No evidence for large causal effects of six air pollutants on non-suppurative otitis media: a two-sample Mendelian randomization study

Background: Observational studies have linked air pollution to otitis media, but confounding and reverse causation preclude causal inference. We used two-sample Mendelian randomization (MR) to assess genetic causal effects of six air pollutants on non-suppurative otitis media, represented by otitis media with effusion (OME). Methods: Exposure GWAS for PM2.5, PM2.5 absorbance, PM2.5-10, PM10, NO2 and NOx were obtained from UK Biobank (MRC-IEU), and the outcome was FinnGen R13 H8_NONSUPPNAS (12,397 cases, 464,237 controls). Inverse-variance weighting was primary, with MR-Egger, weighted median and mode, MR-PRESSO, leave-one-out and Steiger analyses, reverse MR and exploratory multivariable MR (MVMR). Asthma, allergic rhinitis and chronic rhinosinusitis served as positive controls, and three suppurative otitis media phenotypes served as negative controls. Results: After false-discovery-rate correction, none of the six pollutants was significantly associated with OME; the sole nominal signal was PM10 (OR=1.555, 95% CI 0.947-2.553, P=0.081), which reversed with the 88-SNP pooled set (OR=0.64, 0.24-1.69, P=0.365). At 80% power, minimum detectable ORs ranged widely from 1.69 to 32.09; only PM2.5-10 and PM10 had values near 2, so moderate effects (OR 1.2-1.5) cannot be excluded. Confirmatory positive controls using phenotype-specific instruments were strongly positive, whereas negative-control outcomes showed no associations (all FDR>0.85). Every nominal reverse-MR signal had a significant MR-Egger intercept (the PM2.5 absorbance signal also showed significant heterogeneity and MR-PRESSO tests), indicating horizontal pleiotropy rather than reverse causation. MVMR was exploratory only, as conditional F statistics were below 10. Conclusions: We found no evidence for large-effect genetic causal associations between the six air pollutants and OME. These null findings argue against large, lifetime-average causal effects, and the associations reported in observational studies may instead reflect residual confounding, shared comorbidity or differences in exposure windows.