The role of obesity in female reproductive conditions: A Mendelian Randomisation study

20

- Obesity is observationally associated with altered risk of many female reproductive 21 conditions. These include polycystic ovary syndrome (PCOS), abnormal uterine bleeding, 22 endometriosis, infertility, and pregnancy-related disorders. However, the roles and mechanisms of 23 obesity in the aetiology of reproductive disorders remain unclear. 24

and Findings - We estimated observational and genetically predicted causal associations 25 between obesity, metabolic hormones, and female reproductive conditions using logistic 26 regression, generalised additive models, and Mendelian randomisation (two-sample, non-linear, 27 and multivariable) applied to data from UK Biobank and publicly available genome-wide 28 association studies (GWAS). 29 Body mass index (BMI), waist-hip ratio (WHR), and WHR adjusted for BMI (WHRadjBMI) were 30 observationally (odds ratios (ORs) = 1.02 – 1.87 per 1 S.D. obesity trait) and 31 causally (ORs = 1.06 – 2.09) associated with uterine fibroids (UF), PCOS, heavy menstrual 32 bleeding (HMB), and pre-eclampsia. Causal effect estimates of WHR and WHRadjBMI, but not 33 BMI, were attenuated compared to their observational counterparts. Genetically predicted visceral 34 adipose tissue mass was causal for the development of HMB, PCOS, and pre-eclampsia (ORs = 35 1.01 - 3.38). Increased waist circumference also posed a higher causal risk (ORs = 1.16 – 1.93) for 36 the development of these disorders and UF than did increased hip circumference (ORs = 1.06 – 37 1.10). Leptin, fasting insulin, and insulin resistance each mediated between 20% -50% of the total 38 causal effect of obesity on pre-eclampsia. Reproductive conditions clustered based on shared 39 genetic components of their aetiological relationships with obesity. 40 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 3

- In this first systematic, large-scale, genetics-based analysis of the aetiological 41 relationships between obesity and female reproductive conditions, we found that common indices 42 of overall and central obesity increased risk of reproductive disorders to heterogenous extents, 43 mediated by metabolic hormones. Our results suggest exploring the mechanisms mediating the 44 causal effects of overweight and obesity on gynaecological health to identify targets for disease 45 prevention and treatment. 46

47 Obesity is commonly understood as the excess accumulation of body fat which leads to increased 48 health risks. In women, body mass index (BMI) is associated with increased prevalence of 49 gynecological conditions, including excessive and abnormal menstrual bleeding (1, 2), 50 endometriosis and uterine fibroids (UF) (3, 4), polycystic ovary syndrome (PCOS) (5, 6), 51 complications of pregnancy such as pre-eclampsia and eclampsia (7), miscarriage (8, 9), and 52 infertility (10, 11). These are often non-linear and heterogeneous relationships. While the risk of 53 anovulatory infertility and recurrent miscarriages are highest in obese women, underweight 54 women also have increased risk of infertility (9, 12). The association of BMI with endometriosis 55 varies by disease severity, as women with advanced-stage endometriosis have lower BMI than 56 those with minimal disease, and the inverse BMI-endometriosis association is stronger in women 57 with infertility (13, 14). Finally, although the severity of PCOS and menstrual disorders increases 58 with overall obesity, women presenting with these conditions are more likely to store fat in the 59 abdominal region, regardless of their BMI (2, 5). 60 Observational epidemiological studies are limited in assessing causality, due to confounding and 61 reverse causation. The Mendelian randomisation (MR) framework is a genetics-based instrumental 62 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 4 variable approach that relies on the random and fixed assignment of genetic variants at conception 63 to estimate the causal effect size of genetically predicted exposures on an outcome. MR has 64 previously indicated causal effects of genetically predicted BMI on the development of some 65 subtypes of ovarian cancer (odds ratio (OR) = 1.29 per 5 units of BMI) (15), endometrial cancer 66 (OR = 2.06 per 5 units of BMI) (16), and PCOS (OR = 4.89 per 1 S.D. higher BMI) (17). However, 67 the aetiological role of obesity and body fat distribution on many other female reproductive 68 diseases has not been reported. It is especially relevant to investigate the effects of fat distribution, 69 as there are intricate metabolic and endocrine links between adipose tissue and female reproductive 70 organs. Yet, causal investigations of such relationships are lacking. 71 Leptin, which is a hormone secreted by adipocytes, and elevated in individuals with obesity, is 72 increased in women with endometriosis, UF, and adverse pregnancy outcomes, even when 73 adjusted for BMI (7, 18-21). Obesity-induced insulin resistance additionally increases the risk and 74 severity of PCOS and pre-eclampsia by dysregulating steroid hormone and metabolic pathways 75 (5, 22, 23). The dysregulation of sex hormones, including oestrogen and testosterone, is likely to 76 play a role in the obesity-driven development of female reproductive disorders due to its close 77 associations with body fat (22, 24). Yet, the causal impact of these factors in mediating the 78 relationships between obesity and gynaecological diseases has not been detailed. 79 Here, we apply logistic regression, generalised additive models, two-sample, non-linear, and 80 multivariable MR to dissect the relationships of overall obesity and body fat distribution with a 81 range of female reproductive disorders, and investigate the mediating role of metabolic factors 82 including leptin and insulin. 83 84 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 5

85 Observational associations in UK Biobank 86 UK Biobank (UKBB) is a prospective UK-based cohort study with approximately 500,000 87 participants aged 40-69 at recruitment on whom a range of medical, environmental, and genetic 88 information is collected (25). We included 257,193 individuals self-identifying as females of white 89 ancestry in UKBB in our analyses. Baseline measurements of BMI (total body weight (kg) / 90 standing height 2 (m 2)) and waist-to-hip ratio (WHR) (waist circumference (cm) / hip 91 circumference (cm)), and WHR adjusted for BMI (WHRadjBMI) were used to estimate general 92 and central obesity, respectively. Cases of reproductive conditions were identified based on ICD9 93 and ICD10 primary and secondary diagnoses from hospital inpatient data, self-reported illness 94 codes, and primary care records ( Table A in S1 Table). We fitted logistic regression models to 95 estimate the associations of BMI, WHR, and WHRadjBMI with prevalence of endometriosis 96 (7,703 cases, 249,490 controls), heavy menstrual bleeding (17,229 cases, 239,964 controls), 97 infertility (2,194 cases, 254,999 controls), self-reported stillbirth, spontaneous miscarriage or 98 termination (81,102 cases, 176,091 controls), PCOS (746 cases, 256,447 controls), pre-eclampsia 99 (2,242 cases, 254,951 controls), and uterine fibroids (19,192 cases, 238,001 controls). Case 100 definitions for pre-eclampsia included eclampsia cases to capture cases in which the former may 101 have developed into the latter. For each disease, individuals not included in the case group were 102 used as controls. BMI, WHR and WHRadjBMI were adjusted for age, age-squared, assessment 103 centre, and smoking status. The residuals were rank-based inverse normally transformed. Multiple 104 testing correction was applied using the false discovery rate (FDR) to evaluate statistical 105 significance while minimising false negatives (26). 106 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 6 We also tested associations without adjustment for smoking status, as it has previously been 107 suggested that higher BMI increases risk of smoking (27) and adjustment for both could therefore 108 induce collider bias. Adjustment for menopause status was not performed as up to 42% of women 109 with reproductive disorders in UKBB report being unsure of their menopause status as compared 110 to 16% of women who do not have a recorded history or presence of a reproductive condition 111 (Table A in S1 Table). 112 To evaluate the presence of non-linear observational associations between obesity and each 113 reproductive trait, fractional polynomial regression following the closed test procedure was 114 performed using the mfp v1.5.2 R package (28). This algorithm tests for the presence of an overall 115 association, the likelihood of non-linearity, and selects the best-fitting fractional polynomial 116 function. We also fitted generalised additive models (GAM) to the same data, allowing for 117 smoothing of the obesity trait with splines, using the mgcv 1.8-31 R package (29). All models were 118 adjusted for age, age-squared, assessment centre, and smoking status. Model fits were compared 119 with Akaike’s Information Criterion (AIC) (30). 