Exploring subgroup analysis based on post-treatment factors

Abstract

Statistical analyses are often used in clinical literature to support interpretations, such as employing subgroup analysis based on whether a patient experienced dose reduction to argue that the starting dose could be lower, or using subgroup analysis based on post-progression anti-cancer therapies to explain differences between PFS (progression-free survival) and OS (overall survival). However, are these analyses valid for drawing such conclusions?

Subgroup analysis is a critical component of statistical evaluation in clinical studies, often used to explore treatment effects across different patient populations. However, the use of post-baseline factors—variables measured after randomization or treatment initiation—for subgroup analysis has garnered significant scrutiny due to its potential to introduce bias and misinterpretation. In this talk, we will discuss why subgroup analyses based on post-baseline factors may not be appropriate. We will also explore alternative methods from the literature and summarize their potential applications. 

Ultimately, conclusions drawn from subgroup analyses involving post-baseline factors should be approached with great caution.