Endometriosis diagnosis, staging and typology and adverse pregnancy outcome history

Among an inclusive cohort of women with no prior endometriosis diagnosis, we observed a 10% lower probability of live birth and a 50% higher probability of miscarriage after adjustment among women with endometriosis. Additionally, we observed a higher prevalence of ectopic pregnancies and SGA among women with versus without endometriosis. Women with deep endometriosis had a threefold higher probability of miscarriage compared with women without endometriosis, which increased to over a 3.5-fold higher probability compared with women with a ‘normal’ pelvis or without endometriosis or other gynecologic pathology found at laparoscopy. Our study has several strengths. Notably, no other studies have thoroughly assessed reproductive history prior to endometriosis diagnosis both within a broadly representative sample of women undergoing pelvic surgery for any indication and a population-based sample of women without clinical complaints. Much of the available data focus on pregnancy and associated outcomes after endometriosis diagnosis and/or treatment. Compared with registry-based studies, our sample size is modest but captures women’s complete reproductive history prior to endometriosis diagnosis. Other strengths include the ability to modify the comparison group to women without endometriosis and, more restrictedly, women without any visualised gynecologic pathology, whereby we observed a strengthening of association between endometriosis and adverse pregnancy outcomes. The population cohort enables us to broadly assess the consistency of findings observed in the ‘gold standard’ cohort among women with undiagnosed endometriosis from the surrounding geographic population. This latter group of women, in theory, are served by the same clinical facility as women in the operative cohort and would be ‘at risk’ for diagnosis if they were to seek care. Finally, we provide important information on the relationship between endometriosis staging and typology and pregnancy outcomes, which few studies have previously reported on. Our study has important limitations in the context of its strengths. While originally powered to assess environmental exposures and endometriosis, it has limited power for rarer pregnancy outcomes, especially among the population cohort. Consequently, we were limited in the number of factors we could adjust for in some of our models (i.e. endometriosis staging and typology along with the more rare pregnancy), and thus, residual confounding remains a consideration. A priori we recognised that MRI-diagnosed endometriosis could lead to misclassification on disease status in the light of its largely detecting endometriomas. 2 As such, our estimates may be underestimated though the direction of point estimates was largely similar. While the MRI diagnostic method used in the population cohort may have resulted in underestimation of all endometriosis types due to its known ability to primarily capture endometriomas and deep lesions, but not superficial endometriosis, 2 our findings within the operative cohort support the hypothesis that deep endometriosis, with or without ovarian endometriomas, is primarily responsible for increased risk of adverse pregnancy outcomes. An additional limitation is that we may have encountered misclassification in using self-reported pregnancy complications. 28 However, reproductive history was captured prior to incident endometriosis diagnosis, limiting differential bias. A further limitation is the timing of reproductive history captured before an endometriosis diagnosis. This could mean that early pregnancies were unaffected by endometriosis. However, on average, there is an endometriosis diagnostic delay of 7 (standard deviation = 6) years (5 years for countries with private versus state-funded health care). 4 Therefore, in this study, pregnancies within the last 5–7 years were likely affected by endometriosis. We addressed this limitation of the timing of reproductive history relative to endometriosis diagnosis in our sensitivity analysis, limiting to pregnancies within the last 5 years and our findings were consistent, albeit with less precision. Finally, due to the cross-sectional nature of our study, retrospective recall of reproductive history and inclusion of only gravid women, we are not able to separate the effects of endometriosis on fecundability and live birth. Future research among a cohort of adolescents/young women who could be followed prospectively for gynecologic disorder diagnoses and reproductive events is needed in order to disentangle the direct and indirect role that endometriosis may play in the ability to achieve pregnancy and a complication-free live birth. While endometriosis is known to be associated with a longer time to pregnancy, the majority of women who have endometriosis will go on to achieve a planned pregnancy. 