Machine Learning for Predicting Pulmonary Graft Dysfunction After Double-Lung Transplantation - A Single-Center Study Using Donor, Recipient, and Intraoperative Variables

Q: What complication does the study focus on?

The study focuses on grade 3 primary graft dysfunction at 72 hours (PGD3-T72) following double-lung transplantation.

Q: How was the predictive model developed?

Researchers retrospectively analyzed perioperative data from 477 patients and trained supervised machine-learning models (XGBoost and logistic regression) using grid search and cross-validation.

Q: Which model performed better and what were key predictors?

XGBoost outperformed logistic regression, with its best performance at the second graft implantation. Top predictors included ECMO use, blood lactate levels, PaO2/FiO2 ratio, and total lung capacity mismatch.

Aug 5, 2026

Upload by Genevieve

Grade 3 primary graft dysfunction at 72 h (PGD3-T72) is a severe complication after lung transplantation. The study aimed to create an intraoperative machine-learning tool to predict PGD3-T72 using detailed perioperative data. Retrospective analysis covered 477 double-lung transplant recipients from a single center (2012–2019). Supervised models, including XGBoost and logistic regression, were trained with grid-search and cross-validation, and XGBoost showed superior discrimination.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages14

Reading Time35 min

AccessFree to download

Introduction

Materials and Methods

Study Design

*Correspondence Julien Fessler, [juf4007@med.cornell.edu](juf4007@med.cornell.edu) Received: 26 May 2025 Accepted: 30 September 2025 Published: 22 October 2025 Citation: Fessler J, Gouy-Pailler C, Ma W, Devaquet J, Messika J, Glorion M, Sage E, Roux A, Brugière O, Vallée A, Fischler M, Le Guen M and Komorowski M (2025) Machine Learning for Predicting Pulmonary Graft Dysfunction After Double-Lung Transplantation: A Single-Center Study Using Donor, Recipient, and Intraoperative Variables. Transpl. Int. 38:14965. doi: 10.3389/ti.2025.14965 Machine Learning for Predicting Pulmonary Graft Dysfunction After Double-Lung Transplantation: A Single-Center Study Using Donor, Recipient, and Intraoperative Variables Julien Fessler 1,2 *, Cédric Gouy-Pailler3, Wenting Ma 2, Jerôme Devaquet 4, Jonathan Messika 4, Matthieu Glorion 5, Edouard Sage 5,6, Antoine Roux 6, 7, Olivier Brugière 7, Alexandre Vallée 8, Marc Fischler 1, Morgan Le Guen 1,6 and Matthieu Komorowski 9,10 1Department of Anesthesiology, Hôpital Foch, Suresnes, France, 2Department of Anesthesiology, Weill Cornell Medicine, New York, NY, United States, 3CEA, List, Université Paris-Saclay, Palaiseau, France, 4Department of Intensive Care Medicine, Hôpital Foch, Suresnes, France, 5Department of Thoracic Surgery, Hôpital Foch, Suresnes, France, 6Université Versailles-Saint-Quentinen-Yvelines, Versailles, France, 7Department of Pneumology, Hôpital Foch, Suresnes, France, 8Department of EpidemiologyData-Biostatistics, Delegation of Clinical Research and Innovation, Hôpital Foch, Suresnes, France, 9Intensive Care Unit, Charing Cross Hospital, London, United Kingdom, 10Division of Anaesthetics, Pain Medicine, and Intensive Care, Department of Surgery and Cancer, Faculty of Medicine, Imperial College London, London, United Kingdom Grade 3 primary graft dysfunction at 72 h (PGD3-T72) is a severe complication following lung transplantation. We aimed to develop an intraoperative machine-learning tool to predict PGD3-T72 . We retrospectively analyzed perioperative data from 477 patients who underwent double-lung transplantation at a single center between 2012 and 2019 . Data were structured into nine chronological steps, and supervised machine-learning models (XGBoost and logistic regression) were trained to predict PGD3-T72, with hyperparameters optimized via grid search and cross-validation. PGD3-T72 occurred in 83 patients (17 .3%) . XGBoost outperformed logistic regression, achieving peak performance at second graft implantation with an AUROC of 0.84 IQR: 0.065, p < 0.001, with a sensitivity of 0 .81 and a speciﬁcity of 0 .68. The top predictors included extracorporeal membrane oxygenation (ECMO) use, blood lactate levels, PaO2/FiO2 ratio, and total lung capacity mismatch. Subgroup analyses conﬁrmed robustness across ECMO and non-ECMO cohorts. PGD3-T72 can be reliably predicted intraoperatively, offering potential for early intervention. Keywords: lung transplantation, ECMO, primary graft dysfunction, machine-learning, gradient-boosting Transplant International | Published by Frontiers 1 October 2025 | Volume 38 | Article 14965 |
| | --- | |
GRAPHICAL ABSTRACT | | INTRODUCTION Following double-lung transplantations, grade 3 primary graft dysfunction at 72 h (PGD3-T72) is associated with increased risks of graft failure, bronchiolitis obliterans syndrome, and higher one-year mortality [1, 2] . Its incidence varies widely across centers, ranging from 3% to 25%, underscoring the need toreevaluate its risk factors while considering the evolving clinical practices. For instance, ex vivo lung perfusion has expanded the lung donor pool, extending the grafts’ ischemic times, with favorable outcomes [3, 4] . Likewise, tremendous strides have been made with the wider use of intraoperative extracorporeal membrane oxygenation (ECMO) [5] and its extension into the postoperative period [6] . Such dynamic changes in clinical practice, while beneﬁcial for pat

FAQs

What complication does the study focus on?

The study focuses on grade 3 primary graft dysfunction at 72 hours (PGD3-T72) following double-lung transplantation.

How was the predictive model developed?

Researchers retrospectively analyzed perioperative data from 477 patients and trained supervised machine-learning models (XGBoost and logistic regression) using grid search and cross-validation.

Which model performed better and what were key predictors?

XGBoost outperformed logistic regression, with its best performance at the second graft implantation. Top predictors included ECMO use, blood lactate levels, PaO2/FiO2 ratio, and total lung capacity mismatch.

Aug 5, 2026

Upload by Genevieve