Association Between Laparoscopically Confirmed Endometriosis and Risk of Early Natural Menopause

This cohort study investigates the association of endometriosis and risk of early menopause, adjusting for assesses the anthropometric, behavioral, and reproductive factors, among premenopausal female nurses in the Nurses’ Health Study II cohort.


Introduction
Endometriosis is an often chronic inflammatory disease characterized by the presence of endometrial-like tissue outside of the uterus. 1,2 Although endometriosis affects approximately 10% of women during their reproductive years, 3,4 its association with other health outcomes, including early natural menopause (ENM), remain unclear. [5][6][7] Early menopause is defined as the cessation of ovarian function before age 45 years. Affecting approximately 10% of women in Western populations, 8 endometriosis is associated with greater risk for cardiovascular disease, cognitive decline, osteoporosis, and premature mortality. 9,10 Multiple mechanisms support the hypothesis that endometriosis or, specifically, endometriomas (the endometriosis subtype characterized by ovarian cysts) 2 and the treatments associated with it may have qualitative and quantitative implications for ovarian reserve. 11 Endometriosis has been associated with lower ovarian reserve among women with female-factor infertility, 12,13 and endometriomas 14 have been associated with ovarian aging and menopause timing. 11 In addition, local and systemic increased inflammation has been associated with endometriosis 1,2 and hypothesized to be associated with early menopause; however, these associations were nonlinear and inconsistent among markers. 15 Women with endometriosis compared with women without endometriosis have peritoneal fluid that contains more activated proinflammatory, chemotactic, and oxidative stress factors, 16,17 which could create an environment that is detrimental to follicular and ovarian function and leads to an earlier age at menopause. 18 Determinants of age at menopause include the number of oocytes at birth, the rate of atresia throughout the lifespan, and the threshold number of oocytes needed to produce sufficient hormones to maintain menstrual cyclicity. 8,19,20 Reproductive events that interrupt ovulation (eg, pregnancies and lactation) 21 or modify the rate of atresia slow the depletion of the ovarian follicle pool and are associated with delayed menopause. 21,22 Conversely, events that disrupt ovarian function, such as ovarian surgery, autoimmune oophoritis, cigarette smoking, and androgenizing exogenous hormones, 22 may have implications for ovarian reserve, characterized by decreased follicle numbers and lower oocyte quality and quantity, which are associated with earlier age at menopause. 8 Early natural menopause and endometriosis share several risk factors, including body size, 23,24 earlier age at menarche, 24 shorter menstrual cycles, 24,25 nulliparity, 4,21 and infertility. 4 Furthermore, medical and surgical management for endometriosis, including oral contraceptives (OCs) 26 and ovarian surgeries, 14 may alter early menopause risk.
Within the UK Royal College of General Practitioners' Oral Contraception Study cohort, 27 history of endometriosis was associated with ENM. However, the Royal College of General Practitioners' Oral Contraception Study was limited by its reliance on recalled age at menopause and minimal adjustment for potential confounders. Therefore, in the current prospective cohort study, we sought to investigate the association between endometriosis and the risk of ENM using the US Nurses' Health Study II cohort. We applied a prospective time-to-event analysis that accounted for timevarying anthropometric, behavioral, and reproductive factors. 28 In addition, we investigated whether the association between endometriosis and ENM risk differed by the following risk factors

Assessment of Endometriosis
Participants were asked on each biennial questionnaire from 1993 onward whether they had physician-diagnosed endometriosis. Those who responded yes indicated the year of diagnosis and whether it had been confirmed by laparoscopy, the clinical criterion standard. 2 Among women with self-reported laparoscopic confirmation of their endometriosis diagnosis, medical record confirmation was high in 2 validation studies (100% in the first wave and 95% in the second wave of the study). 29 Endometriosis diagnosis status was treated as a time-varying factor and updated over time. Once a participant reported laparoscopically confirmed endometriosis, she was considered to have endometriosis (contributing person-time to the endometriosis-exposed group) through the remainder of the follow-up period.
Participants who reported physician-diagnosed endometriosis but not surgically confirmed endometriosis throughout the follow-up period did not contribute person-time to either the exposed or the unexposed group, as it could not be definitively ascertained that they had endometriosis or confirmed that they did not. 29 If laparoscopic confirmation was reported at any time after the initial report of physician-only diagnosis, the person-time contribution to the endometriosis-exposed group was reset to the time of the first report because the surgical confirmation validated the initial physician diagnosis. 29 At baseline in 1989, there were 3921 women who had a history of endometriosis diagnosis, whereas 6640 women were diagnosed with endometriosis after Nurses' Health Study II enrollment.

