Ludwig, SebastianSebastianLudwigCoisne, AugustinAugustinCoisneHamzi, KenzaKenzaHamziBen Ali, WalidWalidBen AliScotti, AndreaAndreaScottiKoell, BenediktBenediktKoellDuncan, AlisonAlisonDuncanMakkar, RajRajMakkarAkodad, MariamaMariamaAkodadBleiziffer, SabineSabineBleizifferNickenig, GeorgGeorgNickenigKaneko, TsuyoshiTsuyoshiKanekoRuge, HendrikHendrikRugeAdam, MattiMattiAdamSondergaard, LarsLarsSondergaardDahle, GryGryDahleTaramasso, MaurizioMaurizioTaramassoWalther, ThomasThomasWaltherKempfert, JoergJoergKempfertObadia, Jean-FrançoisJean-FrançoisObadiaChehab, OmarOmarChehabTang, Gilbert H LGilbert H LTangGoel, SachinSachinGoelFam, NeilNeilFamDenti, PaoloPaoloDentiPraz, FabienFabienPrazvon Bardeleben, Ralph StephanRalph Stephanvon BardelebenHausleiter, JörgJörgHausleiterLatib, AzeemAzeemLatibConradi, LenardLenardConradiModine, ThomasThomasModinePezel, ThéoThéoPezelGranada, Juan FJuan FGranada2025-05-302025-05-302025-05-07https://boris-portal.unibe.ch/handle/20.500.12422/210834Aims Although several treatment options are available for patients with severe mitral regurgitation (MR), a significant proportion of patients remain ineligible for any mitral valve (MV) intervention. We aimed to analyze the phenotypic characteristics of surgical high-risk patients ineligible for MV interventions using an unsupervised phenotypic clustering approach.Methods And Results Between 2014 and 2022, the CHOICE-MI registry included 984 patients with MR undergoing screening for transcatheter mitral valve replacement at 33 international sites. For this study, only patients with screening failure receiving medical therapy alone were included. Patients receiving transcatheter or surgical treatment were excluded. A cluster analysis using K-means was performed on baseline clinical, demographic, and imaging variables to identify different patient phenotypes. Among 284 patients with MR (77.4±8.82 years, 56.0% female, EuroSCORE II: 6.6±5.8%) considered ineligible for any MV intervention, two clinically distinct phenogroups (PG) were identified using unsupervised hierarchical clustering of principal components. PG1: elderly women with primary MR, preserved left ventricular function, and annular calcification; and PG2: patients with secondary MR, advanced heart failure, and high prevalence of comorbidities. One-year all-cause mortality did not differ between the phenogroups (PG1: 21.4%, PG2: 23.4%, p=0.89). Predictors of mortality were albumin, renal function, extracardiac arteriopathy for PG1, and albumin, coronary artery disease, and prior myocardial infarction for PG2.Conclusions This study identified two major subgroups among patients ineligible for mitral interventions showing profound differences in clinical and anatomical profiles. Identifying these factors may drive technological evolution to address the unmet clinical need for therapeutic options in MR patients.Clinicaltrials.gov Identifier NCT04688190 (CHOICE-MI Registry).enclusteringmedical therapymitral regurgitationnon-supervised machine learningtranscatheter mitral valve replacement600 - Technology::610 - Medicine & healtharticle10.48620/883244032982610.1093/ehjci/jeaf141