Modern adaptive randomized clinical trials

Modern Adaptive Randomized Clinical Trials Statistical and Practical Aspects © 2016 by Taylor & Francis Group, LLC K23296_FM.indd 6/8/15 12:08 PM Editor-in-Chief Shein-Chung Chow, Ph.D., Professor, Department of Biostatistics and Bioinformatics, Duke University School of Medicine, Durham, North Carolina Series Editors Byron Jones, Biometrical Fellow, Statistical Methodology, Integrated Information Sciences, Novartis Pharma AG, Basel, Switzerland Jen-pei Liu, Professor, Division of Biometry, Department of Agronomy, National Taiwan University, Taipei, Taiwan Karl E Peace, Georgia Cancer Coalition, Distinguished Cancer Scholar, Senior Research Scientist and Professor of Biostatistics, Jiann-Ping Hsu College of Public Health, Georgia Southern University, Statesboro, Georgia Bruce W Turnbull, Professor, School of Operations Research and Industrial Engineering, Cornell University, Ithaca, New York Published Titles Adaptive Design Methods in Clinical Trials, Second Edition Shein-Chung Chow and Mark Chang Adaptive Designs for Sequential Treatment Allocation Alessandro Baldi Antognini and Alessandra Giovagnoli Adaptive Design Theory and Implementation Using SAS and R, Second Edition Mark Chang Advanced Bayesian Methods for Medical Test Accuracy Lyle D Broemeling Advances in Clinical Trial Biostatistics Nancy L Geller Applied Meta-Analysis with R Ding-Geng (Din) Chen and Karl E Peace Basic Statistics and Pharmaceutical Statistical Applications, Second Edition James E De Muth Bayesian Adaptive Methods for Clinical Trials Scott M Berry, Bradley P Carlin, J Jack Lee, and Peter Muller Bayesian Analysis Made Simple: An Excel GUI for WinBUGS Phil Woodward Bayesian Methods for Measures of Agreement Lyle D Broemeling Bayesian Methods in Epidemiology Lyle D Broemeling Bayesian Methods in Health Economics Gianluca Baio Bayesian Missing Data Problems: EM, Data Augmentation and Noniterative Computation Ming T Tan, Guo-Liang Tian, and Kai Wang Ng Bayesian Modeling in Bioinformatics Dipak K Dey, Samiran Ghosh, and Bani K Mallick Benefit-Risk Assessment in Pharmaceutical Research and Development Andreas Sashegyi, James Felli, and Rebecca Noel Biosimilars: Design and Analysis of Follow-on Biologics Shein-Chung Chow Biostatistics: A Computing Approach Stewart J Anderson Causal Analysis in Biomedicine and Epidemiology: Based on Minimal Sufficient Causation Mikel Aickin Clinical and Statistical Considerations in Personalized Medicine Claudio Carini, Sandeep Menon, and Mark Chang © 2016 by Taylor & Francis Group, LLC K23296_FM.indd 6/8/15 12:08 PM Clinical Trial Data Analysis using R Ding-Geng (Din) Chen and Karl E Peace Clinical Trial Methodology Karl E Peace and Ding-Geng (Din) Chen Computational Methods in Biomedical Research Ravindra Khattree and Dayanand N Naik Computational Pharmacokinetics Anders Källén Confidence Intervals for Proportions and Related Measures of Effect Size Robert G Newcombe Controversial Statistical Issues in Clinical Trials Shein-Chung Chow Data Analysis with Competing Risks and Intermediate States Ronald B Geskus Data and Safety Monitoring Committees in Clinical Trials Jay Herson Design and Analysis of Animal Studies in Pharmaceutical Development Shein-Chung Chow and Jen-pei Liu Design and Analysis of Bioavailability and Bioequivalence Studies, Third Edition Shein-Chung Chow and Jen-pei Liu Design and Analysis of Bridging Studies Jen-pei Liu, Shein-Chung Chow, and Chin-Fu Hsiao Design and Analysis of Clinical Trials for Predictive Medicine Shigeyuki Matsui, Marc Buyse, and Richard Simon Design and Analysis of Clinical Trials with Time-to-Event Endpoints Karl E Peace Design and Analysis of Non-Inferiority Trials Mark D Rothmann, Brian L Wiens, and Ivan S F Chan Difference Equations with Public Health Applications Lemuel A Moyé and Asha Seth Kapadia DNA Methylation Microarrays: Experimental Design and Statistical Analysis Sun-Chong Wang and Arturas Petronis DNA Microarrays and Related Genomics Techniques: Design, Analysis, and Interpretation of Experiments David B Allison, Grier P Page, T Mark Beasley, and Jode W Edwards Dose Finding by the Continual Reassessment Method Ying Kuen Cheung Elementary Bayesian Biostatistics Lemuel A Moyé Empirical Likelihood Method in Survival Analysis Mai Zhou Frailty Models in Survival Analysis Andreas Wienke Generalized Linear Models: A Bayesian Perspective Dipak K Dey, Sujit K Ghosh, and Bani K Mallick Handbook of Regression and Modeling: Applications for the Clinical and Pharmaceutical Industries Daryl S Paulson Inference Principles for Biostatisticians Ian C Marschner Interval-Censored Time-to-Event Data: Methods and Applications Ding-Geng (Din) Chen, Jianguo Sun, and Karl E Peace Introductory Adaptive Trial Designs: A Practical Guide with R Mark Chang Joint Models for Longitudinal and Timeto-Event Data: With Applications in R Dimitris Rizopoulos Measures of Interobserver Agreement and Reliability, Second Edition Mohamed M Shoukri Medical Biostatistics, Third Edition A Indrayan © 2016 by Taylor & Francis Group, LLC K23296_FM.indd 6/8/15 12:08 PM Meta-Analysis in Medicine and Health Policy Dalene Stangl and Donald A Berry Mixed Effects Models for the Population Approach: Models, Tasks, Methods and Tools Marc Lavielle Modeling to Inform Infectious Disease Control Niels G Becker Modern Adaptive Randomized Clinical Trials: Statistical and Practical Aspects Oleksandr Sverdlov Monte Carlo Simulation for the Pharmaceutical Industry: Concepts, Algorithms, and Case Studies Mark Chang Multiple Testing Problems in Pharmaceutical Statistics Alex Dmitrienko, Ajit C Tamhane, and Frank Bretz Noninferiority Testing in Clinical Trials: Issues and Challenges Tie-Hua Ng Optimal Design for Nonlinear Response Models Valerii V Fedorov and Sergei L Leonov Patient-Reported Outcomes: Measurement, Implementation and Interpretation Joseph C Cappelleri, Kelly H Zou, Andrew G Bushmakin, Jose Ma J Alvir, Demissie Alemayehu, and Tara Symonds Quantitative Evaluation of Safety in Drug Development: Design, Analysis and Reporting Qi Jiang and H Amy Xia Randomized Clinical Trials of Nonpharmacological Treatments Isabelle Boutron, Philippe Ravaud, and David Moher Randomized Phase II Cancer Clinical Trials Sin-Ho Jung Sample Size Calculations for Clustered and Longitudinal Outcomes in Clinical Research Chul Ahn, Moonseong Heo, and Song Zhang Sample Size Calculations in Clinical Research, Second Edition Shein-Chung Chow, Jun Shao and Hansheng Wang Statistical Analysis of Human Growth and Development Yin Bun Cheung Statistical Design and Analysis of Stability Studies Shein-Chung Chow Statistical Evaluation of Diagnostic Performance: Topics in ROC Analysis Kelly H Zou, Aiyi Liu, Andriy Bandos, Lucila Ohno-Machado, and Howard Rockette Statistical Methods for Clinical Trials Mark X Norleans Statistical Methods for Drug Safety Robert D Gibbons and Anup K Amatya Statistical Methods in Drug Combination Studies Wei Zhao and Harry Yang Statistics in Drug Research: Methodologies and Recent Developments Shein-Chung Chow and Jun Shao Statistics in the Pharmaceutical Industry, Third Edition Ralph Buncher and Jia-Yeong Tsay Survival Analysis in Medicine and Genetics Jialiang Li and Shuangge Ma Theory of Drug Development Eric B Holmgren Translational Medicine: Strategies and Statistical Methods Dennis Cosmatos and Shein-Chung Chow © 2016 by Taylor & Francis Group, LLC K23296_FM.indd 6/8/15 12:08 PM Modern Adaptive Randomized Clinical Trials Statistical and Practical Aspects Edited by Oleksandr Sverdlov EMD Serono USA © 2016 by Taylor & Francis Group, LLC K23296_FM.indd 6/8/15 12:08 PM CRC Press Taylor & Francis Group 6000 Broken Sound Parkway NW, Suite 300 Boca Raton, FL 33487-2742 © 2016 by Taylor & Francis Group, LLC CRC Press is an imprint of Taylor & Francis Group, an Informa business No claim to original U.S Government works Version Date: 20150515 International Standard Book Number-13: 978-1-4822-3989-8 (eBook - PDF) This book contains information obtained from authentic and highly regarded sources Reasonable efforts have been made to publish reliable data and information, but the author and publisher cannot assume responsibility for the validity of all materials or the consequences of their use The authors and publishers have attempted to trace the copyright holders of all material reproduced in this publication and apologize to copyright holders if permission to publish in this form has not been obtained If any copyright material has not been acknowledged please write and let us know so we may rectify in any future reprint Except as permitted under U.S Copyright Law, no part of this book may be reprinted, reproduced, transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying, microfilming, and recording, or in any information storage or retrieval system, without written permission from the publishers For permission to photocopy or use material electronically from this work, please access www.copyright.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc (CCC), 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400 CCC is a not-for-profit organization that provides licenses and registration for a variety of users For organizations that have been granted a photocopy license by the CCC, a separate system of payment has been arranged Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe Visit the Taylor & Francis Web site at http://www.taylorandfrancis.com and the CRC Press Web site at http://www.crcpress.com © 2016 by Taylor & Francis Group, LLC Contents Preface xi Contributors xv I Introduction 1 An Overview of Adaptive Randomization Designs in Clinical Trials Oleksandr Sverdlov II Restricted Randomization 45 Efron’s Biased Coin Design Revisited: Statistical Properties, Randomization-Based Inference and Sequential Monitoring 47 Victoria Plamadeala Adaptive Biased Coins: Achieving Better Balance without Compromising Randomness 55 Alessandro Baldi Antognini and Maroussa Zagoraiou Brick Tunnel and Wide Brick Tunnel Randomization for Studies with Unequal Allocation 83 Olga M Kuznetsova and Yevgen Tymofyeyev III Covariate–Adaptive Randomization 115 Development of Novel Covariate–Adaptive Randomization Designs Wenle Zhao 117 Optimal Model-Based Covariate–Adaptive Randomization Designs Anthony Atkinson 131 Statistical Inference Following Covariate–Adaptive Randomization: Recent Advances D Stephen Coad 155 vii © 2016 by Taylor & Francis Group, LLC viii Contents Covariate–Adaptive Randomization with Unequal Allocation Olga M Kuznetsova1 and Yevgen Tymofyeyev2 IV Response–Adaptive Randomization 171 199 Optimal Allocation Designs for a Multi-Arm Multi-Objective Clinical Trial 201 David Azriel 10 Response–Adaptive Randomization: An Overview of Designs and Asymptotic Theory 221 Li-Xin Zhang 11 Statistical Inference Following Response–Adaptive Randomization 251 Yanqing Yi and Xikui Wang 12 Sample Size Re-Estimation in Adaptively Randomized Clinical Trials with Missing Data Ruitao Lin and Guosheng Yin 269 13 Some Caveats for Outcome Adaptive Randomization in Clinical Trials Peter F Thall1 , Patricia S Fox1 and J Kyle Wathen2 287 V Covariate-Adjusted Response–Adaptive Randomization 14 Efficient and Ethical Adaptive Clinical Trial Designs to Detect Treatment–Covariate Interaction Seung Won Hyun1 , Tao Huang2 and Hongjian Zhu3 15 Longitudinal Covariate-Adjusted Response–Adaptive Randomization: Impact of Missing Data 307 309 327 Tao Huang and Hongjian Zhu 16 Targeted Covariate-Adjusted Response–Adaptive LASSO-Based Randomized Controlled Trials Antoine Chambaz1 , Mark J van der Laan2 and Wenjing Zheng2,3 17 Covariate-Balanced Bayesian Adaptive Randomization: Achieving Tradeoff between Inferential and Ethical Goals in Small and Moderate Size Trials Ying Yuan and Jing Ning © 2016 by Taylor & Francis Group, LLC 345 371 Contents VI ix Randomized Designs with Treatment Selection 387 18 Multi-Arm Multi-Stage Designs for Clinical Trials with Treatment Selection James Wason 389 19 Sequential Elimination in Multi-Arm Selection Trials 411 Christina Yap1 , Xuejing Lin2 and Ying Kuen K Cheung2 20 Accounting for Parameter Uncertainty in Two-Stage Designs for Phase II Dose–Response Studies 427 Emma McCallum1 and Bjă orn Bornkamp2 VII Application and Practical Aspects 451 21 A Single Pivotal Adaptive Trial in Infants with Proliferating Hemangioma: Rationale, Design Challenges, Experience and Recommendations 453 Stephane Heritier1 , Caroline C Morgan-Bouniol2 , Serigne N Lˆ o3 , Stephanie Gautier4 and Jean Jacques Voisard5 22 Practical Implementation of Dose–Response Adaptive Trials Tom Parke and Martin Kimber 483 23 Statistical Monitoring of Data in Response–Adaptive Randomized Clinical Trials Paul Gallo 505 © 2016 by Taylor & Francis Group, LLC © 2016 by Taylor & Francis Group, LLC Practical Implementation of Dose–Response Adaptive Trials 501 One of the common problems of clinical trials, that of accrual, has a bearing on adaptive trials If the accrual rate is unexpectedly slow, there may be a decision to delay interims or make them less frequent If the accrual rate is unexpectedly fast, there is a danger that there will be less scope for adaptation than planned This is one of those occasions where the initial step is to run some new simulations to understand the effect of the unexpected accrual rate If