CLINICAL TRIALS A Methodologic Perspective pdf

CLINICAL TRIALS WILEY SERIES IN PROBABILITY AND STATISTICS Established by WALTER A SHEWHART and SAMUEL S WILKS Editors: David J Balding, Noel A C Cressie, Nicholas I Fisher, Iain M Johnstone, J B Kadane, Geert Molenberghs, Louise M Ryan, David W Scott, Adrian F M Smith, Jozef L Teugels Editors Emeriti: Vic Barnett, J Stuart Hunter, David G Kendall A complete list of the titles in this series appears at the end of this volume CLINICAL TRIALS A Methodologic Perspective Second Edition STEVEN PIANTADOSI Johns Hopkins School of Medicine, Baltimore, MD A JOHN WILEY & SONS, INC., PUBLICATION Copyright 2005 by John Wiley & Sons, Inc All rights reserved Published by John Wiley & Sons, Inc., Hoboken, New Jersey Published simultaneously in Canada No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning, or otherwise, except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of the appropriate per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400, fax 978-646-8600, or on the web at www.copyright.com Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, (201) 748-6011, fax (201) 748-6008 Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best efforts in preparing this book, they make no representations or warranties with respect to the accuracy or completeness of the contents of this book and specifically disclaim any implied warranties of merchantability or fitness for a particular purpose No warranty may be created or extended by sales representatives or written sales materials The advice and strategies contained herein may not be suitable for your situation You should consult with a professional where appropriate Neither the publisher nor author shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages For general information on our other products and services please contact our Customer Care Department within the U.S at 877-762-2974, outside the U.S at 317-572-3993 or fax 317-572-4002 Wiley also publishes its books in a variety of electronic formats Some content that appears in print, however, may not be available in electronic format Library of Congress Cataloging-in-Publication Data: Piantadosi, Steven Clinical trials : a methodologic perspective / Steven Piantadosi—2nd ed p ; cm – (Wiley series in probability and statistics) Includes bibliographical references and index ISBN-13: 978-0-471-72781-1 (cloth : alk paper) ISBN-10: 0-471-72781-4 (cloth : alk paper) Clinical trials—Statistical methods I Title II Series [DNLM: Biomedical Research—methods Research—methods Clinical Trials—methods Statistics—methods W 20.5 P581c 2005] R853.C55P53 2005 610 72—dc22 2005040833 Printed in the United States of America 10 CONTENTS Preface xxi Preface to the First Edition xxv Preliminaries 1.1 Introduction, 1.2 Audience and Scope, 1.3 Other Sources of Knowledge, 1.3.1 Terminology, 1.3.2 Review of Notation and Terminology Is Helpful, 1.4 Examples, Data, and Programs, 1.5 Summary, Clinical Trials as Research 2.1 Introduction, 2.1.1 Clinical Reasoning Is Based on the Case History, 10 2.1.2 Statistical Reasoning Emphasizes Inference Based on Designed Data Production, 12 2.1.3 Clinical and Statistical Reasoning Converge in Research, 13 2.2 Defining Clinical Trials Formally, 14 2.2.1 Mixing of Clinical and Statistical Reasoning Is Recent, 14 2.2.2 Clinical Trials Are Rigorously Defined, 16 2.2.3 Experiments Can Be Misunderstood, 17 2.2.4 Clinical Trials as Science, 18 2.2.5 Trials and Statistical Methods Fit within a Spectrum of Clinical Research, 19 v vi CONTENTS 2.3 Practicalities of Usage, 20 2.3.1 Predicates for a Trial, 20 2.3.2 Trials Can Provide Confirmatory Evidence, 20 2.3.3 Clinical Trials Are Unwieldy, Messy, and Reliable, 21 2.3.4 Other Methods Are Valid for Making Some Clinical Inferences, 23 2.3.5 Trials Are Difficult to Apply in Some Circumstances, 25 2.3.6 Randomized Studies Can Be Initiated Early, 25 2.4 Summary, 26 2.5 Questions for Discussion, 26 Why Clinical Trials Are Ethical 3.1 Introduction, 29 3.1.1 Science and Ethics Share Objectives, 30 3.1.2 Equipoise and Uncertainty, 31 3.2 Duality, 32 3.2.1 Clinical Trials Sharpen, but Do Not Create, the Issue, 32 3.2.2 A Gene Therapy Tragedy Illustrates Duality, 32 3.2.3 Research and Practice Are Convergent, 33 3.2.4 The Hippocratic Tradition Does Not Proscribe Clinical Trials, 36 3.2.5 Physicians Always Have Multiple Roles, 38 3.3 Historically Derived Principles of Ethics, 41 3.3.1 Nuremberg Contributed an Awareness of the Worst Problems, 41 3.3.2 High-Profile Mistakes Were Made in the United States, 42 3.3.3 The Helsinki Declaration Was Widely Adopted, 42 3.3.4 Other International Guidelines Have Been Proposed, 44 3.3.5 Institutional Review Boards Provide Ethical Oversight, 45 3.3.6 Ethical Principles Relevant to Clinical Trials, 46 3.4 Contemporary Foundational Principles, 48 3.4.1 Collaborative Partnership, 48 3.4.2 Scientific Value, 49 3.4.3 Scientific Validity, 49 3.4.4 Fair Subject Selection, 49 3.4.5 Favorable Risk–Benefit, 50 3.4.6 Independent Review, 50 3.4.7 Informed Consent, 51 3.4.8 Respect for Subjects, 52 3.5 Methodologic Reflections, 53 3.5.1 Practice Based on Unproven Treatments Is Not Ethical, 53 3.5.2 Ethics Considerations Are Important Determinants of Design, 55 3.5.3 Specific Methods Have Justification, 57 29 CONTENTS vii 3.6 Professional Conduct, 59 3.6.1 Conflict of Interest, 59 3.6.2 Professional Statistical Ethics, 61 3.7 