Abstract: FR-PO1193
Integrating Donor-Derived Cell-Free DNA for the Detection of Kidney Allograft Rejection Under the Banff 2024 Framework: An Extended Multinational Validation Study
Session Information
- Transplantation: Clinical - Rejection, Biomarkers, and Pharmacology
October 23, 2026 | Location: Exhibit Hall A, Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Transplantation
- 2002 Transplantation: Clinical
Authors
- Brousse, Romain H J, Université Paris Cité Paris Institute for Transplantation and Organ Regeneration, Paris, France
- Sablik, Marta, Université Paris Cité Paris Institute for Transplantation and Organ Regeneration, Paris, France
- Mandelbrot, Didier A., Division of Nephrology, Department of Medicine, University of Wisconsin School of Medicine and Public Health, Madison, Wisconsin, United States
- Parajuli, Sandesh, Division of Nephrology, Department of Medicine, University of Wisconsin School of Medicine and Public Health, Madison, Wisconsin, United States
- Gupta, Gaurav, Division of Nephrology, Virginia Commonwealth University, Richmond, Virginia, United States
- Kumar, Dhiren, Division of Nephrology, Virginia Commonwealth University, Richmond, Virginia, United States
- Lefaucheur, Carmen, Université Paris Cité Paris Institute for Transplantation and Organ Regeneration, Paris, France
- Loupy, Alexandre, Université Paris Cité Paris Institute for Transplantation and Organ Regeneration, Paris, France
Group or Team Name
- Paris Institute for Transplantation and Organ Regeneration
Background
Allograft rejection is poorly detected by standard-of-care biomarkers. The integrative donor-derived cell-free DNA model (iDd-cfDNA) demonstrated strong diagnostic performance under Banff 2019. The Banff 2022 revision introduced microvascular inflammation without DSA or C4d (MVI, DSA-, C4d-) and probable AMR as distinct injury phenotypes. Whether iDd-cfDNA retains its performance against this expanded outcome is unknown.
Methods
We enrolled 4,976 kidney allograft recipients from 29 adult and pediatric centers across four continents. A total of 6,827 biopsies (1,415 derivation, 5,412 validation) were reclassified using the Banff Automation Tool. The pre-specified iDd-cfDNA model was assessed alongside five alternative models, including a DSA-free variant, using discrimination, calibration, and a simulated biopsy-decision analysis.
Results
Reclassification identified 485 additional rejection cases (370 MVI DSA-/C4d-; 115 probable AMR). Applied without coefficient re-estimation, iDd-cfDNA achieved an AUC of 0.833 (95% CI 0.815–0.852) against the Banff 2024 outcome, matching the re-estimated model (0.832, 0.814–0.851), with preserved calibration. dd-cfDNA-based models substantially improved discrimination over standard-of-care across AMR, TCMR, mixed rejection and the newly defined MVI DSA-/C4d- phenotype (Figure). A DSA-free variant retained strong performance (AUC 0.813, 0.795–0.831) and outperformed all other models for MVI DSA-/C4d- and TCMR. In a simulated biopsy-decision analysis, integrative models avoided more biopsies and missed fewer rejections than both fixed dd-cfDNA cutoffs and clinical models without dd-cfDNA.
Conclusion
iDd-cfDNA generalizes to Banff 2024 without recalibration and improves rejection detection across phenotypes, including where DSA testing is unavailable, supporting its broad clinical implementation.
Acknowledgment
We thank Maarten Naesens, Edmund Huang, Stanley Jordan, Renato De Marco, David Wojciechowski, Arnaud Del Bello, Nassim Kamar, Rouba Garro, Julien Hogan, Pranjal Jain, Miklos Molnar, Katalin Fornadi, Raja Dandamudi, Vikas Dharnidharka, Oriol Bestard, Tarek Alhamad, Basmah Abdalla, Magali Giral, Sophie Brouard, Sanjiv Anand, Michael Nunley, Dechu Puliyanda, Darshana Dadhania, Eric Thervet, Evgenia Preka, Fadi Haidar, Klemens Budde, Fabian Halleck, Juhan Lee, Minsun Jung, and Emilio Poggio.
Performance of complete and sparse prediction models, with and without dd-cfDNA, to detect Banff 2024 rejection.