Adaptive design for censored survival data adjusting for covariates

In many clinical trials, it is often of interest to compare time to an event between treatment groups. Clinical trials of this type are often designed with a fixed number of patients followed up for a fixed period of time. Adaptive designs can be useful when the observed incidence rates of the event of interest at interim analyses seem to deviate significantly from the initial assumptions at the planning stage. We extend the adaptive method based on weighted log-rank statistics proposed by Shen and Cai (2003) to one based on the proportional hazards model to take important covariates into consideration. Simulation studies are carried out to show the operating characteristics of the proposed method and to compare with a fixed-sample design and the Shen and Cai method. The method is also illustrated using clinical trial data.