradation due to solvolysis and can contribute to impurities during shelf life evaluations. Impurity analytical profiles from scaled-up and laboratory scale batches should be compared with each other while finalizing NDA or ANDA specifications.

36

R. RAJAGOPALAN

IX. IMPURITIES QUANTITATION POST-APPROVAL Due to changes in the source of the drug substance, in the manufacturing process of the drug substance or drug product, impurity profiles can change for a marketed pharmaceutical product. NDA/ANDA impurity limits and test procedures should be reviewed and/or revised based on new findings in the form of supplemental applications to the Agency. Incorporation of the NDA specifications in a USP monograph can occur after the NDA is approved, to create a public standard for the drug substance and drug product. When this happens, the USP test methods and criteria will become official.

X. ROLE OF SPONSORS Pharmaceutical manufacturers ensure the safety and efficacy of drug products by coupling the tight control of raw materials and manufacturing processes with well-executed toxicological and clinical trials for NDAs. Setting the right specifications for impurities is an evolving process in any drug product’s life cycle. When an Investigational New Drug Application (IND) is filed, the proposed set of specifications for impurities is relatively tentative in nature. However, throughout Agency evaluation prior to NDA approval, the situation changes and additional well-defined controls and limits may be added both by the sponsors and as a result of regulatory review of the process. The establishment of a USP monograph also affects the limits set on the impurities. Safety and efficacy objectives regarding drug products are well recognized by pharmaceutical manufacturers. The manufacturers have been active in the development of the ICH Guidance and participate readily in the FDA’s Guidance development process. By providing valuable comments on draft Guidance (FDA and ICH), when solicited through Federal Register publications, the industry has assisted the Agency. Their interactions with the Agency to ensure the approval of new drug products in shorter time frames, and with the USP to develop official monographs, have benefited the American Public. It is hoped that regulators, compendia, and industry will continue to stay on course to bring high-quality drug products to market.

XI. SUMMARY The United States Food and Drug Administration (FDA) has been involved with the International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use (ICH). The guidance, developed through the joint efforts of regulators and industry representatives from the European Union, Japan and the United States, helps ensure that the different regions have consistent requirements for the data that should be submitted to the various agencies. This chapter reviews the existing information from the ICH Guidance, and compendia requirements with emphasis on impurities for the drug substance and the drug product.

2 REVIEW OF REGULATORYGUIDANCE ON IMPURITIES

37

REFERENCES 1. International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use, ICH Steering Committee, July 2000, Q3B(R) Impurities in New Drug Products, and ANDAs: Impurities in Drug Products, Federal Register, 64, 2 (1/5/ 1999) 516 2. International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use, ICH Steering Committee, March 30, 1995, Q2A-Text on Validation of Analytical Procedures, and Q2B- Validation of Analytical Procedures: Methodology 3. This classification does not apply to biological/biotechnological, peptide, oligonucleotide, radiopharmaceutical, fermentation and semisynthetic products derived therefrom, herbal products, and crude products of animal or plant origin. 4. United States Pharmacopeia, 25, US Pharmacopeial Convention, Rockville, MD 2000, 7, General Notices 5. ANDAs: Impurities in Drug Substances, Federal Register, 64, 232, (12/3/1999), 67917 6. Gombar, V.K. and Enslein, K. Quant. Struct.-Act. Relat. 9:321–325, 1990.

This Page Intentionally Left Blank

3 POLYMORPHIC AND SOLVATOMORPHIC IMPURITIES HARRYG. BRITTAINa AND ALES MEDEKb a

Center for Pharmaceutical Physics, Milford, NJ 08848

b

Pfizer, Inc.,Groton,CT 06340

1. INTRODUCTION II. X-RAY DIFFRACTION A. Single Crystal X-ray Diffraction B. X-ray Powder Diffraction III. THERMAL METHODS OF ANALYSIS A. Thermogravimetry B. Differential Scanning Calorimetry IV. VIBRATIONAL SPECTROSCOPY A. Infrared Absorption Spectroscopy B. Raman Spectroscopy C. Near-Infrared Spectroscopy V. SOLID-STATE NUCLEAR MAGNETIC RESONANCE (SS-NMR) SPECTROMETRY A. Solid-State NMR Techniques B. SS-NMR Studies of Polymorphism C. Quantitative Analysis of Polymorphs REFERENCES

1. INTRODUCTION It had been known since the eighteenth century that many substances could be obtained in more than one crystal form, and so the properties of these solids were studied to the fullest extent possible with the characterization tools (e.g., crystal morphology and melting phenomena) available at that time.1 Eventually the science of crystallography caught up with these studies, and quickly became the methodology of choice to study the various crystal forms of materials. The term polymorphism is the term used to denote crystal systems where a substance can exist in different crystal packing arrangements, all of which have the same elemental composition.2 It is also possible to have crystal systems where the substance exists in different crystal

39

40

H.G. BRITTAIN AND A. MEDEK

packing arrangements, each of which has a different elemental composition, and this phenomenon is known as solvatomorphism.3 These phenomena have been discussed in detail.4,5 The pharmaceutical industry has been required by regulatory authorities to take a strong interest in polymorphism and solvatomorphism once it was realized that the nature of the structure adopted by a given compound upon crystallization would then exert a profound effect on the solid-state properties of that system. For a given material, the heat capacity, conductivity, volume, density, viscosity, surface tension, diffusivity, crystal hardness, crystal shape and color, refractive index, electrolytic conductivity, melting or sublimation properties, latent heat of fusion, heat of solution, solubility, dissolution rate, enthalpy of transitions, phase diagrams, stability, hygroscopicity, and rates of reactions, were all affected by the nature of the crystal structure. It is usually the goal in pharmaceutical manufacturing to produce a drug substance that is phase-pure and which remains in that state as long as the bulk material is stored. It is also a goal to formulate the drug substance in a manner so that it remains in the same phase-pure state during the manufacture of the drug product and during any subsequent storage. This requires the development and validation of assay methodology for the determination of phase composition. The use of suitable techniques for the physical characterization of polymorphic and solvatomorphic solids has been discussed before,6–10 but these works usually focused on the study of the phase present in the largest extent. In this chapter, the question of small amounts of polymorphic and solvatomorphic impurities in a bulk solid will be addressed. The most useful methods for performing such work will be outlined, and illustrated with appropriate examples.

II. X-RAY DIFFRACTION Since polymorphism and solvatomorphism are crystallographic phenomena, it follows that the most important techniques for their detection and any quantification must be crystallography based. The technique of X-ray diffraction is such a method since it represents direct methodology whereby fundamental structural information can be obtained on the structure of a crystalline substance. For example, it is only by pure coincidence that two compounds might form crystals in which the three-dimensional spacing of planes is identical in all directions. One such example is provided by the trihydrate phases of ampicillin and amoxicillin,11 but such instances are uncommon. Typical applications of X-ray diffraction methodology include the determination of crystal structures, evaluation of polymorphism and solvate structures, evaluation of degrees of crystallinity, and the study of phase transitions.12 Bragg13 explained the diffraction of X-rays by crystals using a model where the atoms of a crystal are regularly arranged in space, and that they can be regarded as lying in parallel sheets separated by a definite and defined distance. Then he showed that scattering centers arranged in a plane act like

41

3 POLYMORPHIC AND SOLVATOMORPHIC IMPURITIES

a mirror to X-rays incident on them, so that constructive interference would occur for the direction of specular reflection. Within a given family of planes, defined by a Miller index of (h k l) and each plane being separated by the distance (d), each plane produces a specular reflectance of the incident beam. If the incident X-rays are monochromatic (having wavelength equal to ), then for an arbitrary glancing angle of , the reflections from successive planes are out of phase with one another. This yields destructive interference in the scattered beams. However, by varying , a set of values for  can be found so that the path difference between X-rays reflected by successive planes will be an integral number (n) of wavelengths, and then constructive interference will occur. Hence one ultimately obtains an expression known as Bragg’s law that explains the phenomenon: 2d sin  ¼ n

ð1Þ

Unlike the case of diffraction of light by a ruled grating, the diffraction of X-rays by a crystalline solid leads to observation of constructive interference (i.e., reflection), which occurs only for the critical Bragg angles. When reflection does occur, it is stated that the plane in question is reflecting in the nth order, or that one observes nth order diffraction for that particular crystal plane. Therefore one will observe an X-ray scattering response for every plane defined by a unique Miller index of (h k l).

A. Single Crystal X-ray Diffraction The analysis of single crystal X-ray diffraction data is divided into three parts.14,15 The first of these is the geometrical analysis, where one measures the exact spatial distribution of X-ray reflections and uses them to compute the size and shape of a unit cell. The second phase entails a study of the intensities of the various reflections; this information is used to determine the atomic distribution within the unit cell. Finally, one looks at the X-ray diagram to deduce qualitative information about the quality of the crystal or the degree of order within the solid. This latter analysis may permit the adoption of certain assumptions that may aid in solving of the crystalline structure. The phenomenon of X-ray diffraction has found widespread use as a means to determine the structures of single crystals and represents the most powerful and direct method to obtain bond lengths and bond angles for molecules in the solid state. This information is of extreme importance to workers in pharmaceutics when they encounter the existence of polymorphism or solvatomorphism. There is no doubt that single crystal X-ray diffraction is a powerful technique for the study of polymorphs and solvatomorphs, but it is equally apparent that this methodology is not well suited for the routine evaluation of the crystalline state of powdered solids. It is definitely not a technique that one would want to use in the determination of polymorphic phase impurities in a sample, since by definition only a single crystal is studied at one time. That crystal would certainly be phase-pure, and its quantitative analysis not

42

H.G. BRITTAIN AND A. MEDEK

subject to analysis. However, the powder pattern calculated from the single crystal structure would be of great use in interpreting the powder patterns of real samples.

B. X-ray Powder Diffraction For the crystallographic analysis of phase impurities, X-ray powder diffraction (XRPD) is definitely the method of choice.12 Since a powdered sample will present all possible crystal faces of all crystalline components at a given interface, the diffraction off this powdered surface will therefore provide information on all possible atomic spacings (i.e., defined by the crystal lattice). The powder pattern consists of a series of peaks having varying intensities detected at various scattering angles. These angles, and their relative intensities, are correlated with computed d-spacings to provide a full crystallographic characterization of the powdered sample.16,17 To measure a powder pattern, a randomly oriented powdered sample is prepared so as to expose all the planes of a sample. The scattering angle is determined by slowly moving the sample and measuring the angle of diffracted X-rays (typically using a scintillation detector) with respect to the angle of the incident beam. Alternatively, the angle between sample and source can be kept fixed while moving the detector to determine the angles of the scattered radiation. Knowing the wavelength of the incident beam, the spacing between the planes (identified as the d-spacings) is calculated using Bragg’s Law. In a mixture of crystalline materials, each species will be characterized by its distinctive series of scattering peaks, and the observed powder pattern will consist of a sum of the individual components. The intensities of the peaks associated with each component are proportional to its weight fraction in the mixture, so the performance of a quantitative analysis is based on the accurate and reproducible measurement of peak intensities. The analysis can be performed either with or without an international standard, and the theory associated with each approach has been set forth by Suryanarayanan12 and others.18–21 Quantitative XRPD was used to determine the relative component amounts of carbamazepine anhydrate and its dihydrate phase in powdered mixtures of these.22 Each form was characterized by a unique d-spacing (6.78 A˚ or 13.05 degrees 2- for the anhydrate, and 9.93 A˚ or 8.90 degrees 2- for the dihydrate), as shown in Figure 1, and the ratio of the integrated intensities of these scattering peaks could be correlated to the amount of Form-I in the mixture. In this work, the effects on the results that could be caused by preferred orientation, extinction, and particle size were discussed. A profile analysis method of fitting has been used to determine the relative amounts of prazosin hydrochloride polymorphs (the  and  forms) in powdered mixtures.23 The method was calibrated over -prazosin levels of 0.5–10% using a series of calibration samples, and it was found that the limit of detection was equal to 0.5%. The favorable limit was attainable owing to significant regions of non-overlap in the powder patterns of the two forms.

3 POLYMORPHIC AND SOLVATOMORPHIC IMPURITIES

43

FIGURE 1 Portions of the X-ray powder patterns of carbamazepine anhydrate (upper trace) and dihydrate (lower trace), illustrating the strong scattering peak associated with the dihydrate phase at 8.90 degrees 2-h (9.93 —).The figure was adapted from data presented in reference 22.

The quantitation of cefepine dihydrochloride dihydrate in samples of cefepine dihydrochloride monohydrate has been performed using quantitative XRPD.24 In this study, the working range was established as 1.0–8.0% w/w, with a detection limit of 0.75% and a quantitation limit of 2.5%. In this work, these workers overcame the difficult problem of mixing reference materials whose component crystals were of very different morphologies, which represents a problem not often recognized in work of this type. Since the compound was insoluble in acetone, the reference materials were slurried in the solvent, mixed thoroughly, and allowed to dry. It is evident from the electron microscopic images shown in the paper that very homogeneous powder blends were obtained using the slurry approach.

44

H.G. BRITTAIN AND A. MEDEK

In a system having attained great notoriety owing to patent litigations, excellent sensitivity has been obtained in the use of quantitative XRPD for the quantitation of the polymorphs of ranitidine.25 Owing to significant nonoverlap, diffraction peaks suitable for the determination of either Form-I in bulk Form-II, or for Form-II in bulk Form-I, are easily identified. Quantitative XRPD has been used to study the effects of temperature on the phase transformation of two forms of chlorpropamide during tableting.26 Although these authors did not use an internal standard, they were able to generate calibration curves by taking ratios of a characteristic scattering peak for each form. It was learned that the stability of Form A was affected by compression temperature, while that of Form C was independent of temperature. The pathways of polymorphic transitions for cephalexin, chloramphenicol palmitate, and indomethacin have been studied, and the composition of partially transformed samples determined by XRPD.27 The metastable forms of chloramphenicol palmitate were found to transform into the stable phase when found at room temperature, confirming the results of a previous study.28 In contrast, cephalexin became amorphous upon grinding. Indomethacin exhibited the interesting behavior of becoming amorphous when ground at 4 C, but could be transformed into a metastable phase when ground at 30 C. The kinetics associated with the thermally induced phase transformations of phenanthrene and caffeine monohydrate were studied using hotstate quantitative XRPD.29 Using a single non-isothermal experiment conducted at a constant heating rate, it was possible to obtain the activation parameters for the solid-state reactions. In another study, quantitative XRPD was used to study the tetrahydrate to monohydrate transition of the sodium salt of 5-(4-oxo-phenoxy-4H-quinolizine-3-carboxamide)-tetrazolate.30 Although not a polymorphic phase in the formal sense, the amorphous state can represent an important quantity of measure in a crystalline sample. A general XRPD procedure for such work has been described,31 and similar approaches have been reported for cilastatin sodium,32 caffeine,33 imipenem,34 acetaminophen,35 and a Lumaxis analog.36 It is well known that resolved enantiomers and the racemic mixtures of dissymmetric substances ordinarily crystallize in different space groups,37 so it is natural to use quantitative XRPD as a means to determine the enantiomeric composition. This approach was demonstrated in the case of ibuprofen, taking advantage of the differing powder patterns of the resolved enantiomers and the racemate.38 The detection limits for either form (in the presence of the other form) were very similar, being approximately 3.3% w/w.

III. THERMAL METHODS OF ANALYSIS Thermal analysis methods are defined as those techniques for which one or more properties of a sample are determined as a function of an externally

3 POLYMORPHIC AND SOLVATOMORPHIC IMPURITIES

45

applied temperature. Regardless of the observable parameter measured, the usual practice requires that the physical property and the sample temperature be recorded continually and automatically, and that the sample temperature is altered at a predetermined rate. Measurements of thermal analysis are conducted for the purpose of evaluating the physical and chemical changes that may take place in a heated sample, requiring that the operator interpret the events noted in a thermogram in terms of plausible reaction processes. Thermal reactions can be identified as being either endothermic (melting, boiling, sublimation, vaporization, desolvation, solid–solid phase transitions, chemical degradation, etc.) or exothermic (crystallization, oxidative decomposition, etc.) in nature. The various thermal methods in common use have been discussed at great detail elsewhere.39–40 Such methodology has found widespread use in the pharmaceutical industry for the characterization of compound purity, polymorphism, solvation, degradation, and excipient compatibility.66,67 However, given the utility which thermal microscopy has shown for the characterization of polymorphic systems, it is not surprising that the quantitative applications of thermal analysis have proven to be even more useful. Although a large number of techniques have been developed, the most commonly applied are those of thermogravimetry (TG), differential thermal analysis (DTA), and differential scanning calorimetry (DSC).

A. Thermogravimetry Thermogravimetry is a measure of the thermally induced weight loss of a material as a function of the applied temperature.44–46 TG analysis is restricted to transitions that involve either a gain or loss of mass and is most commonly used to study desolvation processes and compound decomposition. TG analysis is a very useful method for the quantitative determination of the total volatile content of a solid and can be used as an adjunct to Karl Fischer titrations for the determination of moisture. As such, it readily permits the distinction between solvates and the anhydrous forms of a given compound. When performed in conjunction with an auxiliary technique (such as gas chromatography or infrared spectroscopy), one may also obtain compound-specific decomposition information. Thermogravimetry analysis also represents a powerful adjunct to the other methods of thermal analysis, since a combination of either a DTA or DSC study with a TG determination can be used in the assignment of observed thermal events. Desolvation processes or decomposition reactions must be accompanied by weight changes, and can be identified by a TG weight loss over the same temperature range. On the other hand, solid–liquid or solid–solid phase transformations are not accompanied by any loss of sample mass and would not register in a TG thermogram. When a solid is capable of decomposing by means of several discrete, sequential reactions, the magnitude of each step can be separately evaluated. The TG analysis of compound decomposition can also be used to compare the stability of similar compounds. In general, the higher the decomposition temperature of a given compound, the greater would be its stability.

46

H.G. BRITTAIN AND A. MEDEK

The measurement of thermogravimetry is simple in principle and consists of the continual recording of the mass of the sample as it is heated in a furnace. The weighing device used in most instruments is a microbalance, which permits the characterization of milligram quantities of sample. The balance chamber itself is constructed so that the atmosphere may be controlled, which is normally accomplished by means of a flowing gas stream. The furnace must be capable of being totally programmable in a reproducible fashion, whose inside surfaces are resistant to the gases evolved during the TG study. It is most essential in TG design that the temperature readout be that of the sample, and not to that of the furnace. To achieve this end, the thermocouple or resistance thermometer must be mounted as close to the sample pan as possible. Although the TG methodology is conceptually simple, the accuracy and precision associated with the results are dependent on both instrumental and sample factors. The furnace heating rate used for the determination will greatly affect the transition temperatures, while the atmosphere within the furnace can influence the nature of the thermal reactions. The sample itself can play a role in governing the quality of data obtained, with factors such as sample size, nature of evolved gases, particle size, heats of reaction, sample packing, and thermal conductivity all influencing the observed thermogram. Other than its ability to demonstrate the anhydrous nature of genuine polymorphic materials, one extremely useful aspect of TG analysis is in the differentiation and characterization of solvatomorphs. The methodology is particularly useful in the determination of a solvate phase in the presence of its anhydrate phase, or vice versa. The most commonly encountered form of lactose is its -anomer, which is obtained in the form of a monohydrate phase. The anhydrous form of the -anomer is known to be very hygroscopic and difficult to either obtain or handle. The -anomer of lactose is obtained as an anhydrate phase, which apparently has no tendency to form any hydrate phases. The thermal properties of these materials have been discussed.47 The theoretical water content for the -monohydrate phase is 5.0% w/w, so in principle one can use thermogravimetry to determine either the anhydrate phase content in a largely monohydrate sample or the monohydrate phase content in a largely anhydrous sample. To illustrate the utility of TG analysis for such determinations, a series of calibration samples were prepared at levels approximately 90% or 10% in the monohydrate phase, using anhydrous -lactose as a diluent in bulk -monohydrate.48 These were subjected to a standard TG evaluation, and the results of the analysis are shown in Table 1. The results indicate that the TG method yields reasonable estimations of the phase composition, but that the difference between measured and found compositions could be unacceptably large for some applications. The data appear to indicate the existence of a small positive bias, which was found to be worse in the 10% monohydrate samples. In a properly controlled and validated system, however, it is highly likely that method limitations could be overcome and TG analysis appropriately used for quantitative analysis of phase composition.

47

3 POLYMORPHIC AND SOLVATOMORPHIC IMPURITIES

TABLE 1 Thermogravimetric Analysis of Lactose Calibration Standards 48

Actual weighed composition of sample, expressed as monohydrate (%)

Anticipated TG weight loss of sample (%)

Experimentally measured TG weight loss (%)

Calculated monohydrate content, based on weight loss of 5.0% for the monohydrate (%)

(a) Approximately 90% monohydrate phase 89.8

4.490

4.58

91.60

88.4

4.420

4.34

86.80

92.3

4.615

4.72

94.40

91.1

4.555

4.63

92.60

92.6

4.630

4.71

94.20

(b) Approximately 10% monohydrate phase 9.8

0.490

0.68

13.60

11.4

0.570

0.73

14.60

12.8

0.640

0.88

17.60

10.9

0.545

0.75

15.00

8.7

0.435

0.59

11.80

B. Differential Scanning Calorimetry Differential thermal analysis was the first major improvement developed over simple melting point analysis, and in countless studies was used to determine the characteristic temperature ranges associated with a variety of thermally induced reactions. Differential scanning calorimetry subsequently effectively replaces the DTA method, primarily because of its ability to yield quantitative information regarding the magnitude of the heat associated with the thermal reaction. For this reason, DSC has become accepted as the most widely used method of thermal analysis for the pharmaceutical industry. In the DSC method, the sample and reference materials are maintained at the same temperature, and the heat flow required to keep the equality in temperature is measured. DSC plots are therefore obtained as the differential rate of heating (in units of watts/second calories/second or joules/second) against temperature.39,40 The area under a DSC peak is directly proportional to the heat absorbed or evolved by the thermal event, and integration of these peak areas yields the heat of reaction (in units of calories/secondgram or joules/secondgram). Two types of DSC measurements are possible, which are usually identified as power-compensation DSC and heat-flux DSC, and the details of each configuration have been fully described.39,40 In power-compensated DSC, the sample and reference materials are kept at the same temperature by the use of individualized heating elements, and the observable parameter recorded is the difference in power inputs to the two heaters. In heat-flux

48

H.G. BRITTAIN AND A. MEDEK

DSC, one simply monitors the heat differential between the sample and reference materials, with the methodology not being terribly different from that used for DTA. In the DTA measurement, an exothermic reaction is plotted as a positive thermal event, while an endothermic reaction is usually displayed as a negative event. Unfortunately, the use of power-compensation DSC results in endothermic reactions being displayed as positive events, a situation which is counter to the IUPAC recommendations.49 When the heat-flux method is used to detect the thermal phenomena, the signs of the DSC events concur with those obtained using DTA, and also agrees with the IUPAC recommendations. The calibration of DSC instruments is normally accomplished through the use of compounds having accurately known transition temperatures and heats of fusion, and an extraordinarily extensive list of suitable reference materials is available.50 Once the DSC system is properly calibrated, it is trivial to obtain the melting point and enthalpy of fusion data for any compound upon integration of its empirically determined endotherm and application of the calibration parameters. The current state of methodology is such, however, that unless a determination is repeated a large number of times, the deduced enthalpies must be regarded as being accurate only to within approximately 5%. The DSC thermograms of a number of lactose monohydrate samples were determined, and it was ultimately found that the dehydration endotherm was characterized by an enthalpy of desolvation equal to 150.5 J/g.48 This value for the pure monohydrate phase can then be used in conjunction with the measured desolvation enthalpies of actual samples to determine either the anhydrate phase content in a monohydrate sample or the monohydrate phase content in an anhydrate sample. To illustrate the utility of DSC analysis for such determinations, a series of calibration samples were prepared at levels approximately 90% or 10% in the monohydrate phase, using anhydrous -lactose as a diluent in bulk -monohydrate.48 These were subjected to a standard DSC evaluation, and the results of this analysis are shown in Table 2. The results indicate that the DSC method was able to provide very good estimations of the phase composition. It is concluded that the magnitude of the error associated with the use of DSC analysis would be acceptable for quantitative analysis of mixtures of lactose monohydrate and anhydrate. As long as the two forms of the analyte do not interconvert during acquisition of the DSC thermogram, and values for the transition enthalpies have been determined for reference materials, one can always use DSC as a quantitative method for the determination of phase composition. The method will fail, however, when a metastable form converts to a more stable form during the DSC analysis. It is therefore always prudent to establish the phase relationship and order of stability between the various forms51 before undertaking development of a quantitative DSC method. A very interesting procedure has been described where DSC is used to determine the water content in hydrates.52 The method is based on the assumption that the enthalpy of binding associated with n moles of water in

49

3 POLYMORPHIC AND SOLVATOMORPHIC IMPURITIES

TABLE 2 Differential Scanning Calorimetric Analysis of Lactose Calibration Standards 48 Actual weighed composition of sample, expressed as monohydrate (%)

Anticipated DSC enthalpy of desolvation for sample (J/g)

Experimentally measured enthalpy of desolvation (J/g)

Calculated monohydrate content, based on an enthalpy of desolvation of 150.5 J/g (%)

(a) Approximately 90% monohydrate phase 91.4

137.56

137.9

91.63

90.2

135.75

135.0

89.70

88.7

133.49

133.0

88.37

89.3

134.40

133.4

88.66

87.2

131.24

131.9

87.64

(b) Approximately 10% monohydrate phase 8.7

13.09

12.5

8.31

10.2

15.35

14.9

9.90

11.8

17.76

17.2

11.43

9.6

14.45

15.0

9.97

9.3

14.00

14.7

9.77

the hydrate is the same as that associated with n moles of water in bulk water. Knowing the enthalpy of vaporization for liquid water and the enthalpy of desolvation for the hydrate sample enables a facile calculation of the number of moles of water bound in the hydrate. The method was applied to the hydrates of ampicillin, carbamazepine, caffeine, theophylline, and metal salts of nedocromil with good agreement between known hydration numbers and those calculated using the DSC method.

IV. VIBRATIONAL SPECTROSCOPY The energies associated with the vibrational modes of a chemical compound lie within the range of 400–4000 cm1. These modes can be observed directly through their absorbance in the infrared region of the spectrum or through the observation of the low-energy scattered bands that accompany the passage of an intense beam of light through the sample (the Raman effect). In either case, the use of Fourier transform methodology has vastly improved the quality of data that can be obtained.53 Most workers are familiar with the use of mid-infrared spectra for identity purposes, where the pattern of absorption bands is taken to be diagnostic for a given compound. However, it has come to be recognized that the vibrational spectra of solid materials will reflect details of the crystal structure, and hence these methods can be used in the spectroscopic investigation of polymorphs and solvates.54–56 In studies of polymorphic or solvatomorphic systems, the purpose of the vibrational spectroscopic investigation should be to gather information from the observed pattern of vibrational frequencies and to use this data to

50

H.G. BRITTAIN AND A. MEDEK

understand the structural aspects that yield crystallographic differences. Once suitable spectral features are identified from this work, they can be used to develop easily performed methods for the quantitative analysis of one polymorph (or solvatomorph) in the presence of the other. Unfortunately, too many workers are satisfied to merely obtain the spectra of the various polymorphs and/or solvatomorphs and to simply display the differences. In doing so, they miss a great opportunity to gain additional understanding about the system they are trying to characterize.

A. Infrared Absorption Spectroscopy The acquisition of high-quality infrared absorption spectra appropriate for the characterization of polymorphs and solvates is best performed using Fourier transform technology (the FTIR method), since this approach minimizes transmission and beam attenuation problems. Essentially all FTIR spectrometers use a Michelson interferometer, where radiation entering the interferometer is split into two beams by means of a beam splitter. One beam follows a path of fixed distance before being reflected back into the beam splitter, while the other beam travels a variable distance before being recombined with the first beam. The recombination of these two beams yields an interference pattern, where the time-dependent constructive and destructive interferences have the effect of forming a cosine signal. Each component wavelength of the source will yield a unique cosine wave, having a maximum at the zero pathlength difference (ZPD) and which decays with increasing distance from the ZPD. The detector is placed so that radiation in the central image of the interference pattern will be incident upon it, and therefore intensity variations in the recombined beam manifest as phase differences. The observed signal at the detector is a summation of all the cosine waves, having a maximum at the ZPD, and which decays rapidly with increasing distance from the ZPD. If the component cosine waves can be resolved, then the contribution from individual wavelengths can be observed. The frequency domain spectrum is obtained from the interferogram by performing the Fourier transformation mathematical operation. More detailed descriptions of FTIR instrumentation and its methodology is available.57 The acquisition of solid-state FTIR spectra suitable for use in the characterization of polymorphic impurities is performed using either the Nujol mull technique, diffuse reflectance (DRIFT), or attenuated total reflectance (ATR). One should avoid the use of pelleting techniques to eliminate any spurious effects associated with compaction of the KBr pellet. The simplest approach is to prepare a mull of the sample in mineral oil, sandwich this between salt plates, and measure the spectrum using ordinary transmission techniques. The main drawback of the mull technique is that regions in the IR spectrum overlapping with carbon–hydrogen vibrational modes will be obliterated because of absorbance from the oil. A quantitative IR absorption for the two polymorphs of the developmental compound (2R,3S)-2-({(1R)-1-[3,5-bis(trifluoromethyl)phenyl] ethyl}oxy)-3-(4-fluorophenyl) morpholine hydrochloride was reported in which the measurements were made in mineral oil mull preparations.58

3 POLYMORPHIC AND SOLVATOMORPHIC IMPURITIES

51

The mull technique was used to ensure compatibility with wet cake and slurry samples. A calibration curve was established for binary mixtures of the two forms using only an absorbance ratio of two characteristic absorption bands, and a limit of quantitation of 2.7% w/w was established. The method was then applied to the kinetic investigation of the transformation from the metastable form to the stable form. The measurement of diffuse reflectance effectively involves focusing the infrared source beam onto the surface of a powder sample and using an integrating sphere to collect the scattered infrared radiation.59 The technique requires careful attention to sample preparation, and often one must dilute the analyte with KBr powder to reduce the occurrence of anomalous effects.60 In practice, one obtains the spectrum of the finely ground KBr dispersant, and then ratios this to the spectrum of KBr containing the analyte. The relative reflectance spectrum is converted into Kubelka–Munk units using standard equations,61 thus obtaining a diffuse reflectance spectrum that resembles a conventional IR absorption spectrum. Guillory and coworkers were among the first to use diffuse reflectance in conjunction with FTIR technology and the partial least-squares (PLS) regression method.62 The particle size of the reference materials was first homogenized through light grinding in an agate mortar, and then calibration samples prepared by thorough mixing. The spectra of these samples, and their dilutions in micronized KCl were obtained under a variety of conditions to investigate the robustness of the process. As shown in Figure 2, good linearity was found between the Kubelka–Munk units and the composition of the calibration samples. When using the PLS method good agreement between the actual and predicted values of the calibration samples was obtained, with the average absolute error being 0.006% in KCl for Form-I and 0.004% in KCl for Form-II. In addition to the XRPD quantitation of cefepine dihydrochloride dihydrate in samples of cefepine dihydrochloride monohydrate, diffuse reflectance infrared spectroscopy was used to determine the solvatomorphic composition.24 For the IR assay, a working range of 1.0–8.0% w/w, with a detection limit of 0.3% and a quantitation limit of 1.0%, and these analytical limits were superior to those obtained using the quantitative XRPD method. The validity of both assay methods were limited for samples having a particle size range of 125–590 mm. Counting polymorphs and solvatomorphs, delavirdine mesylate has been found in 12 different crystal forms, therefore presenting considerable difficulty in the determination of phase composition.63 Only through the combination of factor analysis with the quantitative IR technique could one develop a method useable in the characterization of research and production lots. The composition of drug lots consisting of mixtures of forms was identified through qualitative factor analysis, and quantitated using principal component analysis. Calibration models were developed for the determination of Form-VIII or Form-XII in Form-XI, and a standard error of prediction of 2.0% of either form was reported, with detection limits of 3–5%. Diffuse reflectance IR spectroscopy has been used, in conjunction with XRPD analysis, to determine the levels of ranitidine Form-II in samples of

52

H.G. BRITTAIN AND A. MEDEK

FIGURE 2 Infrared absorption intensity (expressed in terms of Kubelka^Munk units) of

sulfamethoxazole Form-I (absorbance at 3468 cm1) and Form-II (absorbance at 3476 cm1) as a function of concentration in a KCl matrix.The figure was adapted from data presented in reference 62.

bulk Form-I.25 Three artificial neural network models were trained and used to enhance the analysis. Although the model constructed from the XRPD data was superior in predictions at lower concentrations of Form-II (1% and 2%), the model constructed using diffuse IR reflectance had better overall average prediction. The ATR method of detection is emerging as a highly useful approach to obtaining IR absorption spectra with minimal sample preparation.64 If an IRtransparent crystal is sandwiched by the sample and if the refractive index of the crystal exceeds that of the sample, at certain orientations an infrared beam entering the crystal will undergo multiple internal reflections. At each reflection, some of the incident energy is absorbed by the vibrational modes of the sample, and the degree of this absorption builds with the number of internal reflections. When the beam finally emerges from the crystal, it can be processed in the usual way to obtain a pattern of the IR absorption bands of the sample in contact with the crystal. The feasibility of using FTIR–ATR spectroscopy for qualitative and quantitative analysis was investigated using the three known polymorphs of ganciclovir as a model system.65 Definitive identification of each polymorph was obtained from materials that did not need to be subjected to any sample handling or preparation. Quantitation of mixtures was carried out using a partial least-squares procedure, with mean absolute errors of less than 3% being reported for Form-I and Form-II, and about 6% for Form-III.

3 POLYMORPHIC AND SOLVATOMORPHIC IMPURITIES

53

B. Raman Spectroscopy The vibrational modes of a compound may also be studied using Raman spectroscopy, where one measures the inelastic scattering of radiation by a nonabsorbing medium.66 When a beam of light is passed through a material, approximately one in every million incident photons is scattered with a loss or gain of energy. The inelastically scattered radiation can occur at lower (Stokes lines) and higher (anti-Stokes lines) frequencies relative to that of the incident (or elastically scattered) light, and the energy displacements relative to the energy of the incident beam correspond to the vibrational transition frequencies of both the medium. The actual intensities of the Stokes and anti-Stokes lines are determined by the Boltzmann factor characterizing the vibrational population. For high-frequency vibrations, the Stokes lines are somewhat intense relative to the anti-Stokes lines, so conventional Raman spectroscopy makes exclusive use of the Stokes component. The Raman effect originates from the interaction of the oscillating induced polarization or dipole moment of the medium with the electric field vector of the incident radiation. Raman spectra are measured by passing a laser beam through the sample and observing the scattered light either perpendicular to the incident beam or through back-scatter detection. The scattered light is analyzed at high resolution by a monochromator, and ultimately detected by a suitable device. One key to obtaining good spectra is through the use of a notch filter, which will eliminate the exciting line, since that is required to obtain acceptable signal-to-noise ratios. Although both infrared absorption and Raman scattering yield information on the energies of the same vibrational bands, the different selection rules governing the band intensities for each type of spectroscopy can yield useful information. For the low-symmetry situations presented by the structures of molecules of pharmaceutical interest, every vibrational band will be active to some degree in both infrared absorption and Raman scattering spectroscopies. The relative intensities of analogous bands will differ, however, when observed by either infrared absorption or Raman spectroscopy. In general, symmetric vibrations and nonpolar groups yield the most intense Raman scattering bands, while antisymmetric vibrations and polar groups yield the most intense infrared absorption bands. A discussion on a large number of practical applications of Raman spectroscopy is available,67 and its application to pharmaceutical analysis has been discussed.68,69 In one study, Raman spectroscopy was used to analyze binary mixtures of the two forms of chlorpropamide.70 A complete series of calibration samples ranging from 0% to 100% in each form were prepared, and characterized between 400–1800 cm1 and 2800–3200 cm1. Through the use of factor analysis, good correlation of predicted versus theoretical polymorphic composition was obtained in studies of instrumental and between-sample reproducibilities. An unidentified compound of pharmaceutical interest was reported to be able to exist in two polymorphic forms, each of which exhibited distinctive Raman spectra.71 A series of calibration samples were prepared over the Form-A range of 1.8–15.4% w/w in Form-B, and a calibration curve developed using the characteristic scattering bands at 1716 cm1 (Form-A)

54

H.G. BRITTAIN AND A. MEDEK

FIGURE 3 Concentration profiles measured by in situ Raman spectroscopy during the

phase transformation of progesterone Form-II () to Form-I () The figure was adapted from data presented in reference 73.

and 1724 cm1 (Form-B). Using spectral subtraction, it was determined that as little as 1.8% of one form could be detected in the presence of the other. The solvent-mediated polymorphic transformation of progesterone Form-II to Form-I was followed in real time using in situ Raman spectroscopy.72 Although Form-I is the thermodynamically stable form, the spectroscopic investigation showed that its appearance is always preceded by the formation of Form-II, which appeared to be kinetically favored. In a kinetic study (plotted in Figure 3), the time dependence of the phase transformation at 45 C was followed and was used to determine the end point of the transformation. The advantage of Raman spectroscopy to provide real-time evaluations of progress of phase transformation was found to be significant when compared to the use of other off-line techniques. The in situ Raman spectroscopy method has also been used to study the particle size–dependent molecular rearrangements that take place during the dehydration of trehalose dihydrate.73 Different phases were sieved into fractions < 45-mm and > 425-mm particle size, and the Raman spectra obtained at various times during an isothermal heating at 80 C. After being heated for 210 minutes, the < 45-mm dihydrate material appeared to become amorphous while the > 425-mm dihydrate material transformed into the crystalline anhydrate phase. Ratios of various characteristic scattering peaks were used to follow the kinetics of the phase transformations.

3 POLYMORPHIC AND SOLVATOMORPHIC IMPURITIES

55

Raman spectroscopy has been shown to be a useful method for the determination of degrees of crystallinity and can complement other quantitative methods for such work.74 This was demonstrated using indomethacin as an example, with the quantitative analysis being based on the peak intensity ratios of the 1698 cm1 (crystalline) and 1680 cm1 (amorphous) scattering peaks. A linear correlation curve was obtained across the entire 0–100% range of crystallinity, and as little as 1% amorphous or crystalline phase content could be detected. The largest source of error was found to result from inhomogeneous mixing of the calibration standards, a phenomenon commonly encountered during the validation of quantitative assay methods for powdered solids.

C. Near-Infrared Spectroscopy The absorption bands found in the near-infrared (NIR) region of the spectrum (typically 1000–2500 nm) are all due to overtones and combinations of fundamental molecular vibrational modes.75 The NIR spectrum tends to be dominated by vibrational modes of light atoms having strong bond strengths, typically hydrogen bound to nitrogen, oxygen, or carbon. The energies of the overtone bands are more affected by environmental details than are the energies of their fundamentals, so slight perturbations in the bonding can yield drastic frequency and amplitude changes in the nearinfrared. Discussions of the pharmaceutical applications of NIR spectroscopy are available.76–77 Although the molar absorptivities of these bands tend to be rather small, the instrumental state of the art, combined with superior data deconvolution routines, has progressed to the point where their measurement is relatively straightforward.78 The most important experimental requirements are the use of high-intensity stable light sources, low-noise sensitive detectors, and efficient methods for collection of the diffuse reflectance of the measurement. NIR spectroscopy became much more useful when the principle of multiple-wavelength spectroscopy was combined with the deconvolution methods of factor and principal component analysis. In typical applications, partial least squares regression is used to model the relation between composition and the NIR spectra of an appropriately chosen series of calibration samples, and an optimal model is ultimately chosen by a procedure of cross-testing. The performance of the optimal model is then evaluated using the normal analytical performance parameters of accuracy, precision, and linearity. Since its inception, NIR spectroscopy has been viewed primarily as a technique of quantitative analysis and has found major use in the determination of water in many pharmaceutical materials. Near-infrared spectroscopy was used to quantitatively determine the phase composition of a compound capable of existing in two polymorphic forms even though the NIR spectra were fairly similar.79 Using a fivewavelength calibration model, nearly 100% recovery was obtained in a series of spiked calibration samples, with relative standard deviation values ranging from 0.1% to 0.9%. The authors concluded that the use of NIR spectroscopy

56

H.G. BRITTAIN AND A. MEDEK

FIGURE 4 Near-infrared spectra of a-lactose monohydrate (solid trace) and b-lactose anhydrate (dashed trace), highlighting the strong water combination band at 1940 nm. The figure was adapted from data presented in reference 80.

had a major advantage in that it required large quantities of material (1 g or more) for the measurements, thus minimizing sampling errors. In a wide-ranging study, NIR spectroscopy was used to quantitate the phase composition in various forms of sulfamethoxazole, sulfathiazole, lactose, and ampicillin.80 For instance, as shown in Figure 4, the -monohydrate phase of lactose is easily distinguished from -lactose anhydrate on the basis of the characteristic band at 1940 nm associated with the water combination mode. In all cases, however, properly calibrated NIR methods were able to yield good predictions of phase composition relative to the actual composition of the standards used, and it was concluded that the quantitative NIR was equally effective for such work as other commonly used quantitative methods. Near-infrared spectroscopy has been used to quantitate sulfathiazole Form-I and Form-III in binary physical mixtures in which one form was the dominant component.81 The spectra of each form exhibited sufficient differences that unique wavelengths of absorbance were easily attributable to each form. Excellent linearity in calculated versus actual compositions were obtained over the concentration range of 0–5% for either Form-I in Form-III, or for Form-III in Form-I. After considering appropriate calibration models, a limit of quantitation of approximately 0.3% was ultimately deduced. It is relatively straightforward to apply the NIR method for in situ characterization work, as was reported in the case of the EFGR tyrosine kinase inhibitor 4-(3-ethynylphenylamino)-6,7-bis(2-methoxyethoxy)quinazolinium methanesulfonate.82 NIR spectroscopy was used to monitor the kinetics of transformation between the polymorphs and solvatomorphs, and

3 POLYMORPHIC AND SOLVATOMORPHIC IMPURITIES

57

even facilitated the discovery of a new preferred form. It was determined that the in situ NIR method could be generally used in the study of practically all types of two-phase solid–liquid slurries maintained under isothermal conditions. Near-infrared spectroscopy was used to study the solvatomorphic changes and the state of water during the wet granulation of theophylline.83 The anhydrate was granulated in a planetary mixer using water as granulation liquid and characterized by a variety of methods. At a low level of granulation fluid (0.3 moles of water per mole of theophylline anhydrate), water absorption first yielded NIR maxima (1475 and 1970 nm) characteristic of theophylline monohydrate. At higher quantities of granulation fluid (1.3–2.7 moles of water per mole of theophylline anhydrate), the absorption maxima (1410 and 1905 nm) of free water became evident. Owing to its greater ease of spectral acquisition, it was determined that NIR spectroscopy was superior for the detection of different states of water during the wet granulation process.

V. SOLID-STATE NUCLEAR MAGNETIC RESONANCE (SS-NMR) SPECTROMETRY Nuclear magnetic resonance (NMR) has become one of the most powerful spectroscopic tools, able to provide detailed information not only about the chemical constitution of compounds but also about their threedimensional structures and dynamics. Most NMR applications have traditionally involved samples dissolved in a solution phase. The two key factors precluding widespread use of solid-state NMR in the past were sensitivity and resolution. A boost in solid-state applications appeared after the introduction of a combination of the line-narrowing technique known as magic angle spinning (MAS)84–86 with the sensitivity enhancing technique termed cross-polarization (CP).87 Cross-polarization magic angle spinning (CP-MAS) brought about the necessary sensitivity enhancement to make solid-state NMR spectroscopy practicable for natural abundance samples.88 Even though pharmaceutical applications of solid-state NMR (SS-NMR) are well documented, the full potential of this technique has not yet been realized, and it is anticipated that solid-state NMR will become a standard technique for characterization of pharmaceutical solids.

A. Solid-State NMR Techniques 1. Proton Solid-State NMR In contrast to proton spectroscopy in solutions, 1H-NMR has found only limited application in the solid state, mainly because of the existence of extensive line broadening originating from strong through-space, dipole– dipole coupling. This interaction is completely averaged out in solution as the molecules rapidly tumble on the NMR time scale. In solids, however, molecules are locked in their crystal lattices. In the absence of extensive molecular motions, the dipole–dipole interaction can be very strong and are

58

H.G. BRITTAIN AND A. MEDEK

usually larger than the chemical shielding effect. Moreover, since typical organic molecules contain many protons in close proximity to each other, the through-space dipole–dipole interaction leads to fast exchange of magnetization between the protons, counteracting the MAS line narrowing. Such spin diffusion is fast on the NMR time scale. As a result, a typical proton static line shape of organic molecules consists of a single peak (up to 50 kHz wide), precluding resolution of the chemical shifts (see Figure 5, top trace).89–91 Since magic angle spinning only partially narrows down the proton line shape (Figure 5, middle trace), other line-narrowing techniques have been sought. Combined rotation and multiple pulse spectroscopy (CRAMPS)92 offers a partial solution to this problem (Figure 5, bottom trace). This technique combines conventional magic angle sample spinning with rotation of proton magnetization around the magic angle in spin space.

FIGURE 5 Solid-state proton spectra of (L)-alanine, showing the differences between spectra acquired in static, spinning, and CRAMPS modes. All 500 MHz spectra were acquired at ambient conditions.

3 POLYMORPHIC AND SOLVATOMORPHIC IMPURITIES

59

The CRAMPS experiment puts a large demand on the NMR hardware (especially on NMR probes), since high-power radio frequency (rf ) pulses are applied between each acquisition point. High homogeneity of the irradiation field, as well as careful setup of experimental variables, is required to avoid distortions of the proton peaks. Typical line widths afforded by the CRAMPS experiment are approximately 1 ppm, limiting the application of the CRAMPS experiment to compounds with a small number of well-resolved protons. 2. Heteronuclear Solid-State NMR Solid-state NMR spectra of heteronuclei such as carbons are largely devoid of the limitations encountered in proton applications. Though a strong dipole–dipole interaction between protons and carbons prevails, it is relatively easy to decouple the heteronuclear interaction by high-power proton irradiation.89–90 In natural abundance samples, the degree of homonuclear 13C–13C dipolar coupling is very small, because of the spin dilution of 13C nuclei to approximately 1%. Even a moderate MAS speed will remove this effect completely. Depending on the degree of crystallinity and the residual dynamics of the compound, the 13C solid-state NMR peaks can be very narrow. Compounds with high internal mobility, such as adamantane, can have line widths as narrow as 0.01 ppm. This is why in solid-state NMR, adamantane is used not only as a chemical shift standard, but also for optimizing experimental variables and for shimming probes. A typical line width encountered in crystalline organic compounds is of the order of 0.5 ppm, but the exact value will be strongly dependent on the mobility of the compound. It is not unusual to resolve the majority of carbon resonances in compounds containing 40 or more carbons. Before the development of the cross-polarization technique, the two serious limitations that precluded extensive use of carbon solid-state spectroscopy were low 13C sensitivity and its long T1 relaxation times. Both of these limitations are largely removed by cross-polarization of the proton intensities to carbons. CP relies on a strong pool of proton magnetization that is transferred during the cross-polarization step by spin locking pulses to carbons.87–88 The simultaneous spin locking pulses on proton and carbon channels must be of the specific rf amplitudes required by the Hartman–Hahn matching condition.87 Since the magnetization in the CP experiment starts at protons, it is the proton T1 relaxation that determines the recycle delay between successive scans. Because of dipole–dipole coupling, the proton relaxation is more efficient than carbon relaxation, leading to shorter 1H T1 times. Therefore, the CP experiment improves carbon sensitivity and, at the same time, shortens the necessary recycle delays. 3. Assignment of Spectral Features Knowledge of carbon-specific assignments of 13C-NMR resonances is not absolutely essential to differentiate polymorphic substances. However, if the assignments are available, the chemical shift differences between polymorphs can be related to conformational changes at particular molecular sites (Figure 6). Recently it has been demonstrated that assignments can be made

60

H.G. BRITTAIN AND A. MEDEK

13 C assignments of vitamin B12 , recrystallized from ethyl alcohol (‘‘wet’’ form) 154. In the top right scheme of vitamin B12 (depicted in bold) are those carbons that exhibit shifts between the ‘‘wet’’ and ‘‘dried’’ forms. All 75.8 MHz spectra have been acquired at ambient conditions using (a) the conventional CP-MAS sequence, (b) cross-polarization followed by a period of dipolar dephasing (methyl and nonprotonated carbons observed only), (c) INEPT based spectral editing (methyl, methine and nonprotonated carbons observed only) reference 108, and (d) cross-polarization with short contact time (mainly methine and methylene carbons observed). The solid-state assignments are based on combining known solution assignments, reference 171, with spectral editing techniques.

FIGURE 6

3 POLYMORPHIC AND SOLVATOMORPHIC IMPURITIES

61

by relying solely on solid-state NMR data. For instance, a tripeptide sample has been completely assigned in the solid state.93 However, for small molecules with a limited number of resonances and good spectral dispersion, most of the assignments can be deduced by direct comparison of the liquids and solid-state spectra. In typical applications of pharmaceutical interest, solution NMR assignments are either already available or can be completed before proceeding with the solid-state analysis. Solid-state NMR resonances usually fall within several parts per million (ppm) from solution shifts. The origins of differences between the solution-phase and solid-phase spectra are primarily due to conformational changes and crystal-packing effects. A larger contribution, typically not exceeding 12 ppm, comes from the differences between the average molecular conformation or configuration in solution and the solid state. A smaller contribution, typically not exceeding 3 ppm, comes from the packing of molecules in the crystal lattice.94 The latter effect is nonexistent in solution because of the fast tumbling of the molecules. Most of the assignments can be inferred by comparing the solid-state spectra with their assigned solution counterparts. The remaining resonances can be assigned from the application of simple editing techniques. The two simplest editing experiments, affording complementary information, are short CP and dipolar dephasing techniques.95–97 The short CP experiment limits the duration of the cross-polarization period to a short interval of approximately 100 ms, whereas conventional CP times are in the range of 1–10 ms. Only protonated carbons show significant intensities due to their strong dipolar couplings (Figure 6d). Having no close proton neighbors and relying only on relatively distant proton–carbon dipolar couplings, quaternary carbons do not have enough time to build up sufficient polarization. On the other hand, the dipolar dephasing experiment uses strong heteronuclear coupling to dephase the protonated carbons. Immediately following the cross-polarization step, a short period of typically 50 ms is applied during which the proton decoupling is turned off. The signals from strongly dipolar coupled carbons are completely dephased during this interrupted decoupling period, whereas the weakly coupled quaternary and methyl carbons are largely unaffected (Figure 6b). Despite their short proton–carbon distances, methyl carbons are only weakly coupled due to the self-decoupling effect of fast steering wheellike rotations around the C–C bond, taking place even in the solid state. Assignment ambiguities may arise when methylene carbons resonate close to methines. These groups have similar dipolar couplings and are not differentiated by either dipolar dephasing or short CP editing. More complex dipolar coupling-based editing techniques can be applied in these instances.98–110 Editing techniques, relying on spin-spin J-coupling rather a dipolar coupling, have also been introduced.111–112 These are essentially solid-state versions of the solution attached proton test (ATP) experiment, in which signals from quaternary carbons and methylene groups are of opposite phase from those of methine and methyl groups. For this experiment to work, J-coupled multiplets must be resolved in the solid state. Further help with the assignments can be gained from two-dimensional techniques, and there are many variants of multidimensional correlation

62

H.G. BRITTAIN AND A. MEDEK

experiments. Information-rich homonuclear correlations between carbons, which are based on the solids analog of the solution INADEQUATE experiment, require isotope labeling and as such are only of limited use for pharmaceutical applications.111,113–114 Heteronuclear correlations between protons and carbons or nitrogens can be achieved by through-space dipolar couplings 109,115–117 or through-bond J-couplings.93,113,118–121 The dipolar coupling based solid-state version of the solution HETCOR (Heteronuclear Correlation) experiment includes not only correlations between directly bonded carbons and protons but also correlations between protons and carbons close in space. Therefore, quaternary carbons show correlations in the dipolar version but not in the J-coupled version. In both cases, the resolution in the proton domain of these spectra suffers from the same limitations as for the CRAMPS experiment, restricting the limits of these correlation experiments for assignment purposes. Another group of experiments correlates carbon isotropic shifts with anisotropic contributions from the same nuclei.122–125 The assignment power of these experiments stems from the fact that different types of carbons can have similar isotropic shielding but different anisotropic chemical shifts. Therefore, based on the magnitude of the chemical shift anisotropy, the carbons in question can be assigned. As an added advantage, the chemical shift anisotropy could, in many cases, be more sensitive than isotropic chemical shifts to structural variations between polymorphs. Empirical as well as ab-initio quantum mechanical calculations of chemical shifts are likely to become an important tool for resonance assignments. Specifically, the ab-initio–type calculations are likely to prove useful in predicting conformations of polymorphs by comparing the, derived from guessed 3-D molecular structure inputs, with experimental chemical shifts.

B. SS-NMR Studies of Polymorphism Polymorphs are typically differentiated by SS-NMR based on their chemical shift differences, but other NMR properties (such as relaxation) can be used as well. Chemical shifts are very sensitive to molecular conformations and crystal packing. In order for the polymorphs to be distinguished by SS-NMR, they must show differences in either conformations or crystal packing. Even though many pharmaceutical applications of SS-NMR can be found in the literature,122–123,126–154 the full potential of this technique has not yet been realized. Several review articles have been published offering a comprehensive review of the literature in this field.55,155–160 A powerful technique for the study of polymorphism is single crystal X-ray diffraction, since it provides a detailed description of the crystal structure, including molecular conformation and crystal packing. However, the technique requires good-quality single crystals, which may not be available in many instances. The need for single crystals also precludes application of this technique to the routine analysis of polymorph mixtures. Both of these limitations are avoided through the use of X-ray

3 POLYMORPHIC AND SOLVATOMORPHIC IMPURITIES

63

powder diffraction (XRPD), which has established itself as a standard in polymorph determination. Solid-state SS-NMR and XRPD take advantage of different phenomena to evaluate polymorphism. XRPD is sensitive to differences in unit cell dimensions, whereas SS-NMR is sensitive to conformational changes and differences in magnetic environments.94 One can think of two hypothetical situations. If two polymorphic forms differ exclusively in their unit cell dimensions but the conformation of the molecule is preserved, then the XRPD patterns are likely to be substantially different, whereas SS-NMR may not register any changes. On the other hand, if the unit cell dimensions are preserved while the molecular conformation changes substantially, SS-NMR is likely to pick up the differences and the XRPD patterns might be the same. In this sense, both techniques provide information that is complementary to the other.94 1. Advantages of the SS-NMR Approach Solid-state NMR has several important advantages over other spectroscopic techniques that are commonly used to characterize polymorphism. Among these advantages are its sensitivity and high degree of resolution. Even minor changes of conformation or crystal packing among polymorphs can produce different local environments and consequently different chemical shifts (Figure 7). SS-NMR can often be more sensitive for polymorph detection than other techniques, including XRPD, IR, NIR, or Raman spectroscopies. For most of the solvates shown in Figure 7, XRPD shows virtually identical traces. Solid-state NMR can clearly differentiate these forms based on their 13C shifts. In contrast to more established spectroscopic techniques, SS-NMR spectra are virtually independent of the physical properties of the sample, such as particle size, homogeneity, or residual water content. Therefore, pharmaceutical solids can be studied by NMR without a need for special sample preparations. Samples viable for SS-NMR analysis include a whole range of pharmaceutical formulations such as tablets, lyophilized powders, capsules, suspensions, and ointments. SS-NMR does not suffer from a preferred orientation restriction, which often leads to an incorrect identification of polymorphs when using XRPD. In addition, SS-NMR is a nondestructive technique that allows other analyses to be performed on the same sample after the NMR spectrum is acquired. Another advantage of SS-NMR is that the observed chemical shift differences between polymorphs can be related to particular molecular sites based on known assignments (see Figure 6). Site-specific mobility can also be determined by probing the relaxation properties of the compound. This is an important application, since mobility is usually related to polymorphic interconversions and solid-state reactions.94 Solid-state reactions between active pharmaceutical ingredients and excipients can also be followed by SSNMR. The study of polymorphic transitions can be performed by variable temperature experiments. Probably the most important advantage of SS-NMR for pharmaceutical applications is its suitability for the analysis of complex formulations.

64

H.G. BRITTAIN AND A. MEDEK

FIGURE 7 125 MHz 13C spectra of different solvates of the same compound. Solvents used to prepare the different solvatomorphs were (A) water, (B) water and cyclohexane, (C) water and MTBE, (D) water (different stoichiometry from (A)), (E) water and i-propanol, (F) water and n-propanol, (G) water and ethanol, (H) water and propylene glycol. Spectrum (I) represents an amorphous form. Unlike SS-NMR spectra, the XRPD patterns of forms D through H are virtually identical.

Typically, excipients found in drug formulations resonate in limited spectral regions. Most excipient peaks can be found between 60 and 100 ppm, and as such do not completely overlap the active ingredient resonances. Spectral regions containing only drug signals can often be found even in complex formulation. 2. Limit of Detection A particular form or a mixture of polymorphs is identified based on comparison of the sample NMR spectrum with standard spectra of

3 POLYMORPHIC AND SOLVATOMORPHIC IMPURITIES

65

each component in its pure form. If one or more fully resolved peaks can be found in the spectrum, the limit of detection is dictated solely by the overall signal-to-noise ratio. The well known drawback of NMR is its lack of sensitivity (defined as signal-to-noise ratio, as opposed to sensitivity to changes of electronic environment discussed in the previous paragraph). Assuming sufficiently short acquisition times on the order of one hour, the typical limit of detection of most natural abundance 13C pharmaceuticals is approximately 1% (see Figure 8). In favorable cases, when internal mobility is present and sharp lines are observed, the limit of detection may be even lower. The sensitivity of SS-NMR can also be increased by partial or uniform spin labeling. This, however, would limit the application of SS-NMR only to investigative samples, excluding the analysis of production lots. Nuclei other than carbon can be pursued. Although many pharmaceuticals contain nitrogen, the 15N isotope has low natural abundance and poor relative sensitivity. 15N is, unlike the more abundant quadrupolar 14N, a well-behaved spin 1⁄2 nucleus. Unless isotope enrichment can be applied, its low natural abundance severely limits 15N SS-NMR applications of pharmaceutical utility.

FIGURE 8 125 MHz 13C spectrum of 1% solvate (A) in otherwise pure solvate (D) of the compound shown in Figure 7. Pure solvate (A) is shown as an overlay for comparison. The experimental time was approximately one hour, and 1024 scans were acquired.

66

H.G. BRITTAIN AND A. MEDEK

FIGURE 9 470 MHz 19F spectra of two different polymorphs of the same compound. The compound contains one fluorine atom per molecule, which is directly bonded to an aromatic part of the molecule. 35 kHz spinning speed with no proton decoupling was used in these examples (2.5 mm spinners containing approximately 10 mg of sample). Because of relatively unfavorableT1 relaxation properties, 32 scans were acquired in approximately one hour total time.

However, many active pharmaceutical ingredients contain fluorine. 19F is almost as sensitive a nucleus as 1H, but does not face the same empirical limitations as protons. Since the number of fluorine atoms per molecule is usually small, homonuclear 19F–19F couplings are relatively weak and can be removed by conventional magic angle spinning. However, proton decoupling and/or very fast magic angle spinning have to be applied to cancel the strong 1H–19F dipolar interaction, sometimes necessitating the use of special NMR hardware153 (see Figure 9). The presence of fluorine also enables measurements of fluorine–carbon distances by the rotational echo double resonance161–165 (REDOR) experiment without the need for isotope enrichment. This approach provides information similar to single crystal X-ray data for polymorph conformational analysis. Other sensitive but less frequently occurring pharmaceutical nuclei such as phosphorus, bromine, and sodium can be used. The latter two are characterized by quadrupolar nuclear spin. Advanced experiments, such as multiple quantum magic angle spinning (MQMAS) may be required to obtain narrow signals in these instances.166,167 3. Polymorphs Versus Multiple Sites per Unit Cell A problem often arises when multiple lines are encountered for some or all carbons in the sample. The multiple lines can be interpreted as either

3 POLYMORPHIC AND SOLVATOMORPHIC IMPURITIES

67

evidence for a mixture of several polymorphs, or conversely, they could indicate the presence of multiple nonequivalent molecules in the unit cell. To differentiate between these two instances, relative peak intensities can be used as a first approximation when pure form standards are absent. If the peak ratios of all the carbons appearing as multiplets are approximately equal to one, then multiple sites are likely to be present. This is, however, not a proof of multiple sites since, by coincidence, one could have an equal mixture of different polymorphs. In some cases, the buildup dynamics of CP, which is a dependence of CP-MAS signal intensity on cross-polarization time, can be used to differentiate polymorphs from multiple sites.156 However, it is not straightforward to analyze the CP buildup intensity differences between carbons from the same molecular sites in terms of polymorphs versus multiple sites. In carbon detected proton relaxation experiments, observed differences in proton relaxation typically indicate presence of a polymorphic mixture. In contrast, crystals containing multiple sites show the same relaxation for all protons, as the short distance between the multiple sites makes the proton spin diffusion possible. Positive proof for the existence of polymorphs can be gained by varying the relative concentration of the two forms and observing the varying intensity ratios of corresponding carbons. This in turn requires sample manipulations, such as varying the temperature, humidity, or other conditions to change the ratio of polymorphs.

