Metabolic and Anthropometric Predictors of Paediatric Metabolic Steatotic Liver Disease: Development of a Clinical Risk Prediction Model

Abstract

Background. Paediatric metabolic dysfunction-associated steatotic liver disease (MASLD) is common yet frequently undetected, and access to imaging is limited in many settings. A simple risk tool built from routinely available variables could prioritize children for confirmatory assessment. We aimed to develop and internally validate a multivariable model predicting ultrasound-detected MASLD from metabolic and anthropometric data.

Methods. In a cross-sectional study of 205 children and adolescents (aged 5–17 years) with metabolic risk factors, MASLD was ascertained by abdominal ultrasonography. Candidate predictors comprised demographic, anthropometric, biochemical, and family-history variables. A multivariable logistic regression model was developed using clinical reasoning and parsimony, with continuous predictors transformed where appropriate. Discrimination (C-statistic), calibration (flexible calibration curve, calibration slope, Hosmer–Lemeshow test), and clinical utility (decision curve analysis) were assessed; optimism was estimated by bootstrapping (1000 resamples). A simplified integer points score was derived and stratified into risk categories. The study follows the TRIPOD statement.

Results: Ultrasound-detected MASLD was present in 109 of 205 participants (53.2%). The final model retained three predictors: body mass index (BMI), log-transformed ALT, and family history of dyslipidaemia. BMI was associated with MASLD with an adjusted odds ratio (aOR) of 1.81 (95% CI, 1.25–2.62) per 5 kg/m² increase, while a two-fold increase in ALT was associated with an aOR of 2.62 (95% CI, 1.67–4.10). Family history of dyslipidaemia was also independently associated with MASLD (aOR, 1.87; 95% CI, 1.02–3.43). The apparent C-statistic was 0.741, with an optimism-corrected C-statistic of 0.730 and a bootstrap-corrected calibration slope of 0.95. Graphical assessment suggested generally acceptable calibration across most of the predicted-risk range, although the Hosmer–Lemeshow test indicated statistically significant departure from perfect fit. Decision-curve analysis suggested positive net benefit of the model across clinically relevant threshold probabilities. A simplified points score showed a C-statistic of 0.711 and separated participants into lower-, intermediate-, and higher-predicted-risk groups.

Conclusions: A parsimonious model incorporating BMI, ALT, and family history of dyslipidaemia showed moderate discrimination and generally acceptable calibration for ultrasound-detected MASLD among children and adolescents with metabolic risk factors. The model may have potential as a triage tool to prioritize children for further hepatic assessment where imaging resources are limited. However, the model and its simplified points score require external validation in independent and more representative paediatric populations before clinical application.

Journal of Paediatric Surgeons of Bangladesh (2026) Vol. 17 (2): 52-58