JACC: Advances (Feb 2025)

Simplified Outcome Prediction in Patients Undergoing Transcatheter Tricuspid Valve Intervention by Survival Tree-Based Modelling

  • Vera Fortmeier, MD,
  • Mark Lachmann, MD,
  • Lukas Stolz, MD,
  • Jennifer von Stein, MD,
  • Karl-Philipp Rommel, MD,
  • Mohammad Kassar, MD,
  • Muhammed Gerçek, MD,
  • Anne R. Schöber, MD,
  • Thomas J. Stocker, MD,
  • Hazem Omran, MD,
  • Michelle Fett,
  • Jule Tervooren,
  • Maria I. Körber, MD,
  • Amelie Hesse,
  • Gerhard Harmsen, PhD,
  • Kai Peter Friedrichs, MD,
  • Shinsuke Yuasa, MD, PhD,
  • Tanja K. Rudolph, MD,
  • Michael Joner, MD,
  • Roman Pfister, MD,
  • Stephan Baldus, MD,
  • Karl-Ludwig Laugwitz, MD,
  • Stephan Windecker, MD,
  • Fabien Praz, MD,
  • Philipp Lurz, MD,
  • Jörg Hausleiter, MD,
  • Volker Rudolph, MD

Journal volume & issue
Vol. 4, no. 2
p. 101575

Abstract

Read online

Background: Patients with severe tricuspid regurgitation (TR) typically present with heterogeneity in the extent of cardiac dysfunction and extra-cardiac comorbidities, which play a decisive role for survival after transcatheter tricuspid valve intervention (TTVI). Objectives: This aim of this study was to create a survival tree-based model to determine the cardiac and extra-cardiac features associated with 2-year survival after TTVI. Methods: The study included 918 patients (derivation set, n = 631; validation set, n = 287) undergoing TTVI for severe TR. Supervised machine learning-derived survival tree-based modelling was applied to preprocedural clinical, laboratory, echocardiographic, and hemodynamic data. Results: Following univariate regression analysis to pre-select candidate variables for 2-year mortality prediction, a survival tree-based model was constructed using 4 key parameters. Three distinct cluster-related risk categories were identified, which differed significantly in survival after TTVI. Patients from the low-risk category (n = 261) were defined by mean pulmonary artery pressure ≤28 mm Hg and N-terminal pro–B-type natriuretic peptide ≤2,728 pg/mL, and they exhibited a 2-year survival rate of 85.5%. Patients from the high-risk category (n = 190) were defined by mean pulmonary artery pressure >28 mm Hg, right atrial area >32.5 cm2, and estimated glomerular filtration rate ≤51 mL/min, and they showed a significantly worse 2-year survival of only 52.6% (HR for 2-year mortality: 4.3, P < 0.001). Net re-classification improvement analysis demonstrated that this model was comparable to the TRI-Score and outperformed the EuroScore II in identifying high-risk patients. The prognostic value of risk phenotypes was confirmed by external validation. Conclusions: This simple survival tree-based model effectively stratifies patients with severe TR into distinct risk categories, demonstrating significant differences in 2-year survival after TTVI.

Keywords