Differential Associations of Seven Circulating Complement Proteins with Heart Failure: A Single-Center Cross-Sectional Study

The complement system is associated with heart failure (HF), but patterns across circulating complement proteins and their sensitivity to covariate adjustment remain unclear.

In this cross-sectional study, 392 participants (164 chronic HF: 128 HFpEF, 14 HFmrEF, 22 HFrEF; 228 non-HF controls) were enrolled, and seven circulating complement proteins were measured. Spearman correlations with NT-proBNP and LVEF, sequential multivariable logistic regression, and restricted cubic splines were assessed for each protein. Benjamini-Hochberg false discovery rate (FDR) correction was applied to fully adjusted models. Subgroup and restricted-sample sensitivity analyses were performed. Exploratory quasi-Bayesian statistical decomposition assessed whether the C8-HF and CD55-HF associations overlapped with CRP, eGFR, and hemoglobin.

C7 and CD55 showed the strongest positive correlations with NT-proBNP (ρ≈0.32) and weak negative correlations with LVEF. After full adjustment, odds ratios for prevalent HF per 1-SD increase were 1.39 (95% CI 1.06-1.84) for C1r, 1.72 (1.29-2.29) for C7, and 1.47 (1.13-1.92) for C9; FDR-adjusted P values were 0.045, 0.0017, and 0.014. C1s and CD46 were not significant; C8 and CD55 associations attenuated after adjustment. No non-linear associations were found. C1r, C7, and C9 remained significant when restricting to eGFR ≥60 mL/min/1.73 m² or CRP ≤10 mg/L, but attenuated to non-significance when HF cases were restricted to LVEF ≥50%. Exploratory statistical decomposition suggested that the CD55-HF association partly overlapped with eGFR and hemoglobin, but not CRP; no corresponding overlap was observed for C8.

In this cross-sectional sample, C1r, C7, and C9 were associated with prevalent chronic HF after full adjustment and FDR correction, but estimates attenuated in the LVEF-preserved restricted sample. C8 and CD55 associations attenuated after adjustment, while C1s and CD46 were not significant. These associations require prospective validation and should not be interpreted as causal.