Potential Added Clinical Value of Polygenic Risk Scores in Risk Assessment of Atherosclerotic Cardiovascular Disease
Author(s)
Dina Hassen, MPP1, Alicia M. Johns, PhD1, Michelle Pistner Nixon, PhD1, Alex S. F. Berry, PhD1, Laney K. Jones, PharmD, MPH1, Matthew T. Oetjens, PhD1, Yasser Khalil, MD1, Sam S. Gidding, MD1, Andrea Hattenberger, MS1, Ciaran Fisher, BS1, Cara Mccormick, MPH1, Jinyi Zhu, PhD2, Marc S. Williams, MD1, David L. Veenstra, PharmD, PhD3, Josh F. Peterson, MD, MPH2, Jing Hao, PhD, MD, MS, MPH1;
1Geisinger, Danville, PA, USA, 2Vanderbilt University Medical Center, Nashville, TN, USA, 3University of Washington, Seattle, WA, USA
1Geisinger, Danville, PA, USA, 2Vanderbilt University Medical Center, Nashville, TN, USA, 3University of Washington, Seattle, WA, USA
OBJECTIVES: Patients at risk for atherosclerotic cardiovascular disease (ASCVD) are currently identified using statin benefit groups (clinical ASCVD, severe hypercholesterolemia, diabetes, ASCVD risk based on the pooled cohort equation [PCE]) and familial hypercholesterolemia (FH) status. Our objective was to explore the potential added clinical value of polygenic risk scores (PRS) in risk assessment for ASCVD.
METHODS: We performed retrospective analysis using patients’ electronic health record, and genomic data (PRS and pathogenic/likely pathogenic variants in FH-associated genes) from Geisinger’s MyCode Initiative. The observation period was from January 1, 2005-December 31, 2022 with December 31, 2012 as baseline. A CAD-PRS was adopted from the literature to calculate PRS scores.
RESULTS: A total of 102,507 participants (mean [SD] age: 51.1 [15.7] years; 37,902 male [37%]) were included. At baseline, 30,217 (29.5%) participants were among the four statin benefit groups, and 381 (0.4%) were FH variant carriers. Only one-third of identified participants had PRS scores in the top 5th, 10th or 25th percentile -- common thresholds for increased ASCVD risk-- indicating that about two-thirds of those at increased risk by applying PRS are otherwise not identified. Among a sub-cohort of participants without ASCVD, severe hypercholesterolemia or diabetes and between 40-75 years old at baseline (n=22,681), 5,743 (25.32%) developed ASCVD and 1,182 (5.21%) died during the 10-year follow-up period. Hazard ratios predicting incident ASCVD were 1.23 (95%CI, 1.22, 1.24) (p<0.001) for both PCE and PCE+PRS. Harrell’s concordance statistic of PCE alone was 0.653 vs. 0.645 for PCE+PRS. The correlation between PCE and PRS was -0.045.
CONCLUSIONS: Our preliminary results indicate that adding PRS to existing PCE risk assessment has minimal improvements in discrimination metrics; however, PRS could identify additional patients at increased risk. Further analysis will assess the reclassification of adding PRS to PCE and the effects of PRS in stratified populations (e.g., by age and gender).
METHODS: We performed retrospective analysis using patients’ electronic health record, and genomic data (PRS and pathogenic/likely pathogenic variants in FH-associated genes) from Geisinger’s MyCode Initiative. The observation period was from January 1, 2005-December 31, 2022 with December 31, 2012 as baseline. A CAD-PRS was adopted from the literature to calculate PRS scores.
RESULTS: A total of 102,507 participants (mean [SD] age: 51.1 [15.7] years; 37,902 male [37%]) were included. At baseline, 30,217 (29.5%) participants were among the four statin benefit groups, and 381 (0.4%) were FH variant carriers. Only one-third of identified participants had PRS scores in the top 5th, 10th or 25th percentile -- common thresholds for increased ASCVD risk-- indicating that about two-thirds of those at increased risk by applying PRS are otherwise not identified. Among a sub-cohort of participants without ASCVD, severe hypercholesterolemia or diabetes and between 40-75 years old at baseline (n=22,681), 5,743 (25.32%) developed ASCVD and 1,182 (5.21%) died during the 10-year follow-up period. Hazard ratios predicting incident ASCVD were 1.23 (95%CI, 1.22, 1.24) (p<0.001) for both PCE and PCE+PRS. Harrell’s concordance statistic of PCE alone was 0.653 vs. 0.645 for PCE+PRS. The correlation between PCE and PRS was -0.045.
CONCLUSIONS: Our preliminary results indicate that adding PRS to existing PCE risk assessment has minimal improvements in discrimination metrics; however, PRS could identify additional patients at increased risk. Further analysis will assess the reclassification of adding PRS to PCE and the effects of PRS in stratified populations (e.g., by age and gender).
Conference/Value in Health Info
2025-05, ISPOR 2025, Montréal, Quebec, CA
Value in Health, Volume 28, Issue S1
Code
HSD58
Topic
Health Service Delivery & Process of Care
Disease
SDC: Cardiovascular Disorders (including MI, Stroke, Circulatory), STA: Generics