Genome-wide association analysis and replication in 810,625 individuals with varicose veins

Abstract

Varicose veins affect one-third of Western society, with a significant subset of patients developing venous ulceration, costing $14.9 billion annually in the USA. Current management consists of either compression stockings, or surgical ablation for more advanced disease. Most varicose veins patients report a positive family history, and heritability is ~17%. We describe the largest two-stage genome-wide association study of varicose veins in 401,656 individuals from UK Biobank, and replication in 408,969 individuals from 23andMe (total 135,514 cases and 675,111 controls). Forty-nine signals at 46 susceptibility loci were discovered. We map 237 genes to these loci, several of which are biologically plausible and tractable to therapeutic targeting. Pathway analysis identified enrichment in extracellular matrix biology, inflammation, (lymph)angiogenesis, vascular smooth muscle cell migration, and apoptosis. Using a polygenic risk score (PRS) derived in an independent cohort, we demonstrate its predictive utility and correlation with varicose veins surgery.

Fig. 1. Study design and analysis workflow.

A discovery GWAS was performed in the UK Biobank cohort, with the top independent lead variants tested within the 23andMe replication cohort. Forty-nine independent variants at 46 loci met the Bonferroni-corrected threshold in the replication cohort, and were interrogated further in subsequent analyses.

Fig. 2. Results of genome-wide association study in varicose veins.

Manhattan plot showing genome-wide P-values plotted against position on each of the autosomes. The dark blue, light blue and green dots refer to the discovery UK Biobank Cohort, with the red dots corresponding to the 49 variants from the 23andMe cohort at each replicated locus (shown in Supplementary Data 2). The dark blue peaks correspond to the 46 loci that replicated in the 23andMe cohort at a Bonferroni-corrected threshold of P < 4.72 × 10⁻⁴. Candidate genes at each locus are named above each signal, with previously unreported genetic loci in blue, and previously described loci in black.

Fig. 3. Logistic regression for varicose vein case/control status in UK Biobank against VV-polygenic risk score (VV-PRS) decile.

The VV-PRS was derived in an independent study sample, FinnGen (n = 17,027 VV cases and 190,028 controls). Odds ratio refers to odds ratio of VV in UK Biobank versus decile 1 of VV-PRS (lowest). Error bars signify 95% confidence intervals of the odds ratios.