Abstract
Abstract
Among 209 hospitalized patients with type 2 diabetes mellitus (T2DM), including 65 with sarcopenia defined by the 2025 Asian Working Group for Sarcopenia criteria, we evaluated an age-adjusted multiparametric rectus femoris ultrasound model and compared it with serum and combined models. Ultrasound variables were rectus femoris thickness, echogenicity, and shear-wave elastography; serum variables were the triglyceride-glucose (TyG) and systemic immune-inflammation (SII) indices. The age-adjusted ultrasound model had an apparent AUC of 0.986 (95% CI 0.976-0.997), an optimism-corrected calibration slope of 0.825, and a cross-validated AUC of 0.983 (SD 0.005). The age-only AUC was 0.796; adding the ultrasound variables increased the AUC by 0.190 (DeLong P = 5.0 x 10^-11). Among participants aged 65 years or older (n = 120), the age-only AUC was 0.603. Adding TyG and SII increased the apparent AUC from 0.986 to 0.995 (DeLong P = 0.045), but categorical reclassification was not significant (P = 0.106), decision-curve gains were small, and the combined-model optimism-corrected calibration slope was 0.484. After removal of 27 observations flagged by influence diagnostics, conventional estimation of the combined model did not yield finite estimates for several coefficients; in the full cohort, Firth penalized estimation yielded a TyG_Z odds ratio of 10.86 (95% CI 2.59-88.33). The findings support further evaluation of multiparametric rectus femoris ultrasound as a candidate case-finding approach in this setting. The incremental role of TyG and SII remains uncertain, and external validation is required before clinical use.