A case study of the effect of covariate adjustment in a sequential survival clinical trial

Recent literature has discussed the value of adjustment for important covariates in models involving nonnormal data. It is generally concluded that, when performing traditional "fixed sample size" clinical trials, covariate adjustment influences the magnitude of the treatment effect but has little effect on precision of the estimate. In fact, it may actually reduce it. In this paper, we investigate the effect of covariate adjustment in the sequential setting via a case study of a survival clinical trial. Sequential and fixed sample analyses are compared, with and without covariate adjustment of the treatment effect. It was found that conclusions similar to those for the fixed sample size case also apply in the sequential case, but that incorporation of covariate information can present added complications in this setting.