Causal relationship between uterine fibroids and cardiovascular disease: A two-sample Mendelian randomization study.

Our study provides evidence for a potential causal effect between UF and CES as well as MI, laying the foundation for further mechanistic research. However, we did not find a causal relationship between UF and CHD, AF, or HF. Further research is needed to validate our findings and ultimately draw more conclusive results.

Uterine fibroids (UF), also known as leiomyomas, are the most common benign tumors of the reproductive organs in women of reproductive age. They can affect the reproductive system in various ways, either individually or in combination. The most common initial symptom of fibroids is heavy menstrual bleeding, which can lead to anemia, fatigue, and dysmenorrhea. Other symptoms include noncyclical pain, abdominal distension, pain during intercourse, pelvic pressure, and bladder or bowel dysfunction, resulting in frequent urination, urinary retention, and constipation. UF may also be linked to reproductive issues, such as impaired fertility, pregnancy complications, and adverse obstetric outcomes, including an increased risk of preterm labor, cesarean section, antepartum hemorrhage, poor fetal development, and growth restriction. [ 1 , 2 ] Statistically, between 5.4% and 77% of women are affected by leiomyomas, which are found in 70% of uteri after hysterectomy. In more than 80% of cases, multiple leiomyomas are present. [ 3 ] Genetic factors, nulliparity, obesity, polycystic ovary syndrome, diabetes, and hypertension have all been shown to be associated with an increased risk of developing UF. [ 4 ] Cardiovascular disease (CVD) is a group of conditions that affect the heart and blood vessels, including coronary artery disease, stroke, atrial fibrillation (AF), congestive heart failure (HF), and other heart-related disorders. As the leading cause of death worldwide, the prevalence of CVD nearly doubled between 1990 and 2019, rising from 271 million to 523 million cases. Similarly, the number of CVD-related deaths increased from 12.1 million to 18.6 million. [ 5 ] Numerous studies have demonstrated a strong association between the presence of leiomyomas and the diagnosis of some form of CVD. An observational study found that women with a history of UF had a higher chance of developing CVD, including ischemic heart disease, HF, and stroke. [ 6 ] Similarly, a cross-sectional study that included 104 patients with UF and 624 controls reported that women with UF had a significantly high risk of hypertension and more asymptomatic organ damage compared to the control group, especially in younger women. [ 7 ] Yet another observational study found that the presence of UF was not associated with subclinical CVD, despite the fact that women with UF exhibited more CVD risk factors. [ 8 ] Furthermore, traditional observational studies are subject to various confounding factors and reverse causality. [ 9 ] Therefore, the potential causal relationship between UF and CVD remains unclear. The Mendelian randomization (MR) approach uses genetic variation as an instrumental variable (IV) to provide evidence for a potential causal relationship between modifiable risk factors and disease. [ 10 ] Genetic variants, typically single nucleotide polymorphisms (SNPs), are randomly distributed and independent of other traits, which helps minimize the impact of confounding factors and reverse causality. [ 11 ] In the present study, we explored the causal relationship between UF and various CVDs through an MR study using summary statistics from genome-wide association studies (GWAS). Additionally, the linkage disequilibrium score regression (LDSC) method was used to quantitatively analyze the genetic correlations across the entire genome.

Conceptualization: Jie Cui, Yue-Chen Zhao, Tie-Jun Wang. Formal analysis: Li-Zhen She. Methodology: Jie Cui. Writing – original draft: Jie Cui. Writing – review & editing: Yue-Chen Zhao, Li-Zhen She, Tie-Jun Wang.

