Blood Pressure, Cardiometabolic Traits, and Cardiovascular Events in Women With Uterine Fibroids: A Genetic Correlation and Mendelian Randomization Study

Abstract

Background

Uterine fibroids (UFs) are understudied uterus neoplasms, mainly affecting women of reproductive age and often leading to hysterectomy. Clinical series suggest impaired cardiometabolic features in UFs. We investigated potential genetic links between blood pressure (BP), several cardiometabolic traits, and UFs.

Methods and Results

We used summary statistics of genome‐wide association studies for UFs and 18 traits related to BP, obesity, lipids, and main vascular diseases. We applied linkage disequilibrium score regression to estimate genetic correlations and Genome‐Wide Complex Trait Analysis‐multitrait‐based conditional and joint analysis to perform adjusted correlations. Univariate and bidirectional Mendelian randomization verified potential causal associations with UFs. We found UFs to significantly correlate with systolic BP (genetic correlation coefficient [r_g]=0.08, P=8.7×10⁻⁵) and diastolic BP (r_g=0.12, P=8.2×10⁻⁸), including after adjustment for body mass index. UFs also positively corelated with body mass index (r_g=0.11, P=4.1×10⁻⁴), waist‐to‐hip ratio (r_g=0.09, P=7.3×10⁻³), type 2 diabetes (r_g=0.15, P=1.9×10⁻⁵), and triglycerides (r_g=0.17, P=7.6×10⁻⁷). We identified a negative correlation with sex hormone‐binding globulin (r_g=−0.16, P=3×10⁻⁴), a marker of bioavailability of sex steroids. No evidence for shared genetic basis with vascular diseases was observed, except with migraine (r_g=0.08, P=5.8×10⁻⁷). Mendelian randomization analyses confirmed higher body mass index to increase UF risk (beta‐per‐kg/m²=0.033, P=6.1×10⁻⁵), as did waist‐to‐hip ratio (beta‐per‐unit=0.193, P=3.3×10⁻⁵) and triglycerides (bet‐per‐mmol/L=0.163, P=1.9×10⁻⁵). Higher sex hormone‐binding globulin decreased UF risk (beta‐per‐nmol/L=0.005, P=2.5×10⁻³). No causal effect was found for BP.

Conclusions

Our study shows that UFs share substantial genetic basis with traits related to BP, obesity, diabetes, and migraine, a predominantly female vascular disease. We provide Mendelian randomization‐based evidence for central obesity, visceral fat traits, and sex‐steroid bioavailability as relevant risk factors for UFs.

Nonstandard Abbreviations and Acronyms

IV

instrumental variable

IVW

inverse‐variance weighted

LD

linkage disequilibrium

LDSC

Linkage disequilibrium score regression

MR

Mendelian randomization

mtCOJO

multitrait‐based conditional and joint analysis

PPA

posterior probability of association

SHBG

sex hormone‐binding globulin

T2D

type 2 diabetes

UF

uterine fibroids

WHR

waist‐to‐hip ratio adjusted on body mass index

Research Perspective.

What Is New?

• We show that uterine fibroids, benign uterus tumors affecting reproductive organs outside pregnancy, share a significant portion of their genetic background with cardiometabolic traits, including blood pressure, obesity, diabetes, and migraine.

• Mendelian randomization findings do not support a causal link between genetic predisposition to increased blood pressure and UFs, but suggest central obesity‐related traits, such as waist‐to‐hip ratio, triglycerides, and sex hormone‐binding globulin levels, as genetic risk factors for uterine fibroids.

What Question Should Be Addressed Next?

• Our results set the stage for future studies aiming to explore the biological pathways linking uterine fibroids to cardiometabolic health and assess potential intervention strategies; understanding the genetic and metabolic determinants of uterine fibroids could improve risk prediction and inform personalized health care approaches for affected women.

Uterine fibroids (UFs), also known as uterine leiomyomata, are benign smooth muscle cell tumors of the uterus, affecting mainly women of reproductive age. ¹ Although they are mostly asymptomatic, UFs cause a large range of gynecological symptoms including abnormal uterine bleeding and issues with fertility, and less frequently, constipation and dyspareunia. ² The prevalence of UFs ranges from 4.5% to 68.6% depending on the studies. ³ UFs accounts for up to 60% of hysterectomies, 30% among women <44 years. ⁴ Recently, techniques requiring less invasive surgery, such as myomectomy and uterine arteries embolization, are becoming more popular in the treatment of UFs. ²

