Integrating Metabolomics and Polygenics with SCORE2: A New Frontier in Cardiovascular Risk Stratification

Highlight

Synergistic Prediction Power

Adding 11 clinical biomarkers, NMR metabolomics, and polygenic risk scores (PRS) to the SCORE2 model improved the C-index by 0.024, representing a significant leap in discriminatory power.

Superior Reclassification

Using the combined multi-omic approach resulted in a net case reclassification of 16.66%, effectively identifying high-risk individuals who would have been missed by standard clinical models.

Public Health Impact

Population modeling indicates that at-scale application could increase the number of cardiovascular events prevented from 229 to 413 per 100,000 individuals screened.

Background: The Evolution of Cardiovascular Risk Prediction

For decades, cardiovascular disease (CVD) risk assessment has relied on traditional clinical variables—age, sex, blood pressure, and smoking status—codified in tools like the SCORE2 (Systematic Coronary Risk Evaluation 2) model. While SCORE2 is effective at the population level, it often lacks the granularity needed for personalized medicine, particularly for individuals in intermediate-risk categories where treatment decisions, such as statin initiation, remain ambiguous.

In recent years, two emerging fields have promised to refine these assessments: nuclear magnetic resonance (NMR) metabolomics and polygenic risk scores (PRS). NMR metabolomics provides a snapshot of an individual’s metabolic state, including detailed lipoprotein subfractions and small-molecule metabolites. Conversely, PRS offers a window into the stable, lifelong genetic susceptibility to CVD. Despite their individual promise, their incremental value when integrated into current ESC-recommended frameworks has remained a critical research gap. The study by Ritchie et al., published in the European Heart Journal, addresses this by evaluating whether a multi-omic approach can truly shift the needle in clinical practice.

Study Design and Methodology

This prospective study utilized data from 297,463 participants in the UK Biobank, aged 40 to 69 years. To ensure the results were relevant to primary prevention, the researchers excluded individuals with pre-existing CVD, diabetes, or those already receiving lipid-lowering therapy.

The researchers constructed and compared several predictive models:

1. SCORE2:

The baseline clinical model.

2. Clinical Biomarkers:

A panel of 11 markers, including HbA1c, cystatin C, and C-reactive protein.

3. NMR Metabolomic Scores:

Derived from high-throughput metabolic profiling.

4. Polygenic Risk Scores (PRS):

Quantifying genetic predisposition based on genome-wide association studies.

The primary endpoint was the 10-year risk of fatal and non-fatal CVD (myocardial infarction and stroke). The study assessed performance using Harrell’s C-index for discrimination and categorical Net Reclassification Index (NRI) based on the European Society of Cardiology (ESC) risk thresholds (low, moderate, and high risk).

Key Results: Synergy of Biomarkers

The results underscored the limitations of relying on any single data stream. The baseline SCORE2 model achieved a C-index of 0.719. When individual components were added, discrimination improved incrementally:

• Addition of 11 clinical biomarkers: ΔC-index 0.014
• Addition of NMR metabolomic scores: ΔC-index 0.010
• Addition of PRS: ΔC-index 0.009

However, the most striking finding occurred when all three modalities were combined with SCORE2. This integrated approach yielded a total ΔC-index of 0.024 (95% CI: 0.022-0.027). While these decimal increments may seem small in statistical terms, they translate to profound shifts in clinical classification.

Net Reclassification and Precision Stratification

Using the ESC guideline risk thresholds, the combined model demonstrated a net case reclassification of 16.66%. This means that approximately one in six individuals who eventually suffered a CVD event were correctly moved into a higher risk category where preventative interventions (like statins) would be recommended. This improvement in sensitivity was achieved without significantly increasing the over-prescription of medications to low-risk individuals.

Clinical Impact and Population Modeling

To translate these findings into a public health context, the researchers applied population modeling to estimate the real-world impact of screening 100,000 individuals.

Under current SCORE2-based screening, it is estimated that 229 CVD events are prevented. By incorporating clinical, metabolomic, and polygenic data for targeted risk reclassification, this number nearly doubles to 413 events prevented per 100,000 people screened. Importantly, the number of statins prescribed per CVD event prevented remained essentially stable, suggesting that the multi-omic approach increases the efficiency of healthcare resources rather than simply expanding the treated population.

Expert Commentary: Mechanistic Insights and Implementation

The biological plausibility of these findings lies in the distinct types of information each score provides. PRS captures the inherent, unchangeable genetic risk that is present from birth. Clinical biomarkers and metabolomics, however, capture the dynamic interplay between genetics, lifestyle, and environment. NMR metabolomics, in particular, identifies subtle shifts in lipid metabolism and systemic inflammation that traditional cholesterol panels might miss.

Implementation Challenges

Despite the clear statistical benefits, several hurdles remain for clinical implementation:

• Cost-Effectiveness: While the costs of genomic sequencing and NMR profiling are falling, the infrastructure for routine testing across primary care is not yet universal.
• Complexity: Integrating multi-omic data into electronic health records in a way that is interpretable for general practitioners requires sophisticated bioinformatics support.
• Generalizability: The UK Biobank is predominantly of European ancestry, and further validation is required in more diverse global populations to ensure equitable benefits.

Conclusion

The study by Ritchie et al. provides robust evidence that the future of cardiovascular risk prediction lies in the integration of multi-omic data. By moving beyond traditional risk factors to include clinical biomarkers, NMR metabolomics, and polygenic risk, clinicians can identify high-risk individuals with far greater precision. If applied at scale, this approach could significantly reduce the global burden of cardiovascular disease, moving us closer to the goal of truly personalized preventive cardiology.

References

Ritchie SC, Jiang X, Pennells L, et al. Combined clinical, metabolomic, and polygenic scores for cardiovascular risk prediction. Eur Heart J. 2025 Dec 15:ehaf947. doi: 10.1093/eurheartj/ehaf947.