Seminars

General Departmental Seminar Series

A Group Sequential, Response-Adaptive Design for Randomized Clinical Trials

Theodore Karrison, Department of Health Studies, University of Chicago

CO-AUTHORS: Dezheng Huo and Rick Chappell

Friday, November 8, 2002, 12:00 pm

G5/136-142 Clinical Sciences Center - 600 Highland Avenue

ABSTRACT

There has been considerable methodological research on response-adaptive designs for clinical trials, but they have seldom been used in practice. The many reasons for this are summarized in a paper by Rosenberger and Lachin (1993), but the two main reasons generally cited are the logistical difficulties of implementing the adaptive assignment scheme and the potential for bias due to selection effects, "drift" in patient characteristics or risk factors over time, and other sources. Jennison and Turnbull (2000) consider a group sequential, response-adaptive design for continuous outcome variables that partially addresses these concerns while at the same time allowing for early stopping. The key advantage of constant randomization probabilities within sequential groups is that a stratified analysis will eliminate bias due to drift.

In this paper we consider binary outcomes and an algorithm for altering the allocation ratio that depends on the strength of the accumulated evidence. Specifically, patients are enrolled in groups of size n_{Ak},n_{Bk},k=1,2,...,K where n_{Ak},n_{Bk} are the sample sizes in treatment arms A and B in sequential group k. Patients are initially allocated in a 1:1 ratio. After the k^th interim analysis, if the z-value comparing outcomes in the two treatment groups is less than one in absolute value, the ratio remains 1:1; if the z-value exceeds 1.0, the next sequential group is allocated in the ratio R1 favoring the currently better-performing treatment; if the z-statistic exceeds 1.5, the allocation ratio is R2, and if the z-value exceeds 2.0, the allocation ratio is R3. If the O'Brien-Fleming monitoring boundary is exceeded the trial is terminated. Group sample-sizes are adjusted upwards to maintain equal increments of information when allocation ratios exceed one. The z-statistic is derived from a Mantel-Haenszel test stratified by sequential group.

Simulation studies and theoretical calculations were performed under a variety of scenarios and allocation rules (for example, [R1, R2, R3] = [1.5, 2, 2.5]). Results indicate that the method maintains the nominal type I error rate even when there is substantial drift in the patient population. When a true treatment difference exists, a modest reduction in the proportion of patients assigned to the inferior treatment arm and in the overall proportion of failures can be achieved at the expense of smaller increases in the total sample size relative to a non-adaptive design. Comparisons in terms of the total number of failures are less favorable. Limitations, such as the impact of delays in observing outcomes, are discussed, as well as areas for further research.

Back to General Departmental Seminar Series