Several factors were reported to influence the risk of developing UFs. Increasing parity is reported to be a protective factor for UFs, whereas early menarche and the use of oral contraceptives are suspected risk factors. ⁵ Other risk factors include increasing premenopausal age, estrogen and testosterone levels, and unhealthy diet and lifestyle. ⁶ The precise mechanisms behind the association with ancestry, with relatively low occurrence in women of European ancestry, intermediate in Asian, and higher occurrence in women of African ancestry, are currently being explored. ⁷ , ⁸

Molecular and genetic explorations of UFs smooth muscle cell tumors indicate the role of driver mutations and common genetic variants in recurrent molecular pathways. The most prevalent somatic mutations were reported in the RNA Polymerase II mediator subunit gene (MED12). ⁹ , ¹⁰ In lower frequencies, mutations were also reported in the high mobility group AT‐hook 2 (HMGA2) group, among others. ¹¹ To date, genome‐wide association studies (GWAS) identified several common genetic variants to associate to the risk of UFs. ¹² Interestingly, some of the risk variants locate nearby MED12 and HMGA1, belonging to the same gene family as HMGA2 previously involved in fibroid growth via upregulation of proto‐oncogenes. ¹³

A comprehensive investigation of the potential genetic links between UFs and their cardiometabolic risk factors is missing. UFs have been clinically associated with higher body mass index (BMI) and blood pressure (BP), ⁸ , ¹⁴ suggesting a potential role of cardiometabolic risk factors in their pathogenesis. Nonetheless, the role of genetic determinants of major cardiovascular risk factors, including hypertension, has never been explored so far for the risk of UFs. Our study aims to explore the genetic links between a large panel of cardiometabolic traits and diseases and the risk for UFs through the application of genetic epidemiology methods. Through linkage disequilibrium score regression (LDSC) we estimated genetic correlations with UFs of BP, lipids and metabolic traits and several major cardiovascular events. We also applied Mendelian randomization (MR)‐based methods to examine the potential causal associations of cardiovascular traits and diseases with UFs.

METHODS

All genetic association data used in this work are publicly accessible. Appropriate patient consent and ethical approval had been obtained in the original studies from which they were obtained (Table S1).

Access to Publicly Available GWAS Data

The GWAS summary statistics used in this study were obtained from several publicly available resources, where institutional review board approvals and ethics clearance were provided in the original articles describing these data for the first time (Table S1). Data for UFs were obtained from Gallagher et al., ¹² a large meta‐analysis containing approximately 35 474 cases and 267 505 controls from 4 population‐based studies and 1 direct‐to‐consumer cohort. Summary statistics for the 18 cardiometabolic traits were obtained from several published GWAS and the UK Biobank data sets available on http://www.nealelab.is/uk‐biobank. For each summary statistics, a quality control was applied and only variants fulfilling the following criteria were included: minor allele frequency >0.01, info score (R²) >0.9, Hardy–Weinberg equilibrium test P value >10⁻⁶, and biallelic variants.

Single‐Nucleotide Variant‐Based Genetic Correlation Analyses

We used the LDSC method (v1.0.1, https://github.com/bulik/ldsc) to assess genetic correlations with cardiometabolic traits and diseases, adhering to the standard algorithm recommended by the developers. ¹⁵ , ¹⁶ LDSC is a method used to estimate the genetic traits using GWAS summary statistics. ¹⁵ , ¹⁶ The regression of the GWAS test statistics on the LD scores measures the amount of genetic variation tagged by each single‐nucleotide variant (SNV). To estimate the genetic correlation, the heritability for each trait is estimated first, where the intercept of the LD score regression plays a role in controlling for potential biases due to population stratification and cryptic relatedness. The intercept of the bivariate regression helps protect against sample overlap between 2 GWAS data sets. In our analyses, we carefully evaluated the potential for sample overlap among the GWAS summary statistics used. The UF GWAS summary statistics from Gallagher et al. included participants from studies such as the WGHS (Women's Genome Health Study), BioVU, NFBC (Northern Finland Birth Cohort), and 23andMe, which are independent of the cohorts used in the GWAS summary statistics for other traits. However, some GWAS data sets for other traits included data from the UK Biobank, which could overlap with the UF GWAS data. For traits where sample overlap was possible due to the inclusion of UK Biobank individuals or overlapping cohorts within large consortia (eg, BMI, waist‐to‐hip ratio [WHR], high‐density lipoprotein, low‐density lipoprotein, triglycerides, number of cigarettes per day, smoking initiation, sex‐hormone‐binding globulin [SHBG], and type 2 diabetes [T2D]), we did not constrain the LDSC intercept to account for potential sample overlap. This approach allows the LDSC method to correct for biases introduced by overlapping samples through the intercept adjustment. Conversely, for traits where we were confident that there was no sample overlap with the UF GWAS (systolic BP [SBP], diastolic BP [DBP], coronary artery disease, myocardial infarction, spontaneous coronary artery dissection, fibromuscular dysplasia, migraine, stroke, and intracranial aneurysm), we constrained the LDSC intercept. This constraining assumes no sample overlap and provides more precise estimates of genetic correlation.

