整合多系统代谢指标，结合临床研究数据，建立代谢相关脂肪性肝病并发冠心病的预测模型。

选取2020年1月至2025年1月间在新疆生产建设兵团医院确诊的代谢相关脂肪性肝病（MAFLD）患者共1251例。根据是否合并冠心病分为冠心病组（190例）和非冠心病组（1061例）。对临床常用指标实施单因素分析，筛选出组间差异显著的指标，进一步纳入多因素Logistic回归分析，以确定危险因素。

单因素分析结果显示，冠心病组在年龄（均值差为-12.35岁，P=5.4e-37）、肝功能指标（谷丙转氨酶升高11.72 U/L、谷氨酰转肽酶升高23.58 U/L）、血脂水平（甘油三酯升高0.56 mmol/L、LDL-C升高0.43 mmol/L）及尿蛋白含量（升高49.69 mg/dl）等方面均显著异常（P＜0.05）。利用XGboost模型，结合年龄、尿蛋白、谷草转氨酶及谷氨酰转肽酶等关键预测因子，构建了用于二分类预测的模型。该模型经80%训练集与20%测试集验证，ROC曲线下面积（AUC）达0.798，五折交叉验证AUC为0.85，敏感度0.89，特异度0.91，f1-score 0.9，较传统单一指标评估方法优越性显著。

本研究建立的MAFLD冠心病风险预测模型具有较高临床适用性，为实施个体化筛查和分层管理提供了量化工具。建议对MAFLD患者常规开展尿蛋白、肝酶及年龄联合评估，早期识别高危人群并启动生活方式干预或药物治疗，从而降低心血管事件发生率。

The aim of this study was to integrate multiple metabolic indicators from diverse systems and clinical research data to develop a predictive model for metabolic associated fatty liver disease (MAFLD) complicated by coronary heart disease (CHD).

Patients diagnosed with MAFLD between January 2020 and January 2025 were recruited from Xinjiang Production and Construction Corps Hospital. Those with malignant tumors or cognitive impairments were excluded. The patients were categorized into two groups according to the presence or absence of concurrent coronary heart disease. Univariate analysis was performed on commonly used clinical indicators to screen out those with significant inter-group differences. These selected indicators were then subjected to multivariate Logistic regression analysis to identify risk factors.

Univariate analysis revealed significant abnormalities in the CHD group regarding age (mean difference: -12.35 years, P=5.4e-37), liver function markers (ALT increased by 11.72 U/L, GGT increased by 23.58 U/L), lipid profiles (TG increased by 0.56 mmol/L, LDL-C increased by 0.43 mmol/L), and urinary protein levels (increased by 49.69 mg/dl) (P＜0.05). A binary predictive model was developed using logistic regression, incorporating key predictors including age, urinary protein, AST, and GGT. The model demonstrated superior performance with an area under the curve of 0.798 in the 80% training set and 0.85 in cross-validation, exhibiting a sensitivity of 0.89 sensitivity, specificity of 0.9, and f1-score of 0.9, significantly outperforming traditional single-indicator assessment methods.

The MAFLD-CHD risk prediction model established in this study exhibits high clinical applicability, providing a quantitative tool for individualized screening and stratified management. Routine combined assessment of urinary protein, liver enzymes, and age is recommended for MAFLD patients to facilitate early identification of high-risk populations and initiate lifestyle interventions or pharmacological treatments, thereby reducing cardiovascular event incidence.