these results show that there will be a problem, then these will be the basis for closing some sites to return the accrual to the planned rate If the accrual rate cannot be slowed, then minor modifications to the adaptive design might be made, such as adjusting the stopping criteria and the degree of adaptation—planned and justified with further simulations 22.3.5 Third Parties: The Multiple CRO Ecosystem It is common in our experience for adaptive trials to use a combination of Contract Research Organizations (CROs) or a combination of CROs and sponsor teams and this has rarely been a problem The bulk of the work of EDC, central randomization and trial supplies remains unchanged The changes are that EDC systems need to be able to provide the data extraction for the interims, central randomization systems need to be able to implement the changes in randomization, and clinical supply functions need to be able to cope with the additional uncertainties stemming from the adaptive nature of the trial The bulk of the new processes to support the adaptive trials lie in the three new teams: the trial design and simulation team, the ISC, and the DMC One of the biggest problems for the established teams is typically not the degree of difference from normal, but that there is some difference from normal The data collection, data management, central randomization, and trial supply teams will have become well practiced and highly efficient at executing the standard type of trial, the fixed trial As well as accepting changes in process, the teams need to accept some changes in objectives Sometimes this can be seen as raising their game (speed of data collection) and sometimes this might be seen as having to accept the antithesis of what they have come to see as good practice, such as providing uncleaned data, not accruing as fast as possible, or having a greater overage of supplies at the end of the trial It is this need to change well-established processes, the disruption of high-performing teams, and changing “the rules” of what constitutes a wellexecuted process that can cause reluctance in these teams to engage in supporting adaptive trials One of the beneficial effects of involving these teams in the review of some of the simulation results during the planning stage is that they can gain an appreciation of the overall benefits that the adaptive design brings to the drug development program and to see that these far outweigh the additional costs to their activities It also provides them an opportunity to input possible constraints on the design to make the trial cheaper or less risky © 2016 by Taylor & Francis Group, LLC 502 22.3.6 Modern Adaptive Randomized Clinical Trials Managing the Risks in an Adaptive Design The main risks that must be guarded against in executing a dose–response adaptive design are failing to collect the required data in time, errors in sending it to the interim analysis team, errors in running the interim analysis, failing to keep to the pre-planned adaptation and not having the right supplies at the centres as the trial adapts The data collection needs active monitoring, with rapid follow-up at sites where data are late or missing, but with a particular emphasis on the endpoint data required for the interim analysis and a slightly more relaxed attitude to the rest of the data being collected Having a two-tier approach makes it easier for centers to catch up and submit key data that have become late, and avoids centers becoming frustrated with being vigorously chased for data that they might feel are peripheral to the main aims of the trial The data submission to the interim analysis team, like the interim analysis itself, should be automated as far as is possible, tested before the trial starts and tested again once the trial is running but before a “real” interim is required By automating and testing the data submission and the performance of the interim analysis, the risks of human error are minimized and the dependence on particular key individuals to perform the interim can be removed The risk of deviating from the planned adaptations is managed by the DMC, and this should be clear in their charter The DMC should be familiar with the trial design, and either include members of the design team or they should have “bought in to the design” through involvement in reviews of the simulations during the planning stage Applying the adaptations is of course key to the trial achieving its objectives, but is not a particularly complex task So though the impact of the risk is high, it is usually regarded as very unlikely and it is usually regarded as sufficient for the DMC to monitor that the adaptations have been applied rather than taking additional steps to ensure it happens Any risk of being unable to supply the adaptations is very dependent on the circumstances of the trial In many trials this has not been a risk at all— supplies have been plentiful throughout the trial and the costs of oversupply minor compared to the benefits of the trial If supplies are limited or initial supply limited, supply to a particular region limited, or supplies for a particular arm limited, then sometimes the design can take this into account Limiting the number of centers, capping the allocation to a particular arm, dropping slow recruiting centers and re-allocating supplies are all tactics that can be used to manage supply constraints It is important to include consideration of these limitations during the planning and initial simulation stage, both to allow minor design changes that can reduce the risks, and to avoid the supply team feeling ignored, taken for granted and being set up to fail Lastly, really difficult supply situations can be understood in detail by using clinical trial simulation software in the planning stage and possibly overcome © 2016 by Taylor & Francis Group, LLC Practical Implementation of Dose–Response Adaptive Trials 503 using adaptive re-supply rules in the execution of the trial, including possibly incorporating the new adaptive randomization ratios from the interim analysis in forecasting site requirements as accurately as possible 22.4 In Summary: A Challenge Worth Taking The success of drug development is still low, and the costs are eye-wateringly high The use of adaptive trials, particularly highly adaptive phase II trials to learn better about a drug and its optimal use before phase III, still appears to be the best researched improvement to the process Despite this, the prevalence of adaptive trials, particularly to study a wide dose range in phase II, is currently still quite limited A number of surveys have tried to identify