Endometriosis and Risk of Adverse Pregnancy Outcomes

The Nurses’ Health Study II is an ongoing prospective cohort study following 116,429 female U.S. registered nurses who were 25–42 years of age at enrollment in 1989. At baseline and every 2 years thereafter, participants completed self-administered questionnaires, with questions that were consistent across biennial questionnaires as well as newly added domains that varied by questionnaire, to capture detailed information on a variety of lifestyle and reproductive characteristics and to update health-related outcomes. Cumulative follow-up of the cohort is greater than 90%, and we found no difference in follow-up between women with and without a diagnosis of endometriosis (P5.41). The Nurses’ Health Study II protocol was approved by the institutional review board of the Partners Health Care System, Boston, Massachusetts. Women were asked on each biennial questionnaire from 1993 onward whether they had physician-diagnosed endometriosis. Participants who responded “yes” indicated the year of diagnosis and whether it had been visually confirmed by laparoscopy, the clinical gold standard for endometriosis diagnosis. 28 – 30 Self-reported endometriosis was validated in a random subgroup of Nurses’ Health Study II participants (n=184). For women whose medical records were available, a diagnosis of endometriosis was confirmed in 96% of women reporting laparoscopically confirmed endometriosis but in only 54% of women who self-reported endometriosis without laparoscopic confirmation. 8 , 31 Owing to the strong potential for misclassification of self-reported endometriosis without laparoscopic confirmation, we restricted our endometriosis definition to laparoscopically confirmed endometriosis. Once a woman reported incident laparoscopically confirmed endometriosis, she was considered to have endometriosis through the remainder of follow-up. On the 1993 questionnaire, women could report the date of endometriosis diagnosis (before September 1989, September 1989–May 1991, June 1991–May 1993, after June 1993). Women who reported having a laparoscopically confirmed endometriosis diagnosis before September 1989 and having a pregnancy before 1989 were categorized as having endometriosis for that pregnancy, as has been done previously. 25 All models were adjusted for date of pregnancy in relation to cohort enrollment (1989 or before, after 1989) and an interaction term between date of pregnancy and endometriosis diagnosis. This assumption was additionally tested in sensitivity analyses ( Appendix 1 , available online at http://links.lww.com/AOG/B463 ), restricted to incident pregnancies occurring after the cohort began in 1989. Our main outcomes of interest for this analysis were spontaneous abortion, ectopic pregnancy, stillbirth, GDM, hypertensive disorders of pregnancy (preeclampsia or gestational hypertension), preterm birth, and low birth weight. Women reported their pregnancy history in 1989 at cohort enrollment, when they were between the ages of 25–42 years, and reported subsequent pregnancies on each biennial follow-up questionnaire. The 2009 questionnaire included an overall summary of reproductive history to collect information on all pregnancies and births, including detailed information on year of birth, gestation length, birth weight, pregnancy outcomes, and complications. Gestation length was reported using the following categories: less than 8, 8–11, 12–19, 20–27, 28–31, 32–36, 37–39, 40–42, 43 or more weeks. Birth weight for pregnancies lasting 20 weeks or longer was reported using the following categories: less than 5, 5–5.4, 5.5–6.9, 7–8.4, 8.5–9.9, 10 or more lb. Pregnancy outcomes were singleton live birth, multiple birth, miscarriage or stillbirth, tubal or ectopic pregnancy, or induced abortion. Spontaneous abortion was defined as a fetal loss occurring at less than 20 weeks of gestation. Stillbirth was defined as fetal loss at or after 20 weeks of gestation. Ectopic pregnancy, GDM, and hypertensive disorders of pregnancy diagnoses were based on self-report in 2009. Preterm birth was defined as birth at less than 37 weeks of gestation. Low birth weight was defined as birth weight less than 5.5 lb (less than 2,500 g) and was restricted to singletons. Women who reported pregnancy on previous questionnaires but were missing pregnancy outcome information on the 2009 pregnancy questionnaire (2.5% of pregnancies) or were missing age at pregnancy (fewer than 0.01% of all pregnancies) were excluded. Prior work from our cohort found that self-reported adverse pregnancy outcomes are validly reported (GDM: 94% confirmation; preeclampsia: 89% confirmation; preterm birth: 81% sensitivity; birth weight: 0.74 correlation). 