892 Plasma Proteins Linked to Heart Disease Risk in 53,000-Person Study

Cardiovascular disease remains the leading cause of death globally, yet comprehensive, longitudinal analyses linking the full plasma proteome to diverse CVD subtypes have been lacking. This study addresses that gap by systematically mapping plasma protein associations across 13 incident CVD outcomes in one of the largest proteomics cohorts ever assembled.

Using data from 53,026 UK Biobank participants with a median 14-year follow-up, researchers measured 2,920 baseline plasma proteins and applied Cox proportional hazard models to identify significant associations after stringent Bonferroni correction. They found 3,089 significant protein-CVD associations involving 892 unique protein analytes. The most striking associations included NT-proBNP for atrial fibrillation (P = 6.31×10⁻³¹³), NPPB and GDF15 for heart failure, and LEP and FABP4 as the strongest correlates of cardiac structure and function measured by cardiovascular magnetic resonance (CMR) imaging.

A machine-learning prediction model trained on 257 pre-selected proteins significantly outperformed the standard SCORE2 cardiovascular risk tool across most outcomes, achieving an AUC of 0.86 for abdominal aortic aneurysm. Adding protein data to SCORE2 further improved discrimination, underscoring the additive predictive value of the plasma proteome beyond conventional clinical risk factors.

Two-sample Mendelian randomization analysis, using genetic variants as instrumental variables to minimize confounding, identified 225 proteins causally linked to CVDs. LPA showed the strongest causal association with coronary artery disease (OR = 1.13, P = 2.38×10⁻¹⁵). Critically, many of these causal proteins are already targets of existing drugs, opening concrete drug repurposing opportunities. Mediation analyses revealed broad-spectrum mediators — IGFBP4 and GDF15 each influencing 9 cardiovascular outcomes — with modifiable risk factors such as smoking and BMI serving as the primary mediating pathways between proteins and CVD risk.

The study provides an unprecedented systems-level map of how the plasma proteome interacts with cardiovascular health across disease subtypes, cardiac structure, and function. By combining longitudinal epidemiology, causal inference, and prediction modeling, it offers a foundation for developing protein-based biomarker panels, refining early CVD detection, and identifying novel therapeutic targets.