C. Quantitative Analysis of Polymorphs In contrast to solution-phase or direct polarization solid-state NMR experiments, CP-MAS is not a truly quantitative technique. As a result of the cross-polarization step, the integral intensities of peaks in CP-MAS spectra do not directly reflect the stoichiometry of nuclei in the molecule. CP-MAS signals originate from protons, and its transfer efficiency varies from carbon to carbon, based on proximity of protons and local mobility of the molecule. In most pharmaceutical applications, only relative concentrations of mixtures of two or more polymorphs need to be determined even in complex dosage forms. The absolute concentration of the drug in formulation is usually known or can be determined by other techniques. Several approaches can be followed when quantifying polymorphic composition. The most reliable, though laborious, method is quantification by means of a calibration curve. When pure forms are available, solid-state NMR spectra of mixtures with different known concentrations of polymorphs are acquired, and a calibration curve is constructed. To avoid any dependence on the precise setting of all the experimental parameters, the calibration curve can be plotted as relative signal ratio of the two forms rather than as absolute intensities. Since this method does not rely on absolute signal intensities, the calibration graph constructed on pure drug polymorph mixtures can be applied to formulations, where the active ingredient is diluted by excipients. Logarithmic or other transformation of the resulting nonlinear calibration curves is desirable to optimize the fit.

68

H.G. BRITTAIN AND A. MEDEK

When no pure forms are available or when a quick, approximate estimate of the polymorph content is sufficient, the intensity ratio of similar carbon peaks, each belonging to a different form, can be used for quantification. In general, carbons with very close chemical shifts are likely to have similar chemical environments. This can be explicitly confirmed if chemical shift assignments are known. Therefore, assuming the same number of protons in close vicinity of these carbons, the CP intensities are likely to be similar, reflecting the approximate quantitative ratio of the two forms. More precise results can be achieved by fitting the cross-polarization buildup curve168 in terms of T1 (rotating frame relaxation time during the spin-locking pulse, which characterizes decrease of the signal during CP period due to the spin relaxation) and TCH (cross-polarization time constant, which characterizes the CP signal buildup speed). After the two CP constants T1 and TCH are calibrated on pure standards of all polymorphs present in the mixture, the known CP characteristics can be used to quantify the unknown mixtures. If the total experimental time is not critical, direct observation of carbon signal without the CP step can be used. Although the direct polarization carbon signal is fully quantitative, its signal-to-noise ratio will be inferior to that for CP-MAS because of the reasons discussed above. Not only is the signal per scan weaker, but also the relaxation delay must be set much longer. Alternatively, when the carbon relaxation properties of both polymorphs are known, the intensities can be corrected for relaxation without waiting for full relaxation between successive scans. The easiest and usually most precise quantification of polymorphic composition is achieved when a sensitive heteronucleus (such as fluorine or phosphorus) is present in the molecule, and its spectrum shows chemical shift differences between the polymorphs. Direct polarization experiments of these nuclei are sensitive, and therefore the polymorphs can be quantified quickly directly from the peak ratios. The limit of quantification can be well below 1% of one form in the presence of other forms. As an added advantage, the spectra of these nuclei are usually very simple to analyze, since only a very limited number of resonances are typically present in the spectrum. Figure 9 shows the 19F spectra of two polymorphs of the same compound. If these were both present in a mixture, levels of less than 1% could be reliably quantified with acquisition times of approximately one hour, despite the relatively unfavorable T1 relaxation properties encountered for this compound. Many pharmaceutical applications involve quantification of an amorphous phase in the presence of a crystalline phase of the same compound. Even though amorphous spectra are often substantially different from their crystalline counterparts, it is difficult to precisely quantify small amounts of amorphous phase. The amorphous peaks present in low concentration can be lost in the baseline because of their broad line character. Quantification is generally achieved by modeling the resulting spectrum as a sum of pure form standard spectra. Alternatively, if pure forms are not available, one can take advantage of the different properties of amorphous and crystalline phases

3 POLYMORPHIC AND SOLVATOMORPHIC IMPURITIES

69

such as different relaxation rates169,170 to selectively filter one component out of the spectrum. The glass transition temperature (Tg) of amorphous compounds is likely to be lower than the melting temperature of the crystalline phase. Heating the sample above Tg therefore results in observation of only the crystalline phase. The signal from the amorphous phase is lost, since this phase reaches a liquidlike mobility, and therefore the dipolar coupling that the CP-MAS experiment relies on is averaged out to zero.

REFERENCES 1. Verma, A. R. and Krishna, P. Polymorphism and Polytypism in Crystals, John Wiley & Sons, New York, pp. 1–7, 1966. 2. Byrn, S., Pfeiffer, R., Ganey, M., Hoiberg, C. and Poochikian, G. Pharm. Res. 12:945–954, 1995. 3. Brittain, H. G. Pharm. Tech. 24(9):116–125, 2000. 4. Byrn, S. R., Pfeiffer, R. R. and Stowell, J. G. Solid State Chemistry of Drugs, 2nd ed. SSCI Inc., West Lafayette, IN, 1999. 5. Brittain, H. G. Polymorphism in Pharmaceutical Solids, Marcel Dekker, New York, 1999. 6. Brittain, H. G., Bogdanowich, S. J., Bugay, D. E., DeVincentis, J., Lewen, G. and Newman, A. W. Pharm. Res. 8:963–973, 1991. 7. Brittain, H. G. Physical Characterization of Pharmaceutical Solids, Marcel Dekker, New York, 1995. 8. Threlfall, T. L. Analyst, 120:2435–2460, 1995. 9. Brittain, H. G. Methods for the Characterization of Polymorphs and Solvates, Chapter 6. In Polymorphism in Pharmaceutical Solids (Brittain, H. G., Ed.), Marcel Dekker, New York, pp. 227–278, 1999. 10. Brittain, H. G. Solid-State Analysis, Chapter 3. In Handbook of Modern Pharmaceutical Analysis (S. Ahuja and Scypinski, S., Eds.), pp. 57–84. Marcel Dekker, New York, 2001. 11. Boles, M. O., Girven, R. J. and P. Gane, A. C., Acta Cryst., B34:461, 1978. 12. Suryanarayanan, R. X-Ray Powder Diffractometry, Chapter 7. In Physical Characterization of Pharmaceutical Solids (Brittain, H.G., Ed.) Marcel Dekker, New York, pp. 199–216, 1995. 13. Bragg, W. H. and Bragg, W. L. X-Rays and Crystal Structure, G. Bell & Sons, London, 1918. 14. Woolfson, M. M. X-Ray Crystallography, Cambridge University Press, Cambridge, 1970. 15. Stout, G. H. and Jensen, L. H. X-Ray Structure Determination: A Practical Guide, Macmillan Co., New York, 1968. 16. Klug, H. P. and Alexander, L. E. X-Ray Diffraction Procedures for Polycrystalline and Amorphous Materials, 2nd ed., John Wiley & Sons, New York, 1974. 17. Chung, F. K. and Smith, D. K. Industrial Applications of X-Ray Diffraction, Marcel Dekker, New York, 2000. 18. Shell, J. W. J. Pharm. Sci., 52:24–29, 1963. 19. Kidd, W. C., Varlashkin, P. and Li, C.-Y. Powder Diffraction, 8:180–187, 1993. 20. Pirtimaki, J., Lehto, V.-P. and Laine, E. Drug Dev. Indust. Pharm., 19:2561–2577, 1993. 21. Iyengar, A. S., Phadnis, N. V. and Suryanarayanan, R. Powder Diffraction, 16:20–24, 2001. 22. Suryanarayanan, R., Pharm. Res. 6:1017–1024, 1989. 23. Tanninen, V.P. and Yliruusi, J. Int. J. Pharm., 81:169–177, 1992. 24. Bugay, D. E., Newman, A. W. and Findlay, W. P. J. Pharm. Biomed. Anal. 15:49–61, 1996. 25. Agatonovic-Kustrin, S., Rades, T., Wu, V., Saville, D. and Tucker, I. G. J. Pharm. Biomed. Anal., 25:741–750, 2001. 26. Otsuka M. and Matsuda, Y., Drug Dev. Indust. Pharm. 19:2241–2269, 1993. 27. Otsuka, M., Otsuka, K. and Kaneniwa, N. Drug Dev. Indust. Pharm. 20:1649–1660, 1994. 28. de Villiers, M. M., van der Watt, J. G. and Lotter, A. P. Drug Dev. Indust. Pharm. 17:1295– 1303, 1991. 29. Epple, M. and Cammenga, H. K. Ber. Bunsenges. Phys. Chem. 96:1774–1778, 1992.

70

H.G. BRITTAIN AND A. MEDEK

30. Kitamura, S., Tada, T., Okamoto, Y. and Yasuda, T. Pharm. Res. 9:138–142, 1992. 31. Nakamura, T., Sameshima, K., Okunaga, K., Sugiura, Y. and Sato, J. Powder Diffraction, 4:9–13, 1989. 32. Ryan, J. A. J. Pharm. Sci., 75:805–807, 1986. 33. Pirttimaki, J. Laine, E., Ketolainen, J. and Paronen, P. Int. J. Pharm. 95:93–99, 1993. 34. Crocker, L. S. and McCauley, J. A. J. Pharm. Sci. 84:226–227, 1995. 35. de Villiers, M. M., Wurster, D. E., Van der Watt, J. G. and Ketkar, A. Int. J. Pharm. 163:219–224, 1998. 36. Li, Z. G., Liang, L., Harlow, R. L., Foris, C. M., Olson, R. E., Sielecki, T. M., Liu, J., Vickery, R. D. and Maurin, M. B. J. Pharm. Sci. 89:1237–1242, 2000. 37. Brittain, H. G. Pharm. Res. 7:683–690 (1990) Structural Aspects of Molecular Dissymmetry, Chapter 9. In Polymorphism in Pharmaceutical Solids (Brittain, H.G., Ed.) Marcel Dekker, New York, pp. 363–393, 1999. 38. Phadnis, N. V. and Suryanarayanan, R. Pharm. Res. 14:1176–1180, 1997. 39. Wendlandt, W. W. Thermal Analysis, 3rd ed., Wiley-Interscience, New York, 1986. 40. Dollimore, D. Thermoanalytical Instrumentation, Chapter 25. In Analytical Instrumentation Handbook, (Ewing, G. W., Ed.), pp. 905–960, Marcel Dekker, New York, 1990. 41. Giron, D. J. Pharm. Biomed. Anal. 4:755–770, 1986. 42. Ford, J. L. and Timmins, P. Pharmaceutical Thermal Analysis, Ellis Horwood, Ltd., Chichester, 1989. 43. McCauley, J. A. and Brittain, H. G. Thermal Methods of Analysis, Chapter 8. In Physical Characterization of Pharmaceutical Solids (Brittain, H. G., Ed.), Marcel Dekker, New York, pp. 223–251, 1995. 44. Duval, C. Inorganic Thermogravimetric Analysis, 2nd ed., Elsevier, Amsterdam, 1963. 45. Keattch, C. J. and Dollimore, D. Introduction to Thermogravimetry, 2nd ed., Heyden, London, 1975. 46. Dollimore, D. Thermogravimetry, In Thermal Analysis—Techniques and Applications, (Charsley, E. L. and Warrington, S. B., Eds.) pp. 31–58, Royal Society of Chemistry, London, 1992. 47. Brittain, H. G., Bogdanowich, S. J., Bugay, D. E., DeVincentis, J., Lewen, G. and Newman, A.W. Pharm. Res. 8:963–973, 1991. 48. Brittain, H. G. unpublished results. 49. Mackenzie, R. C. Pure Appl. Chem. 57:1737–1740, 1985. 50. Sabbah, R., Xu-Wu, A., Chickos, J. S., Planas Leitao, M. L., Roux, M. V. and Torres, L. A. Thermochim. Acta, 331:93–204, 1999. 51. Grant, D. J. W. Theory and Origin of Polymorphism, Chapter 1. In Polymorphism in Pharmaceutical Solids (Brittain, H. G., Ed.), Marcel Dekker, New York, pp. 1–33, 1999. 52. Khankari, R. K., Law, D. and Grant, D. J. W. Int. J. Pharm. 82:117–127, 1992. 53. Markovich, R. and Pidgeon, C. Pharm. Res. 8:663–675, 1991. 54. Bugay, D. E. and Williams, A. C. Vibrational Spectroscopy, Chapter 3. In Physical Characterization of Pharmaceutical Solids (Brittain, H. G., Ed.), Marcel Dekker, New York, pp. 59–91, 1995. 55. Brittain, H. G. J. Pharm. Sci. 86:405–412, 1997. 56. Findlay, W. P. and Bugay, D. E. J. Pharm. Biomed. Anal. 16:921–930, 1998. 57. Coates, J. Instrumentation for Infrared Spectroscopy, Chapter 7. In Analytical Instrumentation Handbook (Ewing, G. W., Ed.), Marcel Dekker, New York, pp. 233–279, 1990. 58. Skrdla, P. J., Antonucci, V., Crocker, L. S., Wenslow, R. M., Wright, L. and Zhou, G. J. Pharm. Biomed. Anal. 25:731–739, 2001. 59. Bruno, T. J. Appl. Spect. Rev. 34:91–120, 1999. 60. Brimmer, P. J. and Griffiths, P. R. Anal. Chem. 58:2179–2184, 1986. 61. Frei, R. W. and MacNeil, J. D. Diffuse Reflectance Spectroscopy in Environmental ProblemSolving, pp. 3–19, CRC Press, Cleveland, 1973. 62. Hartauer, K. J., Miller, E. S. and Guillory, J. K. Int. J. Pharm. 85:163–174, 1992. 63. Sarver, W. W., Meulman, P. A., Bowerman, D. K. and Havens, J. L. Int. J. Pharm. 167:105– 120, 1998. 64. Urban, M. W., Attenuated Total Reflectance Spectroscopy of Polymers, American Chemical Society, Washington DC, 1996.

3 POLYMORPHIC AND SOLVATOMORPHIC IMPURITIES

71

65. Salari, A. and Young, R. E. Int. J. Pharm. 163:157–166, 1998. 66. Grasselli, J. G., Snavely, M. K. and Bulkin, B. J. Chemical Applications of Raman Spectroscopy, John Wiley & Sons, New York, 1981. 67. Lewis, I. R. and Edwards, H. G. M. Handbook of Raman Spectroscopy, Marcel Dekker, New York, 2001. 68. Huong, P. V. J. Pharm. Biomed. Anal., 4:811–823, 1986. 69. Frank, C. J. Review of Pharmaceutical Applications of Raman Spectroscopy, Chapter 6. In Analytical Applications of Raman Spectroscopy (Pelletier, M. J., Ed.), Blackwell Science, Oxford, pp. 224–275, 1999. 70. Tudor, A. M., Church, S. J., Hendra, P. J., Davies, M. C. and Melia, C. D. Pharm. Res. 10:1772–1776, 1993. 71. Langkilde, F. W., Sjoblom, J., Tekenbergs-Hjelte, L. and Mrak, J. J. Pharm. Biomed. Anal. 15:687–696, 1997. 72. Wang, F., Wachter, J. A., Antosz, F. J. and Berglund, K. A. Org. Proc. Res. Dev. 4:391– 395, 2000. 73. Taylor, L. S., Williams, A. C. and York, P. Pharm. Res. 15:1207–1214, 1998. 74. Taylor, L. S. and Zografi, G. Pharm. Res. 15:755–761, 1998. 75. Stark, E., Luchter, K. and Margoshes, M. Appl. Spect. Rev. 22:335–399, 1986. 76. Ciurczak, E. W. Appl. Spect. Rev. 23:147–163, 1987. 77. Morisseau, K. M. and Rhodes, C. T. Drug Dev. Indust. Pharm. 21:1071–1090, 1995. 78. Burns, D. A. and Ciurczak, E. W. Handbook of Near-Infrared Analysis, 2nd ed., Marcel Dekker, NY, 2001 79. Gimet, R. and Luong, A. T. J. Pharm. Biomed. Anal. 5:205–211, 1987. 80. Luner, P. E., Majuru, S., Seyer, J. J. and Kemper, M. S. Pharm. Dev. Tech. 5:231–246, 2000. 81. Patel, A. D., Luner, P. E. and Kemper, M. S. J. Pharm. Sci. 90:360–370, 2001. 82. Norris, T. and Santafianos, D. J. Chem. Soc., Perkin Trans. 2:2498–2502, 2000. 83. Rasanen, E., Rantanen, J., Jorgensen, A., Karjaalainen, M., Paakkari, T. and Yliruusi, J. J. Pharm. Sci. 90:389–396, 2001. 84. Andrew, E. R., Bradbury, A. and Eades, R. G. Nature, 182:1659, 1958. 85. Lowe, I. J. J. Phys. Rev. Lett. 2:285, 1959. 86. Maricq, M. M. and Waugh, J. S. J. Chem. Phys. 70:3300, 1979. 87. Hartmann, S. R. and Hahn, E. Phys. Rev. 128:2042, 1962. 88. Pines, A., Gibby, M. G. and Waugh, J. S. J. Chem. Phys. 59:569, 1973. 89. Mehring, M. High Resolution NMR in Solids, 2nd ed., Springer, Berlin, 1983. 90. Schmidt-Rohr, K. and Spiess, H. W. Multidimensional Solid-State NMR and Polymers, Academic Press, NY, 1994. 91. Abragam, A. The Principles of Nuclear Magnetism, Oxford University Press, Oxford, 1985. 92. Taylor, R. E., Pembleton, R.G., Ryan, L. M. and Gerstein, B. C. J. Chem. Phys. 71:4541, 1979. 93. Lesage, A., Charmont, P., Steuernagel, S. and Emsley, L. J. Am. Chem. Soc. 122:9739– 9744, 2000. 94. Padden, B. E., Ph.D thesis, University of Minnesota pp. 221, 2000. 95. Fyfe, C. A. Solid State NMR for Chemists, CFC Press, Ontario, 1983. 96. Alla, M. and Lippmaa, E.L. Chem. Phys. Lett., 37:260, 1976. 97. Opella, S. J., Frey, M. G. and Cross, T. A. J. Am. Chem. Soc. 101:5854, 1979. 98. Wu, X. and Zilm, K. J. Mag. Res. 102:205, 1993. 99. Rossi, P., Subramanian, R. and Harbison, G. S. J. J. Mag. Res. 141:159–163, 1999. 100. Pan, H. J. Mag. Res. 124:1, 1997. 101. Sethi, N. K. J. Mag. Res. 94:352, 1991. 102. Burum, D. P. and Bielecki, A. J. Mag. Res. 95:184, 1991. 103. Webb, G. G. and Zilm, K.W. J. Am. Chem. Soc. 111:1455, 1989. 104. Burns, S. T., Wu, X. and Zilm, K. W. J. Mag. Res. 143:352, 2000. 105. Wu, X., Burns, S. T. and Zilm, K. W. J. Mag. Res. 111:29, 1994. 106. Sangill, R., Rastrup-Andersen, N., Bildse, H., Jakobsen, H. J. and Nielsen, N. C. J. Mag. Res. 107:67, 1994. 107. Hu, J. Z., Harper, J. K., Taylor, C., Pugmire, R. J. and Grant, D. J. W. J. Mag. Res. 142, 326, 2000.

72

H.G. BRITTAIN AND A. MEDEK

108. DeVita, E. and Frydman, L. J. Mag. Res. 148:327–337, 2001. 109. Saalwachter, K., Graf, R. and Spiess, H. W. J. Mag. Res. 148:398–418, 2001. 110. Kumashiro, K. K., Niemczura, W. P., Kim, M. S. and Sandberg, L. B. J. Biomol. NMR, 18:139–144, 2000. 111. Lesage, A., Auger, C., Caldarelli, S. and Emsley, L. J. Am. Chem. Soc. 119:7867–7868, 1997. 112. Lesage, A., Steuernagel, S. and Emsley, L. J. J. Am. Chem. Soc. 120:7095–7100, 1998. 113. Lesage, A., Bardet, M. and Emsley, L. J. J. Am. Chem. Soc. Emsley, 121:10987–10993, 1999. 114. Verel, R., van Beek, J. D. and Meier, B. H. J. Mag. Res. 140:300–303, 1999. 115. van Rossum, B. J., Forster, H. and der Groot, H. J. M. J. Mag. Res. 124:516–519, 1997. 116. Yao, X. L., Schmidt-Rohr, K. and Hong, M. J. Mag. Res. 149:139–143, 2001. 117. Wilhelm, M., Feng, H., Tracht, U. and Spiess, H. W. J. Mag. Res. 134:255–260, 1998. 118. Baldus, M., Iuliucci, R. J. and Meier, B. H. J. Am. Chem. Soc. 119:1121–1124, 1997. 119. Hardy, E. D., Verel, R. and Meier, B. H. J. Mag. Res. 148:459–464, 2001. 120. Lesage, A., Sakellariou, D., Steuernagel, S. and Emsley, L. J. J. Am. Chem. Soc. 120:13194– 13201, 1998. 121. Lesage, A. and Emsley, L. J. J. Mag. Res. 148:449–454, 2001. 122. Harper, J. K. and Grant, D. M. J. Am. Chem. Soc. 122:3708–3714, 2000. 123. Smith, J., MacNamara, E., Raftery, D., Borchardt, T. and Byrn, S. J. Am. Chem. Soc. 120:11710–11713, 1998. 124. Frydman, L., Chingas, G. C., Lee, Y. K., Grandinetti, P. J., Eastman, M. A., Barrall, G. A. and Pines, A. J. J. Chem. Phys. 97:4800–4808, 1992. 125. DeLacroix, S. F., Titman, J. J., Haemeyer, A. and Spiess, H. W. J. Mag. Res. 97:435–443, 1992. 126. Byrn, S., Pfeiffer, R., Ganey, M., Hoiberg, C. and Poochikian, G. Pharm. Res. 12:945–954, 1995. 127. Zell, M. T., Padden, B. E., Grant, D. J. W., Chapeau, M. C., Prakash, I. and Munson, E. J. Am. Chem. Soc. 121:1372–1378, 1999. 128. Isogai, A., Kato, T., Uryu, T. and Atalla, T. Carbohydrate Polymers, 21:277–281, 1993. 129. Stephenson, G. A., Stowell, J. G., Toma, P. H., Dorman, D. E., Greene, J. R. and Byrn, S. R. J. Am. Chem. Soc. 116:5766–5773, 1993. 130. Anulewicz, R., Wawer, I., Piekarska-Bartoszewicz, B. and Temeriusz, A. Carbohydrate Res. 281:1–10, 1996. 131. Anulewicz, R., Wawer, I., Piekarska-Bartoszewicz, B. and Temeriusz, A. Carbohydrate Res. 304:335–340, 1997. 132. Hirai, A., Horii, F., Kitamaru, R., Fatou, J. G. and Bello, A. Macromolecules, 23:2913– 2917, 1990. 133. Markovic, R. J., Evans, C. A., Coscolluela, C. B., Zibas, S. A. and Rosen, J. J. Pharm. Biomed. Anal. 16:661–673, 1997. 134. Zell, M. T., Padden, B. E., Grant, D. J. W., Schroeder, S. A., Wachholder, K. L., Prakash, I. and Munson, E. Tetrahedron, 56:6603–6616, 2000. 135. Padden, B. E., Zell, M. T., Dong, Z., Schroeder, S. A., Grant, D. J. W. and Munson, E. Anal. Chem. 71:3325–3331, 1999. 136. Strohmeier, M., Orendt, A. M., Alderman, D. W. and Grant, D. M. J. Am. Chem. Soc. 123:1713–1722, 2001. 137. McGeorge, G., Harris, R. K., Chippendale, A. M. and Bullock, J. F. J. Chem. Soc. Perkin Trans. 2:1733–1738, 1996. 138. Middleton, D. A., Duff, C. S. L., Peng, X., Reid, D. G. and Saunders, D. J. Am. Chem. Soc. 122:1161–1170, 2000. 139. Gao, P. Pharm. Res. 15:1425–1433, 1998. 140. Rohrs, B. R., Thamann, T. J., Gao, P., Stelzer, D. J., Berger, M. S. and Chao, R. S. Pharm. Res. 16:1850–1856, 1999. 141. Zumbulyadis, N., Antalek, B., Windig, W., Scaringe, R. P., Lanzafame, A. M., Blanton, T. and Helber, M. J. Am. Chem. Soc. 121:11554–11557, 1999. 142. Gustafsson, C., Lennholm, H., Iversen, T. and Nystrom, C. Int. J. Pharm. 174:243–252, 1998. 143. Glaser, R. and Shiftan, D. J. Org. Chem. 64:9217–9224, 1999.

3 POLYMORPHIC AND SOLVATOMORPHIC IMPURITIES

73

144. Schmidt, A., Kababya, S., Appel, M., Khatib, S., Botoshansky, M. and Eichen, Y. J. Am. Chem. Soc. 121:11291–11299, 1999. 145. Glaser, R., Novoselsky, A., Shiftan, D. and Drouin, M. J. Am. Chem. Soc. 65:6345–6353, 2000. 146. Bobek, M. M., Giester, G., Kahlig, H. and Brinker, U. H. Tetrahedron Lett. 41:5663–5667, 2000. 147. Chen, L. R., Padden, B. E., Vippagunta, S. R., Munson, E. and Grant, D. J. W. Pharm. Res. 17:619–624, 2000. 148. Tang, H. R., Belton, P.S., Davies, S. C. and Hughes, D. L. Carbohydrate Res. 330:391–399, 2001. 149. Henck, J. O., Finner, E. and Burger, A. J. Pharm. Sci. 89:1151–1159, 2000. 150. Wenslow, R. M., Baum, M. W., Ball, R. G., McCauley, J. A. and Varsolona, R. J. J. Pharm. Sci. 89:1271–1285, 2000. 151. Crowley, K. J., Forbes, R. T., York, P., Apperley, D. C., Nyqvist, H. and Camber, O. J. Pharm. Sci. 89:1286–1295, 2000. 152. Garbow, J. R., Likos, J. J. and Schroeder, S. A. J. Agric. Food Chem. 49:2053–2060, 2001. 153. Campbell, S. C., Harris, R. K., Hardy, M. J., Lee, D. C. and Busby, D. J. J. Chem. Soc. Perkin Trans. 2:1913–1918, 1997. 154. Medek, A. and Frydman, L. J. Am. Chem. Soc. 122:684–691, 2000. 155. Bugay, D. E. Adv. Drug Del. Rev. 48:43–65, 2001. 156. Stephenson, G. A., Forbes, R. A. and Reutzel-Edens, S. M., Adv. Drug Del. Rev. 48:67–90, 2001. 157. Bugay, D. E. Pharm. Res. 10:317–327, 1993. 158. Holzgrabe, U., Diehl, B. W. K. and Wawer, I. J. Pharm. Biomed. Anal. 17:557–616, 1998. 159. Brittain, H. G., Bogdanowich, S. J., Bugay, D. E., DeVincentis, J., Lewen, G. and Newman, A. W. Pharm. Res. 8:963–973, 1991. 160. Potrzebowski, M. J. New Adv. Anal. Chem. P1:359–404, 2000. 161. Guillon, T. and Schaefer, J. Adv. Mag. Res. 13:57–83, 1989. 162. Wooley, K. L., Klug, C. A., Tasaki, K. and Schaefer, J. J. Am. Chem. Soc. 119:53–58, 1997. 163. Tong, G. and Schaefer, J. J. Am. Chem. Soc., 121:5238–5248, 1999. 164. Huang, H., Wooley, K. L. and Schaefer, J. Macromolecules, 34:544–546, 2001. 165. Brouwer, E. B., Gougeon, R. D. M., Hirschinger, J., Udachin, K. A., Harris, R. K. and Ripmeester, J. A. Phys. Chem. Chem. Phys. 1:4043–4050, 1999. 166. Harwood, J. S. and Frydman L. J., J. Am. Chem. Soc. 117:5367–5368, 1995. 167. Medek, A., Harwood, J. S. and Frydman, L. J. Am. Chem. Soc. 117:12779–12787, 1995. 168. Gao, P. Pharm. Res. 13:1095–1104, 1996. 169. Mao, J. D., Hu, W. G., Schmidt-Rohr, K., Davies, G., Ghabbour, E. A. and Xing, B. Soil Sci. Soc. Am. J. 64:873–884, 2000. 170. Hu, W. G. and Schmidt-Rohr, K. Polymer 2000:2979–2987, 2000. 171. Pagano, T. G., Marzilli, L. G., Flocco, M. M., Tsai, C., Carrell, H. L. and Glusker, J. P. J. Am. Chem. Soc. 113:531, 1991.

This Page Intentionally Left Blank

4 IMPURITIES IN DRUG PRODUCTS KENNETH C.WATERMAN, ROGER C. ADAMI, AND JINYANG HONG Pfizer, Inc.,Groton,CT 06340

I. II. III. IV. V. VI. VII. VIII. IX. X. XI. XII.

INTRODUCTION WATER PEROXIDES ALDEHYDES METAL IMPURITIES SMALL MOLECULE CARBOXYLIC ACIDS LEACHABLES/EXTRACTABLES ALCOHOLS AS IMPURITIES BIOLOGICAL IMPURITIES ADDITIVES IN EXCIPIENTS FINAL OBSERVATIONS SUMMARY REFERENCES

I. INTRODUCTION Impurities in drug products can come from the drug or from excipients or can be brought into the system through a process step or by contact with packaging. Though numerous impurities could be detected in any drug product sample, there are relatively few impurities that have an influence on the stability or performance of the dosage form. In general, most of the detrimental impurities are small molecules. This is especially true in solid dosage forms where the limited mobility restricts the reactivity of larger molecules. For most drugs, the reactive species consist of water (which can hydrolyze some drugs or affect the dosage form performance), small electrophiles (e.g., aldehydes and carboxylic acid derivatives), peroxides (which can oxidize some drugs), and metals (which can catalyze oxidation and other drug degradation pathways). These impurities are discussed below, with a focus on how these impurities are likely to affect dosage forms via chemical reactivity and physical changes to the systems.

75

76

K.C.WATERMAN et al.

Additionally, some impurities cause toxicological problems. These impurities may not directly affect the performance or stability of a dosage form, but must be controlled to make a safe drug product.

II. WATER Water is the most commonly found impurity in drug products. Water is prevalent in excipients and drug substances and is used in dosage form preparation. In addition, water can be brought into a system from the atmosphere. Not only is water omnipresent in drug products, but it is often detrimental to chemical stability or dosage form performance. Chemical stability issues with water are generally associated with hydrolysis, particularly of carboxylic acid derivatives (esters, amides), acetals, and other susceptible functionalities.1 Even in nonaqueous liquid formulations, water can be present in amounts sufficient to lead to drug degradation. For some drugs, the moisture level will determine if a hydrate can form. In some cases, hydrate formation or generation of an anhydrous form (by having a sufficiently low humidity level) can lead to a loss in crystallinity, which in turn can lead to greater chemical instability of the drug. In addition, changes in crystal morphology may lead to changes in drug dissolution rates. Water is present in a number of excipients as supplied by the vendor. This moisture content can be important when a dosage form is packaged such that equilibration with the environment does not occur. Under these conditions, the moisture brought into the system through the excipients can be sufficient to bring about hydrolysis. Typical moisture contents for a number of solid excipients are listed in Table 1. What is not known is how readily each excipient will give up its moisture; that is, how tightly the water is bound. Water that is tightly bound as a crystalline hydrate is generally less likely to influence hydrolysis than water that is only loosely associated with an excipient. Ultimately, the water activity at the drug surface will determine the hydrolysis rate. This water can also induce hydrolysis of excipients, which in turn can adversely affect the performance of a dosage form.1 When excipients and dosage forms are exposed to various humidity conditions, water vapor sorption will occur. The relative water vapor sorption levels for a number of excipients have been studied.2 It was found that the total water vapor sorption in solid dosage forms can be predicted from the individual contributions of the excipients as powders. The moisture sorption tendency of an excipient will not only depend on the chemical nature of the excipient, but also on the particle size.3 Because of these effects, without packaging protection, water sorption can be significant for a number of dosage forms. Water is often used in the process of preparing dosage forms. For example, lyophilization involves the dissolution or dispersion of a drug and excipients in water, freezing of the mixture and removal of the water by sublimation. The residual moisture level in lyophiles is typically 1–5% by weight, an amount adequate to induce chemical instability with a number

77

4 IMPURITIES IN DRUG PRODUCTS

TABLE 1 Typical Moisture Content of CommonTableting Excipients1 Excipient

Moisture content (wt%)

Comments

Calcium phosphate, dibasic anhydrous

0.1–0.187

Moisture surface adsorbed; does not form hydrate

Calcium phosphate, dibasic dihydrate

20

H2O lost < 100 C

Microcrystalline cellulose (MCC)

5

Moisture content varies with grade; hygroscopic

Colloidal silicon dioxide

0–15 (< 80% RH); 80 (> 80% RH)

Absorbs H2O without liquification; hydrophobic surface treatments decrease hygroscopicity

Crospovidone

 60%

Hygroscopic

Hydroxypropylmethylcellulose

0–10 (< 80% RH); 35 (> 80% RH)

Moisture content depends greatly on previous storage conditions

Lactose, anhydrous

1

Varies with manufacturer and grade

Lactose, monohydrate

4.5–5.5

Varies with manufacturer and grade

Magnesium stearate

5–15 (< 75% RH); 35 (> 75% RH)

Varies with manufacturer and grade

Mannitol

0–1 (< 75% RH); 10 (> 75%RH)

Should be stored tightly closed under low RH

Methylcellulose

5

Slightly hygroscopic; store in cool, dry area; tightly closed

Sodium lauryl sulfate (SLS)

5

Nonhygroscopic

Sodium starch glycolate (Explotab)

5–20 (< 75% RH); 60 (> 75% RH)

Cakes if exposed to high humidity

Stearic acid

< 0.1

Nonhygroscopic

of drugs. The effects of moisture on the lyophile stability include direct hydrolysis reactions, affecting the microenvironmental pH and polarity and affecting the molecular mobility in the solid.1 These effects in lyophiles can lead to hydrolysis rates with more than linear dependency on water concentration.4 Other examples of water used for processing include wet granulations and film coatings. In both cases, process water is later largely removed by a drying step. In some cases, the residual moisture can be an important factor in stability or dosage form performance. The moisture content within dosage forms can influence not only chemical stability but also the dosage form performance. For example, tablet hardness, tensile strength, and porosity vary with moisture content.5–7 The resulting dosage forms can show different properties, especially for controlled-release dosage applications. For example, theophylline release from amylodextrin tablets shows dramatic effects of moisture content.8 These moisture effects are explained as resulting from different degrees of porosity introduced from decompression during tableting, because of different degrees of elastic relaxation. Since water plasticizes amylodextrin,

78

K.C.WATERMAN et al.

the lower Tg as a function of water content allows greater relaxation and formation of denser tablets for a given compression force. In general, higher moisture resulted in lower tablet porosity and less tablet disintegration during dissolution. The moisture content of the component excipients can affect important properties in making the dosage form itself. For example, the moisture content of excipients can affect the flow of powders9 and the compactibility of tablet blends.3,9,10 Many of these effects are explained as resulting from changes to the interparticle and intermolecular forces that are due to moisture sorption and distribution.3,11 For lyophilized dosage forms, there is sometimes an optimum moisture content for stability, especially for proteins. The choice of excipient can determine the moisture content of the lyophilized cakes based on the moisture sorption levels.12 Controlling moisture levels to this optimum also requires that the dosage form be protected from environmental moisture, usually by packaging in glass.

III. PEROXIDES Peroxides are reactive materials present in a number of excipients.13 Peroxides can be found either in the form of hydrogen peroxide (HOOH) or as organic peroxides (ROOH). These species can act to oxidize susceptible drugs via direct reaction with the drugs.14 In addition, peroxides can thermally generate radicals (alone or with metals), which can act to initiate radical chain processes or themselves react with drug. This latter problem is particularly significant with organic peroxides. Even when acting noncatalytically, peroxides can induce significant drug degradation, especially in situations with high excipient-to-drug ratios (low dose). With some excipients, oxidative degradation of the excipient generates peroxide products, which can in turn react with the drug. Peroxides can be found in excipients either as a result of the excipient manufacturing process or due to oxidative instability of the excipient. In both cases, the issue is most prevalent with polymeric excipients. Peroxides are sometimes present as initiators in polymerization processes, though which excipients have this source of peroxides is difficult to determine because of the proprietary nature of the processes used to prepare a number of excipients. In some cases, peroxides can form in the process of transforming natural or synthetic polymers into the desired final products. Two classes of excipients are most associated with peroxide impurities, although peroxides may be found in other excipient classes, generally at lower levels. The first group consists of polymeric ethers. These ethers include polyethylene glycols (PEGs),15,16 polyethylene oxides (PEOs), polysorbates, polyoxyethylene alkyl ethers, polyoxyethylene stearates, and other ethylene oxide-based materials.13,17 Such materials generally have some level of peroxides present as supplied by the vendor. Polymeric ethers are also subject to autoxidation, which results in the formation of more peroxides.14 To minimize this degradation pathway, the above materials are generally

4 IMPURITIES IN DRUG PRODUCTS

79

supplied with an antioxidant, typically butylated hydroxytoluene (BHT). BHT itself can be a concern for some pharmaceutical formulations due to the formation of strongly colored by-products. Under some storage or packaging conditions, the BHT will volatilize, resulting in reduced levels in the dosage form and correspondingly greater instability.18 The second group of excipients commonly associated with peroxide impurities consist of polyvinyl pyrrolidone (PVP)–based excipients. These include Povidone and Crospovidone.19,20 Both excipients commonly contain 100–200 ppm of peroxide present as process impurities. Peroxides can form by autoxidation of PVP, and greater amounts can form under such shear conditions as found during granulation and tableting. Peroxide formation is relatively slow for solid dosage forms under standard aging conditions; however, the age and storage conditions of the excipient can lead to variable peroxide levels.20 Some organic solvents and oils have also been reported to contain peroxides.21 The level of peroxides in susceptible solvents can increase with time, depending on the storage conditions.

IV. ALDEHYDES Aldehydes, even in trace amounts, are often found to adversely affect the stability and efficacy of drugs via direct reaction with the active pharmaceutical ingredient (API). Formaldehyde, acetaldehyde, furfural, and glyoxal are the aldehydes most often responsible for stability issues in drug products. In solid dosage forms, these aldehydes are sufficiently small to be mobile enough to react effectively with APIs. Many commonly used pharmaceutical excipients contain trace low molecular weight aldehydes. Aldehyde impurities in excipients are often due to excipient degradation. With organic excipients, degradation to aldehydes is generally associated with oxidation. For example, benzyl alcohol oxidation generates benzaldehyde,22 and autoxidation of unsaturated fats, polyethylene glycol, and polysorbates generates small molecule aldehydes.13 In some cases, aldehydes are produced via hydrolytic reactions. This is seen in the formation of furfural and its adducts in the acid-catalyzed degradation of hemicellulose and other sugar-based excipients.23 For example, 5-hydroxymethylfurfural is found in spray-dried lactose because of thermal decomposition occurring during the spray-drying process.24 A second source of aldehydes involves functional additives present in the excipients, either as aldehydes themselves or as materials that oxidize or hydrolyze to give aldehydes. Examples include preservatives, cross-linking agents, and flavoring agents. Corn starch often contains hexamethylenetetramine as a preservative, which hydrolyzes to give ammonia and formaldehyde.25 Glyoxal can be found in hydroxyethylcellulose as a crosslinking reagent22 and is found as an impurity in hydroxypropylmethyl celluloses.13 Many dye and flavoring reagents (for example, vanillin) contain aldehyde impurities.17 A third source of aldehydes involves materials leached from elastomeric stoppers26,27 (see Section VII).

80

K.C.WATERMAN et al.

A number of excipients contain reducing sugar impurities (aldehydes) such as glucose, maltose, and lactose,28,29 which can react with the API or other excipients.30 The mechanism of the deleterious effects of aldehydes on drug products is generally well understood. For drug substances with primary and secondary amines, nucleophilic attack on the aldehyde produces an -hydroxyl amine, which can dehydrate to give an imine (Schiff’s base) as shown below:

With nucleophiles such as alcohols (enolates) and thiols, reaction with aldehydes yields the corresponding addition products.31

Aldehydes can react with many excipients in drug products, which in turn can affect the performance of the drug by altering drug release and delivery. An example of this is seen with gelatin,25,32 where formaldehyde reacts with the amino groups on lysine residues causing protein cross-linking, which in turn changes the dissolution characteristics of gelatin capsules. A similar situation is seen when the aldehyde decomposition product of the antioxidant BHT, 2,6-di-t-butyl-4-hydroxybenzaldehyde, cross-links gelatin capsules, affecting the overall dissolution rate.33

V. METAL IMPURITIES Metals are found in almost all excipients. The types of metals and levels present in excipients can vary significantly depending on the excipient type, source, and production process used to extract or produce the excipient.

4 IMPURITIES IN DRUG PRODUCTS

81

Metals are deleterious to drug products for the following reasons: inherent toxicity of the metal, formation of insoluble metal complexes, and oxidative14 and hydrolytic1 catalysis. Common metals found in excipients include arsenic, bismuth, cadmium, chromium, copper, iron, lead, mercury, nickel and sodium.22,34,35 Metals can also be brought into drug products during processing or can be leached from packaging (see Section VII). Certain excipients inherently contain high levels of metals, such as minerals (talc or kaolin) or inorganic compounds derived from minerals such as buffers and oxides (e.g., phosphates, silicates, titanium dioxide, etc). Additionally, excipients produced using metal catalysts, for example, hydrogenated oils and fats, may contain high concentration of metals such as platinum or palladium. The process equipment and storage container may lead to contamination by leached metals22 (see also Section VII). Levels are usually reported in parts per million (ppm) concentration. Since some of the issues associated with metals involve catalysis, even relatively low levels of metals may impact dosage form performance. Numerous processes for producing dosage forms involve contact of materials with metal surfaces. The degree of metal incorporation into dosage forms will depend on a number of factors. In liquid dosage forms, one is often concerned with direct dissolution of metal ions by contact of a liquid with a metal surface. This will tend to be a greater issue for lower pH solutions. With solid dosage forms, not only is direct dissolution an issue, but also abrasion of metals can enable incorporation of metal particles into the dosage forms. Toxicity that is due to heavy metals is the primary safety concern stemming from metal impurities in excipients. Toxicology studies have demonstrated that the most toxic metals are cadmium, mercury, chromium, lead, nickel, and arsenic. Depending on the level and route of ingestion, these metals can cause death, cancer, damage brain and nerve function, and specifically poison the liver and kidneys.36 Because of their toxicity, these metals have pharmacopeial specifications to limit the amount present in solutions.22 There is, however, variability in pharmacopeial specifications, even with a single country’s monographs. For example, lead is restricted to 5 ppm in USP/NF monographs, yet is limited to 3 ppm as an impurity in calcium carbonate.35 Harmonization of these limits has been attempted to facilitate drug development.34 Insoluble heavy metal salts may be an issue, depending on the solubility product (Ksp) of the drug being formulated in the presence of metals. Although precipitates in liquid dosage forms can sometimes be removed by filtration, this approach is often unacceptable for either quality or cost concerns. Precipitates adversely impact the performance of injectable dosage forms and can cause in vivo reactions such as phlebitis.37 Different salts may precipitate from solution, depending on the buffers used and their concentrations. Calcium salts may be very insoluble and can lead to precipitate or particulate formation. Insoluble chlorides, hydroxides, and sulfates are commonly formed with calcium, mercury, iron, and magnesium because of low solubility products.38 These insoluble precipitates will lead to undesirable dosage form properties and can be quite harmful, depending on the route of administration. Trace metal impurities in excipients can lead to oxidative catalysis resulting in drug substance degradation.14 The metals most commonly

82

K.C.WATERMAN et al.

associated with oxidation are iron and copper. Both of these metals act as catalysts by facilitating the reduction of molecular oxygen and thus rendering it more reactive. There are three distinct oxidative reactions that can occur with metals: (1) direct metal catalysis, whereby the metal acts as an electron exchange to reduce oxygen, (2) metal binding to drug and oxygen simultaneously, and (3) Fenton-type reactions where transition metal ions reduce peroxide to generate the highly reactive hydroxyl radical. Metals are also associated with catalysis of hydrolysis.1

VI. SMALL MOLECULE CARBOXYLIC ACIDS Small molecule carboxylic acids can be found in many polymeric excipients, sugars, and unsaturated fats. Generally, the most reactive carboxylic acids found as impurities in excipients include formic acid, acetic acid, and glyoxalic acid. In polymers made from carboxylic acids, unreacted monomeric carboxylic acids can be carried over from the polymerization process. For example, with poly(lactide), lactic acid is present at concentrations of 0.2–1.0%.39,40 The autoxidation of many polymers such as polyethylene glycols and non-ionic surfactants generates carboxylic acids, particularly formic acid.41,42 Under humid conditions, slow hydrolysis of polylactide-co-glycolide (PLGA) and polyanhydrides results in formation of carboxylic acids. Unsaturated fats also undergo autoxidation easily and generate carboxylic acid degradants in addition to peroxides.43 Oxidation of aldose sugar produces aldonic acid.44–46 Because of the higher oxidation propensity of aldehydes, carboxylic acid impurities are often found in systems that generate aldehydes (see Section IV). Small molecule carboxylic acids interact with drug products by three mechanisms: (1) they induce a change in acid content of a formulation, which can shift the formulation into a less stable pH region and initiate or accelerate the degradation of the API; (2) carboxylic acids can react with drug molecules or excipients with nucleophilic functional groups such as primary or secondary amines or with hydroxyls to form amides and esters, respectively;47 (3) salt formation between carboxylic acids and basic drugs can influence the solubility, dissolution, or chemical stability of the drug.48,49

VII. LEACHABLES/EXTRACTABLES Leachable /Extractable chemicals from packaging are a major source of contaminants in drug products, especially in liquid dosage forms.

4 IMPURITIES IN DRUG PRODUCTS

83

Leachable/Extractables can come from glass, rubber stoppers, and plastic packaging materials. Metal oxides such as Na2O, SiO2, CaO, MgO, and Al2O3 are the major components leached/extracted from glass.50 Manganese and iron can also be leached from amber glass vials. Because of the complexity of different synthetic materials, the leachable/extractable profiles from rubber and plastic are far more complicated than from glass. Generally, most synthetic materials contain leachable oligomers/monomers, vulcanizing agents, accelerators, plasticizers, and antioxidants.51 Some examples of leachables/extractables from synthetic materials include styrene from polystyrene,52 diethylhexylphthalate (DEHP, plasticizer in PVC),3 dioctyltin isooctylmercaptoacetate (stabilizer for PVC),53 zinc stearate (stabilizer in PVC and polypropylene),52,54 2-mercaptobenzothiazole (accelerator in rubber stopper),55 and furfural from rayon.27 Pharmaceutical and analytical scientists should obtain detailed composition information from packaging vendors, since such information is often proprietary and varies from vendor to vendor. The detrimental effects of leachables/extractables vary, depending on the nature of the extracted material. For example, glass leachables/extractables can decrease the formulation pH.50 Specific metals such as iron catalyze certain degradation reactions.14 Antioxidants, when combined with metal ions, can accelerate drug oxidation.56 Many accelerators are free radical initiators. Some leachables/extractables, such as 2-mercaptobenzothiazole, also have toxicological concerns.55

VIII. ALCOHOLS AS IMPURITIES Alcohols as impurities are generally not a problem in most drug products. The reactivity of alcohols is low under typical formulation conditions. Levels of alcohol impurities in solid dosage forms are usually not sufficiently high to cause significant issues. Alcohols are often used in granulation steps and can therefore be expected to be found at some level in the final products. In liquid dosage forms, alcohol impurities may be present in ethanol and long-chain poly-alcohol polymers.

IX. BIOLOGICAL IMPURITIES Biological impurities in excipients consist of bacteria, fungi, viruses, endotoxins, and prions. These impurities can enter excipients in a variety of ways and have different consequences, depending on the dosage form. Safety is the main concern of biological contamination. Infectious microorganisms can cause severe illness or death, depending on the quantity and type of organism. Chemically, certain organisms can degrade drug substances and excipients in the drug product via direct metabolism, although this is of less concern than the primary safety issue. Parenteral dosage forms require extensive testing of sterility and biological impurity contamination because of the high sensitivity of this administration route. These same sterility standards do not apply to solid and

84

K.C.WATERMAN et al.

oral liquid dosage forms, yet it may sometimes be important to assess certain types of biological impurities (e.g., prions) in some excipients regardless of the final dosage form. The term referring to the degree of biological contamination by microorganisms is bioburden and is quantified by measuring the number of colony forming units (CFU) present in an excipient. For example, in the EU the current limit is less than or equal to 100 CFU per 100 mL of sterile product. Microorganisms are introduced as contaminants in excipients at several stages of production including both manufacturing and processing. This can occur, for example, because of poor water quality control or handling during the manufacturing. Contamination can also occur by handling of the final excipient when received for dosage form manufacturing. Non-aseptic storage of excipients may contaminate dry powder excipients that are opened for sampling via spores from the environment. Limits have been set for a variety of excipients. For example, the USP/NF microbial limits are not more than (NMT) 100 CFU/g for lactose monohydrate and NMT 1000 CFU/g for magnesium stearate and microcrystalline cellulose. Limits on yeasts and molds are usually about half the levels allowed for bacteria.34 Excipients derived from bacterial transformation or enzymatic processes, such as sugars (e.g., mannitol or dextrose), pose special concerns due to potential contamination by endotoxins. Endotoxins, or pyrogens, are lipopolysaccharides from gram-negative bacteria that can induce severe fever upon parenteral administration. Parenteral products have different limits for allowable endotoxin levels.57 Prions are found in animal-derived excipients. These infectious proteins induce transmissible spongiform encephalopathy (TSE) and cause degradation of bovine brain tissue. They may also lead to a human disease variant known as Creutzfeldt-Jacob disease.22 Excipients derived from animal sources, primarily bovine, sources are at the highest risk for prion contamination. Excipients such as gelatins or those derived from animal fats, such as magnesium stearate, are most likely to be contaminated by prions. Additionally, lanolin, lanolin derivatives, and milk-based excipients are at risk of TSE contamination. Limits on the risk of transmission were set in 1992 in the CPMP guidelines for minimizing the risk of transmission of agents causing spongiform encephalopathy via medicinal products.22 In addition to microorganism-based impurities, there are other biological impurities that may pose a concern for drug products. For example, allergens from peanuts can contaminate excipients by cross-contamination with processing equipment used to process peanuts and excipient raw materials. These contaminants can cause severe allergic reactions in sensitive individuals.58 Additionally, gluten present in excipients manufactured from wheat starch may cause celiac disease.22

X. ADDITIVES IN EXCIPIENTS A number of excipients contain additives that serve as stabilizers or to improve material properties. Stabilizers can themselves affect drug stability,

4 IMPURITIES IN DRUG PRODUCTS

85

either by direct interaction or by affecting excipient decomposition rates. As such, it is often important to control such additives. Examples of these stabilizers include antioxidants and antimicrobials. For example, PEOs typically contain 100–500 ppm of BHT;18 anhydrous lanolin and certain fat derivatives contain approximately 200 ppm of BHT; soybean lecithin and certain fats and fat derivatives contain 0.1–0.2% -tocopherol; and some aqueous dispersions used in tablet coating contain a few ppm of hydrogen peroxide (as an antimicrobial).22 Additives associated with material properties often involve materials designed to plasticize polymeric excipients. Latex dispersions often have additives to promote coalescence of particles upon drying. These often include surfactants and plasticizers. These additives can be of concern when they interact with the drug. For example, if they dissolve some of the drug, that material may become more reactive toward chemical decomposition.

XI. FINAL OBSERVATIONS In a number of cases, chemical stability issues associated with an excipient are due to impurities in the excipient rather than the excipient itself. The most reactive impurities tend to be small molecules. This is especially true in solid dosage forms, where the mobility of large molecules is severely restricted. In addition to chemical stability, impurities in drug products can cause issues associated with toxicity or dosage form performance. It is often valuable to quantitatively determine the level of important impurities in drug products, and to trace the origin of those impurities to their source. If the source is from an excipient, variability from lot to lot may make a marginal product unacceptable for reliability. If the source is related to the dosage form preparation process, it may be worthwhile considering technologies and processes that provide less of the undesirable impurity.

XII. SUMMARY This overview has focused on those impurities brought into the drug product system by excipients or dosage form processing likely to cause stability or performance issues in the dosage form. The sources of the impurities are discussed as well as how they affect dosage forms or represent product issues. In particular, those impurities that lead to drug stability issues are reviewed.

REFERENCES 1. Waterman, K. C., Adami, R. C., Alsante, K. M., Antipas, A. S., Arenson, D. R., Carrier, R., Hong, J., Landis, M. S., Lombardo, F., Shah, J. C., Shalaev, E., Smith, S. W. and Wang, H. Hydrolysis in pharmaceutical formulations. Pharm. Dev. Technol. 7:233–266, 2002. 2. Dalton, C. R. and Hancock, B. C. Processing and storage effects on water vapor sorption by some model pharmaceutical solid dosage formulations. Int. J. Pharm. 156;143–151, 1997.

86

K.C.WATERMAN et al.

3. Malamataris, S., Tsiri, K. and Goidas, P. Sorption of moisture of different size fractions of some direct compression excipients and tensile strength of the corresponding tablets. Congr. Int. Technol. Pharm. 6th, 5:195–204, 1992. 4. Bell, L. N. and Labuza, T. P. Aspartame degradation as a function of ‘‘water activity.’’ In Water Relationships in Food. (H. Levine and L. Slade, Eds.) pp. 337–349, Plenum Press, NY, 1991. 5. Schepky, G. and Fischer, M. Effect of production-related variations in hardness and moisture content on the sorption characteristics of tablets. Eur. J. Pharm. Biopharm. 39:53– 60, 1993. 6. Matsunaga, Y., Ohta, R., Bando, N., Yamada, H., Yuasa, H. and Kanaya, Y. Studies on internal structure of tablets IV: Effects of water content on physical and chemical stability of tablets containing an anticancer drug TAT-59. Chem. Pharm. Bull. 41:720–724, 1993. 7. Nyqvist, H., Nicklasson, M. and Lundgren, P. Studies on the physical properties of tablets and tablet excipients III: Water sorption and its effect on hardness and disintegration. Acta Pharm. Suec. 18:305–314, 1981. 8. Steendam, R., Eissens, A. C. E., Frijlink, H. W. and Lerk, C. F. Plasticization of amylodextrin by moisture. Consequences for drug release from tablets. Int. J. Pharm. 204:23–33, 2000. 9. Hoon, T. S. and Pathak, Y. V. Evaluation of maltodextrins as tablet excipients  3 Moisture absorption studies. Congr. Int. Technol. Pharm. 6th, 5:185–194, 1992. 10. Kawashima, Y., Takeuchi, H., Hino, T., Niwa, T., Lin, T. L., Sekigawa, F. and Ohya, M. The effects of particle size, degree of hydroxypropyl substitution and moisture content of low-substituted hydroxypropyl cellulose on the compactibility of acetaminophen and the drug release of the resultant tablets. S. T. P Pharma Sci. 3:170–177, 1993. 11. Malamataris, S., Goidas, P. and Dimitriou, A. Moisture sorption and tensile strength of some tableted direct compression excipients. Int. J. Pharm. 68:51–60, 1991. 12. Fakes, M. G., Dali, M. V., Haby, T. A., Morris, K. R., Varia, S. A. and Serajuddin, A. T. M. Moisture sorption behavior of selected bulking agents used in lyophilized products. PDA J. Pharm. Sci. Technol. 54:144–149, 2000. 13. Crowley, P. and Martini, L. Drug-excipient interactions. Pharm. Technol. Europe, 15:26–34, 2001. 14. Waterman, K. C., Adami, R. C., Alsante, K. M., Hong, J., Landis, M. S., Lombardo, F. and Roberts, C. J. Stabilization of pharmaceuticals to oxidative degradation. Pharm. Dev. Technol. 7:1–31, 2002. 15. McGinity, J. W., Hill, J. A. and LaVia, A. L. Influence of peroxide impurities in polyethylene glycols on drug stability. J. Pharm. Sci. 64:356–357, 1975. 16. Johnson, D. M. and Taylor, W. F. Degradation of fenprostalene in polyethylene glycol 400 solution. J. Pharm. Sci. 73:1414–1417, 1984. 17. Kibbe, A. H. (Ed.) Handbook of pharmaceutical excipients, 370–371, 407–411, 576–577. American Pharmaceutical Association: Washington, D. C. 2000. 18. Union Carbide Corporation Sentry2 Polyox2 water soluble resin storage stability. Personal Communication, 1996. 19. Laicher, A., Junger, H. and Klemm, F. H. Stability behaviour of molsidomine-containing pellet formulations. J. Pharm. Pharmacol. 49:131–134, 1997. 20. Hartauer, K. J., Arbuthnot, G. N., Baertschi, S. W., Johnson, R. A., Luke, W. D., Pearson, N. G., Rickard, E. C., Tingle, C. A., Tsang, P. K. S. and Wiens, R. E. Influence of peroxide impurities in povidone and crospovidone on the stability of raloxifene hydrochloride in tablets: identification and control of an oxidative degradation product. Pharm. Dev. Technol. 5:303–310, 2000. 21. Shertzer, H. G. and Tabor, M. W. Peroxide removal from organic solvents and vegetable oils. J. Environ. Sci. Health, A20:845–855, 1985. 22. Ja¨kel, D. and Keck, M. Purity of Excipients. Excipient Toxicity & Safety, 103:21–58, 2000. 23. Janicki, C.A. and Almond, H.R. Jr., Reaction of haloperidol with 5-(hydroxymethyl)-2furfuraldehyde, an impurity in anhydrous lactose. J. Pharm. Sci. 63:41–43, 1974. 24. Brownley, C. A., Jr. and Lachman, L. Browning of spray-processed lactose. J. Pharm. Sci. 53: 452–454, 1964.

4 IMPURITIES IN DRUG PRODUCTS

87

25. Digenis, G. A., Thomas, B. and Shah, V. P. Cross-linking of gelatin capsules and its relevance to their in vitro–in vivo performance. J. Pharm. Sci. 83:915–921, 1994. 26. Battersby, J. E., Hancock, W. S. and Lawlis, V. B., Jr. Container system for storage stable pharmaceuticals. PCT Int. Appl. WO 9210747, 1992. 27. Hartauer, K. J., Mayer, R. F., Schwier, J. R., Bucko, J. H., Cooke, G. G. and Sullivan, G. R. The effect of rayon coiler on the dissolution stability of hard-shell gelatin capsules. Pharm. Technol. 17, 76, 78, 80:82–83, 1993. 28. Kitson, G. E., Hudson, H. E., and Dickinson, N. A. The effect of various sugars on the stability of 2-amino-6-methyl-5-oxo-4-n-propyl-4,5-dihydro-s-triazolo [1,5-a] pyrimidine. Pharm. Acta. Helv. 51:181–185, 1976. 29. Kitson, G. E., Hudson, H. E. and Dickinson, N. A. The effect of various sugars on the stability of 2-amino-6-methyl-5-oxo-4-n-propyl-4,5-dihydro-s-triazolo [1,5-a] pyrimidine Part II: The use of aldehydic and ketonic compounds to investigate the functional groups involved. Pharm. Acta. Helv. 54:359–362 1979. 30. Wirth, D. D., Baertschi, S. W., Johnson, R. A., Maple, S. R., Miller, M. S., Hallenbeck, D. K., and Gregg, S. M. Maillard reaction of lactose and fluoxetine hydrochloride, a secondary amine. J. Pharm. Sci. 87:31–39, 1998. 31. March, J. Advanced Organic Chemistry: Reactions, Mechanisms, and Structure, 4th ed., McGraw-Hill Book Co., New York. 1992. 32. Doelker, E. and Vial-Bernasconi, A. C. Interactions contenant-contenu au sein des capsules gelatineuses et evaluation critique de leurs effets sur la disponibilite des principes actifs. STP. Pharma, 4:296–306, 1988. 33. Desai, D. S., Ranadive, S. A., Lozano, R., and Varia, S. A. Dissolution instability of encapsulated marketed tablets. Int. J. Pharm. 144:153–158, 1996. 34. Chowhan, Z. T. Harmonization of Excipient Standards, Chapter 14. In Excipient Toxicity and Safety, M.L. Weiner and L.A. Kotkoskie (Eds.) Marcel Dekker, Inc., NY, pp. 321–354. 1999. 35. Chowhan, Z. T. A rational approach to setting limit tests and standards on impurities in excipients. Pharm. Tech. 19:43–48, 1995. 36. Duffus, J. and Worth, H. 2001. The Science of Chemical Safety Essential Toxicology, An Educational Resource. IUPAC. 37. Yalkowsky, S. H., Krzyzaniak, J. F. and Ward, G. H. Formulation-related problems associated with intravenous drug delivery. J. Pharm. Sci. 87:787–796, 1998. 38. Ebbing, D. D. General Chemistry, 5th ed. Houghton Mifflin Company, Boston, MA. 1996. 39. Cam, D., Bracci, B. and Marucci, M. Influence of residual monomers and metals on poly D,L-(lactide) thermal stability. Front. Biomed. Polym. Appl. 2:189–193, 1999. 40. Hyon, S.-H., Jamshidi, K. and Ikada, Y. Effects of residual monomer on the degradation of D,L-lactide polymer. Polym. Int. 46:196–202, 1998. 41. Hamburger, R., Azaz, E. and Donbrow, M. Autoxidation of polyoxyethylenic non-ionic surfactants and of polyethylene glycols. Pharm. Acta Helv. 50:10–17, 1975. 42. Sakharov, A. M., Mazaletskaya, L. I. and Skibida, I. P. Catalytic oxidative deformylation of polyethylene glycols with the participation of molecular oxygen. Kinetics and Catalysis, 42:662–668, 2001. 43. Gardner, H. W. Oxygen radical chemistry of polyunsaturated fatty acids. Free Radical Biol. Med. 7:65–86, 1989. 44. Volt, D. and Volt, J. G. Biochemistry, p. 250. John Wiley & Sons, New York, 1990. 45. Krebs, A. The role of heavy metals in the autoxidation of sugar solutions. Biochem. Z. 180: 377–394, 1927. 46. Wehmeier K. R. and Mooradian, A. D. Autoxidation and antioxidation potential of simple carbohydrates. Free Radical Biol. Med. 17:83–86, 1994. 47. Johnson, D. M. and Taylor, W. F. Degradation of fenprostalene in polyethylene glycol 400 solution. J. Pharm. Sci. 73:1414–1417, 1984. 48. Bastin, R. J., Bowker, M. J. and Slater, B. J. Salt selection and optimization procedures for pharmaceutical new chemical entities. Org. Process Res. Dev. 4:427–435, 2000. 49. Gould, P. L. Salt selection for basic drugs. Int, J. Pharm. 33:201–217, 1986. 50. Sanga, S. V. Review of glass types available for packaging parenteral solutions. J. Parenteral Drug Assn. 33:61–64, 1979.