what the key obstacle is to the greater use of adaptive designs We suspect that there is not one key obstacle but the fact that adaptive trials pose challenges across many of the teams involved in implementation is the problem [5] We hope that this account of our experience, particularly in the last 10 years of 30 or so trials with over 10 different sponsors, will further demystify how these trials are different from fixed trials and embolden teams to use them more widely Bibliography [1] European Medicines Agency Committee for Medicinal Products for Human Use (CHMP) (2007) Reflection paper on methodological issues in confirmatory clinical trials planned with an adaptive design CHMP/EWP/2459/02 [2] Fardipour, P., Littman, G., Burns, D D., Dragalin, V., Padmanabhan, S K., Parke, T., Perevozskaya, I., Reinold, K., Sharma, A., Krams, M (2009) Planning and executing response–adaptive learn-phase clinical trials: The process Drug Information Journal 43(6), 713–723 [3] Food and Drug Administration (FDA) (2010) Guidance for industry Adaptive design clinical trials for drugs and biologics (draft document) [4] Gallo, P., Chuang-Stein, C., Dragalin, V., Gaydos, B., Krams, M., Pinheiro, J (2006) Adaptive designs in clinical drug development—An executive summary of the PhRMA Working Group Journal of Biopharmaceutical Statistics 16(3), 275–283 © 2016 by Taylor & Francis Group, LLC 504 Modern Adaptive Randomized Clinical Trials [5] Gaydos B., Anderson, K M., Berry, D., Burnham, N., Chuang-Stein, C., Dudinak, J., Fardipour, P., Gallo, P., Givens, S., Lewis, R., Maca, J., Pinheiro, J., Pritchett, Y., Krams, M (2009) Good practices for adaptive clinical trials in pharmaceutical product development Drug Information Journal 43, 539–556 [6] Getz K., Stergiopoulos, S., Kim, J Y (2013) The adoption and impact of adaptive trial designs Tufts Center for the Study of Drug Development, Tufts University [7] Krams, M., Lees, K R., Hacke, W., Grieve, A P., Orgogozo, J M., Ford, G A (2003) Acute Stroke Therapy by Inhibition of Neutrophils (ASTIN): An adaptive dose–response study of UK-279,276 in acute ischemic stroke Stroke 34, 2543–2548 [8] Packer, M., Bristow, M R., Cohn, J N., Colucci, W S., Fowler, M B., Gilbert, E M., Shusterman, N H (1996) The effect of carvedilol on morbidity and mortality in patients with chronic heart failure U S Carvedilol Heart Failure Study Group New England Journal of Medicine 334(21), 1349–1355 [9] Shen, J., Preskorn, S., Dragalin, V., Slomkowski, M., Padmanabhan, S K., Fardipour, P., Sharma, A., Krams, M (2011) How adaptive trial designs can increase efficiency in psychiatric drug development: a case study Innovations in Clinical Neuroscience 8(7), 26–34 © 2016 by Taylor & Francis Group, LLC 23 Statistical Monitoring of Data in Response–Adaptive Randomized Clinical Trials Paul Gallo Novartis Pharmaceuticals CONTENTS 23.1 23.2 23.3 23.4 23.5 23.1 Introduction Interim Analysis Motivations in Adaptive Randomization Trials Interim Monitoring Confidentiality Concerns Interim Monitoring Issues in Adaptive Randomization Trials 23.4.1 Exploratory Trials 23.4.2 Confirmatory Development: Seamless Phase II/III Trials 23.4.3 Observing Adaptations and Reverse Engineering Summary and Conclusion Bibliography 505 507 508 509 509 510 512 513 514 Introduction One of the critical operational issues associated with implementation of the adaptive designs discussed throughout this volume involves the process by which accruing data are collected, evaluated, and reviewed in order to make adaptations This issue arises in any adaptive design, and certainly in trials where there is potential for adaptive randomization Making mid-trial adaptations extends a process of interim data monitoring and decision making that has become quite familiar in recent clinical trials practice Very commonly, accruing data are examined with a main goal of deciding whether or not a study should continue—perhaps trial objectives have been achieved based upon the results reaching a formal group sequential boundary, or a study might be stopped for futility if it seems clear that it will not meet its objectives, or might be terminated because of an unacceptable safety risk Even in these cases where the envisioned action is conceptually 505 © 2016 by Taylor & Francis Group, LLC 506 Modern Adaptive Randomized Clinical Trials simple, appropriately implementing an interim monitoring plan can be quite challenging One aspect of concern is the possibility that knowledge of interim trial results among trial participants or those managing the trial might compromise objective management of the trial, or introduce biases into the trial conduct and its final results There is particular sensitivity to this concern in trials that aim to provide confirmatory evidence of effectiveness and be the basis for product approvals; but of course in any trial avoidance of bias is desirable These issues are described in depth in an FDA guidance document [14] In trials with an adaptive randomization plan, interim monitoring is done for an intention different than simply deciding whether to stop or continue the trial: data are examined to decide upon changes which will govern a fundamental aspect of the trial—namely, the randomization scheme—as it continues As in other adaptive trial settings, this raises questions that must be carefully addressed and issues that must be carefully implemented in order not to compromise the interpretability of the results: • What are the relevant data flow and data preparation processes? • Which individuals will review the interim results to make the adaptive randomization recommendation or decision? • Must a pre-specified algorithm be rigidly adhered to or the decision makers have flexibility based on unanticipated study issues or outcome patterns? • Does access to unblinded interim results need to be carefully restricted? • What unblinded information needs to be communicated, and to whom, in order to implement the adaptation? • Even if the specific unblinded results remain tightly restricted, could knowledge of the adaptation that was implemented indirectly convey information about those results? (and if so, how much of a concern is this)? These issues have been prominently mentioned in regulatory guidance documents pertaining to adaptive trials [5, 15] as well as discussed in recent literature [1, 6, 8–10, 13] Most of this discussion has taken place in the context of confirmatory studies, or “adequate and well-controlled” trials according to the terminology of [15] Among the types of adaptive randomization designs discussed in this volume, some and some not fall into this category In this chapter, we