Association of Smoking and Alcohol Consumption with Female Reproductive Diseases: A Mendelian Randomization Study on the Mediating Role of Circulating Metabolites

Disorders of the female reproductive system represent a major public health and societal challenge, encompassing benign and malignant tumors, infections, and endocrine abnormalities. Gynecological disorders ranked as the second most common newly diagnosed non-communicable diseases among women of reproductive age globally in 1990 and 2019, and they accounted for the highest burden in terms of disability-adjusted life years. 1 These conditions collectively exert a profound negative impact on women’s quality of life. Abnormal uterine bleeding (AUB), a common clinical symptom, is frequently associated with various gynecological disorders. At present, the lack of specific pharmacological treatments for adenomyosis imposes a significant burden on both individuals and healthcare systems, with hysterectomy remaining the primary therapeutic option. 2 , 3 Therefore, there is an urgent need to develop innovative treatment strategies. Ovarian cysts exhibit a high prevalence of 21.2% among healthy postmenopausal women in Europe, with approximately 7% of women worldwide affected at some point in their lives. 4 Notably, ovarian cancer remains the deadliest gynecological malignancy. 5 Other female pelvic inflammatory diseases and inflammatory diseases of female pelvic organs are particularly prevalent in this population. Delayed treatment for these conditions can lead to various adverse reproductive health outcomes, including infertility, chronic pelvic pain, and ectopic pregnancy. 6 The substantial personal and societal costs associated with these disorders underscore the urgent need for improved prevention and treatment strategies. Women who smoke or consume alcohol in excessive amounts face significantly higher risks to their reproductive health. 7 , 8 These behaviors are also well-established as major contributors to premature mortality. 9–11 A cross-sectional study has shown that combined exposure to smoking and alcohol consumption is significantly associated with an increased risk of dysmenorrhea. 12 In addition, both active smoking and exposure to secondhand smoke exhibit a clear dose–response relationship with the risk of early menopause. Mechanistic studies have further indicated that polycyclic aromatic hydrocarbons present in cigarette smoke can directly induce apoptosis in human ovarian follicular cells through activation of the aryl hydrocarbon receptor, thereby accelerating the decline of ovarian function. 13 In addition, nicotine, a principal bioactive component of tobacco, has been implicated in adverse effects on ovarian and female reproductive tract function. Studies have shown that nicotine not only induces oxidative stress (OS) and apoptotic processes but may also lead to chromosomal segregation abnormalities, impair oocyte development, and damage ovarian morphology and function. 14 Furthermore, it inhibits estrogen production in women, potentially causing physiological dysfunction and increasing the risk of tumor development. 15 Alcohol consumption has been identified as a significant risk factor for chronic diseases and injuries. 16 However, the evidence from traditional observational studies remains inconsistent and is often unable to definitively establish causality. For example, while some studies suggest smoking increases the risk of gynecological infections, 8 the direct causal link is weakened by potential confounding factors such as socioeconomic status and other lifestyle choices. Consequently, whether these harmful behaviors are primary causal factors for specific gynecological disorders remains a critical, unanswered question. Smoking, as an important and potentially modifiable environmental exposure, has been shown to induce extensive systemic metabolic remodeling, primarily involving key pathways related to amino acid metabolism, lipid metabolism, and mitochondrial energy metabolism. 17 Untargeted metabolomics studies have further demonstrated that smoking profoundly disrupts carnitine-mediated fatty acid β-oxidation and activates the tryptophan–kynurenine pathway as well as sphingolipid metabolic networks, 18 suggesting a substantial impact on the regulation of energy homeostasis and oxidative stress. Notably, these smoking-associated metabolic perturbations have been repeatedly observed across multiple gynecological disorders, indicating a potential shared metabolic basis. For instance, elevated circulating acylcarnitine levels are considered a hallmark metabolic feature of endometriosis, reflecting impaired mitochondrial fatty acid oxidation. 19 In addition, tryptophan metabolism plays a pivotal immunoregulatory role in infectious diseases of the female reproductive tract: host cells can restrict tryptophan availability via the IFN-γ–indoleamine 2,3-dioxygenase pathway to suppress the growth of pathogens such as Chlamydia trachomatis, whereas under conditions of vaginal dysbiosis, including bacterial vaginosis, indole produced by anaerobic bacteria can facilitate pathogen-driven tryptophan resynthesis, thereby promoting immune evasion and increasing the risk of persistent infection and adverse reproductive outcomes. 20 Placental levels of polyunsaturated fatty acids are significantly elevated in patients with preeclampsia compared with normotensive pregnant women, and that the degree of lipid metabolic dysregulation is closely associated with disease severity. 21 Although substantial evidence indicates that smoking profoundly reshapes the circulating metabolomic profile and that metabolic dysregulation represents a shared feature of multiple gynecological disorders, it remains unclear whether these metabolic alterations function as causal mediators or merely reflect secondary consequences of disease states. Conventional observational studies are inherently limited in their ability to disentangle mediation effects from confounding or reverse causation, and typically require prolonged follow-up periods and substantial resource investment. Mendelian randomization (MR)–based mediation analysis provides a unique framework for formally testing whether genetic susceptibility to smoking-related behaviors influences diseases of the female reproductive system through specific circulating metabolites. By leveraging genetic variants identified from genome-wide association studies (GWAS) as instrumental variables (IVs), this approach effectively circumvents confounding bias and reverse causation that commonly affect observational research, thereby substantially strengthening the robustness of causal inference. 22–24 By overcoming biases inherent to conventional observational studies, MR offers clear advantages in controlling for confounding and reverse causation. To date, such metabolite-centered mediation pathways have not been systematically evaluated in the context of female reproductive health. Therefore, in the present study, a two-step MR design was employed to quantitatively assess the extent to which the effects of smoking- and alcohol-related behaviors on gynecological diseases are mediated through circulating metabolites, with the aim of identifying novel, biologically plausible pathways linking behavioral risk factors to the onset and progression of gynecological disorders. In this study, we hypothesize that potential causal relationships exist among health risk behaviors, metabolites, and gynecological disorders. Our objective is to elucidate these associations, aiming to identify modifiable risk factors that influence the development of these diseases/symptoms and to determine metabolite targets with potential value for early diagnosis and treatment.

