Multiple comparisons in interim analysis

Abstract

Interim analysis schemes typically relate to accumulating acquisition of clinical data where statistical decisions facilitate medical or clinical studies with respect to primary endpoint and cost-benefit analysis, as well as examination of other therapeutic aspects. Basically, an interim analysis scheme aims to have multiple looks into accumulating datasets where conventional fixed-sample or sequential setups cannot be totally validated; even the group-sequential procedures may not be basically appropriate. Multiple endpoints, auxiliary and or concomitant variables and clinical designs often invalidate the regularity assumptions underlying conventional (linear) statistical inference tools in interim analysis schemes, and therefore alternative avenues incorporating suitable stochastic processes based on progressively censored schemes are explored. Gaussian approximations are quite helpful in this respect, and a general overview of multiple comparisons in such (time-sequential) setups is presented here with an eye on applications in clinical trials.