BMJ Open · Nov 11, 2026

Machine learning-based prediction of non-alcoholic fatty liver disease in Chinese perimenopausal women: a retrospective observational study

By Yang CM, Liu YF, Gao HY et al.

Objective This study aimed to compare the predictive performance of different machine learning (ML) models for non-alcoholic fatty liver disease (NAFLD) in perimenopausal women and to interpret the optimal model using Shapley additive explanations (SHAP). Design Retrospective observational study. Setting A health management centre of China. Participants A total of 5637 perimenopausal women aged 45–55 years who underwent health examinations were included between February 2022 and August 2023. Methods Physical examination data were collected from perimenopausal women at the Health Management Center of the First Affiliated Hospital of the University of Science and Technology of China. Least absolute shrinkage and selection operator regression was employed for feature selection. Eight ML models were developed with hyperparameter tuning. The optimal model was selected based on the area under the receiver operating characteristic curve (AUC), which was calculated on the training set and was subsequently interpreted. Results Among 5637 participants, 1324 (23.5%) were diagnosed with NAFLD. The Extreme Gradient Boosting (XGBoost) model exhibited the best predictive performance, with an AUC of 0.907 and an accuracy of 0.822 in the validation set. SHAP analysis identified that body mass index, triglycerides, high-density lipoprotein, random blood glucose, alanine aminotransferase, hyperuricaemia, platelet count and albumin were the most influential predictors of NAFLD in this population. Conclusion The XGBoost model demonstrated superior NAFLD prediction performance compared with other ML methods in the perimenopausal women and might improve health management for this population.

doi.org/10.1136/bmjopen-2026-116467Open accessBiomarkers

AgingResearch.News