Prognostic value of the triglyceride-glucose index combined with glycemic variability for all-cause mortality in patients with sepsis: A retrospective cohort study
Ao Wang, Yuechen Fu, Ran Li, Yiwen Wang, Mengmeng Zhuang, Qing Xu, Luqi Qiao, Jiayi Cao, Kehui Li, Simiao Chen, Chao Meng, Qian Hong, Wendi Chen, Wensheng Wang, Yong SunObjective
This study aimed to evaluate the predictive value of the combined triglyceride-glucose index (TyG) and glycemic variability (GV) for all-cause mortality in critically ill patients with sepsis.
Methods
This retrospective cohort study used MIMIC-IV v3.1 as the development cohort and the temporally independent MIMIC-III CareVue subset as the validation cohort. Patients were divided into six groups based on the median TyG index (9.34) and GV tertiles (≤20.54, 20.54–28.11, and >28.11). The primary outcome was 90-day all-cause mortality. Analyses were performed using Kaplan–Meier survival curves, Cox proportional hazards models, and restricted cubic splines, supplemented with subgroup analyses and external validation.
Results
This study included 1792 patients from MIMIC-IV and 947 patients from the MIMIC-III CareVue subset. In the fully adjusted Cox proportional hazards model, the G6 group (TyG > 9.34 and GV > 28.11) exhibited the highest 90-day all-cause mortality risk compared with the G1 group (TyG ≤ 9.34 and GV ≤ 20.54) (HR = 2.366, 95% CI: 1.595–3.509). The absolute 90-day mortality rate increased from 10.3% in G1 to 27.2% in G6. Kaplan–Meier analysis revealed significant differences in cumulative mortality across the TyG–GV risk groups during the 90-day follow-up (log-rank P < 0.001), with the lowest survival rate observed in G6. The combined TyG and GV model showed a modest but statistically significant improvement in discrimination compared with the baseline model for 90-day mortality, with the AUC increasing from 0.730 (95% CI, 0.700–0.760) to 0.743 (95% CI, 0.714–0.773) (ΔAUC = 0.013, P = 0.048). The continuous net reclassification improvement (cfNRI) and integrated discrimination improvement (IDI) further supported the incremental value of TyG and GV added to the baseline model. Subgroup analyses showed generally consistent directions of association without significant interactions, and external validation supported the robustness of these findings.
Conclusions
The combined phenotype of elevated TyG index (>9.34) and elevated GV (>28.11) was associated with increased short- and long-term all-cause mortality risks and provided supplementary prognostic information in critically ill patients with sepsis; prospective validation is needed.