基于主成分分析-Logistic回归方法的人体农兽药化学污染物残留与高胆固醇血症关系的研究  

Principal component analysis based Logistic regression analysis of the association between pesticide and veterinary chemical pollutant residues and hypercholesterolemia in humans

作　　者：李梦梦 牛然 宋琪哲 苏畅 王惠君 张兵 武振宇[1] LI Mengmeng;NIU Ran;SONG Qizhe;SU Chang;WANG Huijun;ZHANG Bing;WU Zhenyu(School of Public Health,Fudan University,Shanghai 200032,China;National Institute for Nutrition and Health,Chinese Center for Disease Control and Prevention/Key Laboratory of Trace Element Nutrition of National Health Commission,Beijing 100050,China)

机构地区：[1]复旦大学公共卫生学院,上海200032 [2]中国疾病预防控制中心营养与健康所/国家卫生健康委微量元素与营养重点实验室,北京100050

出　　处：《中国食品卫生杂志》2023年第5期639-644,共6页Chinese Journal of Food Hygiene

基　　金：国家重点研发计划(2019YFC1605100);国家自然科学基金(81573155,82173613);上海市卫生健康委科研项目(202140018)。

摘　　要：目的基于主成分分析-Logistic(PCA-Logistic)回归组合分析策略探索人体农兽药化学污染物残留与高胆固醇血症之间的关系。方法研究数据来源于“降低成年超重者营养相关慢性病风险的适宜身体活动量研究”2018—2019年调查数据。选择有人口学信息、体格测量和血清农兽药化学污染物及血脂检测等数据的496名成年居民作为研究对象。采用常规Logistic与PCA-Logistic分别建立回归模型分析高胆固醇血症的危险因素,利用受试者工作特征曲线下面积(AUC)评价模型判别效果。结果PCA-Logistic回归分析结果显示对高胆固醇血症具有显著影响的农兽药化学污染物包括呋喃它酮、2-乙基己基二苯基磷酸酯(EHDPP)、全氟己烷磺酸(PFHxS)和4-氯苯氧乙酸(4-CPA)。PCA-Logistic回归模型对高胆固醇血症预测性能(AUC=0.736)优于常规Logistic回归模型的预测性能(AUC=0.695)。结论呋喃它酮、EHDPP、PFHxS、4-CPA等农兽药化学污染物残留是影响高胆固醇血症的重要危险因素,PCA-Logistic回归分析方法可有效应用于探索农兽药化学污染物残留与高胆固醇血症的关联性分析。Objective To explore the association of human pesticide and veterinary chemical pollutant residues on hypercholesterolemia based on the principal component analysis(PCA)-Logistic regression model.Methods The data were collected from a survey done in a 2018 to 2019 study on the appropriate physical activity to reduce the risk of nutrition-related chronic disease in overweight adults.A total of 496 adult residents with demographic information,physical measurements,serum pesticides and veterinary chemical pollutants,and blood lipid testing data were selected as the research subjects.Logistic and PCA-Logistic regression analyses were used to establish the models separately to explore the risk factors of hypercholesterolemia.The discrimination ability of the models was evaluated by the area of under curve(AUC).Results PCA-Logistic regression analysis revealed that pesticide and veterinary chemical pollutant residues,including furaltadone,2-ethylhexyl diphenyl phosphate(EHDPP),perflurohexane sulfonate(PFHxS),and 4-chlorophenoxyacetate(4-CPA),had significant impacts on hypercholesterolemia.The PCA-Logistic regression model showed better distinguishing performance than the ordinary Logistic model(AUC 0.736 vs 0.695).Conclusion Pesticide and veterinary chemical pollutant residues that include furaltadone,EHDPP,PFHxS,4-CPA are important risk factors for hypercholesterolemia.PCA-Logistic regression analysis can be effective to explore the association between human pesticide and veterinary chemical pollutant residues and hypercholesterolemia.

关 键 词：高胆固醇血症 农兽药化学污染物残留 主成分分析 模型预测 

分 类 号：R155[医药卫生—营养与食品卫生学]