Summary, 63 3.8 Questions for Discussion, 63 Contexts for Clinical Trials 65 4.1 Introduction, 65 4.1.1 Some Ways to Learn about Trials in a Given Context, 66 4.1.2 Issues of Context, 67 4.2 Drugs, 68 4.2.1 Are Drugs Special?, 70 4.2.2 Why Trials Are Used Extensively for Drugs, 71 4.3 Devices, 72 4.3.1 Use of Trials for Medical Devices, 73 4.3.2 Are Devices Different from Drugs?, 74 4.3.3 Case Study, 76 4.4 Prevention, 76 4.4.1 The Prevention versus Therapy Dichotomy Is Overworked, 77 4.4.2 Vaccines and Biologicals, 78 4.4.3 A Perspective on Risk–Benefit, 79 4.4.4 Methodology and Framework for Prevention Trials, 81 4.5 Complementary and Alternative Medicine, 82 4.5.1 The Essential Paradox of CAM and Clinical Trials, 84 4.5.2 Why Trials Have Not Been Used Extensively in CAM, 85 4.5.3 Some Principles for Rigorous Evaluation, 87 4.6 Surgery and Skill-Dependent Therapies, 88 4.6.1 Why Trials Have Not Been Used Extensively in Surgery, 90 4.6.2 Reasons Why Some Surgical Therapies Require Less Rigorous Study Designs, 92 4.6.3 Sources of Variation, 92 4.6.4 Difficulties of Inference, 93 4.6.5 Control of Observer Bias Is Possible, 94 4.6.6 Illustrations from an Emphysema Surgery Trial, 95 4.7 A Brief View of Some Other Contexts, 101 4.7.1 Screening Trials, 101 4.7.2 Diagnostic Trials, 103 4.7.3 Radiotherapy, 103 4.8 Summary, 104 4.9 Questions for Discussion, 105 Statistical Perspectives 5.1 Introduction, 107 107 viii CONTENTS 5.2 Differences in Statistical Perspectives, 108 5.2.1 Models and Parameters, 108 5.2.2 Philosophy of Inference Divides Statisticians, 108 5.2.3 Resolution, 109 5.2.4 Points of Agreement, 110 5.3 Frequentist, 112 5.3.1 Binomial Case Study, 113 5.3.2 Other Issues, 114 5.4 Bayesian, 115 5.4.1 Choice of a Prior Distribution Is a Source of Contention, 116 5.4.2 Binomial Case Study, 117 5.4.3 Bayesian Inference Is Different, 119 5.5 Likelihood, 120 5.5.1 Binomial Case Study, 121 5.5.2 Likelihood-Based Design, 122 5.6 Afterthoughts, 123 5.6.1 Statistical Procedures Are Not Standardized, 123 5.6.2 Practical Controversies Related to Statistics Exist, 123 5.7 Summary, 125 5.8 Questions for Discussion, 125 Clinical Trials as Experimental Designs 6.1 Introduction, 127 6.1.1 Trials Are Relatively Simple Experimental Designs, 128 6.1.2 Study Design Is Critical for Inference, 129 6.2 Goals of Experimental Design, 129 6.2.1 Control of Random Error and Bias Is the Goal, 129 6.2.2 Conceptual Simplicity Is Also a Goal, 130 6.2.3 Encapsulation of Subjectivity, 131 6.2.4 Leech Case Study, 131 6.3 Trial Terminology, 132 6.3.1 Drug Development Traditionally Recognizes Four Trial Designs, 132 6.3.2 Descriptive Terminology Is Broader and Recognizes More Types of Trials, 133 6.4 Design Concepts, 134 6.4.1 The Foundations of Design Are Observation and Theory, 134 6.4.2 A Lesson from the Women’s Health Initiative, 136 6.4.3 Experiments Use Three Components of Design, 137 6.5 Survey of Developmental Trial Designs, 143 6.5.1 Early Development, 143 6.5.2 Middle Development, 144 6.5.3 Late Development, 148 127 CONTENTS ix 6.6 Special 6.6.1 6.6.2 6.6.3 6.6.4 6.6.5 Design Issues, 151 Placebos, 151 Equivalence or Noninferiority, 154 Nonuniformity of Effect, 154 Randomized Discontinuation, 155 Hybrid Designs May Be Needed for Resolving Special Questions, 156 6.6.6 Clinical Trials Cannot Meet Some Objectives, 156 6.7 Importance of the Protocol Document, 157 6.7.1 Protocols Have Many Functions, 158 6.7.2 Deviations from Protocol Specifications Are Common, 159 6.7.3 Protocols Are Structured, Logical, and Complete, 160 6.8 Summary, 164 6.9 Questions for Discussion, 165 Random Error and Bias 167 7.1 Introduction, 167 7.2 Random Error, 169 7.2.1 Hypothesis Tests versus Significance Tests, 169 7.2.2 Hypothesis Tests Are Subject to Two Types of Random Error, 170 7.2.3 Type I Errors Are Relatively Easy to Control, 171 7.2.4 The Properties of Confidence Intervals Are Similar, 171 7.2.5 Using a One- or Two-Sided Hypothesis Test Is Not the Right Question, 172 7.2.6 P -Values Quantify the Type I Error, 173 7.2.7 Type II Errors Depend on the Clinical Difference of Interest, 173 7.2.8 Post hoc Power Calculations Are not Helpful, 175 7.3 Clinical Biases, 176 7.3.1 Relative Size of Random Error and Bias Is Important, 176 7.3.2 Bias Arises from Numerous Sources, 176 7.3.3 Controlling Structural Bias Is Conceptually Simple, 179 7.4 Statistical Bias, 182 7.4.1 Some Statistical Bias Can Be Corrected, 183 7.4.2 Unbiasedness Is Not the Only Desirable Attribute of an Estimator, 183 7.5 Summary, 185 7.6 Questions for Discussion, 185 Objectives and Outcomes 8.1 Introduction, 187 187 672 AUTHOR INDEX Tosteson, T D., 631 Treasure T., 88, 615 Trimble, E L., 650 Troy, G., 640 Tuffs, A., 557, 655 Tukey, J W., 32, 655 Tunis, S., 619 Turner, J G., 84, 85, 105, 655 Tyce, F A., 642 Tygstrup, N., 311, 655 Ungerleider, R S., 21, 646 Urbach, P., 16, 57, 334, 655 Urban C., 658 Vacanti, J P., 645 Valsecchi, M G., 6, 440, 453, 640 Valtonen, S., 449, 451, 655 Van Belle, S., 634 Van Den Eeden, S K., 525, 641 van der Aart, Q J M., 645 van Gilse, H A., 619 Van Glabbeke, M., 619 Van Houwelingen, H C., 185, 656 van Kammen, D P., 651 Vandervoort, M K., 644 Varley, S A., 5, 649 Varmus, H., 43, 655 Vawter, D E., 627 Venet, L., 651 Venet, W., 651 Venner, P., 650 Verhoeven, E L G., 648 Verweij, J., 206, 655 Verweij, P J M., 185, 656 Vesell, E S., 324, 651 Vetterling, W T., 648 Vietti, T., 652 Villanueva, C., 500, 