Severe acute kidney injury (AKI) occurs frequently in intensive care unit (ICU) settings and correlates strongly with adverse patient outcomes. Whether renal replacement therapy (RRT) can effectively improve mortality outcomes in these patients remains controversial, primarily due to interference from confounding by indication. Within nephrology and critical-care practice, identifying which severe-AKI subgroups derive a survival benefit from RRT is a question of direct bedside relevance, and observational analyses with rigorous bias control are particularly informative when balanced randomised data remain limited. This retrospective cohort analysis utilized the MIMIC-IV database (version 2.2). The study population comprised adult patients (age ≥ 18 years) diagnosed with KDIGO Stage 3 AKI during their first ICU admission. Vascular causes of AKI (renal-artery stenosis, renal infarction, abdominal-aortic-aneurysm associated ischaemia) and clearly post-obstructive AKI were excluded a priori, as these phenotypes carry distinct trajectories that may not respond to RRT in a comparable manner. Patients were stratified into treatment and control groups based on RRT receipt. The clinical indication for RRT was operationalised through a composite set of pre-specified criteria (refractory hyperkalaemia, severe metabolic acidosis, refractory volume overload, uraemic complications, and oligo-anuria) that were entered into the propensity-score model in addition to severity scores and laboratory derangement, in order to reduce residual operator-dependent variation in the initiation decision. To address baseline characteristic differences, propensity score matching (PSM) was conducted using 1:1 nearest neighbor matching with a caliper of 0.02. The primary endpoint was in-hospital all-cause mortality; secondary endpoints encompassed ICU mortality, 28-day mortality, ICU length of stay, total hospital length of stay, RRT-related complications (catheter-associated bloodstream infection, mechanical line complications, bleeding events, and intradialytic hypotension), and renal recovery. Cox proportional hazards regression models were employed to calculate hazard ratios (HRs) with corresponding 95% confidence intervals (CIs). Pre-specified subgroup analyses also stratified by chronic-kidney-disease status, RRT modality, and the qualifying KDIGO Stage-3 criterion (creatinine, urine output, or both). A total of 5,847 patients with KDIGO Stage 3 AKI were enrolled, of whom 2,156 (37%) underwent RRT. Pre-matching analysis revealed substantially greater disease severity in the RRT group (SOFA score: 12.4 ± 4.2 vs. 8.6 ± 3.8, P < 0.001). Following 1:1 PSM, 1,842 matched pairs were established, with all covariate standardized mean differences (SMDs) below 0.1, indicating satisfactory matching. Within the matched cohort, in-hospital mortality was lower in the RRT group than in the controls (35% vs. 42%, P = 0.002), and the corresponding mortality risk was reduced (HR = 0.78, 95%CI: 0.68–0.89, P < 0.001). The direction and magnitude of the association were consistent across CKD versus non-CKD strata, across the qualifying KDIGO Stage-3 criterion, and across RRT modality, and RRT-related complications were observed but did not exceed rates previously reported in the literature. Subgroup analyses confirmed consistent findings of RRT regardless of sepsis status, mechanical ventilation use, age, or disease severity. After the largest confounders for indication were controlled through propensity score matching, RRT was, at a minimum, non-inferior to non-RRT management with respect to in-hospital mortality, and the matched-cohort estimates were consistent with an association with reduced mortality, among patients with severe AKI. These findings should not be read as proof of a causal protective effect, but they support the position that, in appropriately selected critically ill patients with KDIGO Stage 3 AKI – particularly those with refractory hyperkalaemia, severe acidosis, oligo-anuria, or refractory volume overload – RRT should not be withheld on the basis of perceived futility.
Tian et al. (Mon,) studied this question.