Abstract 9844: Dynamic Heart Failure Risk Trajectories in Patients with Fontan Circulation: From Population Data to Individual Risk
Abstract 9844: Dynamische Herzinsuffizienz-Risikotrajektorien bei Patient:innen mit Fontan-Zirkulation: von Populationsdaten zu individuellem Risiko
Circulation · 02.11.2026 · 154 (Suppl_1)
DOI: 10.1161/circ.154.suppl_1.9844 ↗Auf einen Blick
In 456 Fontan-Patient:innen sagte ein Ensemblemodell das Risiko einer Herzinsuffizienz-Hospitalisierung über die Lebensspanne voraus. Die Modellgüte war moderat.
Fragestellung
Kann ein dynamisches Risikomodell das individuelle Risiko für Herzinsuffizienz-Hospitalisierung nach Fontan kontinuierlich aktualisieren?
Methode
Daten aus der Quebec Congenital Heart Disease Database; 456 Fontan-Patient:innen; 43 Variablen; GRU und GBM, kalibriert und als Ensemble kombiniert; unabhängiger Testsatz.
Ergebnisse
63 Patient:innen (13,8 %) hatten eine Herzinsuffizienz-Hospitalisierung. Das Ensemble erreichte eine AUC von 0,782 und eine AUPRC von 0,254 bei einer Basis-AUPRC von 0,09.
Schlussfolgerung der Autor:innen
Das deep-learning-gestützte Ensemble bietet einen Rahmen zur Beurteilung des sich wandelnden HFH-Risikos nach Fontan und ermöglicht personalisierte Überwachung.
Kernaussagen
- 63 Patient:innen (13,8 %) hatten eine HFH
- AUC 0,782; AUPRC 0,254
- Wichtige Prädiktoren: Alter bei Fontan, Glenn-Prozedur, Vorhofflimmern-Last
Deutsche Zusammenfassung: KI-gestützt aus dem Original-Abstract erstellt; alle Zahlen wurden automatisch mit dem Original abgeglichen. Maßgeblich ist der Original-Abstract.
Kurz gefasst
Fachartikel in der Zeitschrift Circulation, erschienen am 02.11.2026. Verfasst von Harry Moroz und 7 weiteren Autor:innen.
Themen: Herzinsuffizienz
Abstract (Originalsprache)
Background
Although the Fontan procedure has enabled survival into adulthood for individuals born with univentricular hearts, the resulting circulation, characterized by elevated venous pressure and limited cardiac output, predisposes patients to progressive multi-organ dysfunction and hemodynamic failure. Heart failure hospitalisation (HFH) is a primary sentinel event of clinical decompensation. Current guidelines recommend lifelong specialist surveillance but lack validated tools to identify high-risk patients. This study aims to develop a dynamic risk stratification tool that updates individual HFH probability as clinical events accumulate over a patient's lifetime.
Methods
We utilized the Quebec Congenital Heart Disease Database, which includes 456 Fontan completion patients with up to 35 years of follow-up (median: 10.2 years). We constructed annual landmark feature vectors consisting of 43 variables, comprising static pre-Fontan characteristics and time-varying post-Fontan clinical events. We trained two models in parallel, a Gated Recurrent Unit (GRU) with self-attention to process sequential data, and a Gradient Boosting Machine (GBM) for tabular interactions. Models were calibrated using isotonic regression and integrated into a weighted ensemble to stabilize and maximize the predictive performance, measured using area under the precision-recall curve (AUPRC). By predicting HFH risk at each annual landmark, we generated lifespan HFH risk trajectories. Using an independent test sample, the evaluated model performance, identified the most important features and timeframes for prediction, and assessed the clinical application of our framework for individual patient risk assessment.
Results
HFH was experienced by 63 (13.8%) patients. The ensemble model achieved an area under the receiver operating characteristic curve of 0.782 and an AUPRC of 0.254 (baseline AUPRC of 0.09). Our model successfully identified divergent lifespan risk trajectories between patients with and without HFH (Figure 1A). The model identified key clinical predictors, including age at Fontan, pre-Fontan Glenn procedure, and atrial fibrillation burden (Figure 1B). High-risk vulnerability time windows were identified at 3-5 years and 10-11 years post-Fontan (Figure 1C). Clinical utility was demonstrated by generating individualized, projected risk trajectories, illustrating how acute clinical events (e.g., atrial fibrillation onset) trigger divergence from baseline risk (Figure 2).
Conclusion
Our deep learning-driven ensemble model provides a framework for assessing the evolving risk of HFH post-Fontan procedure. By integrating longitudinal clinical data, this approach enables personalized surveillance and the potential for timely, preventative interventions to shift patients toward lower-risk trajectories.
Zitieren
APA
Moroz, H., Liu, A., Ordonez, M. V., Li, C., Guo, L., Therrien, J., Li, Y., & Marelli, A. J. (2026). Abstract 9844: Dynamic Heart Failure Risk Trajectories in Patients with Fontan Circulation: From Population Data to Individual Risk. Circulation, 154(Suppl_1). https://doi.org/10.1161/circ.154.suppl_1.9844
Vancouver
Moroz H, Liu A, Ordonez MV, Li C, Guo L, Therrien J, et al. Abstract 9844: Dynamic Heart Failure Risk Trajectories in Patients with Fontan Circulation: From Population Data to Individual Risk. Circulation. 2026;154(Suppl_1). doi:10.1161/circ.154.suppl_1.9844
BibTeX
@article{Moroz2026,
author = {Moroz, Harry and Liu, Aihua and Ordonez, Maria Victoria and Li, Chao and Guo, Liming and Therrien, Judith and Li, Yue and Marelli, Ariane J},
title = {Abstract 9844: Dynamic Heart Failure Risk Trajectories in Patients with Fontan Circulation: From Population Data to Individual Risk},
journal = {Circulation},
year = {2026},
volume = {154},
number = {Suppl_1},
doi = {10.1161/circ.154.suppl_1.9844}
}
EUVIMED