The investigators identified 1,377 adult intensive care admissions for acute ischemic stroke from the MIMIC‑IV electronic critical‑care database. The cohort included patients who were at risk during their ICU stay; 680 of those patients developed KDIGO stage 2 - 3 acute kidney injury while hospitalized. The abstract did not report additional case‑selection details such as exclusion criteria, timing of admission relative to stroke onset, or the precise definition applied to identify an ICU episode.
The study team first applied univariable logistic regression to screen potential predictors of KDIGO stage 2 - 3 AKI. They then used a bidirectional stepwise regression process to identify confounding variables that would enter adjusted models. The abstract reports those analytic steps without presenting the full list of candidate variables or stepwise selection criteria; the precise selection thresholds and model diagnostics were not described in the abstract.
Investigators reported a threshold effect for base excess (BE). Specifically, BE values below −5.124 mmol/L were associated with higher odds of developing KDIGO stage 2 - 3 AKI. After adjustment for selected covariates, the association persisted: adjusted odds ratio (OR) = 1.51 with a 95% confidence interval of 1.05–2.16. The abstract provides that point estimate and interval but does not include additional measures such as model discrimination or calibration metrics.
The team also examined alactic base excess (aBE), reporting an initial unadjusted association for aBE < −3.444 mmol/L (OR = 1.57, 95% CI 1.12–2.20). In the fully adjusted model, aBE below that threshold remained associated with elevated odds of KDIGO stage 2 - 3 AKI (adjusted OR = 1.43, 95% CI 1.00–2.04). The abstract presents these estimates and confidence bounds; it does not provide the number of events within each threshold stratum or tests for interaction.
To evaluate dose–response behavior, the investigators plotted restricted cubic splines (RCS) and reported non‑linear associations between both BE and aBE and the risk of KDIGO stage 2 - 3 AKI. They also performed subgroup analyses stratified by age, Glasgow Coma Scale (GCS), SOFA, and comorbid conditions, summarizing subgroup relationships with a forest plot. The abstract does not include the spline curves, subgroup point estimates by stratum, or formal heterogeneity statistics; those visualizations were referenced but not shown in the abstract text.
The adjusted analyses, as described in the abstract, included SOFA score, serum sodium, and exposure to several medications: anticoagulants, furosemide, and aspirin. Those covariates were listed as confounders retained by the stepwise procedure. The study team therefore treated acid–base indices (BE, aBE) as independent predictors after accounting for organ‑failure burden and selected laboratory and medication exposures. The abstract does not specify whether other common AKI risk factors—such as baseline estimated glomerular filtration rate, contrast exposure, or vasopressor use—were evaluated or excluded.
The authors concluded that low base excess or low alactic base excess were associated with increased risk of KDIGO stage 2 - 3 AKI in ICU patients admitted with AIS. The abstract documents key associations but leaves several methodological questions unanswered: it does not report timing of BE/aBE measurement relative to ICU admission or stroke onset, it does not present the full covariate list considered during selection, and it does not show external validation or predictive performance metrics. The abstract also lacks stratified event counts, model fit statistics, and details on how missing data were handled. These gaps constrain interpretation of causality, timing, and generalizability and represent potential next steps for readers and investigators seeking to apply or validate the findings.