Predictive models for abdominal aortic aneurysms using polygenic scores and PheWAS-derived risk factors

Area of Science:

• Cardiovascular Genetics
• Medical Informatics
• Epidemiology

Background:

• Abdominal aortic aneurysms (AAA) are life-threatening aortic enlargements.
• Current screening relies on limited risk factors (age, sex, smoking history).
• Genome-wide association studies have identified genetic risk loci for AAA.

Purpose of the Study:

• To develop and compare predictive models for AAA risk.
• To integrate genetic risk, clinical diagnoses, and known risk factors.
• To evaluate model performance against existing screening recommendations.

Main Methods:

• Developed polygenic risk scores (PRS) using large-scale genetic data.
• Identified candidate diagnoses from electronic health records (EHR) using phenome-wide association studies.
• Combined PRS, phecodes, and covariates in predictive models.
• Validated models in diverse cohorts (White and Black participants).

Main Results:

• The full predictive model achieved an Area Under the Curve (AUC) of 0.883.
• Models significantly outperformed USPSTF screening criteria (AUCs 0.613-0.632).
• Models demonstrated good calibration and substantial net reclassification improvement.

Conclusions:

• PRS strongly associate with AAA risk and enhance predictive model performance.
• Developed models substantially improve AAA risk identification compared to current guidelines.
• Models show potential applicability across diverse populations, including minorities.