First, the genetic correlation between exposure and outcome was assessed using LDSC. This study then conducted univariable MR (UVMR) analysis assuming UF as the exposure variable and 5 CVDs including coronary heart disease (CHD), myocardial infarction (MI), AF, HF, and cardioembolic stroke (CES) as the outcome variables. The selection of IVs for UF was based on 3 key assumptions: (1) the genetic variation is strongly associated with exposure. (2) The genetic variant is not associated with any confounders of the exposure–outcome association. (3) The genetic variant affects the outcome only through its association with the exposure. [ 12 ] To further explore the possibility of reverse causality, a reverse MR analysis was also performed. Since hypertension, diabetes, smoking, and sedentary behavior may act as confounders in the analysis of exposures versus outcomes, multivariable MR (MVMR) was applied to further explore the direct effect of exposures on outcomes (Fig. 1 ). It is important to note that all original studies included in this analysis received ethical approval, and participants provided informed consent. As a result, this study did not require additional ethical approval. Study design flowchart of Mendelian randomization. AF = atrial fibrillation, CES = cardioembolic stroke, CHD = coronary heart disease, E2 = estradiol, EC = endometrial cancer, EMS = endometriosis, HF = heart failure, MI = myocardial infarction, MR = Mendelian randomization, PCOS = polycystic ovary syndrome, SNPs = single nucleotide polymorphisms, UF = uterine fibroids. Genome-wide association studies (GWAS) associated with UF pooled data from the European Bioinformatics Institute database. To the best of our knowledge, this is the largest published GWAS meta-analysis of European pedigrees to date, [ 13 ] which includes a total of 258,718 samples (21,024 cases and 237,694 controls) from the UK BioBank (UKB) and FinnGen. The UKB is a prospective, population-based study that recruited approximately 500,000 people in the United Kingdom with an average age between 40 and 69 years, and collected in-depth genetic and phenotypic data on them. [ 14 ] As a public-private research project, FinnGen combines estimated genotype data generated from newly collected and legacy samples from the Finnish Biobank and digital health record data from the Finnish Health Registry, aiming to provide novel insights into medicine. [ 15 ] The diagnostic criteria for UF were based on International Classification of Diseases (ICD)-10, with a specific code of D25. We selected 5 CVDs as outcome variables for MR analysis, including CHD, MI, AF, HF, and CES. To increase the confidence of the analysis results, the GWAS summary data for the outcomes were obtained from databases other than UKB and FinnGenn. Additionally, we restricted the study population to Europeans to minimize possible pedigree mismatches. Pooled statistics for CHD were obtained from the Coronary ARtery DIsease Genome wide Replication and Meta-analysis (CARDIoGRAM) consortium, which conducted a meta-analysis of 14 GWAS for coronary artery disease, including 22,233 cases and 64,762 controls. [ 16 ] Summary statistics for MI were from another GWAS analysis, [ 17 ] including 14,825 cases and 2680 controls. Summary statistics for AF were from a GWAS meta-analysis of 6 studies that included 60,620 cases and 970,216 controls. [ 18 ] For HF, summary statistics were derived from 47,309 cases and 930,014 controls from 26 studies. [ 19 ] Finally, summary statistics for CES were obtained from the MEGASTROKE consortium, which includes 7193 cases and 406,111 controls. [ 20 ] Detailed information is provided in Table 1 . The diagnosis of CVD was based on ICD-10. CHD was defined as “ICD-10-I20, I21, I22”; MI as “ICD-10-I21, I22”; AF as “ICD-10-I48”; HF as “ICD-10-I50”; and CES as “ICD-10-I63.4.” Detailed information for the genome-wide association study data. AF = atrial fibrillation, CES = cardioembolic stroke, CHD = coronary heart disease, GWAS = genome-wide association studies, HF = heart failure, MI = myocardial infarction, PMID = PubMed identifier, UF = uterine fibroids. We selected SNPs as IVs for the analysis. First, we extracted SNPs that were strongly correlated with exposure ( P  < 5.0 × 10 -8 ). Second, to minimize bias caused by linkage disequilibrium (LD), the screened SNPs had to fulfill the following 2 conditions: r 2  < 