Abstract:
|
The Cox proportional hazard model (1972) is a popular method and is being used extensively in studying the relationship between survival times and explanatory variables. The main question that needs to be addressed with respect to the adequacy of the Cox PH model is whether the proportional hazard assumption is met. In the current study we illustrate detection of violation of the proportional hazard assumption using methods proposed by [1] and [2] using breast cancer data from the Surveillance, Epidemiology and End Results database (2009). We discuss the impacts of using the subject model with violations that will result misleading conclusions. Correction of the violations is illustrated using Cox model with time-varying effects Therneau et al. (2000). The prognostic factors considered in this study are age, tumor size, historical stage, tumor grade and race. Tumor stage, grade age and tumor size indicate violation of proportional hazards assumption. Our study shows the importance of adjusting the model for non-proportional hazards since the estimates for the initial Cox model and the adjusted Cox model vary drastically.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.