refer to principles that are cited more generally in some of the references mentioned above, and discuss some of these issues more specifically in the context of their application to adaptive randomization trials © 2016 by Taylor & Francis Group, LLC Statistical Monitoring of Response–Adaptive Randomized Clinical Trials 507 23.2 Interim Analysis Motivations in Adaptive Randomization Trials The uses of adaptive randomization designs span different stages of drug development, including both exploratory and confirmatory trials This will have implications for how interim analysis and data monitoring processes are implemented in particular trials, as we will see Early development of adaptive randomization studies often aim to restrict the range of feasible dosages in an efficient manner and lead to selection of a smaller number of doses within a narrower range for subsequent further studies Adaptive dose-ranging studies may change the randomization allocation across doses based on interim results according to a pre-specified plan to most efficiently estimate the dose–response curve, or a particular quantity of interest (e.g., an ED50) Studies might drop dose arms which seem non-viable for further development, or for which it is judged in the context of the current trial that further information is not required Dose arms can be added if the nature of the treatment allows it and if according to plan it is judged that this will help achieve the study objectives Depending on the details and objectives of a particular study, the number of times that data are reviewed for making adaptations might range from a single timepoint to many adaptation points (potentially even on a continuous basis, that is, with the randomization scheme changing after each patient response) Dose-ranging studies often primarily meet their objectives through modeling approaches, rather than arriving at conclusions about a particular dose based solely on data from that dose A special case of adaptive randomization with confirmatory ramifications has been commonly referred to under the term seamless phase II/III design; see, for example, Maca et al [12] These studies generally would be implemented at a point of development where an experimental treatment is viewed as ready for a definitive comparison versus a control, but there remains some uncertainty regarding the optimal dose Such designs generally start out with a small number of doses of the investigation treatment (perhaps 2–5), and at a selection point, a subset of those doses is chosen to continue in the trial based on the results so far In inferentially seamless designs, data from the selected doses both before and after the selection point are included in the main analyses Conceptually, such a trial may be viewed as similar in intent to a phase III trial that includes multiple dose arms with appropriate multiplicity adjustment to control type I error rate However, in a seamless trial, poorly performing arms might be dropped for various motivations (e.g., saving resources, completing the trial sooner, exposing fewer patients to ineffective treatments) A description of a particular example of a seamless trial in a confirmatory setting can be found in Barnes et al [2] © 2016 by Taylor & Francis Group, LLC 508 Modern Adaptive Randomized Clinical Trials At any stage of development, in addition to changes in randomization allocation, the interim monitoring can also incorporate some of the more traditional or familiar interim monitoring features For example, a trial might be stopped for futility if the interim data suggest that the trial will not achieve its objectives (e.g., perhaps no doses will demonstrate sufficiently favorable results), or if safety risks are found that would make it unethical to continue Other adaptations could potentially be incorporated within a plan, for example, sample size re-assessment if the data showed higher variability than originally assumed 23.3 Interim Monitoring Confidentiality Concerns Concerns about confidentiality of interim results in order to avoid biasing influences originally arose, and are viewed as most relevant, for confirmatory studies that attempt to provide definitive evidence of product efficacy and safety These pivotal trials aim to be an important part of the basis by which treatments receive regulatory approval and reach the marketplace We now briefly review these current conventions and their motivation It is common practice in confirmatory trials that access to interim results and unblinded data should be carefully restricted, and in particular, not available to trial management personnel, investigators, or other study participants The rationale for the regulatory viewpoints underlying these conventions is well described in an FDA guidance document [14]; other relevant references include [4, 7, 11] Main points of concern can be summarized as follows: • Trial leadership personnel have various types of decisions to make regarding the management of the conduct of an ongoing trial based on objective scientific reasoning Access to interim results diminishes their ability to make certain decisions in a manner that can be seen to be totally objective • Knowledge of interim results by trial personnel (e.g., investigators and their staff) could introduce subtle, unknown biases into the conduct of the trial and the study results, perhaps causing changes in characteristics of the patients recruited, specific details of administration of the intervention or concomitant therapies, assessment of endpoints, etc On the basis of such concerns, it has become common practice to address the familiar interim monitoring objectives through the use of a Data Monitoring Committee (DMC), a group of experts possessing experience and expertise required to perform the intended monitoring responsibilities In confirmatory trials, DMC members usually play no role within the trial other than to perform their monitoring functions, and are typically external to the trial sponsor organization to maximize their independence and objectivity Access to unblinded study data and results is restricted to the DMC, and a small set of © 2016 by Taylor & Francis Group, LLC Statistical Monitoring of Response–Adaptive Randomized Clinical Trials 509 individuals providing statistical and programming support to them, until such time as the DMC undertakes a major action, such as a recommendation to terminate In exploratory stages of drug development, the use of independent DMCs is far less frequent As described in [14], in certain phase I or early phase II studies, any needed monitoring is often adequately provided by individuals