Causal pathways linking polycystic ovary syndrome to distinct breast cancer subtypes through mediator factors: a multivariable mendelian randomization analysis

PCOS is a prevalent endocrine and metabolic disorder among women of reproductive age. Chronic anovulation and hyperandrogenism are hallmarks of PCOS, leading to clinical features such as amenorrhea, obesity, infertility, and hirsutism, which are also considered to be risk factors for BC development. The causality between PCOS and BC remains controversial, with conflicting results from previous epidemiological studies. While a Danish cohort study suggests that PCOS may increase postmenopausal BC risk [1], another retrospective cohort study indicates that it may elevate the risk of BC generally [2]. In contrast, a Taiwan cohort study has found no significant causality between PCOS and BC [3], and the largest and most recent meta-analysis conducted to date has produced consistent results [4]. In studies that have concluded PCOS leads to an increased risk of developing BC, the exact mechanisms involved are still unclear. Thus far, there have been no large-scale prospective longitudinal cohort studies carried out to investigate this issue.

The rapid development of genetics has led to the emergence of mendelian randomization (MR) in medical research. For one thing, MR studies can produce more reliable results than observational studies. First, MR follows Mendel's second law of inheritance, which states that alleles are randomly assigned at conception, avoiding the impact of residual confounding factors and reverse causation for being affected by other factors in the disease process. Second, MR can provide the impact of lifetime exposure of risk factors on outcomes. Third, genetic variants are measured with sufficient precision to avoid bias due to measurement error. For another, MR and randomized controlled trial (RCT) work on similar principles. High-quality RCTs represent the highest level of evidence. However, RCTs are often quite difficult and expensive to perform due to harsh implementation conditions and possible ethical issues. Therefore, when observational studies fail to come a consistent conclusion and RCTs are not feasible, MR may be the best method to explore causal relationships.

Two-sample MR was conducted to estimate causal relationships between one exposure and one outcome from two different samples. Estrogen receptor-positive breast cancer (ER + BC) is a hormone-dependent tumor, whereas PCOS, as an endocrine disruption disease, has proven by previous two-sample MR studies that PCOS may be a factor in the development of BC, particularly in ER + BC [5‐7]. While in situations where multiple exposures are likely to be present and interact with each other, multivariable MR (MVMR) may be a more useful tool for investigating complex relationships, and can provide more robust and informative results than traditional two-sample MR. The core of MVMR relies on the fact that some genetic variants correlate more strongly with some exposures than others. MVMR pay more attention to the respective effect of each exposure in the overall effect, and correct the true causal relationship through the competition between each exposure.

We selected a series of biological markers describing the clinical manifestation of PCOS, including sex hormone binding globulin (SHBG), testosterone levels, estrogen receptor (ER) and anti-Müllerian hormone (AMH) to evaluate endocrine disorder, infertility, anovulation, age at menarche and menopause to evaluate ovarian dysfunction, insulin-like growth factor-1 (IGF-1), homeostatic model assessment for insulin resistance (HOMA-IR), triglycerides and type 2 diabetes to evaluate insulin resistance. We hypothesize that some PCOS-related traits play a mediating role in the pathway. The aim of this study is firstly to examine the independent causal associations between PCOS and distinct subtypes of BC using two-sample MR, and secondly to assess the intermediary impacts of adjustable PCOS-related risk factors in the pathogenesis of BC using MVMR. By identifying these traits and inquiring the underlying mechanisms, this study could shed light on the etiology of PCOS and BC, providing insights into prevention and intervention strategies.

In this study, we have reaffirmed the causal relationship between PCOS and BC, particularly in ER + BC. Further, our research has found one mediator out of 13 PCOS-related traits, age at menopause, with mediating proportion of -4.82% in the association between PCOS and ER + BC.

One of the causes of endocrine disruption in PCOS is abnormal regulatory function of the hypothalamic-pituitary-ovarian (HPO) axis. Due to the increased sensitivity of the pituitary gland to gonadotropin-releasing hormone (GnRH), it secretes excessive LH, stimulating the cells of the ovarian mesenchyme and follicular membrane to produce excessive androgens. Hyperandrogenism in the ovary inhibits follicular maturation and prevents the dominant follicles formation, but small follicles can still secrete estradiol equivalent to the early follicular phase levels, coupled with the conversion of androstenedione to estrone under the impact of aromatase in peripheral tissues, generating the formation of hyperestrogenemia. Therefore, patients with PCOS not only have excess androgen levels, but also have high estrogen levels. Continuous secretion of estrone and a certain level of estradiol acts on the pituitary gland and hypothalamus, and positively feedbacks on LH secretion, resulting the amplitude and frequency of LH secretion increase with a sustained high level of no cyclicity, and no formation of mid-menstrual LH peaks, so no ovulation occurs. Estrogen in turn exerts negative feedback on follicle-stimulating hormone (FSH) secretion, resulting a relative decrease in FSH levels and an increase in the LH/FSH ratio. High levels of LH boost the secretion of androgens, while low levels of FSH stop the small follicles development, with no dominant follicle formed, thus starting up a vicious cycle of excessive androgens and continuous anovulation, contributing to polycystic changes in the ovary.