32 – 35 Although the validity of maternal recall of pregnancy loss has not been assessed in this population, the sensitivity in the general population is estimated to be approximately 75%. 36 , 37 On the 1989 baseline questionnaire, participants reported a number of characteristics, including their height; current weight; weight at age 18 years; physical activity; smoking history; age at menarche; menstrual cycle length and pattern between ages 18 and 22 years; oral contraceptive use; parity (number of pregnancies lasting 6 months or longer); history of infertility (more than 12 months trying to conceive without success); menopausal status; history of type 2 diabetes, cardiovascular disease, or cancer; and their race and ethnicity. All time-varying characteristics were updated every 2–4 years. On the 1991 questionnaire, nurses recalled their own birth weight, gestational age, diet, and alcohol intake. Participants reported their physical activity between ages 18 and 22 years on the 1997 questionnaire and reported their alcohol consumption during high school on a supplemental 1998 dietary questionnaire. Pregnancy was the unit of observation for our analyses. Analyses in which the outcomes of interest were stillbirth, GDM, hypertensive disorders of pregnancy, preterm birth, and singleton birth weight were restricted to pregnancies resulting in live birth or stillbirth. 38 , 39 Individual pregnancies were excluded if year of pregnancy was unknown or the woman had been diagnosed with diabetes, cardiovascular disease, or cancer before the pregnancy. We estimated the relative risks (RRs) and 95% CIs for history of laparoscopically confirmed endometriosis and adverse pregnancy outcomes using log-binomial regression models. 40 In a few instances, the models did not converge and log-Poisson regression was used. Log-Poisson regression provides consistent estimates of the RR when empirical standard errors are applied. 41 Generalized estimating equations with an exchangeable working correlation structure were used to account for the within-person correlation between pregnancies. Covariates with missing values were assigned to the largest category, or missing indicator variables were created where appropriate. The primary model was adjusted for age at pregnancy (years), date of pregnancy in relation to cohort enrollment (1989 or before, after 1989), and an interaction term between date of pregnancy and endometriosis diagnosis. Multivariable models were additionally adjusted for race (white, other race; 1.5% missing assigned to white race), age at menarche (younger than 11, 12–13, 14 or more years of age; 0.3% missing assigned to largest category [12–13]), menstrual cycle length between ages 18 and 22 years (less than 26, 26–31, 32–39, 40 or more days or irregular; 0.3% missing assigned to largest category [26–31]), body mass index (calculated as weight in kilograms divided by height in meters squared) at age 18 years (less than 18.5, 18.5–22.4, 22.5 or greater; 0.8% missing assigned to largest category [18.5–22.4]), smoking status (never, past, current; 1.6% missing assigned to largest category [never smokers]), alcohol intake (0, less than 5, 5 or more g/d,unknown or missing; 0.6% missing included in indicator variable), parity (continuous; 0 missing by exclusion criteria described above), and history of infertility (yes, no, unknown or missing; 0.4% missing included in indicator variable). Analyses of birth weight were additionally adjusted for gestational age at delivery (less than 37 weeks of gestation, 37 weeks of gestation or greater). Analyses of hypertensive disorders of pregnancy and preterm birth were additionally adjusted for multiple gestation (singleton, multiples). Given the established higher risk of adverse pregnancy outcomes in women with advanced maternal age, 42 – 44 we estimated risks separately for pregnancies in women aged younger than 35 years compared with those in women aged 35 years or older ( Appendix 2 , available online at http://links.lww.com/AOG/B463 ). We also stratified by parity, given the associations between parity and many adverse outcomes. 