88

K.C.WATERMAN et al.

51. Paskiet, D. M. Strategy for determining extractables from rubber packaging materials in drug products. PDA J. Pharm. Sci. 51:248–251, 1997. 52. Hamon, M. Plastic materials in pharmaceutical packaging I: Behavior in relation to various parameters. Sci. Tech. Pharm. 10:277–281, 1981. 53. Aignasse, M. F., Prognon, P., Stachowicz, M., Gheyouche, R. and Pradeau, D. A new and rapid HPLC method for determination of DEHP in PVC packaging and releasing studies. Int. J. Pharm. 113:241–246, 1995. 54. Fras, I., Cassagnau, P. and Michel, A. Influence of processing conditions on the leaching of thermal stabilizers from plasticized poly(vinyl chloride) in the presence of water. J. Appl. Polym. Sci. 70:2391–2400, 1998. 55. Reepmeyer, J. C. and Juhl, Y. H. Contamination of injectable solutions with 2-mercaptobenzothiazole leached from rubber closures. J. Pharm. Sci. 72:1302–1305, 1983. 56. Hovorka, S. W. and Scho¨neich, C. Oxidation degradation of pharmaceuticals: theory, mechanism and inhibition. J. Pharm. Sci. 90:253–269, 2001. 57. Center for Drug Evaluation and Research. Guidance for Industry: Guideline on Validation of the Limulus Amebocyte Lysate Test as an End-Product Endotoxin Test for Human and Animal Parenteral Drugs, Biological Products, and Medical Devices, p. 10. CDER FDA. Rockville, MD, 1997. 58. Schappi, G. F., Konrad, V., Imhof, D., Etter, R. and Wuthrich, B. Hidden peanut allergens detected in various foods: Findings and legal measures. Allergy, 56:1216–1220, 2001.

5 STRATEGIES FOR INVESTIGATION AND CONTROL OF PROCESS- AND DEGRADATION-RELATED IMPURITIES BERNARD A.OLSENa AND STEVEN W. BAERTSCHIb a

Eli Lilly and Company, Lafayette, IN 47909 Eli Lilly and Company, Indianapolis, IN 46285

b

I. INTRODUCTION II. GOALS AND STRATEGIES A. Goals B. Chemistry-Guided VersusTechnique-Oriented Strategy C. Screening Methods D. Targeted Methods E. Focused Methods for Control Use F. Pharmacopeial Methods III. PROCESS-RELATED IMPURITIES IV. DEGRADATION-RELATED IMPURITIES A. Definitions B. Strategy C. Thermolytic Degradation D. Hydrolytic Degradation E. Photolytic Degradation F. Oxidative Degradation G. Common Problems Encountered During StressTesting with HPLC-UV H. Detectors I. Drug Product StressTesting V. SUMMARYAND CONCLUSIONS VI. ACKNOWLEDGMENTS REFERENCES

I. INTRODUCTION The investigation of impurities is a critical part of drug development. The overall goal of these investigations is to provide information that will contribute to reproducible production of a high quality product that has been demonstrated to be safe through toxicological and clinical trials.

89

90

B.A. OLSEN AND S.W. BAERTSCHI

Understanding the origin of impurities and the mechanisms for their minimization or removal are necessary to develop drug substance and drug product manufacturing processes and storage conditions that will assure quality at release and throughout the allowed storage period. The analytical evaluation of impurities is used to guide development and is required for drug registration. As mentioned in Chapters 1 and 2, the International Conference on Harmonisation (ICH) guidelines on impurities in drug substances1,2 and drug products3,4 classify impurities as organic or inorganic. Specific thresholds for identification, safety qualification, and reporting of organic impurities have been set in these guidelines (Table 1). These thresholds define the level of significance of an impurity that can be used to guide development efforts. The investigation of impurities covers topics such as method development and impurity identification techniques, both of which have been reviewed in the literature.5–10 These reviews have often focused on specific examples or particular analytical techniques. This chapter describes strategies for the investigation of process-related and degradation-related impurities in drug substances and drug products. Particular emphasis will be given to a chemistry-guided approach and the iterative nature of investigations as knowledge is gained. The delineation of impurity type and source that will be used for this discussion is given in Table 2. Although the scope of the discussion will be limited to drugs prepared by chemical synthesis, the strategies described may also apply to biosynthetic or semisynthetic molecules. Considerations for investigating and controlling crystal forms,11 residual solvents12 and microbiological impurities11 have been addressed in other ICH guidelines and will not be discussed here.

TABLE 1 ICH Reporting, Identification, and QualificationThresholds Thresholds* Maximum daily dose

Reporting

Identification

Qualification

Drug substance 2g

0.05%

0.10% or 1.0 mg/day intake

0.15% or 1.0 mg/day intake

>2g

0.03%

0.05%

0.05%

Drug product 1g

0.1%

>1g

0.05%

< 1 mg

1% or 5 mg/day intake

1% or 50 mg/day intake

1–10 mg

0.5% or 20 mg/day intake

1% or 50 mg/day intake

10–100 mg

0.2% or 2 mg/day intake

0.5% or 200 mg/day intake

100 mg–2 g

0.2% or 2 mg/day intake

0.2% or 2 mg/day intake

>2g

0.1%

0.1%

*When a percentage and amount/day intake are both given, use whichever is lower.

91

5 INVESTIGATION AND CONTROL OF IMPURITIES

TABLE 2 Impurity Descriptions Impurity type Process-related drug substance

Process-related drug substance

Degradation drug substance or drug product Degradation drug product

Impurity source 8 Organic > > > > < Starting material Intermediate > > > By-product > : Impurity in starting material 





Organic or Inorganic Reagents; catalysts; etc: Organic Degradation products Organic Excipient interaction products

Analytical method development is at the heart of impurity investigations. Since appropriate methods must be used to detect and quantify impurities, a key question for investigating impurities is How much investigation is enough? The answer determines how many techniques and conditions will be used to provide confidence that all impurities of significance have been detected. With a threshold of 0.10% for the drug substance, one can never say with 100% confidence that all impurities have been detected in a drug. However, with an appropriate strategy, a high level of assurance can be provided that all significant impurities have been detected.

II. GOALS AND STRATEGIES A. Goals The goals for the investigation of impurities are outlined in Table 3. A common goal for investigation of both process- and degradation-related impurities is to determine which of the many potential impurities are, in fact, produced in the manufacturing process and which occur under a given set of storage conditions. During new drug development this knowledge can be used to modify the process in order to eliminate or minimize levels of impurities. Knowledge of stability characteristics can be used to store the drug substance and drug product in appropriate packaging and under appropriate environmental conditions to minimize or eliminate unacceptable degradation. Knowledge of the degradation behavior of the drug and interactions with excipients is also useful in developing the formulation. Information from impurity investigations is essential in establishing specification tests and acceptance limits at various control points in the manufacturing process.

92

B.A. OLSEN AND S.W. BAERTSCHI

TABLE 3 Goals of Impurity Investigations Process-related impurities

Degradation-related impurities

Identify significant impurities and demonstrate the absence of potential or postulated impurities

Identify potential degradation products through stress testing and actual degradation products through stability studies

Determine origin of impurities and methods for elimination or reduction

Understand degradation pathways and methods to minimize degradation

Establish a control system for impurities involving:

Establish a control system for impurities involving:

1. Processing/manufacturing conditions 2. Suitable analytical methods/ specifications

1. Processing/manufacturing conditions 2. Suitable analytical methods/ specifications 3. Long-term storage conditions, including packaging 4. Formulation

B. Chemistry-Guided Versus Technique-Oriented Strategy One strategy for impurity investigation might be termed a ‘‘techniqueoriented approach.’’13,14 The main feature of this approach is to use as many techniques as possible to search for impurities. This would include various types of chromatographic techniques with multiple detection schemes. Of course the characteristics of the drug substance, process intermediates, and any obvious by-products or degradation products would strongly influence the primary method(s) employed. Quite often the primary technique is reversed-phase HPLC (RP-HPLC) with UV detection. The search for unknown impurities might continue with orthogonal separation techniques such as normal phase HPLC, TLC, CZE, GC, and SFC and multiple detection modes such as PDA, MS, ELSD, CLND, RI, FID, TC, and multiple visualization reagents for TLC. This approach is sometimes focused on analysis of the drug substance and may not include investigation of impurities in synthetic intermediates and reaction mixtures. Determining when enough investigation has been performed and when the point of diminishing returns is reached are the main issues with this approach. We advocate a ‘‘chemistry-guided’’ approach for the investigation of impurities. For process-related impurities, the chemistry of the process is carefully examined to postulate potential by-products or other impurities at each step. These, in addition to the synthetic intermediates, can form a basis for method development and further impurity investigation. For degradation products, well-designed stress degradation and excipient interaction studies serve to direct development efforts. Some degradation products can also be predicted based on the structure of the molecule. Evaluation of stress-degraded samples using a broad screening method can reveal potential degradation products, and an evaluation of the degradation

5 INVESTIGATION AND CONTROL OF IMPURITIES

93

pathways can help assess whether degradation products have been investigated adequately.

C. Screening Methods A method capable of screening various samples for a broad range of impurities is usually desired. Reversed-phase HPLC with UV detection and gradient elution is recommended to cover a broad polarity range and is often employed for impurity screening. Screening laboratory samples with gradient elution over a wide polarity range is useful in establishing the presence or absence of both polar and nonpolar impurities. The gradient range can then be adjusted accordingly to provide elution of the most nonpolar (i.e., the mostretained) impurities. When gradient baseline artifacts are problematic, it can be useful to employ dual isocratic screening methods. Low solvent strength systems may be needed to separate relatively polar compounds while a stronger solvent strength system can be used to determine nonpolar impurities. Compatibility of the mobile phase with mass spectrometric detection is highly desirable for providing information to help identify unknown components that may appear in the chromatogram. Other separation or detection modes may be necessary, depending on the nature of the drug substance, but the key concept here is to develop and use methodology that will provide the greatest overall separation and most universal detection.

D. Targeted Methods It is ideal when a screening method is capable of detecting all potential impurities, but this is rarely the case. An obvious example is encountered with single-enantiomer drug substances where a chiral method must be used to determine low levels of the undesired isomer. Such methods can be considered ‘‘targeted’’ toward a specific impurity rather than used for general screening purposes. There are abundant examples of chiral methods in the literature so they will not be discussed further here. Determination of other impurities that are not related in structure to the drug substance, such as metal catalysts or inorganic salts used in the synthetic process, usually requires targeted methods since they are not likely to be detected with the screening method. Targeted methods are often used during development to help define the overall control strategy even if the methods are not ultimately needed for routine use. Impurities that are known or suspected to be toxic present a special concern during development. The goal for investigation of such impurities is to demonstrate that they are undetectable or well below an appropriate level of concern. A targeted method is usually required for investigation of these impurities.

E. Focused Methods for Control Use One of the main goals of impurity investigations is to determine which impurities need to have specification limits for routine monitoring.

94

B.A. OLSEN AND S.W. BAERTSCHI

Knowledge gained with the screening and targeted methods described above can be used to establish the points in a process at which significant impurities need to be controlled. When the synthetic process and product formulation are finalized and impurity investigations are completed, the analytical methods needed to monitor impurities can be optimized. For example, a broad polarity range gradient might be used as a screening method in the search for impurities. If there are no significant late-eluting impurities, the gradient range and time can be optimized to reduce run time and provide ruggedness across different laboratories. An isocratic method might even be possible if it can determine all of the important impurities. These focused methods can then be registered in marketing applications and used in quality control laboratories for product release and stability. Although the conditions of targeted methods used during development may not require focusing, the number of targeted methods to carry forward for routine control laboratory use needs to be determined. This decision is based on the results obtained during development. Targeted methods are usually not required if they were used to demonstrate the absence of a postulated impurity in several development batches, including validation batches. Good manufacturing practices and maintenance of a validated process provide assurance that such postulated impurities will not suddenly appear. Determination of some impurities using targeted methods may need to be continued routinely because of regulatory expectations. The set of broad screening and targeted methods can be used as necessary to investigate impurities in connection with process changes or deviations. As recommended above for initial impurity investigations, chemical considerations should first be used to guide the investigation by postulation of new impurities that may require modified or new methods. If new impurities are highly unlikely, the existing methods can be used to establish comparability of material before and after a change or a deviation from the established process. These results can also demonstrate that focused methods for routine use are still valid, i.e., the important impurities are detected.

F. Pharmacopeial Methods Specifications (tests, methods, and acceptance limits) for impurities given in pharmacopeial monographs provide public standards for generic suppliers of drug substances and drug products. Since drug substances can usually be prepared using many different synthetic routes with varying degrees of purification, the suitability of monograph methods to monitor the purity of a drug substance from a particular supplier is an important issue. The monograph method for related substances or compounds is often focused on the impurities found to be important in material from the monograph sponsor, usually the innovator company. For example, fluoxetine hydrochloride may be synthesized using routes that can produce different impurity profiles.15 Some of these differences can be discerned by the isocratic monograph impurity method while others require the use of a nonofficial gradient method.16 These differences highlight the need to conduct an

5 INVESTIGATION AND CONTROL OF IMPURITIES

95

impurity investigation as described above for drug substances that may be of different origin than that used to establish the monograph methods. The investigation may indeed show that the methods are suitable or it may show the need for different or additional methods. This approach is taken in the European Pharmacopoeia process for certification of suitability of drug substances from different sources. There are certainly areas of overlap of the technique-oriented and the chemistry-guided approaches. For example, chemistry-guided approaches still require a multitechnique or multidimensional approach, although the selection of which additional techniques are used is driven by scientific evaluation rather than a predetermined scheme. It should be noted that neither approach can guarantee that all impurities have been accounted for. More detailed descriptions of the chemistry-guided approach are given below in separate sections on process impurities and degradation products. Examples illustrate how this approach can provide assurance that impurities have been investigated in a thorough and efficient manner.

III. PROCESS-RELATED IMPURITIES A generalized scheme for the synthesis of a drug substance is shown in Figure 1. Every reactant, reagent, solvent, intermediate, and by-product can be considered a potential impurity and should be addressed. Not all byproducts are known when development begins, so efforts to discover and identify them must be included in the investigation of impurities. As described above, one of the main goals of process-impurity investigations is to determine which of the many potential impurities are important and how these should be controlled during the synthetic process to produce a high-quality drug substance. A simplified way of looking at this goal is shown in Figure 2. Knowledge of impurities introduced through starting materials and reagents, those formed during processing, and those that carry through to the drug substance enables development of reaction conditions, criteria for forward processing and final specifications. Stability

FIGURE 1 Generalized synthetic scheme for a drug substance.

96

B.A. OLSEN AND S.W. BAERTSCHI

FIGURE 2 Development of an impurity control strategy for drug substances.

characteristics will dictate appropriate storage conditions needed to maintain the quality of the drug substance. As with the technique-oriented approach, a screening method that detects the compounds involved in the synthesis is typically the starting point for the chemistry-guided approach to impurity investigations. In addition to the drug substance, each isolated intermediate is usually examined to determine impurities present at that point in the synthesis. The nature of the intermediates will determine whether a similar analytical screening method may be used at each step or whether quite different methods/conditions are required. Unknowns are often discovered using the screening method. The conditions of the screening methods can be modified if needed for different steps of the synthesis. More specifically, tailored methods can be developed to focus on known or postulated impurities not detected using the screening conditions. Process-related impurities resulting from impurities in reagents, solvents, and starting materials are also possible. The development of selective methods to detect these impurities serves to broaden the scope of the impurity investigation beyond just the known intermediates and by-products and increases the likelihood of discovering other unknowns. A key aspect of the chemistry-guided approach is the determination of impurity origin and fate at each synthetic step. Analysis of reaction mixtures, filtrates, and isolated intermediates can provide valuable information concerning these questions. Determination of the identity and mechanism of formation of an impurity can lead to modification of reaction conditions that can minimize or eliminate the impurity. The effectiveness of purification procedures such as crystallization and washing can be determined by analysis of filtrates and wash solutions. Showing effective control of an impurity at an intermediate step can make it possible to rule out the impurity as a possibility in the final drug substance. For example, referring to Figure 1, a negligible amount of starting material B present in intermediate C eliminates the need to monitor B in the drug substance. Conversely, if an impurity is difficult to remove, the investigation will determine what acceptance limits must be instituted at intermediate steps to assure that acceptable levels will be present in the drug substance. As impurity investigations proceed, laboratory and scale-up samples from throughout the process should be analyzed for impurities. Significant unknowns that are discovered are identified. As knowledge about impurities

5 INVESTIGATION AND CONTROL OF IMPURITIES

97

increases, the synthetic process may be modified to reduce impurity levels and improve reproducibility. These efforts will, in turn, generate further samples for evaluation. Depending on the nature of the process change, modifications of the screening method or new methods may need to be considered. This iterative cycle continues as the process and methods approach their final forms for registration and commercial production. The following example illustrates this strategy. The synthetic scheme for LY297802 tartrate is shown in Figure 3. The intermediates and drug substance were used to develop initial separation conditions compatible with mass spectrometric detection (Figure 4). Samples

FIGURE 3 Partial synthetic scheme for LY297802 tartrate.

FIGURE 4 Separation of LY297802 and intermediates, using LC-MS compatible conditions. Conditions:YMC basic column, 25 cm 3 4.6 mm id, 5 m; A 5 0.05% trifluoroacetic acid in water, B 5 0.05% trifluoroacetic acid in acetonitrile, gradient elution from 25^60% B in 15 min, hold 10 min; 1.0 mL/min, 280 nm.Unlabeled peaks were unknowns.

98

B.A. OLSEN AND S.W. BAERTSCHI

of isolated intermediates, reaction solutions, and filtrates were also amenable to analysis using similar LC-MS conditions. Analysis of laboratory samples revealed several unknowns. Molecular weight information obtained via mass spectrometry was used to tentatively identify unknown impurities. Ultraviolet spectra obtained from photodiode array detection were also used to support the identification in some cases. Some of the impurities detected had been predicted by examination of the process chemistry and the identities of others were postulated from the mass spectral data. The suspected impurities were then synthesized for comparison with the components observed in samples. Identical retention times, UV spectra, and mass spectra provided strong evidence for positive impurity identification. A summary of related substance impurities investigated for LY297802 is given in Figure 5. Two impurities were identified in the reaction of I1 to form I2 (Figure 6). The hydroxylated and elimination by-products were able to undergo the next reaction in the synthesis in the same way as the desired monochloro intermediate. Therefore, the impurities at this step would lead to the analogous impurities in the drug substance unless the initial impurities were less reactive or purification was effected at some point in the process. To determine the significance of the hydroxy and elimination by-products at the intermediate step, it was necessary to determine at what levels the downstream impurities, BP1 and BP2, were present in the final drug

FIGURE 5 LY297802-related substance impurities.

5 INVESTIGATION AND CONTROL OF IMPURITIES

99

FIGURE 6 Impurities formed during preparation of I2. BP1 and BP2 are the corresponding impurities in the drug substance.

substance. Impurities BP1 and BP2 were synthesized and their chromatographic retention properties examined using the LC-MS screening conditions. The hydroxy by-product, BP1, was separated from the drug substance, but the elimination by-product, BP2, coeluted with the drug substance. Therefore, the low pH LC-MS screening conditions could not be used to determine levels of BP2 in the drug substance. Determination of BP2 coeluted with the drug substance peak might have been possible with mass spectrometric detection, but a simpler method with separation of the impurity was desired. The effect of pH was investigated to check the potential to improve separation of the drug substance and BP2. Mobile phase pH can be a powerful tool for changing the separation selectivity of acidic or basic compounds. This was indicated in a preliminary evaluation of retention versus pH for I1, I2, and LY297802. Based on the potential for a change in selectivity, a pH 5.0 mobile phase was used for the separation of BP2 and other potential impurities. Figure 7 shows that BP2 as well as almost all other impurities were separated using these conditions. The pH 5 conditions revealed that significant levels of BP2 appeared in laboratory samples of drug substance. Additional process optimization was performed to reduce the hydroxy and elimination by-products formed in the intermediate reaction to insignificant levels. Further control could be provided with limits on the impurities in isolated I2. Response factors and detectability of potential impurities must also be considered during impurity investigations. A potential impurity in the thiolformation reaction was the hydrido compound, BP5. Although well separated using the pH 5.0 screening conditions, this impurity had a very low response at 280 nm. The response at 250 nm was much higher and provided a better means of tracking the impurity. Figure 8 shows chromatograms recorded at 250 and 280 nm for crude and recrystallized drug substance. While the hydrido impurity was detectable at 280 nm in crude material, detection at

100

B.A. OLSEN AND S.W. BAERTSCHI

FIGURE 7 Screening for LY297802 impurities using pH 5 mobile phase. Conditions: YMC basic column, 25 cm 3 4.6 mm id, 5 mm; A 5water/acetonitrile/acetic acid/triethylamine 950/ 50/3/5, B 5 water/acetonitrile/acetic acid/triethylamine 100/900/3/5, gradient elution from 0 to 100% B in 30 min; 1.0 mL/min, 280 nm.

FIGURE 8 Impurity profiles demonstrating superior detectability of BP5 at 250 nm. Conditions as in Figure 7 with extended hold time at final conditions.

250 nm afforded a more sensitive means of establishing the effectiveness of the recrystallization in removing this impurity. The need for a targeted method arose because of the n-butyl chloride reactant used in the synthesis of LY297802. This starting material contains as impurities the n-propyl and n-pentyl chloride homologs as well as the isobutyl isomer. Since these impurities will react to form the corresponding impurities in the drug substance (SM1–3), the quality of the n-butyl chloride

5 INVESTIGATION AND CONTROL OF IMPURITIES

101

FIGURE 9 Targeted method for determination of the isobutyl isomer of LY297802. Conditions:YMC ODS column, 25 cm 3 4.6 mm id, 5 mm; 77% trifluoroacetic acid (0.05%) in water, 23% tetrahydrofuran,1.0 mL/min.

must be controlled. It is important in terms of specifications and cost to determine what levels of reactive impurities can be tolerated in the n-butyl chloride to provide acceptable drug substance. This is usually done by spiking the starting material impurities at various levels into reactions and checking the extent of formation and carry-through of the corresponding reaction product impurities. In the case of n-butyl chloride, the homologous reaction products (SM1 and SM3) were separated from LY297802 in the pH 5 screening method, but the isobutyl isomer impurity (SM2) was not. The spiking study required a method capable of separating the isobutyl analog of LY297802. The necessary selectivity was obtained using tetrahydrofuran as a mobile phase modifier rather than acetonitrile (Figure 9). These conditions could be used to determine the relationship between the amount of isobutyl chloride in the starting material and the amount of SM2 in the drug substance, thereby establishing a sound basis for a starting material impurity specification limit. Appropriate control of isobutyl chloride in the starting material eliminates the need to routinely monitor SM2 in the drug substance. For impurities that are suspected to be toxic, it is desirable to demonstrate their absence in an intermediate compound early in the process, if possible. A specification for the impurity in the drug substance may be necessary if the impurity is introduced late in the process or can carry through to a late intermediate or final step. Methods for toxic impurities are usually targeted to determine the specific impurity at trace levels. An appropriate detection limit must be based on toxicological considerations of the impurity as well as the dose, formulation, and indication of the drug. For example, a toxic impurity in a drug for chronic use may be of concern at lower levels than in a drug for an acute indication. An example of use of a targeted method is the determination of a trace impurity in paroxetine HCl (Figure 10).17 The impurity is very similar in

102

B.A. OLSEN AND S.W. BAERTSCHI

FIGURE 10 Paroxetine and trace level impurity.

structure to a neurotoxin known to cause Parkinsonism.18 A method utilizing HPLC (70-minute gradient elution profile) with dual-stage mass spectrometric detection was described. The daughter ion with mass-to-charge ratio 44 produced from the parent ion with mass-to-charge ratio 192 was monitored. The limit proposed for the impurity was 5 ppm. Another example is the determination of mesylate esters as impurities in drug substances isolated as mesylate salts. Figure 11 shows GC-MS chromatograms obtained using single ion monitoring for methyl and isopropyl mesylate in a drug substance. Standards at concentrations corresponding to conservative safety limits gave readily detectable signals. No responses for the mesylate esters were observed in several samples, helping to establish the absence of the impurities in the process.

IV. DEGRADATION-RELATED IMPURITIES As outlined in Table 3, the goals of degradation-related impurity investigations are parallel to the goals of process-related impurity investigations. ‘‘Potential’’ degradation products, i.e., those formed upon stressing or forcing conditions, are analogous to ‘‘potential’’ or postulated process-related impurities resulting from the synthetic reactions utilized to create the drug substance compound. Understanding how the ‘‘potential’’ degradation products form (via specific degradation pathways or mechanisms) provides a basis for minimizing or eliminating degradation (e.g., through processing conditions, formulation, storage conditions, or packaging). Degradation studies also provide for the development of analytical methods that are ‘‘stability indicating.’’ Thus, degradation-related impurity investigations need to start early in development to enable a clear understanding of stability issues and to provide for a safe and efficacious product during clinical trials, eventually leading to a marketed product.

A. Definitions It is important to have a clear definition of terms to facilitate the discussion. The ICH guideline on stability19 differentiates the terms ‘‘stress testing’’ and ‘‘accelerated testing,’’ which have historically been used interchangeably. The guideline defines accelerated testing as ‘‘Studies designed to

5 INVESTIGATION AND CONTROL OF IMPURITIES

103

FIGURE 11 GC-MS determination of mesylate esters in a drug substance. A 5 isopropyl

mesylate, B 5 methyl mesylate; chromatograms (from bottom to top): blank, standard solution, drug substance sample solutions. Conditions: DB-WAX, 30 m 3 0.32 mm 3 0.25 mm film; oven temperature held at 90 C for 3 min, increased to146 C at 8 C/min, increased to 180 C at 20 C/min, and then held isothermal for 2 min; injector temperature 5 250 C; flow rate 2 mL/min splitless with purge on at 0.2 min. Isopropyl mesylate was monitored at 79 and 123 m/z and methyl mesylate was monitored at 80, 95, and 110 m/z.

increase the rate of chemical degradation or physical change of an active drug substance or drug product using exaggerated storage conditions as part of the formal, definitive, storage program . . . .’’ For drugs to be stored at room temperature, i.e., 25 C, accelerated testing is defined as 40 C / 75% relative humidity. For other storage conditions, accelerated testing is to be carried out at 15 C above the long-term storage temperature. Accelerated testing, along with long-term testing (formal stability studies under normal storage conditions) is part of the formal stability studies required for regulatory submission. In contrast, the guideline differentiates stress testing as an investigation of the stability of the drug molecule that provides the foundation for the formal, definitive stability studies. Stress testing is distinguished by both the severity of the conditions and the focus or intent of the results. Thus, the guideline indicates that stress testing is intended to ‘‘. . . determine the intrinsic stability of the molecule by establishing degradation pathways in order to identify the likely degradation products and to validate the stability-indicating power of the analytical procedures used.’’ A longer, more detailed definition is also given in the guideline. Determination of the ‘‘intrinsic stability’’ characteristics of the drug molecule

104

B.A. OLSEN AND S.W. BAERTSCHI

includes developing an understanding of (1) conditions leading to degradation, (2) rates of degradation, (3) the chemical structures of the degradation products, and (4) the degradation pathways. Understanding the ‘‘intrinsic stability’’ characteristics of the drug molecule allows for prediction of stability-related concerns. This information can help direct both formulation and packaging development.

B. Strategy The overall strategy for investigation and control of degradation-related impurities can be illustrated as shown in Figure 12. Stress testing studies provide the foundation for the overall strategy. Various parts of the strategy are explored in more detail below. Stress testing studies involve exposure of the drug substance (and/or formulated drug product or drug-excipient mixtures) to the stress conditions of heat, humidity, photostress (UV and VIS), oxidative conditions, and aqueous conditions across a broad pH range. The intent is to induce 10–20% degradation of the parent drug. It is acceptable to conclude that a molecule is stable if no degradation occurs under reasonably stressing conditions. The inherent dilemma here is how to measure the degradation and detect degradation products prior to development of appropriate methods for the purpose. Without stress testing there is no way to assess whether or not the method will resolve and detect the degradation products. As discussed previously in the chapter, there are two major analytical approaches to the search for degradation-related impurities: chemistry-guided

FIGURE 12 Illustration of the overall strategy for investigating and controlling degradationrelated impurities.

5 INVESTIGATION AND CONTROL OF IMPURITIES

105

and technique-oriented. The technique-oriented approach involves the use of multiple analytical techniques to increase the chances of detecting unknown impurities. The chemistry-guided approach involves a scientific evaluation of the possible degradation pathways of the drug substance and choosing analytical techniques appropriate for the proposed degradation chemistry. The chemistry-guided approach should be iterative, i.e., it should involve multiple evaluation points as more information is gathered about the conditions leading to degradation and about the structures and spectroscopic characteristics of the resulting degradation products that are detected. As degradation products and pathways are elucidated, this information is used to reevaluate the relevance of the analytical methods. For example, if a drug degrades under a certain condition to yield a product whose structure indicates that a cleavage has occurred in the parent molecule resulting in (at least) two products but only one is detected, then the analytical conditions may need to be modified or a different analytical technique (or detector) may need to be employed as part of the investigation. Ideally, a method that resolves and quantitatively detects the parent drug and all the degradation products is desired. In reality, RP-HPLC with UV detection is by far the most common analytical technique currently used for detection of degradation and impurities. A good starting point for analyzing stressed samples is a broad screening RP-HPLC method with UV detection. Characteristics of such a method would include photodiode array UV detection, UV-transparent buffers for monitoring at low wavelengths (to increase universality of detection), and gradient elution to cover a wide polarity range. Such a method maximizes the chances of resolving, eluting, and detecting both polar and nonpolar degradation products. The broad screening method can be developed/optimized by analysis of partially degraded samples and the use of standard method development procedures and tools (e.g., optimization of the gradient, mobile phase solvent system, choice of column). The analysis of the stressed samples should reveal the degradation products formed under the various conditions. As information on the structures of the degradation products becomes available, degradation pathways can be proposed and evaluated. It should be remembered that the information about the stability and degradation of the drug substance from stress testing studies is predictive in nature (as opposed to definitive). That is, the degradation products observed during stress testing may or may not be relevant to actual storage conditions of the drug substance and/or to the degradation chemistry of the formulated product. As illustrated in Figure 13, the degradation products observed are ‘‘potential’’ degradation-related impurities. The ‘‘actual’’ or ‘‘significant’’ degradation-related impurities that occur during long-term storage or shipping (as revealed by accelerated testing and long-term stability) are the impurities that must be controlled. If the stress-testing studies are thorough and well designed, the ‘‘significant’’ degradation-related impurities should be a subset of the ‘‘potential’’ degradation products observed during stress testing. Focused methods can then be developed and optimized to monitor the ‘‘significant’’ degradation products for regulatory registration in marketing applications and use in quality control laboratories for product release

106

B.A. OLSEN AND S.W. BAERTSCHI

FIGURE 13 Cartoon illustration of hypothetical stress testing and accelerated or long-term storage stability HPLC chromatograms. Degradation products A^I are the degradation products formed during stress testing and are therefore classified as ‘‘potential’’ degradation products. Products B, C, D, E, and G are the products formed at significant levels during formal stability studies and are classified as the ‘‘actual’’or ‘‘significant’’degradation products.

and stability. The validity of this approach is apparent from an evaluation of the ICH definition of stress testing: . . .It is recognized that some degradation pathways can be complex and that, under forcing conditions, decomposition products may be observed that are unlikely to be formed under accelerated or long-term testing. This information may be useful in developing and validating suitable analytical methods, but it may not always be necessary to examine specifically for all degradation products if it has been demonstrated that in practice these are not formed.

The information gathered during stress testing of the drug substance is also used to guide the formulation of the drug product. As shown in Figure 12, drug-excipient compatibility studies should be performed to determine whether excipient blends or individual excipients have any adverse interactions with the parent drug. A broad screening method such as that developed for drug substance stress testing should be part of the analytical strategy to examine for excipient-catalyzed drug degradation. Once a suitable lead formulation has been developed, stress testing studies can be performed on the formulation leading to the detection/identification of ‘‘potential’’ degradation products. In an analogous manner to the strategy for the drug

107

5 INVESTIGATION AND CONTROL OF IMPURITIES

TABLE 4 Recommended Conditions for Assessing Hydrolytic,Thermolytic, and Photolytic Degradation Pathways

Solid state

Thermolytic

Hydrolytic

Photolytic

55, 70, 85 C, Low humidity (e.g., < 30% RH or no humidity control); 4–6 weeks

55, 70, 85 C, High humidity (e.g., > 70% RH); 4–6 weeks

5–10  the ICH min. confirmatory exposure; (1) thin layer of powder; (2) thin film (optional)

Aqueous solutions or slurries, pH 1–13, RT-70 C, 2–4 weeks

Aqueous solutions or slurries, 5–10  the ICH minimum confirmatory exposure; at pH’s above and below relevant pK’s.

Solution

substance, the ‘‘actual’’ or ‘‘significant’’ degradation products can be determined during accelerated and long-term stability studies. The key to the strategy outlined above is to have well-designed stress testing studies that form all potential degradation products. Thorough stress testing studies need to evaluate the four main degradation pathways of pharmaceuticals: (1) hydrolytic, (2) thermolytic, (3) photolytic, and (4) oxidative. Uncovering the potential hydrolytic, thermolytic, and photolytic degradations is relatively straightforward. The conditions outlined in Table 4 are suggested as practical stress conditions for assessing these pathways. The oxidative pathways are a bit trickier, and will be discussed in more detail later in this chapter.

C. Thermolytic Degradation To evaluate thermolytic pathways, elevated temperatures (e.g., 50–80 C) in the solid state and/or in solutions can be used. Above 80 C many compounds begin to degrade via different mechanisms, giving rise to degradation products that may not be observed under ordinary storage or shipping conditions. For solid-state stressing, the use of high- and lowhumidity atmosphere at the elevated temperatures is appropriate, and both thermolytic and hydrolytic pathways may be observed under such stressing conditions. Similarly, when aqueous solutions are stressed at elevated temperatures, both thermolytic and hydrolytic pathways may be observed.

D. Hydrolytic Degradation To evaluate hydrolytic pathways, aqueous solutions (using 100% aqueous conditions when solubility permits or adding an inert co-solvent such as acetonitrile) can be prepared at various pH conditions (e.g., pH 1–13)

108

B.A. OLSEN AND S.W. BAERTSCHI

and samples can be stored at elevated temperatures if no (or slow) degradation is observed at room temperature.

E. Photolytic Degradation The evaluation of photolytic pathways should be performed in both the solid state and in solution. Samples should be photostressed by exposure to irradiation significantly beyond the ICH minimum recommended confirmatory levels,20 both in the ultraviolet and visible spectral regions. The photoexposure should mimic a worst-case scenario, and thus we recommend exposure to simulated outdoor sunlight (D65), as per Option 1 of the ICH Q1B photostability guideline. An excellent resource for conducting such studies can be found in a two-part review article by Thatcher et al.21,22

F. Oxidative Degradation As mentioned previously, oxidative degradation pathways can be complex, and it is useful to consider the chemistry of oxidative degradation. The oxidative degradation of pharmaceuticals has been discussed in the literature,23–27 and we assert that there are three major pathways, (1) autoxidation or radical-mediated oxidation, (2) peroxide-mediated, and (3) photochemically induced (see Table 5). Traditionally, dilute aqueous peroxide solutions have been used for oxidative stress testing of pharmaceuticals. Landmark papers by Boccardi23,24 in 1992 and 1994 identified autoxidation as the major oxidative degradation pathway for pharmaceuticals. Boccardi showed that the use of radical initiators such as

TABLE 5 Three Major Oxidative Degradation Pathways 1. Autoxidation

2. Peroxide-mediated oxidation

3. Photochemically induced oxidation

5 INVESTIGATION AND CONTROL OF IMPURITIES

109

azobisisobutyronitrile (AIBN) gave results more predictive of autoxidative degradation than dilute peroxide solutions. It is worth noting here that radical-initiated oxidative chemistry can also be induced by transition metals, and stressing of the drug substance with iron(III) and copper(II) (in separate stress experiments) is recommended. In addition to autoxidative processes, peroxide-mediated oxidative degradation can occur that may not be observed using radical initiators as described above. For example, tertiary nitrogens are often observed to oxidize to N-oxides during long-term storage, but may or may not be observed during stress testing using radical initiators in aqueous/acetonitrile solutions. Since most tertiary nitrogens would likely be in the cationic protonated form under neutral aqueous conditions, the protonation undoubtedly helps to protect the nitrogen from oxidation. The use of pressured oxygen (e.g., 150 psi) in conjunction with radical initiators appears to aid in the formation of N-oxides.28 The formation of N-oxides can occur via direct oxidation29 as a result of the presence of peroxides in the excipients (e.g., PEG,30,31 polysorbates,32 and povidone33). Therefore, the use of dilute aqueous hydrogen peroxide to oxidatively stress the drug substance is appropriate. Room temperature is sufficient for such stressing, and elevated temperatures are not recommended because the oxidative degradation pathways catalyzed by peroxides at high temperature may not be predictive of real-world degradation. The third major oxidative pathway is photochemically induced oxidation (see Table 5). Such oxidation can occur from direct reaction of ground-state oxygen with the electronically excited state of the drug molecule or from photosensitization of triplet oxygen to singlet oxygen and direct reaction with the drug molecule. Photolytic stressing of the drug substance, as described above for assessing photolytic degradation pathways, is effective for the prediction of these oxidative pathways.

G. Common Problems Encountered During Stress Testing with HPLC-UV Although RP-HPLC with UV detection is by far the most common analytical technique used to search for and analyze degradation-related impurities, it presents significant problems faced by virtually every analytical researcher. Questions such as (1) Is everything being eluted off the column?, (2) Are there nonchromophoric products?, (3) What are the responses of the unknown impurities?, and (4) Are all the impurities amenable to the separation technique (e.g., solubility, volatility, stability)?, need to be asked. These questions all revolve around the concept of ‘‘mass balance,’’ i.e., The process of adding together the assay value and levels of degradation products to see how closely these add up to 100% of the initial value, with due consideration of the margin of analytical precision.34 Since during impurity investigations (e.g., stress testing/degradation studies), the response factors and other characteristics of the impurities are usually unknown, the researcher will undoubtedly confront questions about mass balance. One practical example of a problem related to mass balance is shown in Figure 14, where aqueous solutions (at pH 7) of drug substance A were

110

B.A. OLSEN AND S.W. BAERTSCHI

FIGURE 14 RP-HPLC-UV chromatogram (using an isocratic method) of an aqueous solution of drug substance A after stressing at 70 C for 14 days at pH 7.

stressed at 70 C for up to 14 days. RP-HPLC-UV analyses were performed using a previously developed isocratic method and the results indicated a loss of parent drug with only a minor increase in related substances (day 14 showed 4.9% degradation with only a 1.4% increase in impurities, calculated using the UV response factor of the parent drug). Close examination of the parent peak revealed a shoulder on the ‘‘tail’’ of the peak, indicating partial co-elution of a degradation product. To investigate this mass balance discrepancy, a broad-screening gradient HPLC method was developed. Analysis of the day-14 sample using this method (see Figure 15) revealed five additional degradation products not observed using the isocratic method, accounting for an increase in detected impurities of approximately 3%. This result is a dramatic example of the importance of using gradient HPLC methodology for impurity investigations. The use of highly resolving broadscreening gradient methods, with monitoring at low wavelengths using a PDA, can help to maximize the chances of resolving and detecting a wide range of impurities. Another example of a real-world problem encountered during degradation studies is illustrated in Figure 16. HPLC analysis of an aqueous solution of a drug substance (LY297802, Figure 18) after exposure to cool white fluorescent light during a photostability study revealed a substantial loss of parent drug with no apparent degradation products. PDA-UV analysis of all wavelengths between 200 and 400 nm did not reveal any degradation-related peaks. Obviously there is a significant ‘‘mass balance’’ issue! Closer examination of the photostressed solution revealed a hazy, insoluble film adhering to the surface of the glass solution container. The film was collected and analyzed by probe EI-MS, which quickly revealed that the film was elemental sulfur (S8). Since LY297802 contains a sulfur atom in the thiadiazole ring, it was apparent that the chromophoric thiadiazole moiety

5 INVESTIGATION AND CONTROL OF IMPURITIES

111

FIGURE 15 RP-HPLC-UV chromatogram (using a gradient method) of an aqueous solution of drug substance A after stressing at 70 C for 14 days at pH 7.

FIGURE 16 HPLC-UV chromatograms of LY297802 after 3 days and 7 days in a photostability chamber (cool white fluorescent lamps, 17,000 lux). Conditions: Zorbax RX-C8 column, 25 cm 3 4.6 mm id, 5 mm; A 5 buffer/acetonitrile 95/5, B 5 buffer/acetonitrile, 25/ 75; buffer 5 25 mM potassium phosphate with pH adjusted to 6.5 with NaOH,UVdetection using photodiode array withWaters Maxplot2 200 ^ 400 nm; gradient elution from 0% Ato100% B in 30 min, 1mL/min flow rate.

was being photodegraded to nonchromophoric and/or volatile products. When LC-MS analysis failed to detect any new peaks, another separation technique that was amenable to volatile compounds (GC) was employed. Hexane extraction of the aqueous light-degraded solution, previously made basic to give the free base of LY297802, followed by GC-FID analysis revealed two major degradation products in the degraded samples (Figure 17). GC-EI-MS analysis quickly revealed the ‘‘missing’’ degradation products as n-butyl-thiocyanate and 1-aza-bicyclo[2.2.2]octane-3-carbonitrile (see Figure 18).

112

B.A. OLSEN AND S.W. BAERTSCHI

FIGURE17 GC^FID analysis of a hexane extract from a photodegraded sample of LY297802.

Conditions: DB-1, 30 m 3 0.53 mm, 3 mm film thickness; oven temperature held at 40 C for 2 min, ramp to 70 C at 5 C/min, then ramp to 260 C at 10 C/min, hold isothermal for 2 min; detector FID at 300 C, helium carrier at 20 cc/min, helium makeup at 10 cc/min; 1 mL injection volume.

FIGURE 18 Structures of LY297802 and its photodegradation products.

The examples described here are illustrative of the ‘‘chemistry-guided approach.’’ An initial analysis is made based on good analytical chemistry, the results are evaluated carefully, and further steps are guided by scientific judgment. As the structures of degradation products are determined, the pathways of degradation can usually be proposed. The pathways help determine whether all the degradation products have been detected and what other possible products might be formed. These pathways and the conditions under which they occur form the basis for understanding the ‘‘intrinsic stability’’ characteristics of the compound, which once developed can be used to guide analytical and other development activities.

5 INVESTIGATION AND CONTROL OF IMPURITIES

113

H. Detectors The basic problem related to the determination of mass balance is the differential response of detectors to different compounds. As outlined in the ICH guidelines on impurities, it is acceptable to use the response factor of the drug substance to estimate levels of unknown impurities.1–4 In many cases, however, the response factor differences can be substantial and therefore, even with results that appear to show ‘‘mass balance,’’ uncertainty exists. The use of alternate detectors is one practical strategy for investigating issues related to mass balance. HPLC with MS detection is commonly employed as an alternate detector to UV. Mass spectrometry is generally regarded as a universal detector, but the response per unit weight depends greatly on the ionization type (e.g., positive or negative electrospray, atmospheric pressure chemical ionization, etc.) and on the ionization efficiency of the analyte under the given conditions. Three HPLC detectors that are viewed as giving a response that is proportional to mass are chemiluminescent nitrogen (CLN), evaporative light scattering (ELS), and refractive index (RI) detectors. RI detectors are useful for isocratic HPLC, but they suffer from variability in response, depending on the mobile phase composition, temperature, and dissolved gases, and are not very sensitive.35,36 ELS detectors should detect most nonvolatile substances, but the response will depend on the quantity and characteristics of the particles formed during solvent evaporation. For compounds of similar structures, one can expect response factors to be similar (þ/ approximately 10–20%). When compounds are substantially different, however, the responses can vary more than 10-fold.37 The CLN detector is based on combustion of the HPLC effluent in an oxygen-rich furnace to produce nitric oxide from nitrogen-containing compounds. The nitric oxide is reacted with ozone to produce nitrogen dioxide in an excited state, and photons are emitted upon return to the ground state. This chemiluminescent response is proportional to the number of moles of nitric oxide, and correspondingly to the number of moles of nitrogen in the analyte (with the exception of N2 and N ¼ N bonds). Nitrogen-containing compounds will therefore produce a signal that is directly related to the number of moles of nitrogen present. As long as the molecular formula of the unknown is known, the mass of the unknown can be determined. Recent articles by Nussbaum et al.37 and Korner38 describe the capabilities and limitations of this detector in more detail. The use of alternate detectors is an important tool to consider when investigating degradation-related impurities. If mass balance can be achieved using a detector that provides a response that is proportional to the mass of the analyte, such as the CLN, further investigations using other analytical techniques can be avoided.

I. Drug Product Stress Testing Developing an understanding of the intrinsic stability of the drug substance does not eliminate the need for specific stress testing of the

114

B.A. OLSEN AND S.W. BAERTSCHI

formulated product. Indeed, the literature is replete with examples of unexpected and unprecedented surprises from the formation of degradationrelated impurities specifically related to the excipients or packaging components, a few of which are cited here.39–41 Drug product stress testing begins with drug-excipient compatibility testing, although a discussion of this topic is beyond the scope of this chapter. An article by Serajuddin et al. provides an excellent discussion of drug-excipient compatibility testing.42 Stressing of the formulated product, both inside and outside of the packaging material is critical to avoiding surprises during formal stability studies that can affect the development time line. Stressing of the drug product should involve evaluation of elevated temperature, humidity, and photolytic stress as a minimum. It is also prudent to evaluate the effects of the solvents and pH conditions used during analytical sample preparation for assay and related substances testing. An interesting example of the identification of a formulation-related impurity is described by Jansen et al. for the case of duloxetine hydrochloride.43 Duloxetine hydrochloride is a secondary amine that is a potent inhibitor of both serotonin and norepinephrine reuptake. Since the drug is unstable in acid, oral dosing requires encapsulation within an enteric coating that consists of hydroxypropylmethylcellulose (HPMC) and HPMC acetate succinate (HPMCAS). During the stress-testing studies, a significant level of the succinamide impurity was formed, as shown in Figure 19. This formation occurred in spite of the fact that duloxetine was physically separated from the enteric coating by a barrier layer (see Figure 20) and did not react with succinic acid in stress studies. It was postulated that succinic anhydride was

FIGURE 19 Structures of duloxetine hydrochloride and a low-level impurity formed upon exposure of enteric-coated pellets to accelerated conditions of 40 C/75% relative humidity.

FIGURE 20 Cartoon representation of a cross-sectional view of the duloxetine hydrochloride enteric-coated pellets.

5 INVESTIGATION AND CONTROL OF IMPURITIES

115

FIGURE 21 Proposed pathway for formation of duloxetine succinamide.

cleaving from the HPMCAS backbone and that migration of either the drug substance or the anhydride facilitated the reaction (see Figure 21). The levels of this impurity that were forming under accelerated conditions threatened the viability of the formulation and the development time line, but the researchers were able to minimize the formation of the impurity by increasing the thickness of the barrier layer. It is clear that knowledge of the structure of this degradation product was critical to the development of an acceptable formulation for the marketed product.

V. SUMMARY AND CONCLUSIONS While the overall strategy for controlling process and degradation-related impurities should involve multiple analytical techniques, a chemistry-guided approach using the results in an iterative fashion can help determine which of the techniques will provide useful results in impurity-related investigations. For both process- and degradation-related impurities, the approaches described in this chapter can differentiate between those impurities that are ‘‘relevant’’ and those that are ‘‘potential.’’ A thorough knowledge of which potential process and degradation-related impurities are likely to be present after synthesis or upon storage (for both the drug substance and product) can be used to guide the development of efficient and robust ‘‘focused’’ control methods. Such knowledge also guides the selection of appropriate packaging and long-term storage conditions. The ultimate goal of controlling impurities in drug products is to provide safe, pure, and effective drug products for the patient.

ACKNOWLEDGMENTS The authors gratefully acknowledge the technical contributions of Patrick Jansen, W. Kimmer Smith and Liu Yang.

116

B.A. OLSEN AND S.W. BAERTSCHI

REFERENCES 1. International Conference on Harmonisation, Impurities in New Drug Substances, Q3A, January 1996. 2. International Conference on Harmonisation, Impurities in New Drug Substances, Draft, Q3A(R), February 2002. 3. International Conference on Harmonisation, Impurities in New Drug Products, Q3B, November 1996. 4. International Conference on Harmonisation, Impurities in New Drug Products, Draft, Q3B(R), October 1999. 5. Identification and Determination of Impurities in Drugs, Go´ro´g, S. (Ed.), Elsevier, Amsterdam, 2000. 6. Niessen, W. M. A. Chimia 53:478–483, 1999. 7. Hilhorst, M. J., Somsen, G. W. and de Jong, G. J. Electrophoresis 22:2542–2564, 2001. 8. Hilhorst, M. J., Derksen, A. F., Steringa, M., Somsen, G. W. and de Jong, G. J. Electrophoresis 22:1337–1344, 2001. 9. Ermer, J. and Vogel, M. Biomedical Chromatography 16:373–383, 2000. 10. Go´ro´g, S. Current Trends in Analytical Chemistry 1:11–18, 1998. 11. International Conference on Harmonisation, Specifications: Test Procedures and Acceptance Criterial for New Drug Substances and New Drug Products: Chemical Substances, Q6A, October 1999. 12. International Conference on Harmonisation, Impurities: Residual Solvents, Q3C, July 1997. 13. Go´ro´g, S., Babja´k, M., Balogh, G., Brlik, J., Csehi, A., Dravecz, F., Gazdag, M., Horva´th, P., Lauko´, A. and Varga, K. Talanta 44:1517–1526, 1997. 14. Giron, D. Boll. Chim. Farmaceutico 133:201–220, 1994. 15. Wirth, D. D., Miller, M. S., Boini, S. K. and Koenig, T.M. Org. Proc. Res. Dev. 4:513–519, 2000. 16. Wirth, D. D., Olsen, B. A., Hallenbeck, D. K., Lake, M. E., Gregg, S. M. and Perry, F.M. Chromatographia 46:511–523, 1997. 17. Pharmacopeial Forum, 26:193, 2000. 18. Markey, S. P. and Schmuff, N. R. Med. Res. Rev. 4:389–429, 1986. 19. International Conference on Harmonisation, Stability Testing of New Drug Substances and Products, Q1A, September 1994. 20. International Conference on Harmonisation, Stability Testing: Photostability Testing of New Drug Substances and Products, Q1B, November 1996. 21. Thatcher, S. R., Mansfield, R. K., Miller, R. B., Davis, C. W. and Baertschi, S. W. Pharmaceutical Technology, March 98–110, 2001. 22. Thatcher, S. R., Mansfield, R. K., Miller, R. B., Davis, C. W. and Baertschi, S. W. Pharmaceutical Technology, April 50–62, 2001. 23. Boccardi, G., Deleuze, C., Gachon, M., Palmisano, G., and Vergnaud, J. P. J. Pharm Sci. 81:2, 183, 1992. 24. Boccardi, G. Il Farmaco. 49:6, 431, 1994. 25. Hovorka, S. W. and Schoneich, C. J. Pharm. Sci. 90(3):253–269, 2001. 26. Johnson, D. M. and Gu, L. C. Autoxidation and antioxidants, In Encyclopedia of Pharmaceutical Technology (Swarbrick, J. and Boylan, J.C. Eds.) Vol. I, Marcel Dekker, Inc., New York, pp. 415–449, 1998. 27. Waterman, K. C., Adami, R. C., Alsante, K. M., Hong, J., Landis, M. S., Lombardo, F. and Roberts, C. J. Pharmaceutical Development and Technology 7:1, 1–32, 2002. 28. Unpublished results, Karen Alsante, Pfizer. 29. March, J. Advanced Organic Chemistry: Reactions, Mechanisms, and Structure, 3rd ed. John Wiley & Sons, New York, p. 1088, 1985. 30. Ginsburg, E. J., Stephens, D. A., West, P. R., Buko, A. M., Robinson, D. H., Li, L. C. and Bommiredi, A. R. J. Pharm. Sci. 89:6, 766–770, 2000. 31. McGinity, J. W., Hill, J. A. and LaVia, A. L. J. Pharm. Sci. 64:2, 356–357, 1975. 32. Chafetz, L., Hong, W.-H., Tsilifonis, D. C., Taylor, A. K. and Philip, J. J. Pharm. Sci. 73:8, 1186–1187, 1984.

5 INVESTIGATION AND CONTROL OF IMPURITIES

117

33. Hartauer, K., Arbuthnot, G., Baertschi, S. W., Johnson, R., Luke, W., Pearson, N., Rickard, E., Tsang, P. and Wiens, R. Pharmaceutical Development and Technology 5:3, 303–319, 2000. 34. International Conference on Harmonisation, Stability Testing of New Drug Substances and Products, Q1A, September 1994. 35. McNabb, T. J., Cremesti, A. E., Brown, P. R. and Fischl, A. A. Seminars in Food Analysis, 4:53–70, 1999. 36. Snyder, L. R., Kirkland, J. J. and Glajch, J. L. Practical HPLC Method Development, 2nd ed. John Wiley & Sons, NY, pp. 80–81, 1997. 37. Nussbaum, M.A., Baertschi, S.W. and Jansen, P.J. J. Pharm. Biomed. Anal., 27:983–993, 2002. 38. Korner, A., LC/GC, 20(4):364–373, 2002. 39. Qin, X.-Z, Ip, D. P., Chang, K. H.-C, Dradransky, P. M., Brooks, M. A. and Sakuma, T. J. Pharm. Biomed. Anal. 12(2): 2212–2233, 1994. 40. Wirth, D. D., Baertschi, S. W., Johnson, R. A., Maple, S. R., Miller, M. S., Hallenbeck, D. K. and Gregg, S. M. (1998). J. Pharm. Sci. 87(1): 31–39, 1998. 41. Schildcrout, S. A., Risley, D. S. and Kleeman, R. L. Drug Development and Industrial Pharmacy 19(10):1113–1130, 1993. 42. Serajuddin, A. T. M., Thakur, A. B., Ghoshal, R. N., Fakes, M. G., Ranadive, S. A., Morris, K. R. and Varia, S. A. J. Pharm. Sci. 88(7):696–704, 1999. 43. Jansen, P. J., Oren, P. L., Kemp, C. A., Maple, S. R. and Baertschi, S. W. J. Pharm. Sci. 87(1):81–85, 1998.

This Page Intentionally Left Blank

6 REFERENCE STANDARDS PAUL A. CULBERT a AND BRUCE D. JOHNSONb a

Eli Lilly Canada Inc.,ON, M1N 2E8 Pfizer, Inc., Ann Arbor, MI 48105

b

I. INTRODUCTION II. DEFINITIONS III. LIFE CYCLE A. Discovery B. Lead Declaration C. Exploratory Development D. Full Development E. New Drug Application (NDA) Submission F. Product Launch G. Product Maturation IV. GOVERNANCE A. Control and Use Procedures B. Qualification Procedures V. QUALIFICATION PROCESS A. CharacterizationTechniques B. Form Selection C. Assignment of Purity Case1: DS Information: Case 2: DS Information: D. Bridging Standards VI. SUMMARY REFERENCES

I. INTRODUCTION Pharmaceutical reference standards are a critical aspect in all phases of drug research, development, and commercialization. Reference standards serve as the basis of evaluation for both process and product performance and are the benchmarks for the assessment of drug potency for patient consumption. Reference standards may be required not only for the active ingredients in dosage forms but also for impurities, degradation products, starting materials, preservatives, process intermediates, and excipients. At the same time, the availability of reference standards and the degree to which they are characterized and governed is often dependent on the stage of

119

120

P.A. CULBERT AND B.D. JOHNSON

the drug development process. The FDA has not provided practical guidance on the topic of reference standards,1 but information does exist in the form of best practice documents from a variety of sources. Therefore, this chapter is intended to provide clarity to the overall role, qualification, and governance of reference standards used in the pharmaceutical environment. The scope of this chapter will be limited to the discussion of nonbiological (i.e., synthetic small molecule) reference standards and will not include the discussion of compendial reference standards. Additionally, for reasons of clarity, the discussion will focus on reference standards of active pharmaceutical ingredients. However, the same considerations should generally be made for reference standards of synthetic impurities, degradation products, and the others mentioned above.

II. DEFINITIONS Terms used to describe reference standards differ among industries and also among companies in the pharmaceutical industry. The following definitions, consistent with literature references,2–13 will be used throughout this chapter. Drug substance: The drug substance (DS), also known as a new chemical entity (NCE) or the active pharmaceutical ingredient (API), in early phases of drug development consists of the following components: (a) active moiety, (b) counter-ion, and (c) water or solvent molecules that are known components of the DS crystal structure. If the single crystal structure of the DS is unavailable, the DS can be defined as a hydrate or solvate based on the results of other analytical tests and the scientific judgment of the scientist. Active Moiety: The active moiety is the pharmacologically active species, expressed in terms of the neutral molecule (e.g., free base or free acid). For chiral substances, the active moiety is the enantiomer specified for clinical development. When racemates are specified for development, however, the active moiety refers to both enantiomers. For pro-drugs, the active moiety refers to the pro-drug and not the parent drug. Drug Product: The drug product (DP), also known as finished product or dosage form, is the DS formulated with inactive ingredients (excipients) for appropriate delivery to either human or veterinary patients. Reference Standard: A reference standard is broadly defined as certified material or substance, supplied by a certifying body, which exhibits one or more properties that are sufficiently well established (and assigned) that it may be used for calibration of an apparatus, assessment of a measurement method, and assigning values to materials. Reference standards in pharmaceutical sciences generally fall within three categories: analytical reference standards, working standards, and authentic materials. Analytical Reference Standard: An analytical reference standard (ARS), also known as primary standard or gold standard, is defined as a batch of drug substance, whose purity is independently established and accepted without reference to other standards. The ARS is the benchmark against

6 REFERENCE STANDARDS

121

which secondary standards are judged. The ARS is also used to determine the amount of active moiety present in the DS or DP. The material is of the highest purity (> 99.5%) that can be obtained by reasonable effort. Typically the material is obtained either by purification of existing production material or, if necessary, specifically prepared by independent synthesis. The material is thoroughly characterized to assure its identity, strength, quality, and purity. The structural identity is performed via independent spectral interpretation, and all impurities should be determined on a weight basis if possible in order to establish full mass balance accountability. Additionally, there should be a complete understanding of the material’s inherent stability, appropriate storage containers and conditions, and other potential chemical transformations that may occur in stressed samples. Ultimately, the ARS must satisfy all identity, assay, and purity aspects of the API specification and have a satisfactory physicochemical measurement profile. Working Standard: The working standard (WS), also known as secondary standard, in-house standard, or laboratory standard, may be used for routine analytical comparison of batches of API and/or dosage forms in cases where a high-purity ARS is difficult to obtain. The WS is generally specific to the phase of drug development in which it is being used. During early stages of drug development, the WS may be the only reference standard available for an NCE. The purity of a WS is typically very high (> 95%) but may be assigned without the same rigor as applied to the qualification of an ARS. In fact, reference standards of higher purity at this early stage of drug development may not be preferred, as the existing impurities may be used to establish chromatographic system suitability. During later stages of drug development, the working standard is typically of higher quality and purity. In fact, qualification of the material is usually based on identity and assay comparisons to the established ARS. The WS is often widely used in place of the ARS to conserve material. Ideally, the WS should be prepared by the normal synthetic route used for the manufacture of the API and subjected to further purification if necessary. Authentic Materials: Authentic materials (AM) are reference standards, which are qualified for identity and approximate purity. The chromatographic purity of an AM need only be 80% or greater. They are not used in quantitative assays but are generally used to establish chromatographic system suitability and as identity comparators for spectroscopic assays. Authentic materials are usually available in small amounts and are frequently obtained by preparative chromatography.

III. LIFE CYCLE As every NCE brought forward into development is unique, the exact role that reference standards take for a specific drug candidate will vary but is generally keyed to drug development milestones. At any phase of drug development, the reference standard should be assessed versus its

122

P.A. CULBERT AND B.D. JOHNSON

intended use and a balance should be struck between resource commitments, scientific judgment, and regulatory requirements. The following descriptions are not meant to be definitive in nature but are general practices.

A. Discovery Discovery is generally considered the phase of drug development where numerous lead compounds are generated that are active against an established biological target. The predominant reference standards, if needed during this phase, are AMs. Time and resource commitments render impractical the generation and full characterization of analytical reference standards for the multitude of NCEs. The principal analytical measurements made during this phase are relative in nature. The AMs generated during this phase are typically characterized for identity by NMR, IR, or mass spectrometry. Approximate purity of the AM is commonly achieved via high-performance liquid chromatography (HPLC) with UV detection; however, automated high-throughput procedures such as a chromatographic system connected to an array of detection modes14 may also be employed. Additionally, only small amounts of the AM are readily available from a firm’s compound distribution system.

B. Lead Declaration Lead declaration is generally considered the phase of drug development where formal preparations are made to introduce an NCE into clinical studies. Prior to this phase, a single NCE that is believed to be highly effective in treatment of the targeted indication has been identified in preclinical studies. The reference standard for analytical quantification during this phase is typically a WS. Often the DS used in the dosage form candidate is frequently utilized as the WS, as it is readily available. Absolute versus comparative methods of analysis should be predominantly relied upon to determine the purity of the DS, because patient safety is especially imperative as the lead compound moves toward first-in-human studies. Therefore, the widespread need for an ARS is minor during this phase of development.