Summary statistics for each trait were filtered to include only HapMap III SNVs panel, aiming to minimize bias from suboptimal imputation quality. Our analysis was limited to summary statistics from European ancestry GWAS, using files of LD score in Europeans derived from the 1000 Genomes reference panel, as provided by the developers. To address potential sample overlap from the comprehensive UK Biobank data set and prevent inflated false‐positive rates, we did not constrain the LD score regression intercept in our estimates. Given the extensive range of traits examined, we applied the Bonferroni correction to account for multiple testing across 18 clinically independent traits, setting a strict significance threshold of P<0.0027 (0.05/18). This is a conservative approach that controls the family‐wise error rate, thus reducing the likelihood of false positives but also potentially increasing the risk of missing true associations. Results with P values between 0.05 and the Bonferroni‐adjusted threshold of 0.0027 are considered “suggestive” and warrant further confirmation. In this context, the use of the term significant refers to the statistical significance complying to the Bonferroni threshold. We also implemented the multitrait‐based conditional and joint analysis (mtCOJO) tool from the Genome‐Wide Complex Trait Analysis pipeline ¹⁷ to adjust GWAS summary statistics for UFs by factoring in SBP and BMI. These adjusted summary statistics were then employed to reassess the genetic correlations between UFs, conditioned on SBP or BMI, and other traits of interest. This step served as a sensitivity analysis to evaluate the impact of SBP and BMI on the genetic correlation of UFs with these traits.

Mendelian Randomization

We applied univariate and bidirectional MR settings as summarized in Figure S1 using MendelianRandomization (v0.6.0) and TwoSampleMR (v0.5.6) R (v4.0.4) packages, respectively. To ensure the robustness of our MR analyses, we implemented a meticulous selection process for instrumental variables, adhering to the 3 core MR assumptions: (1) a strong association with the exposure, (2) independence from confounders affecting both the exposure and outcome, and (3) an effect on the outcome solely through the exposure. This was achieved by employing LD clumping with stringent criteria (P value threshold <5×10⁻⁸, LD r² <0.001 within a 10 000 kb window) based on the European population data from the 1000 Genomes Project.

For the estimation of associations between genetically predicted risk factors and outcomes, we used the multiplicative random‐effects inverse variance‐weighted (IVW) method. The multiplicative random‐effects model was preferred as it accounts for potential heterogeneity among the genetic instruments, providing valid causal estimates in the absence of evidence of pleiotropy in MR analyses, ¹⁸ with estimates scaled to reflect clinically meaningful changes. This approach is recommended unless there are few variants or all variants are from the same gene region, where heterogeneity cannot be estimated reliably. This method was used as it is considered the main 2‐sample MR method, with estimates scaled to reflect clinically meaningful changes. ¹⁹

Sensitivity analyses, including weighted median estimation, MR‐Egger regression, and MR‐PRESSO, ¹⁸ , ²⁰ , ²¹ were conducted to assess the consistency of our findings under different assumptions regarding genetic pleiotropy. MR‐Egger regression tests for directional pleiotropy, where the genetic variants have direct effects on the outcome that are not mediated through the exposure. The weighted median estimation provides a robust causal estimate even if up to 50% of the instruments are invalid. MR‐PRESSO detects and corrects for horizontal pleiotropy by identifying outlier genetic variants. Furthermore, heterogeneity among the estimates was evaluated using Cochran's Q test, ensuring the homogeneity of instrumental variable effects across different variants.

To correct for multiple testing, given our assessment of 9 risk factors, we applied a Bonferroni correction, setting a stringent significance threshold of P<0.0055 (0.05/9) for cardiovascular disease traits and P<0.006 (0.05/8) for bidirectional MR involving 8 diseases. Results yielding P values between 0.0055 and 0.05 were considered suggestively significant, indicating trends that may warrant further investigation.

Overlap Between UFs Top Associated Loci With Cardiovascular Traits and Diseases

We implemented the gwas‐pw method that allows the identification of variants or regions that influence pairs of traits. ²² After applying standard quality control to the cardiometabolic trait summary statistics (minor allele frequency >1%, autosomal SNVs only), we used the recommended European bed file ²³ to divide the genome into 1703 approximately independent blocks. The results are shown as posterior probabilities of association (PPA) based on 4 models: PPA1, the region is associated only with phenotype 1; PPA2, the region is associated only with phenotype 2; PPA3, there is a shared association with both phenotypes; and PPA4, there are 2 distinct associations in the regions, 1 to each phenotype.