During the selection of IVs, 11 to 248 SNPs were identified for smoking- and alcohol-related phenotypes. The F-statistics of these SNPs ranged from 67.06 to 5608.46, strongly indicating the absence of weak instrument bias in this study ( Supplementary Tables S3 , S7 and S8 ). We conducted a two-sample MR analysis to investigate the impact of related to smoking and alcohol consumption on emale reproductive disorders. Figure 2 shows three methods that were used to estimate IVs: IVW, WM, and MR-Egger regression. By analyzing the interactions between these inputs and outcomes, the study seeks to confirm the reliability of the inferred causal relationships. The genetic predisposition to smoking initiation was found to have a positive causal association with the risk of five female reproductive disorders: AUB (OR = 1.33, 95% CI [1.10–1.61], P = 0.003), endometrioid ovarian cancer (OR = 1.50, 95% CI [1.03–2.19], P = 0.034), ovarian cysts (OR = 1.38, 95% CI [1.20–1.57], P = 3.31E-06), other female pelvic inflammatory diseases (OR = 1.92, 95% CI [1.51–2.43], P = 8.21E-08), and inflammatory diseases of female pelvic organs (OR = 1.53, 95% CI [1.35–1.73], P = 8.98E-12). Cigarettes per day was also positively associated with adenomyosis (OR = 1.69, 95% CI [1.20–2.39], P = 0.003), whereas weekly alcohol consumption exhibited a negative association with endometrioid ovarian cancer (OR = 0.48, 95% CI [0.31–0.74], P = 0.001). Figure 2 Mendelian randomization plots for the relationship of smoking- and alcohol-related phenotypes with female reproductive disorders. ( A ) Circular heatmap comparing results across three primary analytical methods. ( B ) Forest plot illustrating causal effect from the IVW method. Abbreviations : OR, odds ratio; CI, confidence interval. SmkInit, smoking initiation; AgeSmk, age of smoking initiation; CigDay, cigarettes per day; SmkCes, smoking cessation; DrnkWk, drinks per week; AUB, abnormal uterine bleeding; EnOC, endometrioid ovarian cancer; OTHFEMPELINF, other female pelvic inflammatory diseases; FEMALEGENINF, inflammatory diseases of female pelvic organs; IVW, inverse-variance weighted; WM, Weighted Median. Mendelian randomization plots for the relationship of smoking- and alcohol-related phenotypes with female reproductive disorders. ( A ) Circular heatmap comparing results across three primary analytical methods. ( B ) Forest plot illustrating causal effect from the IVW method. To identify candidate mediators, we screened 168 circulating metabolites measured as absolute concentrations using a stepwise MR-based approach ( Figure 3 ). First, UVMR analyses were performed to evaluate the associations between each metabolite and female reproductive disorders. Metabolites showing no evidence of an association were excluded at this stage. Second, MVMR analyses were conducted to assess whether the remaining metabolites exerted direct effects on female reproductive disorders independent of smoking-related behaviors (smoking initiation or cigarettes per day). Metabolites whose associations were fully attenuated after adjustment were excluded. Third, only metabolites that were causally associated with smoking-related behaviors were retained, ensuring consistency with the hypothesized exposure–mediator pathway. Fourth, directional consistency between the exposure–mediator and mediator–outcome associations was evaluated. Following the mediator selection process, 10 variables were identified from 168 candidates as meeting all criteria. These variables were selected as mediators linking cigarettes per day to adenomyosis and smoking initiation to inflammatory diseases of female pelvic organs. Figure 3 Screening process for mediators in the causal relationship between smoking and alcohol-related phenotypes and female reproductive disorders. 10 mediators qualified for all criteria were included in the mediation analyses. Abbreviations : VLDL, very low-density lipoprotein; LDL, low-density lipoprotein; HDL, high-density lipoprotein; HDL-C, high-density lipoprotein cholesterol; IDL, intermediate-density lipoprotein. Screening process for mediators in the causal relationship between smoking and alcohol-related phenotypes and female reproductive disorders. 