0.001, kb = 10,000. After going through these 2 steps, we recorded the resulting SNPs along with their variant identifier, effect size of allele, standard error of effect size, and P -value. To assess the strength of association between IVs and exposure factors, we used the F statistic. It was calculated as F =  R 2 ( N  - K - 1)/[K (1 -  R 2 )]. Where N is the sample size of the exposed dataset, K is the number of SNPs, and R 2 is the proportion of phenotypic variance explained by genetic variation. A value of F > 10 was considered sufficient to avoid bias from weak IVs. [ 21 ] We harmonized SNP statistics for both the exposure and the outcomes to ensure that the allele assignments were consistent across all SNPs. Depending on the heterogeneity, an inverse variance weighted (IVW) approach with different models was used as the primary method of analysis. The IVW method relies on asymptotic estimates of the standard error of the causal (ratio) estimate from each variant, making it essential for each IV to be valid. [ 22 ] At the same time, we used several other MR analysis methods, including weighted median, MR-Egger, simple mode, and weighted mode. The weighted median method provides consistent estimates of causal effects when up to half of the IVs are not valid. [ 23 ] MR-Egger regression was used to detect and correct for the possible presence of pleiotropy. [ 24 ] The simple mode and weighted mode methods were also applied as supplementary techniques to evaluate the causal effects of individual SNPs and create ensemble estimates. [ 25 ] Furthermore, we applied the MR-PRESSO method to detect and remove potential outliers, ensuring more accurate MR analysis results. [ 26 ] To correct for multiple comparisons in this study, we applied the Benjamini–Hochberg-based false discovery rate (FDR). The association between exposure and outcome was considered statistically significant if the IVW corrected P -value was <.05 and the results from the other methods were consistent in direction. An association was deemed suggestive if the initial IVW P -value was .05, with the corrected P -value being larger than the initial value. [ 27 ] For causal associations with significant results in UVMR, we further performed MVMR to adjust for potential confounders, including hypertension, diabetes, smoking, and sedentary behavior. Sedentary behavior encompassed time spent using a computer and driving. The primary method of analysis was the IVW method. Multiple sensitivity analyses were used in this study, with the P -values from all sensitivity tests compared against a significance threshold of .05. The intercept of MR-Egger regression was used to estimate the magnitude of horizontal pleiotropy. [ 28 ] Cochran Q -value assessed the heterogeneity among individual SNPs. If the P -value was >.05, it indicated that there was no heterogeneity, and the IVW approach with a fixed-effects model was chosen; conversely, if the P -value was <.05, it indicated that there was heterogeneity, and the IVW approach with a random-effects model was chosen. [ 29 ] Leave-one-out sensitivity analyses were performed to identify any individual SNPs that might have a disproportionate impact on the results. In addition, the results of the MR analyses were visualized using scatter plots, forest plots, and funnel plots. All statistical analyses were performed using the “TwoSampleMR” and “MR-PRESSO” packages in R software version 4.3.2, and all P -values were two-sided. MR estimates were represented by the odds ratio (OR) with 95% confidence intervals (CI). We employed the LDSC approach to assess genetic correlations between UF and CVD. Polygenicity as well as confounding bias may produce spurious distributions of test statistics in GWAS. LDSC allows us to detect the relationship between the test statistic and the LD to quantitatively analyze the effect of each component. [ 30 ] In addition, LDSC can identify genetic correlations between complex traits and diseases to better explain etiology as well as prioritize possible causal relationships. This method is not biased by sample overlap. [ 31 ] This method reflects whether the GWAS test statistic for a variant is associated with a nearby high chain imbalance variant by generating scores. [ 32 ] The z-score for each variant in trait 1 was multiplied by the z-score for each variant in trait 2, and this product was regressed against the LD score to estimate genetic covariates. Genetic covariance normalized by SNP-heritability indicates genetic correlation. [ 33 ]

