Systolic blood pressure strata and acute kidney injury risk in heart failure: a retrospective cohort study

In patients with acute heart failure (HF), both hypotension and hypertension may compromise renal perfusion, yet the association between systolic blood pressure (SBP) and acute kidney injury (AKI) risk is controversial and poorly quantified. This study aimed to define this relationship. In this retrospective cohort analysis of 19,604 HF patients from the MIMIC-IV database, individuals were categorized into three SBP strata based on their mean SBP during the first 24 hours of intensive care unit admission: strict (<110 mmHg), standard (110–139 mmHg), and lenient (≥140 mmHg). We employed multivariable logistic regression, propensity score weighting, and restricted cubic spline (RCS) analysis to evaluate associations. To further address confounding, we incorporated standardized ICU severity scores (OASIS, SAPS II, APACHE III) into the models. Sensitivity analyses were performed using alternative SBP cutoffs and in a subcohort with available BNP measurements. A predictive model was developed and internally validated via bootstrapping. The overall incidence of AKI was 36.6%. After multivariable adjustment, compared with the lenient SBP stratum (≥140 mmHg), both strict (<110 mmHg; aOR 0.74, 95% CI 0.67-0.82) and standard (110-139 mmHg; aOR 0.77, 95% CI 0.69-0.86) SBP strata were associated with significantly lower odds of AKI. These associations remained robust after adjustment for OASIS (strict: aOR 0.71, 0.64–0.79; standard: aOR 0.75, 0.68–0.84), SAPS II (strict: aOR 0.67, 0.60–0.75; standard: aOR 0.75, 0.67–0.83), and APACHE III (strict: aOR 0.69, 0.62–0.77; standard: aOR 0.76, 0.68–0.85). RCS analysis revealed a significant non-linear relationship, with the lowest predicted AKI risk at an SBP of approximately 120 mmHg. Key independent risk factors included chronic kidney disease (aOR 2.64) and elevated 48-hour creatinine ratio (aOR 2.07 per unit). The full prediction model demonstrated moderate discrimination (area under the curve AUC = 0.726, 95% CI: 0.718–0.733) but significant calibration issues (Hosmer-Lemeshow P < 0.001). Internal validation yielded an optimism-corrected AUC of 0.726. Notably, in sensitivity analyses focusing on severe AKI, both standard and lenient strategies were associated with significantly lower risk compared to strict control. The simplified model (6 variables) showed poor calibration (Hosmer–Lemeshow P < 0.001, Brier 0.205) but provided positive net benefit in decision curve analysis across thresholds 0–0.7; however, its poor calibration limits clinical application, and the net benefit estimates should be interpreted with caution. Compared with the lenient SBP stratum, both strict and standard SBP strata were associated with significantly lower odds of overall acute kidney injury in heart failure patients. However, the relationship is non-linear, with the lowest predicted risk observed at approximately 120 mmHg. These findings suggest that an observed SBP of approximately 120 mmHg was associated with the lowest risk of AKI in this population, warranting prospective validation.