10 mediators qualified for all criteria were included in the mediation analyses. MVMR was applied to assess the independent effects of 10 plasma metabolites on two female reproductive disorders. After adjustment for smoking-related phenotypes, the effects of these metabolites remained statistically significant ( Figure 4 ). Sensitivity analyses confirmed the validity of these IVW estimates ( Supplementary Tables S4 – S6 ). Specifically, five metabolites were identified as significantly associated with adenomyosis based on genetic predictions: glycoprotein acetyls (GlycA) (OR = 1.25, 95% CI [1.09–1.50], P = 0.005), total lipids in large VLDL (OR = 1.25, 95% CI [1.14–1.54], P = 0.019), concentration of large VLDL particles (OR = 1.19, 95% CI [1.14–1.54], P = 0.03), cholesteryl esters in very large VLDL (OR = 1.22, 95% CI [1.13–1.49], P = 0.015), and concentration of very large VLDL particles (OR = 1.19, 95% CI [1.17–1.58], P = 0.026). Similarly, five metabolites were significantly associated with inflammatory diseases of female pelvic organs: triglycerides in intermediate-density lipoprotein (IDL) (OR = 1.07, 95% CI [1.03–1.14], P = 0.002), triglycerides in large low-density lipoprotein (LDL) (OR = 1.08, 95% CI [1.04–1.15], P = 0.001), triglycerides in LDL (OR = 1.10, 95% CI [1.05–1.17], P = 2.85E-04), triglycerides in medium LDL (OR = 1.08, 95% CI [1.03–1.16], P = 0.003), and triglycerides in very small VLDL (OR = 1.07, 95% CI [1.03–1.14], P = 0.001). All estimates obtained through MV-IVW were supported by sensitivity analyses. Figure 4 Mediating role of each mediator in the causal association between smoking and alcohol-related phenotypes and the female reproductive disorders. The left panel presents UVMR estimates for the causal effects of smoking and alcohol-related phenotypes on each mediator. The right panel shows MVMR estimates for the causal effects of each mediator on female reproductive disorders, adjusted for smoking. MR estimates were obtained using the IVW method for UVMR and the MV-IVW method for MVMR. The data are reported as β coefficients and OR, accompanied by their respective 95% CIs. Results with P  < 0.05 were considered statistically significant. Abbreviations : CigDay, cigarettes per day; SmkInit, smoking initiation; VLDL, very low-density lipoprotein; IDL, intermediate-density lipoprotein; LDL, low-density lipoprotein; FEMALEGENINF, inflammatory diseases of female pelvic organs. Mediating role of each mediator in the causal association between smoking and alcohol-related phenotypes and the female reproductive disorders. The left panel presents UVMR estimates for the causal effects of smoking and alcohol-related phenotypes on each mediator. The right panel shows MVMR estimates for the causal effects of each mediator on female reproductive disorders, adjusted for smoking. MR estimates were obtained using the IVW method for UVMR and the MV-IVW method for MVMR. The data are reported as β coefficients and OR, accompanied by their respective 95% CIs. Results with P  < 0.05 were considered statistically significant. GlycA was notably identified as a mediator in the causal relationship between cigarettes per day and adenomyosis, with a mediation proportion of 7.61%. Additional mediation proportions were observed for total lipids in large VLDL (3.05%), the concentration of large VLDL particles (2.76%), cholesteryl esters in very large VLDL (2.86%), and the concentration of very large VLDL particles (3.04%). Furthermore, triglycerides in IDL mediated 1.08% of the causal relationship between smoking initiation and inflammatory diseases of female pelvic organs. Mediation proportions were also identified for triglycerides in large LDL (1.39%), LDL (1.77%), medium LDL (1.34%), and very small VLDL (1.12%) ( Figure 5 ). An increase in cigarettes per day was significantly associated with elevated levels of GlycA, total lipids in very large VLDL, cholesteryl esters in very large VLDL, and the concentration of both large and very large VLDL particles ( P < 0.05). Similarly, an increase in smoking initiation was significantly associated with elevated triglyceride levels in IDL, large LDL, medium LDL, and very small VLDL ( P < 0.05) ( Figure 4 ). Figure 5 Two-Step MR framework and mediation proportions of circulating metabolites in the causal associations between smoking and alcohol-related phenotypes and female reproductive disorders. Abbreviations : CigDay, cigarettes per day; SmkInit, smoking initiation; VLDL, very low-density lipoprotein; IDL, intermediate-density lipoprotein; LDL, low-density lipoprotein; FEMALEGENINF, inflammatory diseases of female pelvic organs. Two-Step MR framework and mediation proportions of circulating metabolites in the causal associations between smoking and alcohol-related phenotypes and female reproductive disorders.

