Machine learning-based prediction of clinical outcomes and left ventricular ejection fraction improvement after percutaneous coronary intervention in patients with reduced ejection fraction

Q: What is the clinical evidence from this study?

Study design: Cohort. Population: Patients with left ventricular dysfunction (baseline LVEF <45%) undergoing percutaneous coronary intervention (PCI) (n=2793). Intervention: CatBoost machine learning prediction model (top 10% highest predicted risk decile) vs. Remaining 90% of predicted risk cases. Primary outcome: 1-year composite of major adverse cardiovascular events (MACE: death, myocardial infarction, target vessel revascularization, or coronary artery bypass grafting) (OR 10.74, 95% CI 0.73-0.83, p=7.4 × 10−11).

Q: What does this research mean for the field?

Machine learning models trained on routinely available clinical and procedural data predict 1-year major adverse cardiovascular events (AUC 0.78) and LVEF improvement following percutaneous coronary intervention in patients with reduced ejection fraction. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

Q: What question did this study set out to answer?

To develop and evaluate machine learning models predicting 1-year adverse clinical outcomes and left ventricular ejection fraction improvement after percutaneous coronary intervention in patients with reduced ejection fraction.

Lioz Steinbuch; Edward Itelman; Yuval Altman; Bar Yacobi; Osnat Goren-Peyser; Tamir Bental; Yael Shostak; Pablo Codner; Chen Gurevitz; Amos Levi; Guy Witberg; Yeela Talmor‐Barkan; Ran Kornowski; Leor Perl

doi:10.1093/ehjdh/ztag156

October 19, 2026European Heart Journal - Digital HealthOpen Access

Machine learning-based prediction of clinical outcomes and left ventricular ejection fraction improvement after percutaneous coronary intervention in patients with reduced ejection fraction

View Full Paper

Ask AI

Bookmark

Share

Key result

CatBoost machine learning models predicted 1-year MACE after PCI in patients with LVEF <45% with an AUC of 0.78 (95% CI 0.73–0.83) and 83.7% vs 32.4% event rates in the top decile (OR 10.74).

Why the study?

The expected clinical and functional trajectory after PCI in patients with left ventricular dysfunction remains uncertain. Substantial heterogeneity in both clinical outcomes and LVEF improvement underscores the need for improved post-PCI prognostic risk stratification in this high-risk population.

Can machine learning models using routinely available pre- and intra-procedural data accurately predict 1-year major adverse cardiovascular events and post-procedural LVEF improvement in patients with baseline LVEF <45% undergoing PCI?

Key Points

To develop and evaluate machine learning models predicting 1-year adverse clinical outcomes and left ventricular ejection fraction improvement after percutaneous coronary intervention in patients with reduced ejection fraction.
Retrospective single-centre study of 3336 procedures from 2793 patients with baseline LVEF <45% undergoing PCI between 2004 and 2025.
Developed two CatBoost models utilizing 53 pre- and intra-procedural predictors, trained on a development set of 2669 procedures and validated on a held-out test set of 667 procedures.
Evaluated discrimination for a 1-year primary composite endpoint (death, myocardial infarction, target vessel revascularization, or CABG) and a secondary functional endpoint (absolute LVEF increase ≥10 percentage points).
For the primary composite endpoint (n = 2152; event rate 37.5%), the model attained an AUC of 0.78 (95% CI: 0.73–0.83) and a PR-AUC of 0.70 (95% CI: 0.63–0.77), with event rates of 83.7% in the highest-risk decile compared to 9.3% in the lowest-risk decile.
For LVEF improvement (n = 2551; 34.7% improved), the model achieved an AUC of 0.73, with outcome enrichment reaching 76.9% in the highest-probability decile.

Study Design

Type

Cohort (n=2,793)

Multicenter

No

Structured PICO

Can machine learning models using routinely available pre- and intra-procedural data accurately predict 1-year major adverse cardiovascular events and post-procedural LVEF improvement in patients with baseline LVEF <45% undergoing PCI?

P

Population

2,793 adults (3,336 procedures; median age 69 years, 20.0% female) with baseline LVEF <45% (median 35%) undergoing PCI at a single centre in Israel, followed for 1 year.