Androgens, as precursors to estrogens, can contribute to excessive estrogen production and subsequent breast cell proliferation when present in excess. Testosterone is converted into two antagonistic metabolites: estradiol binds to estrogen receptor to stimulate breast epithelial cell proliferation, while dihydrotestosterone binds to androgen receptor to inhibit this process. At the beginning of hormone-dependent tumor growth, the levels of both hormones are higher than in normal cells, and their effects are counterbalanced. However, if excess androgen levels are not eliminated, the proliferative effects of estrogen will ultimately prevail. Excess androgens can also result in elevated production of epidermal growth factor, a recognized stimulant of breast epithelial cell proliferation. Several prospective studies conducted in postmenopausal females have demonstrated a correlation between elevated circulating androgen levels and serum testosterone levels with an increased susceptibility to breast cancer [27‐29]. PCOS is linked to significantly elevated AMH levels compared to those found in normally ovulating women. The identification of a correlation between elevated circulating levels of AMH and the occurrence of BC lends support to the feasibility of utilizing AMH as a biomarker for the detection of BC [30].

Another pathogenesis of metabolic disorder of PCOS is insulin resistance and hyperinsulinemia. Reduced sensitivity of peripheral tissues to insulin and lower than normal biological efficacy of insulin is known as insulin resistance. About 50% of patients have varying degrees of insulin resistance and compensatory hyperinsulinemia. Excess insulin acts on insulin receptors in the pituitary gland to enhance LH release and promote androgen secretion from the ovaries and adrenal glands, which in turn increases free testosterone by inhibiting hepatic SHBG synthesis. Elevated insulin levels may encourage cell growth and division while affecting estrogen synthesis, metabolism, and signaling pathways, with increasing IGF-1 levels, which promote tumor growth and dissemination.

The pharmacological management of PCOS mainly focuses on alleviating metabolic abnormalities, including the use of oral contraceptive pills to regulate menstrual cycles, steroids to reduce blood androgen levels, metformin to relieve insulin resistance, clomiphene and FSH to induce ovulation, and additionally, IVF to restore fertility. Ovulation-inducing medications can result in increased estradiol levels. Furthermore, clomiphene may diminish estrogen receptor activity in certain tissues and has demonstrated a direct pro-apoptotic impact on BC cell lines, indicating a possible anti-cancer effect. IVF may heighten the BC risk due to hormonal fluctuations during the IVF procedure. Additionally, IVF could lead to the development of more blood vessels in breast tissue, which might offer a route for cancer cells to disperse. In women with impaired glucose homeostasis, metformin could potentially reduce BC risk by lowering insulin levels.

In the pathway from PCOS to ER + BC, -4.82% of the mediation effect was explained by age at menopause. When both direct and indirect effect are significant, but in opposite direction, previous study has described it as suppression effect [31]. According to the results of two-sample MR, genetically predicted PCOS is associated with later age at menopause, and later menopausal age is associated with higher risk of developing BC, which are both consistent with current epidemiological research perspectives [32]. However, after adjusting for the interaction between the two exposures with MVMR method, a weak protective effect against BC was detected, which is worth further investigation. One study has found age at menopause associated SNPs strongly enriched in DNA damage response (DDR) pathways [33]. As menopausal age is delayed, there is an increase of the DDR gene mutation and the BRCA1 gene expression, which promotes the repair of DNA double-strand breaks through homologous recombination. In other words, later menopausal age plays a role in tumor inhibition and BC risk reduction. This seems to be paradoxical with the epidemiological view. However, the study noted that the association between menopausal age variants and BC risk in DDR genes is weaker compared to those in non-DDR genes set, so hormone exposure duration plays a dominant role, producing a greater risk effect to offset the weak protective effect. MVMR may have corrected the interaction of the hormones effect between PCOS and age at menopause (i.e., the shared pathway between two exposures), and the calculated mediating proportion is the uncorrected effect produced by menopausal age variants in DDR pathways, playing a role as homologous recombination repair and tumor inhibition.