Genetic instruments for smoking- and alcohol-related traits were obtained from large-scale genome-wide association studies conducted by the GWAS and Sequencing Consortium of Alcohol and Nicotine (GSCAN), which provides well-characterized summary statistics for behavioral exposures. Outcome GWAS data were derived from the FinnGen R11 consortium, selected for its large sample size and relatively homogeneous European ancestry, which helps minimize population stratification bias. To systematically explore potential causal relationships, a two-sample MR approach was applied to examine the impact of smoking and alcohol consumption on various female reproductive disorders, including AUB, adenomyosis, endometrioid ovarian cancer, ovarian cysts, other female pelvic inflammatory diseases, and inflammatory diseases of female pelvic organs. Alcohol consumption, measured as drinks per week, was included as a related behavioral exposure because smoking and alcohol use are genetically and behaviorally correlated risk factors. This design allowed us to assess whether the observed associations were specific to smoking-related traits rather than reflecting shared lifestyle effects. The mediating role of circulating metabolites was also assessed. Since this study relied exclusively on summary-level data, no ethical approval was required. The study framework is shown in Figure 1 , highlighting the three fundamental assumptions required for causal inference in MR. 25 Specifically, to estimate causal effects using genetic variation, three fundamental assumptions of IVs must be met: (1) demonstrate a strong association with the exposure; (2) influence the outcome exclusively through the exposure pathway; and (3) be independent of confounders in the exposure-outcome relationship. Our MR study was performed following the STROBE-MR guidelines, 26 ( Supplementary Table S1 ). Data were extracted between June 1, 2024, and October 30, 2023. Figure 1 Overview of the study design. A two-sample MR framework was used to assess the causal effects of smoking- and alcohol-related behaviors on female reproductive diseases and to evaluate circulating metabolites as potential mediators. Smoking- and alcohol-related traits were analyzed as exposures, female reproductive disorders as outcomes, and 168 circulating metabolites were screened within a two-step MR mediation framework. Detailed data sources and sample characteristics are provided in Supplementary Table S2 . Abbreviations : SmkInit, smoking initiation; AgeSmk, age of smoking initiation; CigDay, cigarettes per day; SmkCes, smoking cessation; DrnkWk, drinks per week; AUB, abnormal uterine bleeding; EnOC, endometrioid ovarian cancer; OTHFEMPELINF, other female pelvic inflammatory diseases; FEMALEGENINF, inflammatory diseases of female pelvic organs; IVW, inverse-variance weighted; WM, Weighted Median; MR, Mendelian randomization. Overview of the study design. A two-sample MR framework was used to assess the causal effects of smoking- and alcohol-related behaviors on female reproductive diseases and to evaluate circulating metabolites as potential mediators. Smoking- and alcohol-related traits were analyzed as exposures, female reproductive disorders as outcomes, and 168 circulating metabolites were screened within a two-step MR mediation framework. Detailed data sources and sample characteristics are provided in Supplementary Table S2 . Five smoking- and alcohol-related phenotypes were included as exposure variables: smoking initiation, smoking cessation, cigarettes per day, age of smoking initiation, and drinks per week. The data were sourced from the GWAS and Sequencing Consortium of Alcohol and Nicotine. 27 The GWAS of smoking- and alcohol-related phenotypes were conducted predominantly in populations of European ancestry. There was no overlap among participants across the datasets included in this study. Additional details regarding these phenotypes are provided in the Supplementary Materials ( Supplementary Table S2 and Supplementary Box 1 ). Given the large sample sizes of the exposure and outcome GWAS datasets (ranging from approximately 40,000 to over 2 million participants), this study was well powered to reliably estimate causal effects using MR. Previous studies 28 have demonstrated that MR analyses based on large-scale GWAS data can yield precise and robust causal inferences. GWAS data for circulating metabolites, used as mediators, were sourced from the Nightingale Health Study, 29 which involved 12,100 participants of European descent. This dataset includes 249 circulating metabolites, encompassing lipoproteins, lipids, fatty acids, fatty acid composition, and various low-molecular-weight metabolites, such as amino acids, ketone bodies, and glycolytic metabolites. Absolute concentrations of 168 biomarkers were analyzed, whereas ratios for 81 biomarkers were excluded, consistent with the study by Hu et al. 30 Complete GWAS statistics are publicly available through the IEU Open-GWAS project database. Outcome GWAS data were obtained from the FinnGen R11 consortium, covering adenomyosis (ICD-10 code N800, 5130 cases, 119,468 controls), AUB (ICD-10 code N93, 12,229 cases, 119,468 controls), ovarian cysts (ICD-10 code N83[0-2], 25,564 cases, 119,468 controls), other female pelvic inflammatory diseases (ICD-10 code N73, 6181 cases, 226,439 controls), and inflammatory diseases of female pelvic organs (ICD-10 code N7, 28,179 cases, 226,439 controls). Data for endometrioid ovarian cancer were retrieved from the IEU Open-GWAS project database. 