656 Virtamo, J., 632 Viscoli, C M., 636 Vogel, G., 557, 656 Vogelzang, N., 650 Vogelzang, N J., 656 Volkers, N., 24, 635 Vollset, S E., 259, 656 Wade, N., 539, 540, 551, 618, 656 Wald, A., 374, 375, 656 Waldhausen, J A., 481, 656 Waller, R A., 116, 640 Walter, G., 116, 644 Walter, S D., 103, 432, 481, 622, 656 Walters, D E., 261, 263, 656 Ward, J., 225, 656 Ware, J H., 16, 154, 343, 441, 631, 637, 645, 656 Warner, B., 476, 656 Webb, C F., 5, 649 Weber, C., 144, 242, 652 Weber, W W., 444, 656 Wedderburn, R W M., 458, 644 Weeks, B M., 651 Weeks, J C., 89, 656 Wei, L J., 333, 340, 342, 343, 637, 656 Weigel-DiFranco, C., 617 Weijer, C., 50, 656 Weinmann, G G., 96, 656 Weinstein, A S., 622 Weiss, G H., 340, 633 Weiss, R., 550, 656 Weiss, R B., 555, 559, 656 Wells, F., 48, 545, 657 Wells, G A., 493, 642 Wendler, D., 624 West, K P., Jr., 653 Westervelt, S D., 541, 628 Westphal, M., 95, 449, 657 Wetherill, G B., 231, 657 Weymuller, E A., 481, 657 Whisnant, J P., 618 Whitehead, A., 657 Whitehead, J., 116, 356, 382, 653, 657 Whittier, F C., 95, 657 Whitworth, J A., 139, 657 Wieand, H S., 628 Wiebers, D O., 618 Wigner, E., 6, 15, 657 Wilde, D J., 231, 657 Wilkens, C L., 651 Willan, A., 525, 657 Willet, W., 617, 653 Willett, W C., 640 Williams, B., 652 Williams, C J., 29, 657 Williams, D H., 491, 657 Williams, E D., 619 Williams, G., 198, 635 Williams, M., 655 Wilson, E O., 18, 130, 657 Wilson, F R., 14, 657 Wilson, J K., 231, 630 Wilson, S R., 639 Winer, B J., 127, 280, 657 Winget, M D., 488, 657 Wise, R., 638 Wittes, J., 203, 357, 390, 482, 633, 638, 657 Wittes, R E., 4, 160, 309, 499, 638, 652, 657 Wolf, J C., 646 Wolfe, C R., 628 Wolfe, M J., 646 Wolfe, S M., 43, 80, 616, 640 Wolfram, S., 7, 567, 657 Wolfson, S., 151, 614 Wong, J., 89, 638 AUTHOR INDEX Wong, W K., 225, 658 Wooding, W M., 4, 658 Woodward, B., 29, 658 Woolf, P K., 546, 658 Wootten, J C., 54, 653 Wright, P W., 622 Wu, M., 304, 658 Yank, V., 648 Yates, F., 282, 501, 658 Young, J H., 69, 658 673 Yuan, C.-S., 82, 658 Yusuf, S., 148, 314, 537, 658 Zahurak, M L., 239, 630 Zamecnik, P C., 632 Zeger, S L., 405, 407, 439, 623, 639, 653, 658 Zeigler, D., 641 Zelen, M., 342, 351, 482, 658 Zhou, Y., 657 Zoubek A., 148, 658 Zwischenberger, J B., 616 SUBJECT INDEX B (notation), 570 (notation), 570 (notation), 570 (notation), 570 (notation), 571 (notation), 570 α level, 382 α (notation), 570 α spending, 386 β (notation), 570 χ (notation), 570 δ (notation), 570 (notation), 570 γ (notation), 570 λ (notation), 570 µ (notation), 570 ν (notation), 570 φ (notation), 570 π (notation), 570 (notation), 570 ρ (notation), 570 σ (notation), 570 (notation), 570 τ (notation), 570 B -values, 390 abbreviations, 587, 589 ACAPS Group, 614 accountability, 541 accrual assessing, 318 def (rate), 572 estimating, 318 estimating the number of events, 319 low, 318 modeling, 319 power, 288 quantitative estimates, 319 run-in period, 318 ACS, 54, 614 active control, 572 active sampling, 310, 326 Ad hoc Working Group for Critical Appraisal of the Medical Literature, 481, 613 adaptive randomization, 340 adherence, 82, 299, 403 def., 572 adjusted analyses, 124, 471 def of adjustment, 572 example, 473 reporting, 492 AIDS trials, 15 prognostic risk group analysis, 466 references, allocation ratio, 279 power, 288 alternative hypothesis, 174 Alzheimer’s Disease Anti-inflammatory Prevention Trial, 80 American Association for Thoracic Surgery, 100 American Medical Association, 38, 614 American Statistical Association, 61, 614 code of conduct, 61 analysis, 129 adjusted, 470 Clinical Trials: A Methodologic Perspective, 2E, by S Piantadosi Copyright  2005 John Wiley & Sons, Inc 675 676 SUBJECT INDEX analysis (continued) based on adherence, 403 basis of, 409 covariance, 427 preplanned, 411 prognostic factors, 453 treatment received, 405 antiplatelet therapy meta-analysis, 535 area under the curve, 240, 258, 405, 414, 415 def., 572 ASCO, 225, 615 assessment bias, 178 ATBC, 510, 511, 615 auditing, 559 authoritarianism, 178 authorship, 494 governance, 498 inclusion and ordering, 495 models, 497 responsibilities, 496 autonomy, 46 AZT See zidovudine size, 176 sources, 176 statistical, 182 structural, 179 vs random error, 167 binary variable def., 573 binomial distribution def., 573 bioassay, 229 biologicals, 78 Biostatistica (journal), blinding, see masking142 block def., 573 blocked randomization def., 573 blocking, 140, 336, 337, 339 maximum imbalance, 338 proper analysis, 340 size, 339 variable size, 337 Breast Cancer Action, 80 background information, 161 balanced design, 341 def., 572 barriers to participation, 322 baseline, 573 Bayes’s theorem, 117, 175 def., 573 Bayesian methods, 108, 109, 115, 240, 264, 356 Bayesian versus frequentist inference, 119 confidence interval, 263 data-dependent stopping, 372, 378 example, 379 difficulties, 116 dose-finding, 230 example, 117 inference, 115 posterior distribution, 119 SA trials, 264 Belmont Report, 45, 545 beneficence, 47 beta distribution, 312 betting, 411 bias, 22, 129, 130, 167, 168, 331 assessment, 178 biased randomization, 342 clinical, 176 control, 181 def., 573 dose-ranging designs, 234 early stopping, 183, 387 estimation, 183 in incompletely specified models, 464 publication, 177, 483 selection, 93, 177, 483 Cancer and Leukemia Group B, 445 cancer trials, 15 cardiovascular disease trials, 15 case report, 134 def., 573 series of, 135 case series, 93, 135 def., 574 case-control design, 24 