internal to the study sponsor and/or investigator Exploratory trials are not the basis by which treatments will definitively demonstrate their merits and reach the marketplace Generally the amount of information they provide would not be sufficient to provide such evidence; their aim is to produce information and answer questions to a sufficient degree to justify whether and how a product should proceed through further development Statistical rigor (for example, tight type I error control) may not be a priority in certain early development trials Nevertheless, a DMC could in some circumstances help provide independent expert counsel, enhancing the safety of study participants and the credibility of the product development And of course, avoidance of biasing influences is desirable in any clinical investigation The need for confidentiality and the possible use of a DMC in exploratory trials should be evaluated caseby-case, but often these studies will legitimately not require nearly the degree of independence of interim monitoring such as associated with practices in confirmatory trials 23.4 Interim Monitoring Issues in Adaptive Randomization Trials As we extend from the more familiar monitoring motivations to interim decision making such as described in Section 23.2 for adaptive randomization trials, it is natural to consider what similarities and differences from current conventions and practices in non-adaptive trials might be warranted For example, is it important to maintain confidentiality of interim data and results, and should the party reviewing those results to make the adaptation decision be a similarly constituted DMC? Or if not, then who should perform this review? Not surprisingly, decisions in particular studies will be situationspecific and will depend on the nature of the trial and the stage of product development 23.4.1 Exploratory Trials As mentioned in the previous section, in non-adaptive exploratory trials it is infrequent that independent DMCs are viewed as being needed, and often any necessary monitoring perspective can be provided by personnel who are © 2016 by Taylor & Francis Group, LLC 510 Modern Adaptive Randomized Clinical Trials not independent of the sponsor or investigator, or even without other responsibilities in the trial In exploratory adaptive randomization trials there is frequently an additional motivation for the involvement of trial personnel in handling of unblinded accruing data: for the adaptive design to achieve its desired efficiencies, it is important that data be collected, processed, and acted upon quickly Setting up the type of infrastructure typically needed for DMCs or other independent bodies could well compromise the ability of the trial to achieve its intended efficiencies In addition, sponsor or investigator personnel might possess the most relevant knowledge of the trial and the adaptation plan and be better positioned to make optimal decisions In deciding whether independent monitoring might be warranted in exploratory adaptive randomization trials, there are a number of study aspects that could be considered that might tend to argue in one direction or the other: • Placement in the development program: A phase II trial to identify doses for phase III could potentially play a very strong supportive role in a regulatory submission if the results are sufficiently convincing The potential for a phase II trial to be strongly supportive in a submission could lead to independent monitoring being considered, to enhance the interpretability of its results • Safety and ethical concerns: Potential serious safety issues associated with a treatment or the inclusion of a frail population of high-risk patients could justify including independent experts in the interim review to better ensure objectivity in decision making relative to patient welfare • Nature of the endpoints: Endpoints that are short-term and “hard” (that is, straightforward to assess objectively) would tend to decrease the motivation for independent review, as there would be lessened potential for biasing influences to operate (Note: This applies to both main study endpoints and the endpoints that are the basis for the randomization change; these are often, but not always, the same.) • Nature of the adaptive plan: For designs that achieve their efficiencies through frequent ongoing review and decisions (play-the-winner schemes being an illustrative special case), the infrastructure required for independent review might compromise the advantages that the design offers 23.4.2 Confirmatory Development: Seamless Phase II/III Trials A standard illustration of an adaptive randomization trial in a pivotal setting is the so-called seamless phase II/III design [12] Studies with this type of design aim to provide confirmatory evidence of product effectiveness, so the concerns mentioned previously about the integrity of trial results and the © 2016 by Taylor & Francis Group, LLC Statistical Monitoring of Response–Adaptive Randomized Clinical Trials 511 avoidance of biasing influences are fully relevant as in other phase III trials Such studies might typically have an independent DMC in place for familiar purposes, for example, safety monitoring In line with current conventions, it would usually be interpreted that access to interim data and results should be strictly controlled, and in particular, not known to investigators or trial management personnel A challenging question in such trials often involves the precise identification and composition of the group that will review the interim results and make the dose selection, and where the DMC fits into this process In a traditional clinical development program, doses in a phase III trial are typically chosen by the sponsor organization, with assistance from other expert parties if needed, frequently based largely on the results of phase II trials Sponsor perspectives are very important to such major product development decisions, which are often quite complex and not lend themselves well to a simple algorithmic approach Current conventions in non-adaptive trials reflect that objectivity in reviewing interim data is maximized by use of an independent DMC While an independent DMC constituted for familiar motivations may already be in place in a seamless trial, there might be concern that a group entirely independent of the sponsor may not possess all relevant perspectives for a potentially complex decision, one that can have strong and long-lasting business implications for the sponsor There may also be concerns that DMC members experienced in other monitoring contexts might not have experience in this particular