Causal association between uterine fibroids and cardiovascular disease: A Mendelian randomization study

Abstract

Uterine fibroids (UF), the most common benign tumors in women, have been associated with cardiovascular disease (CVD) in observational studies, yet causal inference remains unclear due to confounding. This Mendelian randomization (MR) study leveraged genetic variants as instrumental variables (IVs) to evaluate the causal relationship between genetically predicted UF and CVD risk. Exposure data were derived from a UK Biobank genome-wide association study of 462,933 Europeans (5168 UF cases). Outcome data for CVD subtypes were obtained from FinnGen and EBI consortia (sample sizes: 180, 862–977, 323). Fourteen independent single-nucleotide polymorphisms strongly associated with UF (P < 5 × 10⁻⁶) were selected as IVs, pruned for linkage disequilibrium (r²<0.001; F-statistic > 10). The inverse-variance-weighted method was used for primary analysis, supplemented by sensitivity approaches (weighted median, MR-Egger, MR-PRESSO). Robustness was evaluated via Cochran Q test (heterogeneity), MR-Egger intercept (pleiotropy), and leave-one-out analysis. Genetically predicted UF showed no causal effects on ischemic heart disease (OR = 2.00E−02, 95% CI: 2.21E−04 to 2.47E+00, P = .11), heart failure (OR = 5.20E−01, 95% CI = 1.03E−02 to 2.57E+01, P = .74), venous thromboembolism (OR = 1.65E+00, 95% CI: 1.13E−03 to 2.41E+03, P = .89), or stroke (OR = 3.70E−01, 95% CI = 1.82E−03 to 7.48E+01, P = .71). Sensitivity analyses confirmed consistency (all P > .05), with no heterogeneity (Cochran Q P > .05) or horizontal pleiotropy (MR-Egger intercept P > .05). Leave-one-out analysis indicated stable estimates. This MR study found insufficient evidence to support a causal link between UF and CVD, contrasting previous observational reports.

1. Introduction

A report by the American Heart Association reveals that cardiovascular disease (CVD) caused 19 million global deaths in 2020, marking an 18.7% surge from 2010, while remaining the predominant killer of women by accounting for 35% of female mortality in 2019.^[1] Due to differences in the pathophysiological basis of cardiovascular health between males and females, there are distinct variations in the manifestation of cardiovascular disease symptoms. Studies have indicated that women have a higher 1-year mortality rate following acute myocardial infarction, with fewer obstructions but greater diffusivity compared to men.^[2] CVD typically arises from a combination of multiple etiological factors.^[3] The occurrence and progression of CVD may be driven by interactions between genetic factors, environmental triggers, and immune dysregulation.^[4]

Uterine fibroids (UF) are the most common benign tumors in women, with nearly 80% to 90% of women being diagnosed with fibroids by the age of 50. UF can cause pain, excessive menstrual bleeding, or infertility.^[5] Epidemiological and molecular investigations consistently demonstrate that Black women experience a 2-to-3-fold elevated incidence of UF compared to White women, with concomitant racial disparities manifesting through distinct molecular signatures encompassing differential gene regulatory networks, proteomic remodeling patterns, and miRNA-mediated epigenetic modulation.^[6,7]

In recent years, the association between UF and cardiovascular disease has gained considerable attention from epidemiological studies. Data from a stratified probability sample survey of the civilian non-institutionalized population in the United States indicated that women with self-reported history of fibroids had a higher prevalence of cardiovascular disease, including ischemic heart disease (IHD), heart failure, and stroke.^[8] A study based on the National Health and Nutrition Examination Survey database demonstrated that the incidence of cardiovascular disease was higher in UF patients compared to non-UF individuals.^[9] A long-term epidemiological trial revealed that UF were one of the most common risk factors for IHD.^[10] A case-control study showed an increased risk of VTE in patients with UF.^[11] However, the question of whether UF increases the risk of CVDs remains controversial. Research indicates that while women with UF have more CVD risk factors, the presence of UF is not associated with subclinical CVD.^[12] There is overlap between the risk factors for UF and cardiovascular disease, and some fibroids may be severe enough to require uterine removal with or without oophorectomy, which may independently increase the risk of CVDs beyond the fibroid status. Studies assessing the association between UF and CVDs outcomes have reported inconsistent results, likely due to limitations in sample size. Furthermore, observational epidemiological studies are prone to confounding and reverse causality.^[13] Further investigation is needed to determine the causal relationship between UF and CVD outcomes. Randomized controlled trials (RCTs) can mitigate selection bias and enhance the reliability of results through blinded analysis. However, RCTs are challenging or impractical due to high economic costs, labor intensity, resource and time requirements, and ethical constraints.^[14]