14 , 20 , 29 Whether women with endometriosis, or a specific endometriosis phenotype, have an increased risk of miscarriage, ectopic pregnancies or stillbirth, and consequently decreased probability of live birth, has been under debate for many years. Given that implantation failure and progesterone resistance of the endometrium have been found to be greater among women with endometriosis, 30 , 31 an increased risk of miscarriage for women with endometriosis is plausible. Earlier studies reported no significant association between endometriosis and the frequency of miscarriage; however, these prior studies were limited in sample size and did not differentiate between early versus late losses 32 or did not take into account repeated pregnancy losses. 33 In this study, we were able to demonstrate that endometriosis is associated with a higher probability of losses after 8 weeks and with repeated pregnancy loss. Moreover, recent research in larger populations has shown a modest but consistent relationship between endometriosis and miscarriage, 14 , 15 , 34 – 37 with a recent systematic review of nine cohort studies finding a 1.75 higher risk (95% CI: 1.29, 2.37). 17 , 36 While this is in line with our finding, our lower adjusted pooled estimate of 1.32 may be due to our inclusive population of women seeking and not seeking gynecologic care. Few prior studies have assessed endometriosis typology and pregnancy loss among women with proven fertility with one reporting that women with superficial peritoneal endometriosis had a higher prevalence of miscarriages compared with women with deep lesions or endometriomas; 19 while another study reported little difference in miscarriage risk between the three endometriosis subtypes. 18 Given our nearly fourfold increased prevalence among women with deep endometriosis compared with women without endometriosis or other gynecologic pathology, further research among larger representative cohorts that can prospectively assess endometriosis diagnosis and typology and subsequent pregnancy complications will provide valuable insight into potential biological pathways. Early studies reported no association between endometriosis and ectopic pregnancies. 38 , 39 However, more recent studies support our findings of a higher risk of ectopic pregnancy among women with endometriosis. 14 , 34 , 40 , 41 Indeed, the reported abnormalities of uterine contractility or blockage due to lesions and adhesions in women with endometriosis could mechanistically contribute to both ectopic pregnancies and miscarriages. 42 While our 95% CIs are within the range of the twofold–threefold increased risk found in prior studies, 17 , 43 caution is warranted in extrapolating our findings given that ectopic/tubal pregnancies were reported in only 1.4% ( n = 13) of pregnancies in the operative cohort and in no pregnancies in the population cohort. We were also limited in the number of stillbirths; however, our 1.28 adjusted prevalence ratio is in line with a meta-analysis of seven prior studies that reported a pooled odds ratio of 1.29 (95% CI 1.10, 1.52). 44 Future research among larger samples of women with follow-up including repeated pregnancies to assess lifetime risks is needed for a more complete understanding. 45 In addition to abnormalities of uterine contractility, local and systemic inflammation and deficient placentation found in women with endometriosis are hypothesised to increase the risk of pregnancy complications such as SGA and preterm birth. 17 While there was no consistent pattern for increased prevalence of preterm birth within the ENDO study, we did observe a higher prevalence of SGA, especially among the population cohort. Prior meta-analyses have consistently shown a modest association for endometriosis and SGA (ranging from 16% to 28% elevated risk), which might be accounted for by the lean body habitus 46 reported for women with endometriosis. Our finding of an appreciably higher increased probability of SGA in the population compared with the operative cohort may be due to chance, or possibly due to the differential makeup of the populations in regards to socio-economic status. Further research in larger population-based samples is needed to better understand potential disparities in the relationship between endometriosis and pregnancy outcomes among women with and without access to care.

