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

Background: Obesity is observationally associated with altered risk of many female reproductive conditions. These include polycystic ovary syndrome (PCOS), abnormal uterine bleeding, endometriosis, infertility, and pregnancy-related disorders. However, the roles and mechanisms of obesity in the aetiology of reproductive disorders remain unclear. Methods and Findings: We estimated observational and genetically predicted causal associations between obesity, metabolic hormones, and female reproductive conditions using logistic regression, generalised additive models, and Mendelian randomisation (two-sample, non-linear, and multivariable) applied to data from UK Biobank and publicly available genome-wide association studies (GWAS). Body mass index (BMI), waist-hip ratio (WHR), and WHR adjusted for BMI (WHRadjBMI) were observationally (odds ratios (ORs) = 1.02 - 1.87 per 1 S.D. obesity trait) and causally (ORs = 1.06 - 2.09) associated with uterine fibroids (UF), PCOS, heavy menstrual bleeding (HMB), and pre-eclampsia. Causal effect estimates of WHR and WHRadjBMI, but not BMI, were attenuated compared to their observational counterparts. Genetically predicted visceral adipose tissue mass was causal for the development of HMB, PCOS, and pre-eclampsia (ORs = 1.01 - 3.38). Increased waist circumference also posed a higher causal risk (ORs = 1.16 - 1.93) for the development of these disorders and UF than did increased hip circumference (ORs = 1.06 - 1.10). Leptin, fasting insulin, and insulin resistance each mediated between 20% - 50% of the total causal effect of obesity on pre-eclampsia. Reproductive conditions clustered based on shared genetic components of their aetiological relationships with obesity. Conclusions: In this first systematic, large-scale, genetics-based analysis of the aetiological relationships between obesity and female reproductive conditions, we found that common indices of overall and central obesity increased risk of reproductive disorders to heterogenous extents, mediated by metabolic hormones. Our results suggest exploring the mechanisms mediating the causal effects of overweight and obesity on gynaecological health to identify targets for disease prevention and treatment.

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  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 (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 license It 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/10.1101/2021.06.01.21257781 doi: 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).

128
. CC-BY 4.0 International license It 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/10.1101/2021.06.01.21257781 doi: 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)    CC-BY 4.0 International license It 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. 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  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     Table D in S1 Table). All three obesity traits displayed U-shaped relationships with 255 PCOS.

256
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

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 license It 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/10.1101/2021.06.01.21257781 doi: medRxiv preprint 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.  is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

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 < 0.03.

289
SNPs identified in female-only GWAS and with female-specific weights for BMI, WHR, and 290 WHRadjBMI (29) were found to be the strongest instruments, with F-statistics > 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 license It 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.  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 license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

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 license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

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 license It 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/10.1101/2021.06.01.21257781 doi: medRxiv preprint 20 . CC-BY 4.0 International license It 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.    . CC-BY 4.0 International license It 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/10.1101/2021.06.01.21257781 doi: medRxiv preprint 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  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  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 license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

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

362
. CC-BY 4.0 International license It 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/10.1101/2021.06.01.21257781 doi: medRxiv preprint 25 . CC-BY 4.0 International license It 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. 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  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/10.1101/2021.06.01.21257781 doi: 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 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 license It 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.  CC-BY 4.0 International license It 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/10.1101/2021.06.01.21257781 doi: medRxiv preprint (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 limitation 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  CC-BY 4.0 International license It 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.

580
. CC-BY 4.0 International license It 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.

612
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

643
. CC-BY 4.0 International license It 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/10.1101/2021.06.01.21257781 doi: medRxiv preprint