南方医科大学学报2026，Vol.46Issue(4)：770-784,15.DOI:10.12122/j.issn.1673-4254.2026.04.06
全球非酒精性脂肪性肝病负担的关键决定因素:基于GBD数据的机器学习联合孟德尔随机化验证
Key determinants of global burden of non-alcoholic fatty liver disease:machine learning combined with Mendelian randomization analysis based on GBD data
摘要
Abstract
Objective To analyze the global trends,drivers,and health inequalities of non-alcoholic fatty liver disease(NAFLD)burden to identify key predictors of NAFLD-related mortality.Methods Using data from the Global Burden of Disease(GBD)Study 2021,we extracted global measures of NAFLD from 1990 to 2021,and the trends were analyzed using joinpoint regression.Decomposition analysis was used to quantify the contributions of population growth,aging,and epidemiological changes.The health inequality was assessed using the concentration index.Using XGBoost-SHAP machine learning,the mortality predictors were identified,and two-sample Mendelian randomization was employed to test the causality for the key factors.All the analyses were conducted with data stratification by sex and the socio-demographic index(SDI).Results The global age-standardized disability-adjusted life years(DALYs)rate showed an increasing trend in both males(average annual percentage change[AAPC]=+0.34%)and females(AAPC=+0.05%).Decomposition analysis revealed that population growth was the primary driver of the global increase in DALYs,while population aging contributed to 52.37%of male deaths in high-SDI regions.Health inequality analysis showed a concentration index of-0.05 for DALYs in 2021,indicating a concentration of burden among low-SDI populations.Machine learning identified smoking(relative importance=100%)and advanced age(70-74 years:60%)as the most critical predictors of mortality,and the model demonstrated good fit on the test set(R2=0.98).SDI-stratified analysis showed smoking and aging are the top two predictors across all SDI regions.Mendelian randomization further confirmed positive causal associations of smoking initiation(OR=1.35,P<0.05)and aging(proxied by frailty index,OR=2.01,P<0.05)with NAFLD risk.Conclusion NAFLD burden is heavy globally with significant sex and socioeconomic inequalities.Smoking and advanced age are key risk factors for NAFLD,calling for integrated interventions for tobacco control,geriatric health management,and health equity promotion.关键词
非酒精性脂肪性肝病/全球疾病负担/机器学习/孟德尔随机化/社会人口指数Key words
non-alcoholic fatty liver disease/Global Burden of Disease/machine learning/Mendelian randomization/socio-demographic Index引用本文复制引用
陈浩,李振汉,纪梦佳,汪鑫诚,陈博峰,管谦,武嫚,卢林明..全球非酒精性脂肪性肝病负担的关键决定因素:基于GBD数据的机器学习联合孟德尔随机化验证[J].南方医科大学学报,2026,46(4):770-784,15.基金项目
国家自然科学基金(82500322) （82500322）
广西自然科学基金(2025GXNSFHA069262) （2025GXNSFHA069262）
国家级大学生创新创业训练计划(202510368012) （202510368012）
安徽省大学生创新创业训练计划(S202410368116) Supported by National Natural Science Foundation of China(82500322). （S202410368116）