We applied the LDSC approach to assess the genetic correlation between UF and 5 CVDs. The results indicated that there was no genetic correlation between UF and CHD ( r g  = 0.007, P  = .933), MI ( r g  = -0.068, P  = .244), AF( r g  = -0.021, P  = .665), HF ( r g  = 0.233, P  = .068), and CES ( r g  = -0.011, P  = .923). We extracted SNPs strongly associated with UF from GWAS ( P  < 5.0 × 10 -8 ). Afterward, we removed SNPs in LD ( r 2  < 0.001, kb = 10,000). Finally we excluded palindromic sequences with moderate allele frequencies. In total, 30 SNPs, 44 SNPs, 45 SNPs, 41 SNPs, and 45 SNPs were used to perform MR analyses of UF with CHD, MI, AF, HF, and CES, respectively. In addition, we did not observe any bias from weak IVs, as all F-statistics were >10 (Tables S1–S5, Supplemental Digital Content, http://links.lww.com/MD/O452 ). The results, shown in Figure 2 , indicate that the IVW approach in MR analysis revealed a significant positive correlation between genetically determined UF and CES (OR = 1.113, 95% CI: 1.018–1.218, P  = .019, P FDR  = .047). The causal relationship between UF and MI was suggestive (OR = 0.943, 95% CI: 0.899–0.989, P  = .015, P FDR  = .075). The causal relationship between UF and HF (OR = 1.003, 95% CI: 0.969–1.037, P  = .874, P FDR  = .874), AF (OR = 1.018, 95% CI: 0.980–1.058, P  = .358, P FDR  = .597), and CHD (OR = 1.018, 95% CI: 0.940–1.102, P  = .667, P FDR  = .834) were not significantly associated. The scatterplot visualizing these results is shown in Figure S1, Supplemental Digital Content, http://links.lww.com/MD/O451 . No outliers were identified using the MR-PRESSO method, suggesting that the analysis was reliable. In the test of heterogeneity, the Cochran Q statistic was >0.05 for all 4 CVDs except for UF and AF, where the Cochran Q statistic was <0.05. Therefore, except for AF, we used the IVW method with fixed-effects modeling as the main analysis method. Meanwhile, MR-Egger regression indicated that horizontal pleiotropy did not influence the causal relationship between UF and CVDs (Table 2 ). The leave-one-out method confirmed that no single SNP significantly impacted the causal relationship between UF and CVDs (Figure S2, Supplemental Digital Content, http://links.lww.com/MD/O451 ). Forest plots, shown in Figure S3, Supplemental Digital Content, http://links.lww.com/MD/O451 provided a visual representation of the effect size of individual SNPs and the overall analysis. Lastly, the funnel plots were roughly symmetrical, further supporting the absence of significant horizontal pleiotropy (Figure S4, Supplemental Digital Content, http://links.lww.com/MD/O451 ). Heterogeneity test and horizontal pleiotropy test. MR = Mendelian randomization, SE = standard error. Mendelian randomization analysis of UF on the risk of CVD. AF = atrial fibrillation, CES = cardioembolic stroke, CHD = coronary heart disease, CI = confidence interval, FDR = false discovery rate, HF = heart failure, MI = myocardial infarction, OR = odds ratio, SNP = single nucleotide polymorphism. The results of reverse MR analysis indicated that CHD (OR = 0.997, 95% CI: 0.954–1.043, P  = .905, P FDR  = .905), MI (OR = 1.018, 95% CI: 0.979–1.060, P  = .371, P FDR  = .618), AF (OR = 1.014, 95% CI: 0.982–1.046, P  = .399, P FDR  = .499), HF (OR = 0.883, 95% CI: 0.758–1.029, P  = .111, P FDR  = .558), and CES (OR = 1.026, 95% CI: 0.976–1.079, P  = .307, P FDR  = .769) were not significantly associated with the risk of UF (Table 3 ). Mendelian randomization analysis of cardiovascular diseases on the risk of uterine fibroids. AF = atrial fibrillation, CES = cardioembolic stroke, CHD = coronary heart disease, CI = confidence interval, FDR = false discovery rate, HF = heart failure, MI = myocardial infarction, OR = odds ratio, SNP = single nucleotide polymorphism. Considering that hypertension, diabetes, smoking, and sedentary behavior may act as confounders affecting the causal analysis between UF and CVD, we applied the MVMR method to explore the direct effect of UF on the relationship with various CVD. The IVW analysis results (Table 4 ) indicated a significant causal association between UF and both CES (OR = 1.104, 95% CI = 1.012–1.205, P  = .027) and MI (OR = 0.935, 95% CI = 0.882–0.992, P  = .025), consistent