Development and Cross-Site Validation of ML Models for ANOCA Diagnosis and Prognosis

08 Apr 20265 min read0 viewsJournal Feed

GIST

Introduction Angina with no obstructive coronary artery disease (ANOCA) affects millions and is frequently under-recognised because diagnostic pathways and risk tools predominantly target obstructive coronary artery disease (CAD). This protocol describes shared methods for two machine-learning (ML) studies: (1) differentiating ANOCA from stable angina with obstructive CAD and (2) predicting long-term mortality among patients with ANOCA and obstructive CAD.

Methods and analysis We will develop and cross-site validate ML classification models using a multicentre retrospective cohort drawn from the Alberta Provincial Project for Outcome Assessment in Coronary Heart Disease registry and institutional datasets from the University of Ottawa Heart Institute and the University Health Network. Eligible participants are adults (≥18 years) undergoing initial cardiac catheterisation for chest pain/anginal equivalents since 1995, excluding prior revascularisation, major structural heart disease and predefined non-anginal indications.

Outcomes are (1) ANOCA (0% to Model development will use nested cross-validation with stratified k-fold inner-loop tuning and leave-one-site-out cross-validation for repeated external validation.

Clinical Editorial

Summary

Machine-learning aims for ANOCA differentiation and prognosis: study design and implications

Context and objective

The first aims to distinguish angina with no obstructive coronary arteries (ANOCA) from stable angina with obstructive CAD; the second seeks to forecast long-term mortality among patients with ANOCA and obstructive CAD.

The setting spans Alberta databases and institutional datasets from the University of Ottawa Heart Institute and the University Health Network.

The protocol outlines two multicentre ML studies addressing stable angina with and without obstructive CAD.

Population and data sources

Eligible adults (18 years or older) undergoing initial cardiac catheterisation for chest pain or anginal equivalents since 1995 are included.
Exclusions apply to prior revascularisation, major structural heart disease, and predefined non-anginal indications.
Data sources combine provincial registry data with institutional records, with a plan for harmonization across sites.

Study design and analytic approach

The study employs nested cross-validation with a stratified k-fold inner loop for tuning and leave-one-site-out cross-validation to support external validation.
Candidate predictors will be harmonised across sites, screened for missingness, and refined using expert-driven DAG-guided selection in combination with Boruta and LASSO methods.
Preprocessing steps include appropriate encoding, multivariate imputation by chained equations for missing data, and feature scaling.

Modeling and validation specifics

Algorithms under consideration comprise elastic-net logistic regression, random forest, LightGBM, and multilayer perceptron networks.
Hyperparameters will be optimized through Bayesian optimization.
Model performance will be reported with threshold tuning, and explainability will be addressed via SHAP (SHapley Additive exPlanations) analyses.
A focus on subgroup fairness is noted, with explainability and fairness appraisals integrated into the evaluation.

Outcomes and deployment

Primary outcomes are ANOCA status for the diagnostic model and long-term mortality for the prognostic model.
Final models are intended to be deployed as a web-based clinical risk calculator.

Ethics, limitations, and dissemination

Ethical approvals have been obtained from the University of Calgary and the University Health Network; analyses will use deidentified data within secure environments, with results reported in aggregate form.