Study On Risk Prediction Of Esophageal Cancer (EC) Of Kazak In Xinjiang Uygur Autonomous Region

Objective: The aim of this paper is to research the risk factors of esophagus cancer and establish riskmodel of esophagus cancer among Kazak;as a result, providing more evidence for predicting esophaguscancer.Method:1. Data of this case-control study is collected from6hospitals in northern Xinjiang from March2005to October2010;2.There are two databases established based on the data. Database1includes casesand controls those are collected at the initial stage of the research. There are166cases and370controls,excluding undetected genes in both groups. Database2includes cases that are gathered at the beginning ofresearch,and controls are health people. There are251esophagus cancer cases and4878cases withoutesophagus cancer;3. Logistic regression model is used to screen relative risk factors in order to ascertainvariables of the discrimination model;4. Esophagus cancer discrimination model with various risk factors isbuilt by logistic regression model. Sensitivity and specificity are utilized to evaluate accuracy of the model,calculate predictive incidence rate and set up risk predictive model.Results：1. Risk factors of esophagus cancer,According to the single factor analysis of Kazak group,15riskfactors including drinking river water, pace of eating and etc. are related to esophagus cancer; besides,11risk factors including being crapulent, eating irregularly are bound up with esophagus cancer from thevarious factors analysis.There are12factors including age、gender and so on are related to esophaguscancer according to single analysis based on group control. The results gained from various factors analysisshowed that8factors including eating irregularly、age and so on are dangerous factors causing esophaguscancer;2. Discrimination model I, model II,III and IV are built with different control groups and variouscombination of risk factors, the accuracy of each group is75.9%、70.1%、77.2%and97.3%respectively,and the accuracy of model IV is the highest. Four sub-models are established basing determination modelIV with deleting variables gradually.The accuracy of model IV1-4is97.2%、97.3%、97.0%、97.1%, andsensitivity is67.5%、68.7%、69.9%、68.1%respectively. To compare with others, discrimination model IV2and IV3are better；3. Risk predictive model1which includes age, drinking, regular diet, hot food, lessintake fruits and history of stomach disease and model2which contains the same factors in model1excepthot food are both established in accordance with model IV3and IV4.Conclusion:1. The incidence of esophagus cancer among Kazak is related with unhealthy environment,dietand gene polymorphism;behavior and heredity have less effect on esophagus cancer. Drinking, irregularlydiet,being crapulent, less intake fruits and history of stomach disease are statistically significant with higherOR in logistic model;2. Discrimination model IV which is based on group control, including drinking,irregular diet, age, less intake fruits, hot food, history of stomach disease and being crapulent is better thanother models;3. Risk predictive model based on age, drinking or not, regular diet or not, suffering stomachdisease or not, intake of fruits and so on is suitable for population.The risk of esophagus cancer is50percent higher for those people who have irregularly diet, stomach diseases, intake less fruits and drinking.