Using Integrated Difference Between Two Kaplan-Meier Curves for Quantifying Treatment Contrast in Comparative Clinical Trials with Event Time Observations

• Activity Number: 668
• Type: Contributed
• Date/Time: Thursday, August 5, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Lihui Zhao*+
• Companies: Harvard School of Public Health
• Keywords: Confidence interval ; Integrated survival rate difference ; Martingale ; Perturbation-resampling method ; Proportional hazards

Abstract

For a typical comparative clinical trial to assess relative efficacy or safety between two treatment groups with an event time as the endpoint, often we assume that the ratio of two hazard functions is constant over the study time. One then utilizes such a proportional hazards parameter to quantify the treatment contrast. In this paper, we propose an alternative metric, the integrated survival rate difference over a specific time interval of interest, for quantifying the treatment contrast. The confidence interval estimate for this metric can be constructed using the perturbation-resampling method discussed in Parzen, Ying and Wei (1997). We show that when the event rates for two treatment groups are small and similar, the integrated survival rate difference is preferable than the proportional hazards parameter. The data from a recent trial is used for illustration.