Improved Hazard Ratio Estimation in Small Time-to-Event Trials

• Activity Number: 15
• Type: Topic Contributed
• Date/Time: Sunday, July 29, 2007 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Devan V. Mehrotra*+ and Arthur Roth
• Companies: Merck & Co., Inc. and Pfizer Inc.
• Keywords: Cox model ; Efron approximation ; Generalized logrank test ; Proportional hazards ; Survival Analysis ; Ties

Abstract

The Cox proportional hazards (CPH) model is the standard tool for hazard ratio estimation in time-to-event trials. Mehrotra and Roth (2001) proposed an alternate approach based on a generalized logrank (GLR) statistic, and showed it to be notably more efficient than CPH for small trials (< 200 patients), and equally efficient for large trials. However, a limitation of the GLR approach is that it assumes no (or very few) tied event times, which is restrictive because continuous event times are often rounded to the nearest day, week, etc., resulting in ties. The best tie-handling methods for the Cox model are those of Kalbfleisch and Prentice [KP] (1973) and Efron (1977). Two extensions of the original GLR approach, GLR-KP and GLR-E, will be developed and shown (via simulations) to yield estimators that are more efficient than those based on the KP and Efron methods, respectively.