Large-scale proteomics to identify novel biomarkers linking social determinants of health to heart failure risk

Figures

• 

  Fig. 1.. Correlation of SDOH measures and derivation and prognostic relevance of SDOH factor at Visit 3. SDOH measures were coded as higher being worse. (A) Spearman correlation matrix among SDOH variables; (B) principal axis (PA) exploratory factor loadings; (C) association of factor score with risk of incident HF through (year) in bivariate and multivariable Cox proportional hazard models adjusted for age, sex, BMI, eGFR, smoking status, CAD, AF, DM, and hypertension.

• 

  Fig. 2.. Associations of proteins with SDOH factor and risk of incident HF among Black and white participants. (A) Volcano plots of the beta coefficients for cross-sectional associations of proteins with SDOH (left plots) and hazard ratios for associations of SDOH-associated proteins with incident HF (right plots) among Black and white participants. (B) Venn diagram showing overlap between SDOH- and HF-associated proteins by race group. (C) Forest plot showing effect estimates for associations with SDOH factor and incident HF for the 12 shared SDOH- and HF-associated proteins between race groups.

• 

  Fig. 3.. Mediation by candidate proteins of the SDOH factor association with incident HF stratified by race group. Forest plot shows estimated average causal mediation effect (ACME) with associated 95% confidence intervals. Shading indicates statistical significance of mediation effect after Bonferroni correction.

• 

  Fig. 4.. Results of proteogenomic analyses to assess for potential causal associations of proteins with cardiac outcomes. (A) MR. Top: MR P values for causal effect of protein levels on HF and cardiac structure and function. Hatched line indicates nominal significance (P < 0.05), red line indicates significance after Bonferroni multiple testing correction for number of pQTLs tested. Red pQTLs, nonsignificant; blue, nominally significant; green, significant after multiple testing correction. Bottom: Forest plots for significant MR instrumental variables of direction and magnitude of effect of genetically higher protein levels on each outcome. Orange, trans pQTL; brown, multi-pQTL instrumental variable (IV) comprising both cis and trans SNPs. (B) Results of colocalization analyses for significant MR associations based on single-SNP trans instruments.

• 

  Fig. 5.. Associations of UKB proteins with SDOH factor score and HF. (A) Forest plot showing effect estimates for associations with SDOH factor for the 10 available candidate proteins. (B) Forest plot showing effect estimates for associations between the available 10 candidate proteins and incident HF. (C) Forest plot demonstrating mediation by candidate proteins of the SDOH factor association with incident HF. CI, confidence interval.

• 

  Fig. 6.. Study design. (A) Study design of the analysis in ARIC. (B) Study design of the replication analysis in UKB.