CAST Investigators, 206, 619 censoring, 194, 195, 399, 420 competing risks, 199 def., 574 independent, 196 informative, 196 interval, 195 left, 195 right, 195 time to good outcomes, 199 type I, 195 type II, 195 type III, 195 vs loss to follow-up, 197 Center for Health Care Technology Assessment, 96 Center for Medicare and Medicaid Services, 66, 96 Christmas tree design, 383 Civilized Software, 565, 620 classification, 189 clinical reasoning, 10, 11 clinical research spectrum including trials, 19 clinical significance, 424 clinical trial application, 25 SUBJECT INDEX as a research design, comparative treatment efficacy (CTE) trials, 148 complexity, 129 contexts, 65 def., 16, 574, 584 devices, 74 DF, 143 dose-finding (DF) trials, 143 drawbacks, 21 ethics, 29 expanded safety (ES) trials, 150 Hippocratic Oath, 37 history, 15 in complementary and alternative medicine, 83 in device context, 73 in drug context, 71 in surgery, 88, 90 large-scale (LS), 148 of drugs, 68 phase I, 143 (See TM or DF) phase II, 145 (See SA) phase III, 132, (See CTE) phase IV, 150, (See ES) populist view, 17, 34, 67 protocols, 157 purpose, 130 risk, 35 safety and efficacy (SA) trials, 144 TM, 143 translational, 143 treatment mechanism (TM) trials, 143 types, 132 use of, 65 clinical utility efficacy and toxicity, 250 cluster randomization, 297 Cochrane Collaboration, 67 coding of variables, 457 COMMIT, 139, 621 COMMIT trial, 139 community intervention trials, 139 competing risks, 199 complementary and alternative medicine characteristics, 86 def., 82, 83 rigorous evaluation, 87 use of trials in, 85 compliance monitoring, 319 computer software frequentist stopping boundaries, 392 concept sheet, 158 conditional power, 387 confidence interval, 116, 171, 259 Bayesian, 263 def., 574 event rate, 268 exact binomial, 260 mean, 267 677 confirmatory trials, 20 conflict of interest, 59 conflicts of obligation, 32 confounder, 331 def., 574 consent, 51 CONSORT statement, 479 constant infusion, 415 constraints, 128 contexts, 65 complementary and alternative medicine, 83 devices, 73 diagnostic trials, 103 drugs, 68 overview, 68 prevention, 76 radiotherapy, 104 screening, 101 surgery, 88 continual reassessment method, 224, 237 example, 238 for two interacting drugs, 247 pharmacokinetic measurements, 240 Controlled Clinical Trials (journal), controls concurrent, 181 historical, 24 convenience sampling, 326 Coronary Drug Project, 403 Coronary Drug Project Research Group, 403, 621 correlation coefficient def., 574 cost optimization, 298 Council for International Organizations of Medical Sciences, 45, 621 covariates, 453, 470 correlated, 463 models for, 457 time-dependent, 455 credible interval, 266 critical value, 170 crossover trial, 515 advantages and disadvantages, 517 analysis, 519–521, 525 applications, 520 carryover effects, 518 def., 515, 574 dropouts, 519 example, 525 linear model, 521 multiple treatments, 516 precision, 517 prerequisites, 520 recruitment, 518 test of carryover effect, 522 treatment by period interaction, 521 variance of treatment effect estimate, 517 washout period, 520 678 SUBJECT INDEX CTE trial, 145, 423 adjuvant, 149 bias, 179 choosing α and β, 277 def., 574 equivalence, 150 multiple institutions, 149 pivotal, 149 power, 277 reporting, 490 single modality, 149 Cytel Software, 392, 622 Dalton, 539 data, 565 def., 575 data-dependent stopping, 355, 577 approaches to evaluating evidence, 371 def., 575 DSMC review, 369 frequentist methods, 382 group sequential, 381, 384 likelihood methods, 373 pros and cons, 358 reasons, 357 SA trials, 271 sequential methods, 382 statistical guidelines, 372 tension, 358 data imperfections, 395, 396 correctable vs permanent, 395 data management, protocols, 163 data monitoring NCI policy, 599 NIH policy, 605 data torturing, 442 databases, 24, 135 decision theory, 108 data-dependent stopping, 373, 381 demonstration trials, 59, 156 design assumptions monitoring, 367 Bayesian, 115 benefits from, 129 blocking, 140 comparative trials, 135 components, 137 concepts, 134 def., 16 error control, 137 essentials, 130 ethics, 55 experimental units, 139 factors, 139 flaws, 129 foundations, 134 goals, 130 hybrid, 156 importance, 129 limitations, 156 nonexperimental, 23 observational units, 139 phase I trials, 224, 230 randomization, 140 sampling, 137 simplicity, 130, 131 surgical trials, 88, 92, 94 NETT, 96 treatment, 137 types of trials, 132 deterministic def., 575 device, 73 development, 74 safety, 75 device regulation, 73 device trials, 74 device function vs clinical outcome, 165 DF trial, 224 bias, 179 def., 575 DHHS, 53, 326, 327, 543, 553, 623 diagnostic trials, 103, 128 dichotomous def., 575 dosage modifications, 162 dose-finding, 179, 223, 224, 236, 258 efficacy vs safety, 226 general, 242 dose-limiting toxicity, 231 dose optimality, 226 dose-ranging, 225 dose–response, 227 logistic model, 230 model, 237 dose–response model, 248 double masking, 180 drop-in def., 575 dropout def., 575 drug context vs device context, 74 drug distribution, 413 drug information, 161 dual dose-finding, 244 duality, 30, 32, 39 due process (in misconduct), 550 early stopping, 355 EBCTCG, 533, 624 ECMO, 22, 57, 342, 351, 359 economics, 156 effect size, 10 SUBJECT INDEX effectiveness def., 575 method, 396 vs efficacy, 323, 396 efficacy, 405 def., 575 vs effectiveness, 323 eligibility criteria, 309, 314, 324, 368, 401 protocol, 161 reporting failures, 491 def., 575 errors, 400, 401 Endophthalmitis Vitrectomy Study Group, 625 endpoint, 409 ascertainment, 181 def., 575 dichotomous, 282 hazard ratios, 283 enthusiastic prior, 380 entropy, 215 empirical, 215 bias, 216 variance, 217, 218 epidemiologic studies, 135 equipoise, 31 equivalence trials, 150 error control, 149 random, 167 ES trial, 295 def., 575 purposes, 150 estimation, 188 def., 575 unbiased, 183 ethics, 29 Belmont Report, 45 data-dependent stopping, 355 double standards, 34 epidemiologic, 32 examples of competing obligations, 39 for the statistician, 61 Hippocratic Oath, 36 international guidelines, 41, 44 IRBs, 45 IRBs and TEMCs, 362 lapses, 42 overview for trials, 30 physician obligation, 32 abortion, 39 managed care, 40 organ donation, 40 quarantine, 40 teaching and training, 39 triage, 39 vaccination, 39 physician roles, 38 practitioner lack of knowledge, 54 principles, 46 autonomy, 46 beneficence, 47 collaborative partnership, 48 fair subject selection, 50 favorable risk–benefit, 50 implications, 48, 53 independent review, 51 informed consent, 51 investigator competence, 48 justice, 47 respect for subjects, 53 scientific validity, 49 scientific value, 49 study design, 48 questionable treatments, 55 randomization, 30 research vs practice, 33 selection criteria, 316 study design, 49 trial design, 55 evaluability criteria, 397, 401 def., 575 event rates cause-specific, 398 data-dependent stopping, 376 event times, 194 events counting, 395 uncounted, 398, 399 evidence, 121, 254 misleading, 255 weak, 255 exclusions, 178, 404 criteria, 314, 316 post hoc, 182 reporting, 493 experiment, 17, 127 characteristics, 19 def., 16, 575 experimental design, 9, 127, 128 experimental method, 14 experimental unit def., 576 explanatory perspective, 396 exponential distribution def., 576 external validity, 130, 177, 319 selection criteria, 317 factor def., 576 levels, 139 factorial designs, 81, 516 by by design, 502 by design, 502 679 680 SUBJECT INDEX factorial designs (continued) advantages, 501, 504 characteristics, 502 estimating interactions, 505 estimating main effects, 504 examples of trials, 509 incomplete, 512 interactions, 505, 507 interactions and scale of measurement, 507 linear models, 508 main effects, 507 partial or fractional, 511 precision, 505 FDA, 520, 626 Fibonacci, 231, 244 dose-ranging, 230, 236 follow-up, 576 active vs passive, 181 loss, 398 Food and Drug Administration, 69 Food, Drug, and Cosmetic Act of 1938, 72 fraud, 60, 539, 541, 544 frequentist methods, 108, 112, 382 inference, 119 gain score, 427 Galen, 178 Galileo, 6, 539 ganciclovir, 23 GAO, 445, 628 Gavarret, 14 gender representation, 325 gene therapy, 32 generalization, 11 generalized linear models, 458 golden ratio, 231 gout, 11 Greek letters, 570 group sequential designs SA trials, 272 Hawthorne effect, 192 Haybittle-Peto boundary, 386 hazard estimate, 284 hazard rate, 422 confidence intervals, 423 def., 576 hazard ratio, 284, 382 confidence limits, 284 def., 577 power and sample size, 285 Helsinki Declaration, 42, 44, 545, 593 heterogeneity two-way control, 141 hierarchy of evidence, 137 Hipparchus of Rhodes, 539 Hippocratic Oath, 36 historical controls, 24 hydrazine sulfate, 88, 444, 445 hypothesis test, 112, 114, 167, 169 one- or two-sided, 172 random error, 170 ICMJE, 67, 481 IEC, 51 imputation of missing values, 399 inclusion criteria excessive, 316 inclusiveness, 322 incremental improvement, 10, 68, 91, 102 IND investigator responsibilities, 561 sponsor responsibilities, 562 Independent Ethics Committee See IEC indicator variables, 457 ineligible patients, 400 inference, 107 Bayesian, 115 def., 577 frequentist, 112 likelihood, 120 methods other than trials, 23 informed consent, 46, 51, 58, 323 elements of, 51 in emergency settings, 52 Inspector General Act Amendments, 548, 634 Institute of Medicine, 541, 634 Institutional Review Board, 45, 50, 51 integrity, 540 intention to treat, 124, 194, 368, 402–404 def., 402, 577 interaction, 455, 465, 505, 506, 511, 519 in factorial designs, 505 representation, 327 treatment by period, 519 interim analysis, 355 def., 577 intermediate outcomes, 204 International Agency for Research on Cancer, 444, 634 International Chronic Granulomatous Disease Cooperative Study Group, 209, 634 International Committee of Medical Journal Editors, 479, 634 International Conference on Harmonization, 158, 484, 634 essential documents for a clinical trial, 158 trial reports, 484 International Society for Clinical Biostatistics, International Statistical Institute, 61 interval scales, 189 investigator competence, 49 investigator responsibilities, 560 ISIS-2 Collaborative Group, 359, 444, 634 ISIS-4 Collaborative Group, 500, 510, 511, 634 SUBJECT INDEX Jewish Chronic Diseases Hospital, 42 Johns Hopkins University School of Medicine, 549, 635 Judicial Council of the American Medical Association, 42, 635 justice, 47 Laetrile, 36, 138 large-scale trials, 124, 315, 323 last observation carried forward, 400 Latin square, 141 Latrile, 88 Leibniz, 539 lifetables, 301 likelihood, 120, 254, 266, 281, 293, 296 def., 577 design, 122 event rate, 271 example, 122 function, 110, 117 monitoring methods, 373 ratio test, 374 sample size, 267, 281 support intervals, 122 linear model, 427 Linnæus, 11 local control, 127 Local Research Ethics Committee, 51 logistic model, 458 logistic regression def., 578 logrank statistic, 382 lost to follow-up, 399 def., 578 vs censoring, 197 