Abstract P279: Proteomic Profiling Identifies Mortality-Associated Markers in Coronary Artery Disease

Abstract

Background: Coronary artery disease (CAD) is a leading global cause of mortality. Proteomics has emerged as a promising tool for understanding CAD prognosis. We explored proteomic profiles in CAD patients to identify protein markers which predict mortality. Methods: The discovery cohort included 2,072 participants from the UK Biobank (UKB) with proteomics data and prevalent CAD at enrollment, defined by the ICD-10 and OPCS-4 codes from the in-patient Hospital Episode Statistics data. All-cause mortality was ascertained based on the Death Register records. Plasma proteomic profiling of 1,463 unique proteins at enrollment was conducted using the Olink platform. The validation cohort included 91 participants from the Mental Stress Ischemia Prognosis Study (MIPS), a prospective study of CAD patients enrolled from Emory University hospitals in whom plasma proteomic profiling of 6,159 proteins was measured by the SomaScan Assay. We analyzed 1,203 common proteins in both cohorts with missing rate < 20%. In the UKB, we examined the proteome-wide associations between the rank-based inverse normal transformed protein levels and mortality, using Cox proportional hazards models adjusting for race, sex, age, body mass index, diabetes, hypertension, heart attack history, frequent drinking (≥ 3 times /week), smoking history, high- and low-density lipoprotein, triglycerides, C-reactive protein and estimated glomerular filtration rate. Mortality-associated proteins at Bonferroni corrected p < 0.05 were analyzed for pathway enrichment. LASSO regression was used to further select the best performing proteins in predicting mortality. Prediction performance of the proteins was evaluated using the C statistic. The mortality-associated proteins were validated in MIPS, using Cox models adjusting for race, sex, and age. Results: The UKB cohort was 69% male, 96% White, with mean age of 62 years, and there were 646 (31%) deaths during median follow-up of 13.3 years. The MIPS cohort was 79% male, 24% Black, with mean age of 64 years, and 11 (12%) died during median follow-up of 6.3 years. A total of 274 proteins were significantly associated with mortality in the UKB, representing enriched pathways such as inflammatory response. Seven proteins, including GDF15 , TNFRSF10B , PLAUR , SPON1 , CXCL17 , ANGPT2 , and ICAM5 were identified by LASSO model as key predictors. Adding these proteins to the model with traditional risk factors improved the prediction of mortality by 5.3% (75.8% vs 70.5%, 95% CI 3.7% - 6.9%). All associations except for CXCL17 were validated in the MIPS. Conclusion: The findings suggest the potential added value of plasma proteins for predicting mortality in CAD patients beyond traditional factors. The functions of mortality-associated proteins indicate underlying biological mechanisms, such as inflammation and cell growth regulation, linked to CAD progression.