Association between glycemic variability and the risk of acute kidney injury in patients with traumatic brain injury: a retrospective cohort study with independent cohort analysis

Background: Acute kidney injury (AKI) is a common complication among critically ill patients with traumatic brain injury (TBI) and is associated with adverse clinical outcomes. Glycemic variability (GV), reflecting short-term fluctuations in blood glucose, may contribute to organ dysfunction; however, its relationship with AKI in ICU patients with TBI remains incompletely characterized. Methods: We conducted a retrospective cohort study using the MIMIC-IV database, including 2,151 adult ICU patients with TBI, and performed replication in an independent cohort of 265 patients to evaluate reproducibility and incremental prognostic value. GV was quantified as the coefficient of variation (CV) of glucose measurements obtained during ICU stay until the occurrence of AKI or ICU discharge. Multivariable logistic regression models were used to examine the association between GV and AKI, with stepwise adjustment for potential confounders. Predictive performance was assessed using receiver operating characteristic (ROC) analysis, while dose-response relationships were explored with restricted cubic spline models. The incremental prognostic utility of adding GV to SOFA- and APACHE II-based models was evaluated using calibration plots, decision curve analysis, Integrated Discrimination Improvement (IDI), and Net Reclassification Improvement (NRI). Results: AKI occurred in 59.1% (1,271/2,151) of patients in the MIMIC-IV cohort. Higher GV was independently associated with AKI across all adjusted models (fully adjusted model: OR 1.16, 95% CI 1.02-1.34). ROC analysis yielded an area under the curve of 0.73 (95% CI 0.71-0.75) with 66% sensitivity and 69.7% specificity. Restricted cubic spline analyses suggested an approximately linear increase in AKI risk with rising GV. The association was consistent across prespecified subgroups, robust to adjustment for glucose monitoring intensity, and replicated in the independent cohort. Incorporating GV into SOFA- and APACHE II-based models led to modest but statistically significant improvements in risk stratification. Conclusion: In ICU patients with TBI, greater glycemic variability is independently associated with subsequent AKI and provides incremental prognostic information beyond established severity scores. These findings highlight the potential utility of GV as a complementary marker for risk stratification. Prospective studies using standardized GV definitions and harmonized glucose monitoring strategies are warranted to confirm these results and clarify their clinical implications.