For this secondary data analysis of the ENDO study (2007–2009), we include 392 gravid women (316 from the operative cohort and 76 from the population cohort). The operative cohort included menstruating women ages 18–44 years, without a prior history of endometriosis, who underwent a diagnostic and/or therapeutic laparoscopy or laparotomy regardless of clinical indication at participating hospital-based surgical centres located in Salt Lake City, Utah, or San Francisco, California. The population cohort was matched to the operative cohort by age and residence within the geographic catchment areas for the participating surgical centres ( Figure S1 ). Women in the operative cohort were interviewed by trained research assistants approximately 2 months before surgery, while women in the population cohort were interviewed before undergoing magnetic resonance imaging (MRI) for the detection of endometriosis. Specifically, computer-assisted personal interview techniques captured socio-demographic, lifestyle, psychosocial, health and reproductive data. Poverty line was based on the 2007 Health and Human Services Poverty Guidelines. Physical activity was assessed using the International Physical Activity Questionnaire-Short Form, categorised into low, moderate and high activity. 21 Following the in-person interviews, anthropometric assessments were conducted according to standardised procedures. 22 For the operative cohort, surgeons completed a standardised report after surgery to capture post-operative diagnosis(es) including a ‘normal’ pelvis, endometriosis, uterine fibroids, pelvic adhesions, benign ovarian cysts, neoplasms and/or congenital Müllerian anomalies. Surgeons used the revised American Society for Reproductive Medicine’s (rASRM) staging criteria 23 to report endometriosis severity. In addition to rASRM staging, we used the rASRM standardised form, which has information on lesion location and size, to assess the location of endometriosis lesions. 24 , 25 Women with only superficial lesions in the ovary or peritoneum were considered to have superficial endometriosis (SE), women with deep lesions (>5 mm invasion) 26 noted in the peritoneum or with obliteration of the posterior cul-de-sac were considered to have deep endometriosis (DE), and women with deep lesions of any size noted on the ovary were considered to have ovarian endometriomas (OE). Women who had deep ovarian and peritoneal lesions were considered OE + DE. Among participants in the population cohort, a pelvic MRI was performed using a standardised methodology to identify the presence/absence of endometriosis. 2 Women provided in-depth information about each pregnancy: pregnancy intention, time to pregnancy, use of infertility treatment and pregnancy outcome: single live birth, multifetal pregnancy (and number of live births), stillbirth, miscarriage, ectopic or tubal pregnancy and medical or elective termination. For each pregnancy, whether a loss or live birth(s) occurred, women reported on pregnancy duration. Women reporting live birth(s) were queried on caesarean versus vaginal delivery, whether pregnancy was considered full-term or preterm, and infant(s) weight (asked for each infant of twin pregnancies). Small for gestational age (SGA) was calculated via US-based birth weight for gestational age references. 27 Based on provided information regarding total number of miscarriages and duration of gestation at time of miscarriage, we also were able to assess early (≤8 weeks) versus late (>8 weeks) losses and, at the woman level, recurrent pregnancy loss (≥2 miscarriages). Women’s characteristics by endometriosis diagnosis were compared using descriptive statistics. We used generalised linear mixed models, taking into account multiple pregnancies per woman and multiple offspring per pregnancy, specifying a Poisson distribution and log link function with an autoregressive correlation structure, to determine unadjusted and adjusted prevalence ratios (aPR) between endometriosis and pregnancy outcomes. Multinomial models were used at the woman level for recurrent pregnancy loss (with a three-category outcome: no loss, 1 loss and ≥2 losses). We performed analyses separately for each cohort and combined using a fixed effects meta-analysis. The association between endometriosis staging (I–IV versus no endometriosis as reference) and typology (SE, OE, DE, OE + DE, versus no endometriosis as reference) and pregnancy outcomes were assessed for the operative cohort only. For the fixed effects meta-analysis, staging and typology models, we restricted our assessments to live birth and miscarriage given low prevalence for other adverse outcomes. Potential confounders considered included age, BMI, measured waist circumference, body habitus (scale 1 to 9) at adolescence (ages 15–19), race/ethnicity, marital status, household income, education, site, smoking and alcohol intake, daily caffeine intake, physical activity and time-varying age of each reported pregnancy. All analyses were performed in SAS version 9.4 (SAS Institute, Cary, NC, USA), and the meta-analysis fixed effects models were performed in STATA (STATACorp, College Station, TX, USA). Complete covariate data were obtained for site and race/ethnicity. Missing data were minimal (<1%) for age, pain, education, smoking status, marital status, alcohol consumption, BMI, waist circumference, income and subfertility. Caffeine and physical activity data were missing for 4% and 11% of women, respectively. Among the 1172 reported pregnancies by the 392 gravid women, 0.17% were missing values for pregnancy outcome information. And among women with a live birth(s), birthweight was missing for 3.5%. For our fully adjusted models that took into account age at enrolment, study site, BMI, income, alcohol consumption, smoking status, caffeine consumption, physical activity and time-varying age of pregnancy adjusted results, we performed multiple imputations (with 50 imputations) to address potential missing data bias, notably for lifestyle factors of caffeine consumption and physical activity. We conducted a sensitivity analysis among all gravid women, restricting pregnancy history to the 5 years prior to ENDO Study enrolment corresponding to the average length of time for an endometriosis diagnosis. 