3 RESULTS
Key characteristics of each GWAS data used in the study are presented inTable 1 . MR analysis showed that higher genetically predicted serum estradiol levels were suggestively associated with a decreased risk of SARS-CoV-2 infection in women (odds ratio (OR), 0.88 for log odds of estradiol detection; 95% CI, 0.78, 0.98; P = 0.021; OR 0.66 per standard deviation (SD) increased estradiol; 95% CI, 0.47 to 0.94; P = 0.021) (Figure 1, panel A ). However, little evidence was found to support a causal effect of genetically predicted serum estradiol levels on COVID-19 hospitalization and critical illness using the binary exposure (i.e., estradiol levels were detected or not). In contrast, suggestive inverse causal associations of continuous estradiol levels with COVID-19 hospitalization (OR 0.50 per SD increased estradiol; 95% CI, 0.25 to 0.98; P = 0.044) and COVID-19 critical illness (OR 0.69 per SD increased estradiol; 95% CI, 0.48 to 0.99; P = 0.043) in men were observed (Figure 1, panels B and C ).
A causal effect of genetic predisposition to higher levels of serum BT on SARS-CoV-2 infection was observed in women (OR, 1.08 per SD higher BT; 95% CI, 1.02 to 1.14; P = 0.005), where SNPs mapped to androgen signaling genes were used as genetic instruments (Figure 2, panel A ). Furthermore, there was no evidence of a causal association between serum testosterone levels and COVID-19 severity outcomes, such as hospitalization and critical illness (Figure 2, panels B and C ).
Notably, potential causal effects of serum estradiol and BT levels were mostly observed in women but not men, except for a suggestive inverse causal association of estradiol levels with COVID-19 severity in men, suggesting a sex-specific causal nature in relation to COVID-19 outcomes. Furthermore, there is no evidence supporting a causal association of serum SHBG levels with COVID-19 susceptibility and severity outcomes, whether including or excluding obesity-related SNPs (Figure 3 ).