120 Two-sample Mendelian Randomisation 121 Genetic instruments for BMI, WHR, and WHRadjBMI were selected based on the sentinel variants 122 at genome-wide significant loci ( P < 5E-9) reported in the largest publicly available European 123 ancestry GWAS of Genetic Investigation of ANthropometric Traits (GIANT) and UKBB (max N 124 individuals = 806,801) (31). Similarly, genetic instruments for predicted visceral adipose tissue 125 (VAT) mass (N individuals = 325,153) (32), waist circumference (N individuals = 462,166) and 126 hip circumference (N individuals = 462,117) (33), and waist-specific and hip-specific WHR (N 127 individuals = 18,330) (34) were selected based on the largest publicly available GWAS. 128 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 7 Three instrument weighting strategies were considered where sex-stratified GWAS results were 129 available: (i) SNPs from combined-sexes GWAS with combined-sexes weights (effect sizes), (ii) 130 combined-sexes SNPs with female-specific weights, or (iii) female-specific SNPs with female-131 specific weights. The method of female-specific SNPs with female-specific weights produced the 132 strongest instruments as evaluated by F-statistics and was thus chosen for analysis (Table B in S1 133 Table). Additionally, due to concerns of ascertainment bias in UKBB (35, 36), sensitivity analyses 134 with combined-sexes instruments (combined-sexes SNPs with combined-sexes weights) were also 135 performed. 136 Associations of the genetic instruments for obesity traits with female reproductive diseases were 137 obtained by performing a fixed-effect inverse-variance weighted meta-analysis of publicly 138 available GWAS summary statistics from two large biobank projects - FinnGen and UKBB (37). 139 The meta-analysis was performed using METAL (38) by matching the relevant ICD codes (Table 140 C in S1 Table) for the following traits: infertility (4,996 cases, 421,223 controls), pre-eclampsia 141 (2,711 cases, 480,373 controls), and uterine fibroids (21,835 cases, 456,551 controls). For 142 endometriosis, summary statistics were obtained by request from a recent European ancestry 143 GWAS (39) and meta-analysed as above with publicly available FinnGen and UKBB summary 144 statistics (12,210 cases, 450,183 controls). For heavy menstrual bleeding (HMB) (9,813 cases, 145 210,946 controls), sporadic miscarriage, i.e. 1-2 miscarriages (50,060 cases, 174,109 controls), 146 and multiple consecutive miscarriage, i.e. >= 3 consecutive miscarriages (750 cases, 150,215 147 controls), publicly available summary statistics were obtained from recent European ancestry 148 GWAS that include UKBB individuals (3, 40). For PCOS, estimates were based on a fixed-effect 149 inverse variance-weighted meta-analysis of published GWAS summary statistics (38), publicly 150 available GWAS results by FinnGen, and a European-ancestry GWAS run in UKBB using SAIGE 151 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 8 (11,186 cases, 273,812 controls). As a sensitivity analysis, all MR tests were performed using 152 disease association estimates based on FinnGen only, where available, to alleviate bias due to 153 sample overlap between the exposure and outcome GWAS sources. 154 Power to detect MR associations was calculated using two methods, one designed for general two-155 sample MR (41) and the other for MR performed on binary outcomes (42). Briefly, these methods 156 calculate power by accounting for GWAS sample size, proportion of cases in case-control GWAS, 157 and variance explained by genetic instruments for the exposure. The power to detect a true odds 158 ratio (OR) association of 1.1 or more extreme at an unadjusted significance level of 0.05 was 159 estimated. 160 Instrument SNPs were extracted from the outcome GWAS results, harmonised for consistency in 161 the alleles, and MR was performed using the TwoSample MR v0.5.4 R package (43). Three 162

for MR - inverse-variance weighted (IVW), MR-Egger, and weighted median - were 163 evaluated, and the best method was selected via Rucker's framework (44). Briefly, this framework 164 advises to choose the MR method with least heterogeneity as assessed by Cochran's Q-statistic, 165 while accounting for the trade-off between power and pleiotropy (45). Inverse-variance weighted 166 results, which were the best method chosen by Rucker’s framework for all tested associations, are 167 reported in the main text, but results from all methods are calculated for robustness and displayed 168 in the supplementary information. Multiple-hypothesis testing correction was applied with the 169 FDR method and significance established at FDR < 0.05. MR-Egger intercept tests were performed 170 to detect horizontal pleiotropy, and single-SNP and leave-one-out analyses were used to identify 171 outlier SNPs driving relationships (43). 172 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 9 Reverse MR for obesity traits regressed on female reproductive conditions was performed as 173 detailed above. Genetic instruments for endometriosis (14,926 cases, 189,715 controls) (46), 174 PCOS (10,174 cases, 103,164 controls) (47), and uterine fibroids (20,406 cases, 223,918 controls) 175 (3) were constructed from index variants identified by the largest European ancestry GWAS for 176 each trait. Instrument strength was assessed by F-statistics (endometriosis, 16 SNPs, F = 5.13; 177 PCOS, 14 SNPs, F = 41.6; UF, 29 SNPs, F = 11.1). Associations of genetic instruments for these 178 reproductive conditions with BMI, WHR, and WHRadjBMI were obtained from female-specific 179 summary statistics from the above-mentioned GIANT-UKBB meta-analysis (31). 180 Non-linear Mendelian Randomisation 181 For non-linear MR analyses, we selected female UK Biobank participants of white British ancestry 182 with no second-degree or closer relatives in the study, as identified by the UKBB team (48), to 183 avoid violation of the MR assumption of random assignment of genetic variants; 207,705 women 184 were retained following this selection. Genetic instruments for BMI were constructed for each 185 individual using female-specific index variants from Pulit et al.'s GIANT-UKBB meta-analysis 186 (31). The instruments for BMI explained 4.15% of trait variance after adjustment for age, age-187 squared, smoking status, assessment centre, genotyping array, and the first ten genetic principal 188 components to account for population stratification. Binomial non-linear MR, a method designed 189 to assess causal relationships in different exposure strata while avoiding collider bias, was 190 performed using the fractional polynomial method with 100 quantiles and the piecewise linear 191

with 10 quantiles (49). Outcomes were restricted to female reproductive disorders with 192 prevalence > 5% in UKBB (i.e. HMB, miscarriage, and UF) to maintain sufficient sample sizes in 193 each quantile to estimate localised average causal effects. We assessed non-linearity with the 194 fractional polynomial non-linearity and Cochran's Q tests, and tested heterogeneity of the 195 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 10 instrumental variable (IV) with the Cochran's Q and trend tests. All analyses were performed with 196 the nlmr v2.0 R package (49). 197 MR with Mediation Analysis 198 To investigate the extent to which obesity affects female reproductive disorders via hormone-199 related mediators, two-step MR by the product of coefficients method was performed using GWAS 200 summary statistics. This method was chosen as female reproductive disease phenotypes are binary 201 outcomes with disease prevalence < 10% in UK Biobank, for which two-step MR provides the 202 least biased estimates of mediation (50). Summary statistics for leptin (N = 33,987) (51), fasting 203 insulin (N = 51,750) (52), and insulin sensitivity (N = 16,753) (53) were obtained from publicly 204 available European-ancestry GWAS sources that do not include samples from UKBB to minimise 205 bias from sample overlap (Table B in S1 Table). 206 In the first step of two-step MR, the mediators were regressed on obesity-related exposures using 207 summary statistics MR methods described above. The direction of causality for all relationships 208 was confirmed with the MR-Steiger directionality test (54) and reciprocal MR with mediator 209 instruments and obesity-related exposures as outcomes were performed to ensure correct direction 210 of causality. In the second step, multivariable MR (MVMR) was performed using combined 211 genetic instruments for each obesity trait and hormone to estimate the independent effect of the 212 mediator on each outcome after adjusting for the value of the exposure; and to estimate the 213 independent effect of the exposure on outcome when adjusted for the value of each mediator. This 214 was only done for traits where the total unadjusted effect of the exposure on the outcome was 215 significant (FDR < 0.05). Odds ratios (ORs) for binary outcomes were converted to log ORs to 216 calculate mediated effect by the product of coefficients method. The proportion of effect mediated 217 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 11 was calculated by dividing indirect effect over total effect. Standard errors were estimated with the 218 delta method (55). 219 Disease and SNP Clustering 220 To assess similarities in the aetiological relationships of different reproductive conditions with 221 obesity traits, we projected single SNP causal estimates for BMI, WHR and WHRadjBMI on the 222 reproductive traits, estimated using the Wald ratio, in a two-dimensional space using UMAP. SNPs 223 were annotated to their nearest gene with SNPsnap (56). 224 To identify the genetic instruments driving the causal obesity-reproductive trait association, and 225 identify clusters of SNPs with distinct causal effect sizes, we clustered SNPs by the magnitude of 226 their causal effect using mixture model clustering in the MR-Clust v0.1.0 R package (56). For each 227 obesity trait-reproductive disease pair, the algorithm distinguishes the genetic instruments for the 228 obesity traits that do not have an effect on the disease (“null cluster”), from those which have a 229 similar scaled effect on the disease (the “substantial clusters”), and those that have a scaled effect 230 that cannot be grouped with other variants (“junk cluster”). 231 Code availability 232 All scripts used in analyses are deposited at 233 https://github.com/lindgrengroup/obesity_femrepr_MR 234 235 236 237 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 12

