Genetically Proxied Circulating C-Reactive Protein and Selected Cytokines in Relation to Register-Defined Sensorineural Hearing Loss: A Two-Sample Mendelian Randomization Study

Abstract

Background: Circulating inflammatory biomarkers are associated with sensorineural hearing loss (SNHL), but observational studies cannot establish causality.

Objective: To evaluate whether genetically proxied C-reactive protein (CRP) and eight cytokines are associated with SNHL.

Methods: We used European-ancestry GWASs for CRP (N = 575,531) and eight cytokines (maximum N = 74,783), with FinnGen R13 outcomes of 48,388 SNHL cases and 432,132 controls; sudden idiopathic hearing loss (4,082 cases) was secondary. Variants were harmonized and pruned using 1000 Genomes Finnish linkage disequilibrium (r² < 0.001 within 10 Mb). Random-effects inverse-variance weighting or a Wald ratio was primary. Sensitivity analyses included MR-Egger, weighted median and mode, MR-RAPS, RadialMR, Steiger directionality, cis-only estimates, source-assay filters, and false-discovery-rate correction.

Results: Of 301 unique exposure rsIDs, 279 were retrieved and 276 exposure rows passed harmonization. Finnish LD pruning left 238 CRP instruments. CRP was compatible with the null (OR per unit increase in ln[CRP mg/L], 1.005; 95% CI, 0.948–1.065; p = 0.868), consistent with weighted median, MR-Egger, weighted mode, and MR-RAPS estimates. No cytokine survived FDR correction (all q ≥ 0.393), source-assay filtering yielded no supported association, and secondary-outcome estimates were null. Steiger analyses favored the exposure-to-outcome direction.

Conclusion: Results support no material effect of lifelong genetically proxied circulating CRP on register-defined SNHL. Cytokine evidence remains less definitive because instrument sets were small, predominantly trans acting, heterogeneous, and assay sensitive. The findings neither exclude acute or cochlea-specific inflammation nor evaluate corticosteroid effectiveness. Independent replication and cis-pQTL colocalization are required before therapeutic inference.