Abstract: Objective To construct a prediction model of non-alcoholic fatty liver disease in non-obese population based on lifestyle-related indicators. Methods A prospective cohort study was conducted to select 992 non-obese subjects who participated in physical examination in a tertiary grade-A hospital in Shanghai from April to October 2021 and were followed up until October 2023. The subjects were confirmed whether they had developed non-alcoholic fatty liver disease through abdominal ultrasound during the follow-up and then they were divided into non-occurrence group (n=834) and occurrence group (n=158). General information, results of physical examination and laboratory examination and lifestyle-related indicators were compared between the two groups. LASSO regression and logistic regression were used to analyze the influencing factors and build a prediction model, and a nomogram was drawn. Results Through multi-factor analysis, male (OR=2.375), age (OR=1.044), high income level (OR=0.512), body mass index of 24.0~27.9(OR=3.556), intake of eggs (OR=1.008), vegetable (OR=0.998) and sugar-sweetened beverage (OR=1.002), sitting time (OR=1.199) and sleep time (OR=0.661) were used to construct the prediction model. The area under ROC curve of the model was 0.846, and the area under ROC curve of internal validation was 0.819. The Hosmer-Lemeshow test results of χ²=7.478, P=0.486 indicated good fit. Calibration curve and decision curve analysis showed that the model had good calibration degree and was clinically applicable. Conclusion The prediction model constructed in this study is scientific and practical, with good prediction efficiency, and can easily and effectively predict the risk of non-alcoholic fatty liver disease in non-obese population.

Key words: non-obese, non-alcoholic fatty liver disease, lifestyle, prediction model, nomogram