45 – 48 Additionally, in women with endometriosis, those with a history of infertility may represent a different disease phenotype 12 and may have a different likelihood of laparoscopic evaluation than women without a history of infertility, so we estimated risks separately by history of infertility. Women with endometriosis and no history of infertility are more likely to have presented clinically with pain rather than infertility symptoms, whereas women with endometriosis and infertility may have been diagnosed with endometriosis incidentally through an infertility evaluation. 8 We tested for differences between groups using cross-product terms in the final models, with calculation of a two-sided Wald test for heterogeneity. Most pregnancies in our population occurred before cohort enrollment in 1989 (n=153,965, 77%). Adjustment for year of pregnancy and an interaction term between endometriosis diagnosis and pregnancy before or after cohort enrollment were included in all models. In sensitivity analyses, we restricted our population to pregnancies occurring after 1989 ( Appendix 1 , available online at http://links.lww.com/AOG/B463 ). Additionally, to reduce misclassification, pregnancies in women who reported endometriosis without laparoscopic confirmation were excluded in the main analysis but were later included as endometriosis diagnoses in sensitivity analyses. 31 Lastly, there is a known delay between endometriosis symptom onset and subsequent disease diagnosis; in the Nurses’ Health Study II, the average diagnostic delay was approximately 4 years, 4 and international multicenter studies have observed an average delay of 7 years. 49 Thus, sensitivity analyses were conducted that predated endometriosis diagnoses by 4 and 6 years. We used SAS 9.3 software for all analyses.

Among women with eligible pregnancies, 8,875 (4.5%) had laparoscopically confirmed endometriosis. Compared with pregnancies in women without a history of endometriosis, pregnancies in women with a history of endometriosis were more likely to occur in women with a younger age at menarche and a history of oral contraceptive use and who experienced infertility and were underweight at age 18 years ( Table 1 ). All covariates presented in Table 1 were statistically significantly different between women with and without a diagnosis of endometriosis ( P <.001). These may be statistically significant owing to large sample size and may not reflect meaningful clinical differences. At the end of study follow-up, the mean [SD] and median number of pregnancies were similar between women with (mean 1.49 [1.18], median 2.0) and without endometriosis (mean 1.97 [1.21], median 2.0). Laparoscopically confirmed endometriosis was associated with a greater risk of pregnancy loss in multivariable adjusted models ( Table 2 ) (spontaneous abortion: RR 1.40, 95% CI 1.31–1.49; ectopic pregnancy: RR 1.46, 95% CI 1.19–1.80; absolute risk of spontaneous abortion: 19.3% in women with endometriosis vs 12.3% in women without endometriosis; absolute risk of ectopic pregnancy: 1.8% in women with endometriosis vs 0.8% in women without endometriosis). There was a suggestion of an association between pregnancies in women with endometriosis and stillbirth (stillbirth: RR 1.27, 95% CI 1.01–1.60, P =.04; absolute risk of stillbirth: 1.9% in women with endometriosis vs 1.4% in women without endometriosis). However, the relationship with stillbirth should be interpreted cautiously given the small sample size. Indeed, using a more stringent statistical significance threshold and 99% CIs, given the multiple comparisons in the analysis, this would not meet the dichotomized threshold for statistical significance. In supplemental analyses, our data suggested that relation with spontaneous abortion was stronger for pregnancies in women at younger ages (younger than 35 years) ( P test for heterogeneity=.02) and in first pregnancies ( P test for heterogeneity=.02), whereas the relation with ectopic pregnancy was stronger for pregnancies in women without a history of infertility ( P test for heterogeneity=.03) ( Appendix 2 , http://links.lww.com/AOG/B463 ). Laparoscopically confirmed endometriosis was associated with higher risk of adverse outcomes during pregnancy. Women with a history of laparoscopically confirmed endometriosis had a 35% greater risk of GDM (RR 1.35, 95% CI 1.11–1.63; 4.3% absolute risk) and a 30% greater risk of hypertensive disorders of pregnancy (RR 1.30, 95% CI 1.16–1.45; 9.5% absolute risk) compared with women without endometriosis (GDM: absolute risk 3.0%; hypertensive disorders of pregnancy: absolute risk 6.6%) ( Table 2 ). In supplemental analyses, our data suggested that the relation between laparoscopically confirmed endometriosis and GDM was stronger in pregnancies in