238 Obesity traits are observationally associated with female reproductive diseases in UK 239 Biobank 240 BMI at baseline assessment (age 40-69) was positively associated with the prevalence of most 241 female reproductive disorders in UKBB, with the strongest association observed between BMI and 242 pre-eclampsia (odds ratio (OR) per 1 S.D. higher BMI = 1.87, P = 1.90E-64). Associations with 243 WHRadjBMI were null or lower than those for WHR (ORs for WHRadjBMI v. WHR for PCOS 244 = 1.06 v. 1.48, pre-eclampsia = 1.02 v. 1.13, endometriosis = 1.02 v. 1.08, HMB = 1.06 v. 1.14, 245 and UF = 1.02 v. 1.08), indicating that BMI may be driving many of the associations between 246 WHR and female reproductive diseases (Figures 1 & 2, Table 1). Infertility was the only disorder 247 for which BMI (OR = 0.894, P = 2.16E-07) and WHR (OR = 0.927, P = 4.08E-04) were inversely 248 associated with disease. 249 Non-linear models explained the associations of BMI with many reproductive disorders better than 250 linear models. We observed inverted-U and plateau relationships with endometriosis (linear AIC 251 = 67091, generalised additive model (GAM) AIC = 67051), uterine fibroids (linear AIC = 134160, 252 GAM AIC = 134094), HMB (linear AIC = 116687, GAM AIC = 116636), miscarriage (linear AIC 253 = 314828, GAM AIC = 314819), and pre-eclampsia (linear AIC = 24826, GAM AIC = 24814) 254 (Figure 1, Table D in S1 Table ). All three obesity traits displayed U-shaped relationships with 255 PCOS. 256 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 13 Observational estimates between all obesity traits and female reproductive disorders did not differ 257 with or without adjustment for smoking status (Table E in S1 Table). Statistical significance after 258 multiple-testing correction was established at FDR < 0.05, unadjusted P < 0.04. 259 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 14 Table 1: Observational and genetic associations between obesity traits and female reproductive disorders. Diagnosis Obesity trait Logistic regression Mendelian randomisation OR (95% CI) Unadj. p- value # SNPs OR (95% CI) per 1 S.D. higher obesity trait Unadj. p- value Endometriosis BMI 1.14 (1.12 - 1.16) 4.45E-29 264 1.04 (0.902 - 1.19) 0.606 WHR 1.07 (1.05 - 1.10) 1.00E-09 190 1.24 (1.05 - 1.47) 1.00E-02 WHRadjBMI 1.02 (0.99 - 1.04) 0.188 250 1.24 (1.08 - 1.41) 1.67E-03 Heavy menstrual bleeding BMI 1.20 (1.18 - 1.22) 7.78E-117 268 1.01 (1.004 - 1.013) 3.62E-04 WHR 1.15 (1.13 - 1.16) 8.35E-69 191 1.01 (1.005 - 1.02) 1.42E-04 WHRadjBMI 1.06 (1.04 - 1.07) 5.30E-13 251 1.01 (1.002 - 1.011) 5.20E-03 Infertility BMI 0.894 (0.852 - 0.936) 2.16E-07 267 0.982 (0.810 - 1.19) 0.856 WHR 0.927 (0.884 - 0.969) 4.08E-04 191 1.10 (0.901 - 1.34) 0.355 WHRadjBMI 0.971 (0.929 - 1.01) 0.176 251 1.21 (1.03 - 1.43) 0.0214 Miscarriage BMI 1.03 (1.02 - 1.04) 4.28E-14 265 1.06 (1.01 - 1.12) 0.0238 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 15 (sporadic) WHR 1.04 (1.04 - 1.05) 3.82E-24 190 0.998 (0.947 - 1.05) 0.933 WHRadjBMI 1.03 (1.02 - 1.04) 1.87E-13 250 0.996 (0.953 - 1.04) 0.878 Miscarriage (multiple consecutive) BMI 254 0.917 (0.570 - 1.48) 0.720 WHR 184 1.20 (0.743 - 1.92) 0.462 WHRadjBMI 240 0.978 (0.662 - 1.44) 0.911 PCOS BMI 1.87 (1.80 - 1.94) 1.90E-64 268 1.13 (1.08 - 1.19) 7.60E-08 WHR 1.48 (1.41 - 1.55) 3.32E-26 191 1.07 (1.02 - 1.11) 4.30E-03 WHRadjBMI 1.06 (0.986 - 1.13) 0.124 251 1.02 (0.990 - 1.06) 0.222 Pre-eclampsia BMI 1.25 (1.21 - 1.29) 3.85E-25 266 2.09 (1.60 - 2.73) 5.16E-08 WHR 1.13 (1.09 - 1.17) 4.97E-09 191 1.57 (1.16 - 2.10) 2.92E-03 WHRadjBMI 1.02 (0.982 - 1.07) 0.272 250 1.43 (1.13 - 1.80) 2.46E-03 Uterine fibroids BMI 1.14 (1.12 - 1.15) 2.43E-63 268 1.21 (1.08 - 1.35) 9.93E-04 WHR 1.08 (1.06 - 1.09) 2.75E-23 191 1.24 (1.10 - 1.41) 6.20E-04 WHRadjBMI 1.02 (1.01 - 1.04) 2.94E-03 251 1.17 (1.06 - 1.29) 1.95E-03 Body fat distribution is causally related to risk of female reproductive diseases 260 Two-sample MR indicated that higher genetically predicted WHR and/or WHRadjBMI are causal 261 for higher risk of pre-eclampsia (OR per 1 S.D. higher WHR = 1.57, P = 2.92E-03; WHRadjBMI 262 = 1.43, P = 2.46E-03), endometriosis (WHR = 1.24, P = 1.00E-02; WHRadjBMI = 1.24, P = 263 1.67E-03), uterine fibroids (WHR = 1.24, P = 6.20E-04; WHRadjBMI = 1.17, P = 1.95E-03), 264 infertility (WHRadjBMI = 1.21, P = 2.14E-02), and PCOS (WHR = 1.07, P = 4.30E-03) (Table 265 1, Figure 2). The causal estimates of WHR and WHRadjBMI on most reproductive disorders were 266 higher than their observational counterparts. While genetically predicted BMI also increased risk 267 of most female reproductive disorders (ORs per 1 S.D. higher BMI = 1.01 for HMB to 2.09 for 268 pre-eclampsia), MR estimates of associations between BMI and HMB (OR = 1.01, P = 3.62E-04), 269 endometriosis (OR = 1.04, P = 0.606), and PCOS (OR = 1.13, P = 7.60E-08) were much attenuated 270 compared to observational results. 271 Genetically predicted visceral adipose tissue (VAT) mass was causal for the development of pre-272 eclampsia (OR per 1 kg increase in predicted VAT mass = 3.08, P = 6.65E-07), PCOS (OR = 1.15, 273 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 16 P = 3.24E-05), and HMB (OR = 1.01, P = 0.0125) ( Figure 3 and Table F in S1 Table ). The 274 differential effect of body fat distribution on female reproductive traits was further reflected in the 275 heterogeneous causal effects of waist circumference (WC) and hip circumference (HC) on disease 276 development. Increased WC posed a higher risk than did increased HC for pre-eclampsia (ORs per 277 1 S.D. higher WC = 1.93 v. HC = 1.40, heterogeneity P-het = 0.0373), uterine fibroids (WC = 1.32 278 v. HC = 1.12, P-het = 7.70E-03), and PCOS (WC = 1.16 v. HC = 1.10, P-het = 0.0325). We did 279 not see this heterogeneity in observational associations (all P-het > 0.164) (Figure 3 and Table F 280 in S1 Table). 281 No significant causal effects were found when restricting MR analyses to genetic instruments with 282 a specific effect of waist but not hip circumference, or on hip but not waist circumference ( Table 283 H in S1 Table and S1 Figure), but the power based on these instruments to detect odds ratios 284 more extreme than 1.1 was limited to 5% - 20% (Table I in S1 Table). No non-linear MR models 285 explained the causal effects of BMI on any reproductive disorder better than linear MR models 286 (S3 Figure). However, the power to detect non-linear effects was severely limited by the lower 287 number of cases in each quantile of the BMI distribution in which analyses were run. Statistical 288 significance after multiple-testing correction was established at FDR < 0.05, unadjusted P 60; instrument 291 strength for waist- and hip-circumference was > 45 ( Table B in S1 Table ). We found MR 292 estimates to be consistent between the different MR methods (heterogeneity P > 0.321), when 293 based only on FinnGen summary statistics (heterogeneity P > 0.163), or with combined-sex 294 instruments (heterogeneity P > 0.999), suggesting that the findings were not dependent on the 295 adopted MR method, or substantially biased due to sample overlap between exposure and outcome 296 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 17 GWAS sources or ascertainment bias in UKBB (35, 36) (Tables F, J, K in S1 Table & S1 & S2 297 Figure). 298 We did not find evidence for reverse causal effects of endometriosis, PCOS, or uterine fibroids on 299 BMI, WHR, and WHRadjBMI ( Table L in S1 Table ). However, these estimates may be biased 300 by weak genetic instruments for endometriosis (F-statistic = 5.13) and UF (F-statistic = 11.1), and 301 high heterogeneity for all associations (Cochran’s Q P < 4.71E-06). We were limited in assessing 302 reverse causality of other female reproductive conditions on obesity traits by the lack of large-303 scale publicly available GWAS summary statistics. 