Online Program Home
My Program

Abstract Details

Activity Number: 660
Type: Contributed
Date/Time: Thursday, August 4, 2016 : 8:30 AM to 10:20 AM
Sponsor: Biopharmaceutical Section
Abstract #320879
Title: Statistical Testing in the Presence of Nonproportional Hazards
Author(s): Amarjot Kaur* and Yabing Mai and Ziliang Li and Xia Xu and Wen-Chi Wu
Companies: Merck and Merck Research Laboratories and MRL and MRL and MRL
Keywords: Cox Proportional Hazards Test ; Time-to-Event ; Non-proportionality ; Log-rank Test ; Piecewise Exponential ; Kaplan-Meier

When the proportional hazards (PH) assumption does not hold for the time-to-event type of survival data, the score test based on Cox PH model is no longer optimal. It has been found that that loss of efficiency increases with the magnitude of non-proportionality and could be large in some situations. We carried out an extensive review on alternative methods that account for the non-proportional hazards (NPH) while retaining relatively high efficiency and providing interpretable results. The following three classes of methods were selected from the preliminary review and are focused in later review: piecewise exponential model; weighted log-rank test; weighted Kaplan-Meier test / Restricted Mean Survival Time. Performances of these methods were assessed for a variety of scenarios of non-proportionalities. The NPH methods under review are found to be superior to Cox PH model and log-rank test in most of the NPH scenarios and remain robust under PH or under minor deviation from non-proportionality.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2016 program

Copyright © American Statistical Association