Comparing Cox Model and Parametric Models in Estimating the Survival Rate of Patients with Prostate Cancer on Radiation Therapy

Background and Prostate cancer is the second most common malignant cancer in men and radiotherapy is one of the treatments for this disease. The aim of this study was to determine the effect of demographic, clinical and pathology factors in survival rate of patients on radiotherapy and comparing different survival models to determine an efficient model. In a historical cohort study 422 patients with prostate cancer were included. The subjects were chosen among those undergone radiotherapy and were followed up during 2007-2012 in the Cancer Institute of Tehran University. Effects of various factors on the survival rate of radiation treatment were determined using Cox model and three parametric models including log-normal, log-logistic and Weibull. We used AIC criteria and also standardized variability of regression coefficient. To analyze the data the Survival was applied in R version 2.14.1. Patients who received higher total dose of radiation and those undergone hormone therapy simultaneously, indicated higher survival rate (both p≤0.001). Univariate and multivariate models based on AIC criteria were more efficient than the Cox model for log-normal, log-logistic and Weibull, respectively. The findings were confirmed by the standardized variability. Parametric models in univariate and multivariate models showed higher efficiency compared with Cox model.