Machine Learning Algorithms in Controlled Donation After Circulatory Death Under Normothermic Regional Perfusion - A Graft Survival Prediction Model

Aug 5, 2026

Machine learning classifiers are evaluated for predicting graft survival in controlled donation after circulatory death (cDCD) livers recovered using normothermic regional perfusion (NRP), addressing inadequate performance of existing graft-loss risk scores in the Spanish population. A retrospective multicenter cohort study analyzes 539 donor-recipient pairs with 20 donor, recipient, and NRP variables. Logistic regression showed the best discrimination for 3- and 12-month outcomes, enabling a 3- and 12-month risk score integrated into a donor-recipient matching framework to support liver allocation.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages9

Reading Time23 min

AccessFree to download

Background

Methods

Results

Conclusions

Introduction

Original Clinical Science—Liver Machine Learning Algorithms in Controlled Donation After Circulatory Death Under Normothermic Regional Perfusion: A Graft Survival Prediction Model Rafael Calleja, MD,1 Marcos Rivera, BSc,2 David Guijo-Rubio, PhD,2 Amelia J. Hessheimer, MD, PhD,3 Gloria de la Rosa, MD, PhD,4 Mikel Gastaca, MD, PhD,5 Alejandra Otero, MD, PhD,6 Pablo Ramírez, MD, PhD,7 Andrea Boscà-Robledo, MD, PhD,8 Julio Santoyo, MD, PhD,9 Luis Miguel Marín Gómez, MD, PhD,10 Jesús Villar del Moral, MD, PhD,11 Yiliam Fundora, MD, PhD,12 Laura Lladó, MD, PhD,13 Carmelo Loinaz, MD, PhD,14 Manuel C. Jiménez-Garrido, MD, PhD,15 Gonzalo Rodríguez-Laíz, MD, PhD,16 José Á . López-Baena, MD, PhD,17 Ramón Charco, MD, PhD,18 Evaristo Varo, MD, PhD,19 Fernando Rotellar, MD, PhD,20 Ayaya Alonso, MD,21 Juan C. Rodríguez-Sanjuan, MD, PhD,22 Gerardo Blanco, MD, PhD,23 Javier Nuño, MD, PhD,24 David Pacheco, MD, PhD,25 Elisabeth Coll, MD, PhD,4 Beatriz Domínguez-Gil, MD, PhD,4 Constantino Fondevila, MD, PhD,3 María Dolores Ayllón, MD, PhD,1 Manuel Durán, MD, PhD,1 Ruben Ciria, MD, PhD,1 Pedro A. Gutiérrez, PhD,2 Antonio Gómez-Orellana, MSc,2 César Hervás-Martínez, PhD,2 and Javier Briceño, MD, PhD1 Background. Several scores have been developed to stratify the risk of graft loss in controlled donation after circulatory death (cDCD) . However, their performance is unsatisfactory in the Spanish population, where most cDCD livers are recovered using normothermic regional perfusion (NRP) . Consequently, we explored the role of different machine learning-based classifiers as predictive models for graft survival. A risk stratification score integrated with the model of end-stage liver disease score in a donor-recipient (D-R) matching system was developed. Methods. This retrospective multicenter cohort study used 539 D-R pairs of cDCD livers recovered with NRP, including 20 donor, recipient, and NRP variables. The following machine learning-based classifiers were evaluated: logistic regression, ridge classifier, support vector classifier, multilayer perceptron, and random forest. The endpoints were the 3-and 12-mo graft survival rates. A 3-and 12-mo risk score was developed using the best model obtained. Results. Logistic regression yielded the best performance at 3 mo (area under the receiver operating characteristic curve = 0 .82) and 12 mo (area under the receiver operating characteristic curve = 0 .83) . A D-R matching system was proposed on the basis of the current model of end-stage liver disease score and cDCD-NRP risk score. Conclusions. The satisfactory performance of the proposed score within the study population suggests a significant potential to support liver allocation in cDCD-NRP grafts. External validation is challenging, but this methodology may be explored in other regions. (Transplantation 2025;109: e362–e370) . Received 27 February 2024. Revision received 21 September 2024. Accepted 17 October 2024. 1 Hepatobiliary Surgery and Liver Transplantation Unit, Maimonides Biomedical Research Institute of Cordoba (IMIBIC), Hospital Universitario Reina Sofía, University of Córdoba, Córdoba, Spain. 2 Department of Computational Sciences and Numerical Analysis, University of Córdoba, Córdoba, Spain. 3 General and Digestive Surgery Department, Hospital Universitario La Paz, Madrid, Spain. 4 Organización Nacional de Trasplantes, Madrid, Spain. 5 Hepatobiliary Surgery and Liver Transplantation Unit, Biocruces Bizkaia Health Research Institute, Cruces University Hospital, University of the Basque Country, Bilbao, Spain. 6 General and Digestive Surgery Department, Complejo Hospitalario Universitario deA Coruña, A Coruña, Spain. 7 General and Digestive Surgery Department, Hospital Clínico Universitario Virgen de la Arrixaca, IMIB, El Palmar, Spain. 8 General and Digestive Surgery Department, Hospital Universitari i Politècnic La Fe, Valencia, Spain. 9 General and Digestive Surgery Department, Hospital Regional Univer

FAQs

Why were existing graft-loss risk scores re-evaluated for the Spanish cDCD population?

Their performance is unsatisfactory in Spain, where most cDCD livers are recovered using normothermic regional perfusion (NRP). The study therefore explores new predictive modeling approaches.

Which machine learning classifiers were tested in the retrospective multicenter cohort?

The study evaluates logistic regression, ridge classifier, support vector classifier, multilayer perceptron, and random forest using donor-recipient pairs recovered with NRP.

What model performed best for predicting graft survival, and for which time points?

Logistic regression achieved the best performance at 3 months and 12 months, with area under the ROC curve values reported as 0.82 and 0.83, respectively.

Aug 5, 2026

Upload by Seraphina