31 The selection of IVs for analysis followed these steps: First, genetic instruments for the exposures were required to achieve genome-wide significance ( P < 5e-8) and exhibit independence from others (linkage disequilibrium r 2 < 0.001 within a 10,000-kb window). When shared SNPs between the exposure and outcome were unavailable, proxy SNPs (r 2 ≥ 0.8) from the 1000 Genomes European reference panel were used. Second, SNPs significantly associated with the outcome ( P < 5e-8) were excluded. Third, data were harmonized to ensure consistent allelic effects between exposure and outcome by excluding palindromic SNPs with intermediate allele frequencies. Finally, SNPs with an F-statistic < 10, indicating weak instrument strength, were excluded to minimize weak instrument bias. 32 This study aimed to evaluate the following: (1) the relationship between smoking and drinking behaviors and female reproductive disease outcomes; (2) the association of smoking and drinking behaviors with individual circulating metabolites; (3) the link between each circulating metabolite and female reproductive disease outcomes; and (4) the direct association of metabolites with female reproductive disease outcomes after adjusting for smoking-related exposure factors (cigarettes per day, smoking initiation). Univariable MR (UVMR) was used for the first three analyses, while Multivariable MR, (MVMR) was employed for the fourth. The validity of the MR estimates relies on three core assumptions: relevance, independence, and exclusion restriction. In this study, the relevance assumption was addressed by selecting genetic instruments that reached genome-wide significance and excluding weak instruments (F-statistic < 10). The independence and exclusion restriction assumptions were evaluated by systematically assessing heterogeneity and horizontal pleiotropy using Cochran’s Q statistics, MR-Egger intercept tests, Radial MR, and leave-one-out analyses, thereby supporting the robustness of the causal inferences. In the UVMR analysis, the random-effects inverse-variance weighted (IVW) method served as the primary approach for causal inference. This method assumes all genetic variants are valid instruments, offering high efficiency in MR estimation; however, it is sensitive to pleiotropy-related bias. To strengthen result robustness, four complementary methods were applied: Weighted Median (WM), MR-Egger, Weighted Mode, and Simple Mode. For MVMR analysis, the multivariable IVW (MV-IVW) method was used to assess the independent causal effects of mediators on outcomes. Results were considered statistically significant when P < 0.05. For binary outcomes, odds ratios (OR) with 95% confidence intervals (CIs) were reported, while for continuous variables, β coefficients and standard errors (SE) were presented. To assess the robustness and potential biases of our findings, we implemented several sensitivity analyses: the Cochran Q statistic, the MR-Egger intercept test, the Radial MR method, and a leave-one-out SNP analysis. Initially, the Cochran Q test was used to assess heterogeneity in the results, with P ≥ 0.05 indicating no significant heterogeneity. 33 Horizontal pleiotropy was assessed using the P-value of the MR-Egger intercept test, where P ≥ 0.05 indicated no evidence of horizontal pleiotropy. 34 To examine the influence of individual SNPs on causal associations, a leave-one-out analysis was performed by sequentially removing each SNP and reanalyzing the data. The Radial MR method was then used to detect and exclude outliers potentially affected by pleiotropic bias. 35 Observations impacted by such biases were detected and removed ( Supplementary Figures S1 – S16 ). All analyses were conducted using the TwoSampleMR (v0.6.8), RadialMR (v1.1), and MendelianRandomization (v0.10.0) packages within R software (v4.4.0). The mediating role of circulating metabolites in the causal pathway connecting smoking, alcohol-related phenotypes, and female reproductive diseases was assessed using a two-step MR analysis. 36 In the first step, UVMR was used to estimate the causal effects of each smoking and drinking phenotype on circulating metabolites, with the estimates denoted as β 1 . In the second step, MVMR was employed to evaluate the causal effects of each mediator on female reproductive diseases, with separate adjustments for the effects of smoking initiation and cigarettes per day ( β 2 ). This analysis was performed only if the mediator had shown a causal association with female reproductive diseases in UVMR. The mediation proportion of each metabolite in the relationship between smoking, alcohol-related phenotypes, and female reproductive diseases was calculated as the product of β 1 and β 2 , divided by the total effect of smoking and alcohol-related phenotypes on female reproductive diseases. The standard error was calculated using the delta method. 37

This MR study provides novel genetic evidence supporting potentially causal roles of smoking- and alcohol-related behaviors in a range of female reproductive disorders. The principal contribution of this work lies in the identification of specific circulating metabolites as mechanistic intermediates, highlighting mediator specificity across distinct disease contexts. In particular, GlycA and VLDL-related lipid markers partially mediated the effect of cigarette consumption on adenomyosis, suggesting a potential role of inflammation-related lipid remodeling, whereas triglyceride-rich lipoprotein subclasses mediated the association between smoking initiation and inflammatory diseases of the female pelvic organs, pointing to a metabolically driven inflammatory pathway. From a clinical perspective, these findings extend beyond established epidemiological associations by suggesting that smoking-related reproductive risk may be stratified according to distinct metabolic signatures. Such profiles may aid in the identification of high-risk individuals and support more targeted prevention strategies that integrate smoking cessation with monitoring of downstream inflammatory and lipid metabolic dysregulation. Although both smoking and alcohol consumption were included as exposure variables in the present study, potential interaction effects between these behaviors were not explicitly evaluated. Future studies applying interaction-aware or stratified MR frameworks may help to clarify whether combined behavioral exposures modify the risk of female reproductive disorders.