304 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 18 Leptin and insulin mediate the causal effects of obesity on female reproductive disorders 305 We applied a series of MR-based mediation analyses (58, 59) to study the role of hormonal factors 306 - leptin and insulin resistance - in mediating the causal relationships between obesity and female 307 reproductive health (Figure 4A). The effects of BMI, WHR, and WHRadjBMI on endometriosis, 308 PCOS, pre-eclampsia, and UF were attenuated (95% CIs of ORs all contain 1) when adjusted for 309 leptin, fasting insulin, or insulin sensitivity as measured by the modified Stumvoll Insulin 310 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 19 Sensitivity Index (ISI) (Figure 4B, Table M in S1 Table). Furthermore, these hormones influence 311 risk of pre-eclampsia independently of obesity. After adjustment for BMI, leptin (ꞵ = 0.887, P = 312 1.28E-04), fasting insulin (ꞵ = 1.42, P = 1.29E-03), and ISI (ꞵ = -0.503, P = 2.99E-03) were all 313 associated with risk of pre-eclampsia (Figure 4C, Table 2). Similarly, fasting insulin (ꞵ = 1.27, P 314 = 7.15E-03) and ISI (ꞵ = -0.793, P = 1.46E-05) had causal effects on pre-eclampsia upon 315 adjustment for WHR. Leptin, fasting insulin, and ISI did not have significant causal effects on 316 endometriosis, PCOS, or UF after adjustment for obesity traits. Statistical significance after 317 multiple-testing correction was established at FDR < 0.05, unadjusted P < 0.01. 318 We calculated the proportion of total obesity effect mediated by the above hormones for disorders 319 where the effects of obesity traits and mediators were significant at unadjusted P < 0.05. We found 320 that leptin (50.2% of effect of BMI on pre-eclampsia), fasting insulin (between 27.7% - 36.6% of 321 different effects), and ISI (between 19.1% - 50.1% of different effects) each mediated the total 322 causal effect of obesity traits on female reproductive disorders (Table 3). 323 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 20 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 21 Table 2: Multivariable MR estimates of female reproductive disorders regressed on metabolic hormones, adjusted for obesity traits. Outcome Exposure Adjusted for ꞵ +/- S.E. per 1 S.D. higher exposure Unadjusted P- value Endometriosis Leptin Unadjusted -0.0745 +/- 0.188 0.692 BMI -0.0457 +/- 0.102 0.654 WHR -0.131 +/- 0.126 0.297 WHRadjBMI -0.0964 +/- 0.128 0.453 Uterine fibroids Leptin Unadjusted -0.132 +/- 0.323 0.683 BMI 0.0881 +/- 0.102 0.389 WHR -4.12E-03 +/- 0.138 0.976 WHRadjBMI -0.0834 +/- 0.14 0.550 PCOS Leptin Unadjusted -0.134 +/- 0.0701 0.0567 BMI 0.0491 +/- 0.0409 0.231 WHR 8.44E-03 +/- 0.0476 0.859 WHRadjBMI -0.0310 +/- 0.0473 0.511 Pre-eclampsia Leptin Unadjusted 0.181 +/- 0.501 0.718 BMI 0.887 +/- 0.232 1.28E-04 WHR 0.468 +/- 0.309 0.129 WHRadjBMI -0.102 +/- 0.333 0.760 Endometriosis Fasting insulin Unadjusted 0.150 +/- 0.247 0.544 BMI 0.0996 +/- 0.209 0.634 WHR 0.186 +/- 0.254 0.465 WHRadjBMI 0.423 +/- 0.239 0.0770 Uterine fibroids Fasting insulin Unadjusted 0.0263 +/- 0.342 0.939 BMI 0.311 +/- 0.183 0.090 WHR 0.313 +/- 0.209 0.135 WHRadjBMI 0.262 +/- 0.202 0.195 PCOS Fasting insulin Unadjusted -0.0334 +/- 0.158 0.833 BMI 0.0283 +/- 0.0772 0.714 WHR 0.0932 +/- 0.0788 0.237 WHRadjBMI 0.0289 +/- 0.0726 0.690 Pre-eclampsia Fasting insulin Unadjusted -0.0379 +/- 0.873 0.965 BMI 1.42 +/- 0.441 1.29E-03 WHR 1.27 +/- 0.473 7.15E-03 WHRadjBMI 0.476 +/- 0.467 0.308 Endometriosis Insulin sensitivity Unadjusted -0.0940 +/- 0.133 0.478 BMI -0.101 +/- 0.0864 0.244 WHR -0.131 +/- 0.100 0.192 WHRadjBMI -0.102 +/- 0.106 0.336 Uterine fibroids Insulin sensitivity Unadjusted -0.223 +/- 0.137 0.103 BMI -0.0951 +/- 0.0716 0.184 WHR -0.202 +/- 0.0830 0.0151 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 22 WHRadjBMI -0.120 +/- 0.0855 0.165 PCOS Insulin sensitivity Unadjusted -0.0239 +/- 0.0616 0.698 BMI -0.0257 +/- 0.0290 0.376 WHR -0.0355 +/- 0.0293 0.226 WHRadjBMI -0.0443 +/- 0.0294 0.132 Pre-eclampsia Insulin sensitivity Unadjusted -0.354 +/- 0.295 0.229 BMI -0.503 +/- 0.169 2.99E-03 WHR -0.793 +/- 0.183 1.461E-05 WHRadjBMI -0.519 +/- 0.205 0.0110 Table 3: Proportion of effect mediated for exposure - mediator - outcome relationships. Exposure Mediator Outcome Log OR (S.E.) per 1 S.D. higher exposure Unadj. pval Proportion of Effect Mediated (95% CI) Exposure - Outcome Mediator - Outcome Exposure - Mediator BMI Leptin Pre-eclampsia 0.737 (0.135) P = 5.16E-08 0.887 (0.232) P = 1.28E-04 0.417 (0.0262) P = 7.94E-57 50.2% (18.2% - 82.2%) Fasting insulin 1.42 (0.441) P = 1.29E-03 0.144 (0.0141) P = 1.61E-24 27.7% (7.40% - 48.0%) Insulin sensitivity -0.503 (0.169) P = 2.99E-03 -0.281 (0.0490) P = 9.83E-09 19.1% (3.33% - 35.0%) WHR Fasting insulin Pre-eclampsia 0.449 (0.151) P = 2.92E-03 1.27 (0.473) P = 7.15E-03 0.129 (0.0159) P = 5.04E-16 36.6% (0% - 73.7%) Insulin sensitivity -0.793 (0.183) P = 1.46E-05 -0.283 (0.0601) P = 2.44E-06 50.1% (4.98% - 95.3% Insulin sensitivity Uterine fibroids 0.218 (0.0637) P = 6.20E-04 -0.202 (0.083) P = 1.51E-02 -0.283 (0.0601) P = 2.44E-06 26.2% (0% - 54.4%) WHRadjBMI Insulin sensitivity Pre-eclampsia 0.358 (0.118) P = 2.46E-03 -0.519 (0.205) P = 1.14E-02 -0.168 (0.0427) P = 8.21E-05 24.4% (0% - 51.8%) . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 23 Other metabolic and hormone pathways may drive the aetiological relationships of obesity 324 with female reproductive diseases 325 We assessed the similarities in the aetiological relationships of different reproductive conditions 326 with obesity, by projecting the single SNP causal estimates for BMI, WHR and WHRadjBMI on 327 the reproductive traits in a two-dimensional space using UMAP ( Figure 5A ). The UMAP 328 projections based on all obesity traits clustered endometriosis and UF together with infertility and 329 HMB, which were further separated from miscarriage (sporadic and multiple consecutive). While 330 PCOS and pre-eclampsia were grouped closely in UMAP plots of the effect of WHR and 331 WHRadjBMI variants, they were separated by BMI-associated variants. This reflects a shared 332 genetic component of the aetiological role of general and central obesity in the three groups of 333 reproductive conditions. 334 We further examined if different aspects of obesity play an aetiological role in different 335 reproductive conditions. For each obesity trait-reproductive disease pair, we grouped the genetic 336 instruments for the obesity traits by those that do not have an effect on the disease (“null cluster”), 337 those which have a similar scaled effect on the disease (the “substantial clusters”), and those that 338 have a scaled effect that cannot be grouped with other variants (“junk cluster”) using MRClust 339 (57). One substantial cluster was identified for each pair of obesity traits and reproductive 340 conditions. The only exception to this was with WHRadjBMI and UF, for which two substantial 341 clusters were identified, one with positive causal effect and the other with negative effect ( S4 342 Figure and Table N in S1 Table ). Of the 4 SNPs in the negative effect cluster, rs2277339 343 (missense variant in PRIM1 and upstream of HSD17B6, involved in steroid biosynthesis) is 344 associated with primary ovarian insufficiency, early menopause, and PCOS (60, 61), and 345 rs11694173 is intronic to THADA, which is also associated with PCOS (47). On the other hand, 4 346 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 24 of 10 SNPs in the positive effect cluster are associated with metabolic traits - rs12328675 and 347 rs2459732 with circulating leptin (62), rs6905288 with type 2 diabetes and thyroid stimulating 348 hormone (63, 64), and rs4686696 is intronic to insulin-like growth factor IGF2BP2. 349 SNPs with high probability of belonging to the substantial cluster (≥ 80% probability) were 350 generally unique to each obesity-disease relationship, with no more than 2 variants shared between 351 any two clusters ( Figure 5B). However, 6 BMI index SNPs had positive causal effect estimates 352 for both PCOS and pre-eclampsia, including rs1121980 in the adipose-associated gene FTO and 353 rs7498665 in SH2B1, linked to insulin resistance in obesity. The BMI-associated variant 354 rs7084454 (intronic to MLLT10) was shared by substantial clusters for PCOS, endometriosis, and 355 UF, while rs114760566 (mapped to HMGA1, associated with type 2 diabetes and multiple 356 lipomatosis) was shared by endometriosis and UF. We evaluated the biological effect of the top 357 SNPs in each substantial cluster with the DEPICT algorithms for pathway enrichment and gene 358 prioritisation. We recapitulated the known associations of GEMIN5 subnetwork enrichment in 359 SNPs causal for BMI-PCOS, which has previously been implicated in the aetiology of PCOS (65). 360 Gene prioritisation for WHR-endometriosis causal SNPs highlighted TBX15, an important 361 mesodermal transcription factor with roles in endometrial and ovarian cancer (66, 67). 362 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 25 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 26

363 In this first systematic genetics-based causal investigation of the aetiological role of obesity in 364 female reproductive health, we report evidence that common indices of obesity increase risk of a 365 broad range of reproductive conditions whose effects may be non-uniform across the obesity 366 spectrum. The strongest effect of generalised obesity was found for pre-eclampsia, while more 367 modest effects were observed for nearly all other studied conditions. We identified endocrine 368 mechanisms, including those related to leptin and insulin resistance, as potential drivers of 369 aetiological relationships of both generalised and central obesity with female reproductive health. 370 Finally, we found genetic evidence that certain groups of reproductive conditions, such as UF and 371 endometriosis, may share a mechanistically similar relationship with obesity. 372 Our findings highlight that the relationships between obesity and female reproductive disorders 373 are (i) non-uniform in their nature and strength, and (ii) observationally non-linear across the 374 obesity spectrum. We report substantial differences in the causal effect estimates of BMI on 375 reproductive diseases, with each S.D. increase in BMI doubling the risk of pre-eclampsia, but more 376 moderately (ORs = 1.01 - 1.25 for PCOS, miscarriage, UF, HMB) or not at all (infertility, 377 endometriosis) affecting other conditions. Conversely, central fat distribution independent of BMI 378 showed substantial genetically predicted effects on both infertility and endometriosis (ORs per 1 379 S.D. increase in WHRadjBMI = 1.21 - 1.46) as well as on pre-eclampsia and UF (ORs = 1.17 - 380 1.43), but not on PCOS, HMB, and miscarriage. These findings highlight that the aetiological role 381 of obesity in female reproductive diseases is heterogeneous in its effect strength, and may be driven 382 by overall adiposity (PCOS, HMB, and miscarriage), isolated central obesity (infertility and 383 endometriosis), or by both generalised and central obesity (pre-eclampsia and UF). 384 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 27 For several reproductive conditions, we found substantial differences between the observational 385 and genetically predicted causal effect estimates, which may indicate a bi-directional relationship 386 between obesity and reproductive health. For instance, while the observational analyses suggested 387 an 87% increase in PCOS risk per S.D. higher BMI, the MR analyses indicated that each S.D. 388 higher BMI causally increases PCOS risk by only 13%. Similarly, 1 S.D. higher WHR and 389 WHRadjBMI causally increase endometriosis risk by 24%, while the observational analyses 390 suggest a more modest increase in risk of 7% and 2% respectively. This discrepancy may in part 391 be due to reverse causality, which we were not powered to detect in this study, as the number and 392 strength of the available genetic instruments for reproductive conditions is substantially lower than 393 those for BMI and WHR. The obesity traits upon which the observational analyses were based 394 were measured at ages 44-67, which was for most conditions likely to be several years or decades 395 after women developed the condition, and often post-menopause. While our observational analyses 396 adjusted for the effect of age on obesity traits, adjusting for menopause status proved to be 397 unreliable as up to 42% of women with reproductive diseases in UKB were unsure of their 398 menopause status, as opposed to 16% of female participants without a diagnosis for any of the 399 studied conditions. The observational estimates may therefore capture both the effect of obesity 400 on disease risk as well as any downstream effects of the disease or commonly used treatments on 401 body weight and fat distribution. For instance, the large observational effect of BMI on PCOS 402 prevalence may reflect both a causal effect of obesity on disease risk (68), as captured by the 403 genetically predicted effect, as well as weight gain as a consequence of PCOS (69). Other potential 404 contributing factors to the differences between genetic and observational estimates are 405 confounding by unmeasured variables which lead to inflated observational associations (70, 71), 406 referral bias wherein obesity status affects the likelihood of receiving a diagnosis (72, 73), or 407 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 28 differences in pre- and post-menopausal weight and body fat distribution not captured by age (74). 408 Finally, while the observational relationships between obesity and some female reproductive 409 disorders were non-linear, we did not find non-linearity in the causal effects of BMI on these 410 diseases. The non-linear MR analyses were likely under-powered to detect associations with few 411 cases in each quantile of the BMI spectrum. 412 We noted that genetic estimates for the effect of fat distribution were not similarly attenuated when 413 compared to BMI effects. This disparity may be due to the differing impacts of overall and 414 abdominal (central) adiposity, as the latter is thought to be biologically more directly linked to 415 female reproductive health than generalised obesity, via pathways including insulin resistance and 416 hyper-androgenaemia (5, 75-77). Supporting the stronger effect of central body fat, we also 417 reported higher causal effects of waist than hip circumference with HMB, PCOS, pre-eclampsia, 418 and UF. Genetically predicted visceral adipose tissue mass increased risk of PCOS and pre-419 eclampsia, in line with observational studies (78). VAT mass is also observationally associated 420 with uterine fibroids (76), yet we did not find a significant causal effect of genetically predicted 421 VAT mass on development of UF, which may suggest a bi-directional or reverse causal 422 relationship. 423 Endometriosis and infertility were the only reproductive conditions which did not show a 424 consistently positive link with obesity. The modest observational associations of both BMI and 425 WHR with higher endometriosis prevalence in UKBB contradict previous studies, including 426 prospective cohort studies, which reported that lower BMI was associated with increased disease 427 prevalence (14, 79, 80). The positive association with endometriosis may in part be due to weight 428 gain as a consequence of the disease, for instance due to hormonal treatments (81-83), chronic pain 429 (84), inflammation (85), or earlier onset of menopause (86). We however did not find evidence 430 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 29 that generalised obesity plays a causal role in the aetiology of endometriosis, which suggests that 431 the observational finding reflects a reverse causal relationship. Conversely, the positive genetically 432 predicted effect of WHRadjBMI on endometriosis risk indicates a causal role for abdominal fat 433 distribution. For infertility, we observe a similar divergence between the observational and 434 genetically predicted effects of obesity traits, with BMI showing a negative observational 435 association, but WHRadjBMI a genetically predicted positive association. The causes of female 436 infertility are multiple, ranging from PCOS (87) and anovulation (88), to tubal disease (89), 437 endometriosis (90), low oocyte quality (91), hormonal and immunological dysfunction (92-95), 438 and yet unknown mechanisms. Each of these may have distinct and complex relationships with 439 obesity, which cannot be captured by studying the links with infertility of any cause. Nonlinear 440 effects, such as the increased association of under- and overweight with incidence of infertility 441 (12, 96), may also obscure these estimates, although our observational analyses did not provide 442 evidence for a non-linear relationship. 443 We conducted the first genetics-based investigation of the mediating hormonal pathways 444 underlying the causal relationships between obesity and female reproductive health. We identify 445 mechanisms related to insulin resistance and leptin as mediators of the effects of obesity traits on 446 UF and pre-eclampsia. The latter is consistent with hypotheses that obese women with metabolic 447 dysregulation are at highest risk of developing hypertensive disorders of pregnancy via angiogenic 448 and pro-inflammatory mechanisms. Increased circulating leptin may have a vasoconstructive, 449 hypertensive effect, which may be worsened by attenuation of insulin-induced vasorelaxation and 450 increased levels of TNF-alpha and IL6 (7, 21). 451 Finally, genetic clustering of female reproductive conditions revealed common genetic causes of 452 obesity on endometriosis, UF, and HMB, which are known to share mechanisms of development 453 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 30 (3, 97). The projection of infertility with these diseases merits following up on the genetic basis of 454 endometriosis-related infertility with an eye to prevention and treatment. 455 The main strength of our work is the systematic approach to characterising the relationship 456 between a broad range of obesity traits and common female reproductive conditions using both 457 observational and genetic approaches. All observational associations were estimated in the same 458 large-scale cohort study, which tends to lead to less biased estimates than case-control studies upon 459 which most previous results were based. Moreover, we conducted the first genetics-based 460 mediation analyses to pinpoint the mechanisms driving the causal effect of obesity on risk of 461 reproductive diseases. 462 Reproductive conditions remain underdiagnosed and underreported in the UK, which was reflected 463 in their low prevalence among female UKBB participants ( Table A in S1 Table ). This posed a 464

to our analyses in UK Biobank by reducing power to identify significant associations. 465 For this reason, we opted to use broad case categories, such as infertility of any cause, as we had 466 insufficient power and information to examine conditions by sub-types. Secondly, we restricted 467 our analyses to women of genetically European ancestry, due to a lack of genetic data on women 468 of other ancestries. Many of the reproductive diseases included here, with uterine fibroids being 469 the most notable example (98, 99), are more prevalent in non-European populations and our results 470 may not be transferable to women of other ancestries (100-102), which emphasises the urgent need 471 to set up large-scale studies similar to UK Biobank on participants of non-European ancestry. We 472 were further limited in investigations of metabolic, hormonal, and inflammatory mediating 473 mechanisms by a lack of publicly available GWAS summary statistics for these traits. Finally, the 474 lack of data on BMI and WHR prior to disease onset, and limited information on the age at which 475 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 31 reproductive conditions were first diagnosed, complicated the interpretation of our findings from 476 observational analyses in UK Biobank. 477 Key priorities for the future are the further exploration and validation of the pathways through 478 which obesity increases risk of female reproductive disease. Notably, our finding that insulin 479 resistance may be an important mediating mechanism warrants further attention, as cheap and safe 480 treatments are available to increase insulin sensitivity. This is demonstrated by the successful use 481 of metformin as a first-line treatment in women with PCOS (103), but such a treatment strategy 482 has not yet been explored for other reproductive conditions linked to obesity. More generally, 483 better and more detailed diagnostic information on reproductive health in large-scale cohort studies 484 is urgently required for future research on the causes, consequences and aetiological mechanisms 485 of female reproductive illnesses. 486 In conclusion, we provide genetic evidence that both generalised and central obesity play an 487 aetiological role in a broad range of female reproductive conditions, but the extent of this link 488 differs substantially between conditions. Our findings also highlight the importance of hormonal 489 pathways, notably leptin and insulin resistance, as mediating mechanisms and potential targets for 490 intervention in the treatment and prevention of common female reproductive conditions. 491

492 We acknowledge the participants and investigators of FinnGen for their contribution to this study. 493 This research has been conducted using the UK Biobank Resource under Application Number 494 10844. The views expressed are those of the author(s) and not necessarily those of the NHS, the 495 NIHR or the Department of Health. 496 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 32 Funding 497 The research was supported by the Wellcome Trust Core Award Grant Number 203141/Z/16/Z 498 with additional support from the NIHR Oxford BRC. 499 S.S.V. is supported by the Rhodes Trust (https://www.rhodeshouse.ox.ac.uk/), Clarendon Fund 500 (http://www.ox.ac.uk/clarendon/about), and the Medical Sciences Doctoral Training Centre 501 (https://www.medsci.ox.ac.uk/) at the University of Oxford. S.B. is supported by the Li Ka Shing 502 Foundation. M.V.H. works in a unit that receives funding from the UK Medical Research 503 Council and is supported by a British Heart Foundation Intermediate Clinical Research 504 Fellowship (FS/18/23/33512) and the National Institute for Health Research Oxford Biomedical 505 Research Centre. C.M.L. is supported by the Li Ka Shing Foundation, NIHR Oxford Biomedical 506 Research Centre, Oxford, NIH (1P50HD104224-01), Gates Foundation (INV-024200), and a 507 Wellcome Trust Investigator Award (221782/Z/20/Z). L.B.L.W. is supported by the Wellcome 508 Trust (221651/Z/20/Z). 509 Competing Interests 510 C.M.B. reports grants from Bayer AG, AbbVie Inc, Volition Rx, MDNA Life Sciences, Roche 511 Diagnostics Inc., and consultancy for Myovant. He is a member of the independent data monitoring 512 board at ObsEva; I.G. reports grants from Bayer AG; K.T.Z. reports grants from Bayer AG, 513 AbbVie Inc, Volition Rx, MDNA Life Sciences, Roche Diagnostics Inc, and non-financial 514 scientific collaboration with Population Diagnostics Ltd, outside the submitted work; M.V.H. has 515 consulted for Boehringer Ingelheim, and in adherence to the University of Oxford’s Clinical Trial 516 Service Unit & Epidemiological Studies Unit (CSTU) staff policy, did not accept personal 517 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 33 honoraria or other payments from pharmaceutical companies; C.M.L. reports grants from Bayer 518 AG and Novo Nordisk and has a partner who works at Vertex; no other relationships or activities 519 that could appear to have influenced the submitted work. 520 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 1

521 1. Wei S, Schmidt MD, Dwyer T, Norman RJ, Venn AJ. Obesity and menstrual irregularity: 522 associations with SHBG, testosterone, and insulin. Obesity (Silver Spring). 2009;17(5):1070-6. 523 2. Douchi T, Kuwahata R, Yamamoto S, Oki T, Yamasaki H, Nagata Y. Relationship of 524 upper body obesity to menstrual disorders. Acta Obstet Gynecol Scand. 2002;81(2):147-50. 525 3. Gallagher CS, Makinen N, Harris HR, Rahmioglu N, Uimari O, Cook JP, et al. Genome-526 wide association and epidemiological analyses reveal common genetic origins between uterine 527 leiomyomata and endometriosis. Nat Commun. 2019;10(1):4857. 528 4. Yi KW, Shin JH, Park MS, Kim T, Kim SH, Hur JY. Association of body mass index 529 with severity of endometriosis in Korean women. Int J Gynaecol Obstet. 2009;105(1):39-42. 530 5. Diamanti-Kandarakis E. Role of obesity and adiposity in polycystic ovary syndrome. Int 531 J Obes (Lond). 2007;31 Suppl 2:S8-13; discussion S31-2. 532 6. Glueck CJ, Goldenberg N. Characteristics of obesity in polycystic ovary syndrome: 533 Etiology, treatment, and genetics. Metabolism. 2019;92:108-20. 534 7. Spradley FT. Metabolic abnormalities and obesity's impact on the risk for developing 535 preeclampsia. Am J Physiol Regul Integr Comp Physiol. 2017;312(1):R5-R12. 536 8. Lashen H, Fear K, Sturdee DW. Obesity is associated with increased risk of first 537 trimester and recurrent miscarriage: matched case-control study. Hum Reprod. 538 2004;19(7):1644-6. 539 9. Metwally M, Saravelos SH, Ledger WL, Li TC. Body mass index and risk of miscarriage 540 in women with recurrent miscarriage. Fertil Steril. 2010;94(1):290-5. 541 10. van der Steeg JW, Steures P, Eijkemans MJ, Habbema JD, Hompes PG, Burggraaff JM, 542 et al. Obesity affects spontaneous pregnancy chances in subfertile, ovulatory women. Hum 543 Reprod. 2008;23(2):324-8. 544 11. Wise LA, Rothman KJ, Mikkelsen EM, Sorensen HT, Riis A, Hatch EE. An internet-545 based prospective study of body size and time-to-pregnancy. Hum Reprod. 2010;25(1):253-64. 546 12. Grodstein F, Goldman MB, Cramer DW. Body mass index and ovulatory infertility. 547 Epidemiology. 1994;5(2):247-50. 548 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 2 13. Missmer SA, Hankinson SE, Spiegelman D, Barbieri RL, Marshall LM, Hunter DJ. 549 Incidence of laparoscopically confirmed endometriosis by demographic, anthropometric, and 550 lifestyle factors. Am J Epidemiol. 2004;160(8):784-96. 551 14. Shah DK, Correia KF, Vitonis AF, Missmer SA. Body size and endometriosis: results 552 from 20 years of follow-up within the Nurses' Health Study II prospective cohort. Hum Reprod. 553 2013;28(7):1783-92. 554 15. Dixon SC, Nagle CM, Thrift AP, Pharoah PD, Pearce CL, Zheng W, et al. Adult body 555 mass index and risk of ovarian cancer by subtype: a Mendelian randomization study. Int J 556 Epidemiol. 2016;45(3):884-95. 557 16. Painter JN, O'Mara TA, Marquart L, Webb PM, Attia J, Medland SE, et al. Genetic Risk 558 Score Mendelian Randomization Shows that Obesity Measured as Body Mass Index, but not 559 Waist:Hip Ratio, Is Causal for Endometrial Cancer. Cancer Epidemiol Biomarkers Prev. 560 2016;25(11):1503-10. 561 17. Brower MA, Hai Y, Jones MR, Guo X, Chen YI, Rotter JI, et al. Bidirectional Mendelian 562 randomization to explore the causal relationships between body mass index and polycystic ovary 563 syndrome. Hum Reprod. 2019;34(1):127-36. 564 18. Mahutte NG, Matalliotakis IM, Goumenou AG, Vassiliadis S, Koumantakis GE, Arici A. 565 Inverse correlation between peritoneal fluid leptin concentrations and the extent of 566 endometriosis. Hum Reprod. 2003;18(6):1205-9. 567 19. Markowska A, Rucinski M, Drews K, Malendowicz LK. Further studies on leptin and 568 leptin receptor expression in myometrium and uterine myomas. Eur J Gynaecol Oncol. 569 2005;26(5):517-25. 570 20. Brannian JD, Schmidt SM, Kreger DO, Hansen KA. Baseline non-fasting serum leptin 571 concentration to body mass index ratio is predictive of IVF outcomes. Hum Reprod. 572 2001;16(9):1819-26. 573 21. Plowden TC, Zarek SM, Rafique S, Sjaarda LA, Schisterman EF, Silver RM, et al. 574 Preconception leptin levels and pregnancy outcomes: A prospective cohort study. Obes Sci Pract. 575 2020;6(2):181-8. 576 22. AlAshqar A, Patzkowsky K, Afrin S, Wild R, Taylor HS, Borahay MA. Cardiometabolic 577 Risk Factors and Benign Gynecologic Disorders. Obstet Gynecol Surv. 2019;74(11):661-73. 578 23. Hauth JC, Clifton RG, Roberts JM, Myatt L, Spong CY, Leveno KJ, et al. Maternal 579 insulin resistance and preeclampsia. Am J Obstet Gynecol. 2011;204(4):327 e1-6. 580 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 3 24. Butler MG, McGuire A, Manzardo AM. Clinically relevant known and candidate genes 581 for obesity and their overlap with human infertility and reproduction. J Assist Reprod Genet. 582 2015;32(4):495-508. 583 25. Sudlow C, Gallacher J, Allen N, Beral V, Burton P, Danesh J, et al. UK biobank: an open 584 access resource for identifying the causes of a wide range of complex diseases of middle and old 585 age. PLoS Med. 2015;12(3):e1001779. 586 26. Glickman ME, Rao SR, Schultz MR. False discovery rate control is a recommended 587 alternative to Bonferroni-type adjustments in health studies. J Clin Epidemiol. 2014;67(8):850-7. 588 27. Carreras-Torres R, Johansson M, Haycock PC, Relton CL, Davey Smith G, Brennan P, et 589 al. Role of obesity in smoking behaviour: Mendelian randomisation study in UK Biobank. BMJ. 590 2018;361:k1767. 591 28. Ambler GB, Axel. mfp: Multivariable Fractional Polynomials. In: Luecke S, editor. 1.5.2 592 ed: CRAN; 2015. 593 29. Wood S. mgcv: Mixed GAM Computation Vehicle with Automatic Smoothness 594 Estimation. In: Wood S, editor. 1.8-31 ed: CRAN; 2021. 595 30. H. A. Information Theory and an Extension of the Maximum Likelihood Principle. In: 596 Parzen E. TK, Kitagawa G., editor. Selected Papers of Hirotugu Akaike. Springer Series in 597 Statistics (Perspectives in Statistics). New York, NY: Springer; 1998. 598 31. Pulit SL, Stoneman C, Morris AP, Wood AR, Glastonbury CA, Tyrrell J, et al. Meta-599 analysis of genome-wide association studies for body fat distribution in 694 649 individuals of 600 European ancestry. Hum Mol Genet. 2019;28(1):166-74. 601 32. Karlsson T, Rask-Andersen M, Pan G, Hoglund J, Wadelius C, Ek WE, et al. 602 Contribution of genetics to visceral adiposity and its relation to cardiovascular and metabolic 603 disease. Nat Med. 2019;25(9):1390-5. 604 33. Elsworth B, Lyon M, Alexander T, Liu Y, Matthews P, Hallett J, et al. The MRC IEU 605 OpenGWAS data infrastructure. bioRxiv. 2020:2020.08.10.244293. 606 34. Lotta LA, Wittemans LBL, Zuber V, Stewart ID, Sharp SJ, Luan J, et al. Association of 607 Genetic Variants Related to Gluteofemoral vs Abdominal Fat Distribution With Type 2 Diabetes, 608 Coronary Disease, and Cardiovascular Risk Factors. JAMA. 2018;320(24):2553-63. 609 35. Munafo MR, Tilling K, Taylor AE, Evans DM, Davey Smith G. Collider scope: when 610 selection bias can substantially influence observed associations. Int J Epidemiol. 611 2018;47(1):226-35. 612 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 4 36. Pirastu N, Cordioli M, Nandakumar P, Mignogna G, Abdellaoui A, Hollis B, et al. 613 Genetic analyses identify widespread sex-differential participation bias. bioRxiv. 614 2021:2020.03.22.001453. 615 37. Zhou W, Nielsen JB, Fritsche LG, Dey R, Gabrielsen ME, Wolford BN, et al. Efficiently 616 controlling for case-control imbalance and sample relatedness in large-scale genetic association 617 studies. Nat Genet. 2018;50(9):1335-41. 618 38. Willer CJ, Li Y, Abecasis GR. METAL: fast and efficient meta-analysis of genomewide 619 association scans. Bioinformatics. 2010;26(17):2190-1. 620 39. Painter JN, Anderson CA, Nyholt DR, Macgregor S, Lin J, Lee SH, et al. Genome-wide 621 association study identifies a locus at 7p15.2 associated with endometriosis. Nat Genet. 622 2011;43(1):51-4. 623 40. Laisk T, Soares ALG, Ferreira T, Painter JN, Censin JC, Laber S, et al. The genetic 624 architecture of sporadic and multiple consecutive miscarriage. Nat Commun. 2020;11(1):5980. 625 41. Deng L, Zhang H, Yu K. Power calculation for the general two-sample Mendelian 626 randomization analysis. Genet Epidemiol. 2020;44(3):290-9. 627 42. Brion MJ, Shakhbazov K, Visscher PM. Calculating statistical power in Mendelian 628 randomization studies. Int J Epidemiol. 2013;42(5):1497-501. 629 43. Hemani G, Zheng J, Elsworth B, Wade KH, Haberland V, Baird D, et al. The MR-Base 630 platform supports systematic causal inference across the human phenome. Elife. 2018;7. 631 44. Bowden J, Del Greco MF, Minelli C, Davey Smith G, Sheehan N, Thompson J. A 632 framework for the investigation of pleiotropy in two-sample summary data Mendelian 633 randomization. Stat Med. 2017;36(11):1783-802. 634 45. Bowden J, Spiller W, Del Greco MF, Sheehan N, Thompson J, Minelli C, et al. 635 Improving the visualization, interpretation and analysis of two-sample summary data Mendelian 636 randomization via the Radial plot and Radial regression. Int J Epidemiol. 2018;47(4):1264-78. 637 46. Sapkota Y, Steinthorsdottir V, Morris AP, Fassbender A, Rahmioglu N, De Vivo I, et al. 638 Meta-analysis identifies five novel loci associated with endometriosis highlighting key genes 639 involved in hormone metabolism. Nat Commun. 2017;8:15539. 640 47. Day F, Karaderi T, Jones MR, Meun C, He C, Drong A, et al. Large-scale genome-wide 641 meta-analysis of polycystic ovary syndrome suggests shared genetic architecture for different 642 diagnosis criteria. PLoS Genet. 2018;14(12):e1007813. 643 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 5 48. Bycroft C, Freeman C, Petkova D, Band G, Elliott LT, Sharp K, et al. The UK Biobank 644 resource with deep phenotyping and genomic data. Nature. 2018;562(7726):203-9. 645 49. Staley JR, Burgess S. Semiparametric methods for estimation of a nonlinear exposure-646 outcome relationship using instrumental variables with application to Mendelian randomization. 647 Genet Epidemiol. 2017;41(4):341-52. 648 50. Carter AR, Sanderson E, Hammerton G, Richmond RC, Smith GD, Heron J, et al. 649 Mendelian randomisation for mediation analysis: current methods and challenges for 650 implementation. bioRxiv. 2020:835819. 651 51. Kilpelainen TO, Carli JF, Skowronski AA, Sun Q, Kriebel J, Feitosa MF, et al. Genome-652 wide meta-analysis uncovers novel loci influencing circulating leptin levels. Nat Commun. 653 2016;7:10494. 654 52. Manning AK, Hivert MF, Scott RA, Grimsby JL, Bouatia-Naji N, Chen H, et al. A 655 genome-wide approach accounting for body mass index identifies genetic variants influencing 656 fasting glycemic traits and insulin resistance. Nat Genet. 2012;44(6):659-69. 657 53. Walford GA, Gustafsson S, Rybin D, Stancakova A, Chen H, Liu CT, et al. Genome-658 Wide Association Study of the Modified Stumvoll Insulin Sensitivity Index Identifies BCL2 and 659 FAM19A2 as Novel Insulin Sensitivity Loci. Diabetes. 2016;65(10):3200-11. 660 54. Hemani G, Tilling K, Davey Smith G. Orienting the causal relationship between 661 imprecisely measured traits using GWAS summary data. PLoS Genet. 2017;13(11):e1007081. 662 55. Doob JL. The Limiting Distributions of Certain Statistics. The Annals of Mathematical 663 Statistics. 1935;6(3):160-9, 10. 664 56. Pers TH, Timshel P, Hirschhorn JN. SNPsnap: a Web-based tool for identification and 665 annotation of matched SNPs. Bioinformatics. 2015;31(3):418-20. 666 57. Foley CN, Mason AM, Kirk PDW, Burgess S. MR-Clust: clustering of genetic variants in 667 Mendelian randomization with similar causal estimates. Bioinformatics. 2021;37(4):531-41. 668 58. Burgess S, Thompson DJ, Rees JMB, Day FR, Perry JR, Ong KK. Dissecting Causal 669 Pathways Using Mendelian Randomization with Summarized Genetic Data: Application to Age 670 at Menarche and Risk of Breast Cancer. Genetics. 2017;207(2):481-7. 671 59. Relton CL, Davey Smith G. Two-step epigenetic Mendelian randomization: a strategy for 672 establishing the causal role of epigenetic processes in pathways to disease. Int J Epidemiol. 673 2012;41(1):161-76. 674 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 6 60. Perry JR, Corre T, Esko T, Chasman DI, Fischer K, Franceschini N, et al. A genome-675 wide association study of early menopause and the combined impact of identified variants. Hum 676 Mol Genet. 2013;22(7):1465-72. 677 61. Jones MR, Mathur R, Cui J, Guo X, Azziz R, Goodarzi MO. Independent confirmation of 678 association between metabolic phenotypes of polycystic ovary syndrome and variation in the 679 type 6 17beta-hydroxysteroid dehydrogenase gene. J Clin Endocrinol Metab. 2009;94(12):5034-680 8. 681 62. Ortega-Azorin C, Coltell O, Asensio EM, Sorli JV, Gonzalez JI, Portoles O, et al. 682 Candidate Gene and Genome-Wide Association Studies for Circulating Leptin Levels Reveal 683 Population and Sex-Specific Associations in High Cardiovascular Risk Mediterranean Subjects. 684 Nutrients. 2019;11(11). 685 63. Miranda-Lora AL, Cruz M, Aguirre-Hernandez J, Molina-Diaz M, Gutierrez J, Flores-686 Huerta S, et al. Exploring single nucleotide polymorphisms previously related to obesity and 687 metabolic traits in pediatric-onset type 2 diabetes. Acta Diabetol. 2017;54(7):653-62. 688 64. Nielsen TR, Appel EV, Svendstrup M, Ohrt JD, Dahl M, Fonvig CE, et al. A genome-689 wide association study of thyroid stimulating hormone and free thyroxine in Danish children and 690 adolescents. PLoS One. 2017;12(3):e0174204. 691 65. Ramly B, Afiqah-Aleng N, Mohamed-Hussein ZA. Protein-Protein Interaction Network 692 Analysis Reveals Several Diseases Highly Associated with Polycystic Ovarian Syndrome. Int J 693 Mol Sci. 2019;20(12). 694 66. Wu TI, Huang RL, Su PH, Mao SP, Wu CH, Lai HC. Ovarian cancer detection by DNA 695 methylation in cervical scrapings. Clin Epigenetics. 2019;11(1):166. 696 67. Makabe T, Arai E, Hirano T, Ito N, Fukamachi Y, Takahashi Y, et al. Genome-wide 697 DNA methylation profile of early-onset endometrial cancer: its correlation with genetic 698 aberrations and comparison with late-onset endometrial cancer. Carcinogenesis. 699 2019;40(5):611-23. 700 68. Barber TM, Franks S. Obesity and polycystic ovary syndrome. Clin Endocrinol (Oxf). 701 2021. 