LS trials def., 579 lung cancer, 425 lung volume reduction surgery, 25 See also National Emphysema Treatment Trial masking, 142, 180 double, 180 triple, 180 matrix def., 579 maximum likelihood estimate, 111 maximum nontoxic dose, 227 maximum tolerated dose, 227 def., 579 drug combinations, 244 mean squared error, 184 measurement scales, 189 measurement error, 441 Medical Research Council, 510, 641 Mendel, 539 mentoring, 545 mesothelioma, 418 meta-analysis, 124, 157, 483, 529 cumulative, 534 example, 535 history, 529 limitations, 536 methods, 531 observed minus expected, 534 precision, 530 quality rating system, 534 reviews, 530 statistical analysis, 533 study eligibility criteria, 533 study retrieval, 532 method effectiveness, 396 middle development, 144 minimization, 140, 341 minimum effective dose, 227 def., 579 minority representation, 325 misconduct, 539 basic sciences cases, 540 case histories, 551 causes, 546 clinical sciences cases, 540 def., 542, 544 detecting, 558, 563 example of university procedures, 549 frequency, 545 fringe science, 539 impact, 555, 556 importance in clinical sciences, 541 institutional approach, 549 investigations, 547 recent, 540 NAS definition, 542 NSF definition, 544 NSF policies, 548 ORI, 547 outside the U.S., 545 PHS definition, 543 PHS policies, 547 Piltdown man, 540 potential problem areas, 558, 559 problem areas, 550 psychology, 547 RCP definition, 544 recognizing, 557 statistical sciences cases, 540 missing data, 461 MLAB, 565 model, 13, 108, 440, 456, 457 building, 459 characteristics, 456 def., 579 generalized linear, 458 incomplete specification, 464 681 682 SUBJECT INDEX model (continued) interactions, 465 logistic, 458 missing data, 461 multiple regression, 463 proportional hazards, 458 random effects, 459 risk groups, 466 model fitting, 565 monitoring, 359 Bayesian methods, 379 components, 357 def., 579 ethics, 58 motives, 356 report, 360 most likely to succeed dose, 227 MPS Research Unit, 392, 643 MTD, 229 multiple comparisons def., 579 multiple treatments, 128 multiplicity, 171, 442, 443 multivariable def., 580 multivariate def., 580 National Academy of Sciences, 541, 545, 546, 643 National Cancer Institute, 15, 361, 445, 643 National Commission for Protection of Human Subjects of Biomedical and Behavioral Research., 45, 46, 643 National Death Index, 399 National Emphysema Treatment Trial, 25, 66, 94, 95, 157, 398 results, 97 National Heart Lung and Blood Institute, 96 National Research Act of 1974, 45 National Science Foundation, 544, 545, 644 NCI Press Office, 553, 643 negative findings, 484 reporting, 492 nesting, 140 def., 580 models, 464 NETT Research Group, 66, 97–99, 157, 644 neural networks, 476 Newton, 539 NHLBI, 361, 644 NIH, 553, 644 NIH Revitalization Act of 1993, 324 Non–small Cell Lung Cancer Collaborative Group, 500, 644 nonadherence, 395, 400, 401, 405 power, 301 sample size inflation, 299 treatment, 402 noncompliance See nonadherence nonexperimental designs, 23 noninferiority, 154 confidence limits, 291 likelihood, 292 sample size, 290 nonuniformity of treatment effect, 155 nonvalidated practice, 32 norms of science, 542 North Central Cancer Treatment Group, 445 notation, 6, 569 Greek, 570 Roman letters, 571 NSABP, 553, 554, 556, 644 null hypothesis, 173, 277 def., 580 Nuremberg Code, 41, 44, 591 Nuremberg trials, 41 O’Brien-Fleming boundary, 386 objectives, 187 internal, 187 protocol, 161 selection, 189 objectivity in monitoring, 365 observation, 134 def., 580 observational study def., 580 observational unit def., 580 observer bias, 67, 70, 93, 94 odds def., 580 odds ratio def., 580 Office for Human Research Protections, 56 Office of Research Integrity, 545, 548, 553, 645 Office of Science and Technology Policy, 544, 645 Office of Technology Assessment, 54, 65, 645 operating characteristic dose-ranging designs, 232 optimal biological dose, 226, 581 optimization joint for safety and efficacy, 247 ordered categories, 193 ordering, 189 outcome, 192 as predictor, 190 censoring, 195 counts, 193 dichotomous, 194 event times, 194 hard, 191 intermediate, 204 measured values, 192 ordered categories, 193 SUBJECT INDEX patient reported, 207 qualitative versus quantitative, 192 quality of life, 207 repeated measures, 207 surrogate, 145, see surrogate outcome200, 204 survival, 198 translational trials, 212 unordered categories, 193 outcomes research, 24 overview See also meta-analysis def., 157, 529, 581 p-value, 114 def., 583 early stopping, 384 meaning, 173 poor properties, 432, 433 reporting, 487, 493 statistical significance, 426 strength of evidence, 170 summarizing data, 172 type I error, 173 vs confidence intervals, 433 parallel design def., 581 parameter, 108, 456 def., 581 Parkinson Study Group, 646 Parkinson’s disease, 76, 94 partial sums, 388 patient reported outcome, 207 peer review, 482, 558 penalized likelihood, 185 permutation test, 346 example, 347 permutations number of, 346 personal care principle, 35 pharmacokinetic model, 240, 410, 412, 413, 416 reporting, 488 phase I trial, 132, 223 def., 581 designs in use, 230 healthy volunteers, 242 ideal design, 229 MTD, 242 objectives, 188 purposes, 225 reporting, 488 phase II trial, 132 def., 581 phase III trial, 132, 133 def., 581 phase IV trial, 132, 133 def., 582 physician behavior trial reports, 480 Physician Data Query, 66 Physicians’ Health Study, 509 Piltdown man, 539 pivotal trials, 149 placebo, 43, 58, 67, 70, 91, 151 