In this large-scale MR study, we provide novel genetic evidence that strengthens the link between smoking, alcohol consumption, and the risk of several female reproductive disorders. Our analysis makes three primary contributions. First, moving beyond traditional observational data, we offer more robust support for a potential causal relationship between smoking behaviors and specific gynecological diseases. For instance, genetically predicted daily cigarette consumption was associated with an increased risk of adenomyosis, while smoking initiation was linked to higher risks of abnormal uterine bleeding, endometrioid ovarian cancer, and ovarian cysts. Second, and most notably, we identified and quantified the role of specific circulating metabolites as mediators in these pathways. We found that glycoprotein acetyls (GlycA) and VLDL-related lipids partially mediated the effect of smoking on adenomyosis, while triglycerides in various lipoprotein subclasses mediated the link between smoking initiation and inflammatory diseases. From a public health perspective, our findings suggest several actionable implications. First, the robust causal associations observed for smoking initiation across multiple gynecological outcomes underscore the importance of preventing smoking uptake, particularly among adolescents and young women, as a priority for reproductive health promotion. Second, the dose-dependent association between cigarette consumption and adenomyosis indicates that reducing smoking intensity may still confer potential benefits for women who already smoke, supporting harm reduction strategies alongside smoking cessation efforts. Importantly, the identification of inflammatory and lipid-related metabolites as partial mediators highlights systemic metabolic dysregulation as a potential biological pathway linking smoking to gynecological disease risk. These findings suggest that monitoring inflammatory and lipid profiles may help identify women at elevated risk and inform more targeted lifestyle or metabolic interventions. Collectively, our results support integrated public health strategies that address both smoking behaviors and downstream metabolic consequences, rather than focusing on smoking exposure alone. Adenomyosis is estimated to affect 15% to 20% of young women, with up to 80% of those with infertility due to endometriosis also presenting with adenomyosis. 38 Its high prevalence underscores the critical importance of disease screening and prevention. Smoking has been recognized as a significant risk factor for adenomyosis, but the evidence supporting this association remains controversial. For instance, a cross-sectional study conducted in Nepal 39 found that smokers had a 1.4-fold higher risk of adenomyosis, particularly among individuals who had smoked for more than 10 years ( P = 0.008). Similarly, a retrospective study reported that former smoking increased the likelihood of adenomyosis by 40% (case group vs. hysterectomy control group: OR = 1.4, 95% CI [1.0–2.0]; case group vs. population control group: OR = 1.4, 95% CI [1.0–1.9]). 40 In contrast, another study suggested that smoking women were less likely to develop adenomyosis 41 or that no association existed between smoking and the condition. 42 The heterogeneity of findings across observational studies may partly reflect differences in study design and population selection. Diagnosis of adenomyosis often depends on imaging or surgical criteria and is therefore prone to bias arising from healthcare access, health-seeking behavior, and surgical decision-making, while key confounders such as reproductive history and prior uterine surgery are not consistently controlled. Using a MR framework to proxy lifelong smoking exposure, our study mitigates these limitations and provides genetic evidence supporting a positive causal association between daily cigarette consumption and increased adenomyosis risk. Uterine adenomyosis is a potential cause of AUB, accounting for 20% to 35% of AUB cases. 43 It is not uncommon for a woman to suffer from AUB at some point during her lifetime. 44 Vascular endothelial growth factor(VEGF) and angiogenesis appear to play a critical role in the pathogenesis of adenomyosis. 45 Furthermore, studies have shown that smoking increases the levels of hormones such as dehydroepiandrosterone, progesterone, 17-hydroxyprogesterone, and cortisol, 46 , 47 while no significant changes in urinary cortisol levels have been observed 24 hours after smoking. This discrepancy may be attributed to adaptive changes in metabolic clearance. Previous studies have suggested that smoking-related alterations in the hormonal milieu may compromise endometrial homeostasis, promoting aberrant inflammatory responses and cytokine secretion, thereby increasing the risk of AUB. 48 However, such mechanistic evidence is largely derived from experimental models or observational data and remains limited in its ability to disentangle the direct effects of smoking from those of other endocrine or behavioral factors. Using a MR approach, our study demonstrates a consistent positive association between genetically predicted smoking initiation and increased AUB risk, providing genetic support for a potential causal role of smoking in the development of AUB and suggesting that smoking may influence endometrial homeostasis through multiple biological pathways, including angiogenesis and hormonal regulation. Previous studies have reported that smoking is associated with an increased risk of invasive and borderline mucinous ovarian cancer, whereas its associations with other histological subtypes remain less clear. 