Model-based simultaneous inference for multiple subgroups and multiple endpoints

TL;DR

This paper evaluates a flexible model-based method for simultaneous inference across multiple subgroups and endpoints, comparing it to traditional approaches through simulations and real data analysis.

Contribution

It introduces and assesses the performance of the multiple marginal models (mmm) approach for complex subgroup and endpoint analysis in clinical studies.

Findings

mmm controls familywise error rate effectively

mmm demonstrates higher power than Bonferroni in simulations

Method handles overlapping subgroups and various endpoint types

Abstract

Various methodological options exist on evaluating differences in both subgroups and the overall population. Most desirable is the simultaneous study of multiple endpoints in several populations. We investigate a newer method using multiple marginal models (mmm) which allows flexible handling of multiple endpoints, including continuous, binary or time-to-event data. This paper explores the performance of mmm in contrast to the standard Bonferroni approach via simulation. Mainly these methods are compared on the basis of their familywise error rate and power under different scenarios, varying in sample size and standard deviation. Additionally, it is shown that the method can deal with overlapping subgroup definitions and different combinations of endpoints may be assumed. The reanalysis of a clinical example shows a practical application.