Predicting left main stenosis in stable ischemic heart disease using logistic regression and boosted trees

Area of Science:

• Cardiology
• Medical Informatics
• Predictive Analytics

Background:

• The ISCHEMIA trial indicated similar outcomes for conservative versus invasive strategies in stable ischemic heart disease without left main stenosis.
• Left main coronary artery stenosis is a critical condition requiring accurate identification.

Purpose of the Study:

• To evaluate the feasibility of predicting significant left main coronary artery stenosis.
• To assess prediction models using extensive clinical, laboratory, and non-invasive test data.

Main Methods:

• A population-based cohort study included 150,423 adult patients undergoing stress testing before elective coronary angiography for stable ischemic heart disease.
• Predictors included demographics, comorbidities, lab tests, and stress test data; the outcome was left main coronary artery stenosis (≥50%).
• Logistic regression and boosted trees machine learning algorithms were employed to build prediction models.

Main Results:

• 6.1% of patients had left main stenosis.
• The logistic regression model achieved an optimism-adjusted c-statistic of 0.72 with adequate calibration.
• The boosted trees algorithm demonstrated similar accuracy, with a c-statistic of 0.72 and adequate calibration.

Conclusions:

• Predicting left main coronary artery disease in stable ischemic heart disease patients using extensive clinical data yielded only modest results.
• Both traditional logistic regression and machine learning (boosted trees) showed similar, limited predictive capabilities.
• Further research is needed to improve prediction models for left main coronary artery stenosis.