C. Exploratory Development Exploratory development generally encompasses the investigational new drug (IND) submission and phase I stage in the drug development life cycle. It is during this phase of drug development that patient safety is demonstrated in clinical trials for the new chemical entity. It is also during this phase of drug development that an ARS is prepared. Information gained on the various forms of the DS from the lead declaration phase as well as from the ongoing DS stability studies should be critically scrutinized. Methodologies to fully characterize the proposed ARS should be developed and any nuances of the material should be explored. However, the reference standard for analytical quantification during this phase is still predominantly the WS.

6 REFERENCE STANDARDS

123

D. Full Development Full development is generally considered as phases II and III in the drug development life cycle. It is during these phases that efficacy is demonstrated for the desired indication and large population clinical trials are performed. It is also during these phases of drug development that the ARS reaches maturity in terms of scientific assessment. The physical form and appropriate storage conditions for the ARS have now been selected as a result of extensive characterization of the material by fully established analytical methodologies and an understanding of its inherent stability. Routine use of the ARS is implemented for analytical testing, which demonstrates ruggedness of the material, and the ARS is placed on a formal recertification program. Procedures to generate bulk amounts are fully established by this point. Widespread use of a WS occurs in this phase only if there is a need to conserve the amount of ARS used and the WS has been fully certified against the ARS.

E. New Drug Application (NDA) Submission The testing and qualification of reference standards should continue such that the necessary documentation (internal and external reports, certificates of analysis, stability reports, supporting raw data) is complete from both a regulatory and scientific standpoint at the time of the NDA filing. A key analytical contribution to an NDA filing is the methods validation package. A firm must provide all materials—including reference standards— necessary to perform the analytical methods that will be registered with the NDA.2 The FDA may, in fact, choose to verify the performance of the submitted methods within their laboratories. Some firms may choose to set aside the necessary reference standards at the time of filing, while other firms may wait until the validation package is requested. A slow response to the FDA’s request for the validation package may delay approval of the application. It is therefore incumbent upon the project team to ensure that the necessary materials—reference standards, samples, columns, and special reagents if applicable—can be supplied to the FDA on very short notice. Additionally, care should be taken to ensure that the reference standards are within their validated shelf lives when supplied to the FDA.

F. Product Launch The drug development team must ensure that reference standards will be available to support the launch of a newly approved product once FDA approval is obtained. The quality control function is typically separate from the analytical development function in most firms, so a high level of cooperation and communication are critical. As launch approaches, the drug development team must work closely with the commercial manufacturing quality control unit to coordinate the supply of reference standards to the firm’s compound distribution system. It is useful to involve marketing forecasts to predict the approximate amounts of reference standards—based on projected batches—that will be necessary to support a worldwide launch.

124

P.A. CULBERT AND B.D. JOHNSON

Grams of ARS Requested

Total ARS Requested

FIGURE 1 Atorvastatin Analytical Reference Standard Disbursements,1992^1997.

As an example, Figure 1 shows the disbursements for Atorvastatin ARS,15 the reference standard for LipitorÕ , in the years prior to and immediately after its launch in 1996 and gives an indication of the amounts of ARS required for a widely accepted prescription drug.

G. Product Maturation Several years after the successful launch of a new product, it is typical for North American pharmaceutical firms to become involved in the United States Pharmacopeia-National Formulary (USP-NF) revision process to allow inclusion of their API and formulations in the USP-NF. The USP-NF and their supplements are a compilation of monographs prepared under the authority of the Pharmacopeial Convention. The purpose of the USP-NF and supplements is to provide authoritative standards and specifications for materials and substances and their preparations that are used in health care or for the improvement or maintenance of health.16 The revision process— as outlined in the USP-NF—begins with ‘‘inquiries, comments and suggestions for revision in the USP-NF text’’ mailed directly to the USP.17 Participation in this process ensures consideration of the innovator’s methods, specifications, and reference standards for inclusion in the USP-NF. Approval for inclusion in the USP-NF is gained after consideration by an Expert Committee of the USP Council of Experts. If the Expert Committee so recommends, the proposed revision is published in the Pharmacopeial Forum (PF) for review and comments by interested readers of the PF. The comments and data submitted by interested readers are reviewed and addressed, and the Expert Committee then decides whether to recommend to the USP Council of Experts that the proposed revision should be adopted in the USP-NF. The development of USP Standards through the cooperation of industry ‘‘promotes uniform quality of drugs as an aid to the public health.’’18

6 REFERENCE STANDARDS

125

The USP typically obtains candidate ARS material for a newly approved article from the pharmaceutical innovator but may also source the material from other manufacturers. Occasionally, the USP may synthesize sufficient material for use as an ARS. Once sufficient material to meet the needs of the USP for several years is sourced, a collaborative study is performed between the laboratories of the USP, the innovator, and one or more neutral or interested parties to fully characterize the material. The data obtained from the collaborative study are then submitted to the USP Reference Standards Committee who reviews them to ultimately determine whether the Reference Standard is acceptable for its intended use in support of the relevant USP-NF tests.

IV. GOVERNANCE The requirements for the laboratory control of reference standards was embedded in the Code of Federal Regulations. Title 21 CFR Part 211 section 194 subsection 8b states ‘‘Complete records shall be maintained of any testing and standardization of laboratory reference standards, reagents, and standard solutions.’’ It falls to the individual firm to develop procedures to give inspectors the best possible assurance of compliance to the regulation. The central role of reference standards in the assurance of patient safety makes them a common focal point for regulatory inspections. Development of sound procedures and the strict adherence to such procedures is therefore critical. The pharmaceutical firm must define adequate qualification of reference standards through one or more standard operating procedures. Reference standards SOPs typically fall into two categories: Reference standard control and use procedures and reference standard qualification procedures.

A. Control and Use Procedures A Reference Standards Control and Use Standard Operating Procedure (SOP) would typically include assigned responsibilities and procedures to ensure that: .

.

.

A scope for the reference standards covered by the procedure has been defined. For instance, reference standards used for the calibration of equipment may specifically be excluded from the SOP. Reference standards enter the laboratories under adequate control. These procedures typically identify approved sources and specify the documentation (certificates of analysis, qualification reports) that must accompany the reference standard into the laboratory. The accompanying documentation must be available during a GMP inspection. All reference standards entering the laboratory are properly logged. Information typically logged is identification (compound number or name), lot number, storage conditions, retest date, and the analyst to whom the standard is assigned.

126

P.A. CULBERT AND B.D. JOHNSON .

.

. .

Reference standards are labeled such that their identity, potency, correct use, and correct storage can be assured at all times. Reference standard labeling must also ensure that laboratory personnel handling the material are properly informed of safety risks associated with potentially harmful chemicals. Reference standard users are kept informed of changes to the status of standards currently in use in the laboratory. These changes may include updated potencies based on new test results, the reporting of unusual results or any other results that might compromise the integrity of the data generated using the reference standard in question. The reference standard is removed from the lab and properly disposed of on or before the date of requalification or expiry. Corrections for solvents and moisture are correctly applied. The drying of the reference standard before use, if stated in the method, will eliminate residual solvent(s), unbound moisture, and sometimes bound moisture (depending on the drying conditions). The drying step is always included for hygroscopic compounds. On the other hand, drying can result in the loss of a hydrate or cause degradation in heatsensitive compounds.

B. Qualification Procedures To ensure the compilation of consistent, complete and scientifically defensible regulatory submissions, the qualification of reference standards must be governed by a standard operation procedure. A reference standard qualification SOP will typically assign responsibility for, and include procedures to ensure that .

. .

. .

. . .

A scope has been defined that specifically states what types of standards will be qualified according to the procedure and which types of standards are excluded. Approved sources for reference standards are identified. Reference standards are properly categorized for their intended use. A definition section that clearly links the type of reference standard to its intended use is critical. A minimum basis set of analytical tests is described for each type of analytical standard. A consistent, scientifically defensible process is followed when purity values are assigned. A clear rationale for the calculation of purity values as well as clear instructions on the use of purity values must also be included. Retest, requalification, and expiry dates are appropriately assigned. A well-defined retest schedule is applied to reference standards used for quantitative purposes. The test results are recorded in a consistent and traceable manner. These procedures may give details on the method for reporting results for submission to a regulatory agency, in the correct format

6 REFERENCE STANDARDS

.

127

for certificates of analysis, and even the correct storage area for electronic and paper copies of the supporting reports and raw data. The labeled potency of the reference standard remains valid from the time of first use through to the requalification or expiry date indicated. The project scientist must ensure that the stability of the chemical is understood and that the assigned requalification or expiry dates are scientifically defensible.

V. QUALIFICATION PROCESS The critical parameter in the discussion of reference standards is the process of qualification. FDA guidance on this topic states that ‘‘A noncompendial standard should be thoroughly characterized to assure its identity, strength, quality and purity.’’3 It is therefore left to the pharmaceutical firm to establish ‘‘thorough characterization’’ for the ARS of each NCE. The overall qualification process for an ARS includes characterization of the material, proper form selection, and assignment of purity. Some general considerations in this process include .

.

. .

The expectation that the tests and methods used to characterize an ARS will be different from, and more extensive than, those used to control the new drug substance The desirability of characterization methods which are not comparative (i.e. dependent upon the availability of a previously designated reference standard of the same drug substance) The necessity of some degree of blinding when testing is performed by an outside laboratory, in order to assure unbiased results The requirement that the results generated during characterization be cross-referenced with each other and the results achieve a full accountability of mass balance

Taken together, the data from a series of experiments must show in a scientifically defensible manner that the labeled purity and the identity of the reference standard are both accurate and appropriate for the reference standards intended use.19,20

A. Characterization Techniques An examination of the literature,10,21–24 authoritative guidances,6–9 and current industrial best practices, suggests that the analytical techniques in Table 1 be considered for the characterization of reference standards. Other techniques are occasionally employed but are not discussed here. These may include: particle size analysis,25 nephelometry, heavy metals analysis,26 surface area,27 bulk density,28 pH,29 dissociation constants, microbiological testing30 and other spectroscopic measurements (e.g., NIR, fluorescence, CD, etc.). What follows is a discussion of the test methods commonly used to qualify reference standards and the pertinent information gained from these techniques. Frequently referenced textbooks have been cited and

128

P.A. CULBERT AND B.D. JOHNSON

TABLE 1 Techniques Commonly Applied to the Qualification of Reference Standards Appearance (physical description) Visual inspection Optical microscopy Identification (proof of structure) Elemental analysis UV/Visible spectroscopy Infrared spectroscopy Raman spectroscopy X-ray diffraction High-resolution mass spectrometry Nuclear magnetic resonance spectroscopy Optical rotatory dispersion Purity Loss on drying Karl Fischer titration Residue on ignition Thermogravimetric analysis Differential scanning calorimetry High-performance liquid chromatography Ion chromatography Electrophoretic separations Supercritical fluid chromatography Thin-layer chromatography Chiral chromatography Gas chromatography Assay Titration Phase solubility analysis

can provide a broader scope of technical details. Many of these techniques are described in the United States Pharmacopoeia, the European Pharmacopoeia, and the Japanese Pharmacopoeia. Specific pharmaceutical applications of these techniques can also be found in other chapters of this book, volume 3 of this series,31 or current literature reviews.32 1. Appearance (Physical Description) Visual Inspection: The visible characteristics—color, texture, and morphology, as well as visible contamination—are important and sensitive

6 REFERENCE STANDARDS

129

measures of chemical purity. Substances that are reactive in the presence of light or moisture will nearly always undergo an easily observed color or texture change when exposed to these agents. Appearance is thus an important measure to be included in the testing of reference standards. Color chips are available33 to assist in describing the reference standard. Optical Microscopy: Optical microscopy deals with the examination of material and identification of crystallinity under a microscope. A crude assessment of reference standard morphology, homogeneity, and birefringence, a qualitative aspect of light refraction by crystals, are easily observed using this technique. Crystalline particles will appear to change from bright to dark (or change colors) and noncrystalline (amorphous) particles will not change as the stage of a polarizing microscope is rotated.34 2. Identification (Proof of Structure) Elemental Analysis: The carbon, hydrogen, and nitrogen contents of reference standards may be determined by combustion analysis. Combustion analysis thus provides an orthogonal determination of the molecular formula and the purity of the reference standard. Poor agreement between the theoretical and experimental elemental compositions is generally an indication of an impure substance or a discrepancy between the theoretical and actual molecular formulas. Elemental analysis may also identify the presence of solvates and inorganic contaminants. Elemental analysis is typically performed in specialized laboratories under blinded conditions. Ultraviolet-Visible Spectroscopy: Ultraviolet-visible (UV-VIS) molecular absorption spectrophotometry (often called light absorption spectrophotometry or just UV-visible spectrophotometry) is a technique based on measuring the absorption of near-UV or visible radiation (180–770 nm) by molecules in solution.35,36 Reference standard characterization by UV-VIS spectophotometry includes determining the absorption spectra and the molar extinction coefficient. These two spectral characterizations are used as identifiers of reference standards. Infrared Spectroscopy: Infrared (IR) spectroscopy deals with the interaction of IR radiation with matter. Most commonly, the spectrum is obtained by measuring the absorption of IR radiation (10–12,900 cm1).35,37–39 Because of the rich spectral information that can be obtained and multiple spectroscopic modes that do not require dissolution of the material, IR is the most commonly used identification technique for reference standards. IR is also applied to characterize the presence of polymorphs and solvates. Raman Spectroscopy: Raman spectroscopy is based on the measurement of scattered electromagnetic radiation as a result of the irradiation of matter. Raman spectroscopy is considered complementary to IR spectroscopy, as the two techniques provide a complete vibrational picture of material. Raman spectroscopy is not as widely used for reference standard identification purposes as IR spectroscopy because of the lack of familiarity. However, Raman spectroscopy is an extremely powerful tool in characterizing the presence of polymorphs.35,39

130

P.A. CULBERT AND B.D. JOHNSON

X-ray Diffraction: When applied to single crystals, the technique is known as single crystal X-ray diffraction; applied to powders, the technique is known as X-ray powder diffraction (XRPD). Single crystal X-ray diffraction provides definitive structural information in terms of bond lengths and angles for reference standards, while X-ray powder diffraction yields information on the degree of crystallinity of the material as well as helping to identify polymorphs.38,40–42 Where suitable crystals can be obtained, single crystal X-ray diffraction provides the most definitive proof of structure for a reference standard. Mass Spectrometry: Mass spectrometry refers to the application of any of a number of mass-separative techniques to charged molecules in the gas phase. Techniques that are capable of mass resolution in the low parts per million (ppm) range are considered high resolution (HRMS)—the two most common instruments are magnetic sector and time-of-flight mass spectrometers. Time-of-flight mass spectrometry (TOFMS) has emerged as a highly accurate, relatively inexpensive technique for obtaining high-resolution mass spectra with errors of less than 5 ppm.43,44 Thus, HRMS is a powerful tool for confirming the molecular formula of a reference standard by providing an accurate molecular mass. HRMS also provides orthogonal structural information to techniques such as NMR but may not be appropriate for nonvolatile molecules and molecules that are prone to fragmentation under mass spectrometric conditions. Nuclear Magnetic Resonance: 13C and 1H nuclear magnetic resonance (NMR) spectroscopy and the associated multidimensional experiments can provide definitive evidence for determining the structure of reference standards. Fully assigned NMR spectra are an integral part of the registration package for new chemical entities and their associated impurities. Solid-state NMR is also valuable for the identification and monitoring of polymorphism in drug substances and products.40,45,46 Optical Rotatory Dispersion: Optical rotatory dispersion (ORD) is the measurement of the angle of rotation of the plane of linearly polarized radiation by an optically active (chiral) species. The dependence of the optical rotation on wavelength is termed the ORD. Specific rotation typically involves measuring the rotation of light at a fixed wavelength under controlled temperature of a 1% reference standard solution.35,47 The specific rotation is used to confirm the identity of a chiral reference standard. ORD can also be used to determine chiral purity but is generally limited by sensitivity. 3. Purity Loss on Drying: Loss on drying is a technique that determines the amount of volatile matter of any kind that is driven off under the temperature and pressure conditions specified. The technique is based upon accurate mass measurements, before and after drying. Ideally, the percent of material lost on drying should correlate to the total amount of solvent and moisture obtained by other measurements, when determining mass balance accountability for reference standard material. However, solvents/moisture that forms solvates/hydrates with the reference standard are often difficult

6 REFERENCE STANDARDS

131

to quantify by this technique, as decomposition of the material occurs before the solvent/moisture is driven off. Karl Fischer Titration: Moisture content of reference standards is commonly and accurately determined by Karl Fischer (KF) titration. KF titration determines total water content, including both the free water and water of hydration trapped in the crystal structure. Coulometric KF-titration is the most sensitive application of the method and is normally used for the small amounts of water found in reference standards. A volumetric technique is available for samples containing large amounts of water. Interference is a relatively common problem in KF titration and thus, due caution must be exercised in the interpretation of KF results. Reagents have been designed to cope with interfering functional groups.48–50 Residue on Ignition: Residue on ignition (ROI) is a pharmacopeial method51 that complements chromatographic techniques by providing information on inorganic impurities in reference standards. The technique involves the charring, digestion, and ignition of a reference standard sample. The resulting residue is then weighed to determine the amount of inorganic impurities present as their sulfate salts. ROI can be coupled with spectroscopic techniques such as atomic absorption, flame emission, and inductively coupled plasma (ICP) spectroscopy to provide identification and quantification of inorganic impurities. Thermogravimetric Analysis: Thermogravimetric analysis (TGA) measures the change in the mass of sample as the temperature is changed and is most commonly used to study the loss of solvents or other solid ! solid þ gas reactions. Therefore, as part of a reference standard characterization, TGA can often provide information regarding moisture and solvent levels, an indication if the material is a hydrate or solvate, and the rate of reaction Asolid ! Bsolid þ Cgas.42,52,53 Additionally, if a spectroscopic instrument (i.e., IR) is placed inline with the effluent atmosphere, the identity of the solvent lost during heating from the material can often be determined. As with the other techniques of purity determination, the information gained with TGA should be compared with information from other reference techniques such as gas chromatography and single crystal X-ray diffraction. Differential Scanning Calorimetry: Differential scanning calorimetry (DSC) is a method that measures the difference in energy (heat flux or heat flow) between a reference and a sample. The result of a DSC analysis is a thermogram, a plot of temperature difference versus temperature and represents the enthalpies of various processes occurring in the heated sample, such as solvent loss, crystallization, polymorphism, and chemical reactions. Additionally, DSC is an absolute method and with proper calibration can be used to accurately measure the melting point and purity of the reference material.52,53 High-Performance Liquid Chromatography: High-performance liquid chromatography (HPLC) is the preeminent analytical separation technique in pharmaceutical chemistry that delivers the fundamental impurity information for reference standard qualification. Numerous excellent reviews and texts are available54–58 as is a general chapter in the USP.59

132

P.A. CULBERT AND B.D. JOHNSON

The predominant mode of HPLC, reversed phase, involves the separation of material based on the partitioning between a relatively polar mobile phase and a nonpolar stationary phase. Normal phase HPLC—nonpolar mobile phase and polar stationary phase—is considered an orthogonal technique to reversed-phase HPLC when qualifying reference standards. In fact it is common for the elution order to be entirely reversed when switching an analysis from reversed to normal phase. Therefore, highly nonpolar impurities can be easily characterized by normal phase separations. HPLC remains a powerful tool for the qualification of reference standards because of the possibility of coupling the separation process with a myriad of complementary detection techniques. In 1998, five detection methods were reported to account for at least 95% of all HPLC analyses. The dominant detection techniques were UV-VIS, fluorescence, refractive index, electrochemical, and conductivity detection.60 UV-VIS with diode array detection (DAD) has become increasingly common in pharmaceutical analysis owing to the development of sensitive and low-cost equipment. DAD detectors allow real-time collection of the entire UV-VIS spectrum of each peak in an HPLC chromatogram. DAD software packages can typically provide peak purity assessments and three-dimensional representations of chromatographic experiments. The peak purity algorithms available with these software packages are particularly useful in the qualification of reference standards. The homogeneity of the peak can be determined61 by comparing the UV-VIS spectra at different points in the peak under study against reference spectra. The degree to which the two spectra match can be assessed mathematically to yield a numerical value or ranking for peak purity. The peak purity value is judged versus preestablished criteria that are indicative of an appropriate degree of purity. Care must be taken when using this technique as co-eluting impurities often have UV-VIS spectra that are nearly indistinguishable from the peak of interest. In fact, this is an inherent weakness in UV-VIS detection and highlights the importance of complementary detection techniques. Examples of other detection modes that are either commercially available or have been recently reported in the literature include evaporative light scattering, chemiluminescent, IR, Raman and inductively coupled plasma spectroscopy. New and improved detectors are continuously being developed and these new technologies are reviewed on a regular basis.61 Additionally, the rapid development of hyphenated techniques, such as liquid chromatography-mass spectrometry and liquid chromatographyNMR instruments, has rapidly advanced these techniques as tools for reference standard qualification. The overriding consideration when qualifying reference standards is the importance of using complementary detection techniques. Ion Chromatography: Ion chromatography (IC) is a mode of HPLC in which ionic analyte species are separated on cationic or anionic sites of the stationary phase. The detection techniques largely fall under three categories: electrochemical, spectroscopic, or post-column reactions. In general, IC provides an orthogonal separation mechanism to traditional reversed-phase HPLC (RP-HPLC).54,63 This technique can be exploited to quantify ionic

6 REFERENCE STANDARDS

133

species such as polar impurities, residual metals or counter-ions of the main analyte.59 Electrophoretic Separations: The popularity of electrophoretic separations has exploded in the last decade with the advent of capillary electrophoresis. Capillary electrophoresis (CE) has many practical advantages over standard gel electrophoresis and has been broadly diversified into a family of specialized techniques. The most universal of these techniques is capillary zone electrophoresis (CZE) or free-solution CE. The principal of separation is based upon the charge-to-size ratio of the analytes, where the sample mixture in a free-flowing background electrolyte (BGE) solution is exposed to both an electrophoretic and electroosmotic flow in an open capillary via a high-voltage power supply. Additionally, most classical RP-HPLC detection techniques are compatible with CZE.63,64 CZE provides a completely orthogonal separation mechanism to traditional RP-HPLC. Similar to IC, this technique can be exploited to quantify ionic species in reference materials such as polar impurities, residual metals, or counterions of the main analyte. However, by adding selected modifiers to the BGE solutions, separations based on both RP and enantiomeric modes can supplement the charge-to-size mechanism. Supercritical Fluid Chromatography: Supercritical fluid chromatography (SFC) is a column chromatographic technique in which a supercritical fluid is used as a mobile phase. A supercritical fluid is a gas or liquid brought to a temperature and a pressure above its critical point. Because of the unique properties of the mobile phase and the optional addition of organic modifiers, the SFC separation process is generally considered a blend of both gas chromatography (GC) and liquid chromatography (LC). In fact, the SFC column may either be a packed column, comparable to an HPLC column, or an open capillary column, comparable to a GC column. Detection is performed either on-line (i.e., UV-VIS) or after the expansion of the fluid [i.e., flame ionization detection (FID)]. SFC is generally performed in the normal phase (NP) mode and often NP-TLC (thin-layer chromatography) or NP-HPLC methodologies can be readily adapted to SFC methodologies.66,67 SFC generally provides an orthogonal separation mechanism to traditional RP-HPLC. Because of the similarity in the chromatographic measurement process to HPLC, this technique can be exploited to accurately quantify nonpolar impurities of reference materials. Thin-Layer Chromatography: Thin-layer chromatography (TLC) coupled with densitometric detection is a highly sensitive—though often ignored68,69—method for assessing the purity of reference standards. HP-TLC is an improved version of TLC that employs stationary phases of decreased thickness and particle size resulting in improved resolution over shorter elution distances. Components separated by TLC can be visualized by a number of techniques.70 Short-wave UV irradiation is one of the most convenient and universal of these detection techniques. Densitometers are available that measure the relative amounts of the separated components. In contrast to HPLC, all components in the sample mixture remain and are visualized on the TLC plate. TLC with densitometric detection has shown good linearity and limits of quantification in the low nanogram range when

134

P.A. CULBERT AND B.D. JOHNSON

coupled with post-chromatographic derivitization.71 TLC with densitometric detection has been directly compared with HPLC with favorable results.72,73 It is an ideal complementary chromatographic technique to HPLC because of the universal visualization techniques available and the often-complete accountability for all impurities. Chiral Chromatography: Chiral pharmaceutical active moieties are increasingly common. Therefore, the determination of enantiomeric purity is imperative for the qualification of reference standards. The basis for the majority of enantiomeric chromatographic separations is the formation of a transient diastereomer, which is commonly performed either by using a chiral stationary phase or by adding a chiral selector to the mobile phase. Numerous chromatographic techniques including those discussed previously have been demonstrated to achieve chiral separations through one of these routes.74 Chiral separation technology has now matured to the extent that sensitivity to 0.1% of the minor enantiomer is routinely achieved. In general, the minor enantiomer is considered an impurity of the reference standard. Gas Chromatography: The role of gas chromatography (GC) in the qualification of reference standards is primarily for the determination of organic volatile impurities (OVI) with FID, according to the USP.75 The polar nature and presumed poor volatility of most pharmaceutically important molecules relegates GC far down the list of techniques used in pharmaceutical analysis. Nevertheless, GC is an excellent analytical technique for special cases when volatile impurities are suspected to be present in a reference standard or for cases where a simple derivitization reaction may impart sufficient volatility to a molecule to enable analysis. GC is complementary to loss on drying (LOD) and KF titration in the determination of total volatile impurities in reference materials. 4. Assay Titration: The determination of assay values for reference standards, counter-ions, or impurities can often be independently determined via titration. While titration assays generally have less selectivity in comparison to chromatographic methods, the advantages of a broad spectrum of classical titration techniques that exist for organic functional groups is often overlooked. The methodologies include not only classical potentiometric acid/base titrations but also nonaqueous, redox, indirect, precipitation, and derivatization titrations.76–79 Phase-Solubility Analysis: Phase-solubility analysis is the quantitative determination of the purity of a substance through the application of precise solubility measurements. Phase-solubility analysis is applicable to all species of compounds that are crystalline solids and that form stable solutions. Phase-solubility analysis is an absolute method that provides a totally independent assessment of purity and does not require a fully characterized reference standard or identity of impurities. Instead, this technique exploits the difference in solubility of the reference standard and its impurities. Ideally, the results from phase-solubility analysis should correlate quite closely with the sum of impurities from chromatographic analysis of the reference standard.80,81

6 REFERENCE STANDARDS

135

B. Form Selection During the course of extensive characterization of an analytical reference standard (ARS), attention should be given to gaining an understanding of the material’s full physicochemical profile. The ARS should demonstrate excellent chemical stability under a wide range of storage conditions. This knowledge may be gained by challenging the hygroscopicity, and thermaland photostability of the proposed reference standard in a variety of closure/ container systems. This ensures that the appropriate salt, solvate, hydrate or polymorphic form of the active moiety as well as the closure/container system has been selected for the ARS. This up-front experimentation does not replace a rigorous and ongoing stability program of the ARS but ensures that surprises during development will be minimized. Ideally, it is preferred for the ARS to be of the same drug substance (DS) form as the material to be routinely identified and quantified. However, the form of the DS-ARS to be used for routine identification and quantification may be inherently unstable under the desired storage and testing conditions. The selection of an ARS in the same form as the DS is not a requirement but a preference because direct spectral comparisons cannot typically be applied for identification when using an alternate form of the active moiety as an ARS. Additionally, during comparative assay calculations a molecular weight correction is required when the DS and the ARS are of different forms. To illustrate the ARS form selection process, two pharmaceutical examples of ARS form selection are provided. Indinavir sulfate is the API for Crixivan2, a specific and potent inhibitor of the HIV-1 protease used in the treatment of AIDS. Indinavir sulfate is produced as a crystalline ethanolate sulfate salt. If the material is stored in double polyethylene liners within fiber containers or repeatedly exposed to ambient conditions changes occur in both crystallinity and solvation. Using XRPD, KF, and RP-HPLC, conversion of the crystalline ethanolate to amorphous material or to a hydrate crystal form has been detected and degradation is observed. However, the material is stable if stored in a tightly sealed container impermeable to ethanol/moisture transport under an inert nitrogen atmosphere at a controlled room temperature.82,83 These storage conditions are not practical for a routinely used ARS. Therefore, the free base monohydrate form of indinavir sulfate was evaluated and selected as the ARS. This form of the API was demonstrated to be extremely stable under ambient conditions needed for routine analysis. Levothyroxine is another example of an active pharmaceutical ingredient with relatively poor stability that is available as a USP analytical reference standard in the free acid form, while the drug is formulated as the sodium salt in most commercial preparations.84 Levothyroxine is prone to extensive photochemical decomposition85 that is thought to be exacerbated by the facile ionization of the phenolic hydroxyl group.86 Supply of the USP analytical reference standard as the free acid provides a more stable form through suppression of the ionization of the phenolic hydroxyl.

136

P.A. CULBERT AND B.D. JOHNSON

C. Assignment of Purity Once a thorough characterization of the reference standard has been completed, purity and assay values are assigned to the ARS. The purity, also known as potency or activity, is expressed on a theoretical and experimental basis. The theoretical purity (puritytheory) is the calculated amount (weight percent) of the API present in the DS based on the chemical structure. The puritytheory is used to calculate the assay value. Puritytheory % ¼ 100% 

MWAPI MWDS

Where MWAPI is the molecular weight of the API, MWDS is the molecular weight of the DS. The experimentally determined purity (purityexp) or activity is the amount of the API present in the material being tested. For chiral drug substances, the achiral purityexp is the amount of both enantiomers in the sample, while the chiral purityexp is the weight percent of the desired enantiomer. The purityexp is used to calculate the assay value of the batch and is used for DP manufacture and for the calculation of dosages used for toxicological testing. X X Purityexp ð%Þ ¼ 100%  ½%Related  ½%Water X X  ½%Residual Solvents  ½%Other P where [%related] is the total amount P of related impurities typically determined chromatographically and [%other] is the total amount of inorganic components. The amount of each of the components subtracted in the above equation is expressed on a weight percent basis. In early phases of drug development, it is typically assumed that all related impurities have similar chromatographic response factors in comparison to the main component. In later phases of drug development, the amount of impurities is usually determined on a weight percent basis, using working standards. As the chromatographic area percent determination of related impurities does not account for 100% of the material present (i.e., excludes moisture, solvents, and inorganics), the following equation should be used to more accurately reflect purityexp if the level of related impurities is significant (e.g. >1%) and they are not reported on a weight percent (w/w%) basis. P

½%Related 100 P P P 100%  ½%Water  ½%Residual Solvents  ½%Other  100

Purityexp ð%Þ ¼ 100% 

100% 

137

6 REFERENCE STANDARDS

The assay value is the amount of the active moiety found in the material being tested as a percent of the theoretical value. The assay value is often reported on an ‘‘anhydrous, solvent-free’’ basis and should be consistent with the chromatographic area percent value. "

# Purityexp Assayð%Þ ¼ 100%  Puritytheory   1 P P  100%  100%  ½%Water  %½Residual Solvents When assigning purity and assay values, some general considerations include: .

. .

.

The counter-ion should be expressed in terms of the un-ionized acid or base. A molecular weight correction should be applied (e.g., HCl equivalence). The purity and assay values are typically expressed with three significant figures. Residual solvent contents, water contents, residue on ignition contents, and process-related impurities below their method determination limit of quantitation (LOQ) or below 0.05%, whichever is the greater, are generally not included in the calculation. The LOQ should be 0.05% or less for residual solvents, water contents, and process-related impurity contents and 0.1% for residue on ignition contents. For drug substances defined as hydrates or solvates, an assay value consistent with the ‘‘anhydrous, solvent-free’’ basis can be calculated where the amount of water or solvents subtracted in the denominator is the difference between the measured and theoretical amounts.

In order to gain practical knowledge for these calculations, two examples are provided.

Case 1: DS Information: The DS is not a hydrate or solvate MW of compound X form Y ¼ 419.87 (an amine hydrochloride salt) MW of compound X ¼ 383.41 (free base) Sample characterization results:

Chloride Water Residual solvents

8.5 w/w% 0.3 w/w% < 0.05 w/w%

Residue on ignition (ash) Impurity A

0.1 w/w% 0.08 w/w%

Impurity B

0.3 w/w%

138

P.A. CULBERT AND B.D. JOHNSON

Calculations: Puritytheory % ¼ 100% 

383:41 ¼ 91:3% 419:87

HCl Equivalenceð%Þ ¼ 8:5% 

36:46 ¼ 8:7% 35:45

Purityexp ð%Þ ¼ 100%  ½0:3 þ 0:08%  ½0:3%  ½0:0%  ½8:5% þ 0:1% ¼ 90:7% Assay ‘‘anhydrous; solvent-free’’ value ð%Þ     90:7 1 ¼ 100%   100%  ¼ 99:6% 91:3 100%  ½0:3%  ½0:0%

Case 2: DS Information: The DS is a hydrate. The MW of compound Z form Q ¼ 219.67 (monohydrate, HCl salt) The MW of compound Z ¼ 165.19 (free base) Sample characterization results: Chloride

16.1 w/w%

Residue on ignition (ash)

0.3 w/w%

Water (Theoretical

8.4 w/w% 8.2 w/w%)

Enantiomer

1.0 w/w%

Residual solvents

0.5 w/w%

Impurity A

0.5 w/w%

Calculations: Puritytheory % ¼ 100% 

165:19 ¼ 75:2% 219:67

Purityexp ð%Þ ¼ 100%  ½1:0% þ 0:5%  ½8:4%  ½0:5%  ½16:1% þ 0:3% ¼ 73:2%

Assay ‘‘anhydrous; solvent-free’’ value ð%Þ     73:2 1  100%  ¼ 98:0% ¼ 100%  75:2 100%  ½0:2%  ½0:5%

6 REFERENCE STANDARDS

139

D. Bridging Standards The establishment of an ARS affords the opportunity to qualify new reference standards—both analytical and working—through bridging studies. A new candidate for an ARS can be analyzed against the previous ARS to give rise to a purity value expressed as a percentage by weight versus the qualified ARS. Care should be taken to reconcile the results of other tests against the results of the bridging study. Qualification by bridging establishes a common reference point for subsequent batches and guards against potential drift in results that may arise from the combination of systematic errors present in all experiments.

VI. SUMMARY The intent of this chapter is to provide practical guidance on pharmaceutical reference standards, which are a critical aspect in all phases of drug research, development, and commercialization. The chapter’s key directions include providing clarity to the role, qualification, and governance of reference standards used in the pharmaceutical environment. Reference standards serve as the basis of evaluation for both process and product performance and are the benchmarks for the assessment of drug potency for patient consumption. At the same time, the availability of reference standards and the degree to which they are characterized and governed is often dependent on the stage of the drug development process. Thus, an overriding theme is that at any phase of drug development, a reference standard should be assessed versus its intended use and a balance should be struck between resource commitments, scientific judgment, and regulatory requirements.

ACKNOWLEDGEMENTS The authors would like to acknowledge Dr. Steven Priebe, Dr. Brian Tobias, and Dr. Gerard Hokanson, Pfizer Global Research and Development, Ann Arbor, Michigan for their editorial review of this manuscript.

REFERENCES 1. Current Good Manufacturing Practice for Finished Pharmaceuticals, 21 CFR 211, April 1, 2001. 2. Guideline for Submitting Samples and Analytical Data for Methods Validation, Food and Drug Administration, Center for Drugs and Biologics, Office of Drug Research and Review (HFN-100), Rockville, MD, 1987. 3. Reviewer Guidance—Validation of Chromatographic Methods, Food and Drug Administration, Center for Drugs and Biologics, Office of Drug Research and Review, Rockville, MD, November 1994. 4. Terms and Definitions Used in Connection with Reference Materials 30, 2nd ed. 1992, International Organization for Standardization, Geneva, Switzerland, Ref. No. ISO Guide 30:1992 (E/F).

140

P.A. CULBERT AND B.D. JOHNSON

5. Calibration in Analytical Chemistry and Use of Certified Reference Materials 32 1st ed. 1997, International Organization for Standardization, Geneva, Switzerland, Ref. No. ISO Guide 32:1997 (E). 6. Uses of Certified Reference Materials—Guide 33, 2nd ed. 2000, International Organization for Standardization, Geneva, Switzerland, Ref. No. ISO Guide 33:2000 (E). 7. Certification of Reference Materials/General and Statistical Principles–Guide 35, 2nd ed. 1989, International Organization for Standardization, Geneva, Switzerland, Ref. No. ISO Guide 35:1989 (E). 8. Guidance for Industry—Analytical Procedures and Methods Validation Chemistry, Manufacturing, and Controls Documentation. Food and Drug Administration, Center for Drug Evaluation and Research, August 2000. 9. Chemistry Reviews of DMF’s for Drug Substances/Intermediates (DSI), Manual of Policies and Procedures. Food and Drug Administration, Center for Drug Evaluation and Research, 1998. 10. Jenks, P. J. and Stoeppler, M. The deplorable state of the description of the use of certified reference materials in the literature. Fresenius J. Anal. Chem. 370:164–169, 2001. 11. Pauwels, J. and Lamberty, A. CRMs for the 21st century: new demands and challenges. Fresenius J. Anal. Chem. 370:111–114, 2001. 12. Specifications: Test Procedures and Acceptance Criteria for New Drug Substances and New Drug Products: Chemical Substances. ICH Harmonised Tripartite Guideline (ICH Q6A), 1999. 13. Good Manufacturing Practice Guide for Active Pharmaceutical Ingredients. ICH Harmonised Tripartite Guideline (ICH Q7A), 2000. 14. Taylor, E. W., Qian, M. G. and Dollinger, G. D. Simultaneous on-line characterization of small organic molecules derived from combinatorial libraries for identity, quantity, and purity by reversed-phase HPLC with chemiluminescent nitrogen, UV, and mass spectrometric detection. Anal. Chem. 70:3339–3347, 1998. 15. Jared, Brenda. Personal Communication. Compound Management, Pfizer Global Research and Development, Ann Arbor, Michigan. 16. United States Pharmacopeia Rev. 25 and National Formulary. 20th ed. Constitution, Article 1, Section 3 Name and Objects. p. xxiv, 2002. 17. United States Pharmacopeia Rev. 25 and National Formulary. 20th ed. Continuously Updated Standards and Specifications, p. xlviii, 2002. 18. United States Pharmacopeia Rev. 25 and National Formulary. 20th ed. Preface/Preample Technical Facts on Reference Standards. p. xlvi, 2002. 19. Impurities in New Drug Substances, ICH Harmonised Tripartite Guideline, ICH Q3A(R), February, 2002. 20. Impurities in New Drug Products, ICH Draft Consensus Guideline, ICH Q3B, October, 1999. 21. Ihnat, M. A synopsis of different approaches to the certification of reference materials. Fresenius J. Anal. Chem. 360:308–311, 1998. 22. Crowther, J. B. In Handbook of Modern Pharmaceutical Analysis, Ahuja, S. and Scypinski, S., Eds., Academic Press, San Diego, pp. 415–443 2001. 23. Webster, G. K. and Bell, R. G. Practical approaches to qualifying laboratory standard reference materials. Pharmaceutical Formulation and Quality, January/February, pp. 39–43 1999. 24. Boenke, A. Pure substances and a new generation of CRMs for chemical industry. Fresenius J. Anal. Chem. 360:388–392, 1998. 25. United States Pharmacopoeia Rev. 25 and National Formulary Rev. 20. 2002 General Test h786i Particle Size Distribution Estimation by Analytical Sieving, p. 2044, 2002. 26. United States Pharmacopoeia Rev. 25 and National Formulary Rev. 20. 2002 General Test h231i Heavy Metals, p. 1923, 2002. 27. United States Pharmacopoeia Rev. 25 and National Formulary Rev. 20. 2002 General Test h846i Specific Surface area, p. 2072, 2002. 28. United States Pharmacopoeia Rev. 25 and National Formulary Rev. 20. 2002 General Test h616i Bulk Density and Tapped Density, p. 1981, 2002. 29. United States Pharmacopoeia Rev. 25 and National Formulary Rev. 20. 2002 General Test h791i pH, p. 2052, 2002.

6 REFERENCE STANDARDS

141

30. United States Pharmacopoeia Rev. 25 and National Formulary Rev. 20. 2002 General Test h61i, Microbial Limit Tests, p. 1873, 2002. 31. Handbook of Modern Pharmaceutical Analysis, Vol. 3, Ahuja, S. and Scypinski, S., Eds., Separation Science and Technology Series. Academic Press, NY, 2001. 32. Gilpin, R. K. and Pachia, L. A. Pharmaceuticals and related drugs. Anal. Chem. 73:2805– 2816, 2001. 33. United States Pharmacopoeia Rev. 25 and National Formulary Rev. 20. 2002 General Test h631i Color and Achromicity, p. 1995, 2002. 34. Byrn, S. R., Pfeiffer, R. R. and Stowell, J. G. In Solid-State Chemistry of Drugs, 2nd ed., SSCI, Inc., West Lafayette, Chapter 4, pp. 69–78 1999. 35. Ingle, J. D. Jr. and Crouch, S. R. Spectrochemical Analysis, Prentice-Hall, Inc., A Division of Simon & Schuster, Englewood Cliffs, NJ, 1988. 36. United States Pharmacopoeia Rev. 25 and National Formulary Rev. 20. 2002 General Test h851i Spectrophotometry and Light Scattering, p. 2075, 2002. 37. Socrates, G. Infrared Characteristic Group Frequencies–Tables and Charts, 2nd ed., John Wiley & Sons, NY, 1994 38. Byrn, S. R., Pfeiffer, R. R. and Stowell, J. G. In Solid-State Chemistry of Drugs, 2nd ed., SSCI, Inc., West Lafayette, Chapter 8, pp. 111–117, 1999. 39. United States Pharmacopoeia Rev. 25 and National Formulary Rev. 20. 2002 General Test h851i Spectrophotometry and Light Scattering, p. 2074, 2002. 40. Polymorphism in Pharmaceutical Solids, Brittain, H.G., Ed., Dekker, New York, 1999. 41. United States Pharmacopoeia Rev. 25 and National Formulary Rev. 20. 2002 General Test h941i X-Ray Diffraction, p. 2088. 42. Physical Characterization of Pharmaceutical Solids, Brittain, H.G., Ed., Dekker, New York, 1995. 43. Guilhaus, M., Mlynski, V. and Selby, D. Perfect timing: time-of-flight mass spectrometry. Rapid Communications in Mass Spectrometry, 11:951–962, 1997. 44. United States Pharmacopoeia Rev. 25 and National Formulary Rev. 20. 2002 General Test h736i Mass Spectrometry, p. 2029, 2002. 45. United States Pharmacopoeia Rev. 25 and National Formulary Rev. 20. 2002 General Test h761i Nuclear Magnetic Resonance, p. 2034. 46. Padden, B. E., Zell, M. T., Don, Z., Schroeder, S. A., Grant, D. J. W. and Munson, E. J. Comparison of solid-state 13C NMR spectroscopy and powder X-ray diffraction for analyzing mixtures of polymorphs of neotame. Anal. Chem., 7:3325–331, 1999. 47. United States Pharmacopoeia Rev. 25 and National Formulary Rev. 20. 2002 General Test h781i Optical Rotation, p. 2043. 48. MacLeod, S. K. Moisture determination using Karl Fischer titrations, Anal. Chem. 63:557–566, 1991. 49. Wieland, G. Water Determination by Karl Fischer Titration, Git Verlag GMBH, p. 17. 1987. 50. United States Pharmacopoeia Rev. 25 and National Formulary Rev. 20. 2002 General Test h921i Water Determination, p. 2085. 51. United States Pharmacopoeia Rev. 25 and National Formulary Rev. 20. 2002 General Test h281i Residue on Ignition. p. 1927. 52. Byrn, S. R., Pfeiffer, R. R. and Stowell, J. G. In Solid-State Chemistry of Drugs, 2nd ed., SSCI, Inc., West Lafayette, Chapter 5, pp. 81–82, 1999. 53. United States Pharmacopoeia Rev. 25 and National Formulary Rev. 20. 2002 General Test h891i Thermal Analysis, p. 2081. 54. Encyclopedia of Chromatography, Cazes, J., Ed., Marcel Dekker, Inc., NY, 2001. 55. Chromatographic Analysis of Pharmaceuticals, Adamovics, A., Ed., Chromatographic Science Series, vol. 49 & 74, John Marcel Dekker, Inc., NY, 1990/1997. 56. Snyder L.R. and Kirkland J.J. Introduction to Modern Liquid Chromatography, 2nd ed., John Wiley & Sons, Inc., NY, 1979. 57. Caude, M. and Jardy, A. In Chromatographic Science Series Volume 78: Handbook of HPLC, Katz, E., Eksteen, R. and Schoenmakers, P., Eds., Marcel Dekker, NY, pp. 325–363, 1998.

142

P.A. CULBERT AND B.D. JOHNSON

58. Snyder, L. R., Kirkland, J. J. and Glajch, J. L., Practical HPLC Method Development, 2nd ed., John Wiley and Sons, New York, pp. 240–242, 1997. 59. United States Pharmacopoeia Rev. 25 and National Formulary Rev. 20. 2002 General Test h621i Chromatography, p. 1982. 60. Kok, W. T. In Chromatographic Science Series Volume 78: Handbook of HPLC, Katz, E., Eksteen, R. and Schoenmakers, P., Eds., Marcel Dekker, New York, pp. 143–168, 1998. 61. Castledine, J. B. and Fell, A. F. Strategies for peak purity assessment in liquid chromatography. J. Pharm. Biomed. Anal., 11:1–13, 1993. 62. LaCourse, W. R. Column liquid chromatography: equipment and instrumentation, Anal. Chem., 72:37R-51R, 2000. 63. Small, H. Ion Chromatography, Plenum Press, NY, 1989 64. Capillary Electrophoresis Technology, Guzman, N. A., Ed. Marcel Dekker, NY, 1993. 65. Handbook of Capillary Electrophoresis, 2nd ed., Landers, J. P., Ed., CRC Press, Boca Raton, FL, 1997. 66. Supercritical Fluid Chromatography with Packed Columns: Techniques and Applications in Chromatographic Science Series, vol. 75, Anton, K. and Berger, C., Eds., Dekker, NY, 1998. 67. Packed Column SFC. Berger, T. A. and Smith, R. M., Eds. The Royal Society of Chemistry. 1995. 68. Modern Thin-Layer Chromatography, Grinberg, N., Ed., Chromatographic Science Series, Vol. 52, Dekker, NY, 1990. 69. Wilson, I. D. Thin layer chromatography: a neglected technique. Therapeutic Drug Monitoring, 18:484–492, 1996. 70. Gordon, A. J. and Ford, R. A. The Chemist’s Companion, John Wiley & Sons, New York, pp. 377–379, 1973. 71. CAMAG Application Note A-27.4, Determination of Vitamin B1 in Pharmaceutical Products. CAMAG Scientific Inc. Wilmington, NC, 1998. 72. Zivanovic, L., Agatonovic-Kustrin, S., Vasiljevic, M. and Nemcova, I. Comparison of highperformance and thin-layer chromatographic methods for the assay of lidocaine. J. Pharm. Biomed. Anal. 14:1229–1232, 1996. 73. Naidong, W., Hua, S., Roets, E. and Hoogmartens, J. Assay and purity control of minocycline by thin-layer chromatography using UV and fluorescence densitometry—a comparison with liquid chromatography. Pharm. Biomed. Anal. 13:905–910, 1995. 74. Chromatographic Chiral Separations, Zief, M. and Crane, L. J., Eds., Chromatographic Science Series, Vol. 40, Marcel Dekker, Inc., NY, 1988. 75. United States Pharmacopoeia Rev. 25 and National Formulary Rev. 20. 2002 General Test h467i Organic Volatile Impurities, p. 1943. 76. Huber, W. Titrations in Nonaqueous Solvents, Academic Press, Inc., NY, 1967. 77. Kolthoff, I. M. and Stenger, V. A. Volumetric Analysis, Vol. 1–3 Interscience Publishers, NY, 1942, 1946, 1947. 78. Laitinen, H. A. and Harris, W. E. Chemical Analysis—An Advanced Text and Reference, 2nd ed., McGraw-Hill Book Company, NY. 1975. 79. Siggia, S. and Hanna, J. G. Quantitative Organic Analysis via Functional Groups. 4th ed., John Wiley & Sons, Inc., NY, 1978. 80. William J. Mader. Phase Solubility Analysis, Critical Reviews in Analytical Chemistry, 2:193–215, 1970. 81. United States Pharmacopoeia Rev. 25 and National Formulary Rev. 20. 2002 General Test h1171i Phase Solubility Analysis, p. 2122. 82. Indinavir Sulfate, Analytical Profiles of Drug Substances and Excipients, Brittain, H., Ed., 26:319–357, Academic Press, 1999. 83. Lin, J. H., Ostovic, D. and Vacca, J. P. The integration of medicinal chemistry, drug metabolism and pharmaceutical research and development in drug discovery and development story of Crixivan, an HIV protease inhibitor, Pharm. Biotechnol. Integration of Pharmaceutical Discovery and Development, 11:233–255. 1998. 84. USP 25-NF 20, Levothyroxine Sodium–Official Monograph. 2002 The USP, Rockville Maryland, p. 1001, 2002.

6 REFERENCE STANDARDS

143

85. Kazemifard, A. G., Moore, D. E. and Aghazadeh, A. Identification and quantitation of sodium-thyroxine and its degradation products by LC using electrochemical and MS detection, J. Pharm. Biomed. Anal. 25:697–711, 2001. 86. Post, A. and Warren, R. J. Sodium Levothyroxine, In Analytical Profiles of Drug Substances, Vol. 5. Florey, K., Ed., Academic Press, 1976.

This Page Intentionally Left Blank

7 SAMPLE SELECTION FOR ANALYTICAL METHOD DEVELOPMENT HUGH J. CLARKE AND KENNETH J. NORRIS Pfizer, Inc.,Groton,CT 06340

I. INTRODUCTION II. COMPONENTS OF THE KEY PREDICTIVE SAMPLE SET (KPSS) A. Counter-ions III. STEREOISOMERS IV. MATRIX COMPONENTS V. PROCESS-RELATED IMPURITIES (PRI’s) VI. PURPOSEFUL DEGRADATION SAMPLES A. Sample Screening Recommendations for Purposeful Degradation Samples B. Interpretation of Kinetic Purposeful Degradation Data C. Chiral Compound Screening Recommendations for Purposeful Degradation Samples D. Elimination of Degradation Samples from KPSS VII. STABILITY SAMPLES VIII. PHASE-SOLUBILITYANALYSIS IX. SAMPLE SELECTION STRATEGIES A. Experimental Considerations B. Sample Selection Strategies X. SUMMARY GLOSSARY REFERENCES

I. INTRODUCTION Analytical method development for the quantitation of low level impurities present in pharmaceuticals can be thought of as a three step process. 1. Sample set selection for analytical method development. 2. Screening of chromatographic conditions and phases, typically using the linear-solvent-strength model of gradient elution.1 3. Optimization of the method to fine-tune parameters related to ruggedness and robustness. This can be accomplished using a factorial optimization approach.2–5

145

146

H. J. CLARKE AND K. J. NORRIS

FIGURE 1 Achiral method development process.

Figure 1 illustrates the primary method development process. Item 1, the identification and selection of the sample set for method development is the focus of this chapter. Before beginning a method development experiment, the first task of the chromatographer is to select the set of samples that accurately represents the purity and degradation profiles of the active pharmaceutical ingredient (API). Regulations set forth by the US Food and Drug Administration (FDA) and guidance documents published by the International Conference on Harmonisation necessitate that methods that measure the purity of a drug substance or drug product must be specific or selective for analytes.6–8 If a key degradant or impurity is omitted during method development, the resulting method may not resolve an important impurity. Therefore, the identification of the set of compounds (degradants, impurities, and matrix components), which will be referred to as the essential component set, is critical for effective method development. Each compound may exist in purified form, but it is likely that many will be one compound in a mixture or solution. The set of samples that encompasses all the elements of the essential component set is referred to as the key predictive sample set, or KPSS. In the analysis of pharmaceutical molecules, there are two distinct elements of the essential component set: One element is specific for drug substance and the other is specific for drug product. The drug substance peak set includes process-related substances; therefore, it is process-specific. If the synthetic process changes, the component set may need to be updated. The drug product component set contains the drug substance set. The inclusion of the drug substance peak set for drug product applications does not necessitate the separation of all peaks present, but rather drug substance impurities must be tracked for potential interference with the drug product components of interest (degradants, preservatives, etc.). The drug product peak set is formula-specific, so formulation changes can require an update to the drug

7 SAMPLE SELECTION FOR ANALYTICAL METHOD DEVELOPMENT

147

FIGURE 2 Key Predictive Sample Set (KPSS) concept.

product peak set. The components and samples, which minimally define the KPSS, are summarized in Figure 2 and Table 1. There may be other samples or stressed conditions in addition to those listed in Figure 2, which should be addressed on a case-by-case basis.

II. COMPONENTS OF THE KEY PREDICTIVE SAMPLE SET (KPSS) A. Counter-ions Counter-ions, usually small polar or ionic compounds, are routinely used to enhance the aqueous solubility and/or stability of the API. Because of their polarity, counter-ions are rarely resolved from the chromatographic solvent front in reversed-phase HPLC and have characteristically poor chromophores which makes detection difficult. The counter-ion can be omitted from the achiral method development sample set with minimal risk when this holds true. Analysis of counter-ions is normally performed using ion chromatography.9,10 This assay is separate from the reversed-phase assay performed to measure organic impurity levels. Note: Benzoate, besylate, and tosylate salts are examples of counter-ions that are sufficiently retained and should be included in the method development set.

III. STEREOISOMERS Isomers are compounds that have the same formula as the parent, but different molecular orientation. Because of potential differences in therapeutic

148

TABLE 1 Component Selection Rules Separation priority

What

Why

Isomers

Need to determine whether the achiral method is specific for diastereomers or geometric isomers. If the method is not specific, chiral method(s) must be employed to determine isomeric purity.

MUST

Matrix components

For drug product applications, sample matrix is an important component of the KPSS. Typically, a representative placebo, placebo blend, or excipient ‘‘soup’’ is included in the KPSS for each formulation.

MUST

Antioxidants, flavours, and preservatives are commonly monitored for during stability. Isolated samples of these compounds are recommended for inclusion in the method development sample set, where applicable.

MUST

 Antioxidants  Coatings, dyes and inks  Excipients (‘‘soups’’)  Flavors Process-related impurities (PRI)

Any PRI > 0.05% in the finished goods (known or unknown) must be included.

 Regulatory raw materials

1. PRI controlled at raw materials stage < 0.1% or > 0.1%

N/A/LOW WANT

 Intermediates

2. PRI controlled at the in-process stage

WANT

 Known and unknown synthetic by-products

3. PRI controlled at finished goods

MUST

 Mother liquor components

The mother liquor of the last bond-forming step and first recrystallization are recommended for inclusion in the KPSS. Analysis of both these samples can reduce the KPSS versus that based on a single mother liquor.

H. J. CLARKE AND K. J. NORRIS

 Preservatives

LOW WANT

2. Found in bond-forming liquor and first recrystallization liquor at < 10% of total impurities

WANT

3. Found in bond-forming liquor and first recrystallization liquor at 10% of total impurities

MUST

4. Have purging data to show it is controlled in the recrystallization scheme.

LOW WANT

 Impurity grid components

The impurity grid acts as a hypothetical model to define what could potentially form during the drug substance synthesis. Since any material from this grid which actually forms will be a PRI or in the mother liquors, these are the least important component of the KPSS. If any of the impurity grid substances are included in the KPSS, they may need to be synthesized.

LOW WANT

Degradants Acid, base, light, oxidative, thermal, thermal humidity

Purposeful degradation experiments determine major degradation pathways and flags marker compounds for those pathways.

Stability sample components

1. Meets criteria for KPSS, i.e., present at a level of > 10% of total degradation.

MUST

2. Meets criteria for KPSS, not seen in real stability (6 months @ 40 C/75% RH or 2 years 30 C/60% RH)

LOW WANT

Stability samples act as a reality check for purposeful degradation and they can be used to determine if a kinetically favored secondary degradant is more likely in actual stability. 1. Seen in drug substance stability samples at > 0.05%, but not in purposeful degradation

MUST

2. Seen in drug product stability samples at > 0.1%, but not in purposeful degradation

MUST

7 SAMPLE SELECTION FOR ANALYTICAL METHOD DEVELOPMENT

Evaluation of liquors from two different lots is also desirable. 1. Found in bond-forming liquor but not in first recrystallization liquor

Because of availability and time constraints, some of the peaks or samples listed may not be included in the method development sample set. However, during validation these peaks should be considered for the specificity assessments if available.

149

150

H. J. CLARKE AND K. J. NORRIS

index and safety among stereoisomers, the FDA requires the evaluation of the API isomeric purity profile. Applications for enantiomeric and racemic drug substances should include a stereochemically specific identity test and/or a stereochemically selective assay method. The choice of the controls should be based upon the substance’s method of manufacture and stability characteristics.11

Stereoisomers can be classified into two types: enantiomers and diastereomers. Enantiomers (mirror images) have identical physical and chemical properties and therefore are not separated on the conventional reversed-phase stationary phases. Their separation will not be discussed. Diastereomers are isomers which are not mirror images of the parent. They have slightly different physical and chemical properties and can often be separated on conventional stationary phases. There are two classes of diastereomers: optically active isomers when the API has two or more stereocenters and non-optically active geometric isomers, such as cis–trans, syn–anti, etc. Stereoisomers of chiral molecules must be included in the peak set. Either isolated diastereomers or racemic mixtures are preferred for the KPSS. If no isolated samples are available, a sample that contains an enriched level of the diastereomers may be used, but peak tracking may be hampered as diastereomers are challenging to separate from the API. These molecules are subject to elution order changes.

IV. MATRIX COMPONENTS In pharmaceutical analysis, the majority of matrix components (excipients) are present to improve manufacturing characteristics like bulk powder flow, tablet hardness, or to enhance tablet disintegration.12 Other excipients are present in a formulation to preserve the active pharmaceutical ingredient13 (antioxidant, for example) or to act as an antimicrobial agent14 (methylparaben, for example). Still other excipients may be present to increase the palatability of the formulation15 (flavors, for example). It may be necessary to include individual excipients and/or mixtures as part of the KPSS when analyte/excipient interferences occur. Preservatives and antioxidants are normally assayed for during release testing and stability evaluation of a formulation. Therefore, isolated samples of these excipients should be included in the KPSS in addition to the representative placebo or excipient ‘‘soup’’ samples to facilitate peak tracking. Inclusion of these samples is more critical for high-potency compounds, where the ratio of excipient to drug can be very large.

V. PROCESS-RELATED IMPURITIES (PRIs) A stability-specific-drug-substance-purity method must be able to separate any process-related impurity that is not purged or controlled prior to the final step and all of the API degradation products. Ideally, one would

7 SAMPLE SELECTION FOR ANALYTICAL METHOD DEVELOPMENT

151

like to develop a method that is capable of monitoring the quality of the API throughout the entire synthetic process (from raw materials to finished goods) and is also stability-specific (separates the API degradants). This method could be used to assess the quality and stability (where applicable) for the raw materials, isolated intermediates (in-process control testing), and finished goods. Because of polarity extremes of the various analytes and the sheer number of compounds that must be separated, the ideal method is rarely attainable and trade-offs must be made. For the finished product, impurities that are purged or controlled early in the synthetic process do not necessarily need to be included in the method development sample set (see Table 1). If a viable separation cannot be obtained for compounds that are purged or controlled, they may be dropped from the essential peak set. When a compound is dropped from the peak set, the scientific rationale should be documented. Table 1 provides general guidance on where trade-offs can be made by rating the compounds as a MUST for compounds that absolutely need to be separated, a WANT for compounds that may be present in finished goods, and a LOW WANT for compounds that only show up in early stages of the synthetic process and are controlled. Any process-related impurity, known or unknown, that has been observed in the finished goods at a level 0.05% must be included in the peak set. Experimental campaign bulk or in-process control samples (e.g., prior to recrystallization) are excellent sources of process-related impurities and are a vital component of the KPSS. Isolated fractions collected from a preparative or semi-preparative liquid chromatography (LC) system are also excellent KPSS samples. Sometimes small-scale synthesis of the impurity or degradant is possible and less time-consuming than the isolation techniques. During early stages of drug candidate development, raw materials and isolated process intermediates may be included in the finished goods essential peak set until the appropriate analytical and process-control strategies are implemented. Once the appropriate controls are in place as the candidate approaches commercialization, it may be possible to eliminate some of these components from the component set. An impurity grid can also be used to predict likely synthetic by-products. The grid is an impurity-tracking tool used to formulate the overall impurity control strategy for the API. Each synthetic step is tabulated with reactants and products. The synthetic product of each molecule is charted. If a material is not purged or not completely reacted, the impurity grid can predict potential synthetic by-products. By-products predicted by the grid may be included in the essential component set depending upon availability and at the discretion of the project analyst. Mother liquors from the key bond-forming step and last recrystallization can be used as surrogate challenge samples for the KPSS in the absence of purging data. The differences in impurity profiles of these two liquors can be used to identify potential synthetic by-products that do not purge and therefore must be controlled at finished goods. Screen the mother liquor samples using the current purity method and an orthogonal screen to reduce the risk of having missed any potential impurities before eliminating them

152

H. J. CLARKE AND K. J. NORRIS

FIGURE 3 Drug substance peak tracking example.

from the KPSS. The scientific rationale for this decision should be documented (see Figure 3). It is common upon scale-up that recrystallization steps will be eliminated from the synthetic route; therefore, it is prudent to include the mother liquor from the bond-forming step in the method development sample set even if there is evidence that the major synthetic by-products are being purged in the recrystallization steps.

VI. PURPOSEFUL DEGRADATION SAMPLES Representative challenged samples (typically 10–20% degradation) are a vital component of method development sample set.16 The generation of these key predictive degradation samples is beyond the scope of this text; however, general guidance on the screening and handling of purposefully degraded samples for use in the method development process is provided. For method development purposes, a reduction in the degradant peak set may be justified if the degradant(s) are not observed in the long term and accelerated stability samples (see Table 1). However, this data is usually not available until the time of regulatory filing.