E

Exposure

Two CatBoost gradient boosting machine learning models using 53 routinely available pre- and intra-procedural demographic, clinical, laboratory, echocardiographic, and angiographic predictors evaluated at completion of the index PCI

C

Comparator

Held-out test set stratification comparing highest-risk/highest-probability prediction subsets (top 5%, 10%, 20%, 30%) and lowest-risk prediction subsets (bottom 5%, 10%, 20%, 30%) against remaining cases

O

Outcome

1-year composite of major adverse cardiovascular events (MACE), defined as death, myocardial infarction (MI), target vessel revascularization, or coronary artery bypass grafting (CABG) within 365 days after the index PCI procedurecomposite

Main Result

Odds Ratio: 10.74

Absolute Event Rate: 83.7% vs 32.4%

p-value: p=7.4 × 10−11

Limitations

Retrospective single-centre study lacking external validation
Study period spanned over two decades (2004-2025) without explicit evaluation of temporal effects from evolving clinical practice and therapy
Observational design precludes causal inference regarding the effect of PCI on clinical events or LVEF improvement
Timing of baseline and follow-up echocardiographic assessments was not standardized, and intermediate LVEF changes (>0% and <10%) were excluded
Detailed lesion-level anatomy, post-procedural coronary flow, natriuretic peptides (BNP/NT-proBNP), and cardiac magnetic resonance viability imaging were not available
Retrospective single-centre design requiring external and prospective validation before clinical adoption
Study period spanned more than two decades (2004–2025) without explicit evaluation of temporal effects from evolving clinical practice, device technology, and medical therapy
Observational design precludes causal inference regarding the benefit of PCI versus conservative management
Longitudinal post-procedural medical therapy patterns, dose intensity, and adherence were not explicitly modelled
Non-standardized timing of baseline and follow-up echocardiography and exclusion of intermediate LVEF changes (>0 and <10%) may introduce selection bias
Lack of detailed lesion-level anatomy, post-procedural coronary flow, natriuretic peptide levels (BNP/NT-proBNP), and cardiac MRI viability data

Ask about this paper

Answers are grounded in this paper and cite their sources.

Implication

CatBoost machine learning models using routine clinical and procedural data effectively stratify 1-year MACE risk (AUC 0.78) and likelihood of ≥10% LVEF improvement (AUC 0.73) in patients with LVEF <45% undergoing PCI.

Abstract

Aims The expected clinical and functional trajectory after percutaneous coronary intervention (PCI) in patients with left ventricular dysfunction remains uncertain. Recent trials have highlighted substantial heterogeneity in both clinical outcomes and left ventricular ejection fraction (LVEF) improvement, underscoring the need for improved post-PCI prognostic risk stratification in this high-risk population. This study aimed to develop machine learning models predicting clinical outcomes and LVEF improvement after PCI in patients with reduced LVEF. Methods and results This retrospective single-centre study included adults undergoing PCI between 2004 and 2025 with baseline LVEF <45%. Two CatBoost models were developed using 53 routinely available pre- and intra-procedural predictors. The primary endpoint was a 1-year composite of death, myocardial infarction, target vessel revascularization, or coronary artery bypass grafting. The secondary endpoint was an absolute LVEF increase of ≥10 percentage points. Performance was evaluated on a held-out test set using area under the receiver operating characteristic curve (AUC), area under the precision–recall curve (PR-AUC), and enrichment analysis. The dataset included 3336 procedures from 2793 patients (median LVEF 35%). For the primary endpoint (n = 2152; event rate 37.5%), the model achieved an AUC of 0.78 (95% CI: 0.73–0.83) and a PR-AUC of 0.70 (95% CI: 0.63–0.77), with event rates of 83.7% in the highest-risk decile and 9.3% in the lowest-risk decile. For LVEF improvement (n = 2551; 34.7% improved), performance was moderate (AUC 0.73), with enrichment to 76.9% in the highest-probability decile. Conclusion Machine learning models using routinely available patient and procedural data enabled meaningful prognostic stratification in patients with left ventricular dysfunction undergoing PCI, supporting individualized estimation of clinical and functional trajectories after revascularization.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Steinbuch et al. (2026) conducted a cohort in Patients with left ventricular dysfunction (baseline LVEF <45%) undergoing percutaneous coronary intervention (PCI) (n=2,793). CatBoost machine learning prediction model (top 10% highest predicted risk decile) vs. Remaining 90% of predicted risk cases was evaluated on 1-year composite of major adverse cardiovascular events (MACE: death, myocardial infarction, target vessel revascularization, or coronary artery bypass grafting) (OR 10.74, 95% CI 0.73-0.83, p=7.4 × 10−11). CatBoost machine learning models predicted 1-year MACE after PCI in patients with LVEF <45% with an AUC of 0.78 (95% CI 0.73–0.83) and 83.7% vs 32.4% event rates in the top decile (OR 10.74).

synapsesocial.com/papers/6aee6bb371a7594806220a96 https://doi.org/10.1093/ehjdh/ztag156

View Full Paper

Ask AI

Bookmark

Share

Machine learning-based prediction of clinical outcomes and left ventricular ejection fraction improvement after percutaneous coronary intervention in patients with reduced ejection fraction

Key Points

Study Design

Structured PICO

Main Result

Limitations

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

Machine learning-based prediction of clinical outcomes and left ventricular ejection fraction improvement after percutaneous coronary intervention in patients with reduced ejection fraction

Key Points

Study Design

Structured PICO

Main Result

Limitations

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

Keep reading

Keep reading