Answers are grounded in this paper and cite their sources.
Objective To systematically summarise and critically appraise multivariable diagnostic prediction models for MAFLD and related fatty-liver phenotypes in adults with type 2 diabetes mellitus (T2DM). Methods Eight databases (four international and four Chinese) were searched from inception to May 2026. Two reviewers independently screened studies and extracted data using the CHARMS framework. Risk of bias and applicability were assessed with PROBAST. Because of substantial clinical and methodological heterogeneity, results were synthesised descriptively rather than pooled. Results Eighteen diagnostic studies (study-level: 7 MAFLD, 8 NAFLD, and 3 MASLD) reporting 52 distinct diagnostic model specifications were included. No eligible prognostic studies or prognostic models were identified. Studies on NAFLD and MASLD are regarded as indirect supporting evidence; their performance evaluations neither were combined nor directly compared with the MAFLD model. Feng Leyi et al. (2024) used samples recruited from the same hospital at different times, but the validation period came before the development period, thus the independence of the participants could not be confirmed. It was not classified as an independent external validation. The remaining studies reported significant development-set performance or internal validation; No study provided clear independent external validation. BMI, TG, and HDL-C were the most frequently included predictors, but this recurrence is descriptive and does not establish stable predictor effects or clinical utility. AUC values were not directly comparable across heterogeneous populations, outcome definitions, and validation categories. Therefore, no summary AUC range was calculated, and model-level values are categorised where attribution to a specific model and dataset was supported by the source report. PROBAST indicated a high overall risk of bias, predominantly related to identifiable limitations in the analysis domain; insufficient reporting alone was not equated with high risk of bias. Conclusions Existing diagnostic models show discrimination within their own study settings, but apparent, internally validated, and externally validated AUCs are not directly comparable, and cross-framework extrapolation is not justified. Future research should develop and validate prognostic models for incident disease, use appropriate sample-size planning and missing-data methods, model continuous predictors appropriately, report full model specifications and calibration, assess clinical utility, and conduct rigorous external validation in representative populations. Systematic review registration: https://www.crd.york.ac.uk/PROSPERO , identifier CRD420261439469.
No takes yet. Share an insight, caveat, or question.
Li et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: