Machine Learning Improves Cardiovascular Risk Definition for Young, Asymptomatic Individuals.

Area of Science:

• Cardiology
• Artificial Intelligence in Medicine
• Preventive Medicine

Background:

• Clinical guidelines recommend imaging for subclinical atherosclerosis (SA) in intermediate-risk individuals.
• Standard risk tools may not adequately identify all at-risk young, asymptomatic individuals.
• Early detection of SA is crucial for cardiovascular disease (CVD) prevention.

Purpose of the Study:

• To develop a machine-learning model (EN-PESA) for predicting SA presence and extent in young, asymptomatic individuals.
• To utilize routine, quantitative, and easily measured variables for SA risk estimation.
• To refine CVD risk assessment and optimize the use of imaging for SA detection.

Main Methods:

• An Elastic Net (EN) model was developed to predict SA extent using coronary artery calcification and vascular ultrasound metrics.
• Model performance was compared against traditional CVD risk scores.
• Validation was performed using an independent external cohort.

Main Results:

• The EN-PESA model achieved a c-statistic of 0.88 for predicting generalized SA.
• EN-PESA identified a significantly larger proportion of individuals with intermediate to high cardiovascular risk compared to atherosclerotic CVD and SCORE.
• 86.8% of individuals flagged by EN-PESA showed baseline SA or significant 3-year progression.

Conclusions:

• The EN-PESA model effectively identifies young, asymptomatic individuals at increased risk of CVD due to SA.
• The model integrates age, blood pressure, routine lab tests, and dietary data.
• Individuals identified by EN-PESA may benefit from further imaging or pharmacological treatment for CVD prevention.