Abstract
OBJECTIVE: To explore the association between the triglyceride-glucose (TyG) index at admission and subsequent acute kidney injury (AKI) in adult patients with diabetic ketoacidosis (DKA), and to examine whether a data-driven threshold could aid early risk stratification. METHODS: We conducted a single-centre retrospective cohort study of adults admitted with DKA. Clinical and laboratory variables were extracted and curated from electronic medical records. The TyG index was derived from routine triglyceride and glucose measurements obtained at admission. AKI during hospitalization was defined by KDIGO creatinine criteria. We modelled TyG using restricted cubic splines and in quartiles; multivariable logistic regression adjusted for prespecified confounders. Subgroup analyses evaluated effect modification by sex, age, BMI, diabetes type, and hypertension. RESULTS: A total of 678 adults with DKA were included; 196 (28.9%) developed AKI. The TyG-AKI relationship was non-linear, with risk rising steeply once TyG exceeded 10.92 (Q4). In fully adjusted models, Q4 vs Q1 was associated with higher odds of AKI (OR 2.478, 95% CI 1.285-4.780; p = 0.007). No significant interactions were detected across prespecified subgroups. In sensitivity analyses, findings were directionally consistent. The association was observed across strata without evidence of effect modification. CONCLUSIONS: We observed a non-linear association between admission TyG and subsequent AKI in adults with DKA, with an inflection around TyG ≈ 10.92. These results may help identify patients who warrant closer renal surveillance, yet they should be regarded as exploratory and hypothesis-generating given the single-centre, retrospective design; causality cannot be inferred, and prospective multicentre validation is required before routine clinical adoption.