49 This subtype-specific pattern suggests that smoking-related carcinogenic effects may not be uniformly present across different ovarian cancer types. In the present study, MR analyses revealed a significant positive association between genetically predicted smoking initiation and the risk of endometrioid ovarian cancer. Although this finding is not entirely concordant with some prior observational evidence, it underscores the potential contribution of methodological differences. Conventional epidemiological studies are often influenced by reproductive history, hormonal exposure, metabolic status, and screening behaviors, and subtype-stratified analyses are frequently constrained by limited sample sizes, which may obscure true associations within specific subtypes. By contrast, MR leverages genetic instrumental variables to capture lifelong susceptibility to smoking behavior, thereby providing complementary genetic evidence to support a potential causal relationship between smoking and endometrioid ovarian cancer risk. For the first time, our MR analysis provides genetic evidence suggesting a risk-increasing effect of smoking initiation on ovarian cysts. Previous investigations have suggested that tobacco consumption is associated with an increased formation of ovarian cysts. 50 Follicular phase levels of luteinizing hormone and follicle-stimulating hormone are reported to be higher in smokers than in non-smokers. 51 Follicle-stimulating hormones crucial in the early stages of cyst formation, and sustained stimulation by follicle-stimulating hormone and luteinizing hormone has been shown to induce ovarian follicular cysts in rats. 52 OS can impact various physiological functions within the reproductive system, such as the follicular fluid environment, folliculogenesis, and steroidogenesis. 53 Among smokers, significantly elevated levels of antioxidant enzyme expression have been detected in granulosa cells, suggesting that smoking may induce OS responses. 54 An animal study has shown that an imbalance between oxidants and antioxidants can disrupt normal ovulatory mechanisms. 55 Furthermore, increased concentrations of 8-iso-prostaglandin F2α have been observed in smokers, 56 , 57 which may be linked to the formation of ovarian cysts. Smoking weakens overall immunity in women, increasing their susceptibility to various infections, 58 , 59 and raises the risk of developing pelvic inflammatory disease. 60–62 Balkus et al 63 proved that smoking is independently associated with an increased risk of genital mycoplasma infection (AHR = 3.02, 95% CI [1.32–6.93]). The microbial profile of the reproductive organs in women who smoke is characterized by an increase in Enterobacteriaceae, Gram-positive cocci, and anaerobes, elevated levels of viral infections, and a significant reduction in protective microorganisms. 64 Numerous smoking-related compounds have been detected in the cervical mucus of smokers. 65 Trace levels of benzo[a]pyrene diol epoxide have also been found in vaginal secretions, where benzo[a]pyrene diol epoxide activates latent bacteriophages in Lactobacillus, decreasing the abundance of protective vaginal Lactobacillus species. 66 Nicotine, a specific agonist of nicotinic acetylcholine receptors, may modulate inflammatory responses by binding to nicotinic acetylcholine receptors in reproductive tract tissues. 67 Consistent with our findings, smoking initiation was associated with increased risks of female pelvic and pelvic organ inflammatory diseases, implicating immune dysregulation and reproductive tract microenvironmental imbalance as potential pathways. A key contribution of our study is the exploration of potential mechanisms linking smoking to adenomyosis through the lens of metabolomics. Our mediation MR analysis identified GlycA, a robust biomarker of systemic inflammation, along with several VLDL-related lipid measures, as potential mediators. This finding provides a molecular bridge connecting a behavioral exposure (smoking) to a disease outcome. From a biological perspective, smoking acts as a pro-inflammatory stimulus that upregulates multiple acute-phase proteins, leading to elevated GlycA levels. 68 , 69 As a composite marker reflecting the glycosylation patterns of acute-phase proteins, GlycA has been widely used to capture chronic low-grade systemic inflammation. 70 Reports of increased inflammatory cytokines, including IL-6 and IL-1β, within adenomyotic lesions. 71 Lend biological plausibility to a link between systemic inflammation, elevated GlycA, and disease risk. However, because our analyses evaluated genetically predicted effects on circulating metabolic profiles, they do not permit direct inference regarding local inflammatory intensity or tissue-specific signaling dynamics at lesion sites. Beyond inflammation, the observed mediation by VLDL-related lipid traits implicates lipid transport and metabolic homeostasis in smoking-associated adenomyosis risk. Prior evidence suggests that hypoxia- and inflammation-related pathways can promote angiogenesis and influence lipoprotein uptake and lipid accumulation, 72–74 offering a potential interpretative framework for the involvement of VLDL measures. Nevertheless, these angiogenic and hypoxic pathways were not directly assessed in the present study and should therefore be regarded as mechanistic hypotheses rather than empirically validated mechanisms. Collectively, our