Development and validation of a predictive model for continuous renal replacement therapy in patients undergoing venovenous extracorporeal membrane oxygenation

Background: Acute kidney injury (AKI) and fluid overload are common complications of venovenous extracorporeal membrane oxygenation (VV-ECMO) and are associated with poor outcomes. Continuous renal replacement therapy (CRRT) is often required to manage these complications, but its initiation significantly increases mortality. Currently, there is no accurate prediction model tailored for CRRT in VV-ECMO patients. Therefore, this study aims to develop and validate a risk prediction model for CRRT initiation in VV-ECMO patients.

Methods: This retrospective, multicenter study included patients who underwent VV-ECMO at four hospitals in China. The derivation cohort comprised patients from one hospital, while data from three other hospitals were utilized for external validation. Candidate predictors were selected using logistic regression, least absolute shrinkage and selection operator (LASSO) regression, and the Boruta algorithm, and subsequently visualized in a nomogram.

Results: A total of 234 patients were included in the study, with 130 patients in the derivation cohort and 104 patients in the external validation cohort. Among these, 57 patients (43.85%) in the derivation cohort and 36 patients (34.62%) in the external validation cohort required CRRT. The final model incorporated the following predictors: coronary artery disease (CAD) {odds ratio (OR) [95% confidence interval (CI)]: 12.58 (3.60–44.03), P<0.001}, Sequential Organ Failure Assessment (SOFA) score [1.28 (1.10–1.48), P=0.001], platelet count (PLT) [0.99 (0.99–0.99), P=0.02], hemoglobin (HB) level [0.98 (0.96–0.99), P=0.03], and blood urea nitrogen (BUN) [1.04 (0.96–1.13), P=0.31]. The model exhibited the area under the curve (AUC) of 0.88 in the derivation cohort, and 0.75 in external validation.

Conclusions: This predictive model, incorporating five key predictors, CAD, SOFA score, PLT, HB, and BUN, serves as a practical and reliable tool for assessing CRRT initiation in VV-ECMO patients, facilitating early risk stratification, timely renal risk assessment and implementation of nephroprotective strategies. Further prospective multicenter validation is needed to confirm its generalizability.

Keywords: Venovenous extracorporeal membrane oxygenation (VV-ECMO); continuous renal replacement therapy (CRRT); risk prediction model; nomogram; multivariable analysis

 

Highlight box

Key findings

• This study establishes the first validated multivariable prediction model that incorporates five clinically accessible parameters—coronary artery disease, Sequential Organ Failure Assessment, platelet count, hemoglobin, and blood urea nitrogen, demonstrating strong discrimination with an area under the curve of 0.88 (derivation) and 0.75 (external validation).

What is known and what is new?

• While extracorporeal membrane oxygenation (ECMO) patients requiring continuous renal replacement therapy (CRRT) demonstrate a higher mortality risk compared to those not, there is a notable lack of tools specifically designed to predict the need for CRRT during venovenous (VV)-ECMO.

• This validated nomogram provides a clinically practical tool for early risk stratification of CRRT need in VV-ECMO patients, potentially enhance patient management and improve outcomes in VV-ECMO patients.

What is the implication, and what should change now?

• This predictive model serves as a practical and reliable tool for assessing CRRT initiation in VV-ECMO patients, facilitating early risk stratification and timely interventions.


Cite this article as: Xie J, Han L, Ouyang Y, et al. Development and validation of a predictive model for continuous renal replacement therapy in patients undergoing venovenous extracorporeal membrane oxygenation. J Thorac Dis 2025;17(10):7559-7570. doi: 10.21037/jtd-2025-735