Abstract A130: Comparative Performance of Metabolomic and Genomic Risk Scores in Predicting Incident Ischemic Stroke

Background: Blood biomarkers (including metabolomics) and genetics have identified risk factors for incident stroke but are not frequently compared together. Methods: We conducted a nested case-control study within the REGARDS cohort (986 ischemic stroke cases, 881 controls) among participants aged 45+ enrolled 2003-2007. Participants with a history of stroke/TIA or hemorrhagic stroke at enrollment baseline were excluded. A polygenic risk score (PRS) for stroke was derived from the GIGASTROKE consortium data using PLINK software. A metabolomic risk score (MRS) was developed from metabolomics data using LASSO regression in a randomly selected training dataset (n=942). We then compared MRS and PRS performance in a testing dataset (n=925) using Cox regression and examined stroke-free survival across four risk groups defined by high/low combinations of both scores using Kaplan-Meier analysis. Results: The training (n=942) and testing (n=925) sets were well balanced, with similar age (68.5 vs. 68.8 years), gender (49.9% vs. 49.2% female), race (59.7% vs. 62% Caucasian), and incident stroke rates (51.3% vs. 54.2%), with no significant differences between groups. The MRS was derived from 17 metabolites which included Guanosine, C3 malonylcarnitine, Gutamate, and Pseudourine. In a model adjusted for age and gender, the MRS had the strongest association with incident stroke (HR = 1.52, 95% CI 1.27–1.81, p < 0.0001) compared with the PRS (HR = 1.35, 95% CI 1.17–1.55, p < 0.0001), and this association remained robust after adjustment for Framingham stroke risk factors. Predictive performance improved from PRS alone (AUC = 0.57) to PRS + MRS (AUC = 0.62), reaching optimal discrimination when the PRS, MRS, and Framingham risk factors were combined (AUC = 0.65). There was a stepwise relationship, with stroke incidence the lowest in the low-risk (MRS low/PRS low) group and highest in the combined high-risk (MRS high/PRS high) group. Conclusion: A metabolomic risk score had independent predictive value for incident ischemic stroke when compared to a polygenic risk score, and models that combined it with Framingham risk factors had the best predictive performance. Patients with high metabolomic risk scores and high polygenic risk scores have the highest risk of developing incident stroke.