Mendelian randomization (MR) is a method that overcomes unmeasured confounding and reverse causality in traditional observational studies.^[15] By utilizing genetic variants as instrumental variables (IVs) to proxy exposures, MR enables causal effect estimation between exposures and outcomes.^[16] The advancement of genome-wide association studies (GWAS) has provided robust IVs for MR investigations. This 2-sample MR study aims to determine whether genetically predicted UF exhibit causal associations with CVDs risk.

2. Methods

2.1. Study design

This 2-sample MR analysis was conducted using summary-level genetic data from publicly available GWAS. The study design adhered to the 3 core assumptions of MR: genetic IVs must be strongly associated with UF; IVs must be independent of confounders influencing both UF and CVD; and IVs must affect CVD outcomes exclusively through UF, with no alternative pathways. The reporting of this study follows the STROBE-MR^[17] guidelines to ensure methodological transparency and rigor. Ethical approval was not required, as the analysis utilized exclusively de-identified, publicly available summary-level genetic data from GWAS. All original GWAS datasets obtained informed consent from participants and received approval from their respective institutional review boards. No additional ethical review was necessary for this secondary analysis of aggregated data under national regulations and institutional policies governing the use of anonymized genomic data.

2.2. Data sources

Summary statistics for UF were obtained from a European-ancestry GWAS meta-analysis comprising 462,933 participants (5168 cases and 331,991 controls) in the UK Biobank (dataset ID: ukb-b-12722). CVD outcome data were sourced from the following GWAS consortia: IHD from the FinnGen study (31,640 cases and 187,152 controls; dataset ID: finn-b-I9_IHD), heart failure from the European Bioinformatics Institute study (47,309 cases and 930,014 controls; dataset ID: ebi-a-GCST009541), venous thromboembolism (VTE) from FinnGen (9176 cases and 209,616 controls; dataset ID: finn-b-I9_VTE), and stroke from FinnGen (18,661 cases and 162,201 controls; dataset ID: finn-b-C_STROKE). Detailed characteristics of exposure and outcome datasets, including sample sizes, ancestry, and phenotype definitions, are provided in Table 1.

Table 1.

Details of the data source.

Trait	Non-cancer illness code, self-reported: uterine fibroids	Ischaemic heart disease, wide definition	Heart failure	Venous thromboembolism	Stroke
ID	ukb-b-12722	finn-b-I9_IHD	ebi-a-GCST009541	finn-b-I9_VTE	finn-b-C_STROKE
Year	2018	2021	2020	2021	2021
Category	Binary	Binary	NA	Binary	Binary
Sub category	NA	NA	NA	NA	NA
Population	European	European	European	European	European
Sex	Males and Females	Males and Females	NA	Males and Females	Males and Females
Ncase	7122	31,640	47,309	9176	18,661
Ncontrol	455,811	187,152	930,014	209,616	162,201
Sample size	462,933	218,792	977,323	218,792	180,862
Number of SNPs	9851,867	16,380,466	7773,021	16,380,466	16,380,350
Unit	SD	NA	NA	NA	NA
Priority	1	NA	NA	NA	NA
Author	Ben Elsworth	NA	Shah S	NA	NA
Consortium	MRC-IEU	NA	NA	NA	NA
Ontology	NA	NA	NA	NA	NA
Build	HG19/GRCh37	HG19/GRCh37	HG19/GRCh37	HG19/GRCh37	HG19/GRCh37
Note	https://gwas.mrcieu.ac.uk/datasets/ukb-b-12722/	https://gwas.mrcieu.ac.uk/datasets/finn-b-I9_IHD/	https://gwas.mrcieu.ac.uk/datasets/ebi-a-GCST009541/	https://gwas.mrcieu.ac.uk/datasets/finn-b-I9_VTE/	https://gwas.mrcieu.ac.uk/datasets/finn-b-C_STROKE/

ID = identity, NA = not available, SD = standard deviation, SNP = single-nucleotide polymorphism.