RESULTS

Uterine Fibroids Share a Common Genetic Basis With Blood Pressure and Several Cardiometabolic Traits

We first explored the genetic correlations between UFs and a set of traits related to BP, body weight, and shape using LDSC ¹⁵ (Figure 1A, Table 1). We found UFs to correlate significantly with SBP (r_g=0.08, P=8.7×10⁻⁵), DBP (r_g=0.12, P=8.2×10⁻⁸), and BMI (r_g=0.11, P=4.1×10⁻⁴). Although we observed a positive genetic correlation between UFs and WHR (r_g=0.09, P=7.3×10⁻³) this result does not meet the Bonferroni‐corrected threshold for multiple testing. SBP and DBP correlations remained unchanged after adjustment on the genetics of BMI using the Genome‐Wide Complex Trait Analysis mtCOJO method (r_gSBPadjBMI=0.09, P=4.2×10⁻⁵ and r_gDBPadjBMI=0.13, P=4.7×10⁻⁸ respectively, Figure 1B and 1C, Table 1). The genetic correlation between UFs and BMI also remained significant when we adjusted for SBP genetic results (r_gBMIadjSBP=0.10, P=2.5×10⁻³).

Figure 1. Forest plots of the genetic correlations between uterine fibroids and cardiometabolic traits.

Rho coefficient of genetic correlation (r_g) is represented on the x axis and range represents the 95% CI. P value of correlation for 18 tests adjusted using Bonferroni method is indicated as: *P.adj <0.05. A, Genetic correlations between uterine fibroids and cardiometabolic traits unadjusted (in black), adjusted for systolic blood pressure (in blue) and adjusted for body mass index (in yellow). B, Genetic correlations between uterine fibroids and cardiometabolic diseases unadjusted (in black), adjusted for systolic blood pressure (in blue) and adjusted for body mass index (in yellow). AS, any stroke; BMI, body mass index; CAD, coronary artery disease; CPD, cigarette‐per‐day; DBP, diastolic blood pressure; FMD, fibromuscular dysplasia; HDL, high‐density lipoprotein; IA, intracranial aneurysm; LDL, low‐density lipoprotein; MI, myocardial infraction; MIG, migraine; SBP, systolic blood pressure; SCAD, spontaneous coronary artery dissection; SHBG, sex hormone‐binding globulin; SI, smoking initiation; T2D, type 2 diabetes; TG, triglycerides; and WHR, waist‐to‐hip ratio adjusted on BMI.

Table 1.

Genetic Correlation Between Uterine Fibroids and Cardiometabolic Traits and Diseases