4 Sensitivity analyses were also conducted assessing the association between endometriosis diagnosis, staging and typology and pregnancy outcomes with the comparison group being women without endometriosis or any other gynecologic pathology. Study participants were remunerated for their time and travel. Full human subjects approval was obtained for the conduct of this study; each of the women provided informed consent before any data collection. 2

One hundred and nine women (34%) in the operative cohort and nine women (12%) in the population cohort were diagnosed with incident endometriosis ( Table 1 ). Both cohorts were predominantly non-Hispanic white. In the operative cohort, women with endometriosis had lower BMIs, higher incomes, higher prevalence of subfertility (requiring ≥6 months to become pregnant) and more reported cyclic or chronic pelvic pain compared with women without endometriosis. The direction of associations was similar in the population cohort, with the exception of socio-economic status, which was lower in women with, versus without, endometriosis. There was no difference in median number of pregnancies among women with, versus without, endometriosis ( Table 1 ). However, women with endometriosis were less likely to have had a live birth compared with women without endometriosis in both the operative (61% vs. 69%) and population (68% vs. 77%) cohorts ( Table 2 ). Miscarriage, ectopic/tubal pregnancy, preterm birth (<37 weeks) and small for gestational age occurred more frequently in women with versus without endometriosis in the operative cohort. In the population cohort, miscarriage, planned termination and small for gestational age occurred more frequently in women with versus without endometriosis. There were no ectopic pregnancies, one stillbirth and 3 very preterm births (<33 weeks) reported from women in the population cohort. Among all planned terminations, 2 were reported to be due to genetic problems or foetal abnormalities. Little difference for caesarean delivery was found by endometriosis diagnosis in either cohort. Women in the operative cohort with visualised endometriosis had a lower prevalence of live births, aPR 0.88 (95% CI 0.77, 1.01) and a higher prevalence of miscarriages, aPR 1.51 (95% CI 1.01, 2.25) compared with women without endometriosis ( Table 3 ) after adjusting for age at enrolment, study site (UT vs CA), BMI (kg/m 2 ), income (above vs below poverty line) and time-varying age of pregnancy. Additional adjustment for education, alcohol and smoking status (never, past and present), caffeine consumption (drink/day) and physical activity (low/moderate vs high) resulted in similar results with women in the operative cohort with visualised endometriosis having a lower prevalence of live births, aPR 0.94 (95% CI 0.85, 1.03) and a higher prevalence of miscarriages, aPR 1.48 (95% CI 1.23, 1.77) compared with women without endometriosis. The direction and magnitude of estimates were similar in the population cohort. The prevalence of ectopic pregnancy was higher in the operative cohort for women with versus without endometriosis. An insufficient number of ectopic pregnancies were reported for the population cohort to evaluate. Women with versus without endometriosis were 1.4-fold more likely to have an SGA birth in the operative cohort and over sixfold in the population cohort. No clear patterns emerged for the association between endometriosis and stillbirth, preterm birth or caesarean delivery among the two cohorts. Prevalence for late loss was higher than for early loss as was recurrent pregnancy loss among women with endometriosis ( Table 3 ). Combining the operative and population cohorts, the pooled prevalence ratio for live birth was 0.89 (95% CI 0.81, 0.97) and 1.32 (95% CI 1.06, 1.66) for miscarriage ( Figure 1 ). We observed no clear trend in the prevalence of live birth or miscarriages by endometriosis staging ( Table 4 ). A clearer pattern emerged with typology ( Table 5 ). Women with DE or DE plus OE had the lowest probability of a live birth and the highest probability of miscarriage compared with women without endometriosis. In sensitivity analyses restricting pregnancies to the 5 years prior to endometriosis diagnosis, results were similar in magnitude and direction, but with wider confidence intervals due to the small sample size ( n = 127 pregnancies among operative cohort; Table S1 ). Associations were slightly stronger for endometriosis diagnosis in sensitivity analyses using women without endometriosis or gynaecological disorder as the reference group ( Tables S2 and S3 ). While results remained largely unchanged with staging ( Table S4 ), in typology sensitivity analyses using women without endometriosis or gynaecological disorder as the reference group, women with DE, versus normal pelvis, were over 3.5-fold more likely to report at least one miscarriage ( Table S5 ).