Given the causal impact and elevated prevalence of PCOS and BC, identification of this association could facilitate the creation of a screening program for individuals at elevated risk, as well as the implementation of primary interventions to mitigate their risk. We can suggest patients with PCOS start regular screening for breast examination, breast ultrasound, and mammography at an earlier age, instead of the guideline-recommended screening once every 6–12 months for those who are 40 years old or older. For patients found having benign breast nodules, closer follow-up is required, and core needle biopsy or vacuum assisted biopsy need to be taken into consideration more sensitively. When using endocrine medications for treatment, it is important to carefully monitor their potential interacting effects on the breast. As an aromatase inhibitor, letrozole is commonly employed in treating anovulatory infertility in PCOS. Compared to clomiphene, letrozole demonstrates better tolerability and fewer adverse effects on endometrium and cervical mucus, which can be used in patients with low response to clomiphene. Furthermore, letrozole is utilized as an adjuvant therapy in postmenopausal ER + BC patients, suggesting the possibility of a decreased risk of hormonal-dependent cancer. Considering the comparable pathogenesis underlying both PCOS and BC, common genes in the pathogenesis have been explored, which could be potential targets of treatment for both PCOS and BC [34].

More than 95% of the effect was not explained by the mediators included in the model. The null findings may be attributed to (1) the explanatory power of the instrumental variables (IVs) for the mediator is insufficient; (2) the mediator from PCOS to BC is inherently rare or has a weak mediating effect; (3) the presence of other unconsidered mediators. Our selection of mediators was limited to endocrine and metabolic pathways. This suggests we need to look for suitable mediators from other perspectives. New insights into genetic recombination and repair may be gained from future large-scale genomic studies. Consequently, we should interpret the results cautiously, and further research should delve into additional potential mediators and utilize supplementary approaches to validate the conclusions.

The study yielded an unexpected finding whereby a subset of potential mediators previously supported by robust observational studies, were found to be non-intermediary variables in the pathogenesis of BC from PCOS. Apart from unobserved confounding factors or other types of biases, null results may be elucidated by the collinearity between PCOS and the mediators. In other words, there may be an interaction between PCOS and mediators. As a whole, the effects induced by PCOS through complex endocrine and metabolic pathways encompass the effects induced by mediating factors. Therefore, when PCOS and mediators are included in the MVMR, their effects may be masked or offset, resulting in their causal relationship with BC no longer being significant. Using MVMR method, we identified possible DDR gene pathways of age at menopause other than endocrine pathways while unable to find positive results in other mediators.

The primary advantage of our study is the novel contribution of being the first MVMR study to identify causal mediators in the pathway connecting PCOS and BC. The MVMR approach has the advantage of accommodating the multiple mediators simultaneously and consider the combined impacts, even in the presence of bidirectional relationships. Meanwhile, we applied rigorous criteria to ensure the validity and plausibility of the model we constructed to explain the mediating effect.

However, we recognize certain drawbacks to our research. The absence of standardized diagnostic criteria for PCOS, including Rotterdam criteria, NIH/NICHD criteria, and self-reported diagnosis, has led to heterogeneity in the generated GWAS data sets of PCOS. Thereby, MR studies employing these data sets have not been able to detect the impacts of different phenotypes. A previous investigation has recognized reproductive and metabolic categories of PCOS, which seem to exhibit separate genetic frameworks, given that the PCOS diagnostic criteria fail to differentiate between biologically distinct disease subtypes [12]. Future research can overcome this concern when information regarding discrete subtypes of PCOS becomes accessible. Third, our BC outcome data were not stratified by menopausal status, limiting the ability to determine the effect of exposure and mediator in women of different menopausal statuses. Forth, the mediator GWAS data was composed of both genders; however, the exposure GWAS data of PCOS and outcome GWAS data of BC consisted solely of female. The MR analyses would have been more robust from a more consistent gender composition across exposure, outcome, and mediators. Lastly, the study population consisted solely of individuals of European descent, raising questions about the generalizability of the results of MR studies employing these genetic instruments to other populations. Future studies with more ethnic populations included are needed.

Since 2009, China has implemented a free public health program for BC screening, which has significantly reduced the morbidity and mortality rates of female BC. Gaining an insight into the relationship between PCOS and BC development will help public health policies to incorporate PCOS patients in the BC high-risk cohort in the future, and to spread the concept of "early prevention, early detection and early treatment" to the general population, especially target populations.