Interim Analyses During Group Sequential Clinical Trials

INTRODUCTION

This JAMA Guide to Statistics and Methods describes why interim analyses are performed during group sequential trials, provides examples of the limitations of interim analyses, and provides guidance on interpreting the results of interim analyses performed during group sequential trials.

In an issue of JAMA, Locatelli et al¹ reported a randomized clinical trial comparing a targeted immunotherapy agent, blinatumomab, with standard consolidation chemotherapy in children with high-risk first-relapse B-cell acute lymphoblastic leukemia that had a primary outcome of event-free survival. The trial used a group sequential design and the accumulating data were analyzed to determine if a conclusion could be drawn at prespecified time points before the planned final analysis. The trial was terminated during the first prespecified interim analysis after only enrolling about half of its maximum sample size. Children receiving blinatumomab had significantly improved event-free survival compared with those receiving standard chemotherapy.¹

WHY ARE INTERIM ANALYSES PERFORMED IN GROUP SEQUENTIAL TRIALS?

There is substantial uncertainty regarding the likely treatment effect (ie, the between-group difference in outcomes) when a clinical trial is being designed. The assumed treatment effect (the magnitude of difference the trial is designed to reliably detect) strongly affects the number of patients needed for the trial.² If the true treatment effect is underestimated, the calculated sample size may be larger than necessary, increasing the time needed to complete the trial and potentially delaying delivery of effective treatments to patients outside the study. Alternatively, if the true treatment effect is overestimated, the calculated sample size may be too small, resulting in an underpowered trial that may fail to detect a real benefit. In addition, because new treatments may be ineffective or even harmful relative to standard care, enrolling a trial to its full sample size without reviewing the data risks could waste resources or even harm patients.

Group sequential trials incorporate interim analyses to allow timely decisions that mitigate the challenges associated with uncertainty in the size and direction of the treatment effect. The data available at the time of the interim analysis are assessed to determine if predefined stopping criteria are met. This allows the trial to be stopped and a valid conclusion to be drawn earlier.

DESCRIPTION OF INTERIM ANALYSES IN GROUP SEQUENTIAL TRIALS

When planning a group sequential trial, investigators must prespecify plans for interim analyses. This generally includes when the interim analyses will occur, what information will be used, and what statistical methods and stopping criteria will be applied. One action that may be triggered by an interim analysis is stopping enrollment due to clear evidence that the investigational treatment is superior (known as early success).³ During the interim analysis, a test statistic is calculated (eg, a P value), and, if the test statistic meets a prespecified threshold, the trial may stop. If the test statistic does ...