Bootstrap simulations to estimate overall survival based on the distribution of a historical control

Abstract

Following the calculation of the median overall survival (OS) in a clinical trial, it is often desirable to put the estimates into perspective by comparing them with the results of other studies reported in the current bibliography. The main limitation of this comparison is the different distribution of prognostic baseline characteristics between studies. A SAS® program to obtain a bootstrap estimation for the median OS, balancing it by the historical distribution, is described herein.

Keywords:

• Bootstrap
• Simulation
• Overall survival