Exploration of mortality-associated creatinine thresholds for acute kidney injury in cardiac surgery patients with impaired kidney function: a single-center retrospective cohort study

Abstract

Background

Acute kidney injury (AKI) is a frequent and severe complication following cardiac surgery, particularly in patients with impaired kidney function. The existing Kidney Disease: Improving Global Outcomes (KDIGO) criteria do not specifically address acute-on-chronic kidney injury in this high-risk population. Previous studies have proposed alternative diagnostic thresholds that identify more AKI cases than KDIGO and are associated with adverse outcomes. However, their association with endpoints including mortality and clinical utility in cardiac surgery patients remain unclear. This study aims to explore optimal perioperative serum creatinine (SCr) change thresholds associated with in-hospital mortality and compare their predictive performance with KDIGO and other proposed thresholds.

Methods

This retrospective cohort study included 1,081 adult cardiac surgery patients with impaired preoperative kidney function (eGFR 15–60 mL/min/1.73 m²). Postoperative SCr changes were assessed as maximum absolute increases within 48 h and maximum fold increases within 7 days. Multivariable Cox regression and restricted cubic spline (RCS) analyses were used to evaluate associations with endpoints including in-hospital mortality, the initiation of KRT, failure of kidney function recovery by hospital discharge (or death), and major adverse kidney events (MAKE), defined as a composite of in-hospital mortality, dialysis dependence at discharge, or non-recovery of kidney function by hospital discharge (or death). Optimal thresholds were derived using receiver operating characteristic (ROC) curve analysis and Youden’s index. The predictive performance and net clinical benefit for in-hospital mortality were compared across KDIGO definition, previously proposed thresholds, and the newly derived thresholds using ROC and decision curve analysis (DCA).

Results

Both the absolute increase in SCr within 48 h and the fold increase within 7 days were independently associated with in-hospital mortality (HR 1.66 and 1.59, respectively), RRT (OR 3.10 and 3.62, respectively), kidney function non-recovery (OR 1.43 and 1.38, respectively), and MAKE (OR 2.32 and 2.24, respectively). For in-hospital mortality, the optimal thresholds identified were 38 µmol/L and 2.177-fold, respectively. ROC analysis showed comparable predictive performance with KDIGO definition and other standards. For in-hospital mortality, decision curve analysis suggested a marginally higher net benefit for the new thresholds within the 10–30% threshold probability range.

Conclusions

This study proposes new SCr thresholds specific to cardiac surgery patients with impaired kidney function. If externally validated, these thresholds may aid in improving risk stratification and guiding perioperative management. Nonetheless, further studies are warranted to refine diagnostic approaches to AKI in this high-risk population.