High-Precision Risk Stratification: Beyond Traditional Biomarkers
Cardiovascular disease (CVD) remains the leading cause of global morbidity and mortality, necessitating constant refinement of risk assessment tools. For decades, clinical practice has relied on traditional risk factors—age, sex, smoking status, blood pressure, and lipid profiles—to guide primary prevention. In Europe, the Systematic Coronary Risk Evaluation 2 (SCORE2) serves as the gold standard for predicting the 10-year risk of fatal and non-fatal cardiovascular events. However, a significant proportion of cardiovascular events occur in individuals classified as ‘low’ or ‘intermediate’ risk by these standard models. This residual risk suggests that current tools fail to capture the complex interplay of genetic predisposition and metabolic dysregulation.
A landmark study by Ritchie et al., recently published in the European Heart Journal, provides a compelling solution. By integrating nuclear magnetic resonance (NMR) metabolomics and polygenic risk scores (PRS) into the existing SCORE2 framework, researchers have demonstrated a substantial improvement in risk discrimination and stratification. This multi-omic approach not only identifies high-risk individuals more accurately but also offers a blueprint for how precision medicine can be implemented at a population scale.
The Search for the Missing Risk: Contextualizing SCORE2
While SCORE2 is highly effective for population-level screening, it is inherently limited by its reliance on a narrow set of clinical variables. It provides a snapshot of current clinical status but often misses the cumulative lifetime burden of genetic risk and the early metabolic shifts that precede overt clinical symptoms.
Polygenic Risk Scores (PRS) have emerged as a powerful tool to quantify this inherited susceptibility, reflecting the additive effect of thousands of small-effect genetic variants. Simultaneously, NMR metabolomics offers a high-throughput method to measure hundreds of circulating metabolites, including specific lipoprotein subclasses, fatty acids, and amino acids. These metabolites represent an ‘intermediate phenotype’—the functional output of both genetic and environmental influences. The central question addressed by Ritchie et al. was whether these advanced biomarkers provide independent, additive value to the current clinical standard.
Study Methodology: A Multi-Omic Approach in the UK Biobank
The researchers conducted a massive prospective analysis involving 297,463 participants from the UK Biobank. The cohort was specifically selected to represent a primary prevention population: individuals aged 40–69 years with no prior history of CVD, diabetes, or lipid-lowering treatment at baseline.
Three distinct scoring systems were evaluated in conjunction with SCORE2:
1. A panel of 11 clinical biomarkers, including cystatin C, HbA1c, and N-terminal pro-B-type natriuretic peptide (NT-proBNP).
2. NMR metabolomic biomarker scores, derived from a comprehensive metabolic fingerprint.
3. Polygenic Risk Scores (PRS), capturing the genomic architecture of coronary artery disease.
The primary endpoint was the first occurrence of fatal or non-fatal CVD over a 10-year follow-up period, during which 8,919 incident cases were recorded. The team utilized Harrell’s C-index to measure improvements in risk discrimination and categorical net reclassification index (NRI) to assess changes in risk stratification based on European Society of Cardiology (ESC) guideline thresholds.
Key Findings: Incremental Gains and Synergistic Power
The results underscore a clear hierarchy of predictive improvement. The baseline SCORE2 model achieved a respectable C-index of 0.719. However, the addition of any of the three components individually led to significant improvements:
Discrimination: The C-index Advantage
The addition of the 11 clinical biomarkers provided the largest individual boost, with a delta C-index of 0.014. NMR metabolomic scores and PRSs followed closely, with improvements of 0.010 and 0.009, respectively. Crucially, the effects were additive. When all three layers—clinical biomarkers, metabolomics, and PRS—were combined with SCORE2, the C-index improved by a total of 0.024 (95% CI: 0.022-0.027). While a 0.024 increase might appear modest in isolation, in the context of large-scale cardiovascular epidemiology, it represents a profound shift in the model’s ability to differentiate between future cases and non-cases.
Reclassification: Moving Patients into the Right Risk Tiers
From a clinical perspective, reclassification is often more meaningful than discrimination. The study found a net case reclassification of 16.66%. This means that nearly 17% of individuals who eventually suffered a cardiovascular event were correctly ‘upgraded’ to a higher risk category (and thus targeted for intervention) compared to the standard SCORE2 model. This precision prevents the clinical ‘blind spot’ where high-risk individuals are undertreated because their traditional risk factors do not yet exceed treatment thresholds.
Population Health Modeling: From Data to Lives Saved
Perhaps the most impactful aspect of the study is its population modeling. The researchers estimated the real-world impact of applying this integrated model to a hypothetical cohort of 100,000 screened individuals.
By using the combined model to target interventions, the number of CVD events prevented rose from 229 (using SCORE2 alone) to 413. This represents a nearly 80% increase in the efficiency of prevention (an additional 184 events prevented per 100,000). Remarkably, this gain in efficacy did not come at the cost of over-medicalization; the number of statins prescribed per CVD event prevented remained essentially stable. This suggests that the multi-omic approach is not just casting a wider net, but a much more accurate one.
Clinical Perspectives and Implementation Challenges
The biological plausibility of these findings is robust. The inclusion of NT-proBNP and cystatin C captures subclinical cardiac strain and renal impairment, while metabolomics captures the nuances of lipid metabolism beyond simple LDL-C measurements. The PRS provides the ‘baseline’ risk that remains constant throughout life.
However, implementation in routine clinical practice faces hurdles. While NMR metabolomics is becoming more cost-effective, it is not yet a standard laboratory offering in most primary care settings. Similarly, the integration of PRS requires standardized genetic testing and bioinformatics pipelines that are currently confined to specialized centers.
Expert commentators note that the ’11 clinical biomarkers’ panel used in the study already includes tests like HbA1c and NT-proBNP, which are widely available. Implementing this clinical panel alongside SCORE2 could serve as an immediate first step toward the ‘precision’ future described by the full multi-omic model.
Conclusion: The Future of Personalized Preventive Cardiology
The study by Ritchie et al. marks a significant milestone in the evolution of cardiovascular risk prediction. It demonstrates that the future of preventive cardiology lies in the synthesis of clinical, metabolic, and genetic data. By refining our ability to identify at-risk individuals, we can shift the paradigm from reactive treatment to proactive, precision-targeted prevention. If applied at scale, these findings suggest that we could nearly double the impact of our current CVD prevention efforts, saving thousands of lives without increasing the overall burden of medication. The challenge now lies in the hands of health policy experts and diagnostic innovators to bring these advanced tools from the research database to the clinic.
References
1. 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.
2. SCORE2 working group and ESC Cardiovascular Risk Collaboration. SCORE2 risk prediction algorithms: new models to estimate 10-year risk of cardiovascular disease in Europe. Eur Heart J. 2021;42(25):2439-2454.
3. Inouye M, Abraham G, Nelson CP, et al. Genomic Risk Prediction of Coronary Artery Disease in 480,000 Adults: Implications for Primary Prevention. J Am Coll Cardiol. 2018;72(16):1883-1893.
