gms | German Medical Science

Subgroup analyses of clinical trials are often carried out to detect signals of differential treatment effects across the subgroups. In some cases, such analyses are requested because (external) additional medical evidence has been established; in other cases the data from the trial at hand raises concerns about potential differential effects.

The Draft EMA guideline [1] requests sponsors that the consistency of effects in well-defined subgroups is evaluated when the clinical trial population was heterogeneous. This request has led to the application of heterogeneity tests for assessing consistency. The drawback of this approach is that “homogeneity” is the null-hypothesis, which can only be rejected, but not confirmed through such heterogeneity tests. Furthermore, the performance of these tests is rather poor [2], e.g. having low power with small sample sizes.

We have investigated alternative methods for assessing the consistency of treatment effects across subgroups. These methods aim to scale the test limits for each subgroup based on the difference and variability between subgroup and overall outcome, or take account for the sample size in the subgroups.

We will present the theoretical derivations of these methods, as well as some of their performance characteristics based on simulations, and will finally discuss their applicability for post-hoc analyses.