The correct bibliographic citation for this manual is as follows: SAS Institute Inc 2015 Modern Approaches to Clinical Trials Using SAS®: Classical, Adaptive, and Bayesian Methods Cary, NC: SAS Institute Inc Modern Approaches to Clinical Trials Using SAS®: Classical, Adaptive, and Bayesian Methods Copyright © 2015, SAS Institute Inc., Cary, NC, USA ISBN 978-1-62959-385-2 (Hardcopy) ISBN 978-1-62960-082-6 (Epub) ISBN 978-1-62960-083-3 (Mobi) ISBN 978-1-62960-084-0 (PDF) All rights reserved Produced in the United States of America For a hard-copy book: 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, or otherwise, without the prior written permission of the publisher, SAS Institute Inc For a web download or e-book: Your use of this publication shall be governed by the terms established by the vendor at the time you acquire this publication The scanning, uploading, and distribution of this book via the Internet or any other means without the permission of the publisher is illegal and punishable by law Please purchase only authorized electronic editions and not participate in or encourage electronic piracy of copyrighted materials Your support of others' rights is appreciated U.S Government License Rights; Restricted Rights: The Software and its documentation is commercial computer software developed at private expense and is provided with RESTRICTED RIGHTS to the United States Government Use, duplication or disclosure of the Software by the United States Government is subject to the license terms of this Agreement pursuant to, as applicable, FAR 12.212, DFAR 227.7202-1(a), DFAR 227.7202-3(a) and DFAR 227.7202-4 and, to the extent required under U.S federal law, the minimum restricted rights as set out in FAR 52.227-19 (DEC 2007) If FAR 52.227-19 is applicable, this provision serves as notice under clause (c) thereof and no other notice is required to be affixed to the Software or documentation The Government's rights in Software and documentation shall be only those set forth in this Agreement SAS Institute Inc., SAS Campus Drive, Cary, North Carolina 275132414 December 2015 SAS® and all other SAS Institute Inc product or service names are registered trademarks or trademarks of SAS Institute Inc in the USA and other countries ® indicates USA registration Other brand and product names are trademarks of their respective companies Foreword Recent years, and perhaps particularly the past decade, have seen a rapid evolution in the statistical methodology available to be used in clinical trials, from both technical and implementation standpoints Certain practices as they might have been performed not too far into the past might in fact now seem somewhat primitive or naïve Much, but certainly by no means all, of the recent development is related to recent interest in adaptive trial designs The term itself is quite broad, and encompasses a wide variety of techniques and applications Many trial aspects are potential candidates for adaptation, including but not limited to: sample size or information requirements, dose or treatment regimen selection, targeted patient population selection, the randomization allocation scheme; and within each of these categories there may be multiple and fundamentally different technical and strategic approaches that are now available for practitioners to consider Classical procedures as well have undergone advancements in the statistical details of their implementation, and their usage in analysis and interpretation of trial results Enhancements in classical approaches, and the progress made or envisioned in utilization of novel adaptive and Bayesian designs and methodologies, are reflective of the current interest in the transition to personalized medicine approaches, by which optimal therapies corresponding to particular patient characteristics are sought A categorization of designs and methods into classical, adaptive, and Bayesian methods is by no means mutually exclusive, as a number of methodologies have aspects of more than one of these classes Just to cite one example, group sequential designs are a familiar feature in current clinical trial practice that fall under both the classical and adaptive headings; this is also certainly an area that has seen an evolution in recent years Aspects of clinical trial or program design such as dose finding or population enrichment may contain aspects that are adaptive, or Bayesian, or both, as is communicated well in this volume The interest in novel adaptive and Bayesian approaches certainly does not preclude the possibility that classical approaches will be preferred in many situations; they maintain the attributes which led to their widespread adoption in the first place As has been pointed out by many authors, the best use of