2.3. Selection of instrumental variables

We obtained 9,851,867 SNPs associated with UF from the IEU Open GWAS project (https://gwas.mrcieu.ac.uk/). These SNPs were identified from a meta-analysis of 462,933 European individuals, including 5168 cases and 331,991 controls. All SNPs were thresholded at P < 5 × 10⁻⁶ and tested for linkage disequilibrium (LD) to identify independent SNPs in LD. These SNPs were pruned within 10,000 kb windows and trimmed with an r² < 0.001 threshold. We searched for any selected SNP’s secondary phenotypes in PhenoScanner V2 to exclude any SNP associated with other phenotypes that might influence the risk of CVD outcomes. To avoid potential weak instrument bias, the strength of IVs was assessed using the F-statistic (F = beta²/se²). If F > 10, it was considered that the IV had a sufficiently strong correlation with the exposure to protect the MR analysis results from weak instrument bias.^[18]

2.4. Statistical analyses

Before conducting the analysis, we harmonized the exposure and outcome data to align the effect alleles with the forward strand, either as specified or inferred based on allele frequencies. We employed the inverse-variance-weighted (IVW) method as the primary approach for calculating causal effects. The IVW model is the most powerful method for detecting causal relationships in 2-sample MR analysis.^[19] Additionally, we used other established MR methods, including MR-Egger regression, weighted median, simple mode, weighted mode, maximum likelihood, and penalized weighted median, for sensitivity analysis. The MR-PRESSO method was used to detect outlier variables in the IVW analysis by comparing the observed distance of genetic variation with the expected distance of regression, assuming the absence of horizontal pleiotropy, and assessing the causal estimates after removing the outliers.^[20] Each method made different assumptions about the validity of IVs. When 50% of the IVs are invalid, median weighting is estimated. Although MR-Egger has lower statistical power, it provides estimates after correcting for pleiotropic effects.^[21] Scatter plots were generated to depict the association of genetic liability for UF with CVD. We employed all these methods to comprehensively investigate the causal relationship. Heterogeneity was detected using both IVW and MR-Egger regression. Cochran Q test was used to quantify heterogeneity, with P < .05 considered statistically significant. The presence of heterogeneity does not necessarily imply the ineffectiveness of the IVW model. The MR-Egger method allows for the presence of nonzero intercept and is used to detect directional pleiotropy.^[22] We employed leave-one-out (LOO) analysis to reevaluate the causal effects by sequentially excluding 1 SNP to determine if the association is driven by a single SNP. Forest plots were generated to directly assess the presence of heterogeneity. All statistical analyses were performed using the “TwoSampleMR,” “MR-PRESSO,” and “forestplot” packages in R software version 4.2.3.

3. Results

3.1. Selection of SNPs in Mendelian randomization

We extracted IVs associated with UF from the GWAS (P < 5 × 10⁻⁶) and pruned them for LD (r² < 0.001, 10,000Kb). Subsequently, SNPs associated with CVD were retrieved from PhenoScanner V2. We excluded 6 SNPs (rs12822345, rs2736100, rs17332320, rs71575922, rs2057178, and rs3820282) that were associated with confounding factors (BMI, coronary artery disease, prior uterine and bilateral oophorectomy). The detailed characteristics of SNPs associated with UF are shown in Table S1, Supplemental Digital Content, https://links.lww.com/MD/P48. The F-statistic for all 14 SNPs exceeded the threshold of 10, indicating strong prediction of UF in the MR analysis.

