新疆哈萨克族食管癌风险的预测  

Prediction of risk for esophageal cancer in Kazak people in Xinjiang

作　　者：李婧[1] 秦江 

机构地区：[1]石河子大学医学院预防医学系,新疆维吾尔自治区石河子市832002 [2]卫生部卫生发展研究中心,北京市100191

出　　处：《世界华人消化杂志》2014年第10期1442-1445,共4页World Chinese Journal of Digestology

基　　金：国家自然科学基金资助项目;No.30660161~~

摘　　要：目的:利用不同危险因素建立食管癌风险预测模型,为新疆哈萨克族食管癌高危人群监测体系的建立提供循证依据和基础信息.方法:利用前期收集对照和当地人群对照环境、行为危险因素和基因检测数据,采用条件Logistic回归筛选危险因素,利用Logistic判别建立预测模型.结果:以哈萨克族对照,饮酒(OR=2.77)、饮食不规律(OR=3.42)、经常暴饮暴食(OR=4.01)、少吃水果(OR=2.65)、胃病变史(OR=2.66)、进食速度快(OR=1.94)、食管癌家族史(OR=2.06)、HLA-DRB1*0901(OR=2.83)、TAP2379(OR=2.09)、CYP2E1(OR=1.60)10个因素进入模型;以人群对照时,年龄(OR=1.10)、饮酒(OR=6.27)、饮食不规律(OR=118.05)、经常热烫饮食(OR=3.02)、经常暴饮暴食(OR=2.11)、少吃水果(OR=6.80)、吃熏制肉(OR=17.14)、胃病变史(OR=5.31)7个因素进入模型;从判别效果看4个模型的判别正确率分别为74.6%、70.1%、76.1%和96.9%.模型四的判别效果最好.结论:以环境、不良行为因素建立的风险预测模型适合基层应用,风险预测模型可以通过该简单、有效的预测概率模型进行风险自我评估.AIM： To develop risk prediction models for esophageal cancer, and provide evidence and in- formation for the establishment of an esophageal cancer monitoring system for high-risk Kazak population in Xinjiang. METHODS： Data of the local population and controls, including environmental, behavioral risk factors, and genetic data, were collected. Risk factors were screened using conditional Logistic regression to establish Logistic discriminant predictive models. RESULTS： Two groups were selected as controls in this study： one was Kazakh population and the other was general population. Compared with the Kazakh control group, values ofodds ratios （ORs） of related risk factors were as follows： alcohol drinking （OR -- 2.77）, irregular diet （OR =- 3.42）, frequent binge eating （OR = 4.01）, fruit eating （OR = 2.65）, gastropathy histo- ry （OR = 2.66）, eating speed （OR = 1.94）, family history of esophageal cancer （OR = 2.06）, HLA- DRBI＊0901 （OR = 2.83）, TAP2379 （OR = 2.09）, and CYP2E1 （OR = 1.60）, and 10 factors were selected into the model. As for the population- based control, values of OR were： age （OR = 1.10）, alcohol drinking （OR = 6.27）, irregular diet （OR = 118.05）, frequently eating hot and burning food （OR = 3.02）, frequent binge eating （OR = 2.11）, fruit eating （OR = 6.80）, smoked meat eating （OR = 17.14）, gastropathy history （OR = 5.31）, and 7 factors were selected into the model. The discriminant accuracy rates of the four models were 74.6%, 70.1%, 76.1% and 96.9%, respectively, with the discriminant model 4 having the best accuracy. CONCLUSION： The models established with environmental and poor behavioral factors are suitable for basic health care facilities. Self- assessment of risk probability could be imple-mented using these simple and effective risk prediction models.

关 键 词：食管癌 风险预测 Logistic判别模型 

分 类 号：R735.1[医药卫生—肿瘤]