Assessment of Early Menopause
On the 1989 and subsequent biennial questionnaires, participants were asked if their menstruation had ceased permanently. Those who reported that their menstruation had ceased were asked to indicate the age at cessation and whether cessation was related to surgery, radiotherapy, or chemotherapy or occurred naturally. Age at menopause was defined as age after 12 consecutive months of amenorrhea. For women who reported being postmenopausal on 1 questionnaire and then subsequently reported being premenopausal, we defined age at menopause as the age after which menses were absent for 12 months or more and then confirmed that this status persisted for at least 3 consecutive questionnaires. Information regarding cessation was provided on each questionnaire, which enabled us to account for transient cycles without menses followed by the return of menstruation and hence to avoid the misclassification of early menopause status. We defined cases of ENM as women who reported natural menopause before the age of 45 years.

Covariates
Factors associated with endometriosis or ENM were identified from previous literature as a priori potential confounders 30  Vegetable protein intake and alcohol consumption were calculated, and these measures were adjusted for energy using the residual method. 30

Statistical Analysis
Participants contributed person-months to the Nurses' Health Study II cohort from the return of the baseline questionnaire until (1) the onset of menopause; (2) 45 years of age; (3) a hysterectomy; (4) a bilateral or unilateral oophorectomy; (5) a cancer diagnosis, not including nonmelanoma skin cancer; (6) death; (7) loss to follow-up; or (8) the end of follow-up (May 2017), whichever occurred first. Baseline characteristics of the participants were examined according to baseline endometriosis diagnosis status: laparoscopically confirmed endometriosis (with or without). We used a Cox proportional hazards regression model stratified by age and questionnaire cycle to estimate age-and calendar time-adjusted hazard ratios (HRs) and 95% CIs of the association between laparoscopically confirmed endometriosis and ENM. Proportional hazards assumptions were tested using the likelihood ratio test, which compared models with and without an interaction term for calendar time.
In addition to adjusting for age and calendar time, based on previous literature, we adjusted for known anthropometric, demographic, and behavioral potential confounders for endometriosis and ENM a priori, with time-varying covariates updated biennially at every questionnaire cycle.
Multivariable model 3 was also adjusted for reproductive risk factors for both endometriosis and ENM. In all multivariable models, we applied the missing indicator method in which an indicator variable was created to address missing values of covariates. 5,33 Stratified analyses assessed heterogeneity by potential effect modifiers, including BMI (<25 or Ն25), cigarette smoking status (never or ever), OC use (never or ever), parity (nulliparous or parous), and history of infertility attributed to ovulatory disorder (no or yes). Likelihood ratio tests were used to identify statistically significant differences between potential effect modifiers on the multiplicative scale.
We conducted several secondary analyses. To evaluate whether hormone use affected recognition and reporting of menopause, we conducted analyses that included censoring at the first report of hormone therapy (HT) and controlling for HT in multivariable models. Oral contraceptive use can also mask menopause; thus, we conducted analyses that censored at the first report of OC use. Given the conflicting evidence in the literature about the association of OC use with ENM, we also performed an analysis that excluded OC use as a confounder. Given the inconsistent evidence in the literature about the association between inflammatory markers and ENM and given that antiinflammatory medications were more likely to be used by women with endometriosis, we evaluated potential confounding and mediation effects of nonsteroidal anti-inflammatory drugs. We applied the Cox proportional hazards regression model with and without adjustment for years of use of aspirin, acetaminophen, ibugesic, nonsteroidal anti-inflammatory drugs, and cyclooxygenase inhibitors (modeled continuously). Mediation by medication use was assessed with the difference method, 34

Results
The characteristics at baseline of the 106 633 female participants (mean [