3.2. Causal estimates of genetic susceptibility to uterine fibroids and cardiovascular disease risk

Mendelian randomization analysis revealed no significant association between UF and CVD outcomes (all P > .05) (Fig. 1). The results were consistent with MR-Egger regression, weighted median, simple mode, weighted mode, maximum likelihood, and penalized weighted median methods (Table S2, Supplemental Digital Content, https://links.lww.com/MD/P48, Fig. 2). MR-PRESSO analysis did not detect any outliers, indicating the reliability of the results.

Figure 1.

Causal estimates of genetic susceptibility to uterine fibroids and risks of cardiovascular diseases. IHD = ischemic heart disease, HF = heart failure, VTE = venous thromboembolism.

Figure 2.

Scatter plots for the association between uterine fibroids and cardiovascular diseases. UF = uterine fibroids, MR = Mendelian randomization, IHD = ischemic heart disease, HF = heart failure, VTE = venous thromboembolism.

3.3. MR sensitivity analysis

Sensitivity analysis indicated no evidence of potential heterogeneity and horizontal pleiotropy in the results (Table 2). In the heterogeneity test, the p-values of Cochran Q statistic were all >.05, indicating no heterogeneity among the SNPs (Table 2). The leave-one-out analysis suggested that the potential causal association between UF and CVD risk was not driven by a single SNP (Fig. S1, Supplemental Digital Content, https://links.lww.com/MD/P49). Forest plots provided a visual representation of heterogeneity, as shown in Figure S2, Supplemental Digital Content, https://links.lww.com/MD/P49. Additionally, the MR-Egger regression intercept test indicated no evidence of horizontal pleiotropy in the IVs for UF in any type of CVD.

Table 2.

Heterogeneity and pleiotropy tests for the associations of uterine fibroids with cardiovascular diseases.

Outcome	Heterogeneity test						Pleiotropy test
	MR-Egger			Inverse-variance-weighted			MR-Egger
	Q	Q_df	Q_pval	Q	Q_df	Q_pval	Intercept	SE	P
IHD	11.42	12	.49	11.87	13	.54	−0.01	0.02	.51
HF	12.72	10	.24	13.21	11	.28	0.01	0.01	.55
VTE	14.96	12	.24	15.31	13	.29	−0.01	0.25	.61
Stroke	7.95	12	.79	9.14	13	.76	0.02	0.02	.3

MR = mendelian randomization, HF = heart failure, IHD = ischemic heart disease, p = P value, Q = statistic Q, Q_df = statistic Q degree of freedom, Q_pval = statistic Q P value, SE = standard error, VTE = venous thromboembolism.

4. Discussion

We systematically explored the potential causal effects of genetic susceptibility to UF on CVD risk using a 2-sample MR approach. By utilizing genetic data, this study circumvented confounding bias, selection bias, and reverse causality commonly present in observational studies. Notably, MR’s core strength lies in its ability to disentangle direct effects from environmental and lifestyle confounders, as genetic instruments are fixed at conception and unaffected by postnatal exposures such as diet, smoking, or socioeconomic status.^[23]

During the same period that this study was conducted, Cui et al^[24] published an MR study on the causal association between UF and CVD, providing an important comparative reference for understanding their relationship. Based on data from 258,718 European populations, the study found that genetic susceptibility to UF was significantly positively associated with the risk of cardioembolic stroke (OR = 1.113, 95% CI = 1.018–1.218, P = .019). The study also suggests that UF decreases the risk of myocardial infarction (OR = 0.943, 95% CI = 0.899–0.989, P = .015), but was not significantly associated with coronary heart disease, atrial fibrillation, or heart failure. Both studies used large-scale GWAS data, and the present study had a larger sample size (462,933 cases). We further excluded IVs related to BMI, coronary artery disease, prior uterine and bilateral oophorectomy, thereby reducing the confounding effects of therapeutic interventions and metabolic factors. In terms of outcome scope, this study added subtypes such as venous thromboembolism, expanding Cui et al focus on cardioembolic stroke and myocardial infarction. Collectively, these findings provide a comprehensive evaluation of arterial and venous thrombotic disorders in the context of UF genetic susceptibility.

