Key result
Digital twin-AI framework identifies HFpEF phenotype showing greater NT-proBNP reduction with accelerated pacing.
Why the study?
Patients with HFpEF exhibit marked heterogeneity in response to pacing, and because diastolic relaxation is energy dependent and impaired myocardial energetics are common, favorable response to accelerated atrial pacing may depend on preserved or improved cardiac efficiency in addition to hemodynamic unloading.
Does a physiology-informed digital twin-AI framework identify hemodynamic and cardiac efficiency response phenotypes that predict clinical benefit (MLHFQ quality of life and NT-proBNP improvement) from accelerated atrial pacing in patients with HFpEF?
Does a physiology-informed digital twin-AI framework identify hemodynamic and cardiac efficiency response phenotypes that predict clinical benefit (MLHFQ quality of life and NT-proBNP improvement) from accelerated atrial pacing in patients with HFpEF?
A physiology-informed digital twin-AI framework bridges mechanistic hemodynamic/energetic modeling and routine clinical data to identify HFpEF patients most likely to experience NT-proBNP and quality-of-life improvements from accelerated atrial pacing.
Answers are grounded in this paper and cite their sources.
May identify HFpEF pacing responders via digital twins; hypothesis-generating and requires prospective validation before practice change.
Background Heart failure with preserved ejection fraction (HFpEF) exhibits marked heterogeneity in response to pacing. Because diastolic relaxation is energy dependent and impaired myocardial energetics are common in HFpEF, we hypothesized that favorable response to accelerated atrial pacing would be defined not only by hemodynamic unloading but also by preserved or improved cardiac efficiency (CE). Methods Patient-specific digital twins of 146 HFpEF patients were used to train a variational autoencoder that generated a virtual population (n ≈ 25,000), in which simulations quantified pacing-induced changes in left atrial pressure (LAP), systolic blood pressure (SBP), cardiac output (CO), and CE, defining four response phenotypes. Classifiers predicting these phenotypes from baseline clinical variables were trained in the virtual cohort and applied to myPACE, where treatment × predicted-response interactions with trial outcomes were evaluated in exploratory analyses. Results Across the virtual population, most patients exhibited reduced LAP (95.6%) and increased CO (93.8%), whereas only 36.1% improved CE. In myPACE, the predicted SBP-reduction phenotype showed a significant treatment × predicted-response interaction for NT-proBNP (p = 0.015). In the pacing arm, the predicted SBP-reduction and CE-improving phenotypes were associated with greater quality-of-life improvement, without a significant treatment interaction. Conclusions A physiology-informed digital twin-AI framework identified a predicted SBP-reduction phenotype associated with differential NT-proBNP response to accelerated pacing in HFpEF. Simulated CE provides a candidate mechanistic explanation for this signal. When routine clinical data are insufficient to identify digital twins, AI trained on twin simulations offers a potentially generalizable route for bringing mechanistic models to such data.
No takes yet. Share an insight, caveat, or question.
Gu et al. (2026) studied Heart failure with preserved ejection fraction (HFpEF) (n=194). Physiology-informed digital twin-AI framework predicting accelerated atrial pacing response phenotypes vs. Predicted non-responder phenotypes or standard backup pacing rate of 60 bpm in myPACE was evaluated on Treatment x predicted-response interaction for NT-proBNP and association with quality of life (MLHFQ) (p=0.015). A physiology-informed digital twin-AI framework identified a predicted SBP-reduction phenotype with a significant treatment-by-response interaction for NT-proBNP reduction in HFpEF pacing.
Synapse has enriched 10 related papers. Consider them for comparative context: