We performed two-sample MR analysis with available summary-level data from the commonly available genome wide association studies (GWAS), The flow chart was shown in Fig. 1. Declaration of Helsinki statement and written informed consent had been obtained in the original publications. The summary-level data has been publicly published at https://gwas.mrcieu.ac.uk website for analysis.

The flow chart about three key assumptions in Mendelian randomization study

Genetic instruments of COVID-19 (ID: ebi-a-GCST011073) were obtained from a large-scale study including 1,683,768 participants (1,644,784 controls vs 38,984 cases) from European and 8,660,177 SNPs [16].

Obstetrician-related diseases refer to diseases with high incidence in obstetrics, including gestational diabetes, other disorders of amniotic fluid and membranes, Intrahepatic Cholestasis of Pregnancy (ICP), birth weight, gestational hypertension, spontaneous miscarriages, stillbirth and placental disorders [17]. The datasets we used were summary-level datasets and included populations from different European countries, therefore the diagnostic thresholds for the following common obstetrician-related diseases were not uniform. Gestational diabetes data (ID: finn-b-GEST_DIABETES) included 5687 cases and 117,892 controls, and 16,379,784 SNPs were obtained. Other disorders of amniotic fluid and membranes data (ID: finn-b-O15_AMNIOT_OTHER) included 1753 cases and 104,247 controls, and 16,379,393 SNPs were obtained. Intrahepatic Cholestasis of Pregnancy (ICP) (ID: finn-b-O15_ICP) included 940 cases and 122,639 controls, and 16,379,784 SNPs were obtained. Placental disorders (ID: finn-b-O15_PLAC_DISORD) included 102 cases and 104,247 controls, and 16,379,357 SNPs were obtained. Birth weight (ID: ukb-b-13378) included 261,932 participants, and 9,851,867 SNPs were obtained. Gestational hypertension/pre-eclampsia data (ID: ukb-b-13535) included 462,933 participants (1864 cases vs 461,069 controls), and 9,851,867 SNPs were obtained. Number of spontaneous miscarriages (ID: ukb-b-419) included 78,700 participants and 9,851,867 SNPs were obtained. Number of stillbirths (ID: ukb-b-6412) included 78,879 participants and 9,851,867 SNPs were obtained. All participants were of European descent.

R packages including TwoSampleMR (v 0.5.6), MendelianRandomization (v 0.7.0), and MRPRESSO (v 1.0) were used in this study. Instrumental variables (IVs) were obtained according to the three assumptions of MR. In the three assumption, we set the threshold of p-value as 1 × 10–5 and the threshold of r2 to include more IVs because some of the MR methods we used are less prone to weak instrument bias [18, 19]. Firstly, we selected SNPs that were closely associated with the COVID-19 at a significance level of p < 1 × 10–5, furthermore, SNPs with linkage disequilibrium (r2 = 0.05, kb = 10,000) and IVs with weak bias (F-statistics < 10) were removed. Secondly, we excluded SNPs that were associated with confounding factors (p < 1 × 10–5) that related to COVID-19 and obstetric-related diseases. Finally, SNPs that were directly related to the outcomes of interest (p < 1 × 10–5) were excluded to obtain the IVs. The formula for calculating R2 and F-statistics is in the form.

While MAF is minor allele frequency, SD = SE \(\times \sqrt\), N and n are the sample size and R2 is a risk factor for the genotype the explanation the proportion of variability.

We used Cochran’s Q test in inverse-variance weighting (IVW) method to assess heterogeneity in the sensitivity analysis. Horizontal pleiotropy was estimated by the intercept of the MR-Egger regression and MR-pleiotropy residual sum and outlier (MR-PRESSO). We also assessed whether individual SNP had biases that independently affected the overall causal effect by leave-one-out methods. Odds ratios (OR) (p < 0.05) in this study was presented to evaluate the cause effects.