A. Sample Screening Recommendations for Purposeful Degradation Samples The purposefully degraded samples should be screened using the current purity methodology, or if a purity method is not available, a gradient of 5–95% organic in 60 minutes could be used. At a minimum, broad-spectrum (200–400 nm) photodiode array UV scans are recommended during the screening of these samples to minimize the risk of missing a degradant due to poor UV absorption. Peak purity assessments of API can then be made to identify co-elutions. Additionally, if a working standard or characterized lot of the drug is available, it can be used as an external standard throughout the screening runs to assess material balance for the challenged samples. This screening process is diagramed in Figure 4. Material balance assessments are critical in confirming that degradants have not been missed and can identify drug/degradant co-elutions. A low bias in the material balance can be due to a number of factors. When performing solution degradation studies, low recovery can be due to poor solubility of

153

7 SAMPLE SELECTION FOR ANALYTICAL METHOD DEVELOPMENT

(KPSS)

–

FIGURE 4 KPSS screening flowchart.

the drug or degradants. When analyzing purposeful degradation samples using UV detection, a low bias may be seen because of alteration of molecular chromophores in one or more degradation products. A low bias may be indicative of retention issue: highly retained or completely unretained (in solvent front) molecules not detected in the analysis. When a low material balance is observed, a thin-layer chromatography (TLC) screen is recommended to assure that no degradants have been missed in the highperformance liquid chromatography (HPLC) analysis.

154

H. J. CLARKE AND K. J. NORRIS

A high bias in the material balance is generally indicative of formation of a degradant with higher detector response relative to the API. In the case of UV detection, this would be a degradant with higher molar absorptivity than the API. Photodiode array (PDA) detectors can be used to examine the UV spectra of the parent peak and degradation products. Any degradants with UV spectra significantly different from the parent will likely have different molar absorptivity, and hence, different UV response factors. High recoveries for the API itself, assay values > 103%, may be indicative of degradant/API co-elution. When this occurs, it is prudent to perform some additional orthogonal screening to separate the degradant from the API. In the case of one development candidate, a 120% high bias was observed in the oxidative degradation samples. The co-elution was detected from a PDA peak purity assessment. An orthogonal screen was performed on the questionable samples. The primary degradant was separated from the API on a phenylhexyl column run in the orthogonal screen. This had a significant impact on the method development strategy for the candidate.

B. Interpretation of Kinetic Purposeful Degradation Data Once the sample screening has been completed for each study time point, the data is generally tabulated by peak identification [compound identification or relative retention time (RRT)] and area%. An easy way to compile these data is to start with the last time point (most degraded sample) and assemble the peak list working backwards to the initial time point adding to the peak list as necessary. Once the peak list (identification) is complete, compile the area% values for each time point. This kinetic data is useful in identifying secondary degradants that would typically need not be included in the method development sample set. Table 2 shows an example for a degradation study of a drug candidate. This example shows that for this candidate, secondary degradation (double dehydration) was occurring at 24 h and beyond. The graph illustrates that the primary degradants at RRT 1.41 and 1.46 have reached steady state at 24 h and RRT 1.17 and 1.30 have firstorder growth after the 24-h timeframe.

C. Chiral Compound Screening Recommendations for Purposeful Degradation Samples For compounds with one or more stereocenters, it is prudent to screen the key samples (10–20% degradation timepoint) with the current chiral purity method to determine if the degradation pathway is stereospecific. From the achiral method development perspective, stereospecific degradation pathways will not affect the outcome of the method development process, but this information can affect the impurity control strategy for the compound.

D. Elimination of Degradation Samples from KPSS It is recommended to put the key degradation samples through an orthogonal reversed-phase and/or TLC screen to ensure that all the

155

7 SAMPLE SELECTION FOR ANALYTICAL METHOD DEVELOPMENT

TABLE 2 Development Compound Degradation Kinetic Study Results @ 70 C

Sample Initial–Drug in water

Pull times

Parent compound

Rr 51.41

Rr 51.46

Rr 51.17

Rr 51.30

(h)

%area

(wt/wt)%

(wt/wt)%

(wt/wt)%

(wt/wt)%

100.0

ND

ND

ND

ND

70 C T0 H2SO4

0.0

100.0

ND

ND

ND

ND

70 C T1hr H2SO4

1.0

99.9

ND

ND

ND

ND

70 C T3hr H2SO4

3.0

98.4

0.14

0.05

ND

ND

70 C T7hr H2SO4

7.0

93.8

0.39

0.14

ND

ND

70 C T24hr H2SO4

24.7

83.0

0.76

0.28

0.45

0.49

70 C T96hr H2SO4

100.0

83.0

0.62

0.26

1.52

2.67

70 C T168hr H2SO4

168.8

77.3

0.60

0.28

2.37

4.55



NA

degradants have been accounted for (see Figure 4 for flowchart). After screening if no unique peaks are present in the key challenge samples or the peaks present in those samples are covered by another sample in the KPSS, the challenge sample may be safely dropped from the method development sample set. The scientific rationale for this decision should be documented (see Figure 6).

VII. STABILITY SAMPLES Stability samples augment the peak set since they are more indicative of the actual degradation profile under typical storage conditions than purposefully degraded samples. Any peak(s) which are observed in the accelerated challenge (40 C/75% RH 6 month or 50 C/20% RH 3 month) or long-term storage samples must be included in the essential peak set. It is critical to demonstrate specificity for these peaks (i.e., any peak > 0.05% and 0.1% of API for drug substance and product, respectively).

156

H. J. CLARKE AND K. J. NORRIS

4-Fluorothiophenol

FIGURE 5 Example elimination rationale form.

When long-term storage stability profiles become available, these profiles can be used to justify a reduction in the degradant peak set. The scientific rationale for this decision should be documented.

VIII. PHASE-SOLUBILITY ANALYSIS Phase-solubility analysis17 (sometimes referred to as phase equilibrium purification) is the quantitative determination of the purity of a substance through the application of precise solubility measurements. At a given temperature, a definite amount of a pure substance is soluble in a definite quantity of solvent. The resulting solution is saturated with respect to the particular substance, but the solution remains unsaturated with respect to other substances even though such substances may be closely related in chemical structure and physical properties to the particular substance being tested. There are examples of the use of this technique in HPLC methods development18 and in the characterization of reference standards,19 but the

7 SAMPLE SELECTION FOR ANALYTICAL METHOD DEVELOPMENT

157

FIGURE 6 HCl salt impurity enhancement example.

technique can also be used to enrich saturated solutions with low-level impurities. The ability to track the key peaks from column-to-column and conditionto-condition is generally the most challenging aspect of the method development process. In particular, tracking impurities or degradants that are present at levels < 0.2% of the API can be difficult, and peak misassignments are common. Photodiode array UV detection and mass spectrometry for tracking of these low level peaks may be of limited utility because of poor sensitivity, baseline noise, and ionization inefficiencies. For these cases, an impurity/degradant enhancement technique can be used to increase the levels present with respect to the API and thereby facilitate the peak tracking process. Of the impurity/degradant enhancement techniques available, the phasesolubility analysis technique requires minimal sample handling from the

158

H. J. CLARKE AND K. J. NORRIS

analyst and the observed enhancement factors can range from 5 to 10000 depending upon the solubilities of the drug and impurities. A saturated solution of drug substance (typically 1 gram in a few milliliters of solvent) is created, and the impurities present are concentrated in the solvent (supernatant) since they typically are not present at their saturation level. The magnitude of this concentration enhancement in general depends on the solubility of the API. Poor API solubility can result in significant enhancement of the drug substance impurities. An example from a development project is shown in Figures 6 and 7. Saturated solutions of two salt forms of a development candidate, hydrochloride salt and citrate salt, were created in acetonitrile. The hydrochloride salt (Figure 6) was readily soluble (> 10 mg/mL) in acetonitrile, whereas the citrate salt (Figure 7) was only slightly soluble (0.04 mg/mL). The impurity enhancement factors were markedly different, 5 for the hydrochloride salt and 1000þ for the citrate salt. Even an increase in the impurity-to-drug ratio of 5 can make a significant difference in terms of

FIGURE 7 Citrate salt impurity enhancement example.

7 SAMPLE SELECTION FOR ANALYTICAL METHOD DEVELOPMENT

159

peak tracking. Impurities that are present at levels approaching 1.0% relative to the API are fairly easy to track from column to column and condition to condition. For cases where an impurity is present at low levels (< 0.5%), the phase equilibrium purification technique is an easy way to concentrate the impurity to a readily trackable level. This technique can be used to concentrate multiple trace impurities in non-recrystallized bulk lots without resorting to the time-intensive semi-preparative fraction collection and/or isolation methods. When utilizing this technique, both the drug substance reference material and the enhanced supernatant from the saturated solution would have to be included in the method development sample set to insure that only the relevant peaks are being tracked during the screening process.

IX. SAMPLE SELECTION STRATEGIES The selection of the set of samples that best covers the impurity and degradation profiles of the API is critical to successful method development. For a given API there are several ways to design the method development sample set some of which are more advantageous, but incur more risk in terms of the ability to track peaks from column to column and condition to condition. This section will address method development sample set design strategies, experimental concerns, and trade-offs.

A. Experimental Considerations 1. Solution and Degradant/Impurity Stability The method development sample solutions need to be stable over the course of the screening runs, which are typically 24 hours or more depending on the number of samples in the method development set. Some instability can be tolerated as long as it does not interfere with the ability to track the relevant peaks. The chance for degradation and impurity interactions increases with solution complexity. Isolated or simple mixtures are less problematic from a solution stability standpoint, as degradation can readily be identified and generally does not interfere with the ability to track the few relevant peaks that are present. If little is known about solution stability or there is a reason to suspect that there may be an issue based on the chemistry, the use of a refrigerated autosampler and amber HPLC vials is recommended. A quick evaluation of solution stability over the course of 24 hours or at the end of each screening run can also be performed to identify stability issues and to insure that peak tracking is not compromised. Sample solutions should be stored at 2–8 C until solution stability has been established and for longterm storage 70 C should be considered. The sample solvent can be the source of instability and rapid degradation of the relevant impurity can occur. In the case of one development compound, an imine process-related impurity was a primary degradation product. The imine was stable in acetonitrile, but would readily hydrolyze in aqueous media. In fact, on-column degradation was observed when screening

160

H. J. CLARKE AND K. J. NORRIS

the imine under acidic conditions. It is important to consider compound/ solvent interactions whenever possible. 2. Solvent Selection The selection of an appropriate sample diluent can have a significant impact on the ability to track peaks during the screening process. Since extreme gradient ranges (5–95% organic) may be employed during the screening process, solvent/mobile phase matching as a way to reduce system artifacts is not possible. Because of differences in solvent purity and mixing properties, the gradient artifacts will vary from diluent to diluent. Therefore, it is convenient to minimize the number of diluents used to generate the sample solutions, as respective blanks for each diluent should be included in the screening exercise. Solvent strength effects such as splitting of poorly retained species (small polar molecules) can occur if a strong solvent is used as the diluent.20 A solvent strength of 65:35 aqueous/organic has been found to work well for most applications. There are cases where a stronger solvent may be necessary because of poor degradant or impurity solubility. Generally, these compounds are retained longer and the diluent/mobile phase mismatch is not as severe, so peak splitting is usually not observed. Additionally, if methanol is used as a component of the diluent, a variable intensity displacement peak is commonly observed at screening wavelengths below 220 nm.21,22 If unexpected peak splitting is observed during the screening process, the diluent solvent strength for the sample in question should be investigated. 3. Handling of Fractions and Phase Equilibrium Supernatants Impurity enhancement techniques such as fraction collection and phase equilibrium purification can be used to provide enriched samples for use in the method development process.23 When using the fraction collection approach, one or more cuts (fractions) of the chromatographic separation of a bulk lot or mother liquor are isolated. The excess solvent in these fractions is then evaporated to achieve the desired concentration enhancement. These fractions typically contain extraneous peaks because of the presence of salts in the mobile phase or sample degradation during the concentration step. The salts can be removed by extraction and/or a LC cleanup step. To insure that these extraneous peaks/artifacts are not identified as key peaks for separation, the original bulk lot or mother liquor should be included in the method development sample set. The same holds true for phase-equilibriumpurification supernatants.

B. Sample Selection Strategies For a given API, there can be numerous ways to mix and match the available samples to construct the method development sample set. During early stages of development, only a few isolated intermediates and purposeful degradation samples may be available, making sample set selection relatively easy. Later in development, more isolated samples of the major degradants

7 SAMPLE SELECTION FOR ANALYTICAL METHOD DEVELOPMENT

161

Individual Components

Cocktails or Complex Samples FIGURE 8 Sample selection trade-offs.

and process-related impurities will be available. The most difficult task during chromatographic method development is accurately tracking the peaks from various columns or conditions.24–26 The use of isolated samples in the method development sample set reduces both the peak-tracking burden and the risk of peak misassignment, while increasing the run time of the screening process. As long as retention time drift is not occurring during method development, the benefit of being able to unambiguously attribute the component outweighs the extra run time, since most screens are automated (See Figure 8). When the sample set is small, the use of isolated samples is strongly recommended, as peak attribution is generally the rate-limiting step in the method development screening process. When the sample set becomes large (15þ samples), the risk for retention time drifts to occur during the screening process increases and cocktailing should be considered to reduce the number of samples. Cocktails are mixtures of KPSS samples that combine two or more compounds to be separated into a single sample. This can save significant time in the method development effort since the amount of injections necessary are reduced by the number of combinations. However, using complex cocktail mixtures can complicate the assignment of peak retention times. The following are guidelines for employing the cocktail technique: .

. .

The preparation of cocktail samples should be documented in a notebook, and that notebook ID should be used to name the sample for injection, along with a convenient naming convention. For example, ‘‘NB-12345-001 PD UV/PD thermal’’ could be used to define a mixture of two purposeful degradation samples into a cocktail sample. Stereoisomers should be kept separate from other samples, as they are easy to misassign. When possible, spike samples at different levels relative to peak response. Differences in peak areas of 30% should allow for

162

H. J. CLARKE AND K. J. NORRIS

.

.

discrimination between closely eluting samples. There are benefits to adding a component to more than one cocktail mixture, or overlapping. This allows you to control-chart the retention times and ensure that the HPLC system was well-equilibrated during the run. It also provides an assignable peak from one sample to another to act as a reference for polarity assessments of unknowns in the samples. The total number of tracked compounds in a cocktail sample should be limited by your confidence in correctly assigning peak IDs.

Drug substance lots with different impurity profiles can be combined for peak-tracking purposes, provided the impurity profiles are relatively simple and that the abundance of each impurity in the final solution is sufficient to obtain accurate photodiode array UV data.

X. SUMMARY Accurate quantification of low level impurities is essential in defining the quality of pharmaceutical products. To that end, a great deal of time is devoted to developing methods to meet these needs. The first step of this development project must be to define and gather a set of samples containing any potential and actual impurities that need to be assessed by the purity method. With this set in hand, subsequent development experiments can assure that a method or methods can accurately and completely determine the purity of a pharmaceutical product. This chapter provides detailed guidance in selecting the set of samples that contain the compounds of interest that must be quantified at low levels for a pharmaceutical product. A list of potential components and their sources is provided. Guidance is given on sample-screening techniques and when to eliminate samples that are redundant or unnecessary. Finally, techniques are outlined to enrich and combine samples in order to minimize the sample set.

GLOSSARY API Active pharmaceutical ingredient. This is the therapeutic agent in the formulation. cocktail A solution comprised of two or more samples. The samples are mixed to increase the number of peaks tracked per injection. Cocktails are usually prepared in order to shorten method development timelines. impurity grid A table of the theoretical impurities that could possibly form during drug substance synthesis. For example, the table would show what would become of a raw material impurity after each step in the synthesis. matrix components All elements of the finished goods drug product including antioxidants, coatings, dyes, excipients, flavors, inks, and preservatives. method development peak set All compounds necessary in a method development project to show that a method meets its specific business need. For a finished product purity assay, this could be all components of the

7 SAMPLE SELECTION FOR ANALYTICAL METHOD DEVELOPMENT

163

essential component set. For an in-process purity determination, this could be a subset of the full essential component set. mother liquor The supernatant solution that is left over after recrystallization of bulk drug substance. Typically, mother liquors are excellent sources for impurities (synthetic by-products) that may be found in trace amounts in the crystallized bulk drug substance. orthogonal RP screen A series of LC experiments utilizing orthogonal reversed-phase columns and mobile phase conditions. The goal of the experiment is to resolve co-eluting peaks that may not be detected on a single chromatographic or mobile phase. PRI process-related impurity. A synthetic by-product of the manufacture of bulk drug substance. PRIs may be related to the API or to raw materials or processing agents used in manufacture. purposeful degradation Purposeful degradation studies of the drug substance or drug product include appropriate solution and solid-state stress conditions (e.g., acid/base hydrolysis, heat, humidity, oxidation, and light exposure, in accordance with ICH guidelines). Specified stress conditions should result in approximately 10–20% degradation of the drug substance or represent a reasonable maximum condition achievable for the drug substance. The specific conditions (intensity and length of time) will depend on the chemical characteristics of the drug substance. soup Drug product blend containing the proposed excipients and API at the maximum excipient/API range. This represents the worst-case example for investigation of drug product stability. stability-indicating methodology A stability-indicating method (generally, potency) must be able to separate the analyte (typically, the API) from the degradants, extractables, impurities, and matrix components. stability-specific methodology A stability-specific-drug-substance-purity method must be able to separate any process-related impurity that is not purged or controlled prior to the final step and all of the API degradation products. A stability-specific-drug-product-purity method must be able to separate the API and its degradants from the matrix components, processrelated impurities, and extractables.

ACKNOWLEDGMENTS The authors thank Cheryl Bye for her contributions to this chapter.

REFERENCES 1. Snyder, L. R. and Dolan, J. W. The linear-solvent-strength model of gradient elution. Adv. Chromatogr. N.Y. 38:115–187, 1998. 2. McGraw, Joel D., Leonard, Jason A. and Madrak, M. K. Design of experimental approach to optimization of an HPLC separation. Abstracts of Papers, 222nd ACS National Meeting, Chicago, IL, August 26–30, 2001. 3. Mahesan, B. and Lai, W. Optimization of selected chromatographic responses using a designed experiment at the fine-tuning stage in reversed-phase high-performance liquid chromatographic method development. Drug Dev. Ind. Pharm. 27(6):585–590, 2001.

164

H. J. CLARKE AND K. J. NORRIS

4. Harang, V. and Westerlund, D., Optimization of an HPLC method for the separation of erythromycin and related compounds using factorial design. Chromatographia 50(9/10): 525–531, 1999. 5. Osborne, L. M. and Miyakawa, T. W. Use of experimental design in the optimization of HPLC methodology for the separation of stereoisomers. J. Liq. Chromatogr. Relat. Technol. 20(4):501–509, 1997. 6. FDA CDER Guidance: Analytical procedures and methods validation chemistry, manufacturing, and controls documentation. http://www.fda.gov/cder/guidance/2396dft.htm 7. ICH Q2A: Text on validations of analytical procedures, federal register, Vol. 60, March 1, 1995, p. 11260. http://www.ifpma.org/ich5q.html#Analytical 8. ICH Q2B: Methodology: federal register, Vol. 62, No. 96, May 19, 1997, p. 27463–27467. http://www.ifpma.org/ich5q.html#Analytical 9. Douglas, T. Gjerde, James, Fritz. S. Ion chromatography, 2nd ed. Heidelberg; NY: A. Huthig, xi, p. 283 1987. 10. Snyder, L. R., Glajch, J. L. and Kirkland, J. J. Practical HPLC Method Development, NY, Wiley, p. 341, 1988. 11. FDA’s Policy Statement for the Development of New Stereoisomeric Drugs. Publication Date: 5/1/92. http://www.fda.gov/cder/guidance/stereo.htm 12. Ainley, W. and Weller, P. J. Handbook of Pharmaceutical Excipients. American Pharmaceutical Association, 1994. 13. Papas, A. M. Antioxidant Status, Diet, Nutrition, and Health. CRC Press, Boca Raton, FL, 1999. 14. S.F. Bloomfield [et al.]. Microbial Quality Assurance in Pharmaceuticals, Cosmetics, and Toiletries. Halsted Press, NY, 1988. 15. Smolinske, Susan C. Handbook of Food, Drug, and Cosmetic Excipients, CRC Press, Boca Raton, FL, 1992. 16. Alsante, K. M., Friedmann, R. C., Hatajik, T. D., Lohr, L. L., Sharp, T. R., Snyder, K. D. and Szczesny, E. J. Degradation and impurity analysis for pharmaceutical drug candidates, In Handbook of Modern Pharmaceutical Analysis, pp. 85–172. Ahuja, S. and Scypinski, S., Eds., Academic Press, San Diego, CA, 2001. 17. USP 24 h1171i Phase-solubility analysis, The United States Pharmacopeial Convention, Inc 2001. 18. Rapado-Martinez, I., Garcia-Alvarez-Coque, M. C. and Villanueva-Camanas, R. M. Liquid chromatographic procedure for the evaluation of -blockers in pharmaceuticals using hybrid micellar mobile phases. J. Chromatogr. A 765(2):221–231, 1997. 19. Grdinic, V., Jaksevac-Miksa, M., Bezjak, A., Radaic, A. and Briski, D. Importance of factors for ruggedness test in phase solubility analysis. Eur. J. Pharm. Sci. 2(4):293–6, 1994. 20. Wilson, T. D. Sample solvent effects in an apparent chiral high-performance liquid chromatographic separation on -cyclodextrin. Pharm. Sci. Dep., Sterling-Winthrop Res. Inst., Rensselaer, NY, J. Chromatogr. 448(1):31–39 1988. 21. McCormick, R. M. and Karger, B. L. Role of organic modifier sorption on retention phenomena in reversed-phase liquid chromatography. Adv. Chromatogr. (Houston) 15:259– 73, 1980. 22. Dekany, I. and Nagy, L. G. Immersional wetting and adsorption displacement on hydrophilic/hydrophobic surfaces. J. Colloid Interface Sci. 147(1):119–28, 1991. 23. Blanchard, A. J., Alsante, K. M., Nickerson, B., Snyder and Kimberly, D. Extraction of low level impurities from tablets using accelerated solvent extraction. Abstracts of Papers, 222nd ACS National Meeting, Chicago, IL, August 26–30, 2001. 24. Strasters, J. K., Billiet, H. A. H., De Galan, L. and Vandeginste, B. G. M. Strategy for peak tracking in liquid chromatography on the basis of a multivariate analysis of spectral data. J. Chromatogr. 499:499–522, 1990. 25. Lankmayr, E. P., Wegscheider, W., Daniel-Ivad, J., Kolossvary, I., Csonka, G. and Otto, M. Recent advances in fuzzy peak tracking in high-performance liquid chromatography. J. Chromatogr. 485:557–67, 1989. 26. Dolan, J. W. Peak tracking. LC-GC 8(6) 1990.

8 SAMPLE PREPARATION METHODS FOR THE ANALYSIS OF PHARMACEUTICAL MATERIALS DAVID T. ROSSI AND KENNETH G. MILLER Pfizer Global Research and Development, Ann Arbor, MI 48105

I. INTRODUCTION II. SOLID-PHASE EXTRACTION (SPE) A. Suppliers B. Sorbents C. Cartridges D. Disks E. Extraction Process F. Nonconventional Techniques III. LIQUID ^LIQUID EXTRACTION (LLE) A. Theoretical Considerations in LLE B. Countercurrent LLE C. Liquid^Liquid Extraction Cartridges D. Emulsions E. Solvent Selection in LLE F. Drying Agents IV. SUPERCRITICAL FLUID EXTRACTION (SFE) A. Fundamental Properties of Supercritical Fluids B. Supercritical Fluid Extraction Instrumentation C. Static/Dynamic SFE D. Pumps Used for SFE E. Extraction Cells F. Modifiers G. Restrictors H. Collection Devices I. Pharmaceutical Applications of Supercritical Fluid Extraction V. ACCELERATED SOLVENT EXTRACTION (ASE) A. Sample Pretreatment B. Solvent Selection in ASE C. The Effects of Temperature in ASE

165

166

D.T. ROSSI AND K.G. MILLER

D. The Effects of Pressure E. CycleTime F. Carry-over VI. CENTRIFUGATION A. Centrifugation for Small Molecule Sample Preparation VII. FILTRATION A. Membrane Selection B. Housing Selection C. Pore Size D. Filtration Device Size VIII. SUMMARY REFERENCES

I. INTRODUCTION Many samples in the pharmaceutical industry cannot be directly processed for chromatographic or spectroscopic characterization for a number of reasons. Biological or excipient-laden pharmaceutical samples are too complex and require a series of sample preparation, cleanup, and concentration steps. Sample preparation can be the most laborious, unreliable, and least reproducible step of the analysis. The characteristics of successful sample preparation steps are (a) that a homogeneous representative sample must be obtained, (b) the sample preparation must be performed without losing any of the target analytes for quantitative analysis, (c) the preparation should result in the sample’s being in the most ideal form for the chosen analytical technique while removing, and not cross-contaminating with interfering components, and (d) the preparation steps should dilute or concentrate so that the target analytes are in the correct concentration range for the chosen analytical technique. There are many literature references that describe sample preparation procedures for individual analytes and pharmaceutical samples, which are too numerous to reference here. Our aim instead is to discuss the characteristics of several widely used sample preparation techniques in their broadest sense and to take a more practical approach rather than a theoretical approach. The references for each section were chosen carefully and are recommended to give the reader a comprehensive link to specific topics. Several vendor and supplier Web sites are mentioned that contain detailed information. These sites were selected because of their usefulness, and we would encourage the reader to examine them for more specific information.

II. SOLID-PHASE EXTRACTION (SPE) The conventional process of solid-phase extraction involves passing a liquid sample over a solid surface in order to separate target analytes from the sample matrix. The attractive interactions of the analyte with the solid

8 SAMPLE PREPARATION METHODS FOR THE ANALYSIS OF PHARMACEUTICAL MATERIALS

167

surface must be greater than the attractive interactions of the analyte with the sample matrix in order for retention and extraction to occur. The technique is based on the same principles as chromatographic separations (partitioning, adsorption, or ion exchange). The intent of this section is to serve as a practical guide to SPE sorbent and device selection, extraction steps, and expected performance.

A. Suppliers At the time of this writing, there are at least six main suppliers of solidphase extraction equipment and supplies. At least three of the suppliers offer technical guides and general solid-phase extraction literature.1–4 Table 1 lists the companies and their Internet sites. Thurman and Mills provide a comprehensive list of SPE product suppliers in their book Solid Phase Extraction (Wiley, 1998).5

B. Sorbents The retention mechanisms in SPE are similar to those of HPLC. The packings used in SPE are derivatives of the same materials used for HPLC and are subject to many of the same functional limitations and production problems.6,7 Table 2 summarizes the most commonly available sorbents for SPE. Typical sorbents are made of 40–200-mm irregularly shaped porous silica gel particles with pore diameters averaging 60 A˚ and ranging as high as 300 A˚. Bonded phases are often attached to the silica particles to extend and amplify their functionality. Carbon loading for typical bonded phases (C2 to C18, phenyl, cyclohexyl) ranges from less than 4% to about 17% carbon by weight. It has been reported that heavy loading of bonded ligands can block or fill pores of the SPE sorbents, effectively creating a pellicular material.6 Chemically bonded silica gels with cyanopropyl, aminopropyl, and diol functional groups are also available and each has been used for normal-phase as well as reversed-phase separations. Polymeric-based packings are available and can be used for ion-exchange or reversed-phase applications. Crosslinked polymeric based packings have been combined with ion-exchange materials to create a more pH-stable stationary phase. Graphitized carbon TABLE 1 Solid-Phase Extraction Equipment and Suppliers Company

Internet site

Supleco (Bellefonte, PA)

www.Sigma-Aldrich.com

J.T. Baker (Phillipsburg, NJ)

www.JTBaker.com

3M Company (St. Paul, MN)

www.3M.com/Empore

Varian (Harbor City, CA)

www.Varianinc.com

Waters (Milford, MA)

www.Waters.com

Jones Chromatography (Lakewood, CO)

www.Joneschrom.com

168

D.T. ROSSI AND K.G. MILLER

TABLE 2 Commercially Available Sorbents for Solid-Phase Extraction (compiled from refs. 1^5, 15,16) Sorbent

Typical mode of separation

Octadecyl

Reversed-phase

Octyl

Reversed-phase

Butyl

Reversed-phase

Propyl

Reversed-phase

Ethyl

Reversed-phase

Methyl

Reversed-phase

Cyclohexyl

Reversed-phase

Phenyl

Reversed-phase

Cyano

Normal-phase/reversed-phase

Amine (1 /2 )

Ion exchange/normal-phase/reversed-phase

Diol

Normal-phase/reversed-phase

Propyl amine

Ion exchange/normal-phase/reversed-phase

Silica gel

Adsorption/normal-phase

Alumina

Adsorption/normal-phase

Magnesium silicate

Adsorption/normal-phase

Graphitized carbon

Adsorption/normal-phase

Styrene-divinylbenzene

Adsorption

Quaternary amine

Ion exchange/normal-phase/reversed-phase

Carboxcylic acid

Ion exchange/normal-phase/reversed-phase

Propyl sulfonic acid

Ion exchange/normal-phase/reversed-phase

Benzenesulfonic acid

Ion exchange/normal-phase/reversed-phase

and styrene-divinyl benzene-based sorbents have been used widely for retention and extraction of aromatic-based compounds. Wide-pore sorbents having pore diameters of about 300 A˚ can utilize a size-exclusion mechanism for extraction. As an example of this, a 1,6 glucose polymer cross-linked with epichlorohydrin is available and has been used to desalt protein solutions.2 Magnesium silicate (FlorisilÕ ), silica gel, and aluminum oxide (alumina) are used as normal-phase sorbents and are available in a variety of packing grades. Primary, secondary, and quaternary amines, sulfonic and carboxylic acid, and cyano-based sorbents can be used in ionexchange modes. The pH and ionic strength of the sample solution and ion-exchange capacity of the packing are important operating parameters for ion-exchange extractions. It is common for SPE users to create their own packings by combining materials with different modes of separation. A common mixed-mode system is a weak cation exchange packing with a typical reversed-phase packing (e.g., C18). Variations of this mixed mechanism approach have been described as shielded hydrophobic phases, semipermeable surfaces, dual-zone phases, and internal-surface reversed phases.8,9 This type of methodology has been used

8 SAMPLE PREPARATION METHODS FOR THE ANALYSIS OF PHARMACEUTICAL MATERIALS

169

extensively for sample cleanup with pharmaceutical compounds as the target analytes. Some packings not described in Table 2 utilize mechanisms that are more specific, such as affinity solid-phase extraction.10 These packings are not generally available commercially but are highly specific for target analytes or classes of compounds. The mechanism is identical to performing an enzyme immunoassay where an analyte binds selectively to an active site on an antibody through multiple non-covalent interactions. In this case the antibody is attached, usually via an aliphatic chain, to a silica-based or carbohydrate (agarose) sorbent. The analyte is dislodged by upsetting the binding site and is eluted with buffer or organic solvent. A reversible denaturation of the antibody is a viable elution strategy. This approach is highly selective and is effective for peptides or other biopolymers.

C. Cartridges The basic design of SPE cartridges has changed little since its introduction in the late 1970s, although the cartridge format has now taken on a wider range of sizes. The typical SPE cartridge (see Figure 1) consists of an open syringe barrel containing a sorbent packed between porous plastic, stainless steel, or titanium frits. The dimensions of the sorbent bed are purposely miniaturized to minimize the difficulty in sample processing using gravity- or suction-aided sample flow. The syringe barrels are constructed of polyethylene, polypropylene, glass, or metal. The open portion above the sorbent packing, generally referred to as the ‘‘sample reservoir,’’ is available with volumes from about 1–20 mL. Larger reservoirs with volumes up to about 100 mL are available and can be connected via an adapter to the inlet of an open syringe barrel device. The cartridge inlet can be shaped like a standard syringe or can be flared to allow for easier sample introduction. Cartridges are also available with female Luer-lock fittings at the inlet for easy connection and use with on-line applications. The standard outlet of the cartridge design is a male Luer-lock fitting. Most open inlet designs have a Luer-lock – type fitting at the outlet. Commercial cartridges are available with sorbent masses of about 25 mg to more than 10 g. The packing density of the cartridges varies widely and the cartridge beds are physically not very stable, which indicates that packing is not tightly controlled in the manufacturing process. The instability of the packing can result in heterogenous flow through the sorbent, which reduces the effectiveness of the extraction. Typical processing flow rates are usually between 3 and 30 mL/min. A cartridge containing about 500 mg of sorbent with a bed height of 6 mm will produce about 5–15 theoretical plates.11 As the primary function of SPE is retention and not separation, such low numbers of theoretical plates are not necessarily detrimental for its intended use. Most suppliers now include certificates of analysis for each lot of cartridges, which include some basic information on the silica gel such as average pore size, surface area, pore volume, surface pH, and average particle diameter. This information proves to be very useful because lot-to-lot variability for sorbent materials is high. It also provides a basis for comparison when switching manufacturers.

170

D.T. ROSSI AND K.G. MILLER

FIGURE 1 A pictorial representation of solid-phase extraction, including conditioning, sample introduction, washing, and elution.

8 SAMPLE PREPARATION METHODS FOR THE ANALYSIS OF PHARMACEUTICAL MATERIALS

171

D. Disks In the early 1990s particle-loaded membranes and particle-embedded glass fiber disks emerged as alternates to the SPE cartridge technology. These products were generically classified as disk technology for SPE and claimed the following advantages over cartridge format: (1) shorter sample processing times because of a larger cross-sectional area and decreased pressure drop, enabling higher sample flow rates to be used, (2) decreased plugging by particles because of the larger cross-sectional area of the disks, (3) reduced channeling and improved kinetic performance resulting from the use of smaller sorbent particles and the greater mechanical stability of the disks, (4) the optimized use of bed mass resulting in a cleaner background and lower interferences in the isolation of the analytes of interest.6,11,12 Disks can be used for the same isolation techniques as cartridges. The disks or membranes are used essentially as filter paper except that the analytes to be isolated are extracted and retained by the membrane. Several suppliers sell specialized glass filtration devices to make their use easier. SPE disks are available with the same sorbent functionality as cartridge packings and come in a variety of sizes ranging from 4 mm to 10 cm in diameter with 47 mm being the most popular for impurity isolation. The major difference between cartridges and disks other than format is particle size of the sorbent. The size of particles ranges from about 8 to 30 mm for disks. The large surface area and the small particle size presented by the disk format allows for rapid mass transfer and reduces problems associated with channeling within the bed. These characteristics make disk technology applicable to situations where large amounts of sample (e.g., water) need to be processed or where small volumes of conditioning and elution solvents would be beneficial. Solid-phase extraction disks are available in several useful formats. The free disk format resembles filter paper and is treated in much the same manner. The syringe barrel format is similar to an SPE cartridge, but includes a single prefilter constructed of glass fiber pressed on top of the particleloaded membrane. The prefilter removes large particles associated with the sample matrix and allows for unimpeded passage of the sample through the disk. A miniature syringe barrel format is used for 96 well micro-titer plate applications and is useful when many small volume (< 500 mL) samples are to be processed. The ‘‘syringe tip’’ format replicates a typical syringe tip filter with Luer-lock fittings and a polypropylene housing. The syringe tip filters are currently available in 13-mm or 25-mm diameters.

E. Extraction Process SPE is a sample preparation or sample clean-up technique used to separate compounds of interest from interfering matrix components. The actual process of SPE has been explained in as little as four steps— conditioning, sample introduction, washing, and elution. In reality it can be significantly more complicated. In order to develop a useful method, the entire analytical scheme should be considered. A decision must be made to use a sample cleanup step, based on the requirements of downstream

172

D.T. ROSSI AND K.G. MILLER

analytical techniques. Considerations should be given to the need for quantitative or qualitative analysis. The analytical technique to be used will affect decisions as to the quality and purity of the final sample generated. Assuming these factors have been considered and SPE is deemed a viable choice for sample processing, the procedure for SPE can be developed. The first considerations are for the target analyte and the sample matrix. Gaseous and liquid samples requiring no chemical modification are generally the easiest to process. Solids and complex liquid samples, in which the target analytes are similar to or bound to the sample matrix can be challenging. Body fluids and tissues are common sample matrices in which SPE has been widely applied. Fatty tissue, proteins, and salts have been handled successfully by homogenization, precipitation with organic solvents, and/or dilution with water. Reducing or increasing the solvent strength of the sample matrix depends on whether the target analytes are to be retained or washed through the SPE device and can be done by pH adjustment or addition of the appropriate co-solvent. The Handbook of Sorbent Extraction Technology4 provides a relatively comprehensive reference to modifications of common sample matrices. Knowing as much as possible about the physical and chemical characteristics of the analytes greatly enhances the chances of successful sample preparations. The solubility and polarity of the compounds of interest as well as the sample matrix determine what type of device and mode of interaction should be used. Most SPE supplier technical bulletins have useful flowcharts for general sorbent selection. If the analyte is organic, the molecular weight is used to determine if size exclusion can be used as an extraction mechanism. If size exclusion is not appropriate, the solubility and polarity of the analyte can be used to determine what separation mechanism is suitable. Ionic water-soluble compounds are most readily handled by ion exchange while non-ionic or ion-paired compounds are managed by adsorption, reversed or normal-phase mechanisms depending on polarity. Ion-exchange sorbents are useful for trace metals and reversed-phase sorbents are useful for metal chelates. The characteristics of the sample, the matrix, and the required processing times determine whether a cartridge or disk is more suitable. The same basic procedures for conditioning, sample addition, washing, and elution apply to both cartridges and disks. Conditioning is required to wet the surface of the sorbent and to remove any impurities that may be present on the unused tube or disk. For reversedphase sorbents, an aqueous buffer with low concentration of organic solvent (e.g., 2% methanol) is typically used. For normal-phase sorbents, the packings are usually conditioned with the organic solvent in which the target analyte or sample exists. For ion-exchange sorbents, if the sample is in a nonpolar organic solvent, the sorbent should be conditioned with a solvent in which the target analyte or sample matrix exists. If the sample is in a polar solvent, a methanol buffer or acetonitrile buffer solution adjusted to the proper elution pH and ionic strength is normally used. The amount of conditioning solvent is usually one full column volume if using a tube or cartridge and about 5 mL for a 47-mm disk. Larger amounts of conditioning

8 SAMPLE PREPARATION METHODS FOR THE ANALYSIS OF PHARMACEUTICAL MATERIALS

173

solvent will generally not affect the performance but can increase the performance in some poorly packed cartridges by redistributing the packing into a more uniform bed. The sorbent bed should not be allowed to dry before the sample is introduced, because buffer salts can precipitate and affect wetting of the stationary phase surface. A general recommendation is to leave just enough conditioning solvent so that the top of the frit or disk is covered. The extraction process using tubes or disks is similar to liquid chromatography (LC). Yet instead of a small injection plug of sample, the sample volume can vary from a few microliters to several liters. Samples are generally processed to become suitable liquid samples and poured quantitatively onto the top tube frit or disk as is or through a prefilter. Many commercial devices are available to aid with this sample introduction. Tubes can be processed individually by simply pouring the liquid sample on top of the frit or disk and allowing the solvent to drip through the packed bed by gravity. Other methods include the use of syringe plungers and single tube processors fitted for the tube design and attached via a tube adapter or Luer-lock fitting. Air or nitrogen lines can be attached to some tube processors to force the sample solution and elution solvents through the packing. Sample solution can be poured on top of a cartridge frit and placed in a centrifuge tube, then spun until the sample liquor is pulled completely through the packing. A vacuum flask apparatus can be constructed for use with tubes or disks, using a holed rubber stopper, collection tube (e.g., test tube), a transfer line (e.g., large-gauge needle connected to the SPE tube), and house vacuum or water aspiration. Several companies manufacture vacuum devices that process multiple tubes. Typical designs process 6–24 tubes using an adjustable vacuum manifold, stainless steel or PTFE transfer lines, and glass test tubes as the collection devices. The sample matrix itself can contribute to the loss of retention capacity by not properly wetting the solid-phase surface. To reduce the risk of sample breakthrough via this mechanism, small amounts of organic solvent can be added (1–5%) to maintain properly wetted conditions. The general rule for retention capacity of silica-based sorbents is about 5% (w/w). In practice, the retention capacity is usually much lower before breakthrough can occur. One trick to test for breakthrough is to attach two or more tubes in series and apply the sample solution. The tubes are separated and processed individually to elute any retained analytes. If analytes are detected in more than the first tube, breakthrough is occurring. The bed volume, flow rate, sample matrix, or packing should be evaluated to eliminate breakthrough. Several researchers have addressed breakthrough in SPE, using liquid chromatographic theory.6,7,10–13 In some cases it is desirable to wash the sorbent bed after application of the sample matrix. If the desired analytes are purposely not retained on the SPE sorbent, a wash solution similar to the sample matrix is usually applied to ensure complete removal of the analytes of interest from the sorbent. The washing step typically uses similar volumes as the conditioning step. If the desired procedure is to retain the analytes of interest on the SPE packing and a wash step is necessary to remove sample matrix components, the solution

174

D.T. ROSSI AND K.G. MILLER

selected should be strong enough to remove the interfering compounds but have minimal impact on the retained analytes. Usually, a similar but weaker solution of eluting solvent is used. Eluting the analytes of interest from the packing is customarily done with as small a volume as necessary to completely remove the analytes while leaving behind any interfering compounds that were retained during the wash step. Several smaller eluting volumes usually remove the analytes better than one larger volume. Depending on disk or tube size, eluting volumes range from around 100 mL to about 5 mL.

F. Nonconventional Techniques Solid-phase extraction devices and applications are evolving rapidly, and novel techniques that stretch the classical definition of SPE are becoming routine. Pawliszyn introduced solid-phase micro extraction (SPME) in 1989,5,14 and a commercial apparatus is available from Supelco (Bellefonte, PA). The SPME apparatus is merely a modified syringe that houses a fused silica optical fiber coated with an immobilized polymer film. The fiber can be exposed for extraction and then retracted for insertion or removal from the sample vial or instrument. Both manual and autosampler devices are available and each can be adjusted for proper fiber depth. Several coatings are available with varying thickness including polydimethylsiloxane, polyacrylate, polydimethylsiloxane/divinylbenzene, and carbowax/divinylbenzene. In contrast to SPE, which is an exhaustive extraction approach, SPME will extract only a fraction of an available analyte, hence it is not suitable for the isolation of impurities and degradants in most applications.15 Extensions of silica fiber coating and derivitization have been described and theorized.16,17 Surface-phase extraction techniques involve coating or derivatizing the surfaces of labware or specialty devices and dipping or passing the liquid sample over the device. Simpson has suggested that a particle-loaded membrane or disk could be modified to become a solid-phase wipe or mop and that SPE sorbents could be placed into a permeable or semipermeable membrane to create a tea bag – like device in which target analytes could diffuse into.16 Although not yet applied to pharmaceutical impurity isolations, this could be the basis of future applications.

III. LIQUID^LIQUID EXTRACTION (LLE) Liquid–liquid extraction is an important sample preparation technique from both historical and practical perspectives.18,19 LLE has been in use for many decades as an analytical sample pretreatment to remove unwanted matrix components20 or to selectively extract components of interest from a mixture, thereby purifying and concentrating them for further workup. It is based on the principles of differential solubility and partitioning equilibrium of analyte molecules between two immiscible phases, usually aqueous and organic. Depicted in Figure 2, LLE initially involves pH adjustment of the original (aqueous) sample with an appropriate buffer. This pH adjustment is

8 SAMPLE PREPARATION METHODS FOR THE ANALYSIS OF PHARMACEUTICAL MATERIALS

175

intended to neutralize the analyte molecules, making them more amenable to extraction into an organic phase. The next step in the process is the addition of an immiscible organic extraction solvent, followed by agitation (vortexing) to facilitate equilibrium partitioning of analyte molecules between phases. The phases are physically separated and the aqueous component is subjected to an additional extraction or, alternatively, discarded. If several extracts have been produced, the organic phases are combined, evaporated to dryness and, possibly, resuspended with a clean solvent, depending on the next step in the process. Back extractions into a secondary aqueous phase, once common, have largely become unnecessary when the higher selectivity of HPLC/tandem mass spectrometric detection is available.21 When performed in the organic chemistry laboratory, LLE is a preparative technique, performed manually and employing separatory funnels. When performed in the pharmaceutical analysis laboratory, LLE can be either preparative-scale (hundreds of milliliters), using a separatory funnel, or analytical-scale (microliters to a few milliliters) and performed using test tubes, vials, and glass or polypropylene pipets. Liquid–liquid extraction is a very good sample cleanup technique for nonpolar or moderately polar analytes that can be deionized in aqueous solution by simple pH adjustment. From a practical perspective, the best recoveries are obtained when an excess of organic solvent (3- to 10-fold excess) or multiple-step extractions are used. Multiple-step extractions have recently fallen into disfavor for routine use because they are time-consuming and labor-intensive. Method development for LLE is usually straightforward, as the approach can have a high probability of some success on the first attempt. The technique can, however, be labor-intensive, and it does not provide good recoveries for highly polar or zwitterionic species unless ion-pair reagents are added to the

FIGURE 2 A pictorial representation of liquid^liquid extraction, including pH adjustment, extraction, phase separation, solvent evaporation and resuspension steps.

176

D.T. ROSSI AND K.G. MILLER

sample, effectively making the analyte neutral. This approach is elaborate and often not worth the trouble, especially because other effective approaches for dealing with ionic compounds, such as ion-exchange SPE, are available. Automated liquid handling systems using a 96-well format have gone a long way toward making LLE an effective tool for microanalysis in situations such as drug discovery, where high sample throughput is required.21

A. Theoretical Considerations in LLE For an electronically neutral analyte that is partitioning between water and organic liquid phases, the equilibrium process can be described as the distribution KD ¼

½A org ½A w

ð1Þ

where KD is the distribution constant for the analyte A, partitioning between organic and aqueous phases. A large distribution constant indicates a preference for the organic phase over the aqueous phase, and a small KD indicates a preference for aqueous over organic. With a large KD and a rapid rate of transfer between phases, quantitative transfer from aqueous to organic phase can be made in a single extraction step. When a theoretical basis for LLE is developed, it is often assumed that KD is independent of concentration, yet strongly dependent on other variables such as pH, temperature, and ionic strength. At some point in the concentration range, however, the solubility limit of the analyte in one or both phases will be reached and relationship (1) will break down. One drawback of this effect is that unpredictable recoveries can occur when an extraction procedure is being scaled up. Exhaustive extraction involves the quantitative removal of a solute, while selective extraction involves the separation of two or more solutes from each other. In developing and/or applying LLE it is sometimes expedient to choose solvents and conditions to gain high recovery (exhaustive), then extend and improve the separation so that it becomes more selective. When Vorg milliliters of organic solvent is used to extract XA moles of the analyte of interest dissolved in V milliliters of aqueous sample, the distribution ratio, D, is defined as D¼

CAorg ðXA  YA Þ=Vorg ¼ CA YA =A

ð2Þ

where YA moles of A remain in the aqueous sample after a single extraction. The fraction of analyte remaining unextracted is given as w ¼

YA 1 V ¼ ¼ ð1 þ DðVorg =VÞÞ ðV þ DVorg Þ XA

ð3Þ

and is assumed to be independent of the initial concentration of analyte molecule.22 It can be seen from equation (3) that some of the ways to obtain

8 SAMPLE PREPARATION METHODS FOR THE ANALYSIS OF PHARMACEUTICAL MATERIALS

177

a high recovery of the analyte (or a low fraction remaining) are to obtain a high D (generally through the selection of solvent, manipulation of pH or ionic strength), to make the ratio of Vorg/V large, or a combination of the two. In the development of an LLE, a common tactic is to optimize recovery by using a large volume of an extracting solvent with a smaller volume of sample to be extracted. This approach allows a nearly quantitative or exhaustive extraction with only a moderate distribution constant. If successive extractions are performed with fresh aliquots of organic solvent, after n extractions, the fraction remaining unextracted, n, can be represented as n  1 þ DV org w ¼ ð4Þ V There is no practical gain for n > 4 or 5 because n approaches zero asymptotically.22

B. Countercurrent LLE When the distribution ratio is not highly favorable, it is still possible to obtain a quantitative and selective separation through the use of a countercurrent liquid extraction approach. Although such approaches are no longer practical, having largely been supplanted by instrumental techniques such as preparative HPLC and continuous solvent extraction, countercurrent separations are conceptually useful. These approaches can be applied to preliminary separation of complex mixtures or in the isolation of compounds that do not perform well in LC because of undesirable interaction with the stationary phase (irreversible adsorption, denaturation, etc.). For these reasons, most applications of countercurrent separations involve the isolation of natural or biochemical products from plant or animal extracts. As will be described below, countercurrent extractions form the theoretical basis for LLE cartridges. Countercurrent separations are performed by the exposure of separated aqueous-phase samples with fresh, separate portions of organic phase in an experimental layout known as the countercurrent distribution23 or the Craig apparatus,24 shown in Figure 3. This figure depicts a countercurrent extraction of five sample containers, each containing 20 units of analyte dissolved in equal volumes of solvent. These samples are extracted sequentially, using five equal-volume aliquots of clean organic solvent. For the sake of simplicity, the distribution ratio is 1.0. The distribution ratio can be higher or lower than this, with proportionally better or poorer recovery of the analyte. At equilibrium, when a volume of aqueous phase is extracted with an equal volume of organic phase, 10 units of analyte will partition into the organic phase and 10 units will remain in the aqueous phase. If all aliquots (5V) of aqueous phase were combined and extracted with a single aliquot of organic phase (5V), and assuming no loss of material and completely efficient phase separations, the recovery of analyte into the organic phase would be 50%.

178

D.T. ROSSI AND K.G. MILLER

FIGURE 3 A pictorial representation of a 10 -step countercurrent extraction process for five aqueous samples and five organic aliquots.

8 SAMPLE PREPARATION METHODS FOR THE ANALYSIS OF PHARMACEUTICAL MATERIALS

179

The right-hand vessel containing aqueous sample (bottom row) is extracted first, using the left-hand vessel containing fresh organic solvent (top row). As the extraction proceeds, the organic phases are used to extract each of the aqueous samples in sequence, so that each organic sample is used for five extractions, and each aqueous sample is extracted five times. In order to accomplish this process, 10 extraction steps (involving 25 total extractions) are needed. Again, it is assumed that each extraction step is completely efficient, with no loss of material. At the end of the extraction process, 75.4% of the analyte is recovered into the organic phase, thereby showing a significant improvement in overall recovery for the countercurrent process relative to the single-step process (50% recovery). If this process were extended to the separation of multiple components using many extraction vessels, it is apparent that this approach could be used to separate multiple component mixtures, as long as the individual components had differing distribution ratios. The larger the difference in distribution ratio (D), the easier the separation becomes. The greater the number of vessels (analogous to the number of theoretical plates in the distillation or chromatography experiments) the easier the separation is to perform. The countercurrent model as shown here is not practical for real-world separations because the material losses sustained from five phase separation steps would make the real recovery closer to 50% than the predicted 75%. To make countercurrent extractions practical, they must be implemented in a flow stream, cartridge, or other geometry, where material losses are not so devastating. One such approach is given below.

C. Liquid^Liquid Extraction Cartridges Historically, the difficulty associated with LLE has been the mechanical act of separating the phases.21 As the number of phase separations in an experiment increases, performing the separation becomes increasingly tedious and labor-intensive. A number of different approaches have been proposed for doing the separation of phases in LLE, including freezing the aqueous phase to facilitate organic phase removal, phase separation by filtration through a silicon impregnated cellulose membrane, and direct separation of phases through automated solvent manipulation. An older approach that has only recently been commercialized and is gaining some popularity involves the use of LLE cartridges containing diatomaceous earth.25 In this approach, an aqueous sample is added to a small gravity-fed column packed with diatomaceous earth. The water, along with the analytes of interest and sample matrix components are retained by the diatomaceous sorbent, which can be unbuffered (pH 10) or buffered (pH 4.5 or 9) to offer pH control of the aqueous phase. Acidic compounds are extracted at pH 4.5, while basic compounds are extracted at pH 9. When organic solvent is added to the columns, the analytes partition between the aqueous and organic phases. The organic phase itself is poorly retained and can pass freely through the column and be collected. Additional fresh organic solvent can be added to the column so that, in essence, a countercurrent LLE system is employed. The cartridge LLE approach is useful because it facilitates the phase separations and

180

D.T. ROSSI AND K.G. MILLER

because it can allow LLE to be adapted with automated SPE systems that are becoming well established.26 An added advantage of LLE cartridges is the ability to avoid emulsion formation. This problem is described below.

D. Emulsions An emulsion is a colloidal suspension of one liquid within another.27 Minute organic droplets are occasionally suspended in an aqueous solution after vigorous mixing of the two solvents, especially if any viscous or gummy material was present in either solution. Emulsions are problematic in LLEs because they can take a long time to separate out and allow the formation of a defined phase boundary, thus facilitating a clean phase separation. Emulsions can be eliminated in several ways, including the addition of a small aliquot of concentrated salt solution (NaCl or KCl), a small amount of water-soluble detergent, or a small amount of a nonpolar organic solvent to the biphasic mixture. It is also sometimes possible to avoid emulsions entirely by filtering the aqueous sample before extraction. This can remove the gummy material that leads to emulsion formation. Emulsions can also be minimized through clever solvent selection because they are less likely to occur with nonpolar solvents such as cyclohexane than with partially watermiscible solvents such as ethyl acetate. This, then, suggests a trade-off between emulsion potential and distribution coefficient (and ultimately, analyte recovery).

E. Solvent Selection in LLE Liquid extraction methods are excellent for extracting drug impurities from drug matrices because they offer easy method development and effective sample cleanup. A wide variety of available organic solvents offer a range of solvating and partitioning properties, thereby allowing the selectivity of the extraction to be customized to the analyte. Some desired features of effective solvents for LLE include immiscibility with water, polarity to match the analytes of interest, high volatility for easy drydown, moderate viscosity for effective volumetric transfer, and lower density than the aqueous phase, so that the organic phase rises to the top and can be siphoned off. This last characteristic is not required in that denser-than-water solvents including methylene chloride and chloroform have been used effectively for numerous analytical-scale LLE. Table 3 gives a list of important physical properties for several of the more desirable organic solvents. Although liquid extraction methods have historically been difficult to automate, recent advances in workstations designed to manipulate liquids in a 96-well format have made highly effective semi-automated approaches possible.21

F. Drying Agents Some water is soluble in all organic solvents to varying degrees. This solubility varies from trace amounts for hexane, heptane, and pentane to more than 3% (w/w) for a polar solvent such as ethyl acetate. Depending on

8 SAMPLE PREPARATION METHODS FOR THE ANALYSIS OF PHARMACEUTICAL MATERIALS

181

TABLE 3 Important Physical Properties of Highly Desirable Liquid^Liquid Extraction Solvents Boiling pt (C)

Polarity

Density (g/ml)

Methyl ethyl ketone

80

4.7

0.80

Ethyl acetate

77

4.4

0.90

MTBE þ5%EtOH

58

3.2

0.75

Diethyl ether

35

2.8

0.71

MTBE

55

2.5

0.74

Butyl chloride

78

1.0

0.89

Pentane

36

0.0

0.63

the application, water can be an unwanted component in the sample extract. For example, if the next processing step in a chemical isolation is to remove the solvent by nitrogen drydown, then the presence of water will drastically increase the time required for this process. If the process must be conducted at low (< 40 C) temperatures because of analyte stability considerations, then the presence of a few tens of microliters of water will increase the solvent evaporation time from minutes to hours. If the extract is to be injected into a normal phase HPLC system, then the presence of water (a strong eluent in normal phase LC) will destroy the chromatographic separation. A drying agent is an anhydrous inorganic compound that acquires water of hydration when exposed to moist air or a wet solution. Drying agents can be used to remove water from the organic layer after a phase separation. The crystals of the salt are added directly to the organic layer in sufficient quantity to make a thin layer at the bottom of the container. After a period of standing, the crystals are removed by filtration or decantation. If the solution is very wet, the process can be repeated. Often the organic solvent goes from opaque or cloudy to clear as the drying agent acquires water. Common drying agents are given in Table 4,27 and are classified in terms of capacity, completeness, rate and use. Magnesium salts sometimes cause rearrangements with epoxides, as magnesium is a strong Lewis acid and should be avoided in these circumstances. Calcium chloride cannot be used with compounds containing nitrogen or oxygen because it forms complexes. Calcium chloride also absorbs methanol and ethanol and can be used effectively in this way. Sodium sulfate is a good all-around drying reagent, but may not completely dry the solvent.

IV. SUPERCRITICAL FLUID EXTRACTION (SFE) It can be argued that the first supercritical fluid extractions (SFE) were performed in 1879 when Hannay and Hogarth investigated the solvating capabilities of ethanol.28 However, it took roughly 100 years before supercritical fluids made any significant impact on industrial processes. The removal of caffeine from coffee beans was reported in the 1970s29 and led to

182

D.T. ROSSI AND K.G. MILLER

TABLE 4 Solid Drying Agents that Can Be Removed by Filtration Acidity

Capacity

Completeness

Rate

Use

Magnesium sulfate

Neutral

High

Medium

Rapid

General

Sodium sulfate

Neutral

High

Low

Medium

General

Calcium chloride

Neutral

Low

High

Rapid

Hydrocarbons, halides

Calcium sulfate

Neutral

Low

High

Rapid

General

Potassium carbonate

Basic

Medium

Medium

Medium

Amines, esters, ketones, bases

Potassium hydroxide

Basic

–

–

Rapid

Amines only

Molecular sieves

Neutral

High

Extremely high

–

General

Anhydrous silica gel

Neutral

Medium

Extremely high

Rapid

General

a large increase in the number of published articles focusing on large-scale extractions using supercritical carbon dioxide. The development of commercial supercritical fluid chromatographs and extraction systems escorted a slew of publications about analytical-scale work in the 1980s. In general, the use of supercritical fluids for extraction was touted as having several advantages over traditional extraction techniques such as LLE. These included lower solvent cost, better selectivity through pressure and temperature modification, and better solvating power because of fluid properties. A complete understanding of SFE and its relation to other extraction methods cannot be made without some knowledge of the basic properties of supercritical fluids and the basic principles of analytical SFE instrumentation. The purpose of this section is to give an introduction to the use of supercritical fluids in analytical-scale extractions while focusing on the application of SFE to pharmaceutical analysis.

A. Fundamental Properties of Supercritical Fluids A typical phase diagram (pressure versus temperature) for a pure substance is shown in Figure 4. The regions where the substance exists as a single phase are bounded by curves indicating the coexistence of two phases, which determine the points of sublimation, melting, and vaporization. These curves intersect at the triple point (Tp) and represent the point where solid, liquid, and gaseous phases coexist in equilibrium. The liquid–gas line breaks at the critical point (Cp). The critical point is defined as a point in the phase diagram designated by a critical temperature and a critical pressure where no liquefaction will occur with increasing pressure and no gas phase will be formed upon raising the temperature. These properties define the supercritical fluid region and are a defining characteristic of a substance. Table 5 displays the characteristic critical pressures, temperatures, and densities of some common solvents.30–35 The density of a supercritical fluid depends on the pressure and temperature to which it is subjected. Critical densities of commonly

183

8 SAMPLE PREPARATION METHODS FOR THE ANALYSIS OF PHARMACEUTICAL MATERIALS

FIGURE 4 Phase diagram of a pure substance. TABLE 5 Critical Properties of Representative Solvents Used in SFE (compiled from refs. 30^35) Solvent

Tc (C)

Pc (bar)

Density (g/mL)

CO2

31.1

72

0.47

N2O

36.5

70.6

0.45

NO2

158

98.7

0.27

Ammonia

132.5

109.8

0.23

Water

374.2

214.8

0.32

SF6

45.5

38

—

Helium

268

2.2

0.07

Hydrogen

240

12.6

0.03

Xenon

17

56.9

1.1

HCl

51

83.3

0.45

Methane

82

46

0.17

Ethane

32.3

47.6

0.2

Propane

96.7

42.4

0.22

Methanol

239

79

0.27

Ethanol

243

72

0.28

IPA

235.3

47.6

0.273

Chlorotrifluoromethane

28.8

214.8

0.58

Tetrahydrofuran

267

51

0.32

Acetone

235

47

0.28

Acetonitrile

275

47

0.25

used solvents used for SFE are typically between 0.2 and 0.8 g/mL. The high densities lead to strong fluid–solute interactions similar to liquids. Table 6 compares the properties of supercritical CO2 and typical gas and liquid values.34 High diffusivity and low viscosity lend to the efficient extraction capabilities of supercritical fluids. The solvating power of

184

D.T. ROSSI AND K.G. MILLER

TABLE 6 Comparison of Critical Properties of Supercritical CO2 and Typical Liquid/Gas Values (adapted from ref. 34) Density (g/mL)

Viscosity (g/cm-S)

Diffusion coefficient (cm2/S)

Gas

0.001

0.0002

0.1–0.4

Liquid

0.6–1.6

(0.2–3)  102

(0.2–2)  105

CO2 at Tc/Pc

0.47

0.0003

0.0007

supercritical fluids at high density approaches that of a liquid. The maximum solubility in most liquids surpasses that of supercritical fluids. Even though supercritical fluids do not offer much advantage over liquid solvents as far as solvating power is concerned, they do have other benefits. Mass transfer determines the rate at which extraction can be accomplished. Because solute diffusivities in supercritical fluids are typically higher than those of liquid solvents by an order of magnitude, their mass transfer properties are more favorable. Other important physical chemical properties are polarity and dielectric constant. Water has a high dielectric constant (78.5 at STP), which would effectively mask ionic charges and lead to high solubility of ionic compounds. The dielectric constant of CO2 at 200 bar and 40 C is approximately 1.5, and CO2 is considered a very non polar solvent. As would be expected, polarity influences solubility for supercritical fluids. Carbon dioxide has a dipole moment of 0.0 Debye, while the value for NH3 is approximately 1.5. Therefore, CO2 by itself is poorly suited for dissolving polar compounds.