findings extend the association between smoking and adenomyosis from the level of behavioral exposure to quantifiable signatures of systemic inflammation and lipid metabolism, supporting a potential role for inflammation-associated lipid remodeling in smoking-related adenomyosis risk. Further analysis revealed that the risk effect of smoking initiation on inflammatory diseases of female pelvic organs is partially mediated by an increase in triglyceride content across five lipoprotein types. Remarkably, the effect sizes for different lipoproteins were similar, likely due to the strong intercorrelations among these subtypes. Previous studies have shown an association between smoking and alterations in various metabolites, 75 , 76 particularly a positive correlation with triglycerides and LDL cholesterol. This phenomenon may result from mitochondrial dysfunction induced by smoke-derived extracts, which suppress sirtuin 3 activity, leading to its inhibition and subsequent hyperacetylation of superoxide dismutase, impairing its ability to eliminate reactive oxygen species. This disruption triggers a systemic OS response. 77 LDL is the primary transport form of cholesterol and is highly susceptible to oxidation under OS, resulting in the formation of oxidized LDL. OxLDL promotes lipid accumulation in macrophages, leading to foam cell formation, and induces the expression of pro-inflammatory cytokines. It also engages the Toll-like receptor 4 signaling pathway, contributing to chronic inflammation and the initiation of infections. 67 , 78 This study has several notable strengths. By leveraging large-scale GWAS datasets and multiple complementary MR approaches, we minimized confounding and reverse causation, thereby strengthening causal inference. Furthermore, the integration of multivariable and mediation MR enabled the identification and quantification of specific circulating metabolites as partial mediators, thereby providing mechanistic insights that extend beyond conventional associations between exposures and disease outcomes. Comprehensive sensitivity analyses consistently supported the robustness of the findings. Nevertheless, several limitations should be acknowledged. First, MR estimates reflect the effects of lifelong genetic predisposition to smoking-related behaviors and may not directly translate to the short-term effects of behavioral modification or clinical intervention. Second, the mediation proportions identified in this study were modest, which is expected given the complex and multifactorial etiology of gynecological disorders. The detected metabolites likely represent only a subset of biological pathways linking smoking to disease risk. Third, circulating metabolites measured in peripheral blood may not fully capture tissue-specific metabolic alterations within the female reproductive organs, and thus should be interpreted as systemic metabolic signatures rather than direct indicators of local pathology. Although alcohol consumption was included as a behavioral exposure in this study, its associations with female reproductive disorders were comparatively limited and less consistent than those observed for smoking-related traits. Specifically, genetically predicted alcohol consumption was inversely associated with endometrioid ovarian cancer risk, while no robust causal associations were detected for other gynecological outcomes. Importantly, alcohol-related traits did not give rise to identifiable metabolite-mediated pathways in the present mediation analyses. These findings suggest that the pathogenic pathways linking smoking to female reproductive disorders may not simply reflect shared lifestyle or behavioral risk patterns, but instead involve smoking-specific biological mechanisms. Given the strong genetic and behavioral correlation between smoking and alcohol use, the inclusion of alcohol consumption as a comparator exposure strengthens the inference that the observed metabolic mediation effects are more likely attributable to smoking-related processes rather than generalized unhealthy behaviors. Several explanations may account for the comparatively weaker alcohol-related signals observed. Genetic instruments for alcohol consumption capture long-term average intake rather than patterns of heavy or episodic drinking, which may be more relevant for gynecological outcomes. In addition, alcohol-related effects on female reproductive health may operate through pathways distinct from the inflammatory and lipid metabolic signatures identified here, or may depend on complex interaction effects that are not readily captured by the current MR framework. Further studies incorporating refined alcohol phenotypes and interaction-aware designs are warranted. In addition, several well-established clinical risk factors for adenomyosis, such as prior caesarean section or other uterine surgical procedures, could not be accounted for in the present analysis, as such information is not available in summary-level GWAS datasets. This represents an inherent limitation of MR studies relying on publicly available epidemiological data. Finally, given that the present study was primarily conducted in populations of European ancestry, the generalizability of our findings to other populations and to more granular clinical subtypes remains to be established. Future studies are therefore warranted to validate these results in larger and more diverse independent datasets, thereby strengthening the robustness and credibility of the conclusions.