Testosterone-related indices and adverse glucose-lipid metabolic status in men: a single-center retrospective study
TL;DR
In male andrology outpatients, lower TT was associated with an adverse glucose-lipid profile, particularly higher glucose and TG and lower HDL-C, and machine learning-assisted validation further supported the interpretive importance of SHBG, although prediction performance remained exploratory.
Abstract
Background Lower testosterone is often observed with abnormal glucose-lipid metabolism, but the clinical value of calculated testosterone fractions in metabolic assessment remains uncertain. This study evaluated cross-sectional associations of total testosterone (TT), calculated free testosterone (FT), free testosterone percentage (FT%), bioavailable testosterone (BioT), bioavailable testosterone percentage (BioT%), and sex hormone-binding globulin (SHBG) with glucose-lipid traits in male andrology outpatients. Methods This single-center retrospective cross-sectional analysis included 420 adult male andrology outpatients with complete records for TT, SHBG, albumin, glucose, total cholesterol (TC), triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C). Age-adjusted partial Spearman correlations, rank-standardized regression, SHBG-adjusted models, restricted cubic spline analysis, age-stratified interaction analysis, sensitivity analyses, and exploratory machine learning-assisted validation were performed. Results TT was inversely associated with glucose and TG and positively associated with HDL-C. FT% and BioT% were positively associated with TG and inversely associated with HDL-C, but these associations were markedly attenuated or reversed after additional adjustment for SHBG. Restricted cubic spline models supported approximately linear associations of TT with glucose, TG, and HDL-C. In machine learning-assisted validation, adverse glucose-lipid metabolic status was present in 285 participants (67.9%). The Age + TT + SHBG model and XGBoost showed modest discrimination, with AUCs of 0.691 and 0.688, respectively. SHAP analysis suggested SHBG and age were the strongest model contributors, but this ranking should be interpreted as exploratory because correlated predictors can affect SHAP-based importance estimates. Conclusions In male andrology outpatients, lower TT was associated with an adverse glucose-lipid profile, particularly higher glucose and TG and lower HDL-C. Percentage-based calculated testosterone indices appeared to reflect SHBG-dependent redistribution of circulating testosterone rather than an independent favorable androgen state. Machine learning-assisted validation further supported the interpretive importance of SHBG, although prediction performance remained exploratory. These findings should be interpreted within a single-center outpatient cohort and should not be generalized directly to the broader male population.