controls, 123 def., 582 effect, 11, 89, 152, 180 experimental, 151 surgical, 89, 94 use of, 153 plagiarism, 544 play the winner, 342 PLCO Project Team, 630 Pocock boundary, 384 Poisson distribution, 295 Poisson rates, 296 polyposis, 424, 426 portacaval shunt, 93 posterior distribution, 117 postmarketing surveillance, 150 Potsdam International Consultation on Meta-analysis, 530, 648 power, 174, 175, 251, 253, 282 calculating, 255 computer programs, 303 CTE trials, 277 curve, 174, 304 def., 582 equivalence trials, 154 event rates, 284 for prognostic factor analyses, 470 nonadherence, 300 nonparametric assumptions, 285 parametric assumptions, 284 sample size, 256 simulation, 304 t-test, 278 practice vs research, 33 pragmatic perspective, 396 precision, 129, 130, 253, 259, 426 direct specification, 253 relative, 270 prevention, primary, 76, 79 secondary, 76 tertiary, 77 prevention trials, 76, 77, 128 error rates, 277 methodology, 81 vs therapeutic trials, 77 primary outcome, 582 prior distribution, 116, 265, 379 clinical, 116 eliciting, 116 enthusiastic, 117 reference, 116 skeptical, 117 683 684 SUBJECT INDEX privacy, 53 probit def., 582 product limit method, 196, 420 professional conduct, 59 prognostic factor, 418, 440 analysis, 453 motivations, 454 neural networks, 476 nonmodel methods, 475 recursive partitioning, 475 balance, 336, 337 characteristics, 455 control, 336 def., 582 sample size, 302 scale of measurement, 457 Project ImpACT, 67 proportional hazards def., 582 proportional hazards model, 458 Prostate Cancer Trialists’ Collaborative Group, 500, 648 protocol, 157, 370 agent information, 162 background information, 161 concept sheet, 158 def., 582 deviations, 159 dosage modifications/side effects, 162 drug information, 161 eligibility criteria, 161 importance, 157 objectives, 161 outline, 160 publishing, 484 purposes, 158 randomization, 162 serial measurements/study calendar, 162 staging criteria, 161 statistical section, 163 treatment evaluation, 162 treatment program, 162 protocol deviations, 159 pseudorandom numbers, 345 def., 582 Ptolemy, 539 publication bias, 21, 177, 483 def., 582 quality assurance DSMC, 368 quality of life, 207 quantile def., 583 radiotherapy, 85, 103 random def., 583 random effects models, 441 random error, 127, 167–169 size, 176 vs bias, 167 randomization, 15, 25, 124, 127, 144, 147, 242, 332, 333, 520 adaptive, 340 administration, 343 benefits of, 334 bias control, 179 blocked, 336 causal inference, 334 cluster, 297, 333 constrained, 336 control over, 335 control over unknown factors, 333 def., 583 design, 140 difficulties with, 26 discontinuation of a subset, 97 ethics, 30, 57 from the first patient, 25 imbalances, 335 in crossover trials, 516 justification of type I errors, 345 need for, 332 pre-randomization, 351 protocol, 162 reasons for early initiation, 26 SA trials, 147 safety and activity trials, 274 sealed envelopes, 344 simple, 335 TM trials, 144, 242 tuberculosis trial, 333 urn designs, 340 randomization tests, 346 randomized discontinuation, 155 rate def., 583 ratio def., 583 ratio scales, 190 reasoning clinical, 10, 13 statistical, 12, 13 recursive partitioning, 475 registration of trials, 66, 67, 481 regression model collinearity, 463 multiple, def., 579 multivariable, 463 univariable, 462 regression to the mean, 151 repeated measures, 207 SUBJECT INDEX replication, 127, 138 reporting, 479 abstracts, 481, 487 abstracts, structured, 487 biological consistency, 494 CTE trial background, 490 CTE trial discussion, 492 CTE trial figures and tables, 493 CTE trial methods, 491 CTE trial results, 492 electronic, 499 eligibility failures, 491 epidemiologic studies, 480 evaluating the literature, 482 exploratory analyses, 493 guidelines, 479, 482 CONSORT, 479 ICH, 480 ICMJE, 479 ICH guidelines, 484 intention-to-treat, 493 negative results, 484 objectives of phase I trials, 488 p-values, 494 peer review, 482 pharmacokinetic models, 488 phase I trials, 482, 488 efficacy, 489 patient characteristics, 488 publication bias, 483 SA event rates, 489 SA feasibility, 489 SA treatment comparisons, 490 SA trials, 489 statistical methods and assumptions, 492 statistical significance, 493 structured abstracts, 481 study design, 491 study population, 490 titles, 487 treatment assignment, 491 treatment failures, 491 uniformity, 481 utility of the report, 480 writing the paper, 481 representation, 318, 322–324 guidelines, 326 NIH and FDA guidelines, 326 race and biology, 329 validity and power, 326 research clinical and statistical reasoning, 13 right to participate, 323 vs practice, 33 Research Ethics Board, 51 research practices, 544 response rates 5-fluorouracil and colon cancer, 309 response surface dose-finding, 246 retroactive definitions, 397 retrospective designs, 160 ridge regression, 185 risk difference, 411 estimates of, 411 groups, 466, 467 example, 468 ratio def., 583 relative, 411 risk–benefit, 79 ratio, 50 Roman letters, 571 Royal College of Physicians, 544, 649 Royal Statistical Society, 61, 649 Royal Statistical Society Code of Conduct, 609 run-in period, 319 safety TEMC review, 368 safety and activity trial, 145, 258, 417 Bayesian methods, 264 bias, 179 data-dependent stopping, 271 external validity, 186 objectives, 188 randomization, 147 reporting, 489 sample size, 258 skipping, 145 two-stage designs, 272 safety and activity trial(SA) def., 583 Salk polio vaccine trial of 1954, 78 sample size, 129, 251, 