Trait or disease	Unadjusted						Adjusted for SBP						Adjusted for BMI
	rg	SE	L95	U95	Z score	P value	rg	SE	L95	U95	Z score	P value	rg	SE	L95	U95	Z score	P value
Systolic blood pressure	0.08	0.02	0.04	0.13	3.92	8.7×10−05	…	…	…	…	…	…	0.09	0.02	0.05	0.13	4.10	4.2×10−05
Diastolic blood pressure	0.12	0.02	0.08	0.16	5.36	8.2×10−08	…	…	…	…	…	…	0.13	0.02	0.08	0.17	5.46	4.7×10−08
Body mass index	0.11	0.03	0.05	0.17	3.53	4.1×10−04	0.10	0.03	0.03	0.16	3.03	2.5×10−03	…	…	…	…	…	…
Waist‐to‐hip ratio (adjusted on BMI)	0.09	0.03	0.02	0.15	2.68	7.3×10−03	0.07	0.04	0.00	0.13	1.84	6.6×10−02	0.08	0.03	0.02	0.14	2.57	1.0×10−02
Type 2 diabetes	0.15	0.04	0.08	0.22	4.27	1.9×10−05	0.12	0.04	0.05	0.19	3.22	1.3×10−03	0.10	0.04	0.03	0.17	2.78	5.4×10−03
High‐density lipoprotein	−0.16	0.03	−0.22	−0.09	−4.73	2.3×10−06	−0.18	0.03	−0.24	−0.11	−5.19	2.1×10−07	−0.15	0.03	−0.21	−0.08	−4.47	7.7×10−06
Low‐density lipoprotein	0.01	0.04	−0.08	0.09	0.12	9.1×10−01	0.01	0.04	−0.07	0.10	0.28	7.8×10−01	0.01	0.05	−0.08	0.11	0.28	7.8×10−01
Triglycerides	0.17	0.04	0.11	0.24	4.95	7.6×10−07	0.16	0.04	0.09	0.23	4.45	8.8×10−06	0.15	0.04	0.07	0.22	3.92	8.8×10−05
Cigarettes per day	0.03	0.05	−0.05	0.12	0.77	4.4×10−01	0.04	0.05	−0.05	0.14	0.88	3.8×10−01	0.01	0.05	−0.09	0.10	0.15	8.8×10−01
Smoking initiation	0.04	0.03	−0.02	0.10	1.16	2.5×10−01	0.04	0.03	−0.02	0.11	1.27	2.1×10−01	0.04	0.03	−0.03	0.10	1.13	2.6×10−01
Sex hormone‐binding globulin	−0.16	0.04	−0.24	−0.07	−3.63	3.0×10−04	−0.16	0.04	−0.25	−0.07	−3.55	4.0×10−04	−0.14	0.04	−0.23	−0.06	−3.34	8.0×10−04
Coronary artery disease	0.07	0.03	0.01	0.12	2.31	2.1×10−02	0.05	0.03	−0.01	0.11	1.66	9.7×10−02	0.05	0.03	−0.01	0.11	1.52	1.3×10−01
Myocardial infraction	0.08	0.03	0.02	0.14	2.49	1.3×10−02	0.06	0.03	0.00	0.12	1.99	4.6×10−02	0.06	0.03	0.00	0.12	1.83	6.8×10−02
Any stroke	0.03	0.05	−0.07	0.14	0.66	5.1×10−01	0.03	0.05	−0.07	0.13	0.52	6.0×10−01	0.03	0.05	−0.07	0.13	0.56	5.8×10−01
Intracranial aneurysm	0.09	0.05	0.00	0.18	1.94	5.2×10−02	0.08	0.05	−0.01	0.17	1.64	1.0×10−01	0.08	0.05	−0.01	0.17	1.76	7.8×10−02
Spontaneous coronary artery dissection	0.03	0.05	−0.07	0.13	0.50	6.2×10−01	0.01	0.09	−0.17	0.19	0.10	9.2×10−01	0.02	0.06	−0.09	0.13	0.38	7.0×10−01
Fibromuscular dysplasia	0.11	0.09	−0.06	0.29	1.31	1.9×10−01	0.13	0.09	−0.04	0.29	1.51	1.3×10−01	0.17	0.09	0.00	0.34	1.96	5.0×10−02
Migraine	0.08	0.02	0.05	0.11	5.00	5.8×10−07	0.08	0.02	0.05	0.12	4.52	6.3×10−06	0.08	0.02	0.04	0.12	4.28	1.9×10−05

The table displays genetic correlation computed using linkage disequilibrium score regression (unadjusted) and the genetic correlation after Genome‐Wide Complex Trait Analysis multitrait‐based conditional and joint analysis adjustments on systolic blood pressure and body mass index respectively (adjusted for SBP and adjusted for BMI). BMI indicates body mass index; L95, lower bound of the 95% CI for genetic correlation; r_g, genetic correlation; SBP, systolic blood pressure; and U95, upper bound of the 95% CI for genetic correlation.

We then investigated the shared genetic basis between UFs and several metabolic diseases and traits. UFs and T2D seem to share a significant proportion of their genetic basis (r_g=0.15, P=1.9×10⁻⁵, Table 1). This correlation was only marginally affected by the adjustment on both SBP and BMI GWAS results (Table 1). A positive genetic correlation was established with triglyceride levels (r_g=0.17, P=7.6×10⁻⁷), a marker of metabolic dysregulation and liver fat accumulation, and a negative correlation with high‐density lipoprotein cholesterol (r_g=−0.16, P=2.3×10⁻⁶) (Figure 1A, Table 1 and Table S2). These correlations remained unchanged after mtCOJO adjustments on both SBP and BMI GWAS results (Figure 1B and 1C, Table 1 and Table S2). A significant negative genetic correlation was detected with SHBG levels (r_g=−0.16, P=3×10⁻⁴), a blood glycoprotein produced in the liver that regulates bioavailability of sex steroids, including after adjustments on SBP and BMI genetics (Figure 1, Table 1 and Table S2).