B. Supercritical Fluid Extraction Instrumentation Supercritical fluid extraction can be performed effectively with very simple systems. Figure 5 displays the basic components of an effective analytical SFE device. There are relatively few commercial suppliers of dedicated supercritical fluid extraction instrumentation. Table 7 shows the companies that promote SFE instrumentation as of the writing of this chapter. Some of the more traditional instrument manufacturers such as Hewlett–Packard (7680T SFE), Dionex (SFE 723), and Supelco (SFE-400) have discontinued their SFE lines. Dionex has invested quite heavily into high-temperature/high-pressure solvent extraction devices, and this will be described in the next section. For most purposes, inexpensive and efficient extraction units can be assembled using the basic components shown in Figure 5.

C. Static/Dynamic SFE Supercritical fluid extractions can be performed statically or dynamically. During static extractions the extraction cell is pressurized and heated to the desired temperature. The supercritical fluid remains in the extraction cell and

185

8 SAMPLE PREPARATION METHODS FOR THE ANALYSIS OF PHARMACEUTICAL MATERIALS

FIGURE 5 Basic components of an analytical supercritical fluid extraction system (adapted from ref. 34).

TABLE 7 Commercially Available Analytical Scale SFE Instruments in the US

Applied Separations 930 Hamilton St. Allentown, PA 18011

Isco 4700 Superior, Lincoln, NE 68505

Durability, Inc. 1872 Pratt Dr., Suite 1630, Blacksburg, VA 24060 - 6344 (Isco-Suprex)

Spe-ed SFE 2

SFX 220

SFE-101

Spe-ed SFE 4

SFX 3560

Spe-ed SFE NP

FastFat HT

Thar Designs, Inc. 730 William Pitt Way, Pittsburgh, PA 15238 Prep and pilot scale systems

does not flow through the restrictor. Extracted components are typically collected by a short dynamic extraction. A vent or shutoff valve must be placed between the extraction cell and the restrictor for static extractions. Dynamic extractions are performed by allowing the fluid to continually flow through the cell and are collected in some sort of collection device.30–35

D. Pumps Used for SFE Pumping liquids (e.g., in HPLC) generally requires less constraint than pumping fluids. Pumps for SFE (or chromatography) must be able to withstand greater pressures (approximately 600 atm) and be able to maintain a constant flow and pressure. It is highly desirable for a pump to be able to recover to full pressure quickly after pressurizing a cell. All of the commercially available instruments identified in Table 7 use syringe pumps. Syringe pumps have the advantage of supplying and maintaining a wide flow and pressure range and can generally withstand pressures in excess of 600 atm, depending on design. They have the disadvantage of a finite volume and must be refilled by the user. Reciprocating pumps have been used successfully, and

186

D.T. ROSSI AND K.G. MILLER

HPLC pumps have been modified to behave as fluid pumps. Probably the simplest and most affordable supply of high-pressure CO2 is that described by Westwood where a 10-mL cell was filled with 9 g of crushed dry ice and heated to 100 C. It provided 170 atm of CO2 for over 30 min.34

E. Extraction Cells The extraction cell is the heart of the SFE process. The characteristics of a good extraction cell include being capable of withstanding the necessary pressures and temperatures of the extraction (i.e., not leak), be capable of withstanding effects from samples, modifiers, or co-solvents; be easy to assemble/disassemble, fill, and clean; and not contribute to carry-over issues. In early work, researchers commonly used modified HPLC fittings or columns. The advent of commercial SFE equipment resulted in the development of analytical-scale high-pressure cells with finger-tight seals that eliminated the need for wrenches. The companies listed in Table 7 supply such cells. Keystone Scientific (Bellefonte, PA) and Thar Designs (Pittsburgh, PA) also supply analytical-scale SFE vessels. The general shape and size of cells used for SFE are cylindrical and contain between 1 and 20 g of sample within an approximate volume of 1–20 mL. Stainless steel is the most common material used for the housing and frits of SFE cells, while TeflonÕ / graphite are common seal and O-ring materials. Isco, Inc. (Lincoln, NE) has a patented cell design constructed of a high-temperature crystalline polymer.30–35

F. Modifiers Carbon dioxide is the most commonly used solvent for SFEs because it is nontoxic, available in high purity grades, and has achievable critical points. It is capable of dissolving and extracting nonpolar to moderately polar compounds but is not particularly useful for extracting polar species. Other solvents such as ammonia have better physical characteristics for extracting polar compounds; however, their toxicity, reactivity, or hazardous nature prohibit them from achieving more widespread use. The easiest approach to solving this problem has been to modify the solvating parameters of the extracting solvent by adding a co-solvent. There is a great deal of literature describing the effects of solvent-modified CO2 extractions versus pure CO2, and the reader is encouraged to research the literature for specific areas of interest. Unfortunately, there is relatively little published information between 1980 and 2000 that reports on the proper selection of a modifying solvent. A common-sense approach would be to begin with a solvent that is compatible with the target analytes and the supercritical solvent. The critical temperature of CO2 with a modifier will be higher than that of pure CO2, and considerations as to the general parameters of the extraction may have to be adjusted (e.g., temperature). The actual mixing of a modifier with a fluid can pose a problem but is usually handled in one of the four ways. The simplest way is to add solvent

8 SAMPLE PREPARATION METHODS FOR THE ANALYSIS OF PHARMACEUTICAL MATERIALS

187

directly to the extraction cell. The results of this can be significant as the co-solvent can disrupt the sample matrix, effectively aiding in the release of target analytes as well as increasing their solubility in the supercritical fluid. The term used for such an extraction is more appropriately named solventassisted SPE and closely resembles the principles of accelerated solvent extraction, described in the next section. When using a syringe pump, another and still less-expensive way to add a modifier to the fluid is to add a known amount of co-solvent directly to the pump chamber. The advantage of this technique is that it is simple and a constant supply of modified fluid can be supplied to the extraction cell. A disadvantage is that the volumes and concentrations of the solvents in the pump chamber vary, which make it difficult to predict the actual composition of the fluid. Another potential problem is contamination of the pump chamber and seals with the co-solvent. A third, and more expensive method for modifier addition, is to purchase the mixed solvent directly from the gas supplier. Custom mixes (e.g., CO2/ methanol or CO2/formic acid) are common but costly. The advantage is a more reliable and predictable solvent composition. The fourth method is to have a second pump, dedicated to supplying a constant amount of solvent, connected to the system. This is likely the most expensive design, but also very common, as most commercially available SFE instruments are designed in this fashion.30–35

G. Restrictors The restrictor controls the flow rate of the SFE system. It is positioned after the extraction cell and ends in a collection device (off-line SFE) or in the injection port of another analytical device (on-line SFE). A shutoff valve is typically placed between the restrictor and extraction cell to enable static extractions to occur. A review of the literature indicates that the restrictor is one of the more problematic aspects of SFE. Restrictors are prone to plugging by ice formation, caused by expansion cooling of the supercritical fluid at the outlet of the restrictor, or by extracted material from the sample matrix. The technology of restrictors as flow-control devices in SFE has made significant advances since initial descriptions30 and has redefined restrictors as either fixed flow or variable flow. Short pieces of small-diameter stainless steel tubing or fused silica (15–50 mm id) are examples of fixed-flow restrictors. Fixed-flow restrictors have also been constructed from tapered fused silica or crimped stainless steel tubing.36,37 The benefits of most fixed-flow restrictors are low cost and simple construction. Stainless steel has the benefit of flexible strength. Small diameter fused silica tubing can become very brittle and fragile when used with modifiers. Flow control is determined by the length and internal diameter of the tubing. Plugging that is due to ice formation can usually be controlled by keeping the temperature of the tip of the restrictor above 0 C. This can be accomplished simply by placing the collection device in a temperature-controlled water bath. Variable-flow restrictors have the advantage of manual or computercontrolled heating and adjustment of internal diameter of the tubing. Applied

188

D.T. ROSSI AND K.G. MILLER

Separations (Allentown, PA) and Isco (Lincoln, NE) have their own versions of variable restrictors, and these are included with their commercial SFE instruments. The cost of variable restrictors can range from hundreds to thousands of dollars.

H. Collection Devices Restrictors have been described as one of the more problematic aspects of SFE. Collection for on-line and off-line SFE has common issues, but on-line SFE is intimately dependent on the connected analytical device. Therefore, only off-line collection techniques with the end result of a chromatographycompatible sample will be summarized here. To have a successful and quantitative extraction, the fluid of choice must solvate the analytes of interest, transport them from the sample matrix, through the extraction cell, tubing, and restrictor, and quantitatively deposit them in some sort of collection device while depressurizing and escaping as a gas. The most common method for analyte collection after SFE is bubbling expanded fluid through a small volume of liquid solvent. This can be viewed as another extraction step in the sample preparation scheme. The proficiency of this method depends on many factors, including the solubility of the analytes in the solvent, the mass-transfer rate of the analytes to the solvent, the flow rate (aerosol formation), and possibly the solvent volume (saturation level) if using a small volume of solvent. Because the expanding fluid can rapidly cool the collection solvent, the collection vial is usually placed in a water bath. Numerous articles can be found on the use of sorbent materials such as silica, stainless steel beads, and common solid-phase extraction materials for collection of supercritical fluid–extracted analytes.38–42 In this scenario the fluid is usually allowed to expand into a gas, which is forced through a packed bed of solid sorbent material. The analytes are bound or simply deposited and cryogenically trapped onto the sorbent and are later eluted with appropriate solvents. The advantage of this technique is that further cleanup can be performed by carefully selecting the eluting solvent.

I. Pharmaceutical Applications of Supercritical Fluid Extraction Supercritical fluids are widely used in manufacturing operations, however analytical-scale SFE has had a relatively modest impact on the isolation of pharmaceutical impurities. Pharmaceutical samples usually consist of complex matrices with polar target analytes, which lead to difficult method development and optimization steps for SFE. Nevertheless, novel laboratory research has successfully used supercritical fluids for the isolation of active pharmaceutical ingredients (API) and pharmaceuticals products. Roston and coworkers found that formic acid–modified CO2 effectively extracted an active enantiomer of misoprosotol, which was covalently linked to the drug product polymer matrix.43 Karlsson, Torstensson, and Taylor

8 SAMPLE PREPARATION METHODS FOR THE ANALYSIS OF PHARMACEUTICAL MATERIALS

189

summarized and demonstrated that SFE could be used with success for extraction of polar substances in tablets, creams, ointments, and aqueous infusions.44 Bonazzi et al.45 extracted four imidazole antimycotic drugs from creams and quantitatively trapped them on a C18 sorbent. The precision of the assay was equivalent to a standard solid-phase extraction procedure and suitable for a quality control assay of commercial cream formulations. Roston and Sun46 showed improved precision and accuracy results for the extraction of an HIV protease inhibitor and benzophenone from animal feed with the use of an internal standard. Simmons et al.47 extracted 13-cis retinoic acid and its photoisomers from cream, gel, and beadlet dosage forms using methanol-modified CO2. Eckard and Taylor48 used 1-heptane sulfonic acid as an ion-pairing agent to extract pseudoephedrine hydrochloride from an inert surface. Khundker, Dean, and Jones49 compared SPE to SFE for the isolation of fluconazole from animal feed. The extraction of polar drugs from chicken livers and meat products was demonstrated by Parks et al.51 and Ezzell et al.50 The general conclusion of both studies was that the solubility of the target analytes in supercritical CO2 does not alone determine the extent of the extraction from a complex matrix. The addition of modifiers played several roles including matrix breakdown and transport. Luque de Castro and coworkers reported for the first time the extraction of vitamin D2 and D3 from multivitamins.52 In this study, the addition of methanol added directly to the sample cell drastically reduced recovery efficiency; however, diethyl ether increased recovery by 10%.

V. ACCELERATED SOLVENT EXTRACTION (ASE) Accelerated solvent extraction (ASE), also referred to as pressurized liquid extraction,53–55 is a relative newcomer to the battery of extraction techniques. This technique has been applied successfully to problems in environmental analysis56–58 and has recently begun to find a few applications in pharmaceutical analysis as well. Some of these applications have involved biological samples59,60 and solid-dosage forms such as transdermal patches.61 As shown schematically in Figure 6, a solvent is selected for ASE with the relative solvating properties in mind. A volume of this solvent is pumped into a sample cell or ‘‘bomb’’ (1–100 mL) containing the sample to be extracted. Usually the volume of solvent added is equal to approximately two-thirds of the volume of the bomb. For a 1-g sample, a bomb of 25–50 mL is a useful size. A photograph of a typical bomb is shown in Figure 7. The bomb is moved via carousel to an oven; pressurized by addition of nitrogen and heated for a predefined time (typically 10–20 minutes). After the prescribed time has elapsed, additional clean solvent is pumped into the bomb and the solvent is purged into a collection vessel with nitrogen gas. The extract is lyophilized, evaporated to dryness with nitrogen, or used directly in the subsequent analytical steps.

190

D.T. ROSSI AND K.G. MILLER

FIGURE 6 Schematic representation of an accelerated solvent-extraction system, showing pneumatic and fluidic pathways.

FIGURE 7 Stainless steel sample cells (5 -mL), or ‘‘bombs,’’ showing screw cap ends and sintered stainless steel frits.

A. Sample Pretreatment An ideal sample consistency for ASE is similar to that desired for liquid extraction, that being a dry, finely divided powder. To achieve this, ASE samples are often ground prior to placing in the bomb. To avoid particle aggregation, a dispersing agent such as Ottawa sand or diatomaceous earth is

8 SAMPLE PREPARATION METHODS FOR THE ANALYSIS OF PHARMACEUTICAL MATERIALS

191

occasionally added to the sample. If the sample contains an appreciable amount of water, it is useful to choose a co-solvent system that is water miscible, such as acetone or alcohol. Although a conventional drying agent such as sodium sulfate, molecular sieves or silica may be added to the sample at this point, this represents an additional step in the extraction process and could be unnecessary. Two reasons why trace water is undesirable in the extraction process are the relative difficulty in performing a dry-down of the extract and the greater potential for degradation of the analyte by hydrolysis. Alternatively, one of the attractive features of ASE is the potential to minimize sample pretreatment prior to extraction. It is therefore desirable to omit as many of these steps as possible.

B. Solvent Selection in ASE A typical solvent or co-solvent system is selected based on the ability to solvate the analyte(s) relative to the undesired matrix components and the ease with which the solvent can be eliminated after extraction. Co-solvent blends are useful because the polarity (or other properties) can be tailored to that of the analyte(s). Traditional aprotic organic solvents are useful because they can be removed quickly and at low temperatures. Because of the dramatic increase in extraction efficiencies, solvents that have only moderate extraction properties at room temperature and atmospheric pressure can perform quite well under ASE conditions. Because organic-aqueous co-solvents can be used, it is often possible to prepare a solvent that can chemically neutralize the analyte molecule, thereby further facilitating the extraction. Dilute organic acids or bases can be employed for this purpose. Strong mineral acids are generally undesirable because they attack and destroy the stainless steel bombs or other instrument system components.

C. The Effects of Temperature in ASE Aside from solvent selection, temperature is one of the most important variables in ASE, as it dictates solvent viscosity, wetting properties and ultimately the equilibrium of analyte partitioning between matrix and solvent. Extraction efficiency and overall recovery will increase nonlinearly to a point and then level off. Higher temperatures are generally more efficient unless thermal degradation of the analytes begins to occur. The temperature range of commercially available ovens can be precisely controlled from nearambient to 300 C or higher, but the higher temperatures may not be appropriate if drug degradation is observed. When developing an extraction method for a drug impurity and optimized recovery is desired, a good approach is to construct a graph of recovery versus temperature over 20 C intervals. A rule of thumb is to operate approximately 20 C below the temperature where degradation is first observed and recovery begins to decline. An idealized recovery experiment that is hindered by thermal degradation is shown in Figure 8. Many environmental applications operate in the 75–125 C range, while pharmaceutical applications operate at lower temperatures because greater potential for degradation exists.62

192

D.T. ROSSI AND K.G. MILLER

FIGURE 8 Graph showing relative recovery versus temperature for a thermally labile compound in accelerated solvent extraction.

D. The Effects of Pressure Increased pressure will allow the solvent to remain as a liquid (although it may be above the atmospheric pressure boiling point) and facilitate its transit through the extraction system. Changes in pressure will have little effect on analyte recovery when the applied pressure is well above the minimum required to maintain the solvent in liquid state. Typical ASE extractions work well between 1000 and 2000 psi.

E. Cycle Time All other variables being equal, a partitioned equilibrium for the analyte between the sample matrix and the extraction solvent is reached more quickly at higher temperature and pressure because the analyte solubilization kinetics are improved. Therefore, cycle time can be much shorter for ASE extractions relative to room-temperature/pressure-solvent extractions. If certain sample variables such as pore size or structure make rapid equilibrium questionable, it is simple to design a recovery versus extraction time experiment (the results of which are shown in Figure 9) so that variability and lower recovery due to a pre-equilibrium phase separation can be avoided. The desirable extraction duration is a trade-off between the recovery and the time required to achieve it and generally runs from 10 to 17 min. It is also possible to introduce fresh solvent into the cell over the course of the extraction or to add additional extraction cycles to the overall extraction process. As fresh solvent is added to the cell, extract is transferred into the collection vessel. This process helps to disrupt and shift the equilibrium toward quantitative completion of the extraction.53–55

F. Carry-over A big advantage of ASE is the ability to extract with reduced volumes of solvent relative to more conventional liquid extraction approaches such

8 SAMPLE PREPARATION METHODS FOR THE ANALYSIS OF PHARMACEUTICAL MATERIALS

193

FIGURE 9 Graph showing percent recovery as a function of extraction time in accelerated solvent extraction.

as Soxhlet extraction.6 A 1-g sample can be extracted in less than 15 minutes using less than 10 mL of solvent using ASE, while this could require an overnight extraction using continuous Soxhlet. This low volume capability does, however, raise some concern about the carry-over performance of the instrumentation. Recently it has been shown with the Dionex ASE-200, that carry-over is minimal and manageable (< 0.13%) for non-ionic analytes.63 Unfortunately, cationic analytes occasionally adhere to metal surfaces (especially the sintered stainless steel frits within the extraction cells), and silation or special washing solvents, such as dilute organic acid, could be required to prepare the surfaces before or after an extraction.63 Compared to older, more-established sample extraction techniques, ASE is an infant. Although commercial instrumentation has been developed to the point where it can be effectively utilized for many difficult tasks in isolation of drug-related substances, only a very few pharmaceutical applications have appeared to date.61 Reasons for this could be that the approach is a more efficient, yet relatively risky alternative to SPE or LLE. Solvent heating is often involved in ASE and sometimes causes degradation of the compound to be isolated if the thermal degradation properties are not defined. These properties are seldom defined in the early stages of impurity isolation. ASE does, however, offer distinct advantages over roomtemperature, atmospheric-pressure techniques. The solvating power of heated, pressurized solvent far surpasses that of room-temperature, atmospheric-pressure solvent and the potential for automatically performing difficult extractions with relative ease is high. As ASE becomes more common in pharmaceutical analysis laboratories and researchers better understand how to contend with the variable of thermal degradation, it is likely that ASE will become more commonly used and more widely applied than it is today for the isolation of impurities.

194

D.T. ROSSI AND K.G. MILLER

VI. CENTRIFUGATION A. Centrifugation for Small Molecule Sample Preparation Centrifugation is a physical method for separating particles from each other in solution. Although it is widely used in biochemistry for the separation of various cellular fractions and components (also as whole cells, subcellular organelles, membranes, or large biomolecules), it also can be used effectively for separating small molecules contained in heterogeneous solution from suspended particulate matter. The centrifugation process, called sedimentation, is governed by Stoke’s Law, which describes the settling of a sphere in a gravitational field.64 From the Stoke’s equation v¼

2r2 ðp  m ÞG 9

ð5Þ

it can be seen that the sedimentation velocity ( ) is directly proportional to the radius (r) of the particle squared, the difference in density between the particle (p) and the medium (m), and the centrifugal force (G). Sedimentation velocity is inversely proportional to the viscosity of the medium (). The Stoke’s equation shows that the sedimentation rate is faster for large particles than for small particles, is faster in low viscosity media than in high viscosity ones, and is faster as the centrifugal force increases. The sedimentation rate is faster when the density difference between the particles and the medium is large. Centrifugal force (G) is calculated from the equation G ¼ !2 r

ð6Þ

where ! is the angular velocity (in radians per second, 2p in a complete revolution) and r is the effective radius of the centrifuge rotor. The calculated centrifugal force is often converted to relative centrifugal force (RCF) in units called x g. Relative centrifugal force can also be calculated directly from the equation RCF ¼ k  rpm  r

ð7Þ

where k is a constant (1.118  10-5), rpm is the rotor speed in revolutions per minute, and r is the effective radius of the rotor (in cm). This effective radius is not the actual radius of the rotor; it is the distance of the sample from the center of the rotor. For low-speed sedimentation applications, RCF of 2000–3000 g and rpm values of 2–3 k are typical. Regular glass centrifuge tubes can be used at these speeds and forces. For biological sedimentation applications, RCF of 10–20 k (ultracentrifugation) are typical. Special glass containers can be used at speeds up to 18 k rpms. At the highest speeds, polypropylene or nitrocellulose tubes are used so that tube shattering is

8 SAMPLE PREPARATION METHODS FOR THE ANALYSIS OF PHARMACEUTICAL MATERIALS

195

avoided. Centrifuge rotors are subjected to great stress during centrifugation, and for this reason they are rated for maximum speeds at which they can be safely used.65 A technique called differential centrifugation is commonly used to fractionate particles into pellet and supernatant. The pellet is an aggregate of all sedimenting components, while the supernatant is a purified portion of the sample containing only the slowest sedimenting components. In this way, differential centrifugation is a purification technique in which large particles are removed from the supernatant. Differential centrifugation65 is commonly used in work involving the isolation of impurities and/or degradants from tablets or other solid dosage forms because, although the active drug components and related substances are often extractable into a simple solvent such as water, alcohol, or acetonitrile, the excipients seldom are. Differential centrifugation can be used to remove these insoluble component particles from the sample preparation before a secondary extraction technique, such as solid-phase extraction, is applied. During the process of differential centrifugation, some of the analyte of interest is lost as it occludes with the pellet, making the yield of this technique less than perfect. Zone velocity centrifugation66 is a technique that can lead to higher yields because a separation gradient of the various analytes is formed. In contrast to differential centrifugation, this gradient technique displays a distinct band for each analyte. It is not often used for the separation of low molecular weight impurities from solid-dosage-form extracts because it is dependent on using a gradient material (solvent) that has a density lower than the analytes of interest, and this is not usually achievable or desirable in low molecular weight (< 2000 Da) impurities isolation.67 In addition to the relatively straightforward particle removal applications described above, ultracentrifugation is also effective at separating, purifying, and fractionating biological materials and for the study and physical characterization of macromolecules. These applications are beyond the scope of this chapter.

VII. FILTRATION Membrane filtration is a widely used but narrowly understood technique for sample preparation in chemical analysis. This section has the goal of providing some basic information to aid in the use of filtration tools with drug impurities. Many of the common sample preparation approaches described elsewhere in this chapter, such as liquid extraction, supercritical fluid extraction, and accelerated solvent extraction are effective at removing the dissolved analytes of interest from the matrix while leaving behind many poorly soluble or insoluble matrix components. In contrast, filtration is designed to remove these suspended particles from the extract prior to subsequent analytical steps. Unfiltered samples can destroy the performance of a downstream analytical technique such as HPLC or optical spectroscopy.68,69 The choice of a filter is dependent on four parameters: the membrane material, the material used to house the membrane, the membrane pore size,

196

D.T. ROSSI AND K.G. MILLER

and the device size or membrane diameter.70 The effects of these parameters are described below.

A. Membrane Selection The membrane material is selected on the basis of compatibility with the solvent being used and the chemical nature of the analyte. Membrane material is selected so that the analyte will not bind to it and so that analytical recovery will be as high as possible. The selection of material may also extend to the size and amount of particulate present in the sample. For example, certain materials such as sintered borosilicate glass are excellent for removing particles from fluids. The potential for interactions of the membrane or filter housing with the sample components is also a consideration. For example, when protein analytes are being considered, a cellulose or cellulose acetate filter would be preferred over a synthetic polymer membrane such as nylon. For removal of large particles in aqueous samples from ground tablets, polysulfone membranes work well. When organic solvents are present and fewer particles need to be filtered out, polypropylene or polyethyltetrafluoroethylene (PTFE) membranes can be effective. Table 8 lists the most common membrane filter materials and typical applications for them.

B. Housing Selection Housing material is generally either acrylic or polypropylene. Acrylic is suitable for aqueous samples. It offers fair to poor solvent resistance. Polypropylene offers high solvent and acid/base resistivity. Either of these housing materials offers extremely low extractables and low protein binding. Analyte binding to either membrane or housing material is difficult to predict and must be empirically evaluated on an individual case basis. Some analytes such as polyamines will be very adsorptive, and their binding can be predicted by the extent of protein binding that is expected.

C. Pore Size Membrane pore size rating refers to the size of a specific particle to be retained by the filter with a specific degree of efficiency. Pore size will affect the flow rate, back pressure, and life of the filter. Membrane pore size is outlined in Table 9 and selected on the basis of the applications requirements. The size and amount of particles to be filtered from the sample can also affect the pore size selection. Most applications in the isolation of drug impurities and degradants would either require prefiltration to remove large clumps of insoluble excipients or a filtration/clarification step prior to HPLC, and would be chosen on these bases.71

D. Filtration Device Size The size (diameter) of the filtration device is selected on the basis of the volume of liquid to be filtered and, to some extent, the cleanliness of the

Membrane type

Application

Compatibility

Borosilicate glass

Prefiltration

For particle-laden samples. Prolongs membrane life and reduces filtration time.

Cellulose acetate

Clarification sterilization

Good for biological applications.

Nitrocellulose

DNA and RNA binding

Good for biological applications with higher protein-binding characteristics.

Regenerated cellulose

Particle filtration

Good for particle filtration.

Nylon

General filtration, sterilization

Displays some protein binding and good solvent inertness.

Polypropylene

HPLC sample preparation, general filtration

Good for both aqueous and organic samples. Has low protein-binding properties.

Polysulfone

General filtration, sterilization

Good flow-rate characteristics. Best for aqueous samples.

Polytetrafluoroethylene

Gas, air, and solvent filtration

Good for chromatographic applications and when protein binding is not an issue.

Organic solvent resistance

Protein binding

þþþþ

þþþ

[ [

þ

–

þ

þþ

[ [

þþþ

—

þþ

þ

[

þþþ

–

þ

–

þþþþ

þþ

Hydrophilicity

[

8 SAMPLE PREPARATION METHODS FOR THE ANALYSIS OF PHARMACEUTICAL MATERIALS

TABLE 8 Properties and Uses of Filtration Membrane Materials

þþþþ indicates strong solvent resistance or high protein binding þþ indicates moderate solvent resistance or protein binding – indicates low solvent resistance or protein binding — indicates very low solvent resistance or protein binding

197

198

D.T. ROSSI AND K.G. MILLER

TABLE 9 Filtration Membrane Pore Sizes and Possible Applications

Application

Components for removal

Pore size (km)

Biological separations

Subcellular particles

0.1

Liquid sterilization and aseptic processing

Microorganisms

0.1–0.2

HPLC samples on < 3 mm columns

Particulates

0.2

General clarification

Particulates

0.45–1.0

HPLC samples on > 3 mm columns

Particulates

0.45

Solvent venting

Airborne particulates

0.2–0.45

Prefiltration

Macro particulates

0.8–5.0

TABLE 10 Relationship Between Filter Diameter and Volume of Fluid to Be Filtered Fluid volume (mL)

Membrane diameter (mm)

Fluid holdup volume (kL)

100,000 in some cases). Yet even these impressive statistics pale in comparison to the capabilities of FT mass spectrometers, some of which can deliver sub-ppm mass accuracies with mass resolving powers on the order of 1,000,000 or higher. Together, high mass accuracy (with precision in the low ppm range) and the ability to uniquely measure a single ionic species (i.e., purity of the analyte ion population of interest is guaranteed by the resolving power) combine to provide a limited list of potential elemental formulas. This information can be extremely valuable in the characterization of unknown ion structures,148 which explains the reemergence of accurate mass determinations in the recent mass spectrometry literature. That being said, the utility of accurate m/z measurements diminishes with increasing mass of the analyte because the number of possible formulas having m/z values that differ by less than the uncertainty of the measurement increases geometrically. It is also important to mention that the execution of such experiments is far from trivial if reliable data are desired. Calibration procedures, internal standardization, and the requirement of appropriate signal levels (necessary to avoid peak skewing due to pulse pileup) all contribute to the complexity of the experiment. While the reemergence of TOF technology has resulted in widespread interest in accurate mass measurements, to date only the m/z values of precursor ions and first generation fragment ions (i.e., MS and MS/MS, or MS2) have been measured using these devices. In this respect, the

11 MASS SPECTRAL CHARACTERIZATION

277

TOF instruments are similar to sector instruments because of the need to have n þ 1 analyzers for every n stage of structural interrogation (i.e., precursor selection, excitation, and product analysis).

VI. ION STRUCTURE INTERROGATION Beginning in the 1950s with chemical ionization and continuing to the present day (a progression that includes field ionization,149 plasma desorption,150 fast-atom bombardment, thermospray, matrix-assisted laser desorption,151 electrospray, and atmospheric pressure chemical ionization), ‘‘soft’’ ionization methods have continually expanded the range, type, and size of molecules that can be analyzed by MS. The ability to ascertain the molecular weight of very large and very fragile molecules (presumably having structures that closely resembled their neutral precursors) has come as a consequence of the minimal amount of energy these techniques impart to newly formed ions. Mass spectra obtained using the techniques mentioned above generally exhibit few signals, which greatly limits their structural information content. In order to recover information about ion structure that traditional electron ionization had always provided, these ‘‘soft’’ ionization techniques were coupled with an activation event designed to increase the internal energy of the analyte ion. The first, and still most prevalent, means of achieving this activation or deposition of internal energy is through the conversion of some portion of the analyte ion’s translational or kinetic energy. The simplest way to accomplish this desired translational/internal energy conversion is through glancing collisions with gaseous target atoms or molecules. Described initially over 30 years ago,152 collision-induced dissociation (CID) or collisionally activated decomposition (CAD) continues to be an important means for probing ion structures. Other techniques for energizing ions have been described, including the use of solid surfaces in lieu of gaseous targets (surface-induced dissociation, SID),153 the use of photon beams (photodissociation, PD)154 and the use of electron beams (electroninduced excitation in organics, EIEIO).155 However, none have gained the commercial acceptance of CID for conducting tandem mass spectrometric experiments. Ideally, an ion activation method would impart a desired (or controllable) quantity of energy in a desired or controllable time frame and the distribution or spread of energies imparted would be narrow. None of the methods available possess all of these characteristics. The precision with which the various methods deposit energy into ions generally is not high. Most of the methods produce an ion population that is best described as possessing a distribution of internal energies. This range of internal energies results in a diversity of fragmentation reactions, which is generally viewed as a beneficial attribute when ion structure elucidation is the goal. Unfortunately, having a range of internal energies also allows for the possibility of isomerization (rearrangements) within the ion during the time between excitation and dissociation. These rearrangement processes and their resulting products are no different from those observed in EI spectra and are no less likely to confuse or complicate the interpretation of fragmentation

278

D. J. BURINSKYAND F. WANG

spectra. The degree of fragmentation achievable in a CID MS/MS experiment is controllable (to the extent allowed by the intrinsic properties of the analyte molecule) in some instrumental configurations through the selection of the kinetic energy of the ions (i.e., collision energy), and the number (collision gas pressure) or type of targets that an ion encounters in the excitation region. The maximum achievable degree of fragmentation will be observed for ions that interact with a large number (high pressure) of massive target particles at relatively high kinetic energy. Due to drastic differences in operating pressures, experimental time scales, and the translational energies of ions as they traverse analyzers, the parameters of the CID experiment can be quite diverse and the terminology somewhat ambiguous. CID experiments intended to probe ion structure for the purpose of compound characterization or identification should maximize fragmentation in terms of both the variety and abundance of ions produced. From the earliest days of MS, it was recognized that the yield and variety of dissociation processes could be enhanced by raising the internal pressure of the instrument.156 Until the popularization of the triple quadrupole mass spectrometer in the 1980s, most of the fundamental investigations into CID processes had been conducted at kilovolt energies using sector instruments. The use of reverse geometry instruments (magnetic sector preceding the electric sector) and the importance of collision gas identity and pressure were the contributions of early research efforts.157 The most common conditions for high-energy CID experiments included the use of helium as collision gas at a localized pressure sufficient to diminish the intensity of the mass-selected precursor ion beam by approximately 30–70%. This degree of beam suppression correlated with multiple collision conditions and a maximum in the signal intensity versus collision gas pressure function for fragment ions.158 Multiple collision conditions are important for enabling access to higher energy processes, particularly when the precursor ions are formed by ionization events that are ‘‘soft’’ and impart very little internal energy. They result in consecutive dissociation reactions for one or more subsequent generations of fragment ions. The use of helium was determined to be optimal because energy deposition is reasonably efficient and losses due to scattering or neutralization of fragment ions are minimal. Increasing the mass of the collision gas target by using nitrogen, argon, or some other gas was found to improve the energy deposition process only marginally while degrading transmission (i.e., increasing scattering or neutralization) unacceptably. A special case where using a collision gas other than helium was advantageous was the investigation of charge permutation reactions, which benefited substantially from the use of oxygen.159 The sector instruments were not amenable to significant adjustments in the translational energy of the ions. Even in cases where this was possible, little if any difference was observed in the appearance of fragmentation spectra as a function of collision energy. The situation is quite different for low-energy CID experiments. The effect of changing either translational energy, collision gas, or collision gas pressure can be quite dramatic. Trapping instruments have greater limitations than quadrupole mass filters with respect to the maximum kinetic energy they can impart to ions, the pressure of collision gas used, or which gases can be used.

279

11 MASS SPECTRAL CHARACTERIZATION

These shortcomings are overcome by the ability of these devices to subject the ions multiple times to the lower energy collisions occurring at lower pressures. Thus, restrictions posed by the need to maintain ions within the trap under stringent pressure requirements (either low pressures in the case of the ICR cell or a constant pressure of helium bath gas for the quadrupole ion trap) are offset by the long residence times of the ions within these analyzers. The linear beam-type instruments, such as the recently popularized quadrupole time-of-flight (QTOF) or the triple quadrupole instrument, are more flexible in terms of the upper limits of translational energy that can be used in CID experiments. The collision gas pressure can be varied in both of the quadrupole-based instruments for the purpose of achieving multiple collision conditions. However, much in the same way that too high a pressure was found to be detrimental in the high-energy regime, so too is excess pressure a liability in the low-energy regime of CID. Diminished intensity of fragment ion signals due to scattering losses is equally undesirable in these experiments. The merits of choosing a more massive gas target over a less massive target for CID experiments are manifested in the degree or extent of fragmentation that can be achieved. The upper limit for the amount of internal energy that can be imparted in a single collision event in the lowenergy regime (i.e., approximately 1–100 eV) is defined by the following expression, Ecom ¼ Elab ðmt =mt þ mp Þ where Ecom is the energy of the ion in the center-of-mass frame of reference (related to the internal energy of the ion), Elab is the kinetic energy of the selected projectile ion in the laboratory frame of reference (i.e., the offset potential of the collision quadrupole in a triple quadrupole instrument), mt is the mass of the target atom or molecule and mp is the mass of the projectile (analyte) ion. The importance and impact of target mass are illustrated by the examples for singly charged ions shown in Table 2. Increasing the mass of the collision target causes the conversion of a larger portion of the projectile ion’s translational energy into internal energy, thus producing more structurally diagnostic fragment ions of higher abundance. The most likely amount of energy imparted by any single collision with a gaseous target is only some fraction of the theoretical limit. Nevertheless, the increase of internal energy as a function of target mass has clear utility in the CID of larger species (having many degrees of freedom among which the collision-induced excitation energy can be dissipated) or small ions having very stable structures. Figure 6 illustrates the effect of varying collision energy and the impact of target mass. The interpretive utility of a fragmentation spectrum is based on (a) the quantity of information available, which is reflected in the number and intensity of the signals in the spectrum and (b) establishing precursor/product (familial) relationships between ions represented by the signals in the spectrum. The derivation of an ion structure requires a fragmentation scheme of most probable structures for the various fragment ions (and neutral products that are eliminated in the fragmentation processes), as well as their relationships to one another. The dissociation processes responsible for the

280

D. J. BURINSKYAND F. WANG

TABLE 2 Illustration of the Relationship Between Target Gas Mass and Energy Deposition in Collision-Induced Dissociation Reactions Elab (eV)

m*t

10

28 (N2)

100

2.2

10

28

500

0.5

10

28

900

0.3

40

28

100

8.8

40

28

500

2.1

40

28

900

1.2

10

131 (Xe)

100

5.7

10

131

500

2.1

10

131

900

1.3

40

131

100

22.7

40

131

500

8.3

40

131

900

5.1

mp

Ecom (eV)

*Increasing the mass of the collision target causes the conversion of a larger portion of the projectile ion’s translational energy into internal energy, thus producing more structurally diagnostic fragment ions of higher abundance. (see Figure 6).

SCHEME 1

formation of product ions in CID experiments are of one of two types, either consecutive or competitive, as illustrated in Scheme 1. Product-ion MS/MS spectra recorded using triple quadrupole or QTOF instruments are composite pictures of all fragmentation reactions (consecutive and competitive) captured in a single experiment. While information rich, these spectra can pose an interpretive challenge because of the ambiguous origins of various fragment ions. Additional information about some of the relationships that exist between ions can be obtained from subsequent experiments. One approach employs a sequential CID experiment in which a first stage of CID is undertaken in the source region of the instrument (so-called in-source or skimmer/cone CID) and followed by the standard product-ion MS/MS experiment. Using this procedure, the fragmentation spectrum for each fragment ion observed in the original product-ion MS/MS spectrum of the analyte ion can be recorded. The precursor/product relationships between various fragment ions can be deduced upon compilation of such data sets. The success of this experiment depends on two critical factors: first, that the fragment ions generated in the source CID

11 MASS SPECTRAL CHARACTERIZATION

281

FIGURE 6 Product-ion spectrum for the [MþH]þ ion (m/z 529) of a therapeutic azasteroid illustrating the dependence of structural information on both collision energy (translational energy of the analyte ion) and the mass of the target (identity of the collision gas). Top panelTranslational energy of the precursor ion was 40 eV (Elab, laboratory frame of reference) with xenon as collision gas. Middle panelTranslational energy of the precursor ion was 20 eV (Elab) with xenon as collision gas. Bottom panelTranslational energy of the precursor ion was 40 eV (Elab) with air as collision gas.

282

D. J. BURINSKYAND F. WANG

stage of the experiment are identical in both mass and structure to the fragment ions observed in the product-ion MS/MS spectrum of the analyte ion; and second, that the intensities of the in-source CID product-ion signals are sufficient to yield second-generation fragmentation spectra of adequate quality (S/N) to interpret. Another approach amenable to triple quadrupole but not QTOF instruments, is to record precursor-ion spectra for each fragment ion observed in the original product-ion spectrum of the analyte ion. Each of these spectra will display all of the precursor ions responsible for the fragments’ formation through either single or multiple step processes. Both of these approaches are time-consuming and labor-intensive. Product-ion MS/MS spectra for many compound classes obtained using trapping instruments tend to have a simpler appearance (i.e., fewer fragment signals) than comparable triple quadrupole or QTOF spectra. The minimal number of signals can be a hindrance to interpretation since only minimal insight can be gained into the ion structure being examined. Fortunately, this shortcoming can be overcome by utilizing a series of sequential CID experiments. These sequential experiments are made possible by the ion trap’s capacity to isolate and energize subsequent generations of precursor ions, each a fragment of a previously mass-selected precursor ion. However, this cascade of ionic dissociation products is not unending. Despite their very high efficiencies for trapping and fragmenting ions, the trapping instruments are constrained by certain practical limits. In general practice, there are few situations that require more than 4 or 5 stages of MS/MS (sometimes denoted as MS4 or MS5), and even these situations are rare. Ion traps generally scan more rapidly than quadrupole instruments, although specialized functions such as those required for increased mass resolving power160 or high mass analysis161 take considerably more time than a conventional analytical scan.

VII. DATA ACQUISITION AND INTERPRETATION The popularity of HPLC in pharmaceutical analysis is due not only to the wide variety of compounds to which it is amenable but also to the variety of detectors that can be used to obtain information about compounds once separated from one another. Several detectors can be employed simultaneously in order to boost both the amount of information and the rate at which it can be obtained. Each detector adds a new dimension of information to the resulting data set. Simple instruments use refractive index or single wavelength UV detectors to produce characteristic two-dimensional response traces for eluting compounds that plot intensity (absorption or refraction of light) versus time. Replacing the single wavelength detector with a diode array adds a third dimension to the data set, associating a complete absorption spectrum with each time/intensity data point. Introduction of a mass spectrometer, either in series or parallel, adds one or more additional dimensions of data, depending on the type of instrument chosen. Single analyzer instruments, be they quadrupole mass filters or TOF analyzers, add

11 MASS SPECTRAL CHARACTERIZATION

283

the fourth dimension to the data set—molecular mass. If tandem instruments such as triple quadrupole or QTOF instruments are used, a fifth (and possibly a sixth) dimension are added to the data set—m/z values derived from a precursor m/z value. Finally, with the use of ion traps (either quadrupole ion traps or ion cyclotron resonance spectrometers), a sixth, seventh, eighth, or higher (in theory up to thirteenth, with currently available instruments capable of 10 stages of analysis) dimension of data is accessible for a single compound eluting from an HPLC column. However, visualization of ndimensional information (where n 4 or 5) poses challenges. Commercially available data-processing software deals effectively with three-dimensional data (either time/intensity at a specific UV wavelength coupled with diode array spectra or time/intensity for all masses coupled with complete mass spectra). The most common representation of chromatographic data when a mass spectrometer is incorporated into the detection scheme commonly is referred to as a total ion chromatogram or TIC. This trace correlates time (x-axis) with a summation of the abundances (y-axis) of all ions detected in the specified scan range (or acquisition range for nonscanning instruments). Because each data point in the TIC is composed of a complete mass spectrum, it is a trivial operation to produce the second most common and possibly the most useful data representation, the mass chromatogram or selected ion trace (also referred to as an extracted ion trace). This trace highlights one or more specific compounds by virtue of their unique molecular masses. The utility and value of mass chromatograms is realized in situations where the molecular weights of related substances can be predicted or inferred from other information. Any and all compounds having the requisite molecular mass (assuming that they ionize to produce an ion with the expected m/z value) will be differentiated from all other responses in the total ion chromatogram. Once the information is obtained, processed, and appropriately displayed in a meaningful fashion, the process of spectral interpretation can begin. Considerable information regarding compound class or the presence of structural analogs can be obtained from the UV absorption spectrum of HPLC peaks if diode array detection (DAD) is employed. Although the UV absorption spectrum of any particular compound may not provide specific information upon which a definitive identification can be based, these spectra often make it possible to differentiate drug-related compounds from excipient compounds. This differentiation can be an important first step in the characterization process because it can eliminate compounds of little or no importance. The molecular mass of an analyte compound is generally reflected in the m/z value of the most abundant ion in the mass spectrum (e.g., [MþH]þ for positive ESI or APCI and [MH] for negative ionization modes). This conclusion can be substantiated by the appearance of other ions such as adducts ([MþNa]þ, [MþK]þ, [MþHþCH3CN]þ, [MHþ HCO2H], [M-HþCF3CO2H], etc.) and dimer species ([MþHþM]þ, [Mþ NaþM]þ, [MHþM], etc.). Aside from the immediate information offered by the signals in a mass spectrum, that is m/z values for ions that represent the intact molecule or its fragments, there are also a number of more subtle details about the analyte that can be extracted upon closer examination.

284

D. J. BURINSKYAND F. WANG

The observation of naturally occurring isotopes was one of the first important accomplishments achieved using early mass spectrometers. Although isotopic patterns are often ignored, they can provide useful information. For compounds containing elements that have a distinct and recognizable distribution of isotopes, such as some halogens and most transition metals (among others), identification or recognition of a compound can be immediate. The majority of molecules encountered in pharmaceutical analysis lack atoms that possess distinctive isotope patterns. However, at the early stages of some synthetic schemes it is not uncommon to encounter synthons that contain atoms such as chlorine, bromine, sulfur, and boron. The appearance of characteristic isotopic signals in a mass spectrum can also aid in the interpretation of fragmentation pathways. When tandem MS experiments are used to obtain structural information, the isotopic information is no less valuable, provided the necessary experimental conditions are used. One of the hallmarks of product-ion MS/MS spectra is the specificity of the information that comes from the selection of ions having a specific m/z value. This monoisotopic (at either unit or high resolving power) selection of precursor ions was initially an important element of the MS/MS experiment because there was no separation technique prior to ionization. The impact of monoisotopic precursor ion selection is a loss of the informative isotopic pattern. The lost information can be recovered through multiple experiments in which each isotopic signal is successively interrogated. The multiple fragmentation spectra are then compared in order to identify the mass shifts that indicate the presence of the isotope in a particular ionic fragment or its elimination in a dissociation process. Alternatively, when a separation technique such as HPLC is used to introduce a compound into the mass spectrometer, guaranteeing the homogeneity of the ion source contents, the importance of monoisotopic selection of precursor ions is reduced. Consequently, the resolving power of the first analyzer (in a beam instrument) or first stage of analysis (in a trapping instrument) can be relaxed so that ions of several m/z values are passed through to the activation stage. The integrity of the isotopic abundances that are ultimately reflected in the fragmentation spectrum may be less than accurate because of imperfections in the fields and functions responsible for discrimination between adjacent masses in the mass-selection stage. Despite less than perfect abundance ratios, the appearance and location of recognizable isotopic signals in a single product-ion MS/MS spectrum can provide valuable insight into fragmentation processes and ion structure. Direct interpretation of a mass spectral fragmentation pattern in the absence of other information about a molecule is challenging. This sometimes laborious process is carried out according to various interpretive rules based on fundamental thermodynamic and kinetic relationships (e.g., quasi-equilibrium162 or RRKM163 theory). The inherent complexity of the processes in combination with the large number of degrees of freedom for average-sized organic molecules makes interpretation or rationalization of anything but the major signals in a spectrum difficult at best. As the population of unique chemical structures continually increases, the number of exceptions to fundamental rules becomes larger and larger making

11 MASS SPECTRAL CHARACTERIZATION

285

interpretation all the more difficult. There has been some recent progress in the development of interpretive software applications based on fundamental rules and empirical observations.164 However, these computerized algorithms cannot interpret mass spectra completely and accurately. This means that many, if not most, fragmentation spectra are interpreted in a manual fashion. The large number of interpretive rules, the significant number of possible exceptions to those rules and the increasing complexity of pharmaceutically relevant molecules has made pattern- or template-based interpretation a widely practiced approach. The tactic employs the parent drug molecule as a template upon which interpretation of unknown structures is based. The identification of compounds from molecular weight information and the fragmentation products observed in the product-ion MS/MS spectrum depends on significant (and characteristic) portions of the parent molecular structure being retained in the impurities, degradants, or metabolites.165–167 Differences in molecular weights of the intact molecules or the m/z values of the fragment ions reflect changes in the substructures leading to a proposed structure for the unknown compound. This paradigm has proven to be highly successful for the identification of drug substance and drug product related substances for both conventional therapeutics168 and biotechnology products (proteins).169-171 From an interpretive perspective, maximum (or complete) information is obtained when the spectrum displays a large number of signals that are distributed at regular intervals across the m/z range. Unfortunately, in many cases the information provided by the mass spectral data is less than complete because of a general lack of interpretable signals in regions of the spectrum that would be indicative of important structural elements. Although the identity or meaning of a particular fragment ion can be quite specific, the information in many ways is similar to that obtained from vibrational spectroscopy in that it indicates the presence of a structural subunit or functional group, but not its specific location within the molecule or its spatial relationship to other components (atoms or other functional groups). For example, a difference in mass of 18 u between the protonated molecule ([MþH]þ) and an adjacent signal in a fragmentation spectrum indicates the presence of an –OH functional group in the molecule. However, the precise nature (e.g., carboxylic acid, alcohol, phenol, etc.) and location of the functional group is not evident from this spectral information. Similarly, signals indicating the loss of 42 u (ketene, CH2CO) from a protonated molecule or the appearance of a complementary ion with m/z 43 are clear indications of an acyl functional group, although the position of the substituent cannot be deduced from this information. The information present in mass spectra takes many forms. Because the unequivocal determination of structure by MS alone is an elusive and sometimes unattainable goal, in many circumstances it can be the more subtle information that contributes to the understanding of a molecule’s structure or identity. In situations where identification is not possible, the mass spectral information can still contribute substantially to the characterization process by providing insight into the points of similarity (or dissimilarity) that the analyte shares with other known compounds or constituents of a given sample. This ability to immediately eliminate a compound (represented by a

286

D. J. BURINSKYAND F. WANG

particular peak in a chromatogram) from further consideration is a critical contributory factor to optimum operational efficiency. Some of the subtler and less quantifiable kernels of information sometimes evident in mass spectra include characteristic isotopic abundances, stable solvation complexes, and multiply charged ions. For example, the presence of chlorine, bromine, sulfur, or other less common elements produces unique and recognizable isotope patterns in a mass spectrum because of abundant naturally occurring isotopes. While the more common elements comprising biologically relevant molecules contain unremarkable isotopic abundances, the relative abundance of the 13C signal (normalized to that of the 12C signal) for the parent ion of a compound provides an approximation of the number of carbon atoms in the molecule. This information can be used to constrain the list of possible elemental formulas generated from an accurate mass measurement or can be correlated with 13C NMR spectroscopic data and knowledge of the molecule’s chemistry to ascertain a structure. In fact, any one of these seemingly benign spectral features could augment other information about the sample (e.g., details about the synthetic scheme, features of the compound’s UV–VIS absorption, IR, Raman, or NMR spectrum, etc.) and result in the determination of compound identity.

VIII. APPLICATIONS The goal of chemical and pharmaceutical development efforts is to establish final synthetic routes and effective formulations as rapidly as possible. These activities provide the materials necessary for safety evaluation (i.e., screening for adverse toxicological effects) and human clinical trials, as well as information leading to large-scale production of commercial medicines. All of the various processes must be optimized by the time the manufacturing batch sizes (typically many kilograms) for the clinical trial supplies are reached. This ensures that the analytical methods used for release testing of manufactured drug substance and drug product are capable of detecting and quantifying known or expected related substances that may be present (i.e., according to approved regulatory specifications). There are numerous examples detailing the contributions of MS to the various processes associated with the development of new pharmaceutical products. Qualitative information about related substances172–183 (see Figure 7) provides a useful baseline against which any future changes in the impurity profile of a product can be compared. The existence of a knowledge base can facilitate the recognition of compounds that may have been encountered under various conditions during synthetic scale-up reactions. In cases where the specific compound has not been encountered, information about structural analogs generally results in rapid identification of the new compound, along with recommendations for the prevention of its formation in the future. The structure elucidation of metabolites in many ways parallels the qualitative analysis efforts that are undertaken for the chemical and pharmaceutical aspects of the drug development process. The determination

11 MASS SPECTRAL CHARACTERIZATION

287

FIGURE 7 Representative data set for the characterization and structure elucidation of low-level related substances in an active pharmaceutical ingredient (API). Top panelMass chromatogram (ESI) representing the 1200 mass spectra (one per second) acquired throughout the HPLC separation. Mass spectra are acquired to determine nominal and/or exact molecular masses183 for all observed analyte ions. Subsequent CID experiments yield fragmentation (product-ion MS/MS) spectra that contain structural information about the ions of interest. Second panelCID product-ion spectrum of related substance A (m/z 132 for [MQH] Q). Third panelCID product-ion spectrum of related substance B (m/z 156 for [MQH]Q). Bottom panelCID product-ion spectrum of the API, peak C (m/z 315 for [MQH]Q).

288

D. J. BURINSKYAND F. WANG

of structure enables a more complete understanding of metabolic pathways, which in turn contributes to knowledge about the potential pharmacological effects of the putative medicine. Realization that many biochemical transformations leave the core structure of the pharmacological agent largely intact has led to widespread use of a rational interpretive philosophy. Once the fragmentation spectra of the parent molecule and the metabolite have been compared, deduction of the site of structural modification can be straightforward.184,185 As with any tandem mass spectrometric analysis, success requires analytes of interest to yield abundant signals (indicative of molecular weight) upon ionization and that these ions can be efficiently (and extensively) dissociated. Spectral interpretation reveals relationships between sites of metabolic modification and systematic fragmentations (producing characteristic ions or neutral losses). Information about metabolic pathways is fed back into the ‘‘design stage’’ therapeutic molecules so that synthetic modifications can be made that will either increase efficacy, decrease toxicity, or prolong the duration of action (by reducing the clearance rate). The characterization of compounds arising from biochemical processes is fraught with many of the same difficulties encountered during the characterization of compounds of chemical or pharmaceutical origin (i.e., related substances emanating from synthetic or formulation activities). However, the relatively limited set of potential biochemical reactions is an advantage in that it restricts the total number of potential structures that might reasonably be encountered. That being said, mass spectral characterization alone is incapable of complete and unequivocal structure determination. Conjugates frequently dissociate by eliminating the endogenous molecule, providing little or no indication as to its site of attachment to the drug molecule. Similarly, primary oxidation products rarely fragment sufficiently to shed any light on the identity of the specific substructure or atom that has undergone the biotransformation. The use of interpretive guides or templates (i.e., the product-ion CID spectra of the parent molecule and other structural analogs) can lead to conclusions regarding the portion of the molecule (e.g., oxidation of an aromatic ring to produce a phenol versus production of an N-oxide or sulfoxide). However, certain metabolic transformations will produce structures that fragment very differently from any of the ‘‘standards.’’ Complementary information from other spectroscopic techniques (primarily NMR spectroscopy) becomes critical for success in such cases.

IX. CONCLUSIONS The characterization (identification or structure elucidation) of chemical compounds at trace levels in complex matrices is a crucial activity during many stages of the pharmaceutical development process. Complete and unequivocal determination of chemical structure is the product of a comprehensive and coordinated process that involves a variety of instrumental techniques and information from various sources (e.g., synthetic scheme, storage conditions, nature, and composition of excipients, published scientific literature, personal experience, etc.). The source, purity, and

11 MASS SPECTRAL CHARACTERIZATION

289

available quantity of sample material are important factors since they influence both data quality and interpretive certainty. Great strides have been made in the past several decades in the development of hybrid techniques that combine chromatography (or other separation techniques) with spectroscopic methods capable of providing insight into molecular structure. These approaches yield valuable information about compounds related to the API, some of which are present in the drug substance or finished product at very low levels. One of the primary tools that has emerged and become a mainstay in this work is HPLC coupled with tandem mass spectrometry (LC/MS/MS). This dynamically linked combination of two powerful analytical techniques was the logical successor to GC/MS and was first highlighted in the 1980s. Since that time it has been adopted in many areas of chemical and biological research as an important means of analyzing (both quantitatively and qualitatively) a wide variety of complex samples not amenable to other techniques. However, hybridization of techniques always comes with compromise. For techniques such as LC/MS or LC/NMR the compromise is acquisition time or less than ideal matrices that degrade (if only slightly) spectral quality. Despite the utility and convenience of these online techniques, they do not eliminate the need for compound isolation. Expertise in separation science (on the preparative, semipreparative, and analytical scales), purification, and sample handling is still a critical component of a successful structural characterization effort. Mass spectral characterization constitutes only a single facet of the pharmaceutical analysis paradigm. The ultimate importance and value of mass spectrometric characterization comes when it is integrated with other information, whether from complementary chemical and physical measurements or from knowledge of and experience with the chemistry at hand. Final confirmation of identity comes only with independent synthesis of the compound suggested by the spectroscopic conclusions. The process of structure elucidation for low-level components of complex mixtures, such as those encountered in the pharmaceutical development process, has been aided tremendously by the availability and utilization of sophisticated instrumentation. However, the endeavor remains largely empirical in nature, requiring experience, dedication, perseverance, and attention to detail.

X. SUMMARY Mass spectrometry has had an increasingly significant impact on the pharmaceutical development process over the past several decades. Advances in the design and efficiency of the interfaces that directly connect separation techniques with mass spectrometers have afforded new opportunities for monitoring, characterizing, and quantifying drug-related substances in APIs and pharmaceutical formulations. Possessing exceptional analytical specificity and sensitivity, MS dramatically reduces the cycle time of chromatographic method development, validation, and sample analysis. The popularity of LC/ MS/MS systems for complex mixture analysis of thermally labile, biologically relevant molecules is largely attributed to the ‘‘soft’’ nature of atmospheric

290

D. J. BURINSKYAND F. WANG

pressure ionization techniques such as electrospray ionization (ESI), atmospheric pressure chemical ionization (APCI) and atmospheric pressure photoionization (APPI). This chapter describes the fundamentals of LC/MSbased techniques for ion structure analysis, including aspects of ion formation in API sources, attributes of various mass analyzers and scan modes used for collision-induced dissociation experiments, and issues surrounding the interpretation of the resulting mass spectra. Although the determination of chemical identity or molecular structure for related substances in pharmaceutical products has continuously benefited from the availability and evolution of modern instrumentation, fundamental knowledge about solution phase chemistry, ionization, and gas-phase processes is still vitally important for achieving success in this endeavor.

ACKNOWLEDGMENTS The authors wish to acknowledge Dr. Kevin Facchine for his encouragement and support of this work.

REFERENCES 1. Gorog, S. Identification and Determination of Impurities in Drugs. Elsevier, Amsterdam, 2000. 2. Norwood, D. L., Prime, D., Downey, B. P., Creasey, J., Sethi, S. K. and Haywood, P. Analysis of polycyclic aromatic hydrocarbons in metered dose inhaler drug formulations by isotope dilution gas chromatography/mass spectrometry. J. Pharm. Biomed. Anal. 13: 293–304, 1995. 3. Paskiet, D. M. Strategy for determining extractables from rubber packaging materials in drug products. PDA J. Pharm. Sci. Technol. 51:248–251. 1997. 4. Tiller, P. R., El Fallah, Z., Wilson, V., Juysman, J. and Patel, D. Qualitative assessment of leachables using data-dependent liquid chromatography/mass spectrometry and liquid chromatography/tandem mass spectrometry. Rapid Commun. Mass Spectrom. 11:1570– 1574, 1997. 5. Wu, S.-L., Wang, Y. J., Hu, J. and Leung, D. The detection of the organic extractables in a biotech product by liquid chromatography online with electrospray mass spectrometry. PDA J. Pharm. Sci. Technol. 51:229–237, 1997. 6. ICH Guideline: Impurities in New Drug Products, Federal Register, 62, No. 96, p. 27454ff. (May 19, 1997). 7. Kapadia, A. J. and Autian, J. The stability of secobarbital sodium solutions. II. Separation and identification of degradation products of secobarbital sodium. J. Am. Pharm. Assoc., Sci. Ed. 49:380–382, 1960. 8. Gruber, M. P., Klein, R. W. and Foxx, M. Infrared and thin-layer chromatography determination of benzoyl peroxide degradation products in pharmaceuticals. J. Pharm. Sci. 58:566–569, 1969. 9. Gyanchandani, N. D., McGilveray, I. J. and Hughes, D. W. Improved thin-layer chromatographic identification of tetracyclines and their degradation products: application to an epimerization study. J. Pharm. Sci. 59:224–228, 1970. 10. Diding, E., Sandstrom, H., Ostelius, J., Karlen, B. and Bogentoft, C. Studies on the application of integrated gas chromatography-mass spectrometry in drug purity control. Impurities in clofibrate, norethindrone, methandienone, haloperidol and meladrazine. Acta Pharm. Suec. 13:55–64, 1976. 11. Johansson, E. and Ryhage, R. Gas chromatographic-mass spectrometric identification and determination of residual by-products in clofibrate preparations. J. Pharm. Pharmacol. 28: 927–929, 1976.

