Genetic links between atopy, allergy, and alopecia areata: insights from a Mendelian randomization study

Our study adopted a two-sample MR approach to investigate the causal effects of atopic and allergic conditions—including hay fever, eczema, asthma, and allergies to cats, dust, and pollen—on AA incidence. We used single nucleotide polymorphisms (SNPs) from GWAS summary data as instrumental variables. These SNPs, reflecting the random allocation of genetics, help address confounders like age and sex [13]. Our MR analysis was grounded on three assumptions: the genetic instruments strongly predict the exposure, are free from confounders, and influence the outcome solely through the exposure [14].

We sourced atopy and allergy GWAS data from the Social Science Genetic Association Consortium (SSGAC) [15], including data from the UK Biobank and 23andMe but excluding FinnGen. We opted for multi-trait GWAS given its superior predictive power over its single-trait counterpart [16]. We focused on six phenotypes: hay fever, eczema, asthma, and allergies to cats, dust mites, and pollen. The effective sample sizes ranged from 254,386 to 1,214,626, with all participants being of European descent. Diagnosis criteria were based on ICD-10 or self-reported surveys [15] (Table 1).

AA data came from the latest FinnGen database release (DF9—as of May 11, 2023) [17], encompassing 682 AA patients and 361,140 controls of European ancestry (Table 1). Diagnoses adhered to ICD revisions 8 through 10. We noted an average patient age of 42.12 years, with a slight age variance between genders. To mitigate population heterogeneity bias, we only used SNPs from European ancestry GWAS data.

Our SNP selection involved multiple rigorous steps. Starting with genome-wide significant SNPs (p < 5 × 10^–8), we excluded those in linkage disequilibrium (R² < 0.001, window size = 10,000 kb) [18] or additionally, weak SNPs (F-statistic < 10) [19]. Post-harmonization, we further eliminated palindromic and ambiguous SNPs (those with a minor allele frequency (MAF) > 0.01) [20]. The MR-Pleiotropy Residual Sum and Outlier (MR-PRESSO) analysis [21], helped re-evaluate and exclude SNPs potentially influenced by pleiotropy, ensuring a robust set of instrumental variables for each atopic phenotype.

Addressing pleiotropy concerns, we utilized five MR methods: inverse variance weighted (IVW), MR-Egger regression, weighted median, maximum likelihood ratio, and weighted mode. And the IVW method, our primary analysis technique. Each method offers unique strengths and mitigates specific biases, ensuring a comprehensive examination of the causal effects [22, 23].

To ensure the validity of our primary results, we performed sensitivity analyses including heterogeneity tests (Cochran’s Q tests and MR-PRESSO global tests) and horizontal pleiotropy tests (MR-Egger intercept tests and leave-one-out analyses). These analyses helped identify sensitivity and the influence of individual SNPs on the causal estimate, enhancing the reliability of our findings [24].

Statistical significance was set at p < 0.05. We expressed MR estimates as odds ratios (ORs) with 95% confidence intervals (CIs), indicating the relative risk of AA associated with each atopic phenotype. Analyses were conducted using ‘Two Sample MR’ (version 0.5.7) and ‘MRPRESSO’ (version 1.0) packages in R (version 4.3.1), ensuring rigorous assessment of the MR estimates and identification of pleiotropic outliers [25].

The IVW method, our primary analysis technique, indicated a significant increase in AA risk associated with hay fever (OR = 1.88, 95% CI 1.10–3.20; p = 0.02), eczema (OR = 6.16, 95% CI 2.09–18.13; p < 0.001), asthma (OR = 2.04, 95% CI 1.12–3.73; p = 0.02), pollen allergy (OR = 1.39, 95% CI 1.16–1.68; p < 0.001), dust allergy (OR = 1.25, 95% CI 1.04–1.49; p = 0.02), and cat allergy (OR = 1.24, 95% CI 1.03–1.50; p = 0.02) (Fig. 1). The corresponding scatter plot was shown in Fig. 2, and all analysis methods showed consistent directionality. These associations were corroborated by the maximum likelihood (ML) model, emphasizing the notable link between these conditions and AA. While the weighted median approach found no significant link for hay fever, it did confirm the connection with the other conditions, aligning with the general trend observed in the IVW model. The MR-Egger and weighted mode methods further identified these six conditions as risk factors for AA, although their statistical significance was not as pronounced.

The MR-Egger intercept did not indicate the presence of horizontal pleiotropy (p > 0.05 for all exposures), suggesting that our results are free from bias related to pleiotropy. Despite the detection of heterogeneity in eczema exposure (p < 0.05), we applied a random effects IVW model to mitigate this issue (Table 2). Despite the detection of heterogeneity in eczema exposure, the random effects IVW was utilized to perform the MR analysis and balance this heterogeneity. The absence of horizontal pleiotropy bias in this context was reaffirmed by leave-one-out analysis, and forest plots (Additional file 1: Figs. S1–S6) confirmed that no single SNP disproportionately influenced our results, further attesting to the reliability of our findings.