基于贝叶斯网络模型的非酒精性脂肪肝发病路径分析  

Analysis of the pathogenesis of non-alcoholic fatty liver disease based on Bayesian network model

作　　者：汤建敏 王璇 陈硕[2] 韩玉梅[2] 孔邻润 杨兴华[1] TANG Jian-min;WANG Xuan;CHEN Shuo;HAN Yu-mei;KONG Lin-run;YANG Xing-hua(School of Public Health,Capital Medical University,Beijing 100069,China;不详)

机构地区：[1]首都医科大学公共卫生学院,北京100069 [2]北京市体检中心

出　　处：《现代预防医学》2025年第1期21-26,共6页Modern Preventive Medicine

基　　金：北京市自然科学基金(7202010)。

摘　　要：目的 识别非酒精性脂肪肝(NAFLD)关键影响因素,并构建贝叶斯网络模型探索其发病路径。方法 研究基于北京健康管理队列。通过LASSO回归筛选NAFLD的相关影响因素,使用10折交叉验证选择最优惩罚参数λ。基于筛选出的变量,应用贝叶斯网络模型进行结构学习,并通过贝叶斯信息准则(BIC)和交叉验证评估模型拟合优度,进行路径分析。结果 5年随访期内,30 001名研究对象中有22.83%发生NAFLD(n=6 850),发病密度为45.7/1 000人年。筛选出6个关键影响因素:AGE、BMI、HDL-C、FPG、ALT、DBP,并基于这些因素构建贝叶斯网络模型。贝叶斯网络分析揭示了23条NAFLD发病路径,以BMI为根节点的路径最多,有12条,其次年龄组有5条。其中AGE→DBP→FPG→NAFLD路径在45岁以下人群中的发病概率显著高于45岁以上人群;BMI为偏低状态下,BMI→HDL→ALT→NAFLD等路径显示NAFLD的发病概率高于其他BMI组。结论 贝叶斯网络模型揭示了NAFLD的发病路径,明确了关键因素对其发病的影响,为NAFLD的早期筛查和干预提供了理论依据。Objective To identify key influencing factors of non-alcoholic fatty liver disease(NAFLD)and construct a Bayesian network model to explore its pathogenesis.Methods The study was based on the Beijing Health Management Cohort.The LASSO regression was employed to screen for relevant influencing factors of NAFLD,and 10-fold cross-validation was used to select the optimal penalty parameterλ.Based on the selected variables,structural learning was conducted using the Bayesian network model,and the model fit was evaluated through the Bayesian Information Criterion(BIC)and cross-validation for path analysis.Results During the 5-year follow-up period,22.83%of the 30001 study participants developed NAFLD(n=6850),with an incidence density of 45.7 per 1000 person-years.Six key influencing factors were identified:AGE,BMI,HDL-C,FPG,ALT,and DBP,and a Bayesian network model was constructed based on these factors.The Bayesian network analysis revealed 23 pathways for the development of NAFLD,with the pathway having BMI as the root node being the most frequent(12 pathways),followed by age groups(5 pathways).The pathway AGE→DBP→FPG→NAFLD showed a significantly higher incidence probability in individuals under 45 years compared to those over 45.In cases of low BMI,the pathway BMI→HDL→ALT→NAFLD indicated a higher probability of NAFLD compared to other BMI groups.Conclusion The Bayesian network model elucidated the pathways of NAFLD development and clarified the impact of key factors on its onset,providing a theoretical basis for early screening and intervention of NAFLD.

关 键 词：非酒精性脂肪性肝病 LASSO回归 贝叶斯网络 发病路径 

分 类 号：R575.5[医药卫生—消化系统]