Comparison of efficiency of Cox Proportional Hazards Model and Accelerated Failure Time Models in Determining Survival Factors in Gastric Cancer Patients

Abstract

Background and purpose: Stomach cancer is a multifactorial disease that may be influenced by many factors, including environmental and genetic factors. Therefore, it is important to investigate and recognize the prognostic factors in the survival of this disease. The purpose of this study was to investigate the factors affecting the survival of gastric cancer by parametric and semi-parametric regression methods and finally to fit the best model among these models. Materials and methods: A historical cohort study on 193 patients with gastric cancer in Mazandaran province 2011-2014 was carried out. Demographic, clinical and therapeutic information of patients were collected. The Schoenfeld test was utilized to evaluate the assumption of proportional hazards, while Cox-Snell residuals were employed to assess the adequacy of the model. The STATA software (v. 14) was used to analyze the data. The significance level of the tests was considered 0.25 for univariate analysis. Results: 30% and 70% of the patients were women and men, respectively. The average age at diagnosis of the patients was 64.92±14.04 years. The mean and median survival time were 21.92 and 8.06 months, respectively, with a standard error of 2.57 and 1.20. Based on Akanke’s information criterion and Cox-Snell's residuals, the log-logistic model was selected as the optimal model. The results of the log-logistic model showed that the variables of body mass index (acceleration factor= 1.11 and P= 0.003) and in terms of the type of treatment, the combination of chemotherapy and surgery compared to surgery (acceleration factor=3.12 and P=0.028) and three types of combined treatment of radiotherapy, chemotherapy and surgery to surgery (acceleration factor=7.58 and P=0.020) and kidney disease (acceleration factor=0.20 and P=0.014) were factors affecting survival. Conclusion: Despite the preference of the majority of researchers to utilize the Cox model, accelerated failure models can serve as a viable alternative to the Cox model in comparable circumstances

Keywords

• cox proportional hazards model
• accelerated failure model
• akanke’s information criterion
• cox snell residuals
• survival
• gastric cancer
• parametric regression