Abstract / Summary
Abstract Background and Objective Obesity is a major global risk factor for multiple chronic diseases. Prediction models have increasingly been developed to support early identification and prevention; however, evidence regarding their performance, methodological robustness, and applicability across different obesity-related diseases remains fragmented. We conducted a systematic review to synthesise prediction models for obesity-related diseases, focusing on model performance, predictors, validation characteristics, and risk of bias. Methods We searched PubMed, Web of Science, Embase, and CNKI from Jan 1, 2020 to Feb 1, 2026. Models developed to estimate the current or future risk of obesity-related diseases were eligible, including cross-sectional risk classification models and longitudinal incident-risk models. Models predicting prognosis, treatment response, or post-diagnosis outcomes were excluded. Two reviewers independently screened studies and extracted data. Model characteristics, performance (discrimination, calibration, and clinical utility), and validation methods were summarised. Risk of bias was assessed using the PROBAST and PROBAST+AI. Results We included 24 studies comprising 43 prediction models across multiple outcomes, including metabolic, cardiovascular, and sleep-related diseases. Most models showed favourable discrimination, with area under the curve values ranging from 0.715 to 0.961. However, methodological limitations were common. Half of the models were at high risk of bias, mainly due to deficiencies in the analysis domain. External validation was rare, with only 4.7% of models validated in independent populations. Predictors were predominantly routine clinical variables, such as age, body mass index, and metabolic biomarkers. Conclusions Although existing models demonstrate promising discriminatory performance, their methodological limitations and lack of external validation may limit reliability and generalisability. Future research should prioritise methodological rigour and external validation to improve clinical applicability.