gms | German Medical Science

The gold standard for the evaluation of efficacy and safety of new treatments are randomized controlled trials (RCT). Analyses of RCT are commonly adjusted for baseline covariates to minimize bias and increase the efficiency of the inference, i.e. higher power for hypothesis tests or shorter confidence intervals [1]. With an increasing interest in personalized treatment strategies more stratified trials investigating subgroups of patients are conducted. As in Placzek and Friede (2017) [2] we consider the analyses of continuous endpoints testing for treatment effects in several nested subgroups, which might be defined by thresholds of a biomarker. Here we extent their work by adjusting the analyses for baseline covariates considering an analysis of covariance (ANCOVA) model. Following Mehta et al. (2014) [3] we divide the population into disjunctive subsets, in which the hypotheses of interest are tested. Hence, the hypothesis tests for subpopulations consisting of one or more disjunctive subsets are tested by combining the p-values of those subsets. This approach was also considered by Graf et al. (2017) [4], however, without adjusting for baseline covariates. A method for sample size determination is presented. Furthermore, we propose a design with blinded sample size re-estimation correcting misspecifications of nuisance parameters in the sample size calculation. The performance of the proposed testing strategy, sample size determination and blinded reestimation are investigated using Monte Carlo simulations. The proposed methods are illustrated by a trial evaluating a fatigue intervention in multiple sclerosis [5].

The authors declare that they have no competing interests.

The authors declare that a positive ethics committee vote has been obtained.