食管癌危险因素及预测模型的研究

Research on risk factors and prediction models of esophageal cancer

目的分析食管癌的相关危险因素并建立预测模型,为食管癌的机会性筛查提供数据基础,探讨更优筛查方案。方法收集仪陇县人民医院2017年1月至2021年12月胃镜筛查者数据,进行多因素Logistic回归分析,将筛选出的影响因素建立预测模型,利用ROC计算曲线下面积(AUC)、特异性及敏感性等指标对预测结果进行评价,并通过十折交叉验证对模型的稳定性进行验证。结果性别、年龄、萎缩性胃炎、肿瘤家族史、吸烟、饮酒、腌制酸菜、烫食是食管癌的危险因素(P<0.05)。构建的食管癌预测模型受试者工作曲线下面积为0.783(0.759~0.807);十折交叉试验后表明模型的稳定好。结论建立的食管癌人群预测模型具有较好的预测价值,可用于选择性接受内窥镜检查的早癌精准筛查,提高筛查效益,节约经济成本。

Objective To analyze the related risk factors of esophageal cancer and establish a predictive model.This was to provide a data base for the opportunistic screening of esophageal cancer and explore a better screening program.Methods Data of patients undergoing gastroscopy screening in our hospital from January 2017 to December 2021 were collected.The data were analyzed by using multivariate logistic regression analysis and a predictive model was established based on selected influencing factors.The area under ROC curve(AUC),the specificity,the sensitivity and other indicators were used to evaluate the prediction results.The stability of the model was verified through 10-fold cross-validation.Results Gender,age,atrophic gastritis,family history of cancer,smoking,drinking,pickled sauerkraut,and hot food were risk factors for esophageal cancer(P<0.05).The area under the subject working curve of the esophageal cancer prediction model was 0.783(0.759-0.807).The ten-fold crossover test showed that the model was stable.Conclusions The established predictive model of esophageal cancer population has good predictive value.It can be used for selective endoscopic screening of early cancer in order to improve the screening efficiency and save the economic cost.