women at younger ages (younger than 35 years) ( P test for heterogeneity=.02), in pregnancies in women without a history of infertility ( P test for heterogeneity=.03), and in second or later pregnancies ( P test for heterogeneity=.02) ( Appendix 2 , http://links.lww.com/AOG/B463 ). Our data also suggested that the relation between laparoscopically confirmed endometriosis and hypertensive disorders of pregnancy was stronger in second or later pregnancies ( P test for heterogeneity=.03) ( Appendix 2 , http://links.lww.com/AOG/B463 ). Relations with GDM attenuated when analyses were restricted to incident pregnancies occurring after cohort enrollment in 1989 ( Appendix 1 , http://links.lww.com/AOG/B463 ). Our data suggested that pregnancies in women with a history of laparoscopically confirmed endometriosis had a 16% greater risk of preterm birth compared with pregnancies in women without endometriosis (RR 1.16, 95% CI 1.05–1.28; absolute risk of preterm birth: 12.0% in women with endometriosis vs 8.1% in women without endometriosis) ( Table 2 ). The data suggested that the relation was stronger in second or later pregnancies ( P test for heterogeneity=.03) ( Appendix 2 , http://links.lww.com/AOG/B463 ) and was strongest in incident pregnancies that occurred after enrollment in 1989 (RR 1.37, 95% CI 1.19–1.57) ( Appendix 1 , http://links.lww.com/AOG/B463 ). Our data also suggested that pregnancies in women with endometriosis were associated with a 16% greater risk of low birth weight (RR 1.16, 95% CI 1.03–1.29; absolute risk of low birth weight: 5.6% in women with endometriosis vs 3.6% in women without endometriosis) ( Table 2 ). Reported relationships for all adverse outcomes were consistent in sensitivity analyses in which endometriosis was predated by 4 or 6 years to account for diagnostic delay and in which the definition of endometriosis was expanded to include non–laparoscopically confirmed endometriosis cases to assess diagnostic bias (data not presented).

We investigated the relationship between a history of laparoscopically confirmed endometriosis and adverse pregnancy outcomes in a large, well-characterized cohort of women. We found that women with a history of endometriosis were at greater risk of pregnancy loss, GDM, hypertensive disorders of pregnancy, preterm birth, and low birth weight. Recently, the relationship between endometriosis and adverse pregnancy outcomes has been an important topic of research 13 , 14 , 21 but has yielded mixed results. The authors of a recent systematic review concluded that there were a great deal of inconclusive and inconsistent findings, 14 possibly influenced by significant heterogeneity across studies in 1) the population under study, 2) the definition of endometriosis, and 3) the source of the data. For example, much of the research to date has been conducted among women attending infertility clinics, which may conflate the influence of advanced maternal age, fertility treatment, and infertility itself with endometriosis, given the known elevated risk of adverse pregnancy outcomes in this population. 22 , 23 , 50 , 51 Additionally, there has been substantial heterogeneity in the definition of endometriosis previously used, with variation in the need for pathologic confirmation, American Society for Reproductive Medicine endometriosis staging, 52 endometriomas, and a focus on treatment for endometriosis. Moreover, registry-based studies, although large, may be prone to misclassification of endometriosis and pregnancy outcomes (eg, miscarriage), because they may represent only the most severe disease phenotypes and may have limited information on confounding and modifying factors. Progesterone resistance in women with endometriosis is hypothesized to lead to dysregulation of genes important in embryo implantation, which may ultimately lead to pregnancy loss. 12 , 14 We found that women with endometriosis had a greater risk of spontaneous abortion and ectopic pregnancy compared with women with no history of endometriosis. Although prior research on this topic has been mixed, 53 – 55 possibly owing to heterogeneity in endometriosis definitions and populations, our data, using a prospective cohort study investigating surgically confirmed endometriosis cases, support the findings of a recent meta-analyses that found that women with endometriosis had a 75% greater risk of spontaneous abortion (n=9 studies) 21 and a 29% greater risk of stillbirth (n=7 studies) 13 compared with women without endometriosis. Our findings are also in accordance with the largest cohort study conducted to date (n=123,335), 56 which reported a 20% greater risk of spontaneous abortion and a 90% greater risk of ectopic pregnancy in naturally conceiving women with endometriosis compared with women with no history of endometriosis. Although there have been mixed results for the relationship of endometriosis and GDM, 57 the most recent meta-analysis of 12 studies reported a 26% greater risk of GDM in women with endometriosis compared with women without endometriosis, which appeared to attenuate when restricted to women with infertility using assisted reproductive technology. 