Abstract
Acute kidney injury (AKI) represents a complex disorder characterized by distinct subphenotypes with varied clinical presentations and prognoses. Categorizing these subphenotypes may facilitate standardization of research cohorts and optimization of therapeutic strategies. The endothelial activation and stress index (EASIX) quantifies thrombotic microangiopathy severity, a pathophysiological hallmark of AKI. Consequently, we utilized EASIX trajectory analysis to identify AKI subphenotypes. AKI patients were identified from the eICU Collaborative Research Database to develop a group-based trajectory model. EASIX scores recorded during the initial seven ICU days were utilized for trajectory modeling. Patients were stratified into distinct subgroups according to the best model. Variable selection was performed using LASSO regression, followed by multivariate Cox regression analyses to calculate hazard ratios (HRs) across the identified subgroups. An independent validation cohort comprised patients from the central ICU of West China Hospital (WCH). The study's primary endpoints included all-cause in-ICU and in-hospital mortality across the identified subphenotypes. The final analysis included 317 patients from the eICU database and 58 patients from WCH. Based on the EASIX trajectories derived from the first seven ICU days, we identified two distinct subphenotypes: a "Stably High" (SH) group and a "Decreasing" (D) group. Compared to the D group, the SH group demonstrated significantly higher mortality risk, with an HR of 2.26 (95% CI 1.14-4.26, p = 0.018) for ICU mortality and 1.85 (95% CI 1.03-3.29, p = 0.038) for 30-day in-hospital mortality. These findings were replicated in the WCH validation cohort. This study identified and validated two distinct AKI subphenotypes through EASIX trajectory analysis, demonstrating significant heterogeneity in clinical characteristics, laboratory findings, comorbidities, and outcomes between these groups. Future research may focus on early subphenotype prediction, differential treatment responses, and molecular mechanisms driving inter-group variation.