Genetic instruments selection
Single nucleotide polymorphisms (SNPs) that were genome-wide significant (P < 5 × 10-8) were extracted from each GWAS of sex hormones. A stringent linkage disequilibrium (LD) clumping strategy was then applied, using a window of 10 Mb and r2 threshold of less than 0.001. We used the 1000 Genomes Project European samples as the LD reference in PLINK 1.9. After LD clumping, statistically independent SNPs were selected as genetic instruments for sex hormones.
If genetic instruments were not available in the summary data of COVID-19 outcomes, we used a proxy variant that was identified in 1000 Genomes Project European samples using LDlink (https://ldlink.nci.nih.gov/?tab=ldproxy). We harmonized the SNP-exposure and SNP-outcome summary datasets using a previously described method 22. During the data harmonization process, SNPs with a minor allele frequency > 0.49 were excluded. The number of genetic variants after LD clumping and IVs for each sex hormone are listed in eTable 1 and eTable 2 .