An official website of the United States government

NIH: National Institute of Diabetes and Digestive and Kidney Diseases
NIDDK Central Repository

Publication Information

Public Release Type
Journal
Publication Year
2025
Affiliation
1Normandie Univ, UNICAEN, CHU de Caen Normandie, Néphrologie, CAEN, France 2Normandie université, Unicaen, UFR de médecine, 2 rue des rochambelles. Caen, France 3ANTICIPE” U1086 INSERM-UCN, Centre François Baclesse, Caen, France 4Skane university hospital, clinical studies Sweden forum south, Remissgatan 4, SE-221 85 Lund, Sweden 5Lund University, Malmö, Sweden 6Department of Public Health and Primary Care, KU Leuven, Campus Kulak, Kortrijk, Belgium 7Itec, imec Research Group, KU Leuven, Kortrijk, Belgium 8Lund University, Box 117, SE-221 00, Lund, Sweden. 9Östra Vallgatan 41, SE-223 61 Lund, Sweden 10 Department of Clinical Chemistry and Pharmacology, Laboratory Medicine, Lund University, SE-22185 Lund, Sweden. 11Karla Healthcare Centre, Faculty of Medicine and Health, Örebro University, 701 85 Örebro, Sweden. 12University Hospital of North Norway (UNN), Breivika, 9038 Troms., Norway. 13Division of Nephrology and Hypertension, Mayo Clinic, Rochester, Minnesota, USA 14Department of Clinical Science, Intervention and Technology, Division of Pediatrics, Karolinska Institutet, Karolinska University. Hospital Huddinge, Stockholm 14186, Sweden 15Department of Medicine Huddinge, Karolinska Institutet, C2:91 Karolinska University Hospital, SE-141 52 Huddinge, Sweden. 16Barnnjursektionen K 88, Astrid Lindgrens Barnsjukhus, Karolinska University Hospital, SE-141 86 Stockholm, Sweden. 17Clinical Chemistry and Pharmacology, Entrance 61, 2nd floor, Akademiska Hospital, SE-751 85 Uppsala, Sweden. 18Service de Physiologie-Explorations Fonctionnelles Renales Hopital Europeen Georges Pompidou, 20 rue Leblanc, 75015 Paris, France 19Exploration Fonctionnelle Renale Pavillon P, Hopital Edouard Herriot, 5 place d’Arsonval, 69437 Lyon cedex 03, France. 20CHU de Bordeaux, Nephrologie-Transplantation-Dialyse, Hopital Pellegrin, Universite de Bordeaux, place Amelie Raba Leon, 33076 Bordeaux, France. 21Renal Transplantation Department, Assistance Publique–Hopitaux de Paris (AP-HP), Hopital Bichat, 46 rue Henri Huchard, 75018 Paris, France. 22Department of nephrology, Clermont-Ferrand University hospital, Clermont-Ferrand, France 23Service de Nephrologie et Immunologie clinique, CHU de Nantes, 30 boulevard Jean Monnet, 44093 Nantes Cedex 1, France. 24Department of Nephrology and Organ Transplantation, CHU Rangueil, 1 avenue J.Poulhes, TSA 50032, 31059 Toulouse Cedex 9, France. 25Transplantation renale, Hopital Necker, 145 rue de sevres, 75015 Paris, France. 26Service de Nephrologie, Hemodialyse, Aphereses et Transplantation Renale, Hopital Michallon, Centre Hospitalier Universitaire Grenoble-Alpes, Boulevard de la Chantourne, 38700 La Tronche, France 27Institute of Public Health, Charité. Universitätsmedizin Berlin, Luisenstrasse 57, 10117 Berlin, Germany 28Boelelaan 1112, Amsterdam 1081 HV, the Netherlands. 29Clinical Biochemistry, Pathology, East Kent Hospitals University NHS Foundation Trust, Canterbury, Kent CT1 3NG, United Kingdom 30Service de Nephrologie, Dialyse et Transplantation Renale, Hopital Nord, CHU de Saint-Etienne, 25 Boulevard Pasteur, 42055, Saint-Etienne Cedex 2, France 31Department of Nephrology-Dialysis-Transplantation, University of Liège, CHU Sart Tilman, Liège, Belgium 32Department of Nephrology-Dialysis-Apheresis, Hôpital Universitaire Carémeau, Nîmes, France.
Authors
Lanot A, Akesson A, Nakano FK, Vens C, Björk J, Nyman U, Grubb A, Sundin P-O, Eriksen BO, Melsom T, Rule AD, Berg U, Littmann K, Åsling-Monemi K, Larsson A, Courbebaisse M, Dubourg L, Couzi L, Gaillard F, Garrouste C, Jacquemont L, Kamar N, Legendre C, Rostaing L, Ebert N, Schaeffner E, Bökenkamp A, Lamb EJ, Mariat C, Pottel H, Delanaye P

Abstract

Rationale & Objective: Creatinine-based estimated glomerular filtration rate (eGFR) equations are widely used in clinical practice but exhibit inherent limitations. We developed and validated machine learning models to assess the trustworthiness (i.e. the ability of equations to estimate measured GFR (mGFR) within 10, 20% or 30%) of the European Kidney Function Consortium (EKFC) equation. Study design: Observational study Setting & Participants: Data from European and US cohorts, comprising 22,343 participants of all ages with available mGFR results were utilized. Predictors: Creatinine, age, sex, height, weight, body mass index, EKFC. Outcome: Probability that EKFC falls within 30% of mGFR. Analytical Approach: Four machine learning and two traditional logistic regression models were trained on a cohort of 8,455 participants to predict whether the EKFC creatinine-derived eGFR fell within 30% (P30), 20% (P20) or 10% (P10) of the mGFR value. The algorithms were internally and then externally validated on cohorts of respectively 3,781 and 10,107 participants. Results: The random forest model proved to be the most robust model. In the external validation cohort, the model achieved an area under the curve (AUC) of 0.652 (95%CI 0.637;0.667) and an accuracy of 0.851 (95%CI 0.844;0.858) for the P30 criterion. Sensitivity was 0.989 (95%CI 0.987;0.991) and specificity was 0.06 (95%CI 0.049; 0.07) at the 50% probability level that EKFC is within 30% of mGFR. A strategy of using the EKFC equation or mGFR according to the trustworthiness assessed by the random forest model yielded a global P30 of 86.1%, not significantly better than the EKFC P30 of 85.2%. Limitations: Observational study, creatinine as the only biomarker. No inclusion of Black participants. Conclusion: A strategy using machine learning model did not significantly improve GFR assessment compared to using the EKFC equation alone.