Integrating restricted polygenic risk scores with SCORE2 enhances cardiovascular risk prediction and reclassification in UK Biobank
T Adane, J van Meurs, A Uitterlinden, J van RooijAbstract
Background/Introduction
Cardiovascular disease (CVD) remains a leading cause of morbidity and mortality worldwide, and accurate risk prediction is essential for prevention strategies. Clinical models such as SCORE2 do not include genetic risk, even though a polygenic risk score (PRS) captures substantial inherited susceptibility. A recent study in UK Biobank demonstrated that integrating genome-wide PRS with SCORE2 improves prediction, but these scores are computationally intensive and difficult to interpret. Whether a simplified and resource efficient genetic approach can provide similar benefit remains uncertain.
Purpose
We propose a restricted PRS based on genome-wide significant variants as a resource-efficient alternative and evaluated whether this approach enhances SCORE2 performance for coronary artery disease (CAD) risk prediction.
Methods
We included 303,556 UK Biobank participants of European ancestry aged 40–69 years without prior cardiovascular disease or diabetes. SCORE2 was calculated using guideline-based predictors. A restricted coronary artery disease (CAD) polygenic risk score (PRS) was derived from 223 genome-wide significant variants and categorized into deciles. Odds ratios (ORs) for each PRS decile relative to the middle two deciles were used to define a PRS factor or risk conveyed through the PRS. The PRS was then integrated with SCORE2 using a multiplicative approach (SCORE2 × PRS factor) to estimate combined risk. We compared two models: SCORE2 with an without the PRS. Model performance was evaluated using the concordance index (C-index) and net reclassification improvement (NRI), overall and stratified by age groups (<50 years and ≥50 years), reflecting SCORE2’s age-specific thresholds.
Results
SCORE2 and the integrated model (SCORE2 × PRS factor) were strongly associated with incident CAD. Integration of PRS improved overall C-index from 0.70 to 0.73, with gains from 0.73 to 0.75 in participants younger than 50 years and from 0.67 to 0.69 in those aged 50 years or older. Net reclassification improved by 10.9% overall, driven mainly by case reclassification, with improvements of 8.6% and 10.7% in participants younger than 50 and aged 50 years or older, respectively.
Conclusion
Integrating a restricted PRS with SCORE2 improves cardiovascular risk prediction and reclassification in UK Biobank participants, offering a resource-efficient alternative to genome-wide scores. Further validation in independent cohorts and across diverse populations is warranted.