Front Med (Lausanne). 2026 Apr 21;13:1807526. doi: 10.3389/fmed.2026.1807526. eCollection 2026.
ABSTRACT
BACKGROUND: Heart transplantation (HT) is an effective treatment for end-stage heart disease, but postoperative acute kidney injury (AKI) requiring continuous renal replacement therapy (CRRT) is associated with poor outcomes. Although risk factors for AKI after HT have been well established, studies specifically focusing on CRRT as a clinical endpoint and employing rigorous predictive modeling remain limited. This study aims to identify perioperative risk factors for CRRT after HT and to develop a validated prediction model.
METHODS: This single-center retrospective study included HT recipients from April 2018 to November 2023. Patients requiring CRRT within 7 days after surgery were compared with those who did not. Candidate predictors were pre-selected based on clinical rationale and previous literature. LASSO regression was used for variable selection to prevent overfitting. A multivariable logistic regression model was then constructed and internally validated using 1,000 bootstrap resamples. Model performance was assessed by discrimination (optimism-corrected AUC), calibration (calibration plot, Hosmer-Lemeshow test), overall fit (Brier score), and clinical utility (decision curve analysis). A time-dependent Cox proportional hazards model was used to evaluate the association between CRRT and postoperative mortality, thereby avoiding immortal time bias.
RESULTS: Among 213 recipients, 30 (14.1%) received CRRT. LASSO regression identified nine key predictors: preoperative hemoglobin, preoperative total bilirubin, preoperative ECMO use, cardiopulmonary bypass time, intraoperative blood loss, red blood cell transfusion volume, mechanical ventilation time, VIS score, and lactate peak. Considering the limited number of events (EPV = 10), three core variables were ultimately included in the multivariable model: preoperative hemoglobin (OR 0.963, 95%CI 0.937-0.986, p = 0.003), VIS score (OR 1.282, 95%CI 1.175-1.423, p < 0.001), and lactate peak (OR 2.032, 95%CI 1.464-2.986, p < 0.001). The model demonstrated good discrimination (optimism-corrected AUC = 0.885) and excellent calibration (Hosmer-Lemeshow test p = 0.678). Decision curve analysis confirmed a positive net benefit across clinically relevant threshold probabilities. After adjusting for confounders, CRRT remained independently associated with increased mortality (adjusted HR = 6.957, 95%CI 3.669-13.192, p < 0.001).
CONCLUSION: Patients requiring CRRT after HT have a markedly poorer prognosis. The internally validated multifactorial model, incorporating preoperative hemoglobin, VIS score, and lactate peak, provides robust predictive value for CRRT and may facilitate early risk stratification. The significant association between CRRT and mortality underscores the critical need for targeted perioperative interventions in high-risk patients.
PMID:42094977 | PMC:PMC13138926 | DOI:10.3389/fmed.2026.1807526