Moreover, we analyzed the genetic correlations between UFs and main forms of cardiovascular disease. Although we found a positive genetic correlation between UFs and coronary artery disease (r_g=0.07, P=0.02), this result does not meet the Bonferroni‐corrected threshold for multiple testing. However, we found no support for a genetic link with stroke, intracranial aneurysm, or other vascular diseases with higher prevalence among women such as spontaneous coronary artery dissection ²⁴ , ²⁵ and fibromuscular dysplasia ²⁶ (Figure 1A, Table 1, and Table S2). The lack of significant genetic correlations was not masked by SBP and BMI genetic factors according to Genome‐Wide Complex Trait Analysis mtCOJO analyses (Figure 1B and 1C, Table 1 and Table S2). Interestingly, we report significant genetic correlation between UFs and migraine (r_g=0.08, P=5.8×10⁻⁷) that was only marginally affected by adjustment on SBP and BMI GWAS results (r_gmigraineadjSBP=0.08, P=6.3×10⁻⁶, r_gmigraineadjBMI=0.08, P=1.9×10⁻⁵, Figure 1, Table S2).

Central Obesity and Visceral Fat Markers as Genetic Risk Factors for Uterine Fibroids

We applied MR to estimate the potential causal links between traits for which we reported significant genetic correlations and UFs. Two‐sample MR analyses showed that genetically predicted higher BMI (beta_IVW‐BMI=0.033±0.008 per kg/m², P=6.1×10⁻⁵) and WHR adjusted BMI (beta_IVW‐WHR=0.193±0.047, P=3.3×10⁻⁵), but not BP traits, were associated with increased risk of UFs (Figure 2 and Table 2). Genetically predicted higher triglyceride levels were also associated with an increased risk for UFs (beta_IVW‐triglycerides =0.163±0.038 per mmol/L, P=1.9×10⁻⁵). Interestingly, genetically predicted higher SHBG levels were associated with lower risk of UFs (beta_IVW‐SHBG=−0.005±0.002 per nmol/L, P=2.5×10⁻³). However, bidirectional MR analyses indicated no evidence of reverse causality of genetic links with any of the diseases with which UFs shared their genetic basis (Figure S2 and Table S3). This likely results from the lack of precision and the low number of instruments involved in the analysis. Sensitivity analyses, including MR‐Egger, weighted median, and MR‐PRESSO, validated the MR results by providing evidence on the robustness of the IVW method's estimates against potential pleiotropic effects. None of the sensitivity tests failed, supporting the robustness of our findings.

Figure 2. Forest plots of the Mendelian randomization between uterine fibroids and their potential risk factors.

Beta coefficient of Mendelian randomization ( β ) is represented on the x axis and range represents the 95% CI. P value of correlation for 9 tests adjusted using Bonferroni method is indicated as: *P.adj <0.05. A, Log odds ratio between uterine fibroids and systolic blood pressure. B, Log odds ratio between uterine fibroids and diastolic blood pressure. C, Log odds ratio between uterine fibroids and body mass index. D, Log odds ratio between uterine fibroids and waist‐to‐hip ratio. E, Log odds ratio between uterine fibroids and high‐density lipoprotein levels. F, Log odds ratio between uterine fibroids and triglyceride levels. G, Log odds ratio between uterine fibroids and cigarettes per day. H, Log odds ratio between uterine fibroids and smoking initiation. I, Log odds ratio between uterine fibroids and sex‐hormone binding globulin. BMI indicates body mass index; CPD, cigarettes per day; DBP, diastolic blood pressure; HDL, high‐density lipoprotein; OR, odds ratio; SBP, systolic blood pressure; SHBG, sex hormone‐binding globulin; SI, smoking initiation; T2D, type 2 diabetes; TG, triglycerides; and WHR, waist‐to‐hip ratio adjusted on BMI.

Table 2.

Mendelian Randomization Analysis Between Uterine Fibroids (Outcome) and Cardiovascular Risk Factors, Hormones, and Blood Traits (Exposure)

