Clinical Trial Optimization Using R is a collaborative effort by nine pharmaceutical industry statisticians to provide guidance on clinical trial optimization. Key questions throughout a trial can include selection of dose and schedule, optimal patient population and treatment strategy, and whether a drug is effective with a reasonable safety profile. There is a need to make correct and efficient decisions based on limited data, and adaptive designs require fast and accurate feedback to make changes quickly mid-trial. When initially planning a trial, various statistical designs and approaches are often tossed around before a final decision is made. But as combinations of these strategies are endless, perhaps there is a more systematic approach to teasing out which assumptions and approaches are optimal for a particular trial. The authors explore the Clinical Scenario Evaluation (CSE) framework, which considers objectives, design and analysis selections, underlying assumptions, and quantitative metrics to facilitate decision making (Benda et al. Citation2010). The authors provide examples in R to systematically make these decisions in a reproducible way with the goal of providing an evidence-based approach when designing trials and programs.
CSE aims to break down the multidimensional process of clinical trials into manageable components: (1) data models, (2) analysis models, and (3) evaluation models. Data models refer to the assumptions and process of generating trial data. Analysis models specify the strategies applied during the analysis of trial data, and evaluation models assess the performance of these strategies. Combinations of data and analysis models build clinical scenarios for which simulations can be performed to systematically assess the operating characteristics of proposed designs and methods, with an emphasis on realistic assumptions for the trial.
The R package Mediana supports a general framework for CSE-based simulations, where you can specify your various models. Design considerations include sample size, events, distribution/type of outcome, and enrollment/follow-up period definitions. Analysis components include the test used, the statistic to be computed, and any multiplicity adjustments. Evaluation looks at combinations of these strategies with common success criteria of marginal, disjunctive, conjunctive, or weighted power. The authors give real trial examples and walk readers step by step through creating the code necessary to run these CSE simulations in Mediana.
Chapter 1 is a basic introduction of the framework, R code, and a couple of traditional and adaptive trial examples. Chapters 2 and 3 move into trials with multiple objectives and subgroup analyses, respectively. I found Chapter 4 most interesting as it builds on all of the previously covered material to explain how to use this decision-making information in clinical development. Specifically, the authors utilize concepts of the Go/No-Go (GNG) paradigm for whether phase II data warrant a definitive phase III trial and probability of success (POS) calculations to quantify how likely future data will deliver success.
All chapters spend a significant amount of time walking through examples with associated R code and results and do a very nice job explaining the initial CSE framework. Examples expand in complexity as the book progresses. As a biostatistician working in an academic setting, I am quite familiar with simulations used to construct new trials. However, the concept of CSE framework was brand new to me, and I think the strategies outlined in this book could definitely improve my approach to designing trial and analysis plans! This would also facilitate discussions with the clinical study team on how to proceed given our results. I would recommend this book to any clinical trial statistician who is interested in exploring simulations to better understand the implications of selected design and analysis strategies within their trials.
Wake Forest School of Medicine
Winston-Salem, NC
[email protected]
Correction Statement
This article has been republished with minor changes. These changes do not impact the academic content of the article.
References
- Benda, N., Branson, M., Maurer, W., and Friede, T. (2010), “Aspects of Modernizing Drug Development Using Clinical Scenario Planning and Evaluation,” Drug Information Journal, 44, 299–315.