There is no MR evidence supporting a potential causal relationship between genetic susceptibility to UF and overall CVD risk. The study findings remained consistent in sensitivity analyses using different Mendelian tools and statistical models, indicating the robustness of our analysis.

Previous studies have often investigated the association between CVD and UF. A descriptive study in a review found that out of 438 women with VTE, 72 (16.4%) had fibroids.^[25] In a multicenter study in Japan, among 470 female stroke patients, 39 (8%) had common benign gynecological diseases, with 24 (62%) of them having UF.^[26] However, previous research has primarily focused on the causal direction from CVD to UF. Our study, on the other hand, aimed to explore the causal relationship between UF and CVD.

The MR analysis did not provide sufficient evidence to support a positive causal relationship between UF and CVD risk. The lack of genetic causal effects between UF and CVD risk suggests that the observed correlation between the presence of UF and CVD risk, as demonstrated in previous observational studies, may be due to residual confounding from shared risk factors. Factors such as the renin-angiotensin-aldosterone system, estrogen, and endothelial dysfunction have been implicated in the pathogenesis of UF, as reported by Kirschen et al.^[27] The expression of cytokines and chemokines with pro-inflammatory and pro-fibrotic characteristics, which play important roles in atherosclerotic plaques and thrombus formation, is elevated in UF patients and may influence the occurrence of CVDs.^[28] Obesity, diabetes, hypertension, and hypercholesterolemia are more common in women with UF, and coincidentally, these are also traditional risk factors for cardiovascular disease.

Genetic susceptibility to UF is influenced by various hormones, including growth hormone, prolactin, and estrogen.^[29] Hormonal therapies such as gonadotropin-releasing hormone analogs, progestogens, selective estrogen receptor modulators, dopamine agonists, prostaglandin analogs, and selective progesterone receptor modulators are used as non-surgical treatments for uterine fibroid patients.^[30] However, oral hormone therapy increases the risk of venous thromboembolism and stroke.^[31] Additionally, vitamin D deficiency is closely associated with the potential development of UF, and vitamin D supplementation can prevent the occurrence of heart failure.^[32,33]

The main strengths of this study include the use of a 2-sample MR analysis to assess the independent effects of UF on multiple CVD outcomes, thereby overcoming several limitations of traditional epidemiological research. Additionally, large-scale GWAS datasets were utilized, with non-overlapping exposure and outcome data, enhancing the reliability of the results. Furthermore, multiple analytical methods were employed to ensure consistent findings.

Nevertheless, there are inherent limitations to this study. Despite using summary-level data, the lack of access to individual-level data prevented a complete analysis of the association between the severity of UF and CVD. Therefore, further MR analyses are still needed to estimate the causal relationship between UF of different severities and CVD outcomes. Additionally, the prevalence and mortality rates of UF vary among different racial and ethnic groups. The participants in this MR analysis were all of European ancestry, which may limit the generalizability of our findings to other populations. Finally, while MR minimizes confounding by environmental factors, it does not explicitly model gene-environment interactions. Residual confounding from unmeasured interactions remains a possibility.

5. Conclusion

This MR analysis, utilizing summary-level data from large-scale GWAS in European-ancestry populations, found insufficient evidence to support a causal relationship between genetically predicted UF and CVD risk. These findings contrast with previous observational reports suggesting an association. Further studies are warranted to reconcile these discrepancies and investigate potential biological mechanisms that may link these conditions in observational settings.

Acknowledgments

We would like to thank supports from Jilin Provincial Scientific and Technological Development Program (No. 20230203069SF) and Jilin Provincial Scientific and Technological Development Program (No. YDZJ202301ZYTS198). We thank all participants, staff, and institutions that have contributed to the publicly available GWAS database.

Author contributions

Conceptualization: Ming Yao, Yiqiang Wang, Lihong Jiang.

Investigation: Dongze Zhang, Yingzi Cui.

Methodology: Ming Yao, Yingzi Cui, Lihong Jiang.

Writing – original draft: Ming Yao, Shulan Zhao.

Writing – review & editing: Ming Yao, Yingzi Cui, Lihong Jiang.