A comparison of multiple approaches to subgroup analysis in clinical trial
Abstract
In randomized clinical trials, medical researchers are interested to determine the effectiveness of a new treatment not only in the overall population but also to some subgroups with possible enhanced treatment effects. However, subgroup analysis may become problematic due to the issue of multiplicities, data dredging etc. Accounting for these issue, we summarized some guidelines on the use and interpretation of subgroup analysis. We reviewed three approaches to subgroup analysis, a tranditional Bayesian regression with interaction terms, the 'Virtual Twins' methods and a Bayesian model selection approach. The advantage and disadvantage of these three approaches are discussed.
Description
text