282, 286, 288, 301, 315 computer programs, 303 CTE trials, 279 def., 583 DF trials, 229 ES trials, 295 event rates, 285 for prognostic factor analyses, 470 nonadherence, 299 power, 256 SA trials, 145, 259 translational trials, 218 vs number of events, 285 sampling design, 310 SAS Institute, 303, 565, 650 science, 18 vs nonscience, 18 scientific method, 84 misconduct, 542 norms of science, 542 685 686 SUBJECT INDEX screening trials, 101 seeding trials, 19 selection randomized SA trial, 274 selection bias, 91, 93, 177, 182, 311, 314 comparative trials, 317 controlled by randomization, 332 def., 583 literature, 484 quantitative example, 312 selection criteria quantitative, 316 sequential methods design, 374 fully, def., 576 group, def., 576 sequential probability ratio test, 374 sequential probability ratio test, 377 sham surgery, 89, 94, 95, 153 side effects, 162 ES trials, 150 signal-to-noise ratio, 168 significance test, 112, 169 simulation, 304 skeptical prior, 379 Society for Clinical Trials, SPLUS, 565, 653 spontaneous improvement, 151 staging criteria, 161 standardization, 123 standards, 123 Standards of Reporting Trials Group, 480, 490, 533, 653 statistic as science, 12 def., 583 different perspectives, 108 Statistical Analysis System (SAS) data and programs, 565 statistical bias, 168, 182 correcting, 183 statistical reasoning, 12, 13, 107 statistical significance def., 583 statistical societies American Statistical Association, 61 Royal Statistical Society, 61 Statistics and Epidemiology Research Corporation, 654 Statistics in Medicine (journal), Steering Committee of the Physicians’ Health Study Research Group, 64, 510, 654 stochastic approximation, 231 stopping boundary, 56, 375, 377, 382, 383 stratification, 124, 141, 337 def., 583 excessive, 339 structure statistical model, 108 use of, 22 study cohort, 309 defining, 310 study duration, 128 subjective probability, 116 subjectivity, 131, 228 subset analysis, 350 support intervals, 122 surgical trials, 88, 90, 101 developmental, 89 surrogate outcome, 145, 200, 204 cancer, 201 cardiovascular diseases, 203 def., 200, 584 disease specificity, 201 efficiency, 204 eye diseases, 203 HIV, 203 limitations, 205 uses, 206 survival curves, 196 nonparametric, 420, 428 time, 195 survivor’s bias, 97 synergy dose-finding, 244 systematic error, 127 systematic review, see meta-analysis529 t-test power and sample size, 278 Taylor series def., 584 terminology, 6, 17, 132, 133, 569 list, 572 theory, 134 therapeutic ignorance ethics, 54 therapeutic ratio def., 584 therapeutic touch, 84 time-dependent covariates, 440, 455 TM trial def., 584 randomization, 144, 242 Toronto Leukemia Study Group, 309, 655 toxicities, 162 toxicity DSMC review, 368 translational trials, 143, 211 bias, 216 characteristics, 213 def., 213 empirical entropy, 216 entropy, 215 SUBJECT INDEX sample size, 218, 256 setting, 212 vs dose-finding, 211 treatment poorly defined, 157 questionable methods in cancer, 54 unproved, ethics, 53 treatment allocation, 331 balance, 332 discovery, 331 haphazard, 335 practical issues, 332 unequal, 286, 349 treatment combinations, 504 treatment cost, 350 treatment differences, 426 treatment effects monitoring, 58, 355 administration, 360 components, 357 data quality, 368 for efficacy, 369 motives, 356 practical questions, 369 statistical issues, 372 tensions, 358 timeliness, 367 weaknesses, 371 treatment effects monitoring committee, 51, 56, 58, 181, 361, 555 baseline comparability, 366 composition, 363 expertise and objectivity, 365 expertise versus objectivity, 365 masking, 363 meeting format, 364 quality assurance, 368 relationship to investigators, 361 relationship to IRBs, 362 weaknesses, 371 treatment failures reporting, 491 treatment preference ethics, 57 treatment program, 162 treatment received, 403 treatment received analyses, 405 trial design, 127 triangular designs, 382 triple masking, 180 trust, 541 Tuskegee study, 42 687 two-compartment model, 413 two-stage designs, 272 type I error, 170, 171, 173, 189, 254, 277, 382, 384, 387 control, 171, 443 CTE trials, 277 def., 584 type II error, 170, 172, 173, 254, 277 CTE trials, 277 def., 584 U.S Congress, 58, 72, 655 U.S Government, 655 U.S Public Heath Service, 42 uncertainty, 128 uncertainty principle, 31 univariable def., 584 univariate def., 585 univariate analyses reporting, 492 unproven cancer treatments, 54 unrecorded data, 398 urn randomization, 341 utility coefficients, 207 vaccines, 78 validity biological, 159, 556 design, 159, 556 variability, 22, 128 def., 585 variable def., 585 variances unequal, 350 vector def., 585 washout period, 520 WHI Investigators Writing Group, 136, 658 Willowbrook State Hospital, 42 window of opportunity, 21 Women’s Health Initiative, 136 World Health Organization, 44 World Medical Association, 42, 658 zidovudine, 23 ... trials and an even-handed treatment of different statistical perspectives Examples from the clinical trials literature are used, and data and computer programs for some topics are available A. .. perspective that encourages it are exactly backward Historically as well as to a large extent today, clinical trials have been an anti-establishment tool, challenging authoritarian views and those aligned... involves a statistician only in the “analysis” of data from a trial can expect a substantially inferior product overall 2.2 DEFINING CLINICAL TRIALS FORMALLY 2.2.1 Mixing of Clinical and Statistical