with the findings from the UVMR analysis. Furthermore, after adjusting for confounders, no significant causal associations were found between UF and CHD, AF, or HF. Meanwhile, we performed heterogeneity tests, and the results indicated that there was no heterogeneity for CHD (Q = 135, P  = .051), HF (Q = 165, P  = .060), and CES (Q = 181, P  = .080), whereas heterogeneity was present for MI (Q = 283, P  < .05) and AF (Q = 135, P  < .05). Results of multivariate Mendelian randomization analysis. AF = atrial fibrillation, CES = cardioembolic stroke, CHD = coronary heart disease, CI = confidence interval, HF = heart failure, MI = myocardial infarction, OR = odds ratio.

For the first time, we systematically investigated the potential causal relationship between genetic susceptibility to UF and the risk of CVDs at the genetic level using MR methods. Our findings revealed a significant positive association between genetic susceptibility to UF and the risk of CES. The causal relationship between UF and MI was suggestive, but not definitive. There was insufficient evidence to support a causal association between UF and CHD, AF, and HF. After adjusting for potential confounders, associations between UF and both MI and CES were still observed. Specifically, UF was positively correlated with CES and negatively correlated with MI. Additionally, reverse MR analysis demonstrated no causal relationships between CHD, MI, AF, HF, CES, and UF. Although both men and women share several common risk factors for CVD, women also have unique risk factors, such as the use of oral contraceptives. [ 34 ] Various gestational conditions, including gestational hypertension, diabetes, and preeclampsia, may also contribute to the development of CVD in women. In addition, polycystic ovary syndrome as well as early menopause have been associated with an increased risk of CVD. [ 35 , 36 ] In recent years, there has been growing interest in the potential role of UF as a new risk factor for CVD in women. One hypothesis suggests that UF and CVD share similarities in their processes of wound healing or arterial plaque formation. Through these mechanisms, the synthesis of extracellular matrix in UF or the formation of arterial plaques may be accelerated, contributing to both UF and the hypertrophy of ventricular and vascular walls. [ 8 ] Moreover, an association has been observed between UF and several known CVD risk factors, including hypertension, obesity, serum lipid abnormalities, and carotid intima-media thickness, which may further increase the risk of CVD. [ 37 ] However, in our study, we did not find significant associations between UF and CHD, AF, or HF. This lack of association may be attributed to confounding factors inherent in observational studies. Furthermore, CVD endpoints in women with UF remain understudied. For example, hysterectomy is the mainstay of treatment for UF, and studies have reported an increased risk of CVD after hysterectomy. [ 38 ] This could have influenced the study results. Additionally, given that UF primarily affects women of reproductive age, and the risk of CVD in postmenopausal women increases over time, it is important to follow women with UF over extended periods, taking into account the normal progression of CVD. [ 8 ] UF are common in women, but ischemic stroke, particularly CES, is not a frequent complication associated with UF. Several case reports have described instances of ischemic stroke in patients with UF. For example, Higuchi et al reported a case of recurrent ischemic stroke in a middle-aged woman with UF, and suggested that UF may cause hypercoagulability in this patient, as evidenced by elevated plasma D-dimer levels and ultrasound Doppler findings of microembolic signals in the right middle cerebral artery. [ 39 ] A multicenter retrospective study from Japan found that 39 of 470 female patients with ischemic stroke or transient ischemic attack had common noncancerous gynecologic diseases, including 24 patients (62%) with UF. They were further found to have cryptogenic strokes (predominantly due to CES), nonbacterial thrombotic endocarditis, and paradoxical embolism. These conditions were largely attributed to the development of hypercoagulable states, with significantly elevated levels of CA125 and D-dimer observed in these patients. [ 40 ] CA125 is a mucin molecule and serves as a serum marker. [ 41 ] Mucins can