Exposure factors	No. SNV	BETA (IVW)	SE (IVW)	P value (IVW)	BETA (Egger)	SE (Egger)	P value (Egger)	BETA (weighted median)	SE (weighted median)	P value (weighted median)	BETA (PRESSO)	SE (PRESSO)	P value (PRESSO)	Intercept	Intercept SE	Intercept P value
Systolic blood pressure	443	0.004	0.002	9×10−02	−0.002	0.007	7×10−01	0.00	0.003	1.0	0.004	0.002	5×10−02	0.002	0.002	3×10−01
Diastolic blood pressure	456	0.006	0.004	2×10−01	−0.01	0.010	2×10−01	0.002	0.006	8×10−01	0.006	0.004	1×10−01	0.004	0.002	4×10−02
BMI	267	0.03	0.008	6×10−05	−0.01	0.023	6×10−01	0.02	0.014	2×10−01	0.04	0.008	5×10−06	0.005	0.002	4×10−02
Waist‐to‐hip ratio adjBMI	333	0.19	0.047	3×10−05	0.29	0.108	7×10−03	0.12	0.062	5×10−02	0.20	0.04	1×10−06	−0.002	0.002	3×10−01
High‐density lipoprotein cholesterol	187	−0.12	0.105	3×10−01	0.12	0.176	5×10−01	0.10	0.143	5×10−01	−0.11	0.09	2×10−01	−0.003	0.002	9×10−02
Triglycerides	147	0.16	0.038	2×10−05	0.14	0.063	2×10−02	0.20	0.055	4×10−04	0.16	0.04	3×10−05	0.001	0.002	7×10−01
Cigarettes per day	12	0.05	0.075	5×10−01	0.08	0.167	6×10−01	−0.05	0.088	6×10−01	0.04	0.07	6×10−01	−0.002	0.008	8×10−01
Smoking initiation	69	−0.05	0.062	4×10−01	−0.12	0.304	7×10−01	−0.07	0.083	4×10−01	−0.07	0.06	2×10−01	0.002	0.008	8×10−01
Sex hormone binding globulin	162	−0.005	0.002	3×10−03	−0.008	0.003	1×10−02	−0.01	0.002	3×10−02	−0.005	0.001	4×10−04	0.002	0.002	3×10−01

BETA indicates effect size obtained from the inverse variance weighted; BMI, body mass index; Egger, weighted median and MR‐PRESSO analyses; IVW, inverse variance weighted; and No. SNV, count of single‐nucleotide variants used as instrumental variables in the MR analysis; For MR‐PRESSO, the results showed are from the outlier‐corrected estimations except for triglycerides, cigarettes per day, and smoking initiation where there were no outliers.

Finally, an overlap between UF loci and those associated with the tested traits and diseases was performed to control for potential overwhelming pleiotropy driving the observed genetic correlations and MR associations. Among the top SNVs at the 23 UFs loci studied, only 6 shared high posterior probability (>80%) of the same SNVs with BP and BMI traits, SHBG (Table S3), and 2 independent diseases (coronary artery disease/myocardial infarction and T2D, Table S4). Overall, we excluded that the association between cardiovascular traits and UFs is overwhelmingly driven by shared and highly pleiotropic loci, consistent with MR_PRESSO and MR_Egger findings, which both consider potential pleiotropy.

DISCUSSION

Our study shows that UFs share a substantial part of their genetic background with cardiometabolic traits related to BP, obesity, and diabetes in addition to migraine. Our MR findings do not support a causal relation between genetic predisposition to increased BP. However, we report evidence for traits related to central obesity, including WHR, triglycerides, and SHBG levels. We report these as relevant genetic risk factors for UFs supporting the role of visceral fat accumulation and its adverse metabolic consequences in UFs, an underresearched women‐specific condition affecting reproductive organs outside the context of pregnancy.

The present study reports a positive genetic correlation between UFs and both SBP and DBP. This result is in line with clinical reports highlighting that women with hypertension are at higher risk for UFs. ⁷ , ⁸ , ²¹ The epidemiological co‐occurrence of UFs with hypertension, and the positive genetic correlations we observed between BP and UFs may mirror the existence of shared genetically determined mechanisms. Vascular cells proliferation and tissue remodeling share many biological mechanisms in the vascular wall and the uterine tissue. Both vascular and uterine smooth muscle cells can be induced by growth factors and vasoactive peptides, ²⁷ , ²⁸ such as angiotensin‐II, which was previously reported to induce the proliferation of UF cells in vitro. ²⁸ We have recently described ultrastructure abnormalities indicating degenerative changes in the contractile apparatus of smooth muscle cells, in resistance vessels of patients with UFs, similar to fibroids abnormalities. ²⁹ However, our results do not support the existence of causal associations between genetically predicted fluctuations in BP with the risk for UFs. The biological pathways linking BP and UFs are complex and involve multiple intermediary factors or conditional metabolic and hormonal dependencies that were probably not captured by the current MR framework. Further population and molecular investigations are needed to disentangle the complex relation that may connect BP to UFs.

