Association of the triglyceride-glucose index with one-year all-cause mortality in patients with spontaneous basal ganglia hemorrhage: a retrospective study

BACKGROUND: The triglyceride-glucose (TyG) index is a marker of insulin resistance, and its relationship with the prognosis of patients with spontaneous basal ganglia hemorrhage (SBGH) has not been clearly established. This study aimed to evaluate the relationship between the TyG index and one-year all-cause mortality. METHOD: This single-center, hospital-based retrospective study included 282 patients with SBGH. Missing values were imputed using a random forest algorithm. The primary outcome was one-year all-cause mortality, which was confirmed by standardized telephone follow-ups. Variables considered for multivariable modeling were prespecified based on established clinical relevance and prior literature, after which bidirectional stepwise selection was used to select the final model covariates. Association between the TyG index and mortality was assessed using Cox proportional hazards models. Additional analyses included Kaplan-Meier survival curves and restricted cubic spline (RCS) analyses. Incremental predictive performance of adding TyG to the baseline model was assessed using the AUC, IDI, and NRI. Robustness was examined using E-values and sensitivity analyses, including complete-case analyses, exclusion of extreme values, further adjustment for AST and INR, and subgroup analyses. RESULTS: Among the 282 participants, 49 (17.4%) died during the one-year follow-up. The results revealed significant associations between the TyG index and mortality. Specifically, for each 1-SD increase in the TyG index, the risk of all-cause mortality increased by 61.1% (HR = 1.611, 95% CI = 1.184-2.191). RCS analyses suggested an approximately linear relationship (P for nonlinearity > 0.05) and identified a cutoff value of 8.499. Patients with TyG ≥ 8.499 had worse one-year survival (log-rank P = 0.002). Adding the TyG index to the basic model increased the AUC from 0.691 to 0.723 (optimism-corrected 0.661 vs. 0.703; ΔAUC was not significant), improved IDI (0.044, P = 0.047), but not continuous NRI (0.090, P = 0.299). Sensitivity analyses supported the robustness of these associations. Subgroup analyses suggested that the association was generally consistent, with limited evidence of interaction. CONCLUSION: The TyG index may aid early risk stratification in SBGH. However, further research is required to determine if improved TyG index control would lead to better clinical results in the future.