Adaptive design of confirmatory trials: Advances and challenges

Abstract

The past decade witnessed major developments in innovative designs of confirmatory clinical trials, and adaptive designs represent the most active area of these developments. We give an overview of the developments and associated statistical methods in several classes of adaptive designs of confirmatory trials. We also discuss their statistical difficulties and implementation challenges, and show how these problems are connected to other branches of mainstream Statistics, which we then apply to resolve the difficulties and bypass the bottlenecks in the development of adaptive designs for the next decade.

Health Analytics, Economics and Medicine toward a 21<sup>st</sup> Century Health Care System

“…The third step is design economically feasible and risk-manageable clinical trials to provide data for the development and verification of personalized therapies. A partnership between the pharmaceutical/biotechnology industry, the regulatory agency, and researchers from academia can make use of the aforementioned and still ongoing advances in adaptive design [13] to break the innovation bottleneck and resolve a long-standing problem in translational medicine.…”

Design of clinical trials with failure-time endpoints and interim analyses: An update after fifteen years

“…Acknowledging that type I error probability guarantees are important to gain regulatory approval of a new treatment, "frequentist twists" are used to satisfy the type I error constraint at some chosen parameter configuration in the null hypothesis by Monte Carlo simulations at the configuration and thereby adjusting the rejection threshold of the Bayesian test [35,36]. However, as pointed out in [37], there is no guarantee that the type I error is maintained at other parameter configurations for a composite null hypothesis, as in semiparametric models for survival outcomes. In contrast, the modified Haybittle-Peto stopping boundary applied to commonly used censored rank statistics in comparing the survival distributions of two treatments has frequentist validity besides efficiency, flexibility and ease of use in time-sequential survival trials, showing its advantages over the preceding Bayesian approach to early stopping.…”

“…As noted in [37], adaptive seamless designs represent an important trend of innovations in clinical trial designs to address the need for more efficient and effective drug development processes in translating the breakthroughs in biomedical sciences into treatments of complex diseases. In particular, whereas traditional Phase III trials are inefficient "stand-alone" trials whose analyses ignore the information from previous phases, [50] argues for combining Phase II and Phase III into a single trial conducted in two stages, which is the basic idea of seamless Phase II-III designs.…”

“…As noted in [37], it is widely recognized that this p-value combination approach is inefficient, and more efficient seamless Phase II-III designs incorporating patient subgroup selection are long-standing open problems. In fact, Section 4.3 of [37] and the recent paper [55] have developed such designs for comparing a new method against standard medical care for stroke patients.…”

“…In fact, Section 4.3 of [37] and the recent paper [55] have developed such designs for comparing a new method against standard medical care for stroke patients. The endpoint of that trial is the Rankin score, which is much easier to handle than the censored failure-time endpoint in survival trials.…”