A Study on the Prognostic Factors of Breast Cancer Survival Time Using Bayesian Cox Model

Breast cancer is one of the common diseases among women with various factors involved in it. The purpose of this study was to investigate the effect of tumor markers, estrogen receptor (ER), human epidermal growth factor receptor 2 (Her2), and Ki67 antigen, and other factors affecting the survival time of patients with breast cancer in Yazd City, Iran, using Bayesian multiple Cox regression analysis.
This was a population-based study of 538 women with breast cancer registered in the clinical database of the Ramezanzade Radiotherapy Center from the April 2005 until March 2012. Comprehensive data on prognostic factors, comorbidity and treatment together with complete follow-up for survival were used to evaluate improvements in mortality. Data was analyzed using R 3.4.2 software. P-value of less than 0.050 was considered as the significance level.
Findings: The mean age of patients with breast cancer and the mean survival time were 48.03 ± 11.16 years and 97.64 ± 4.23 months, respectively. Based on Kaplan-Meier method, the 1, 3, 5 and 8-year cumulative survivals in patients with breast cancer were 0.976, 0.898, 0.823, and 0.737, respectively. Bayesian Cox regression analysis showed that surgery [Hazard ratio (HR): 1.631, 95% Prediction interval (PI): 1.102-2.422)], ki67 (HR: 3.260, 95%PI: 1.6308-6.372), stage (HR: 5.620, 95%PI: 4.079-7.731), lymph node (HR: 1.765, 95%PI: 1.127-2.790), and estrogen receptor (HR: 2.033, 95%PI; 2.023-3.354) were significantly related to survival time.
According to Bayesian multiple cox regression, stage, Ki67, lymph node, estrogen receptor, and surgery variables have a positive effect on death hazard. By combining Bayesian and semi-parametric methods in survival analysis, in order to use prior information and relaxation of parametric assumption, the model gain more flexibility and robustness against misspecification of the probability model; this gives more valuable results.