702 69. Teede HJ, Joham AE, Paul E, Moran LJ, Loxton D, Jolley D, et al. Longitudinal weight 703 gain in women identified with polycystic ovary syndrome: results of an observational study in 704 young women. Obesity (Silver Spring). 2013;21(8):1526-32. 705 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 7 70. Norgaard M, Ehrenstein V, Vandenbroucke JP. Confounding in observational studies 706 based on large health care databases: problems and potential solutions - a primer for the 707 clinician. Clin Epidemiol. 2017;9:185-93. 708 71. VanderWeele TJ, Hernan MA, Robins JM. Causal directed acyclic graphs and the 709 direction of unmeasured confounding bias. Epidemiology. 2008;19(5):720-8. 710 72. Ezeh U, Yildiz BO, Azziz R. Referral bias in defining the phenotype and prevalence of 711 obesity in polycystic ovary syndrome. J Clin Endocrinol Metab. 2013;98(6):E1088-96. 712 73. Luque-Ramirez M, Alpanes M, Sanchon R, Fernandez-Duran E, Ortiz-Flores AE, 713 Escobar-Morreale HF. Referral bias in female functional hyperandrogenism and polycystic ovary 714 syndrome. Eur J Endocrinol. 2015;173(5):603-10. 715 74. Karvonen-Gutierrez C, Kim C. Association of Mid-Life Changes in Body Size, Body 716 Composition and Obesity Status with the Menopausal Transition. Healthcare (Basel). 2016;4(3). 717 75. McCann SE, Freudenheim JL, Darrow SL, Batt RE, Zielezny MA. Endometriosis and 718 body fat distribution. Obstet Gynecol. 1993;82(4 Pt 1):545-9. 719 76. Sun K, Xie Y, Zhao N, Li Z. A case-control study of the relationship between visceral fat 720 and development of uterine fibroids. Exp Ther Med. 2019;18(1):404-10. 721 77. Sato F, Nishi M, Kudo R, Miyake H. Body fat distribution and uterine leiomyomas. J 722 Epidemiol. 1998;8(3):176-80. 723 78. Lord J, Thomas R, Fox B, Acharya U, Wilkin T. The central issue? Visceral fat mass is a 724 good marker of insulin resistance and metabolic disturbance in women with polycystic ovary 725 syndrome. BJOG. 2006;113(10):1203-9. 726 79. Ferrero S, Anserini P, Remorgida V, Ragni N. Body mass index in endometriosis. Eur J 727 Obstet Gynecol Reprod Biol. 2005;121(1):94-8. 728 80. Hediger ML, Hartnett HJ, Louis GM. Association of endometriosis with body size and 729 figure. Fertil Steril. 2005;84(5):1366-74. 730 81. Berlanda N, Somigliana E, Frattaruolo MP, Buggio L, Dridi D, Vercellini P. Surgery 731 versus hormonal therapy for deep endometriosis: is it a choice of the physician? Eur J Obstet 732 Gynecol Reprod Biol. 2017;209:67-71. 733 82. Kim SA, Um MJ, Kim HK, Kim SJ, Moon SJ, Jung H. Study of dienogest for 734 dysmenorrhea and pelvic pain associated with endometriosis. Obstet Gynecol Sci. 735 2016;59(6):506-11. 736 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 8 83. Jeng CJ, Chuang L, Shen J. A comparison of progestogens or oral contraceptives and 737 gonadotropin-releasing hormone agonists for the treatment of endometriosis: a systematic 738 review. Expert Opin Pharmacother. 2014;15(6):767-73. 739 84. Okifuji A, Hare BD. The association between chronic pain and obesity. J Pain Res. 740 2015;8:399-408. 741 85. Engstrom G, Hedblad B, Stavenow L, Lind P, Janzon L, Lindgarde F. Inflammation-742 sensitive plasma proteins are associated with future weight gain. Diabetes. 2003;52(8):2097-101. 743 86. Yasui T, Hayashi K, Mizunuma H, Kubota T, Aso T, Matsumura Y, et al. Association of 744 endometriosis-related infertility with age at menopause. Maturitas. 2011;69(3):279-83. 745 87. Sirmans SM, Pate KA. Epidemiology, diagnosis, and management of polycystic ovary 746 syndrome. Clin Epidemiol. 2013;6:1-13. 747 88. Laven JS, Imani B, Eijkemans MJ, Fauser BC. New approach to polycystic ovary 748 syndrome and other forms of anovulatory infertility. Obstet Gynecol Surv. 2002;57(11):755-67. 749 89. Mardh PA. Tubal factor infertility, with special regard to chlamydial salpingitis. Curr 750 Opin Infect Dis. 2004;17(1):49-52. 751 90. de Ziegler D, Borghese B, Chapron C. Endometriosis and infertility: pathophysiology and 752 management. Lancet. 2010;376(9742):730-8. 753 91. Homer HA. The Role of Oocyte Quality in Explaining "Unexplained" Infertility. Semin 754 Reprod Med. 2020;38(1):21-8. 755 92. Arojoki M, Jokimaa V, Juuti A, Koskinen P, Irjala K, Anttila L. Hypothyroidism among 756 infertile women in Finland. Gynecol Endocrinol. 2000;14(2):127-31. 757 93. Guerin LR, Prins JR, Robertson SA. Regulatory T-cells and immune tolerance in 758 pregnancy: a new target for infertility treatment? Hum Reprod Update. 2009;15(5):517-35. 759 94. Luciano AA, Lanzone A, Goverde AJ. Management of female infertility from hormonal 760 causes. Int J Gynaecol Obstet. 2013;123 Suppl 2:S9-17. 761 95. Sen A, Kushnir VA, Barad DH, Gleicher N. Endocrine autoimmune diseases and female 762 infertility. Nat Rev Endocrinol. 2014;10(1):37-50. 763 96. Ramlau-Hansen CH, Thulstrup AM, Nohr EA, Bonde JP, Sorensen TI, Olsen J. 764 Subfecundity in overweight and obese couples. Hum Reprod. 2007;22(6):1634-7. 765 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 9 97. Nilufer R, Karina B, Paraskevi C, Rebecca D, Genevieve G, Ayush G, et al. Large-scale 766 genome-wide association meta-analysis of endometriosis reveals 13 novel loci and genetically-767 associated comorbidity with other pain conditions. bioRxiv. 2018:406967. 768 98. Catherino WH, Eltoukhi HM, Al-Hendy A. Racial and ethnic differences in the 769 pathogenesis and clinical manifestations of uterine leiomyoma. Semin Reprod Med. 770 2013;31(5):370-9. 771 99. Marsh EE, Ekpo GE, Cardozo ER, Brocks M, Dune T, Cohen LS. Racial differences in 772 fibroid prevalence and ultrasound findings in asymptomatic young women (18-30 years old): a 773 pilot study. Fertil Steril. 2013;99(7):1951-7. 774 100. Bougie O, Healey J, Singh SS. Behind the times: revisiting endometriosis and race. Am J 775 Obstet Gynecol. 2019;221(1):35 e1- e5. 776 101. Kim JJ, Choi YM. Phenotype and genotype of polycystic ovary syndrome in Asia: Ethnic 777 differences. J Obstet Gynaecol Res. 2019;45(12):2330-7. 778 102. Nakimuli A, Chazara O, Byamugisha J, Elliott AM, Kaleebu P, Mirembe F, et al. 779 Pregnancy, parturition and preeclampsia in women of African ancestry. Am J Obstet Gynecol. 780 2014;210(6):510-20 e1. 781 103. Tan X, Li S, Chang Y, Fang C, Liu H, Zhang X, et al. Effect of metformin treatment 782 during pregnancy on women with PCOS: a systematic review and meta-analysis. Clin Invest 783 Med. 2016;39(4):E120-31. 784 785 786 787 788 789 790 791 792 793 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint 10 Supporting Information Captions 794 S1 Figure. Comparison of different Mendelian randomisation methods for regression of 795 female reproductive diseases on various obesity traits. Summary statistics MR was performed 796 with the R package TwoSampleMR using three different methods whose results are compared. 797 Effect sizes are displayed as odds ratios (ORs) with 95% confidence intervals (CIs). P-values are 798 adjusted for multiple testing with the false discovery rate (FDR) correction; solid lines indicate 799 associations that are significant at an adjusted p-value threshold of 0.05. BMI = body mass index, 800 IVW = Inverse-variance weighted, PCOS = polycystic ovary syndrome, WHR = waist-hip ratio, 801 WHRadjBMI = WHR adjusted for BMI, VAT = genetically predicted visceral adipose tissue 802 mass. 803 S2 Figure. Mendelian randomisation sensitivity analyses for female reproductive disorders 804 regressed on obesity traits. (A) Female-specific genetic instruments vs combined-sexes genetic 805 instruments. (B) Reproductive outcomes from meta-analysis of UKBB and FinnGen summary 806 statistics vs those from FinnGen only. Summary statistics MR performed with TwoSampleMR R 807 package and best method (displayed, IVW) chosen via Rucker’s framework. Odds ratios (ORs) 808 with 95% confidence intervals (CIs) displayed. 809 S3 Figure. (A) Non-linear Mendelian randomisation (MR) estimates for relationships 810 between BMI and female reproductive disorders. Localised average causal estimates (LACE) 811 are calculated by dividing the IV-free exposure into 100 quantiles (fractional polynomial 812 method) or 10 quantiles (piecewise linear method) with the odds ratio (OR) and 95% confidence 813 intervals (CI) displayed. The reference point for OR = 1 is mean BMI, 27.0 kg/m2. (B) 814 Heterogeneity across BMI spectrum in instrument variables used for non-linear MR. 815 Instrument variable (IV)-free BMI was divided into quantiles as described in (A). Left: fractional 816 polynomial 100 quantiles, right: piecewise linear 10 quantiles. Proportion of variance in BMI 817 explained by instrument SNPs in each quantile is plotted (β), with 95% confidence intervals 818 shown as standard error bars. Spont. misc. = spontaneous miscarriage. 819 S4 Figure. Single-SNP genetic effect estimates for obesity instruments on female 820 reproductive disorders. SNPs are annotated with their nearest gene by SNPsnap and clustered 821 by obesity-female reproductive disorder relationship. SNPs with >= 80% probability of 822 belonging to a substantial cluster (MRClust) are displayed, with an * if the SNP belongs to 823 substantial clusters for multiple disorders. Effect size estimates are scaled to a variance of 1 824 within each disease. 825 S1 Table. Supplemental Tables A-P. 826 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 1, 2021. ; https://doi.org/10.1101/2021.06.01.21257781doi: medRxiv preprint