Abstract
BACKGROUND: The prevalence of metabolic-associated steatotic liver disease (MASLD) is rising precipitously among children, particularly in regions or countries burdened with high prevalence of obesity. However, identifying those at high risk remains a significant challenge, as the majority do not exhibit distinct symptoms of MASLD. There is an urgent need for a widely accepted non-invasive predictor to facilitate early disease diagnosis and management of the disease. Our study aims to 1) evaluate and compare existing predictors of MASLD, and 2) develop a practical screening strategy for children, tailored to local prevalence of obesity. METHODS: We utilized a school-based cross-sectional survey in Beijing as the training dataset to establish predictive models for screening MASLD in children. An independent school-based study in Ningbo was used to validate the models. We selected the optimal non-invasive MASLD predictor by comparing logistic regression model, random forest model, decision tree model, and support vector machine model using both the Beijing and Ningbo datasets. This was followed by serial testing using the best performance index we identified and indices from previous studies. Finally, we calculated the potential MASLD screening recommendation categories and corresponding profits based on national and subnational obesity prevalence, and applied those three categories to 200 countries according to their obesity prevalence from 1990 to 2022. FINDINGS: A total of 1018 children were included (N(Beijing) = 596, N(Ningbo) = 422). The logistic regression model demonstrated the best performance, identifying the waist-to-height ratio (WHtR, cutoff value ≥0.48) as the optimal noninvasive index for predicting MASLD, with strong performance in both training and validation set. Additionally, the combination of WHtR and lipid accumulation product (LAP) was selected as an optimal serial test to improve the positive predictive value, with a LAP cutoff value of ≥668.22 cm × mg/dL. Based on the obesity prevalence among 30 provinces, three MASLD screening recommendations were proposed: 1) "Population-screening-recommended": For regions with an obesity prevalence ≥12.0%, where MASLD prevalence ranged from 5.0% to 21.5%; 2) "Resources-permitted": For regions with an obesity prevalence between 8.4% and 12.0%, where MASLD prevalence ranged from 2.3% to 4.4%; 3) "Population-screening-not-recommended": For regions with an obesity prevalence <8.4%, where MASLD prevalence is difficult to detect using our tool. Using our proposed cutoff for screening MASLD, the number of countries classified into the "Population-screening-recommended" and "Resources-permitted" categories increased from one and 11 in 1990 to 95 and 28 in 2022, respectively. INTERPRETATION: WHtR might serve as a practical and accessible index for predicting pediatric MASLD. A WHtR value ≥0.48 could facilitate early identification and management of MASLD in areas with obesity prevalence ≥12.0%. Furthermore, combining WHtR ≥0.48 with LAP ≥668.22 cm × mg/dL is recommended for individual MASLD screening. Moreover, linking these measures with population obesity prevalence not only helps estimate MASLD prevalence but also indicates potential screening profits in regions at varying levels of obesity risk. FUNDING: This study was supported by grants from Capital's Funds for Health Improvement and Research (Grant No. 2022-1G-4251), National Natural Science Foundation of China (Grant No. 82273654), Major Science and Technology Projects for Health of Zhejiang Province (Grant No. WKJ-ZJ-2216), Cyrus Tang Foundation for Young Scholar 2022 (2022-B126) and Sino-German Mobility Programme (M-0015).