type of decision Thus, we might view that there is a conflict between the familiar motivation to insulate the sponsor from access to interim results, and the principle of bringing all relevant perspectives to bear in order to make the most fully informed decision The question might be asked whether a sponsor-internal group could be convened to make this type of decision, or whether there should at least be sponsor representation on an otherwise independent DMC for some limited portion of its deliberations, or some potential for a sponsor representative to ratify a recommendation made by the DMC If sponsor personnel are utilized and have access to comparative interim results, it is highly recommended that they have no other involvement in the trial as it proceeds other than performing this single role Though decisions as to how to proceed will depend on situation-specific details, the principles seem fairly straightforward As described previously, study integrity is best maintained if trial personnel remain insulated from interim results Sponsor access raises risks by compromising independence, as discussed in [14] Involvement by any sponsor personnel should require clearly stated and convincing justification, and be minimal to meet the needs— including as a desirable special case if possible, no access As discussed in [8], if some sponsor involvement could be convincingly justified: • The sponsor representation should include the minimum number of individuals possessing the perspectives needed to assist in arriving at the best decision, perhaps just one or two sponsor management representatives © 2016 by Taylor & Francis Group, LLC 512 Modern Adaptive Randomized Clinical Trials • These individuals should not otherwise be involved in trial activities nor participate in discussions of trial management issues while the study is ongoing • These individuals would have access to results only at the time of the dose selection, and will see only information relevant to that decision (e.g., unlike an independent DMC, which may have a broader and ongoing role) • Appropriate firewalls and process documentation should be in place to ensure that access to results is appropriately restricted, and there should be subsequent documentation that the processes were adhered to and information remained confidential Planning can play an important role in bridging the various types of concerns Extensive advance discussions can help satisfy sponsor concerns about allowing an independent DMC to make the adaptation decision without, or with only limited, sponsor involvement Prior to a trial’s start (or at least prior to any DMC access to unblinded data), it is not controversial for a sponsor, trial Steering Committee, and independent DMC to discuss issues openly It is important to iron out differing viewpoints at this stage, as this can be very problematic after the DMC has received access to unblinded data The sponsor can educate the DMC in regard to whatever relevant perspectives it might possess The planning discussions should include raising varied and complex hypothetical outcome scenarios, and discussing what might seem to be the appropriate recommendations in each This might then allow the actual data review and recommendations to be performed by the independent DMC without sponsor access to the results or direct sponsor participation in deliberations 23.4.3 Observing Adaptations and Reverse Engineering A question sometimes arises when considering confirmatory adaptive designs involving the extent to which knowledge of the adaptation made can provide information to observers about the interim results that led to the adaptation Even in the presence of strict confidentiality processes and firewalls, so that the actual results remain confidential within a DMC, it might be asked whether such knowledge could be problematic in terms of trial integrity A basic example would be a sample size re-assessment method in a twoarm trial, such as that of Cui, Hung, and Wang [3], where a protocol-specified plan might be to increase sample size in an algorithmic manner based on an interim treatment effect estimate Someone who knows the plan and becomes aware of the sample size change can potentially invert the algorithm and “back calculate” or “reverse engineer” to determine the estimate that led to the change This is information that would normally be restricted during an ongoing trial As we consider this issue in the context of seamless phase II/III for dose © 2016 by Taylor & Francis Group, LLC Statistical Monitoring of Response–Adaptive Randomized Clinical Trials 513 selection, we might ask about the implications of knowing which doses have been promoted to continue in a trial beyond the selection interim analysis Initially we might consider who needs to know which dose(s) have been selected We can envision a trial with a control and three doses, with the one most favorable at the interim analysis continuing to enroll patients Does it need to be broadly announced which dose has been selected? Perhaps not, but certainly trial logistic personnel at the very least will need this knowledge in order to implement the randomization change and revised drug supply, and sponsor management will probably require knowledge for planning purposes Often such trials are embedded within a development program whereby a second pivotal trial might be undertaken using the dose selected in the seamless design, so that this selection will need to be very widely known Regardless, it will usually be the case that very little information could be inferred about the magnitude of the interim treatment effects, and there would seem to be little potential for introducing biases into the trial Planned adaptations of this sort would probably be judged to convey far less information than types of futility judgments very commonly used and accepted in conventional trials As an example, consider the seamless design described in Barnes et al [2] The selection involved a complex algorithm essentially promoting the smallest dose that showed a certain degree of separation from placebo and an active control according to two separate measures, along with the next lower dose Seeing which doses were selected to continue did not convey to observers even knowledge of which dose was most effective, much less qualitative information about the nature of the dose–response pattern Because seamless phase II/III trials produce confirmatory data, trial planners must be sensitive to this issue and adhere to the relevant principles to the extent possible That is, if, given the nature of the trial and its role within a development program, it