these novel approaches will be realized by a full understanding of their behavior and an objective evaluation of their advantages and relevant tradeoffs in particular situations This point is clearly and objectively conveyed throughout this volume, as approaches of varied types are presented not to promote or endorse their casual routine use, but rather are described with sufficient explanations to help practitioners make the best choices for their situations, and of course to have the computational tools to implement them It seems inevitable that the availability to users of software and computational capabilities is inextricably linked with increased consideration of and interest in alternative design and analysis strategies, and ultimately their implementation Certainly, if a novel methodology is seen as adding value in such an important arena as clinical trials, it will spur development of the computational tools necessary to implement it However, in a cycle, the increased availability to practitioners leads to increased consideration and implementation, which spurs further interest, enables learnings from experience, perhaps motivates further research, and ultimately leads to further methodological and in-practice improvements and evolution Just as a simple illustration of this phenomenon: questions regarding how clinical sites should best be accounted for in main statistical analysis models had undergone some debate in past decades, with occasional flurries of literature activity, but evolution in conventional practices was limited The introduction of SAS’ proc mixed in the early 1990s provided a platform for more widespread consideration and usage of some approaches that were less commonly utilized at that time, which incorporated clinical site as a random effect in analysis models in various manners There were implications for important related issues, such as sample size determination and targeted centersize distributions, and for certain practices that were in use at the time such as small center pooling algorithms Given the presence of the new computational tool available to users in the form of the SAS procedure, it may not be a coincidence that by the latter part of that decade there was vigorous dialogue taking place in the literature on matters involving how best to design multicenter studies and accommodate center in analysis models, and within a relatively short period of time there were notable changes in conventional practices Given the extent of recent methodological advances, and the wide knowledge of and usage of SAS throughout the clinical trials community, a focused volume such as this one is particularly timely in this regard It integrates a broad yet coherent summary of current approaches for clinical trial design and analysis, with particular emphasis on important recently developed ones, along with specific illustrations of how they can be implemented and performed in SAS In some cases this involves relatively straightforward calls to SAS procedures; in many others, sophisticated SAS macros developed by the authors are presented Motivating examples are described, and SAS outputs corresponding to those examples are explained to help guide readers through the most accurate understandings and interpretations This text might well function effectively as a technical resource on state-of-the-art clinical trials methodology even if it did not contain the SAS illustrations and explanations; and it could also fit within a useful niche if it focused solely on the SAS illustrations without the methodological and practical explanations The fact that it contains both aspects, well integrated in chapters prepared by experienced subject matter experts, makes it a particularly valuable resource The ability that the material contained here offers to practitioners to test and compare different design and analysis options to choose the one that seems best for a given situation can help drive the most impactful usage of these new technologies; and, along the lines of the methodology-computational tools cycle described earlier, this perhaps may assist in leading to further experience-driven methodological or implementation advancements Paul Gallo Novartis October 2015 About This Book Purpose Modern Approaches to Clinical Trials Using SAS®: Classical, Adaptive, and Bayesian Methods is unique and multifaceted, covering several domains of modern clinical trial design, including classical, group sequential, adaptive, and Bayesian methods that are applicable to and widely used in various phases of pharmaceutical development Topics covered include, but are not limited to, dose-response and doseescalation designs; sequential methods to stop trials early for overwhelming efficacy, safety, or futility; Bayesian