Discussion
In this large, prospective cohort study, we observed that surgically confirmed endometriosis was associated with a significantly greater risk of ENM. The risk estimate remained robust and statistically significant with adjustment for demographic and behavioral factors. The association was slightly attenuated with additional adjustment for reproductive factors but remained statistically significant.
Residual confounding may remain; however, for residual confounding to alter the conclusions that were drawn from the quantified effect estimates, it would have to exceed the confounding outcomes that were captured by the data that were already included in the multivariable-adjusted models. The observed association between endometriosis and ENM varied between subgroups, with statistically significant effect modifiers identified. Those with endometriosis had the highest risk for ENM among women who never used OCs or were nulliparous.
To our knowledge, no previous study has prospectively investigated the association between endometriosis and ENM with time-varying covariates, time-to-event analyses, and early age at natural menopause (clinically defined as before age 45 years). 35 A cross-sectional study 27 reported that women with ENM (defined as menopause at Յ49 years, the cohort's median age at natural menopause) were more likely to report an endometriosis diagnosis after adjustment for social class, In addition, we examined the association between endometriosis and ENM after stratifying by BMI (<25 or Ն25). Previous studies found a consistent inverse association between overweight or obesity and risk of endometriosis. 4,23,37 In the population of the present study, those with endometriosis were less likely to be overweight or obese at cohort baseline in 1989 and in late adolescence. Previous studies reported that women with a BMI between 25.0 and 29.9 had significant lower odds for ENM. 38 We found that, among participants with a BMI of 25 or higher, endometriosis was associated with an approximately 60% greater risk for ENM. This finding suggests that, among women with a BMI of 25 or higher who may have lower risk for ENM, endometriosis emerges as a risk factor.
Endometriosis was associated with a greater risk for ENM among nulliparous vs parous participants. Parity has been associated with delayed age at menopause, 21 and endometriosis was not a significant risk factor for ENM in this subgroup. Although a meta-analysis found an association between OC use and later age at natural menopause, 39 a recent discovery in the Nurses' Health Study II population 40 did not support a clear association between duration of OC use (decreasing lifetime number of ovulatory cycles) 41 and risk for ENM. In the present study, among participants who never used OCs, endometriosis was associated with a 2-fold greater risk for ENM. It is likely that OC use masks menopause, which is important to consider in this analysis particularly because women may use OCs to control endometriosis-associated symptoms. Similarly, women may use analgesics for endometriosis-associated symptoms; however, none of the analgesics we assessed were significant mediators of the association between endometriosis and ENM.

Limitations
This study has several limitations. Censoring because of loss to follow-up or competing events (eg, surgical menopause) can represent potential selection bias in prospective studies. In the Nurses' Health Study II cohort, both dropout and incidence of competing events were low; simulation studies have shown that, when overall censoring is low, the likelihood of potential bias in cohort studies is also low. 42,43 Although we relied on self-report of endometriosis, which might lead to some exposure misclassification, the endometriosis data were collected prospectively before the outcome and therefore could not be differentially misclassified with respect to the outcome. Furthermore, selfreport of endometriosis has been demonstrated to be highly valid in this Nurses' Health Study II cohort of medical professionals. 29 We also relied on self-report for onset of menopause, which might contribute to some outcome misclassification 44,45 ; however, menopause status also was collected prospectively, repeated across multiple questionnaires, confirmed for reported status consistency, and unlikely to be reported more or less accurately by women with endometriosis compared with those without endometriosis. Self-assessment of menopause was validated in the Nurses' Health Study I cohort, wherein age at menopause was confirmed with exceptional validity using medical records for 99% of participants, and 82% of participants consistently recalled the same age at menopause within 1 year on repeated questionnaires. 45 To further account for potential masking of ENM among participants who were exposed to hormones, we conducted analyses that censored for HT, and the results did not change. The race and ethnicity of the study population were fairly homogenous, yet we would expect that the physiological association between the reproductive factors of endometriosis and ENM would not differ substantially by race and ethnicity. However, confirmation of these results in more diverse populations will be informative. This analysis included more than 100 000 women who contributed more than 1.5 million person-years of follow-up, during which 2542 ENM cases were reported. Nevertheless, low incidence of early menopause presented some limitations with estimating the risk among endometriosis-exposed (n = 6640) compared with unexposed (n = 99 993) participants, highlighting the challenge for longitudinal studies of uncommon outcomes and the need for large sample sizes and long follow-up duration.

Conclusions
This prospective cohort study found a statistically significant association between laparoscopically confirmed endometriosis and risk for ENM. Endometriosis and ENM had many overlapping risk factors, suggesting that an association could be explained by confounding. After time-varying multivariable adjustment, little confounding by known demographic and behavioral risk factors for ENM was observed. There was some confounding impact after adjustment for reproductive factors.