promote thrombosis through adhesion-dependent interactions between neutrophils and platelets, as well as the bidirectional signaling pathway between them. [ 42 ] A retrospective study by Buamah, conducted over a period of 42 months, revealed elevated serum levels of CA125 in 6 patients with nonmalignant gynecological diseases, 2 of whom had UF. [ 43 ] However, some studies have indicated that CA125 is not associated with UF. For instance, Dawood et al conducted a prospective study comparing the serologic indices of 51 women with UF at more than 14 weeks of gestation with those of 30 normal childbearing women, and the results showed no difference in CA125 levels during the follicular and luteal phases between the 2 groups. [ 44 ] D-dimer, a marker of thrombin formation and fibrinolysis, can indicate thrombus formation. [ 45 ] Recent studies have found that high high D-dimer levels may point to an undetected source of cardiac embolism, such as silent AF, as the etiology of stroke in some patients. [ 46 ] In conclusion, we found a causal relationship between UF and CES, providing a theoretical basis for the early diagnosis and treatment of such patients. However, the underlying mechanisms linking UF and CES remain poorly understood. Future research should focus on exploring the potential pathogenesis of CES in patients with UF, which may facilitate timely clinical intervention. Interestingly, our study found a potential association between UF and MI, with UF appearing to reduce the risk of MI. This phenomenon may be closely related to the hormone levels in women. Estrogen and progesterone play significant roles in the development and progression of UF. They can stimulate mature leiomyoma cells to release proliferative signals, thereby promoting the transformation of surrounding undifferentiated cells into those that may support tumor growth. Progesterone is essential for the proliferation of uterine fibroid cells, while estrogen enhances cell sensitivity to progesterone by increasing the availability of progesterone receptors. [ 47 ] Studies have shown that estrogen induces rapid dilation of coronary arteries, offering vascular protective effects. Additionally, estrogen can exert anti-inflammatory effects by reducing the expression of neutrophil chemoattractant factor (CINC-2β) and monocyte chemoattractant protein (MCP-1) and inhibiting the expression of C-reactive protein. [ 48 ] Korkmaz et al, in their prospective studies, explored the risk of CVD in UF patients and found that although the number of UF correlates positively with triglycerides and LDL levels, the intima-media thickness of the carotid artery in UF patients was significantly lower, which may be related to the vascular protective effects of estrogen. [ 49 ] Although existing studies suggest a potential role of hormones in the relationship between UF and MI, further exploration of other potential mechanisms is needed to better understand the impact of UF on MI. To the best of our knowledge, this study is the first MR analysis to explore the causal association between UF and CVD. By utilizing data from a large-scale GWAS, we were able to expand the association analysis between UF and CVD at the genetic level, while avoiding the limitations inherent in traditional observational studies. The application of multiple sensitivity analysis methods further strengthened the robustness and reliability of our findings. Additionally, the MR approach benefits from the process of randomization, which enhances its methodological rigor while also being time- and cost-efficient. Our study has several limitations. First, the MR approach did not allow for the assessment of a potential nonlinear relationship between UF and CVD, and the absence of clinical data prevented us from conducting further subgroup analyses. Second, our study was limited to populations of European ancestry, which may introduce bias due to the homogeneous nature of the sample, thereby limiting the generalizability of the findings to other populations. Third, despite efforts to mitigate pleiotropic effects, some potential pleiotropy could not be excluded, as the biological functions of many genetic variants remain unclear. Finally, while we have made a preliminary assessment of the causal relationship between UF and CVD, the exact mechanisms underlying their interaction remain unclear and warrant further investigation.