Background: In competing risks data, when a person experiences an event other than the one of interest in the study, usually the probability of experiencing the event of interest is altered. Therefore, it is necessary to analyze the competing risk data. The aim of this study was to analyze the breast cancer risk factors using the competing risk model in patients with breast cancer.
Materials and Methods: In this cohort study, 6206 cancerous patients included all women diagnosed with breast cancer were identified during 1990-1999 by the British Columbia Cancer Center and followed until 2010. To compare the competing risk models, the stratified Cox and proportional sub-distribution hazards models were fitted.
Results: Findings showed that for breast cancer death, the hazard ratio increased for age (29%, 40%), radiotherapy (71%, 55%) and hormone therapy (76%, 84%) in the stratified Cox and proportional sub-distribution hazards models, respectively. Surgery decreased the hazard ratio in both models (89% ,80%). For deaths not due to breast cancer, the hazard ratio for age (81%, 91%) and chemotherapy (67%, 61%) decreased in both models, respectively.
Conclusion: The Cox model, which ignores the competing risks, presents the different estimates and results than the proportional sub-distribution hazards model. Thus, in the analysis of competing risks data, the sub-distribution proportion hazards model is more appropriate than the Cox model.