Quantification of treatment effect variability in clinical trials

The effect of treatments is well known to vary largely between individuals. However, the standard statistical model of clinical trials assumes a constant treatment effect across individuals (see e.g. the ICH E9 guideline). The assumption of constant treatment effects is then investigated in the exploratory part of the statistical analysis by a series of subgroup analyses. This provides information on the heterogeneity of the treatment effect across specific subgroups. However, due to random fluctuations in small subgroups, lack of power and multiple testing issues, the interpretation of subgroup analyses is difficult and still a controversial issue.

Treatment effect heterogeneity describes how average treatment effects vary across baseline strata. By its focus on averages, it necessarily ignores treatment effect heterogeneity within a stratum. With the concept of subject-treatment interactions, it is possible to also account for heterogeneity within strata that remains unexplained by the baseline information. This permits to quantify the variance of a generally random treatment effect. Unfortunately, the estimation of treatment effect variance is accompanied with an unidentifiability issue which is difficult to resolve.

The goal of this proposal is to develop model building and variable selection methods that reduce the unidentifiability problem as much as possible and provide more precise estimates of treatment effect variance. In a second step we aim to explore how far such estimates can detect cases with relevant treatment effect heterogeneity as well as identify cases where treatment effect heterogeneity should be of no concern.

Project start: June 2016
Project member: A. Lüschen and W. Brannath (PI)
Research Funding of University Bremen (Focus Project)

Read the press atricle of the "Weser Kurier" (in German) on a talk of W. Brannath at the Haus der Wissenschaft (in the city center of Bremen) about the projekt.