Abstract Body (Do not enter title and authors here): Introduction: Abdominal aortic aneurysm (AAA) is a significant cause of morbidity and mortality in older adults. The AAA screening guidelines from the Society of Vascular Surgery include risk factors such as sex, age, smoking, and family history. This study explored whether integration of a polygenic risk score (PRS) and proteomics with clinical data could improve AAA prediction in the ARIC Study. Methods: Over a median follow-up of 24 years from ARIC visit 2 (1990-92) baseline, we identified 487 clinical AAA cases among 9,373 ARIC participants (7,397 Whites and 1,976 Blacks) through hospital discharge diagnoses or death certificates. We selected AAA-associated clinical risk factors based on literature and our expertise, including age, gender, race, field center, smoking status, smoking pack-years, waist girth, BMI, levels of total and HDL cholesterols, hypertension, diabetes, and eGFR. We calculated the PRS[WT1] based on SNP dosage in ARIC and the latest genome-wide association study for AAA, which reported 141 independent associations from 14 discovery cohorts (PMID: 37845353). ARIC used SOMAscan v4 to measure 4,955 plasma proteins at baseline, of which 24 were significantly associated with clinical AAA (p < 1x10^-5) independent of the clinical risk factors. The prediction equation for AAA risk was constructed in 3 Cox regression models: 1) clinical risk factors measured at baseline, 2) model 1 variables plus PRS, and 3) model 2 variables plus the 24 AAA-associated proteins identified through proteomics analysis. We used the area under the curve (AUC) to evaluate the prediction performance of these models for AAA risk.
Results: Participants in the top quintile of PRS showed significantly higher AAA risk compared to the lowest quintile (HR 1.41, 95% CI: 1.03 – 1.85) after adjustment for clinical risk factors. Adding the PRS to clinical risk factors did not improve the AUC: 0.890 (95% CI: 0.869 - 0.945) in model 1 vs 0.891 (95% CI: 0.853 - 0.961) in model 2. Further adding the 24 AAA-associated proteins substantially improved the performance of the prediction model [AL2] (AUC=0.950, 95% CI: 0.939 - 0.986 in model 3), with 289 AAA events in the top quintile of predicted risk compared to 12 in the lowest quintile. Conclusion: A proteomics-integrated approach that combined clinical risk factors and proteomics data enhanced AAA risk prediction and has the potential to improve risk stratification and early intervention for AAA.