13 Our data support this overall pattern, as well as its heterogeneity. We found that, although endometriosis was associated with GDM in women both with and without a history of infertility, the relationship was stronger in women without a history of infertility. Women with infertility may be at greater risk of GDM overall, 51 thus attenuating a possible relation with endometriosis. Additionally, women with endometriosis without infertility may be more representative of the endometriosis-associated pain phenotype. We found that the relation between endometriosis and GDM was stronger for pregnancies in younger women and also in second or later pregnancies. In our population, the relationship attenuated when restricted to incident pregnancies after 1989. The risk of GDM increases with age; thus, the difference we report may signify that, as age increases, the influence of endometriosis may not be as strong as other competing age-related risk factors. National-level data have also shown a dramatic increase in GDM diagnoses between 1989 and 2004. 58 In our population, the incidence of GDM increased among all pregnancies after 1989. The reported stronger association in pregnancies in younger women and the temporal trends in GDM across time may help explain the attenuation in sensitivity analyses restricted to pregnancies after 1989. The etiology of and risk factors for preeclampsia are not well understood; however, inflammation and abnormal placentation have been hypothesized to contribute. The literature on endometriosis and hypertensive disorders of pregnancy is mixed, with some studies suggesting an increased risk, some no association, and some a decreased risk. 14 Heterogeneity in the relation between endometriosis and hypertensive disorders of pregnancy has been reported by endometriosis disease stage 54 and use of assisted reproductive technology procedures in women with endometriosis. 22 Our findings support the two largest studies to date, which found that women with endometriosis had a 13% greater risk of preeclampsia using Swedish national registries (n=1,442,675) 26 and a 37% greater risk of preeclampsia using Danish national registries (n=82,793) 25 compared with women with no history of endometriosis. A recent meta-analysis of 13 studies also suggests a greater risk of preeclampsia in women with endometriosis (odds ratio [OR] 1.18, 95% CI 1.01–1.39). 13 We found a greater risk of hypertensive disorders of pregnancy in our overall population, with the relation stronger in second or later pregnancies. Women with endometriosis have an inflammatory state both locally and systemically. 15 , 16 , 59 Increased inflammation is one mechanism thought to increase risk of preterm birth. 17 As with many of the prior outcomes discussed, the existing literature has been mixed with regard to the relationship between endometriosis and preterm birth; however, two recent meta-analyses have reported greater risk of preterm birth in women with endometriosis (OR 1.70, 95% CI 1.40–2.06) 13 ; OR 1.49, 95% CI 1.30–1.70 60 ). Additionally, recent findings from the Aarhus Birth cohort reported a greater risk of preterm birth for women with endometriosis, with the risk being highest for very preterm birth. 25 Findings from our study suggested a greater risk of preterm birth for women with endometriosis compared with women without endometriosis. This relationship was stronger in second or later pregnancies. Endometriosis may lead to growth restriction through inadequate uterine contractility and improper placentation. 