Kidney function-dependent plasma proteomic architecture of incident myocardial infarction: a proteome-wide association study of UK Biobank participants

Abstract

Background: Individuals across the spectrum of kidney function experience excess risk of acute myocardial infarction (AMI), but whether the plasma proteome's association with this risk is itself modified by the estimated glomerular filtration rate (eGFR) remains uncharacterized. Methods: Cox proteome-wide association studies of 2,923 Olink proteins in 33,446 UK Biobank participants (fully adjusted model 2: 29,417 complete cases, 1,584 incident AMI events; the full eligible cohort comprised 1,848 events, the complete-case reduction reflecting full covariate adjustment). After fully adjusted screening, protein × eGFR interaction models tested whether each protein association was modified by eGFR (continuous). Incremental prediction was evaluated with a leakage-free elastic-net Cox model. Results: Over a median 13.9-year follow-up, 1,848 AMI events occurred. After multivariable adjustment, 975 proteins were linked to AMI (false discovery rate [FDR]-adjusted p < 0.05). For 176 of these, the AMI association was modified by kidney function (interaction FDR < 0.05)-strengthening as eGFR declined for 154 proteins and at higher eGFR for a smaller set of 22. Exploratory GSEA identified FDR-significant innate immune enrichment in the higher-eGFR direction, whereas lower-eGFR cardiac and matrix signals were nominal. Adding plasma proteins to clinical predictors modestly improved 10-year AMI prediction (Harrell C, 0.748 → 0.780; ΔC = 0.032, 95% confidence interval = 0.021-0.043). Conclusion: Kidney function modifies the plasma proteomic signature of incident AMI: most protein-AMI associations intensify as eGFR declines, while a smaller subset strengthens at preserved filtration. This supports kidney function-aware proteomic cardiovascular risk interpretation.