11 MASS SPECTRAL CHARACTERIZATION

291

12. Ogorka, J., Schwinger, G., Bruat, G. and Seidel, V. Online coupled reversed-phase highperformance liquid chromatography-gas chromatography-mass spectrometry. A powerful tool for the identification of unknown impurities in pharmaceutical products. J. Chromatogr. 626:87–96, 1992. 13. Wessels, P., Ogorka, J., Schwinger, G. and Ulmer, M. Elucidation of the structure of drug degradation products by online coupled reversed phase HPLC-GC-MS and online derivatization. J. High Resolut. Chromatogr. 16:708–712, 1993. 14. Karger, B. L., Kirby, D. P., Vouros, P., Foltz, R. L. and Hidy, B. On-line reversed phase liquid chromatography-mass spectrometry. Anal. Chem. 51:2324–2328, 1979. 15. Arpino, P. J., Guiochon, G., Krien, P. and Devant, G. Optimization of the instrumental parameters of a combined liquid chromatograph-mass spectrometer, coupled by an interface for direct liquid introduction I: Performance of the vacuum equipment. J. Chromatogr. 185: 529–547, 1979. 16. Ashcroft, A. E., Chapman, J. R. and Cottrell, J. S. Continuous-flow fast atom bombardment mass spectrometry. J. Chromatogr. 394:15–20, 1987. 17. Voyksner, R. D. and Haney, C. A. Optimization and application of thermospray highperformance liquid chromatography/mass spectrometry. Anal. Chem. 57:991–996, 1985. 18. Whitehouse, C. M., Dreyer, R. N., Yamashita, M. and Fenn, J. B. Electrospray interface for liquid chromatographs and mass spectrometers. Anal. Chem. 57:675–679, 1985. 19. Henion, J. D., Thomson, B. A. and Dawson, P. H. Determination of sulfa drugs in biological fluids by liquid chromatography/mass spectrometry/mass spectrometry. Anal. Chem. 54: 451–456, 1982. 20. Ahuja, S. Impurity analysis of pharmaceutical compounds. Chem. Anal. 85:195–216, 1986. 21. Berridge, J. C. Impurities in drug substances and drug products: new approaches to quantification and qualification. J. Pharm. Biomed. Anal. 14:7–12, 1995. 22. Goosens, E. C., Stegman, K. H., de Jong, D., de Jong, G. J. and Brinkman, U. A. T. Investigation of online reversed-phase liquid chromatography-gas chromatography-mass spectrometry as a tool for the identification of impurities in drug substances. Analyst. 121:61–66, 1996. 23. Gorog, S. Identification, structure elucidation and determination of related organic impurities 2.1. Strategies in impurity profiling. Prog. Pharm. Biomed. Anal. 4:67–83, 2000. 24. Gorog, S. Selected examples (impurity profiling of some groups of drugs) 9.1. Peptides. Prog. Pharm. Biomed. Anal. 4:639–659, 2000. 25. Ermer, J. and Vogel, M. Applications of hyphenated LC-MS techniques in pharmaceutical analysis. Biomed. Chromatogr. 14:373–383, 2000. 26. Gorog, S. The role of the diode-array UV spectra in the identification of impurities. Prog. Pharm. Biomed. Anal. 4:252–265, 2000. 27. Mak, M., Czira, G. and Brlik, J. Mass spectrometry in impurity profiling. Prog. Pharm. Biomed. Anal. 4:97–108, 2000. 28. Tollsten, L. HPLC/MS for drug impurity identification. Prog. Pharm. Biomed. Anal. 4: 266–298, 2000. 29. Wu, Y. The use of liquid chromatography-mass spectrometry for the identification of drug degradation products in pharmaceutical formulations. Biomed. Chromatogr. 14:384–396, 2001. 30. Alsante, K. M., Hatajik, T. D., Lohr, L. L. and Sharp, T. R. Isolation and identification of process related impurities and degradation products from pharmaceutical drug candidates: part I Am. Pharm. Rev. 4:70–78, 2001. 31. Lohr, L. L., Sharp, T. R., Alsante, K. M. and Hatajik, T. D. Isolation and identification of process related impurities and degradation products from pharmaceutical drug candidates: part II Am. Pharm. Rev. 4:104–113, 2001. 32. Reynolds, D. W., Facchine, K. L., Mullaney, J. F., Alsante, K. M., Hatajik, T. D. and Motto, M. G. Conducting forced degradation studies. Pharm. Tech. 26:48–56, 2002. 33. Nicolas, E. C. and Scholz, T. H., Active drug substance impurity profiling. Part I. LC/UV diode array spectral matching J. Pharm. Biomed. Anal. 16:813–824, 1998. 34. Gorog, S. and Herenyi, B. Analysis of steroids. XXXVIII. The use of high-performance liquid chromatography with diode-array UV detection for estimating impurity profiles of steroid drugs. J. Chromatogr. 400:177–186, 1987.

292

D. J. BURINSKYAND F. WANG

35. Czech, M. A., Meltzer, M. and Mouskountakis, J. Determination of Ro 14–4767 (Loceryl) by LC using automated column switching with ultraviolet and electrochemical detection. J. Pharm. Biomed. Anal. 9:1019–1029, 1991. 36. Gorog, S., Bihari, M., Csizer, E., Dravetz, F., Gazdag, M. and Herenyi, B. Estimation of impurity profiles of drugs and related materials. Part 14: the role of HPLC/diode-array UV spectroscopy in the identification of minor components (impurities, degradation products, metabolites) in various matrixes. J. Pharm. Biomed. Anal. 14:85–92, 1995. 37. Mack, G. D. and Ashworth, R. B. A high performance liquid chromatographic system for the analysis of tetracycline drug standards, analogs, degradation products and other impurities. J. Chromatogr. Sci. 16:93–101, 1978. 38. Graham, K. C., LeBelle, M. J. and Wilson, W. L. High performance liquid chromatographic analysis of rifampin and related impurities in pharmaceutical formulations. J. Liq. Chromatogr. 2:365–371, 1979. 39. Mount, D. L., Miles, J. W. and Churchill, F. C. Liquid chromatographic and spectroscopic characterization of pentamidine isethionate and impurities in bulk drug and injectables. J. Assoc. Off. Anal. Chem. 69:624–629, 1986. 40. Cyr, T. D., Matsui, F., Sears, R. W., Curran, N. M. and Lovering, E. G. Liquid chromatographic methods for assay of carbamazepine, 10,11-dihydrocarbamazepine, and related compounds in carbamazepine drug substance and tablets. J. Assoc. Off. Anal. Chem. 70:836–840, 1987. 41. Wannerberg, O. and Persson, B. Liquid chromatographic method for the determination of bambuterol hydrochloride and related compounds. J. Chromatogr. 435:199–203, 1988. 42. Betto, P., Ciranni-Signoretti, E. and Di Fava, R. Determination of cimetidine and related impurities in pharmaceutical formulations by high-performance liquid chromatography. J. Chromatogr. 586:149–152, 1991. 43. Gorski, R. J., Morgan, D. K., Sarocka, C. and Plasz, A. C. Estimation and identification of non-polar compounds in clarithromycin bulk drug by high-performance liquid chromatography. J. Chromatogr. 540:422–429, 1991. 44. Kersten, B. S., Catalano, T. and Rozenman, Y. Ion-pairing high-performance liquid chromatographic method for the determination of 5-aminosalicylic acid and related impurities in bulk chemical. J. Chromatogr. 588:187–193, 1991. 45. Beaulieu, N., Graham, S. J., Sears, R. W. and Lovering, E. G. Liquid chromatographic determination of alprazolam and related impurities in drug substance. J. AOAC Int. 75: 801–804, 1992. 46. Elrod, L. Jr., Golich, T. G., Shaffer, D. I. and Shelat, B. P. Determination of 6,7-difluoro-1(2,4-difluorophenyl)-1,4-dihydro-4-oxo-3-quinilinecarboxylic acid and related compounds by high performance liquid chromatography. J. Liq. Chromatogr. 15:451–466, 1992. 47. Baran, H. and Schwedt, G. Identification of degradation products from epinephrine (adrenaline) in infusion solutions by HPLC and multidetection. Pharmazie. 48:273–275, 1993. 48. Bauer, J. F., Krogh, S. K., Chang, Z. L. and Wong, C. F. Determination of minor impurities in terazosin hydrochloride by high-performance liquid chromatography. J. Chromatogr. 648:175–181, 1993. 49. Priestner, A. A. Use of radiolabeled drug substance to investigate mass balance during validation of a high-performance liquid chromatography method for impurities. Anal. Proc. 30:374–377, 1993. 50. Golich, T. G., Eldrod, L. Jr. and Morley, J. A. Determination of minor impurities and diastereomers of 6-[3-[(2-amino-1-oxopentyl)amino]-1-pyrrolidinyl]-5-fluoro-3-oxo-3Hpyrido [3,2,1-kl]phenoxazine-2-carboxylic acid hydrochloride by high performance liquid chromatography. J. Liq. Chromatogr. 17:2005–2016, 1994. 51. Golich, T. G., Thomsen, J. R., Green, D. L. and Jersild, J. Determination of minor impurities of (R)-1-acetoxymethyl-3-(7-[(1-N,N-dimethylcarbamoyl)-6-(4-fluorophenyl)-indol-3-oyl]-1H, 3H-pyrrolo[1,2-c]thiazol-3-yl)pyridinium chloride by high performance liquid chromatography. J. Liq. Chromatogr. 18:4107–4116, 1995. 52. Orsi, D. D., Gagliardi, L., Chimenti, F. and Tonelli, D. HPLC determination of beclomethasone dipropionate and its degradation products in bulk drug and pharmaceutical formulations. Anal. Lett. 28:1655–1663, 1995.

11 MASS SPECTRAL CHARACTERIZATION

293

53. Stemm, N. L., Skoug, J. W. and Robins, R. H. Gradient high-performance liquid chromatographic assay for degradation products of adinazolam mesylate in a sustained-release tablet formulation. Pharm. Res. 12:738–745, 1995. 54. Bernal, J. L., del Nazal, M. J., Velasco, H. and Toribio, L. HPLC versus SFC for the determination of salbutamol sulfate and its impurities in pharmaceuticals. J. Liq. Chromatogr. Relat. Technol. 19:1579–1589, 1996. 55. Doub, W. H., Ruhl, D. D., Hart, B., Mehelic, P. R. and Revelle, L. K. Gradient liquid chromatographic method for determination of chlorhexidine and its degradation products in bulk material. J. AOAC Int. 79:636–639, 1996. 56. Olori, L., Pierini, N., Spagnoli, M. A. and Tranfo, G. Determination of impurities in buflomedil hydrochloride by HPLC. Boll. Chim. Farm. 135:342–350, 1996. 57. Faulkner, A. J., Wang, Q., DeMontigny, P. and Murphy, J. S. Stability studies and highperformance liquid chromatographic procedures for L-648,548 and its major degradates in an animal health formulation. J. Pharm. Biomed. Anal. 15:523–536, 1997. 58. Gamiz-Gracia, L. and De Castro, M. D. L. An HPLC method for the determination of vitamin B1, caffeine, acetylsalicylic acid, and the impurities of salicylic acid in a pharmaceutical preparation. J. Liq. Chromatogr. Relat. Technol. 20:2123–2133, 1997. 59. Husain, S. and Nageswara R. R. High performance liquid chromatographic monitoring of process impurities in bulk drugs and formulations. Process Control Qual. 10:41–57, 1997. 60. Panderi, I., Archontaki, H., Gikas, E. and Parissi-Poulou, M. Acidic hydrolysis of bromazepam studied by high performance liquid chromatography. Isolation and identification of its degradation products. J. Pharm. Biomed. Anal. 17:327–335, 1998. 61. Al-Kurdi, Z., Al-Jallad, T., Badwan, A. and Jaber, A. M. Y. High performance liquid chromatography method for determination of methyl 5-benzoyl-2-benzimidazolecarbamate (mebendazole) and its main degradation product in pharmaceutical dosage forms. Talanta 50:1089–1097, 1999. 62. Xu, X. and Stewart, J. T. HPLC analysis of haloperidol and its related compound in the drug substance and a tablet dosage form using a non-porous silica octadecylsilane column. J. Liq. Chromatogr. Relat. Technol. 22:1857–1866, 1999. 63. Vaidya, V. V., Khanolkar, M. and Gadre, J. N. Generation of trace impurity profile for tolnaftate by reverse phase high performance liquid chromatography. Indian Drugs 37: 521–523, 2000. 64. Daraghmeh, N., Al-Omari, M., Badwan, A. A. and Jaber, A. M. Y. Determinaton of sildenafil citrate and related substances in the commercial products and tablet dosage form using HPLC. J. Pharm. Biomed. Anal. 25:483–492, 2001. 65. Gjerde, H., Snotun, A., Hem, H., Larsen, K. H. and Dohl, J. Determination of mangafodipir trisodium and related impurities in bulk substance and pharmaceutical formulation by ion-pair high-performance liquid chromatography. J. Pharm. Biomed. Anal. 25:109–114, 2001. 66. Gomes, A. R. and Nagaraju, V. High-performance liquid chromatographic separation and determination of the process related impurities of mebendazole, fenbendazole and albendazole in bulk drugs. J. Pharm. Biomed. Anal. 26:919–927, 2001. 67. Loregian, A., Gatti, R., Palu, G. and De Palo, E. F. Separation methods for acyclovir and related antiviral compounds. J. Chromatogr., B: Biomed. Sci. Appl. 764:289–311, 2001. 68. Montgomery, E. R., Edmanson, A. L., Cook, S. C. and Hovsepian, P. K. Development and validation of a reverse-phase HPLC method for analysis of efavirenz and its related substances in the drug substance and in a capsule formulation. J. Pharm. Biomed. Anal. 25:267–284, 2001. 69. Ogorka, J., Schwinger, G., Bruat, G. and Seidel, V. Online coupled reversed-phase highperformance liquid chromatography-gas chromatography-mass spectrometry. A powerful tool for the identification of unknown impurities in pharmaceutical products. J. Chromatogr. 626:87–96, 1992. 70. Wessels, P., Ogorka, J., Schwinger, G. and Ulmer, M. Elucidation of the structure of drug degradation products by online coupled reversed phase HPLC-GC-MS and online derivatization. J. High Resolut. Chromatogr. 16:708–712, 1993.

294

D. J. BURINSKYAND F. WANG

71. Dette, C., Waetzig, H. and Uhl, H. Characterization of the important pharmaceutical intermediate pyridine-2,4-dicarboxylic acid using capillary electrophoresis (CE). Pharmazie 48:276–280, 1993. 72. Tjoernelund, J. and Hansen, S. H. Determination of impurities in tetracycline hydrochloride by nonaqueous capillary electrophoresis. J. Chromatogr., A 737:291–300, 1997. 73. Hamdan, M. Pharmaceutical applications of liquid chromatography, capillary electrophoresis coupled to mass spectrometry. Process Control Qual. 10:113–127, 1997. 74. Ellis, D. R., Palmer, M. E., Tetler, L. W. and Eckers, C. Separation of cimetidine and related materials by aqueous and non-aqueous capillary electrophoresis. J. Chromatogr., A 808: 269–275, 1998. 75. Yesilada, A., Tozkoparan, B., Gokhan, N., Oner, L. and Ertan, M. Development and validation of a capillary electrophoretic method for the determination of degradation product in naphazoline HCl bulk drug substance. J. Liq. Chromatogr. Relat. Technol. 21: 2575–2588, 1998. 76. Okafo, G., Tolson, D., Monte, S. and Marchbank, J. Analysis of process impurities in the basic drug SB-253149 using capillary electrophoresis and on-line mass spectrometric detection. Rapid Commun. Mass Spectrom. 14:2320–2327, 2000. 77. Shafaati, A. and Clark, B. J. Determination of azathioprine and its related substances by capillary zone electrophoresis and its application to pharmaceutical dosage forms assay. Drug Dev. Ind. Pharm. 26:267–273, 2000. 78. Altria, K. D. High-speed determination of drug related impurities by capillary electrophoresis employing commercial instrumentation. J. Chromatogr. 636:125–132, 1993. 79. Ahrer, W., Schoftner, R. and Buchberger, W. Identification of unknown degradation products in a new cholesterol-reducing drug by ion-chromatography coupled to mass spectrometry. J. Chromatogr., A 912:91–98, 2001. 80. Gruber, M. P., Klein, R. W. and Foxx, M. Infrared and thin-layer chromatography determination of benzoyl peroxide degradation products in pharmaceuticals. J. Pharm. Sci. 58:566–569, 1969. 81. Gyanchandani, N. D., McGilveray, I. J. and Hughes, D. W. Improved thin-layer chromatographic identification of tetracyclines and their degradation products: application to an epimerization study. J. Pharm. Sci. 59:224–228, 1970. 82. Ponder, G. W. and Stewart, J. T. High-performance thin-layer chromatographic determination of digoxin and related compounds, digoxigenin bisdigitoxoside and gitoxin, in digoxin drug substance and tablets. J. Chromatogr. 659:177–183, 1994. 83. Crecelius, A., Clench, M. R., Richards, D. S., Mather, J. and Parr, V. Analysis of UK-224,671 and related substances by thin-layer chromatography-matrix-assisted laser-desorption ionization-time of flight mass spectrometry. J. Planar Chromatogr. Mod. TLC. 13:76–81, 2000. 84. Aboul-Enein, H. Y. and Abu-Zaid, S. Two-dimensional TLC method for identification and quantitative analysis of salbutamol and related impurities in pharmaceutical tablet formulation. Anal. Lett. 34:2099–2110, 2001. 85. Gorog, S., Balogh, G. and Gazdag, M. Analysis of steroids. XLIII. Estimation of impurity profiles of drugs and related materials. Part VIII: combined application of high-performance liquid chromatography and NMR spectroscopy in the impurity profiling of drugs. J. Pharm. Biomed. Anal. 9:829–833 1991. 86. Roberts, J. K. and Smith, R. J. Use of liquid chromatography-nuclear magnetic resonance spectroscopy for the identification of impurities in drug substances. J. Chromatogr., A 677: 385–389, 1994. 87. Crowe, E. A., Roberts, J. K. and Smith, R. J. 1H and 19F LC/NMR: application to the identification of impurities in compounds of pharmaceutical interest. Pharm. Sci. 1:103–105, 1995. 88. Ruckmick, S. C., Marsh, D. F. and Duong, S. T. Synthesis and identification of the primary degradation product in a commercial ophthalmic formulation using NMR, MS, and a stability-indicating HPLC method for antazoline and naphazoline. J. Pharm. Sci. 84: 502–507, 1995. 89. Mistry, N., Ismail, I. M., Smith, M. S., Nicholson, J. K. and Lindon, J. C. Characterization of impurities in bulk drug batches of fluticasone propionate using directly coupled HPLCNMR spectroscopy and HPLC-MS. J. Pharm. Biomed. Anal. 16:697–705, 1997.

11 MASS SPECTRAL CHARACTERIZATION

295

90. Ryan, T. W. Identification of four process-related impurities in the bulk drug butalbital using HPLC-UV photodiode array detection, particle beam MS, and NMR. Anal. Lett. 31:2447–2456, 1998. 91. Peng, S. X., Borah, B., Dobson, R. L. M., Liu, Y. D. and Pikul, S. Application of LC-NMR and LC-MS to the identification of degradation products of a protease inhibitor in dosage formulations. J. Pharm. Biomed. Anal. 20:75–89, 1999. 92. Lehr, G. J., Barry, T. L., Petzinger, G., Hanna, G. M. and Zito, S. W. Isolation and identification of process impurities in trimethoprim drug substance by high-performance liquid chromatography, atmospheric pressure chemical ionization liquid chromatography/ mass spectrometry and nuclear magnetic resonance spectroscopy. J. Pharm. Biomed. Anal. 19:373–389, 1999. 93. El-Anwar and Fawzy, M. Use of HPLC and NMR for the study of cefonicid sodium stability and identification of a degradation product involved in allergic reactions. J. Drug Res. 23:73–77, 2000. 94. Lindon, J. C., Nicholson, J. K. and Wilson, I. D. Directly coupled HPLC-NMR and HPLCNMR-MS in pharmaceutical research and development. J. Chromatogr., B: Biomed. Sci. Appl. 748:233–258, 2000. 95. Hess, S., Teubert, U., Ortwein, J. and Eger, K. Profiling indomethacin impurities using highperformance liquid chromatography and nuclear magnetic resonance. Eur. J. Pharm. Sci. 14:301–311, 2001. 96. Harrison, A.G. Chemical ionization mass spectrometry Adv. Mass Spectrom. 11A:582–595, 1989. 97. Barber, M., Bordoli, R. S., Sedgwick, R. D. and Tyler, A. N. Fast atom bombardment of solids (F.A.B.): a new ion source for mass spectrometry. J. Chem. Soc., Chem. Commun. 325–327, 1981. 98. Kerns, E. H., Volk, K. J., Hill, S. E. and Lee M. S. Profiling new taxanes using LC/MS and LC/MS/MS substructural analysis techniques. J. Rapid Commun. Mass Spectrom. 9:1539–1545, 1995. 99. Volk, K. J., Klohr, S. E., Rourick, R. A., Kerns, E.H. and Lee, M. S. Profiling impurities and degradants of butorphanol tartrate using liquid chromatography/mass spectrometry and liquid chromatography/tandem mass spectrometry substructural techniques. J. Pharm. Biomed. Anal. 14, 1663–1674, 1996. 100. Biemann, K. High resolution mass spectrometry. Advan. Anal. Chem. Instrum. 8:185–221, 1970. 101. Abramson, F. P. Analytical applications of high resolution mass spectrometry. J. Chem. Educ. 49: A283-A293, 1972. 102. Kondrat, R. High resolution mass spectrometry. J. Am. Soc. Mass Spectrom. 10, 661, 1999. 103. Andren, P. E., Emmett, M. R. and Caprioli, R. M. Micro-electrospray: zeptomole/attomole per microliter sensitivity for peptides J. Am. Soc. Mass Spectrom. 5:867–869, 1994. 104. McLuckey, S. A., Cameron, D. and Cooks, R. G. Proton affinities from dissociations of proton-bound dimers. J. Am. Chem. Soc. 103:1313–1317, 1981. 105. Kissinger, P. T. Electrochemical detection in bioanalysis. J. Pharm. Biomed. Anal. 14:871–880, 1996. 106. Anderson, J. K., Hollaway, W. L. and Mole, J. E. Ion-pair reverse-phase HPLC of tryptic peptides from plasma amyloid P-component using a volatile buffer system. J. High Resolut. Chromatogr. Chromatogr. Commun. 4:417–418, 1981. 107. Patthy, M. and Gyenge, R. Perfluorinated acids as ion-pairing agents in the determination of monoamine transmitters and some prominent metabolites in rat brain by highperformance liquid chromatography with amperometric detection. J. Chromatogr. 449: 191–205, 1988. 108. Ayrton, J., Dear, G. J., Leavens, W. J., Mallett, D. N. and Plumb, R. S. Use of generic fast gradient liquid chromatography-tandem mass spectroscopy in quantitative bioanalysis. J. Chromatogr., B: Biomed. Sci. Appl. 709:243–254, 1998. 109. Leroy, F., Presle, B., Verillon, F. and Verette, E. Fast generic-gradient reversed-phase highperformance liquid chromatography using short narrow-bore columns packed with small nonporous silica particles for the analysis of combinatorial libraries. J. Chromatogr. Sci. 39:487–490, 2001.

296

D. J. BURINSKYAND F. WANG

110. Olah, T. V., McLoughlin, D. A. and Gilbert, J. D. The simultaneous determination of mixtures of drug candidates by liquid chromatography/atmospheric pressure chemical ionization mass spectrometry as an in vivo drug screening procedure. Rapid Commun. Mass Spectrom. 11:17–23, 1997. 111. Smith, R. D., Barinaga, C. J. and Udseth, H. R. Improved electrospray ionization interface for capillary zone electrophoresis-mass spectrometry Anal. Chem. 60, 1948–1952, 1988. 112. Moseley, M. A., Deterding, L. J., Tomer, K. B. and Jorgenson, J. W. Coupling of capillary zone electrophoresis and capillary liquid chromatography with coaxial continuous-flow fast atom bombardment tandem sector mass spectrometry. J. Chromatogr. 480, 197–209, 1989. 113. Lee, E. D., Mueck, W., Henion, J. D. and Covey, T. R. On-line capillary zone electrophoresis-ion spray mass spectrometry for the determination of dynorphins J. Chromatogr. 458:313–321, 1988. 114. Reinhoud, N. J., Niessen, W. M. A., Tjaden, U. R., Gramberg, L. G., Verheij, E. R. and van der Greef, J. Performance of a liquid-junction interface for capillary electrophoresis mass spectrometry using continuous-flow fast-atom bombardment Rapid Commun. Mass Spectrom. 3:348–351, 1989. 115. Gale, D. C. and Smith, R. D. Small volume and low flow-rate electrospray ionization mass spectrometry of aqueous samples. Rapid Commun. Mass Spectrom. 7, 1017–1021, 1993. 116. Chowdhury, S. K. and Chait, B. T. Method for the electrospray ionization of highly conductive aqueous solutions. Anal. Chem. 63:1660–1664, 1991. 117. Wahl, J. H. and Smith, R. D., Comparison of buffer systems and interface designs for capillary electrophoresis-mass spectrometry. J. Cap. Elec. 1:62–71, 1994. 118. Pramanik, B. N., Bartner, P. L., Mirza, U. A., Liu, Y.-H. and Ganguly, A. K. Electrospray ionization mass spectrometry for the study of non-covalent complexes: an emerging technology. J. Mass Spectrom. 33:911–920, 1998. 119. De la Mora, J. F., Van Berkel, G. J., Enke, C. G., Cole, R. B., Martinez-Sanchez, M. and Fenn, J. B. Electrochemical processes in electrospray ionization mass spectrometry. J. Mass Spectrom. 35:939–952, 2000. 120. Thurman, E. M., Ferrer, I. and Barcelo, D. Choosing between atmospheric pressure chemical ionization and electrospray ionization for the HPLC/MS analysis of pesticides. Anal. Chem. 73:5441–5449, 2001. 121. Mochida, Y., Yokoyama, Y. and Nakamura, S. Effect of ammonia addition to the mobile phase on atmospheric pressure chemical ionization mass spectrometry of methyl hydroperoxyoleates. J. Mass Spectrom. Soc. Japan 46:246–249, 1998. 122. Tong, W., Chowdhury, S. K., Chen, J. C., Zhong, R., Alton, K. B. and Patrick, J. E. Fragmentation of N-oxides (deoxygenation) in atmospheric pressure ionization: investigation of the activation process. Rapid Comm. Mass Spectrom. 15:2085–2090, 2001. 123. Burinsky, D. J., Williams, J. D., Thornquest, A. D. Jr. and Sides, S. L. Mass spectral fragmentation reactions of a therapeutic 4-azasteroid and related compounds. J. Am. Soc. Mass Spectrom. 12:385–398, 2001. 124. Robb, D. B., Covey, T. R. and Bruins, A. P. Atmospheric pressure photoionization: an ionization method for liquid chromatography-mass spectrometry. Anal. Chem. 72:3653– 3659, 2000. 125. Hanold, K. A., Evans, M. D., Fischer, S. M., Cormia, P. H. and Syage, J. A. Photoionization interface for LC/MS. Abstr. Pap.—Am. Chem. Soc. 221st ANYL-108, 2001. 126. Rauha, J.-P., Vuorela, H. and Kostiainen, R. Effect of eluent on the ionization efficiency of flavinoids by ion spray, atmospheric pressure chemical ionization, and atmospheric pressure photoionization mass spectrometry. J. Mass Spectrom. 36:1269–1280, 2001. 127. Kemp, D. L., Beynon, J. H. and Cooks, R. G. Charge stripping and E/2 mass spectra. Org. Mass Spectrom. 11:857–866, 1976. 128. Loo, J. A., Loo, R. R. O., Udseth, H. R., Edmonds, C. G. and Smith, R. D. Solvent-induced conformational changes of polypeptides probed by electrospray-ionization mass spectrometry. Rapid Commun. Mass Spectrom. 5:101–105, 1991. 129. McLafferty, F. W. (1980). Interpretation of Mass Spectra, 3rd ed., Chapter 4. University Science Books, Mill Valley.

11 MASS SPECTRAL CHARACTERIZATION

297

130. Deterding, L. J. and Gross, M. L. Tandem mass spectrometry for identifying fatty acid derivatives that undergo charge-remote fragmentations. Org. Mass Spectrom. 23:169–177, 1988. 131. Thomson, B. A., Douglas, D. J., Corr, J. J., Hager, J. W. and Jolliffe, C. L. Improved Collisionally Activated Dissociation Efficiency and Mass Resolution on a Triple Quadrupole Mass Spectrometer System. Anal. Chem. 67:1696–1704, 1995. 132. Williams, J. D., Cox, K. A., Cooks, R. G., Kaiser, R. E. Jr. and Schwartz, J. C. High massresolution using a quadrupole ion-trap mass spectrometer. Rapid Commun. Mass Spectrom. 5:327–329, 1991. 133. Jennings, K. R. Collision-induced decompositions of aromatic molecular ions. Int. J. Mass Spectrom. Ion Phys. 1:227–235, 1968. 134. Beynon, J. H., Morgan, R. P. and Brenton, A. G. New methods of identifying organic compounds. Philos. Trans. R. Soc. London, Ser. A 293:157–160, 1979. 135. Boyd, R. K., Porter, C. J. and Beynon, J. H. Linked-scan laws to detect fragmentations in the second field-free region of a double-focussing mass spectrometer. Int. J. Mass Spectrom. Ion Phys. 44:199–214, 1982. 136. Boyd, R. K., Porter, C. J. and Beynon, J. H. A new linked scan for reversed geometry mass spectrometers. Org. Mass Spectrom. 16:490–494, 1981. 137. Kemp, D. L., Cooks, R. G. and Beynon, J. H. Simulated mass-analyzed ion kinetic energy (MIKE) spectra from a conventional double focusing mass spectrometer. Int. J. Mass Spectrom. Ion Phys. 21:93–101, 1976. 138. Yost, R. A. and Enke, C. G. Selected ion fragmentation with a tandem quadrupole mass spectrometer. J. Am. Chem. Soc. 100:2274–2275, 1978. 139. Bean, M. F., Carr, S. A., Thorne, G. C., Reilly, M. H. and Gaskell, S. J. Tandem mass spectrometry of peptides using hybrid and four-sector instruments: a comparative study. Anal. Chem. 63:1473–1481, 1991. 140. Josephs, J. L. Detection and characterization of fumonisin mycotoxins by liquid chromatography/electrospray-ionization using ion trap and triple quadrupole mass spectrometry. Rapid Commun. Mass Spectrom. 10:1333–1344, 1996. 141. Stafford, G. C., Jr., Kelley, P. E., Syka, J. E. P., Reynolds, W. E. and Todd, J. F. J. Recent improvements in and analytical applications of advanced ion trap technology. Int. J. Mass Spectrom. Ion Processes. 60:85–98, 1984. 142. Baldeschwieler, J. D. Ion cyclotron resonance spectroscopy. Science. 159:263–273, 1968. 143. Comisarow, M. B. and Marshall, A. G. Fourier transform ion cyclotron resonance spectroscopy. Chem. Phys. Lett. 25:282–283, 1974. 144. Winger, B. E. and Kemp, C. A. J. Characterization of pharmaceutical compounds and related substances by using HPLC FTICR-MS and tandem mass spectrometry. Am. Pharm. Rev. 4:55–63, 2001. 145. Price, D. and Milnes, G. J. The renaissance of time-of-flight mass spectrometry. Int. J. Mass Spectrom. Ion Processes 99:1–39, 1990. 146. Gohlke, R. S. Time-of-flight mass spectrometry and gas-liquid partition chromatography. Anal. Chem. 31:535–541, 1959. 147. Mamyrin, B. A., Karataev, V. I., Shmikk, D. V. and Zagulin, V. A. Mass reflectron. New nonmagnetic time-of-flight high-resolution mass spectrometer. Zh. Eksp. Teor. Fiz. 64:82–89, 1973. 148. Beynon, J. H. Qualitative analysis of organic compounds by mass spectrometry. Nature 174:735–737, 1954. 149. Beckey, H. D. Field desorption mass spectrometry: a technique for the study of thermally unstable substances of low volatility. Int. J. Mass Spectrom. Ion Phys. 2:500–503, 1969. 150. Macfarlane, R. D. and Torgerson, D. F. Californium-252-plasma desorption time-of-flight mass spectrometry. Int. J. Mass Spectrom. Ion Phys. 21:81–92, 1976. 151. Karas, M., Bachmann, D., Bahr, U. and Hillenkamp, F. Matrix-assisted ultraviolet laser desorption of non-volatile compounds. Int. J. Mass Spectrom. Ion Processes 78:53–68, 1987. 152. McLafferty, F. W., Bente, P. F., III, Kornfeld, R., Tsai, S.-C. and Howe, I. Metastable ion characteristics XXII: Collisional activation spectra of organic ions. J. Amer. Chem. Soc. 95:2120–2129, 1973.

298

D. J. BURINSKYAND F. WANG

153. Mabud, M. A., Dekrey, M. J. and Cooks, R. G. Surface-induced dissociation of molecular ions. Int. J. Mass Spectrom. Ion Processes 67:285–294, 1985. 154. Harris, F. M., Mukhtar, E. S., Griffiths, I. W. and Beynon, J. H. Design of a high-resolution mass spectrometer for studying the photodissociation of organic ions in the gas phase. Proc. R. Soc. London, Ser. A 374:461–473, 1981. 155. Cody, R. B. and Freiser, B. S. Electron impact excitation of ions from organics: an alternative to collision induced dissociation. Anal. Chem. 51:547–551, 1979. 156. Aston, F. W. A positive-ray spectrograph. Phil. Mag. 38:707–715, 1919. 157. Hemberger, P. H., Laramee, J. A., Hubik, A. R. and Cooks, R. G. Angle-resolved mass spectrometry. Target effects upon energy deposition. J. Phys. Chem. 85:2335–2340, 1981. 158. Glish, G. L. and Todd, P. J. Collision region for mass spectrometry/mass spectrometry. Anal. Chem. 54:842–843, 1982. 159. Ast, T., Porter, C. J., Proctor, C. J. and Beynon, J. H. Charge stripping reactions in mass spectrometry: ionization energies of some mono- and disubstituted benzene ions. Glas. Hem. Drus. Beograd 46:135–151, 1981. 160. Schwartz, J. C., Syka, J. E. P. and Jardine, I. High resolution on a quadrupole ion trap mass spectrometer. J. Am. Soc. Mass Spectrom. 2:198–204, 1991. 161. Kaiser, R. E., Jr., Cooks, R. G., Moss, J. and Hemberger, P. H. Mass range extension in a quadrupole ion-trap mass spectrometer. Rapid Commun. Mass Spectrom. 3:50–53, 1989. 162. Rosenstock, H. M., Wallenstein, M. B., Wahrhaftig, A. L. and Eyring, H. Absolute rate theory for isolated systems and the mass spectra of polyatomic molecules. Proc. Natl. Acad. Sci. U.S. 38:667–678, 1952. 163. Lin, S. H., Lau, K. H., Richardson, W., Volk, L. and Eyring, H. Stochastic model of unimolecular reactions and the RRKM [Rice-Ramsperger-Kassel-Marcus] theory. Proc. Nat. Acad. Sci. U.S. 69:2778–2782, 1972. 164. Hart, K. J., Palmer, P. T., Diedrich, D. L. and Enke, C. G. Generation of substructure identification rules using feature-combinations from tandem mass spectra. J. Am. Soc. Mass Spectrom. 3:159–168, 1992. 165. Kerns, E. H., Volk, K. J., Hill, S. E. and Lee, M. S. Profiling new taxanes using LC/MS and LC/MS/MS substructural analysis techniques. Rapid Commun. Mass Spectrom. 9:1539–1545, 1995. 166. Volk, K. J., Klohr, S. E., Rourick, R. A., Kerns, E. H. and Lee M. S. Profiling impurities and degradants of butorphanol tartrate using liquid chromatography/mass spectrometry and liquid chromatography/tandem mass spectrometry substructural techniques. J. Pharm. Biomed. Anal. 14:1663–1674, 1996. 167. Straub, K. M., Rudewicz, P. and Garvie, C. ‘Metabolic mapping’ of drugs: rapid screening techniques for xenobiotic metabolites with MS/MS techniques. Xenobiotica 17:413–422, 1987. 168. Tyczkowska, K. L., Voyksner, R. D. and Aronson, A. L. Solvent degradation of cloxacillin in vitro. Tentative identification of degradation products using thermospray liquid chromatography-mass spectrometry. J. Chromatogr. 594:195–201, 1992. 169. Pflumm, M. N., Gruber, S. C., Tsarbopoulos, A., Wylie, D., Pramanik, B., Bausch, J. N. and Patel, S. T. Isolation and characterization of an acetylated impurity in Escherichia coliderived recombinant human interleukin-10 (IL-10) drug substance. Pharm. Res. 14: 833–836, 1997. 170. DiPaolo, B., Pennetti, A., Nugent, L. and Venkat, K. Monitoring impurities in biopharmaceuticals produced by recombinant technology. Pharm. Sci. Technol. Today 2:70–82, 1999. 171. Oliva, A., Farina, J. and Llabres, M. Development of two high-performance liquid chromatographic methods for the analysis and characterization of insulin and its degradation products in pharmaceutical preparations. J. Chromatogr., B: Biomed. Sci. Appl. 749:25–34, 2000. 172. Eckers, C., Hutton, K. A., de Biasi, V., East, P. B., Haskins, N. J. and Jacewicz, V. W. Determination of clavam-2-carboxylate in clavulanate potassium and tablet material by liquid chromatography-tandem mass spectrometry. J. Chromatogr., A 686:213–218, 1994.

11 MASS SPECTRAL CHARACTERIZATION

299

173. Weber, J. V., Sampino, K., Dunphy, R., Burinsky, D. J., Williams, T. and Motto, M. G. Characterization of cladribine and its related compounds by high-performance liquid chromatography/mass spectrometry. J. Pharm. Sci. 83:523–531, 1994. 174. Iavarone, L., Scandola, M., Pugnaghi, F. and Grossi, P. Qualitative analysis of potential metabolites and degradation products of a new antiinfective drug in rat urine, using HPLC with radiochemical detection and HPLC-mass spectrometry. J. Pharm. Biomed. Anal. 13:607–614, 1995. 175. Burinsky, D. J., Armstrong, B. L., Oyler, A. R. and Dunphy, R. Characterization of tepoxalin and its related compounds by high-performance liquid chromatography/mass spectrometry. J. Pharm. Sci. 85:159–164, 1996. 176. Nicolas, E. C. and Scholz, T. H. Active drug substance impurity profiling. Part II. LC/MS/ MS fingerprinting. J. Pharm. Biomed. Anal. 16:825–836, 1998. 177. Zhang, H., Wang, P., Bartlett, M. G. and Stewart, J. T. HPLC determination of cisatracurium besylate and propofol mixtures with LC-MS identification of degradation products. J. Pharm. Biomed. Anal. 16:1241–1249, 1998. 178. Taylor, S. and Preece, S. The use of LC-API/MS with photodiode array detection for the determination of impurities in drug synthesis. Am. Biotechnol. Lab. 16:29–30, 1998. 179. Zhao, Z., Wang, Q., Tsai, E. W., Qin, X.-Z. and Ip, D. Identification of losartan degradates in stressed tablets by LC-MS and LC-MS/MS. J. Pharm. Biomed. Anal. 20:129–136, 1999. 180. Myung, S.-W., Chang, Y.-J., Min, H.-K., Kim, D.-H., Kim, M., Kang, T., Yoo, E.-A., Sohn, Y. T. and Yim, Y.-H. Characterization of amiodarone metabolites and impurities using liquid chromatography/atmospheric pressure chemical ionization mass spectrometry. Rapid Commun. Mass Spectrom. 14:2046–2054, 2000. 181. Feng, W., Liu, H., Chen, G., Malchow, R., Bennett, F., Lin, E., Pramanik, B. and Chan, T.-M. Structural characterization of the oxidative degradation products of an antifungal agent SCH 56592 by LC-NMR and LC-MS. J. Pharm. Biomed. Anal. 25:545–557, 2001. 182. Franolic, J. D., Lehr, G. J., Barry, T. L. and Petzinger, G. Isolation of a 2:1 hydrochlorothiazide-formaldehyde adduct impurity in hydrochlorothiazide drug substance by preparative chromatography and characterization by electrospray ionization LC-MS. J. Pharm. Biomed. Anal. 26:651–663, 2001. 183. Eckers, C., Haskins, N. and Langridge, J. The use of liquid chromatography combined with a quadrupole time-of-flight analyzer for the identification of trace impurities in drug substance. Rapid Comm. Mass Spectrom. 11:1916–1922, 1997. 184. Unger, S. E. Using mass spectrometry to determine ADME properties in drug discovery. Annu. Rep. Med. Chem. 34:307–316, 1999. 185. Debrauwer, L. Use of LC-MS/MS for xenobiotic metabolism studies in animals. Analusis 28:914–920, 2001.

This Page Intentionally Left Blank

12 NMR CHARACTERIZATION OF IMPURITIES LINDA L. LOHR, ANDREW J. JENSEN, AND THOMAS R. SHARP Pfizer, Inc.,Groton,CT 06340

I. II. III. IV. V.

VI.

VII. VIII.

IX. X.

INTRODUCTION TO NUCLEAR MAGNETIC RESONANCE (NMR) INFORMATION GATHERING SAMPLE PREPARATION FOR NMR SAMPLE PREPARATION FOR LC-NMR NMR INSTRUMENTATION A. General Hardware Considerations B. Probe Specifications NMR EXPERIMENTS A. Getting Started B. One-Dimensional Experiments C. Two-Dimensional Experiments CHOOSING AN EXPERIMENT SET DATA INTERPRETATION A. Primary Interpretation B. Secondary Interpretation C. Tertiary Interpretation: Spectral Simulations FINAL STEPS SUMMARY REFERENCES

I. INTRODUCTION TO NUCLEAR MAGNETIC RESONANCE (NMR) The ability of NMR to provide information regarding the specific bonding structure and stereochemistry within a molecule has created broad applicability across physics, chemistry, biology, and medicine.1–5 NMR provides a powerful analytical tool for structural elucidations. Unfortunately, NMR has traditionally been sensitivity-limited compared with other analytical techniques. Conventional sample requirements for NMR are on the order of 10 mg, as compared with mass spectroscopy, for example, which requires < 1 mg. Therefore, NMR spectroscopy historically has not been the first approach for an analytical chemist when identifying an unknown compound. Technological advancements in the field of magnetic resonance have made significant strides in improving sensitivity levels.6–8 This is particularly important in the structural characterization of drug impurities and

301

302

L.L. LOHR et al.

degradants, which often are available only in extremely limited quantities.9 The nondestructive, noninvasive nature of NMR spectroscopy makes it a valuable tool for the characterization of low-level impurities and degradants. In addition, NMR can be considered close to a ‘‘universal detector’’ for hydrogen and carbon, as well as for other magnetically active nuclei. This may or may not be considered advantageous, since all signals are detected — those arising from the compound of interest as well as all other components in the sample, such as solvent and starting materials. Quantitation is accurate over a dynamic range of nominally four orders of magnitude, although not as precise as other analytical tools particularly at low levels. This poses a challenge for studying mixtures containing low-level components. It is therefore preferred, if possible, to isolate a given impurity prior to study by NMR rather than to analyze a mixture. This also will dramatically simplify spectral interpretation. The following discussion focuses exclusively on what is termed ‘‘small molecules’’ in the industry, namely compounds with a molecular mass on the order of 1000 Da or less. The study of proteins, polymers, and other such ‘‘macromolecules’’ by NMR warrants an entirely different approach that is beyond the scope of this book. Similarly, we will restrict our discussion to liquid-state NMR spectroscopy, since solid-state NMR techniques are discussed elsewhere in this book (see Chapter 3). The general characterization strategy we will take for NMR is shown schematically in Figure 1. The first step in the NMR characterization process is to determine sample purity, solubility, and quantity. This will enable one to assess the feasibility of an NMR evaluation, as well as estimate time frames for data acquisition. If any of these three criteria is low, then data acquisition times will be long, since only a limited amount of target sample will be in the detection region of the NMR spectrometer. Low sample purity will additionally complicate interpretation of the data. Depending on the importance of the structural elucidation, a practical upper limit for acquisition of mass limited NMR data is typically a weekend run. Longer acquisitions on unknown entities may bring sample stability issues into question. Acquisition times should be discussed with the project team to appropriately organize time allocations. If the degradant or impurity is proposed to be a previously characterized species, then simply matching onedimensional proton and carbon spectra to that of the known structure provides sufficient confirmation of structure. Otherwise, a full NMR analysis must be conducted. In this case, one must first decide whether to run tubebased, nonflow NMR on an isolated sample or alternatively run LC-NMR (HPLC-NMR) on a sample mixture. NMR analysis of an isolated sample is typically more straightforward than that of a mixture. Most obviously, there are not competing signals from multiple components that may be difficult to distinguish. In addition, there is probably not an overwhelming solvent signal, which commonly dominates an LC-NMR spectrum, even if solvent suppression techniques are employed. However, LC-NMR of a sample mixture may be the preferred route if the sample is not readily isolatable (see Chapter 13). This may also be the preferred approach if only a quick analysis is required. One may find it valuable, for example, to first perform an

12 NMR CHARACTERIZATION OF IMPURITIES

303

FIGURE 1 Flowchart for structural elucidation of impurities and degradants by NMR.

LC-NMR run and collect a 1-D proton spectrum, and then decide whether to isolate additional sample, based on the results. The next step is to identify an appropriate hardware configuration. For tube-based NMR, one must select the appropriate NMR probe to match the coil size and frequency to the quantity and concentration of sample to be investigated. An appropriate set of experiments must also be determined. For LC-NMR, on the other hand, in addition to selecting an appropriate set of experiments, one must establish optimized chromatographic conditions for maximizing the chromatographic separation and the quantity of sample in the active NMR coil region. Once the data have been collected, one should compare the results to NMR spectra of any relevant standard samples, such as that of the parent or precursors. This will aid in the structural elucidation process, as will other supporting information such as mass spectral results and possible sources of the impurity or degradant. If appropriately selected, the collection of NMR spectra thus obtained should lead to one or at most just a few plausible structures. These then can be presented to the project team to determine synthetic plausibility and the next steps.

304

L.L. LOHR et al.

This outlines a logical approach for making key decisions to move effectively and efficiently through the characterization process. Each of these steps will be discussed at greater length in the sections that follow. Alternative schemes may be employed, such as the order of procession for interpreting the NMR data. However, the overall logic should be the same as that outlined here, particularly regarding instrumentation and experiment selections.

II. INFORMATION GATHERING Arguably the single most important piece of non-NMR data used by the NMR spectroscopist is the total molecular mass. Although not required for structural elucidations, this information greatly facilitates the characterization process by reducing the number of possible structures. Certain mass changes are indicative of specific structural fragments. Furthermore, in many cases fragmentation patterns can identify which portion of the molecule differs from the parent species, thus pinpointing the degradation site. One should be careful, however, to remember the capabilities and limitations of each technique when identifying possible structures. For example, if LC-MS was performed, one should consider the possibility of co-elution of multiple components, especially stereoisomers. All of this information must be balanced with the NMR data when considering plausible structures. Another fruitful source for identifying possible structures comes from knowing the sample’s origin. Was the sample extracted from a mother liquor, isolated from a bulk lot, or synthesized from precursors? It is useful to know both the original synthetic route as well as the isolation pathway followed to prepare the sample. Not only will this provide clues of the possible structure based on predicted chemistry, it will also reveal any potential contaminants in the NMR sample. This will be important if the sample purity is low. Some examples of common contaminant sources include parent compound, precursors, catalysts, stereoisomers, excipients, extractables, and reaction vessel. Often little is known about the inherent properties of a degradant or impurity before it has been fully characterized. If the information is available, there are several properties, that are useful to know for proper handling of the NMR sample. First, it is essential to have an approximate idea regarding the quantity and purity of the sample, since this will determine which NMR hardware is most appropriate. This is commonly estimated using liquid chromatography. We will discuss specific quantity requirements in the ‘‘NMR Instrumentation’’ section, but for now, let’s say that approximately 1 mg of targeted compound is required. Of course, this value is directly related to the sample purity, since 1 mg of sample with a purity level of 50% would contain only 500 mg of targeted compound. For tube-based NMR, one would prefer a purity level of at least 80%, and 90% or greater should be the target for isolation. Sample mixtures containing components in approximately equal quantities are particularly challenging to investigate, since it is usually difficult to discern which peaks correspond to which component. A 40:60 mixture, for instance, is actually preferred over a 50:50 mixture, since it is more straightforward to distinguish spectral contributions from each

12 NMR CHARACTERIZATION OF IMPURITIES

305

component. The quantity of sample observed by NMR also relates to the solubility in the selected solvent. Preliminary solubility tests therefore are important to identify an appropriate NMR solvent and hence maximize sensitivity. Choosing an amenable solvent system may also improve spectral resolution, since proton chemical shifts are solvent-dependent. Solubility in methanol and acetonitrile are often known beforehand based on chromatographic methods used for isolation. Related to purity is salt content. A large amount of salt in the sample, that is several times the amount of degradant, significantly changes the magnetic susceptibility of the sample. In unfavorable cases, this may be beyond the tuning range of the probe and will therefore make it challenging or impossible to tune and match the probe. The effect on the resulting data is lower resolution, reduced sensitivity, and poor coherence selection, which is essential for indirect detection experiments. If information is available regarding stability, it is useful for appropriate sample handling. Possible sources of instability include light, heat, air, and stability in solution. Unfortunately, this information is usually not known a priori. Therefore, one should minimize risk by taking suitable precautions, as described in the sample preparation section. Since this is an unknown research sample, safety hazards are also unknown, so one should proceed with at least the same level of caution that applies for the parent species. The parent is often a useful starting point to assist in the interpretation of the NMR spectra of the unknown degradant or impurity. One-dimensional proton and carbon NMR spectra of the parent species, along with those of any high-level sample contaminants, are valuable references to expedite the data interpretation process. Ideally, because of the solvent dependency of proton chemical shifts, one should acquire reference proton spectra using the same solvent as for the target degradant or impurity. Unfortunately, this may not be possible because of solubility or stability limitations. Two-dimensional proton–carbon correlation spectra may also be necessary to make all resonance assignments. Additional information may come from a variety of sources. Certainly a proposed structure based on plausible chemistry and the total molecular mass is helpful. However, one must be careful to keep in mind that proposed structures are based on preliminary data only and thus may not be consistent with subsequently collected data. If LC-NMR is to be performed, then it is essential to obtain details of the chromatographic method to be used. Other information may include color to suggest conjugation, IR absorption to detect carbonyl stretches, and relative chromatographic retention time to evaluate polarity compared to the parent and other known compounds.

III. SAMPLE PREPARATION FOR NMR Deuterated solvents are used in NMR for two reasons. First, deuterium provides a signal on which to lock the spectrometer frequency, thus enhancing long-term stability. Second, it dramatically reduces the protonsignal that arises from the solvent, which would otherwise overwhelm the signal of the target compound. Solvents should be 99.9þ% deuterated. It is

306

L.L. LOHR et al.

best to use 0.5–1.0 ml ampoules rather than larger solvent bottles. This ensures the integrity of the solvent and avoids contamination during sample preparation. It also prevents absorption of water and deuterium exchange with hydrogen in air over time. Some common deuterated solvents used for NMR studies include dimethyl sulfoxide (DMSO), chloroform, methanol, water, acetone, and acetonitrile. DMSO is considered by many to be a ‘‘universal solvent,’’ since it provides excellent solubility for a wide variety of compounds. In addition, its high boiling point makes it ideal for variable temperature experiments, which are useful for dynamics studies including rotamer investigations. However, because of this, it is nontrivial to remove DMSO from the sample, so it may be difficult to retrieve the sample for subsequent non-NMR tests. It is also extremely viscous, so sample preparation and retrieval using small diameter tubes, including capillaries and submicro tubes, is not easy. One additional point to note is that DMSO absorbs water over time, so careful storage conditions are suggested. Protic solvents such as water and methanol provide proton sources for rapid exchange with labile protons. This makes NMR detection of species such as N–H and O–H protons impossible in these solvents. Also, some compounds, such as peroxides, may dissociate in protic solvents. When using water as the solvent, the pH may need to be adjusted to maximize solubility. Where possible, use deuterated compounds to adjust the pH to minimize undesirable background proton signals. In general, it is recommended to use the same solvent as used for the parent or other proposed impurity source to facilitate subsequent comparison of spectra. As we will see in our discussion of NMR instrumentation, the volume of solvent depends on the specific probe and tube size chosen. In general, one should minimize the total amount of solvent in order to concentrate the sample in the detection region and hence increase the observed signal. Concurrently, adequate solvent must be used so that the sample height is above the top of the coil to ensure proper field homogeneity across the sample. Ideally the sample should be symmetric about the coil region for maximum field homogeneity. A 1:2:1 ratio is a good rule of thumb for volume of sample below, in, and above the active coil region respectively. To minimize the degree of shimming required to achieve good resolution, ideally the sample height should be roughly equal to that of the standard NMR sample used to create the shim file. Otherwise, there may be a significant difference in the shim values necessary to optimize the line shape. A standard proton sample used to measure resolution is 1% chloroform in deuterated acetone. To properly reference the spectra, tetramethylsilane (TMS) is often added as an internal reference standard. The chemical shift for the TMS resonance is set equal to 0.0 ppm for both protons and carbons. One must be extremely careful to add only a minuscule amount of TMS to the sample to avoid overwhelming the target signals. Typically, not more than vapor is added. Alternatively, one can use the solvent signal itself as an internal reference for chemical shift values. Clean glassware is essential for characterization of low-level impurities and degradants. Remember that one is often working with the ‘‘world’s supply’’ of the sample, so it is imperative to avoid contamination. For this

12 NMR CHARACTERIZATION OF IMPURITIES

307

reason, we recommend not reusing 5-mm NMR tubes for this purpose, but rather use a new tube for each impurity or degradant characterization. The justification is that the price of a 5-mm NMR tube is much less than the value of the sample. Similarly, we recommend not reusing capillary or 1.7-mm tubes for the same reason. In addition, these ultra small diameter tubes are challenging to clean thoroughly. The cost of most 2.5–3.0 mm NMR tubes, however, is more substantial, so often these tubes are reused. For any of these sizes, a high quality tube should be selected. This ensures a high degree of tube colinearity and concentricity, which affects observed line shape. Tubes specifically designated by the vendor as disposable usually do not meet these quality standards. Special NMR tubes with plugs is also available that are susceptibility-matched to particular solvents.10 This can increase sensitivity, although the cost of these tubes and plugs is significantly higher than for traditional glass tubes. These tubes are reused. Rinsing tubes with solvent prior to use washes away any residual contaminants that may be present. Similarly, an LC-NMR system, including the probe and any sample storage loops to be used, should be washed for several minutes prior to use. Disposable syringes are available for LC-NMR injections as an alternative to glass syringes, although one must be certain that the sample is inert to the plastic material of the syringe. Unless otherwise known, proceed with a level of safety precaution at least as rigorous as for the parent species. Samples should be provided in the smallest vessel possible, preferably with a conical bottom, to reduce loss. Once the sample has been transferred to the NMR tube or syringe, inspect the tube for any residual solids. Solids decrease both the sensitivity and the resolution achievable and may produce a broad hump in the spectral baseline. Finally, before placing the sample tube into the magnet or injecting into an LC-NMR system, be sure to eliminate any air bubbles, since bubbles will make shimming more challenging and will cause adverse effects in the LC-NMR system. Based on available stability information, keep the sample out of light by covering it completely with aluminum foil when not in the magnet. To avoid thermal instability, do not heat the sample above ambient conditions. Keep the sample capped or sealed to minimize exposure to air. Dissolve the sample in solvent only immediately prior to running NMR experiments to evaluate its stability in solvent. Run a standard proton spectrum immediately after making up the sample as a stability reference. Reacquire the proton spectrum following the final NMR experiment to verify that sample integrity was maintained. If the sample needs to be stored for possible future use, it is recommended that it be stored under refrigerated conditions and protected from exposure to light. Reacquire the proton spectrum if any additional experiments are subsequently required.

IV. SAMPLE PREPARATION FOR LC-NMR As a technique requiring compatibility with both chromatographic and spectroscopic instrumentation, LC-NMR samples require special considerations beyond those described so far. For example, for the

308

L.L. LOHR et al.

FIGURE 2

Calculation to determine the quantity of sample in the active volume of an LC-NMR flow cell. The conditions in this example result in less than the desired amount of sample in the coil, i.e., nominally one microgram. Note that some sample loss throughout the system may be observed. In this case, a smaller amount of sample will be present in the active region of the coil.

applications described here, NMR is not used quantitatively. However, in order to forecast feasibility of an LC-NMR approach, it is beneficial to estimate sample quantity. Depending on the source of the sample and its stability, the sample can be a solid or liquid. A solid sample is convenient since an accurate weight can be obtained. If the sample is a liquid, it can be compared with a known solution to estimate an approximate concentration. Given the sample concentration, the amount of sample in the active volume of the probe can be calculated and then used to determine if LC-NMR will be a viable technique (see Figure 2). One microgram or more is desirable for a reasonable acquisition time. Sample handling requirements and safety considerations are analogous to nonflow NMR. The minimum amount of sample to target is 1 mg in the active volume of the coil. The concentration and conditions normally used for optimization of LC detection are not relevant for LC-NMR analysis since the sensitivity of NMR is much less in comparison. The optimization of LC conditions for NMR compatibility is extremely important. Normally a sample concentration of 0.6 – 2.0 mg/mL is used for LC detection. For LC-NMR analysis, this concentration needs to be as high as possible, potentially as high as 50 mg/mL depending on percent impurity level. Parallel to sample concentration is amount of sample injected. Attempt to inject the maximum amount possible. Optimize resolution and peak shape to maximize the amount of sample in the active volume of the probe. The impurity concentration occasionally can be increased either by using a mother liquor or other fraction of sample that may be enriched. Using other concentrating techniques, such as column trapping11 or solid-phase extraction (SPE),12 can also enrich the amount of impurity introduced to the NMR system. Finally, peak width is optimized to produce a peak that is as narrow as possible yet fills the active volume of the flow cell. This may involve the use of different columns or changing

12 NMR CHARACTERIZATION OF IMPURITIES

309

mobile-phase conditions such that the peak of interest elutes as soon as possible, since peak width is normally narrower for early eluting peaks. A variety of solvents and buffers or modifiers can be used for LC-NMR. However, one must bear in mind that protonated species used in the chromatographic separation will be observed in the NMR spectrum. The exception is when using sample concentrating techniques such as solid-phase extraction. If LC-NMR-MS13 is used, then any required buffers must be compatible with all three techniques. Examples include formate and acetate buffers. If possible, use deuterated solvents and buffers or modifiers to reduce unwanted proton resonances in the NMR spectrum. Deuterium oxide and protonated acetonitrile have historically been used as mobile phase constituents since deuterated acetonitrile is usually much more expensive than LC-NMR–grade protonated acetonitrile. Alternatively, deuterated methanol may be used, but it is even more expensive than deuterated acetonitrile. In addition, methanol is not recommended for LC-NMR applications since it is retained in the tubing and leaches out over an extended period of time. This causes very noticeable peaks in the proton spectrum. Additionally, non-salt buffers are preferred since organic solvents may evaporate over time during analysis, resulting in the buffer’s potentially salting out and clogging the flow probe. Since protonated acetonitrile has a proton resonance that saturates the dynamic range of the NMR receiver, the signal must be reduced by solvent suppression techniques. Examples include presaturation,14 WET,15 and other shaped pulse sequences. Since there is often residual protonated HOD in D2O, this resonance is also reduced by solvent suppression. The primary disadvantage of solvent suppression is that it dramatically distorts the spectrum at the solvent chemical shift, thereby obscuring any target signals in these regions. When setting solvent suppression parameters, it is best to reduce the residual acetonitrile and water resonances to an intensity below the 13C satellite signals of acetonitrile. Another option is to use all deuterated solvents, thereby eliminating the need for solvent suppression. By using deuterated rather than protonated acetonitrile, the capacity of the receiver is effectively increased by a factor of 10. The primary disadvantage of using fully deuterated solvents is cost. One should also be aware of potential retention time shifts because of deuterated solvents. Comparison of retention time and peak area should provide confidence that the peak of interest is tracked properly. If the peak of interest cannot be easily identified using LC-NMR conditions, then LC-NMR-MS can alternatively be used to track peaks.

V. NMR INSTRUMENTATION A. General Hardware Considerations The three main sample factors that determine the appropriate selection of NMR equipment are quantity, solubility, and stability. The quantity of sample determines the detection limits required. Because we are usually

310

L.L. LOHR et al.

sample-limited, it is important to identify the NMR probe that can accommodate the maximum quantity of sample without unnecessarily diluting the sample in excessive solvent. This then relates to the sample’s solubility in the selected solvent. Obviously one should choose a solvent that yields high solubility to maximize the amount of sample in solution. Stability is important since we wish to ensure that the sample does not change over the course of acquiring the data. For unstable compounds, it may be necessary to separate additional degradants from the target species. For this purpose, chromatographic techniques can be combined with traditional NMR techniques.13,16–18 Regardless of individual sample characteristics, there are several general instrumentation guidelines to perform the experiments described here. Mass and solubility limitations, combined with complex mixtures and the need for rapid problem-solving, demand high-end instrumentation, including a highfield spectrometer with a fast CPU, lots of computer memory, and an assortment of probes or a cryoprobe. Antivibration legs reduce spectral noise that arises from floor vibrations. This is essential at high field strengths, 400 MHz and above. Similarly, temperature control minimizes thermal instabilities, which may produce unwanted ridges in 2-D data sets, particularly during long acquisitions. System temperature control is also useful for certain experiments such as the investigation of rotamers. A minimum magnetic field strength of 400 MHz is recommended to achieve the resolution and sensitivity necessary for reliable low-level impurity characterizations. The experiments described here are for protons and carbons, so two radio frequency (RF) channels optimized for these frequencies are required in addition to the deuterium lock channel. One channel is dedicated to protons, while the other (sometimes designated ‘‘X’’) channel covers a broad range of lower frequencies that include carbon. A single Z-axis gradient channel is recommended both for solvent suppression and to significantly improve coherence selection and hence the overall spectral quality of 2-D experiments. A deuterium switch for deuterium gradient shimming is also desirable for fast, reliable shimming. Another desirable capability is LC-NMR, which combines the mixture separation capabilities of liquid chromatography with the structural elucidation capabilities of NMR.13,16,18 In order to perform LC-NMR experiments, one must have both the LC accessories for the NMR console and an LC-NMR probe. The LC accessory typically consists of an LC system with either a variable wavelength UV detector or a diode array detector. It therefore requires a UV active chromophore for the targeted compound to be detected. An optional loop collector allows individual fractions to be stored for future NMR experiments. Individual fractions are sent to the LC-NMR probe, and NMR experiments that employ solvent suppression are performed. Currently, only LC-NMR probes optimized for proton detection are available since the sensitivity prohibits reasonable carbon detection. If using protonated solvents, single or double solvent suppression techniques must be incorporated into the experiments. Gradient capabilities are required for some solvent suppression techniques. Sample sizes range from hundreds of micrograms down to hundreds of nanograms. LC-NMR is suitable for

12 NMR CHARACTERIZATION OF IMPURITIES

311

compounds that are difficult to isolate or are unstable for prolonged periods when isolated. LC-NMR is also useful for initial NMR screening to determine if subsequent sample isolation is necessary. When preparing to purchase an LC-NMR accessory, one must make a number of choices. They include the following questions: Is mixture analysis required, or can samples be isolated for use with microprobes? What sample quantities will be routinely studied? If isolated samples will be available in greater than 10-mg quantities, then tube-based NMR will probably be a more efficient route for structural elucidations. Will fully deuterated solvents be used for the chromatographic separations? Deuterated acetonitrile can be quite costly. Often deuterated water and protonated acetonitrile are thus used to create the mobile phase for LC-NMR. However, this approach produces a sizeable acetonitrile signal in the proton NMR spectrum, so solvent suppression techniques must be employed. Unfortunately, detection of signals in this spectral region is obscured. It is therefore desirable to use deuterated solvents for the structural elucidation of impurities and degradants. Should mass spectrometry (MS) or solid-phase extraction (SPE) capabilities be added? Combining MS with an LC-NMR system provides mass information and facilitates chromatographic peak tracking. An SPE module is used to concentrate the chromatographic peak of interest in the active region of the NMR coil. Alternatively, column trapping techniques can be employed for a similar effect. Are single or triple axis gradients required? Triple axis gradients occasionally provide improved solvent suppression. However, a triple axis gradient probe typically has reduced sensitivity performance compared with an equivalent single axis gradient probe. The HPLC system contains the following components: degasser, binary or quaternary pump, ultraviolet detector, auto or manual injector, column oven, and loop collection (see Figure 3). A degasser is required to purge the system so gases do not cause bubbles in the tubing that can collect on the pump head or at the detector. A pump should be selected appropriate for the HPLC methods used. Signals of protonated solvents, buffers, and modifiers will be observed in the NMR spectrum. The exception to this is when SPE or column trapping is employed. An SPE unit contains cartridges used to concentrate individual sample fractions, then evaporate solvents, and finally elute the fraction off the cartridge with the desired deuterated solvent. This approach provides nontrivial operating cost savings since protonated solvents can be used for the primary chromatographic separation, and then a much smaller amount of deuterated solvent can be used after sample concentration prior to introduction into the NMR probe. The UV detector must satisfy any chromatographic method requirements. Variable wavelength (VW) or diode array detectors (DAD) are commonly employed for this application. There are three main modes of operation for LC-NMR: on-flow, peak collection, and stopped flow. On-flow LC-NMR involves acquisition of realtime NMR scans during the chromatographic run, but requires high sample concentration since only a small number of scans can be averaged for each NMR spectrum. Stopped-flow LC-NMR is normally used when a single peak of interest is to be analyzed. The peak of interest is selected either manually, by region of time, or by intensity threshold. In stopped-flow mode, the LC

312

L.L. LOHR et al.