designs that incorporate historical data; adaptive sample size re-estimation; adaptive randomization to allocate subjects to more effective treatments; and population enrichment designs Methods are illustrated using clinical trials from diverse therapeutic areas, including dermatology, endocrinology, infectious disease, neurology, oncology, and rheumatology Individual chapters are authored by renowned contributors, experts, and key opinion leaders from the pharmaceutical/medical device industry or academia Numerous real-world examples and sample SAS code enable users to readily apply novel clinical trial design and analysis methodologies in practice Is This Book for You? This book is intended for biostatisticians, pharmacometricians, clinical developers, and statistical programmers involved in the design, analysis, and interpretation of clinical trials Further, students in graduate and post-graduate programs in statistics or biostatistics will benefit from the many practical illustrations of statistical concepts Prerequisites Based on the above audience, users will benefit most from this book with some graduate training in statistics or biostatistics, and some experience or exposure to clinical trials Some experience with simulation may be useful, though this is not required to use this book Some experience with SAS/STAT procedures, SAS/IML, and the SAS macro language is expected 10 progression-free survival (PFS) 2.3.4 Survival Endpoint 11.4.2.5 Flexible Adaptive Seamless (Enrichment) Design promising zone proof of concept (POF) Proschan, M.A 3.3.5 Methods Based on the Conditional Type I Error Principle 3.3.5 Methods Based on the Conditional Type I Error Principle [2] prospective design psoriasis 2.2 Examples of Classical Fixed-Sample Designs 2.4.3 Unified Family Boundaries 2.4.4 Spending Function Boundaries PSS option 2.4.3 Unified Family Boundaries 2.4.3 Unified Family Boundaries [2] 621 R randomized clinical trials (RCT) randomized play-the-winner rule (RPWR) 9.5.2 Randomized Play-the-Winner Rule 10.1.3.1 RAR Urn Models RAR (response-adaptive randomization) Different Types of Randomization Schemes 9.5.1 Play-the-Winner Rule 10.1.2 Response-Adaptive Randomization 10.4.1 Simulating a Single RAR Sequence 10.4.2 Operating Characteristics of RAR designs via Monte Carlo Simulation 10.5 Power and Sample Size for Response-Adaptive Randomization RAR Urn models RCT (randomized clinical trials) REPORT procedure required inputs, for %BSMED response-adaptive randomization (RAR) Different Types of Randomization Schemes 9.5.1 Play-the-Winner Rule 10.1.2 Response-Adaptive Randomization 10.4.1 Simulating a Single RAR Sequence 10.4.2 Operating Characteristics of RAR designs via Monte Carlo Simulation 10.5 Power and Sample Size for Response-Adaptive Randomization response-based adaptive randomizations restricted randomization Different Types of Randomization Schemes 9.3.1 Permuted Block Randomization retrospective design 622 Rho boundary 2.4.1 Calculation of Stopping Boundaries Methods of Deriving Stopping Boundaries 2.4.4 Spending Function Boundaries Robins, J.M Rosenberger, W.F 1.4 Emergence of Adaptive Designs in the 90s 9.5.3 Optimal Adaptive Randomization 10.1.2 Response-Adaptive Randomization 10.2 Optimal Allocation 10.3.1 Preliminaries 10.3.2 Sequential Maximum Likelihood Estimation Design 10.3.3 Doubly Adaptive Biased Coin Design 10.3.5 Theoretical Comparison of Various RAR Designs 10.5 Power and Sample Size for Response-Adaptive Randomization 10.6.4 Time Trends Royston, P RPWR (randomized play-the-winner rule) 9.5.2 Randomized Play-the-Winner Rule 10.1.3.1 RAR Urn Models RSIHR design 623 S sample size calculating considerations for optimal response-adaptive randomization designs sample size re-estimation (SSR) about based on promising zone blinded methods considerations and issues information-based design unblinded methods SAMPLESIZE statement 2.4.2 P-value or Haybittle-Peto Boundaries 2.4.3 Unified Family Boundaries 2.4.3 Unified Family Boundaries [2] 2.4.4 Spending Function Boundaries sampling priors Sargent, D SAWP (Scientific Advice Working Party) SBA (summary basis for approval) Schäfer, H 3.3.5 Methods Based on the Conditional Type I Error Principle 3.3.5 Methods Based on the Conditional Type I Error Principle [2] Scharfstein, D.O Scientific Advice Working Party (SAWP) score allocation selection impact curve SEQDESIGN procedure 624 2.4.2 P-value or Haybittle-Peto Boundaries 2.4.2 P-value or Haybittle-Peto Boundaries [2] 2.4.3 Unified Family Boundaries 2.4.4 Spending Function Boundaries 2.4.4 Spending Function Boundaries [2] 2.5 Special Issues 3.3.3.3 Example of the CHW Method 3.4 Information-Based Design 3.4 Information-Based Design [2] SEQTEST procedure 2.4.3 Unified Family Boundaries 2.4.3 