is feasible to restrict knowledge of the algorithm and the actual selection, then it is preferable to so But regardless, in the vast majority of situations, as long as the actual analysis results remain properly restricted, we see little cause for concern in this regard 23.5 Summary and Conclusion Processes relating to examination of interim data to implement changes within adaptive clinical trials have received a good deal of attention in recent years as the usage of adaptive designs has expanded The focus on this issue largely arises from the viewpoints, reflected in current interim monitoring practices, that knowledge of interim results has potential to introduce biases into trial results that could impair their interpretability The added complexities of actions implemented based on interim results in adaptive design trials raises new © 2016 by Taylor & Francis Group, LLC 514 Modern Adaptive Randomized Clinical Trials challenges, in terms of setting up processes that both maintain trial integrity while bringing the right perspectives to bear to ensure sound decision-making Here we have focused specifically on these issues as they pertain to adaptive randomization trials Not surprisingly, these apply differently depending on clinical development stage and a trial’s objectives, that is, whether it intends to provide confirmatory evidence, or rather is exploratory and aims to advise on certain aspects of the further development of a treatment In studies that aim to provide definitive demonstration of product efficacy and safety, independent review such as might be provided by a DMC is highly desirable Trial designs that have been referred to as seamless phase II/III designs fall into this category The membership of the interim review and decision-making body should reflect all perspectives and experiences relevant to decisions with which it is charged This will include a sufficient understanding of the adaptation methodology and algorithm To the extent that it can be justified that sponsor personnel possess perspectives relevant to arriving at the best adaptation decisions, it can be considered to expand the board to include such personnel in the decision process However, this must be implemented carefully to maintain confidentiality of the interim results and insulation from other trial participants and trial management personnel, so that those involved in this process should have adequate and documented independence from other roles and activities in the trial In exploratory adaptive randomization trials, interim review generally should not require the level of independence associated with the manner in which independent DMCs are utilized in confirmatory settings In planning a trial and its interim monitoring processes, there should be awareness of these issues so that it can be considered whether independent review might have merits, or how otherwise to responsibly minimize the possibility of bias But in general, taking into account the benefits of conducting the trial effectively and efficiently, strict independence will often not be needed in such trials Bibliography [1] Antonijevic, Z., Gallo, P., Chuang-Stein, C., Dragalin, V., Loewy, J., Menon, S., Miller, E R., Morgan, C C., Sanchez, M (2013) Views on emerging issues pertaining to data monitoring committees for adaptive trials Therapeutic Innovation & Regulatory Science 47(4), 495–502 [2] Barnes, P., Pocock, S., Magnussen, H., Iqbal, A., Kramer, B., Higgins, M., Lawrence, D (2010) Integrating indacaterol dose selection in a clinical study in COPD using an adaptive seamless design Pulmonary Pharmacology & Therapeutics 23, 165–171 © 2016 by Taylor & Francis Group, LLC Statistical Monitoring of Response–Adaptive Randomized Clinical Trials 515 [3] Cui, L., Hung, H M J., Wang, S J (1999) Modification of sample size in group sequential trials Biometrics 55, 853–857 [4] Committee for Medicinal Products for Human Use (CHMP) (2005) Guideline on Data Monitoring Committees London: EMEA [5] Committee for Medicinal Products for Human Use (CHMP) (2007) Reflection paper on methodological issues in confirmatory clinical trials planned with an adaptive design London: EMEA [6] Chow, S., Corey, R., Lin, M (2012) On the independence of data monitoring committee in adaptive design clinical trials Journal of Biopharmaceutical Statistics 22(4), 853–867 [7] Ellenberg, S S., Fleming, T R., DeMets D L (2002) Data Monitoring Committees in Clinical Trials: A Practical Perspective Chichester: Wiley [8] Gallo, P (2006) Confidentiality and trial integrity issues for adaptive designs Drug Information Journal 40, 445–450 [9] Gaydos, B., Anderson, K M., Berry, D., Burnham, N., Chuang-Stein, C., Dudinak, J., Fardipour, P., Gallo, P., Givens, S., Lewis, R., Maca, J., Pinheiro, J., Pritchett, Y., Krams, M (2009) Good practices for adaptive clinical trials in pharmaceutical product development Drug Information Journal 43, 539–556 [10] Herson, J (2008) Coordinating data monitoring committees and adaptive clinical trial designs Drug Information Journal 42, 297–301 [11] International Conference on Harmonisation (ICH) Expert Working Group (1998) ICH Harmonised Tripartite Guideline: Statistical Principles for Clinical Trials Federal Register 63, 49583–49598 [12] Maca, J., Bhattacharya, S., Dragalin, V., Gallo, P., Krams, M (2006) Adaptive seamless phase II/III designs—background, operational aspects, and examples Drug Information Journal 40, 463–473 [13] Sanchez-Kam, M., Gallo, P., Loewy, J., Menon, S., Antonijevic, Z., Christensen, J., Chuang-Stein, C., Laage, T (2014) A practical guide to Data Monitoring Committees in adaptive trials Therapeutic Innovation & Regulatory Science 48(3), 316–326 [14] US Food and Drug Administration (FDA) (2006) Guidance for Clinical Trial Sponsors on the Establishment and Operation of Clinical Trial Data Monitoring Committees Rockville MD: FDA [15] US Food and Drug Administration (FDA) (2010) Guidance for Industry for Adaptive Clinical Trials for Drugs and Biologics (draft) Rockville MD: FDA © 2016 by Taylor & Francis Group, LLC ... problem for a randomized trial comparing treatments T1 and © 2016 by Taylor & Francis Group, LLC 20 Modern Adaptive Randomized Clinical Trials Completely Randomized Design Response? ?Adaptive Randomized. .. 1.1 Modern Adaptive Randomized Clinical Trials 1.6.2 Adaptive Optimal Dose-Finding Designs 1.6.3 Randomized Designs with Treatment Selection 1.6.4 Group Sequential Adaptive. .. 2016 by Taylor & Francis Group, LLC xii Modern Adaptive Randomized Clinical Trials stances? What special considerations are required for adaptive randomized trials? What kind of statistical inference