Body mass index accounts for most of the discrimination of the triglyceride glucose–body mass index for newly detected hepatic steatosis in older adults with type 2 diabetes and hypertension: a retrospective cohort study

Background: Composite indices such as the triglyceride glucose–body mass index (TyG-BMI) are widely proposed for stratifying the risk of metabolic dysfunction-associated steatotic liver disease. Reported discrimination is high, but is almost always presented without a benchmark against body mass index (BMI), a component of the index already recorded at every routine visit. We compared TyG-BMI with TyG-WC and six conventional lipid ratios for newly detected, ultrasound-defined hepatic steatosis, then asked how much of its discrimination BMI alone accounts for. Methods: All 898 participants had type 2 diabetes and hypertension and attended three annual examinations (2023–2025); the 453 free of baseline steatosis with complete covariates formed the analytic cohort. The primary estimand was cumulative risk over the fixed window to 2025, from logistic regression excluding the components of TyG-BMI as covariates. The pre-specified comparison ranked TyG-BMI against TyG-WC and the lipid ratios; benchmarking against BMI and waist circumference was added afterwards and is reported as post-hoc, with collinearity, multi-model and incremental-value analyses. Results: Steatosis was newly detected in 203/453 participants (44.8%; a detection rate, not incidence). TyG-BMI had the largest association (OR per 1-SD 2.03, 95% CI 1.62–2.54) and the highest single-index discrimination (AUC 0.682, 0.632–0.731), exceeding TyG, LDL/HDL-C, NHHR and AIP after Holm correction but not TyG-WC or UHR. BMI alone reached 0.670 (ΔAUC 0.011, 95% CI −0.017 to 0.040; DeLong P=0.45); TyG-BMI correlated with BMI at r=0.877, and the two could not be entered jointly (variance inflation factor up to 309). BMI and TyG entered separately performed as well (0.6922 vs 0.6925). Adding TyG-BMI to a BMI-containing clinical model moved the apparent AUC from 0.680 to 0.692 (IDI 0.015, 0.001–0.045), with a net-benefit increment of at most 0.024. Conclusions: The discrimination of TyG-BMI resided in its adiposity component: BMI alone, recorded at every visit at no additional cost, accounted for nearly all of it, and the composite added no demonstrable value beyond it. The comparison was underpowered to exclude a small advantage, so this is absence of demonstrated benefit rather than proven equivalence. Discrimination reported for composite metabolic indices should be interpreted against the simplest measurement they contain.Note: This version has been substantially reframed and retitled. Version 1.0 (10.5281/zenodo.21775796) remains available.