Unified Family Boundaries [2] 2.4.3 Unified Family Boundaries [3] 2.4.4 Spending Function Boundaries sequential analysis, defined sequential maximum likelihood estimation design (SMLE) sequential multiple assignment randomized trials (SMARTs) sequential parallel comparison designs (SPCDs) sequential testing strategy design Shih, W.J Simon, R 11.2.2.3 Sequential Testing Strategy Design 11.3 Efficiency of Classical Enrichment Designs 11.3 Efficiency of Classical Enrichment Designs [2] 11.4.2.1 Adaptive Signature Design 11.4.3.2 Bayesian/Frequentist Design for Biomarker Treatment Interaction 11.4.3.2 Bayesian/Frequentist Design for Biomarker Treatment Interaction [2] simple randomization Different Types of Randomization Schemes 9.2 Simple Randomization 625 simulated data, analysis examples using SLE (systemic lupus erythematosus) sliding-window subgroup plot SMARTs (sequential multiple assignment randomized trials) SMLE (sequential maximum likelihood estimation design) software Song, Y Southwest Oncology Group (SWOG) SPCDs (sequential parallel comparison designs) special issues spending function boundaries spending methods SSR See sample size re-estimation (SSR) Stallard, N statistical design considerations, in classical dose-response study statistical hypothesis, for determining PoC statistical model, for Bayesian dose-response STOP=BOTH option stopping boundaries calculating 2.1 Introduction 2.4.1 Calculation of Stopping Boundaries types of stratified randomization 626 Sugitani, T summary basis for approval (SBA) survival endpoint Sutton, A.J SWOG (Southwest Oncology Group) systemic lupus erythematosus (SLE) 627 T tail-oriented subgroup plot Tanaka, Y Tao, A Taylor's approximation Thall, P.F Thomas, N 1.5.4 Opportunities in the Learning Phase 7.3 Considerations for MCP-Mod at the Design Stage 7.4 Further Considerations on MCP-Mod Thompson, W.R three cohort design time trends, optimal response-adaptive randomization designs and Time-to-Event Continual Reassessment Method (TITE-CRM) Todd, S trials, logistics triangular test 2.1 Introduction 2.4.1 Calculation of Stopping Boundaries Methods of Deriving Stopping Boundaries Tsiatis, A.A 1.3 Emergence of Group Sequential Designs in the 70s and 80s 2.3.1 General Framework 3.4 Information-Based Design 3.4 Information-Based Design [2] Turnbull, B.W 1.3 Emergence of Group Sequential Designs in the 70s and 80s 2.1 Introduction 628 2.5 Special Issues two cohort sequential design TWOSAMPLEFREQ option TWOSAMPLEMEAN option Tymofyeyev, Y 1.5.5 Software 10.6.1 More Than Two Treatments and Non-Binary Outcomes Type diabetes mellitus (T2DM) Type I errors calculating choice of classification of methods to control Type II errors, choice of 629 U unblinded SSR methods unified family (power) methods 2.1 Introduction 2.4.1 Calculation of Stopping Boundaries Methods of Deriving Stopping Boundaries 2.4.3 Unified Family Boundaries 630 V Vandemeulebroecke, M 631 W Wakana, A Wald test 2.3.1 General Framework 3.3.3.1 Overview 10.2 Optimal Allocation 10.2 Optimal Allocation [2] 10.2 Optimal Allocation [3] 10.3.2 Sequential Maximum Likelihood Estimation Design 10.4.2 Operating Characteristics of RAR designs via Monte Carlo Simulation Wang-Tsiatis boundaries 2.4.1 Calculation of Stopping Boundaries Methods of Deriving Stopping Boundaries 2.4.3 Unified Family Boundaries Wang, F Wang, S.J 3.3.3.1 Overview 3.3.3.2 CHW Procedure as a Special Case of the Combination Test 3.3.3.3 Example of the CHW Method 11.2.2.3 Sequential Testing Strategy Design 11.4.3.2 Bayesian/Frequentist Design for Biomarker Treatment Interaction Wang, X Wathen, J.K Wei, L.J 9.3.2 Biased Coin Design 9.5.2 Randomized Play-the-Winner Rule 10.1.3.1 RAR Urn Models Weibull distributions WEIGHT=1 option 632 Western Ontario and McMaster University (WOMAC) Osteoarthritis index Whitehead, J Woodcock, J Wouter, W 633 Z Zhang, L.X zidovudine (AZT) clinical trial 634 Gain Greater Insight into Your SAS® Software with SAS Books Discover all that you need on your journey to knowledge and empowerment support.sas.com/bookstore for additional books and resources © 2013 SAS Institute Inc All rights reserved S107969US.0413 635 ... follows: SAS Institute Inc 2015 Modern Approaches to Clinical Trials Using SAS? ?: Classical, Adaptive, and Bayesian Methods Cary, NC: SAS Institute Inc Modern Approaches to Clinical Trials Using SAS? ?:... October 2015 About This Book Purpose Modern Approaches to Clinical Trials Using SAS? ?: Classical, Adaptive, and Bayesian Methods is unique and multifaceted, covering several domains of modern clinical. .. Carolina: SAS Institute Inc Wicklin R (2 01 3) Simulating Data with SAS? ? Cary, North Carolina: SAS Institute Inc Zink RC (2 01 4) Risk-Based Monitoring and Fraud Detection in Clinical Trials Using JMP® and