[Application of Bayes method in determining of the risk factors on the survival rate of gastric cancer patients]

Gastric cancer is one of the most common cancers in the world. The classical methods such as Cox regression and parametric models are used in most medical researches that their aims is the survival distribution survey, although the Bayes models have some advantages in compared with the classical models. The present study was performed to analyze the survival rate of patients who had gastric cancer and were under treatment in the gastroenterology ward of Taleghani hospital, in Tehran using Bayes models. Weibull distribution was used for modeling in the study. This study was a cohort study and performed in the gastroenterology ward of Taleghani hospital by using gastric cancer patient's data from January 2003 to December 2007. 178 patients were enrolled to the study and their information was collected through telephone contacts. The survival rate of patients were analyzed using Bayes Weibull models by considering variables such as age of diagnosis, gender, tumor size, metastasis of other lymph. For determining of the risk factors on the survival of patients, was used Weibull model in the case that interval censoring. Data analysis was carried out using Winbugs software and significant levels were considered 0.05. The results showed survival rate are dependent on the age of diagnosis and tumor size. Those patients who had early diagnosis, the rate of survival was greater. In addition, the patients who had smaller tumor size, their survival rate was greater. Considering to classical models are based on normal approximation and applicable for big samples, Bayes methods are emphasized for small to medium samples. The results of this study showed that the Bayesian Weibull model is a suitable model. This study also showed that age of diagnosis and tumor size of patients is important factors in regard to the survival rate of these patients. As a result, if gastric cancer is diagnosed early, the relative risk of death would reduce