FIGURE 3

LC-NMR hardware configuration.

pump is physically stopped when the component is in the active volume of the probe. The selected sample fraction is then analyzed by NMR for a duration of time. Peak collection involves an uninterrupted LC run, where the peaks of interest are sent to individual sample storage loops. Then, at the operator’s convenience, each fraction is sent to the NMR probe to be analyzed separately.

B. Probe Specifications Regarding probe specifications, if the console is equipped with gradient, capabilities then the NMR probe requires a corresponding gradient coil to accommodate it. A single z-axis gradient should suffice for these applications. Triple-axis gradients may be used for solvent suppression, although typically at a slight sensitivity loss compared with equivalent single-axis gradient configurations. The strength of the gradient is nominally 60 Gauss/cm. The probe must also have at least two RF channels, one for protons and one for carbons or X nuclei, in addition to the deuterium channel used to lock the signal. Probes are often designed such that one coil is concentrically inside the other. The inner coil yields the highest sensitivity since it is closest to the sample. Therefore, for proton detection experiments, one should choose a probe with the proton coil as the inner coil. This is called a broadband inverse probe, or an H–X probe. Alternatively, for optimal carbon detection, one should designate the X coil to be the inner coil, which is called a

12 NMR CHARACTERIZATION OF IMPURITIES

313

broadband observe probe. If the X channel is limited to carbon detection only, the probe is called a dual probe. When dealing with mass limited samples, carbon detection is often not feasible in a reasonable amount of time. In this case, one should use an inverse probe. If sample quantity and solubility are not limited, then one may decide to spend the extra acquisition time necessary to collect a carbon spectrum. This may be desirable to confirm the number and type of carbons present, especially for quaternary carbons. For these experiments, a broadband or dual probe should be used. The field strength of the probe must match that of the magnet. At least 400 MHz is recommended. Several probe sizes are commercially available, and if one intends to do impurity and degradant characterizations routinely, it is important to have a representative range of sizes. The probe size is expressed in terms of the diameter of the sample tube for which it is designed. A common size is a 5-mm probe. This is an appropriate choice if one is not limited by sample quantity. The sample size is ideally 25–50 mg in 0.5–0.75 mL of solvent, although as little as 1 mg of material can be used. Obviously, increasing the sample concentration will reduce the experiment time required to obtain adequate signal-to-noise ratios. If the amount of sample is limited, a good alternative is the family of ‘‘microprobes.’’ Depending on the vendor, microprobes are available in either 2.5- or 3.0-mm sizes. The sample volume is approximately 200 mL. This is suitable for samples on the order of one milligram down to hundreds of micrograms. For smaller sample quantities, on the order of tens to hundreds of micrograms, a ‘‘submicro-probe’’ is available.6–8 This probe, also known as a ‘‘SMIDGE’’ probe, is an inverse detection probe — in other words, optimized for proton detection. The tube diameter is 1.7 mm and has an active volume of approximately 30 mL. Similarly, 1.0-mm diameter capillary tube probes have been used for even smaller sample volumes.19 It is important to note that decreasing the sample tube size effectively concentrates the sample by reducing the amount of solvent required. Concentrating the sample in the active coil region increases the number of spins detected and thus increases observed signal-to-noise ratios in the NMR spectra. Going to smaller tube sizes, therefore, will yield no significant advantage if there are solubility limitations preventing an increase in concentration. In the case of very poor solubility, the standard 5-mm probe affords the greatest ease of use. LC-NMR requires use of a special probe that contains a flow cell rather than using a conventional tube. Samples flow into the flow cell, halt for extended acquisitions, and then flow out of the probe to either waste or a fraction collector. The LC-NMR probe is interfaced to the HPLC, using standard chromatographic tubing. LC-NMR flow cells are available in various sizes covering the range of 30–240 ml. Cryogen-cooled tube and flow NMR probes offer a significant sensitivity enhancement. In the NMR industry, these are called ‘‘cryoprobes’’12 or ‘‘chiliprobes,’’20,21 depending on the vendor. While the sample remains at ambient conditions, the probe electronics are cooled to cryogenic temperatures to reduce therma noise and, hence, boost observed signal-to-noise

314

L.L. LOHR et al.

ratios. In favorable cases, sensitivity enhancements of up to fourfold have been reported.12, 20 Clearly this is extremely advantageous for the characterization of low-level impurities and degradants. As with conventional NMR probes, cryogenic probes come in several possible configurations. When selecting a probe of this type, one needs to decide on sample size, inverse or dual detection, and tube or flow configurations. The same considerations apply here as for conventional probes. One additional limitation of a cryogenic probe is the potential difficulty in removal from the magnet bore. This is because of the time required for warming up the probe as well as to the sheer weight of the probe. Similarly, when installing the probe, one must also allow sufficient time to install and cool down the probe and establish thermal equilibrium. Additional supporting hardware and software may also be required for cryogenic probe use. For these reasons, it is common to dedicate an NMR spectrometer to cryogenic probe use and not to switch probes as one would do for a conventional configuration.

VI. NMR EXPERIMENTS A. Getting Started Once the sample is loaded into the magnet, set the temperature control and wait for thermal equilibrium to be established. Depending on sample size, approximately 15 minutes should suffice. If the sample does not adequately tune or match, there may be an excessive amount of residual salt present, in which case the sample must be further purified. This may also be exemplified by an unexpected broadening of all observed proton resonances. Spinning the sample improves resolution by reducing field inhomogeneities in the x–y plane. However, instabilities are introduced by the spinning itself, so one should never spin the sample for any long-term experiments, such as nOe measurements or 2-D experiments. The resolution improvement is trivial compared to the line width of carbon signals, so there is no advantage in spinning the sample when acquiring carbon-detected data. The single case, therefore, when one may want to spin the sample is for the simple 1-D proton experiment. However, it is important to remember that spinning sideband signals arise in the proton spectrum at multiples of the spinning speed on either side of each observed resonance.22 The intensity of sideband signals depends on the quality of shimming. Poor shimming gives rise to more intense sideband signals. This may confuse the interpretation of the data, particularly when deciphering low-level contributions to the spectrum. We therefore recommend running all experiments described here without spinning the sample for ease of data interpretation and for optimal stability. Deuterium gradient shimming, or proton gradient shimming if running LC-NMR with protonated solvents, yields reasonable shim values for optimal line shape and resolution. Depending on the amount of available sample, one should allow between 8 and 48 hours to acquire and process a standard set of experiments.

12 NMR CHARACTERIZATION OF IMPURITIES

315

B. One-Dimensional Experiments The most basic NMR experiment is the one-pulse proton experiment.23–25 Proton chemical shifts typically range from 0 to 10 ppm, so the spectral width should be set at least this large. A good approach is to set the spectral width to a larger value, such as 15 ppm, to identify the actual limits of the resonances observed for a given sample. Then the spectral width can be reset to a smaller value specific to the sample. Acquisition parameter values determined for the 1-D proton spectrum can be used as a guideline for other proton-detected experiments, including the proton dimension of twodimensional experiments. Integrating each of the observed proton resonances is useful for subsequent data interpretation. When the proton spectrum is integrated, one can compare integral values of the various resonances to estimate sample purity and determine if further purification is necessary. Ideally, there should be no more than roughly 20% contaminants in the sample. Feasibility of 2-D acquisitions and ‘‘carbon observe’’ experiments can also be assessed, based on the sensitivity and resolution obtained from the proton spectrum. It is good practice to acquire a proton spectrum at the start and end of data collection. By comparing the two resulting spectra, one can confirm if sample integrity was maintained over the course of the experiments. The standard carbon experiment decouples protons from carbons to enhance sensitivity.26 In the absence of proton decoupling, each carbon resonance is split by one more than the number of protons attached to that carbon. Since, unlike for protons, the natural abundance of carbon-13 is only 1.1%, it is critical to maximize sensitivity in order to detect carbon in a reasonable amount of time.27,28 Usually protons are the only nuclei that are decoupled in a standard carbon experiment. Other spin-1/2 nuclei, such as fluorine, are not decoupled and thus split each carbon resonance by one more than the number of attached fluorines (for example, see Figure 4). This can be particularly advantageous for identifying fluorinated carbon entities. In order to simultaneously decouple both protons and fluorines, an additional RF channel is necessary. Since carbon signals are often too weak to observe in a small number of scans, one needs to make a reasonable approximation for the spectral width. A standard carbon spectrum covers the range of 0 to 200 ppm. Because of the relatively long T1 relaxation time of carbons, common practice is to use less than a full 90 pulse, such as a 30 pulse, and a recycle time of 0.2 s. A line broadening of one Hz improves sensitivity without significantly reducing resolution. Acquisition parameter values determined for the 1-D carbon spectrum can be used as a guideline for other carbon experiments as well, including the carbon dimension of 2-D experiments. Because of the inherently poor sensitivity, the acquisition of a carbon spectrum should be considered optional. Deciding whether to collect carbon data depends on the expected sensitivity based on the proton spectrum, the anticipated sample stability over the duration of the experiment, the required degree of carbon resolution necessary, and the importance of directly detecting quaternary carbons. Alternatively, the carbon resonances can be

316

L.L. LOHR et al.

FIGURE 4

One-dimensional carbon spectrum showing splitting due to a tri-fluoro group. The effected resonances appear as quartets due to splitting by three equivalent spin-1/2 nuclei. The size of the coupling constant increases dramatically with proximity to the fluorines.This greatly facilitates the assignment of carbon resonances.

obtained in a fraction of the time indirectly from projections of 2-D heteronuclear spectra. However, 2-D projections exhibit poorer signal-tonoise ratio because of fewer scans, and poor resolution than for direct detection. In addition, some quaternary carbons may not appear in projections because of the presence of neighboring heteroatoms. The DEPT experiment, or distortionless enhanced polarization transfer, is a carbon selectivity experiment.29–35 Based on the pulse length selected, one

12 NMR CHARACTERIZATION OF IMPURITIES

317

can selectively observe different types of carbon entities depending on the number of attached protons. We recommend setting the DEPT proton pulse length to 135 degrees. In this case, quaternary carbons are suppressed, methylenes are inverted, and methine and methyl carbons appear upright in the phased spectrum. Methines and methyls are then distinguished based on chemical shift differences and 2-D proton correlations. Methines usually appear downfield of methyls. If desired, the entire series of DEPT experiments can be performed to conclusively distinguish methine from methyl resonances. Similar experiments to the DEPT include APT36 and PENDANT,37 among others. APT affords the advantage over DEPT of detecting all carbon entities including quaternaries. However, it has lower sensitivity. PENDANT, on the other hand, incorporates both the observation of all carbon entities and an nOe enhancement. However, there is no phase cycle associated with this experiment, and hence phasing can occasionally be problematic. The same considerations apply as for the standard carbon experiment when deciding whether to run one of these optional carbon observe experiments. This information generally can be obtained by close analysis of the HSQC (or equivalent HMQC) spectrum, as we will see in the next section. However, overlapped proton signals may make conclusive assignment of carbon type challenging for some carbon resonances. Alternatively, as discussed in the following section, one can indirectly obtain carbon multiplicity information from a DEPT-HSQC experiment. Although lower in carbon spectral resolution, this approach affords a significant saving in acquisition time, compared with direct carbon detection.

C. Two-Dimensional Experiments 1. Introduction While much information can be gleaned from simple 1-D experiments, a wealth of additional chemical structure information can be obtained from 2-D experiments. The general purpose of expanding to higher dimensions is to establish correlations either between the same nuclear type, known as ‘‘homonuclear’’ experiments, or between different types of nuclei, known as ‘‘heteronuclear’’ experiments. In either case, either short- or long-range interactions can be investigated. In addition to through-bond connectivities, spatial relationships can also be studied. Two-dimensional spectra are commonly displayed as contour plots with either the spectral projections or the corresponding 1-D spectra shown along the axes. For the 2-D experiments described here, the spectral width of the proton dimension should be equal to that determined for the 1-D proton spectrum, and the spectral width of the carbon dimension should be equal to that determined for the 1-D carbon spectrum. If the 1-D carbon spectral width has not yet been determined, then 0 – 200 ppm is a reasonable default range. A good compromise of data set size versus adequate resolution is 2048 points in the t2 dimension and 512 points in the t1 dimension. This data set size is appropriate for all the 2-D experiments described in this section. Linear prediction can be used to enhance the apparent spectral resolution of truncated data.

318

L.L. LOHR et al.

There are several general considerations for identifying appropriate 2-D experiments to use for structural elucidation. First, most of the experiments described here are available both in gradient and nongradient versions. Gradients are used in these experiments to improve coherence selection that is otherwise performed using elaborate phase cycle schemes.38,39 It is therefore possible to achieve improved signal selectivity and reduced spectral artifacts with fewer acquisitions using gradient selection. The primary disadvantage of gradient coherence selection is reduced sensitivity. Therefore, if the low quantity of sample available requires a large number of scans for sensitivity reasons, one may prefer to use nongradient versions of the experiments described and instead rely on phase cycling for signal selectivity. On the topic of sensitivity, the 2-D experiments described here are all proton-detected experiments. The reason is that protons are naturally abundant and have a high gyromagnetic ratio — and hence are easily detected — while 13C nuclei have low natural abundance and gyromagnetic ratio, and thus are more challenging to detect. Experiments that traditionally were based on carbon detection, thus often have modern equivalent versions that transfer the carbon magnetization to protons and detect the corresponding proton signal. This is known as ‘‘inverse detection’’ or ‘‘indirect detection,’’ since the carbon signal is indirectly detected via its effect on the observed proton signal.40–43 Inverse detection experiments are particularly valuable for the characterization of impurities and degradants, since sample quantity is often quite limited. 2. HomonuclearTwo-Dimensional Experiments The experiment traditionally used to detect proton–proton scalar couplings is known as a COSY38,44 for correlation spectroscopy. The 1-D proton spectrum is shown along each of the two axes, and off-diagonal peaks show through-bond correlations between neighboring proton groups. Each unique proton group also exhibits a correlation to itself, which appears along the diagonal axis of the spectrum. These self-correlation peaks do not provide any additional information, and depending on resolution limitations, may obscure cross peaks near the diagonal. It may therefore be advantageous to add a ‘‘double quantum filter’’ to the COSY experiment to reduce contributions along the diagonal. The double quantum filtered COSY, or DQFCOSY,45–47 provides the same information as a standard COSY with improved resolution near the diagonal. COSY-type experiments are useful for showing through-bond connectivity within individual spin systems. However, they cannot provide information to link spin systems separated by nonprotonated groups. COSY-type experiments target geminal or vicinal spin coupling between protons. For longer range proton correlations, the TOCSY is often beneficial.48,49 This twist on the traditional COSY experiment shows through-bond correlations across an entire protonated chain, hence the name ‘‘total correlation spectroscopy.’’ When long spin-lock times are used, spatial correlation artifacts due to ROESY transfers (see ‘‘Additional Experiments’’ section of this chapter) may appear. However, these are easily distinguished, since they are anti-phase to TOCSY correlation peaks. Because of

12 NMR CHARACTERIZATION OF IMPURITIES

319

the additional observed resonances, TOCSY spectra are more complex to interpret than COSY spectra. As we will see in the next section, TOCSY spectra provide information analogous to heteronuclear HMBC spectra. TOCSY has a significant sensitivity advantage over HMBC, since it is a proton–proton rather than a proton–carbon correlation experiment. Therefore, while the HMBC may be easier to interpret, the TOCSY may be the preferred choice for extremely mass-limited samples. This usually would be the case, for example, for LC-NMR applications. 3. Heteronuclear Two-Dimensional Experiments Regarding heteronuclear correlations, HETCOR historically was the standard experiment used to detect proton–carbon scalar couplings.50–52 The disadvantage of HETCOR is that it is a carbon-detection experiment. It is therefore limited in sensitivity as previously discussed. Recalling that the signal-to-noise ratio is proportional only to the square root of the number of acquisitions, it can readily be seen that it becomes unrealistically timeconsuming to attain desired sensitivity levels. For example, a sensitivity loss of a factor of four requires a 16-fold increase in the number of scans to compensate. There are therefore two modern, equivalent experiments commonly used for the indirect detection of short-range proton–carbon scalar couplings. They are heteronuclear single quantum coherence or HSQC,53–55 and heteronuclear multiple quantum coherence or HMQC.56–58 Both experiments show the 1-D proton spectrum along one axis and the 1-D carbon spectrum or carbon projection along the other axis. Cross peaks correlate protons directly bonded to carbons. Two-dimensional correlations can be used to distinguish poorly resolved resonances in either the proton or carbon dimension. This may be the case, for instance, in a complicated aromatic system or for a macrolide. A proton resonance showing no correlations thus must be bonded to a noncarbon entity, e.g., an O–H or N–H group. Conversely, a carbon resonance showing no correlations must be a quaternary carbon. Since quaternary carbons are not observed in this experiment, the carbon spectral width can be set to cover only the protonated carbon region, nominally 0–180 ppm, to improve resolution in this indirect dimension. As previously mentioned, this information is complementary to that obtained from the DEPT spectrum, since one can also match proton integrals to corresponding carbons to distinguish methine, methylene, and methyl carbons. This interpretation may be complicated by the presence of overlapping proton resonances. A version of this experiment, called the DEPT-HSQC (for example, see Figure 5), combines these two experiments into a single proton-detected experiment.59 The cross peaks of the resulting HSQC type experiment are phase-sensitive depending on carbon multiplicity. Methylene-derived cross peaks are negatively phased, while methine- and methyl-derived peaks are phased positively. By incorporating a standard carbon observe DEPT experiment into the proton observe HSQC experiment, a dramatic time saving over running both experiments separately can be achieved without foregoing pertinent structural information. Comparing HSQC and HMQC experiments, the HSQC tends to yield higher resolution,

320

L.L. LOHR et al.

FIGURE 5

DEPT-HSQC spectrum. The four methylene resonances appear as negative correlation peaks, while the methyl resonance in the top right of the spectrum appears as a positive correlation peak.

while the HMQC tends to yield higher sensitivity.60 A 1J(C, H) coupling constant of approximately 145 Hz is recommended for this type of experiment. Long-range correlations can be investigated using the heteronuclear multiple-bond correlation experiment, or HMBC. It detects scalar couplings between protons and carbons that are separated by multiple bonds.43,61–63 We recommend setting the long-range coupling parameter, nJ(C, H), to approximately 10 Hz and the short-range coupling parameter, 1J(C, H) to 145 Hz. The pulse sequence filters out contributions from short-range scalar

12 NMR CHARACTERIZATION OF IMPURITIES

321

couplings. Given these parameter values, for aliphatic systems, peaks are typically observed for protons and carbons separated by two bonds. In aromatic systems, peaks are observed for protons and carbons separated by three bonds. Low-intensity 2J(C, H) and possibly 4J(C, H) correlations may also be observed in highly conjugated systems (see, for example, Figure 6). Because the HMBC experiment detects weak couplings, it is an inherently

FIGURE 6

Comparison of strong and weak long-range scalar correlations in an HMBC spectrum. A strong intensity correlation is observed between carbon 2 and protons 4 and 6, since the parameters are optimized for three-bond separations on an aromatic ring. Only a weak intensity correlation is observed between carbon 2 and protons 1 and 3, since this corresponds to a two-bond separation.

322

L.L. LOHR et al.

insensitive technique compared to the other 2-D experiments described thus far. It is therefore wise to incorporate more scans and allow additional time for acquisition. Occasionally, expected long-range peaks are not observed in the HMBC spectrum. This may be because the chosen coupling values are not optimized for the chemical entity being studied. Unfortunately, for an unknown impurity or degradant being characterized, one usually does not know a priori what the long-range scalar couplings are. To circumvent this obstacle, a class of experiments was created to cover a range of coupling values. The original experiment of this type, called ACCORD-HMBC, splits each HMBC cross peak into its corresponding proton multiplicity.64 Thus for a proton quartet, for example, correlation peaks will arise in sets of four. The advantage of this approach is its ability to cover a range of coupling constants rather than necessitating a single value to be used. Additionally, the more defined peak groupings potentially facilitate spectral interpretation by distinguishing more readily between resonances. The primary disadvantage of this HMBC derivative is its reduced sensitivity compared to the traditional HMBC due to T2 signal loss related to a lengthier acquisition. Apparent signal-to-noise ratio is also diminished, since observed correlation peaks are split into multiplets. As we saw earlier, COSY and TOCSY spectra are used to link proton resonances within a given spin system. However, they are unable to link neighboring spin systems. In this case, the HMBC spectrum is used to observe long-range couplings, particularly across nonprotonated carbons and heteroatoms. HMBC couplings can also be used to confirm correlations within a given spin system. The carbon projection of the HMBC spectrum gives essentially an equivalent to the full carbon spectrum. Clearly this affords a significant sensitivity advantage and time savings to acquiring the standard 1-D carbon spectrum. Limitations include reduced resolution, and the possibility of not observing all carbon resonances. This limitation would arise for neighboring quaternary carbons or other nonprotonated entities since longrange coupling to the nearest proton would be too weak to observe. By interpreting these data interactively with other data such as COSY and HSQC spectra, one can correlate carbons to their attached protons, establish individual spin systems, and then link them together to piece together a plausible chemical structure. 4. Additional Experiments If more information is required, one may need to perform additional experiments. One common example is when information regarding stereochemistry is necessary. Examples include investigation of chiral centers, E versus Z isomers, and overall structural conformation. The 2-D experiments previously described all relate to through-bond scalar correlations and not spatial relationships. One experiment that specifically addresses through-space interactions is known as the nuclear Overhauser effect, or ‘‘nOe,’’ experiment.65–68 The nOe experiment shows correlations between nuclei that are in close proximity to each other. A target signal is irradiated, and this alters the intensity of signals from neighboring nuclei. The resulting

12 NMR CHARACTERIZATION OF IMPURITIES

323

alteration of signal intensity is known as the nuclear Overhauser effect. Taking the difference between the irradiated spectrum and a standard spectrum leaves only the target signal, which is inverted, and the resulting nOe’s. Observations are independent of the number of bonds separating the protons of interest. This experiment requires very high instrument stability, so it is useful to employ temperature control to ensure thermal stability. The nOe interactions are typically very weak, approximately 1 – 3% of the original signal intensity. Therefore, this experiment is commonly limited to proton detection based on sensitivity considerations. For this reason, it is extremely challenging on mass-limited samples and is not considered realistically feasible for LC-NMR. As an added complication, depending on the molecular tumbling rate of the molecule, which relates generally to size, the nOe may be positive, negative, or even zero. Hence the inability to observe an expected nOe does not necessarily mean that two protons are not within close proximity. For small molecules, observed nOe signals are generally positive, while for macromolecules, they are generally negative. To avoid any potential for being in the zero nOe regime, one can use the alternative rotating frame equivalent experiment. In this case, the observed effect is always positive, regardless of molecular tumbling rate. Two-dimensional, homonuclear versions of these experiments are known as NOESY69 and ROESY experiments.70 Other possible 1-D experiments include detection of nitrogen, fluorine, oxygen, and other nuclei of pharmaceutical interest. However, with the exception of fluorine, these active nuclei have very low natural abundance which makes data acquisition prohibitively time-consuming.28 In most cases, impurity and degradant samples are too mass-limited to consider detection of these low abundance nuclei. If pairs of peaks are observed in the 1-D spectra, particularly for protons, this is often indicative of rotamers. Rotamers occur when the rotation about a bond, usually a nitrogen–carbon bond, is slow compared to the NMR time scale. In this case, NMR observes two distinct yet equivalent entities. Elevating the temperature increases the rate of rotation such that, at the coalescence temperature, pairs of peaks coalesce to one, and an average conformation is observed.71 While performing these experiments is not generally limited by sample quantity, sample stability at elevated temperatures should be strongly considered, particularly when dealing with an unknown chemical entity. The solvent’s boiling point should also be taken into consideration. A good choice for elevated temperature studies is DMSO, since its boiling point is quite high. Another useful experiment addresses labile protons, such as N–H and O–H groups. The experiments described thus far detect proton–carbon and proton–proton interactions and do not specifically address protons bonded to other nuclei. As mentioned when discussing NMR solvents, the presence of protic solvents causes labile protons to exchange rapidly, rendering them unobservable by NMR. We can use this to our advantage. A proton spectrum is first acquired using an aprotic solvent. A few drops of deuterium oxide or deuterium chloride are then added to the sample, and the proton spectrum is reacquired. Labile protons will exchange with the D2O and hence no longer be observed in the spectrum.

324

L.L. LOHR et al.

For any of the experiments described thus far, modifications can be made to the individual experiments to adapt to the requirements of a given sample. For example, if a significant amount of residual protonated solvent is present, then solvent suppression techniques can be employed to reduce or eliminate this contribution to the spectrum. Also, shaped pulses can be used to select a desired frequency band within the spectral range.72,73 This enables one to observe only the region of interest and null all other regions. This may be advantageous for LC-NMR, for example, to eliminate the solvent region, or for improved resolution of a target region in the indirect dimension of a 2-D experiment.

VII. CHOOSING AN EXPERIMENT SET With this arsenal of techniques, we are now prepared to obtain sufficient data to begin elucidating a structure. At this point, one must identify an appropriate set of experiments for the sample and issues at hand. The following should be taken simply as guidelines to experiment selection. Alternative approaches may be equally valid. A 1-D proton spectrum should be acquired at the start and end of any investigation. This is used to evaluate sample stability over the course of the experiments. As we have seen, carbondetected experiments should be reserved for those rare instances when sample quantity is not limited and time is not critical. If one is fortunate enough to be in this situation, a suggested set of experiments may include the following: 1-D proton, APT or equivalent for detection of carbon resonances and their corresponding multiplicity, DQFCOSY for establishing spin systems, ACCORD-HMBC for linking spin systems, and HSQC for correlating carbons to their attached protons. Supplemental experiments described in the previous section may also be relevant, depending on the questions to be addressed. This strategy is relevant for sample quantities in excess of approximately 10 mg. If, on the other hand, one has an isolated sample of limited mass and/or a limited amount of time, an alternative approach should be taken. Carbon resonances and their corresponding multiplicity should be detected indirectly to maximize sensitivity and minimize acquisition times. Therefore, instead of acquiring an APT plus an HSQC spectrum, the more time-efficient DEPTHSQC can be used. COSY affords improved sensitivity over DQFCOSY at the cost of resolution along the diagonal. Similarly, ACCORD-HMBC can be replaced with the more sensitive HMBC. This approach is recommended for time-limited situations or for the nominal mass range of 100 mg–10 mg. For yet smaller sample sizes or time frames, a different tactic can be taken. The weak couplings of the HMBC experiment will probably be unattainable in this scenario, so long-range interactions can alternatively be probed with TOCSY. While this provides a much higher level of sensitivity, it will be more challenging to link neighboring spin systems. Therefore, it is possible to acquire a 1-D equivalent to the HMBC spectrum for individually designated resonances. Similarly, if heteronuclear 2-D experiments in general are not

12 NMR CHARACTERIZATION OF IMPURITIES

325

feasible given the amount of sample, then a 1-D version of the HMQC experiment can be employed on targeted resonances. A COSY spectrum will also facilitate the structural elucidation process. Unlike for the previous two cases, nOe measurements will most likely not be within these detection limits. This case applies to samples on the order of 10–100 mg. If one is truly pushing the limits of NMR detection or has only an extremely time-sensitive sample, yet another strategy is preferred. Here a 1-D proton spectrum may be all that is realistically attainable. Possible additional experiments include a 1-D version of TOCSY and a 1-D version of COSY, also known as a homonuclear decoupling experiment. This applies to sample quantities down to hundreds of nanograms, which is currently the practical limit of NMR detection using commercially available equipment. LC-NMR is typically limited to the last two options because of limited sample quantity. Keep in mind that LC-NMR may require incorporating single or double solvent-suppression techniques into these experiments if protonated solvent systems are used. Having selected and acquired a standard set of experiments, the next step is data interpretation as a first pass for structural elucidation.

VIII. DATA INTERPRETATION A. Primary Interpretation Which set of NMR experiments employed will dictate the strategy for interpreting the data? The following discussion is therefore meant as a general guideline. In practice, not all of the spectra mentioned may be available for a given sample. Use the 1-D proton spectrum and either a 1-D carbon spectrum, including any multiplicity-edited version, or the carbon projection of a 2-D heteronuclear experiment as reference spectra for making proton and carbon assignments. Keep in mind that some multiplicity-edited experiments, including DEPT, do not show quaternary carbon resonances. When selecting a carbon projection, it is preferable to take it from a long-range rather than a short-range experiment, since one will be able to detect most, if not all, quaternary carbon signals. If available, begin by identifying the number of protons attached to each carbon. This information can be obtained from several potential sources including DEPT or equivalent, DEPT-HSQC, or by matching proton integrals to carbon resonances in an HSQC or HMQC spectrum. It is useful when interpreting data to label each carbon resonance as quaternary, methine, methylene, or methyl for easy reference. Next, total the number of each type of carbon and compare these numbers to those for the parent or other relevant related structure. Remember that NMR detects only magnetically inequivalent species, so when counting carbons in the parent structure, be sure to count all magnetically equivalent carbons as only one unique carbon entity. The differences in carbon counts between the impurity and the parent will point out any obvious changes from the parent. These numbers may also be compared with

326

L.L. LOHR et al.

the corresponding numbers for any proposed structures, which may rule out some possibilities immediately. A comparison of parent and impurity chemical shifts will also identify any obvious differences. This is true for both proton and carbon resonances. This comparison is especially useful, for example, when determining the oxidation site of an N-oxide or a sulfoxide, since resonances neighboring the oxidation site will be shifted downfield compared to the corresponding peaks in the parent (see Figure 7 for

FIGURE 7

Comparison of carbon spectra for a parent compound and its associated N-oxide.Notice the carbons immediately neighboring the oxidation site are shifted downfield, and carbon 5 is significantly broadened.The upfield shift of carbons 2 and 3 reflects this shift in electron density.

12 NMR CHARACTERIZATION OF IMPURITIES

327

example). When comparing spectra, recall that minor differences may arise if different solvents are used. This is usually insignificant for carbon spectra but is more noticeable for proton spectra. At this point, we can continue by identifying an appropriate starting point for specific assignments. There is inevitably at least one proton or carbon resonance, and usually several, that one can assign immediately. This may be based on chemical shift. For example, a carbonyl should appear as a quaternary significantly downfield in the carbon spectrum or projection, while aliphatic quaternaries appear significantly upfield. Carbon multiplicities that are due to fluorine coupling are also excellent starting identifiers (see Figure 4). Recall that the size of the coupling constant will determine the actual number of bonds separating the carbon and fluorine. In the proton spectrum, a unique peak may be assigned based on chemical shift, multiplicity, and proton integration. If no peaks can be identified using this direct approach, it is necessary to go to the next step by looking at correlations as a unique identifier. For example, there may be two methyl triplets in the proton spectrum, but only one that is expected to neighbor two methines. This can be readily identified using the COSY or DQFCOSY spectrum (see Figure 8). Ideally, a starting point should be identified for each spin system in the molecular structure. It is therefore useful to identify several starting points to facilitate assignments. Once starting points have been identified, one can use 2-D correlations to map out the rest of the structure. First, if starting with a protonated carbon group, use the HSQC or equivalent to identify the carbon resonance directly bonded to the proton group or vice versa. Proton line shape will identify the number of neighboring protons since the observed multiplicity is equal to one more than the number of neighboring protons.74 Labile protons may show only weak couplings, so bear this in mind when evaluating proton multiplicities. Next use a COSY or TOCSY spectrum to identify any immediate protonated neighbors. Recall that TOCSY shows correlations between all protons within a given spin system, while COSY shows only immediate neighbors. Comparing COSY and TOCSY data then distinguishes immediate versus distant neighbors. Alternatively, HMBC-type spectra can be used to identify long-range correlations. This information is invaluable for linking spin systems together, particularly when heteroatoms or quaternary carbons are present. However, it may be challenging to link together molecular fragments separated by more than one heteroatom or quaternary carbon. In this instance, such long-range couplings between carbons and protons are typically too weak to observe. Remember that for two correlated protonated carbons in an HMBC spectrum, a cross peak will exist from the proton of the first to the carbon of the second. An additional cross peak will exist from the carbon of the first to the proton of the second (see Figure 9). This built-in redundancy is particularly useful when limited spectral resolution obscures a correlation peak. Of course, this redundancy is not possible for correlations to quaternary carbons. On the issue of redundancy, once all of the proton and carbon resonances have been assigned, it is a good idea to confirm that all of the observed 2-D correlations are consistent with the proposed assignments. Throughout the initial assignment process, one

328

L.L. LOHR et al.

FIGURE 8

DQFCOSY spectrum of an extended spin system. Methyl 1 correlates to methylene 2, which neighbors methine 3. Methine 3 correlates to methylene 4, which in turn neighbors hydroxyl proton 5.

should refer to an HSQC or HMQC spectrum if available to correlate carbon resonances to their corresponding attached proton resonances. Assignments based on 2-D correlations can often be confirmed by information found in the 1-D spectra. For instance, if two protons neighbor each other as evidenced in the COSY spectrum, their proton resonances should have the same coupling constant. Aromatic ring protons often show an additional hyperfine splitting, which helps to distinguish ring protons from other electron-rich proton resonances. Labile protons found in N–H and

12 NMR CHARACTERIZATION OF IMPURITIES

329

FIGURE 9

This HMBC spectrum shows that for two protonated carbons correlated to each other, a cross peak will exist from the proton of the first to the carbon of the second, and an additional cross peak will exist from the carbon of the first to the proton of the second. Here we see a correlation between carbon 1 and proton 3, and an equivalent correlation between carbon 3 and proton 1.

O–H groups reduce the measured proton integration of these peaks. In addition, these proton resonances are often broader than other resonances in the spectrum (see Figure 10). This may make 2-D correlations to these groups too weak to observe. Summation of proton integrals should be consistent with any proposed structures. When sensitivity-limited, 1-D equivalents to the common 2-D experiments can be employed to selectively probe key correlations. In extreme cases, one may need to rely entirely on

330

L.L. LOHR et al.

FIGURE 10

Labile protons found in N^H and O^H groups reduce the measured proton integration of these peaks. Here we see the integration of the hydroxyl proton is less than the expected value of1.00. Labile protons are also broadened, as seen in the line width of resonance 5 compared to that of 3 and 4.

1-D data for structural elucidation. Any required solvent suppression may further complicate spectral interpretation by dramatically distorting selected proton regions. When all possible proton and carbon assignments have been made, one should have a reasonable picture of the actual structure. Recall that any nuclei other than protons and carbons have not been directly investigated, so keep in mind other possible heteroatoms that may be present. Certainly proton and carbon chemical shifts and correlations (or lack thereof) will provide clues to the presence of other nuclei. Complementary data from other techniques, such as MS, may provide additional insights into the presence of heteroatoms. Also, as previously discussed, NMR is limited to identifying

12 NMR CHARACTERIZATION OF IMPURITIES

331

magnetically inequivalent nuclei.75 Therefore, NMR spectra for a particular molecule often will be indistinguishable from its corresponding dimer, trimer, etc. Possible structures will therefore include these equivalent entities as well. We need to rely on molecular mass and other complementary information to make this distinction. This brings us to the secondary interpretation.

B. Secondary Interpretation As mentioned, for the structural elucidation of impurities and degradants, it is useful to compare the proton and carbon assignments to those of the parent compound. It will be immediately obvious which peaks have changed, disappeared, or are new. This identifies the site of degradation or reactivity. This is a valuable strategy for the determination of oxidation sites for N-oxides and sulfoxides (for instance, see Figure 7). In either case, resonances immediately neighboring the oxidation site should shift significantly downfield to reflect the enrichment of the electronic environment by the addition of oxygen. Also, any protons observed in the parent to be attached to the nitrogen will no longer be observed for the N-oxide. If there is no clear resemblance between the parent and impurity spectra at all, then the impurity may have derived from an alternative source. In this case, one should refer to the list of possible sources previously described to identify a likely candidate. One can then proceed in a similar fashion by comparing the proton and carbon assignments for the impurity to the corresponding chemical shifts for the excipient, precursor, or other identified source. Also bear in mind whether the parent or other source was run as a salt form or a free base, and similarly for the impurity or degradant, since resonances may be observed from the salt. These peaks should be easily identifiable by comparison to proton and carbon chemical shifts for known salts. Any supplemental data such as nOe results or variable temperature studies should also be evaluated at this time. Observation of an nOe between two protons confirms spatial proximity. This information can prove invaluable in distinguishing cis versus trans conformations or other stereoisomers. Coalescence of proton resonances at elevated temperatures is consistent with chemical exchange or the presence of rotamers. Therefore, by acquiring a set of experiments at an elevated temperature, it greatly simplifies the interpretation process for a rotamer system. As we will see subsequently, one may also use spectral simulations to lend additional credibility to a proposed structure.76 However, simulations should not be taken as proof of structure but rather be used to suggest plausible structures.

C. Tertiary Interpretation: Spectral Simulations In certain cases, the experimental NMR data may differ significantly from expected chemical shifts of a proposed structure. When this occurs, one may need to use simulated spectra to provide confidence in the assignments of the experimental chemical shifts. Using the previously obtained MS data, one can propose structures that satisfy the molecular weight requirement and are consistent with the experimental NMR data. These structures are then

332

L.L. LOHR et al.

used to simulate proton and/or carbon spectra. Comparison of the simulated spectrum to the experimental spectrum will hopefully provide a match. If a match is not obtained, then at least the relative chemical environment can be assessed, reported back to the project team, and then used to propose plausible structures related to the original synthesis conditions. These simulations are especially important to assist in the assignment of heteroatom connectivity in the molecule. The simulated spectra may also be used to provide confidence in unexpected chemical shifts. They also prove useful as a comparison with experimental chemical shifts of moieties that have not been previously analyzed. There are two primary methods to calculate simulated spectra, theoretically and empirically. Theoretical simulations establish a molecular model based on quantum mechanical calculations, while empirical simulations rely on a database of experimentally measured chemical shifts and coupling constants for known structures. There are a number of software programs available for either simulation method. Some software packages simulate only 1-D proton and carbon spectra, while others include additional analytical simulations, such as 2-D NMR, HPLC retention times, log P, log D, etc. 1. Empirical Simulations This approach uses a large database of experimental NMR data to simulate 1-D proton and carbon spectra. Databases commonly contain over a million chemical shifts and more than 250,000 coupling constants.77 Some databases can be tailored to specific needs by incorporating the user’s experimental data into the database.77 There are some cautionary notes to these simulation software packages that need to be addressed. One important consideration is that they normally do not accurately simulate tautomers, zwitterions, or charged species. Quadrupolar coupling effects are also difficult to simulate. These are representative instances where one may need to check the references and structures that the software used to simulate the spectra. These cases are a result of situations where too few related examples exist in the database of chemical shifts. Remember that when entering new data into the database, the simulated spectra are only as good as the data in the database. 2. Theoretical Simulations The following discussion focuses on ab initio predictions of 13C chemical shifts.78 Calculations for predictions of proton resonances are possible as well, but are not presented here. In the comparison plots for all compounds, both the x- and y-axes have a range of 200 ppm. The diagonal line drawn from the lower left to upper right of the graph represents a theoretically perfect correlation. The observed experimental shifts, plotted along the x-axis, are the chemical shifts assigned by interpretation of a full NMR spectral data set. Assignments are based on the expected chemical shift ranges for a given type of carbon, experimentally determined proton multiplicities, both homonuclear and heteronuclear correlation experiments, and whatever additional experiments were necessary to assign a spectral line to a particular atom. Calculated chemical shifts are plotted along the y-axis. The line drawn

12 NMR CHARACTERIZATION OF IMPURITIES

333

through the data points is the best linear fit of the data. The units generated by the DFT DGAUSS program have been converted to conventional ppm units by comparison with computations for TMS, used as a reference compound, computed using the same DFT program parameters. A theoretical approach to spectral simulations can, in many cases, provide an extremely accurate match to experimental values. Deviations may arise in the case of charged species, differing stereoisomers or conformations, and potentially the presence of atypical heteroatoms. Bearing these precautions in mind, simulations can provide valuable insights into plausibility of proposed structures. The 13C NMR chemical shift predictions for sertraline (Structure 1) are plotted in Figure 11. The correlation is quite high. The greatest deviations between observed and predicted chemical shifts are for the N-methyl carbon and for the two aromatic ring carbons to which chlorine atoms are attached. Computations for ziprasidone (Structure 2) are shown in Figure 12. A large variety of chemical shifts are present in this molecule. Nonetheless, the predictions are generally good as seen by the r2 value of 0.9899.

STRUCTURE 1 Sertraline

FIGURE 11 Observed versus predicted 13C chemical shifts for sertraline.

334

L.L. LOHR et al.

STRUCTURE 2 Ziprasidone

FIGURE 12

Observed versus expected shift for ziprasidone.

One notable exception is for a quaternary carbon in the five-membered portion of the benzisothiazole ring. These computations took approximately 5 hours to complete. The trovafloxacin (Structure 3) computed chemical shifts are plotted in Figure 13. The three aberrant chemical shifts at approximately 148 ppm are the fluorinated carbons. Computing time for trovafloxacin was 31.8 hours. Figure 14 is a composite comparison of observed versus expected shift for 12 compounds.

IX. FINAL STEPS After the required levels of NMR data interpretation have been completed, it is advantageous to reconvene with the characterization team to discuss the structural information obtained thus far. Specifically, the group should identify areas of consistency or inconsistency of information gathered from other sources compared with the NMR conclusions. For example, are the mass spectral fragment weights consistent with the molecular fragments determined by NMR? Is the degradation site or other deviation from known compounds identified by NMR consistent with the location identified by mass spectral fragmentation patterns? Is the total molecular mass consistent with

12 NMR CHARACTERIZATION OF IMPURITIES

335

STRUCTURE 3 Trovafloxacin

FIGURE 13

FIGURE 14

Observed versus expected shift for trovafloxacin.

Composite comparison of observed versus expected shift for 12 compounds, including data for sertraline, ziprasidone, and trovafloxacin.The three aberrant carbon shifts from trovafloxacin, previously mentioned, are identifiable in this composite plot.

336

L.L. LOHR et al.

structures based on the NMR results? The characterization strategy outlined here does not explicitly detect nuclei other than carbon and hydrogen, and as previously mentioned, NMR cannot distinguish magnetically equivalent nuclei. Consistency between NMR and MS data is thus especially important for distinguishing monomers from dimers, trimers, etc. If a carbonyl was observed in the IR spectrum, was it also observed either directly or indirectly in the NMR data? Is the expected polarity of a proposed structure consistent with the observed chromatographic retention time? If any such inconsistencies are identified, focus on these areas to propose alternative structures. Even if a structure is identified to be consistent with the data, it is important to consider if any other structures may be consistent as well. Clearly this approach requires a high level of interaction among the members of the characterization team. If a structure has passed all the requirements thus far, one needs to determine if it can be derived from a plausible synthetic route. In other words, is it consistent with the reaction pathway and chemical entities present in the reaction vessel? If additional structural information is required, it is necessary at this point to identify appropriate follow-up experiments. For example, if the stereochemistry at a chiral center is critical to the understanding of the system, nOe experiments may be useful to investigate spatial relationships. To improve the confidence level in a proposed structure, it may be necessary to synthesize the proposed structure using a wellestablished synthetic route and then repeat the NMR characterization on the synthesized compound. Comparing the proton and carbon spectra of the unknown with the known compound should confirm consistency of structure. Finally, the team should draw conclusions based on the identified structure, plausible synthetic routes, and ways to prevent its formation.

X. SUMMARY As one can see from this discussion, the NMR characterization process plays a valuable role in the structural elucidation of impurities and degradants. A rigorous decision process enables a logical selection from the many choices of instrumentation and experiments available. The approach outlined in this chapter minimizes data-acquisition time while maximizing relevant information gathered. This becomes critical for rapid characterizations, either because of aggressive project time lines or sample instability. Mass limited samples, as well as low-solubility or low-purity samples provide additional challenges that can be appropriately addressed through prudent hardware and experiment selection. Interpretation of the NMR data requires careful diligence, usually involving an iterative process of studying several spectra simultaneously. This sometimes time-consuming effort is well worth the investment. NMR spectroscopy yields critical structural information that cannot be obtained readily, and in many cases not at all, by any other means. The results of the NMR analysis must be carefully weighed with observations from other analyses, such as MS and HPLC, to ensure a thorough understanding of the

12 NMR CHARACTERIZATION OF IMPURITIES

337

chemical system. NMR spectroscopy is often applied last in the characterization process, and because of the level of structural detail achievable by NMR, it often provides the missing information that ties all other experimental observations together. It is this philosophy of interactively collecting and interpreting complementary information that enables the successful characterization of the most challenging impurities and degradants, a requirement that continues to grow ever more stringent with evolving FDA guidelines.

REFERENCES 1. Derome, A. E. Modern NMR Techniques for Chemistry Research. Pergamon Press, Oxford, UK, 1987. 2. Ernst, R. R. Principles of Nuclear Magnetic Resonance in One and Two Dimensions. Clarendon Press, Oxford, UK, 1987. 3. Abragam, A. Principles of Nuclear Magnetism. Clarendon Press, Oxford, UK, 1989. 4. Slichter, C. P. Principles of Magnetic Resonance, 3rd ed. Springer–Verlag, Berlin, 1990. 5. Harris, R. K. Nuclear Magnetic Resonance Spectroscopy. Longman Scientific & Technical, NY, 1990. 6. Martin, G. E., Crouch, R. C. and Zens, A. P. Magn. Reson. Chem. 36:551–557, 1998. 7. Martin, G. E., Guido, J. E., Robins, R. H., Sharaf, M. H. M., Schiff, P. L. Jr. and Tackie, A. N. J. Natural Products. 61:555–559, 1998. 8. Martin, G. E., Hadden, C. E., Tackie, A. N., Sharaf, M. H. M. and Schiff, P. L. Jr. Magn. Reson. Chem. 37:529–537, 1999. 9. Alsante, K., Friedmann, R., Hatajik, T., Lohr, L., Sharp, T., Snyder, K. and Szczesny, E. Degradation and Impurity Analysis for Pharmaceutical Drug Candidates. In Handbook of Modern Pharmaceutical Analysis, Ahuja, S. and Scypinski, S. (Eds.), Academic Press, San Diego, 2001. 10. Available for example, from Shigemi, Inc., Allison Park, PA; www.geocities.com/shigemi. 11. Kaerner, A. and Jackson, D. Column Trapping LC-NMR-MS, presented at the 43rd ENC Conference, Pacific Grove, CA, 2002. 12. Spraul, M. Recent Advances in LC-SPE-NMR-MS and Cryoprobe Technology, presented at the 2002 Bruker-Biospin Users’ Conference, Pacific Grove, CA, 2002. 13. Kaerner, A., Dow, L., Jackson, D. and Borel, A. Recent Development and Applications of HPLC-NMR for Structural Elucidation, presented at the 41st ENC Conference, Pacific Grove, CA, 2000. 14. Gueron, M., Plateu, P. and Decorps, M. Prog. NMR Spectrosc. 23:135–209, 1991. 15. Smallcombe, S., Patt, S. and Keifer, P. J. Magn. Reson., Ser. A 117(2):295–303, 1995. 16. Kaerner, A. and Maple, S. The Development and Applications of Exact Mass HPCL-NMRMS, presented at the 41st ENC Conference, Pacific Grove, CA, 2000. 17. Farrant, R. D., Mills, G. and Walker, V. Identification of the Species Produced by the Reaction of P5C and Puridoxal Phosphate Using LC-NMR-MS, presented at the 41st ENC Conference, Pacific Grove, CA, 2000. 18. Olson, D. L. Trace Analysis Using Capillary HPLC-NMR, presented at the 2nd SMASH NMR Conference, Argonne, IL, 2000. 19. Available for example, from Protasis/MRM Corp., Savoy, IL; www.protasis.com/MRM. 20. Crouch, R. Chili Probe—An Applications View, presented at the 2002 Varian Users Conference, Palo Alto, CA, 2002. 21. Avizonis, D. Cryo Flow—Why Not? Presented at the 2002 Varian Users Conference, Palo Alto, CA, 2002. 22. Derome, A. E. Modern NMR Techniques for Chemistry Research. Pergamon Press, Oxford, UK, 41, 1987. 23. Sanders, K. K. and Hunter, B. K. Modern NMR Chemistry, 2nd ed. Oxford University Press, Oxford, UK, 1993.

338

L.L. LOHR et al.

24. Friebolin, H. P. Basic One- and Two-Dimensional NMR Spectroscopy, 2nd ed. VCH, Weinheim, 1993. 25. Braun, S., Kalinowski, H.-O. and Berger, S. 150 and More Basic NMR Experiments. Wiley, VCH, Weinheim, 45–49, 1998. 26. Kalinowski, H.-O., Berger, S. and Braun, S. Carbon-13 NMR Spectroscopy, Wiley, Chichester, 1988. 27. Braun, S., Kalinowski H.-O. and Berger, S. 150 and More Basic NMR Experiments, Wiley, VCH, Weinheim, 50–54, 1998. 28. Harris, R. K. Nuclear Magnetic Resonance Spectroscopy, Longman Scientific & Technical, NY, 236, 1990. 29. Bendall, M. R., Pegg, D. T., Doddrell, D. M. and Field, J. J. Mag. Reson. 520–526, 1983. 30. Bendall, M. R. and Pegg, D. T. J. Mag. Reson. 53:40–48, 1983. 31. Bendall, M. R. and Pegg, D. T. J. Mag. Reson. 53:272–296, 1983. 32. Derome, A. E. Modern NMR Techniques for Chemistry Research, Pergamon Press, Oxford, UK, 143–151, 1987. 33. Schenker, K. V. and Philipsborn, W. V. J. Mag. Reson. 66:219–229, 1986. 34. Doddrell, D. M., Pegg, D. T. and Bendall, M. R. J. Mag. Reson. 48:323–327, 1982. 35. Braun, S., Kalinowski, H.-O. and Berger, S. 150 and More Basic NMR Experiments. Weinheim: Wiley, VCH, 180–182, 1998. 36. Patt, S. and Shoolery, J. J. Magn. Reson. 46:535–539, 1982. 37. Homer, J. and Perry, M. J. Chem. Soc. Chem. Commun. 373–374, 1994. 38. Hurd, R. E. J. Mag. Reson. 87:422–428, 1990. 39. Tolman, J. R., Chung, J. and Prestegard, J. H. J. Mag. Reson. 98:462–467, 1992. 40. Tyburn, J. M., Brereton, I. M. and Doddrell, D. M. J. Mag. Reson. 97:305, 1991. 41. Vuister, G. W., Boelens, R., Kaptean, R., Hurd, R. E., John, B. K. and van Zijl, P. C. M. J. Am. Chem. Soc. 113:9688, 1991. 42. Hurd, R. E. and John, B. K. J. Mag. Reson. 91:648, 1991. 43. Willker, W., Leibfritz, D., Kerschebaum, R. and Bermel, W. J. Mag. Reson. 31:287–292, 1993. 44. Jeener, J. Ampere International Summer School, Proposal. Basko Polje, 1971. 45. Piantini, U., Sorensen, O. W. and Ernst, R. R. J. Am. Chem. Soc. 104:6800–6801, 1982. 46. Brereton, I. M., Crozier, S., Field, J. and Doddrell, D. M. J. Magn. Reson. 93:54–62, 1991. 47. Braun, S., Kalinowski, H.-O. and Berger, S. 150 and More Basic NMR Experiments. Wiley, VCH, Weinheim, 168–371, 1998. 48. Braunschweiler, L. and Ernst, R. J. Magn. Reson. 53:521–528, 1983. 49. Kover, K., Uhrin, D. and Hruby, V. J. Magn. Reson. 130:162–168, 1998. 50. Martin, G. E. and Zektzer, A. S. Two-dimensional NMR Methods for Establishing Molecular Connectivity, VCH, Weinheim, C. 3, 1988. 51. Bax, A. and Morris, G. A. J. Mag. Reson. 42:501–505, 1981. 52. Freeman, R. and Morris, G. A. J. Chem. Soc. Chem. Commun. 684–686, 1978. 53. Bodenhausen, G. and Ruben, D. J. Chem. Phys. Lett. 69:185–188, 1980. 54. Braun, S., Kalinowski, H.-O. and Berger, S. 150 and More Basic NMR Experiments. Wiley, VCH, Weinheim, 397–400, 1998. 55. Kay, L. E., Keifer, P. and Saarinen, T. J. Am. Chem. Soc. 114:10663–10665, 1992. 56. Muller, L. J. Am. Chem. Soc. 101:4481–4484, 1979. 57. Braun, S., Kalinowski, H.-O. and Berger, S. 150 and More Basic NMR Experiments, Wiley, VCH, Weinheim, 384–387, 1998. 58. Hurd, R. E. and John, B. K. J. Magn. Reson. 91:648–653, 1991. 59. Willker, W., Leibfritz, D., Kerssebaum, R. and Bermel, W. Magn. Reson. Chem. 31:287–292, 1993. 60. Cavanagh, J., Fairbrother, W., Palmer, A. and Skelton, N. Heteronuclear NMR Experiments from Protein NMR Spectroscopy Principles and Practice, Academic Press, San Diego, 1996. 61. Bax, A. and Summers, M.F. J. Am. Chem. Soc. 108:2093–2094, 1986. 62. Ruiz-Cabello, J., Vuister, G. W., Moonen, C. T., Van Gelderen, P., Cohen, J. S. and Van Zijl, P. C. M. J. Mag. Reson. 100:282–303, 1992. 63. Martin, G. E. and Zektzer, A. S. Two-Dimensional NMR Methods for Establishing Molecular Connectivity. VCH, Weinheim, 267–273, 1988.

12 NMR CHARACTERIZATION OF IMPURITIES

339

64. Kogler, H., Sorensen, W., Bodenhausen, G. and Ernst, R. J. Magn. Reson. 55:157–163, 1983. 65. Kinns, M. and Sanders, J. K. M. J. Mag. Reson. 56:518–520, 1984. 66. Neuhaus, D. and Williamson, M. The Nuclear Overhauser Effect. VCH, Weinheim, 1989. 67. Derome, A. E. Modern NMR Techniques for Chemistry Research. Pergamon Press, Oxford, 97–127, 1987. 68. Noggle, J. H. and Schirmer, R. E. The Nuclear Overhauser Effect—Chemical Applications. Academic Press, San Diego, 1972. 69. Jeener, J., Meier, B., Bachmann, P. and Ernst, R. J. Chem. Phys. 71:4546–4563, 1979. 70. Bothner-By, A., Stephens, R., Lee, J.-M., Warren, C. and Jeanloz, R. J. Am. Chem. Soc. 106: 811–813, 1984. 71. Harris, R. K. Nuclear Magnetic Resonance Spectroscopy. Longman Scientific & Technical, NY, 119–131, 1990. 72. Warren, W. and Silver, M. Adv. Magn. Reson. 12:247–384, 1988. 73. Freedman, R. Prog. NMR Spectrosc. 32, 59–106, 1998. 74. Harris, R. K. Nuclear Magnetic Resonance Spectroscopy. Longman Scientific & Technical, NY, 26, 1990. 75. Abragam, A. Principles of Nuclear Magnetism. Clarendon Press, Oxford, UK, 480–481, 1989. 76. Williams, T., Blinov, K., Elyashberg, M. and Martisrosian, E. Recent Advances in the Use of NMR Prediction and Automated Structure Elucidation Software for Natural Product Characterization, presented at the SMASH NMR Conference. Argonne, IL, 2000. 77. Williams, A. Current Opinion in Drug Discovery & Development, 3(3):298–305, 2000. 78. Koch, W. and Holthausen, M. A Chemist’s Guide to Density Functional Theory, 2nd ed. Wiley, VCH, Weinheim, 2001.

This Page Intentionally Left Blank

13 HYPHENATED CHARACTERIZATION TECHNIQUES THOMAS N. FEINBERG Cardinal Health, ResearchTriangle Park, NC 27709

I. INTRODUCTION II. EXPERIMENTAL EXAMPLES A. Absorption Spectroscopy B. Mass Spectrometry C. Nuclear Magnetic Resonance (NMR) Spectroscopy III. CONCLUSIONS REFERENCES

I. INTRODUCTION The pharmaceutical industry is required by the federal Food, Drug, and Cosmetic Act1 to establish the identity and purity of the products it wishes to market in the United States. The most common techniques used for these purposes are chromatography based, coupled with detectors whose output varies with either the total amount of an eluted species (mass) or the concentration of the eluted species in the flowing stream. For detectors to provide useful identity or purity measurements, the characteristic time constant of the detector signal must be less than the peak width. These basic techniques can be considered single-dimensional, as the detector varies only in time, and the resultant data is a two-dimensional plot (i.e., a chromatogram). The two most common chromatographic techniques, high-performance liquid chromatography (HPLC) and gas chromatography (GC) can provide both identity measurements by retention time and purity measurements by area (or area under the curve, AUC) when coupled with the two most widely used detectors, ultraviolet (UV) absorbance and flame ionization detection, respectively. Retention time matching can be validated to establish identity, but is insufficiently specific for general use. Retention times are not necessarily unique for chemically distinct species; even isomers may or may not co-elute. A change in synthetic process may introduce a different impurity profile; any of these new process impurities has a non-zero chance of co-elution with

341

342

T. N. FEINBERG

another impurity or the desired product. A change in storage condition can also invalidate a previously used retention time identity test as different degradation pathways (heat, light, etc.) may have different intermediates and/ or products. It is possible to determine, a priori, the probability of species co-elution in a chromatogram, using some basic relationships from chromatography. It is easiest to show how this probability can be calculated using the case of isocratic or isothermal chromatography, although it can be generalized by appropriate mapping techniques.2–4 Consider the basic relationship between retention time (t), peak width (w), and chromatographic plates (N) for an isocratic (isothermal) separation:5 N ¼ 16 

h t i2 w

ð1Þ

Over an observation window from tf to ti, the average peak width waverage for peaks with similar interaction with the column (constant N), is given by waverage ¼

2  ðtf þ ti Þ pffiffiffiffiffi N

ð2Þ

Dividing the observation time (tf  ti) by the average peak width, the total number of separated peaks (peak capacity, nc) can be deduced as follows: pffiffiffiffiffi ðtf  ti Þ ðtf  ti Þ  N ¼ nc ¼ waverage 2  ðtf þ ti Þ

ð3Þ

Given an arbitrarily long final observation time (limit of nc as tf approaches infinity), the total peak capacity of a separation system is approximated as pffiffiffiffiffi N ð4Þ nc ðtf ! 1Þ ¼ 2 Table 1 lists some common separation techniques and typical peak capacities. Cursory examination would suggest that a carefully chosen separation technique and a single dimension detector would be sufficient to separate mixtures containing up to nc components. This conclusion however would be incorrect. Components in mixtures will have a distribution of retention times with no a priori way to avoid coelution. The probability of coelution is much like the old party trick of showing that in a roomful of 40 or more people, two people are very likely to share the same birthday (try it!). The statistical argument can be presented for the birthday trick (or for chromatography) as follows: In a room with only two people, the second person has 365 out of 366 (remember leap years) chance of not having the same birthday. Adding an additional person to the room, the chance that the third person does not share either of the other two birthdays is given by 364 divided by 366. But the chance that none in the room share the same birthday is joint, i.e., the

343

13 HYPHENATED CHARACTERIZATION TECHNIQUES

TABLE1 Calculated Peak Capacities for Common Chromatographic SeparationTechniques Separation technique

Theoretical plates (N)

Thin-layer chromatography

10

Flash chromatography Packed column gas chromatography High-performance liquid chromatography Capillary gas chromatography Capillary electrophoresis

Peak capacity (nc)