14 Our results suggested that pregnancies in women with endometriosis may be associated with low birth weight compared with pregnancies in women without endometriosis. A recent meta-analysis of 12 studies shows a borderline association between endometriosis and low birth weight (OR 1.13, 95% CI 1.00–1.27) 13 —similar to our observed modest association (RR 1.16). Although there are many strengths of this study, including the large sample size and detailed exposure and outcome definitions, we also must recognize its limitations. The majority of pregnancies in the Nurses’ Health Study II cohort occurred before the longitudinal cohort began in 1989. Our main analysis included all pregnancies that occurred in our cohort of women whether the pregnancy was before or after cohort initiation. However, for the pregnancies before 1989, the exact timing of endometriosis diagnosis in relation to the pregnancy itself is not known. If a diagnosis of laparoscopically confirmed endometriosis was reported before 1989, we assumed that the pregnancies before 1989 were exposed to endometriosis, as has been done in prior research. 25 This assumption may be robust given the substantial diagnostic delay (4–7 years) between endometriosis symptom onset and disease diagnosis and the lower rate of endometriosis diagnosis after a first pregnancy. 27 Moreover, there is a complete lack of knowledge regarding the natural chronologic history of extrauterine implants and emergence of symptoms. In all analyses, year of pregnancy was adjusted for and an interaction term between endometriosis diagnosis and year of pregnancy was added to account for differences before and after cohort inception in 1989. Additionally (as presented in Appendix 1 , http://links.lww.com/AOG/B463 ), we conducted analyses restricting our population to incident pregnancies occurring after enrollment in 1989. However, this population represents the subset of women having pregnancies at older ages and higher parities and is less generalizable to all pregnancies. We believe that diagnostic bias for pregnancies before 1989 is unlikely given the invasive surgical diagnosis necessary for endometriosis confirmation in our analytic population. However, for women experiencing repeated miscarriage who receive an infertility evaluation or for women experiencing an ectopic pregnancy who have corrective surgery, there may be the possibility of diagnostic bias for pregnancies before 1989 because endometriosis may be incidentally diagnosed under these clinical circumstances. However, we found no difference in the prevalence of joint endometriosis and infertility diagnoses in pregnancies in women with recurrent spontaneous abortion compared with those with normal pregnancies before 1989. Additionally, the vast majority (approximately 90%) of ectopic pregnancies are medically managed without surgery, 61 , 62 which minimizes the risk of diagnostic bias in these outcomes. Endometriosis was defined by surgical visualization. Within this large, geographically diverse cohort across multiple decades, information was not routinely nor uniformly available on endometriosis biopsy-confirmed pathology, lesion location, revised American Society for Reproductive Medicine stage, or adenomyosis status. Additionally, we did not have sufficient sample size nor details to investigate the pregnancy endpoints placenta previa, accreta, or incompetent cervix, which may influence the adverse pregnancy outcomes reported. All models were adjusted for year of pregnancy, and we conducted subanalyses that stratified by calendar year; however, we may not be fully accounting for temporal patterns in recognition of endometriosis and adverse pregnancy outcomes over time. For some endpoints of interest, there were small number of events, which may limit statistical power. Owing to the known diagnostic delay between symptom onset and disease diagnosis in endometriosis, we conducted sensitivity analyses in which endometriosis diagnoses were predated by 4 and 6 years, and the results remained consistent. As with all studies of pregnancy, there is risk of missing outcomes. Women may experience early miscarriages of which they are unaware, and, thus, these losses go unreported. This is even more frequent in women who are not intentionally trying to conceive. Women with endometriosis may represent a unique population of women at greater risk for adverse outcomes across pregnancy and birth. Given the suggestion of subgroup heterogeneity in the relationship between endometriosis and adverse pregnancy outcomes by maternal age, parity, and infertility history, future research should focus on these differences and on rigorously distinguishing temporality between endometriosis diagnosis and pregnancy outcomes. Elucidating mechanisms of association and possible pathways for intervention or screening procedures will be critical to improve the health of women with endometriosis and their children.