Exploring Genetic Associations Between Autoimmune Diseases and Pathological Scars: A Bidirectional Mendelian Randomization Analysis
Xueqi Wang, Xianglin DongABSTRACT
Background
Autoimmune diseases and pathological scars share inflammatory and fibroproliferative features, but observational evidence cannot establish whether their co‐occurrence reflects causation or shared biology. We used bidirectional Mendelian randomization (MR) to test genetic associations between clinically defined autoimmune diseases and pathological scars.
Methods
We analyzed 10 autoimmune diseases and 2 pathological scar phenotypes (keloids and hypertrophic scars) in both directions, comprising 40 primary IVW tests. Autoimmune‐disease instruments were selected at p < 5 × 10 −8 and scar instruments at p < 5 × 10 −6 because few genome‐wide significant scar loci were available. Independent SNPs were obtained by LD clumping ( r 2 < 0.001; 10 000 kb), and variants with F ≤ 10 were excluded. Multiplicative random‐effects inverse‐variance weighted (IVW) analysis was primary, with MR‐Egger, weighted median, mode‐based estimators, Cochran's Q , MR‐Egger intercept, leave‐one‐out analysis, MR‐PRESSO, and radial MR as sensitivity analyses. Benjamini–Hochberg false discovery rate (FDR) correction was applied across the 40 primary tests.
Results
Three associations met nominal significance based on raw p values: hypertrophic scars with celiac disease (OR = 1.099, 95% CI 1.036–1.165, p = 0.00157; 2 SNPs), rheumatoid arthritis with hypertrophic scars (OR = 1.119, 95% CI 1.026–1.220, p = 0.0107; 19 SNPs), and type 1 diabetes with keloids (OR = 1.212, 95% CI 1.048–1.401, p = 0.0090; 12 SNPs). None survived FDR correction (minimum FDR‐adjusted p = 0.0628). The remaining 37 primary tests had raw p ≥ 0.05.
Conclusion
The analyses identified only nominal associations that were not robust to multiple‐testing correction and therefore do not provide reliable evidence of causality. The results offer an initial framework for investigating potential shared genetic signals and should be treated as hypothesis‐generating pending independent validation.