Additionally, our study provides evidence for associations between genetically determined central obesity and visceral fat content and increased risk of UFs, supporting data from observational studies. ²⁷ Our data are confirmatory to recent findings where BMI and WHR were identified as important genetic risk factors for several female reproductive conditions, including UFs. ³⁰ Our data also provide first evidence of the role of genetically determined triglycerides, a marker of visceral fat, in UFs risk in accordance with epidemiological evidence in the NFBC66 study, where 1 millimolar increase of circulating triglyceride levels was associated with 1.27 increase in the risk of UFs. ³¹ Consistently, we found that genetically lowered SHBG levels are causally associated with a higher risk of development of UFs. The increase in visceral fat, especially around the liver where it is more biologically active, ³² , ³³ is a potential additional source of estrogen production. A genetically determined lower availability or synthesis of SHBG by adipose tissue as a counterpart to higher estrogen levels is a potential mechanism of uterine smooth muscle cell proliferation that may deserve deeper investigation. ²⁷ , ³³ , ³⁴

Our results also show evidence for significant genetic correlation between UFs and T2D. Observational clinical data previously described higher risk of T2D among patients with UFs. ¹⁴ , ³⁵ However, the absence of causal association established by MR suggests that the 2 diseases share genetically determined biological mechanisms but the currently established genetic determinants of T2D have little influence on the genetic risk for UFs. One major suspected shared mechanism is central obesity, higher visceral fat content, altered lipid metabolism specifically in the liver, and diminished bioavailability of SHBG. ³⁶ , ³⁷ , ³⁸ Whether female sex‐specific genetic risk factors for T2D may influence UFs onset may clarify this interesting genetic link between T2D and UFs.

Beyond BP, our study does not support genetically driven causal relationship between cardiovascular diseases and UFs, despite evidence for patients with UFs to be at higher cardiovascular risk. ⁸ , ¹⁴ We report only nominally significant genetic correlations with coronary artery disease and myocardial infarction, suggesting a potential common genetic background with atherosclerotic disease. Suspected biological mechanisms are notably hyperlipidemia and smooth muscle cell remodeling, 2 major mechanisms of atherosclerotic disease. ³⁵ Nonetheless, we excluded any evidence of genetic correlations between UFs and fibromuscular dysplasia, a female arteriopathy, where smooth muscle cell proliferation is a suspected mechanism as reported in recent genetic investigation from our group. ²⁰ A potential explanation is the major discrepancy between UFs and fibromuscular dysplasia, with the latter not being influenced by dyslipidemia and overall or central obesity, ³⁹ whereas obesity is a known risk factor for UFs.

We report for the first time a genetic correlation between UFs and migraine. Recently, a significant genetic correlations was reported between endometriosis and migraine. ⁴⁰ These findings linking genetically migraine to several women's reproductive diseases stresses further the need to dig into these shared molecular mechanisms, especially that migraine is a neurovascular disorder predominantly affecting women. The absence of association through MR between migraine and UFs suggests joint occurrence rather than causal relation. One possible explanation could be shared exposure to BP and female‐specific factors, such as hormones fluctuation during the life course. ⁴¹ Of note, ultrastructural phenotypic resemblance in vascular remodeling with UFs involving smooth muscle cells were described in superficial temporal and occipital arteries of patients with migraine. ⁴² , ⁴³

The present study also presents several limitations. First, the UFs GWAS meta‐analysis employed included overlapping samples from the UK Biobank that we used in some of the cardiometabolic data sets in the genetic correlation analyses, although we considered this situation in genetic correlation analyses through nonconstraint on intercept. The lack of association causality through MR reflects only the variance of the traits controlled by genetic variants and not the overall variance of the traits analyzed. This study used only publicly available summary statistics, which limits the possibility to stratify on important potential confounding or mediating factors such as oral contraception intake or parity. Our study was limited to genetic data generated mainly in women of European ancestry and may not be applicable in African or Asian ancestry, where UFs have higher prevalence. At the time of design and analyses of our study, well‐powered genetic studies in these populations were rare and those published did not provide publicly accessible GWAS summary statistics to allow their use for genetic correlation or MR analyses. Overall, our findings need to be confirmed using larger and more diverse sets of patient cohorts.

Conclusions

In conclusion, our study confirms existing and delineates novel genetic links between the risk of UFs development and traits and diseases associated with cardiovascular risk. We demonstrate that UFs exhibit a substantial genetic overlap with BP, central obesity, and T2D. Furthermore, we present evidence supporting mediation by shared mechanisms, notably smooth muscle cell remodeling, dyslipidemia, and impaired levels of SHBG. Notably, we report, for the first time, a genetic correlation between UFs and migraine, likely attributable to shared yet not entirely overlapping genetic mechanisms involved in vascular remodeling with high relevance of BP regulation and potentially shared risk factors related to female sex. These findings unveil multiple avenues for further investigation to elucidate the cause of UFs, an underinvestigated and poorly understood condition of women associated with poor cardiovascular health.

Sources of Funding

This study was funded by the French Society of Cardiology, through Fondation Coeur et Recherche (to Nabila Bouatia‐Naji); La Fédération Française de Cardiologie (to Nabila Bouatia‐Naji).

Disclosures

None.

Supporting information

Tables S1–S4

Figures S1–S2