Modeling the Probability of Breast Cancer and Competing Risks Mortality Using the Hazard Function of the Cumulative Incidence Function (CIF)

• Activity Number: 353
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Yuanyuan Liu*+ and Ellen McCarthy and Long H. Ngo
• Companies: Beth Israel Deaconess Medical Center and Beth Israel Deaconess Medical Center and Harvard Medical School
• Keywords: competing risks ; survival analysis ; cause-specific hazard ; subdistribution ; breast cancer

Abstract

Competing risk data arise naturally in medical research, where a subject may experience an event and this event prevents the outcome of interest from happening. Conventional analytic methods often treat competing risk events as censored events, which could lead to biased results. Cause-specific models and Fine-Gray proportional subdistribution hazards models are two most commonly used methods for competing risk analysis. While cause-specific models are easy to fit, they do not allow a direct interpretation in terms of marginal probabilities for the particular failure type. The increasingly popular Fine-Gray method offers more intuitive clinical interpretation of risk, but researchers have been slow to adopt this method that account for competing risks, largely due to the computational complexity and limitations of statistical packages. In this article, we analyze data from a cohort of older women in a national registry to evaluate breast cancer mortality. Different statistical methods are used to account for competing health risks. We also explore the computational efficiency of different statistical packages in Fine-Gray model estimation.