Endometriosis, characterised by the presence of endometrial glands and stroma outside the uterus, is associated with inflammatory responses 1 including chronic and cyclic pelvic pain. Endometriosis affects at least 11% of women 2 and is an important marker of female fecundity. 3 Diagnosis of endometriosis takes on average 4–7 years between symptom presentation and disease diagnosis. 4 – 6 As such, pregnancy history prior to incident endometriosis diagnosis may be informative about the underlying pathophysiology, especially for women with asymptomatic or undiagnosed disease who do not seek clinical care. Understanding the natural history of endometriosis is important for delineating its relation with later onset diseases such as cardiovascular disease, 7 , 8 cancer 9 , 10 and other chronic disease. 11 – 13 Emerging evidence indicates that endometriosis not only leads to reduced female fertility but may also be a risk factor for adverse pregnancy outcomes. 14 – 17 However, inconsistent research exists, in part, due to choice of sampling framework, which often relies on women seeking clinical care 15 or undergoing operative procedures as captured in hospitalisation registries for convenience and efficiency purposes. 14 Additionally, research on endometriosis staging or typology and adverse pregnancy outcomes is lacking and with contradictory findings. 18 , 19 Given the equivocal body of literature to date, our research objectives were to evaluate whether endometriosis diagnosis, staging and typology are associated with self-reported pregnancy outcomes. We include both an operative cohort of women seeking gynaecological care for multiple indications and an age-and residence-matched population cohort of women who were not seeking care, who participated in the Endometriosis, Natural History, Diagnosis and Outcomes (ENDO) Study. 2 Previously, we reported a 29% reduction in fecundability across all pregnancy-trying attempts, among women reporting ever having planned a pregnancy. 20 In this study, we build on our prior work to explore the association between endometriosis and pregnancy outcomes in the well-phenotyped ENDO cohorts.

We show that among gravid women, while there is no difference in number of pregnancies achieved, women in both an operative cohort (seeking gynecologic care) and population cohort (not seeking gynecologic care) who have endometriosis have a 20–50% higher probability of miscarriage compared with women without endometriosis. Additionally, women with DE have a threefold higher probability of miscarriage compared with women without endometriosis and a 3.5-fold higher probability compared with women without endometriosis or other gynecologic pathology. Future studies are needed to capture endometriosis diagnoses prior to pregnancy complications, preferably via surgically confirmed endometriosis, but as clinical assessment criteria are better refined, 47 this approach also may be useful. Additionally, while we observed an overall association between endometriosis and reduced live birth rates and increased miscarriages, ectopic pregnancies and SGA births, an investigation into potential mediating effects of endometriosis (e.g. via the activation of inflammatory cytokines or chronic pelvic pain) on this association would provide insight into pathophysiology and inform clinical care. Finally, future prospective research is warranted to examine how endometriosis typology (superficial, endometrioma and deep endometriosis) and severity (minimum, mild, moderate and severe), along with the presence of uterine adenomyosis, may impact pregnancy outcomes.