The association between the triglyceride-glucose index and short-term mortality in ICU patients with sepsis-associated acute kidney injury

BackgroundThe triglyceride-glucose (TyG) index has emerged as a novel marker for insulin resistance and is commonly observed in patients suffering from sepsis-associated acute kidney injury (SA-AKI). This study explored the correlation between the TyG index and short-term all-cause mortality among SA-AKI patients.MethodsWe performed a retrospective analysis of ICU patients with SA-AKI using data from the MIMIC-IV database. The primary outcomes were 28-day and 90-day all-cause mortality. Multivariate Cox proportional hazards regression, restricted cubic spline (RCS) models, and Kaplan-Meier (K-M) survival analyses were used to examine the associations between the TyG index and mortality. Subgroup and sensitivity analyses were conducted to ensure the robustness of the findings.ResultsThe study included 4971 SA-AKI patients, with 2873 males (57.8%), an average age of 65.4 years (+/- 15.8), and an average TyG index of 9.10 (+/- 0.70). RCS analysis revealed a U-shaped relationship between the TyG index and mortality. When the TyG index was below 9.04, the risk of mortality at both 28 days and 90 days was reduced (adjusted HRs of 0.695, 95% CI: 0.542-0.890 and 0.691, 95% CI: 0.557-0.858, respectively). In contrast, values above 9.04 were associated with increased mortality, though the relationship was not statistically significant (adjusted HRs of 1.026, 95% CI: 0.855-1.231 and 1.012, 95% CI: 0.863-1.188, respectively). K-M analysis revealed higher mortality rates for patients with either high (T3) or low (T1) TyG indices than for those with moderate (T2) TyG indices. Sensitivity analyses confirmed these associations even after excluding patients with diabetes, cerebrovascular diseases, or ICU stays of less than 2 days.ConclusionThe TyG index is significantly and nonlinearly associated with short-term all-cause mortality in SA-AKI patients; however, establishing a causal relationship between the two requires validation through larger prospective studies.