Risk of congenital anomalies among infants of patients with endometriosis: a population-based cohort study

We conducted a population-based cohort study using Ontario administrative health records held at ICES ( www.ices.on.ca ). ICES is an independent, nonprofit research institute with designation under the Ontario Personal Health Information Protection Act to collect and analyze health care data for Ontario residents, without consent, for health system research. ICES maintains validated data sets that have a unique maternal–infant matching number to link maternal–infant pairs and capture information on pregnancy and pregnancy outcomes; Appendix 1, Supplemental Table 1 (available at www.cmaj.ca/lookup/doi/10.1503/cmaj.250439/tab-related-content ) outlines the databases and variables used. We included all live births at 20 weeks’ gestation or later to patients aged 15 to 50 years in Ontario, Canada, from 2006 to 2021. We obtained pregnancy characteristics and neonatal outcomes from the Better Outcomes Registry & Network Ontario (BORN) available at ICES, which captures 99% of live hospital births in Ontario. 14 We obtained information about the variables used in the mediation analysis from BORN and Canadian Institute of Health Information (CIHI) databases ( Appendix 1, Supplemental Table 1 ). We obtained patient demographic characteristics and pre-existing health conditions through linkage of ambulatory records, hospital admission records, and same-date surgery records, which are mandatory submissions from hospitals to the Ministry of Health. Income quintile was based on median income for the first 3 characters of the postal code of patient residence, which was ranked into quintiles within each census metropolitan area. We excluded births delivered by a gestational carrier, as this population exhibits a different health risk profile. 15 , 16 We also excluded maternal–infant pairs with less than 5 years of Ontario Health Insurance Plan (OHIP) eligibility before the conception date, to prevent under-ascertainment of endometriosis. To obtain diagnoses of endometriosis, we used a validated algorithm using the maternal Discharge Abstract Database. We defined endometriosis as the presence of a medical diagnosis of endometriosis (≥ 2 medical consults billed as International Statistical Classification of Diseases and Related Health Problems, 9th Revision [ICD-9] code 617 OHIP diagnosis of endometriosis or a hospital admission for endometriosis ICD-9 code 617 or ICD-10 code N80 without surgery), or 1 of several obstetric surgical procedures coded with a diagnosis of endometriosis (ICD-9 code 617 or ICD-10 code N80) before the index pregnancy, as previously used. 3 , 17 – 19 The diagnosis of endometriosis in the administrative health data sets had a positive predictive value of 75% (95% confidence interval [CI] 68.5% to 81.5%). 20 Additionally, the sensitivity and specificity were both high (75%, 95% CI 68.5% to 81.5%; 95.2%, 95% CI 93.8% to 96.9%, respectively), suggesting a high level of validity for diagnoses related to endometriosis. 20 We obtained diagnosis codes for congenital anomalies from the CIHI Discharge Abstract Database. We classified anomalies using the algorithm from the Metropolitan Atlanta Congenital Defects Program, a surveillance system of the US Centers for Disease Control and Prevention, 21 as done in other studies. 22 – 25 We defined congenital anomalies using ICD-10 codes (Q000 to Q999) and categorized them as any congenital anomaly, organ system anomaly, or specific anomalies, as previously done in ICES studies. 22 , 26 We used standardized differences to assess differences in baseline characteristics between the exposed and unexposed groups (patient age and income quintile, and those identified a priori as related to both endometriosis and anomalies: chronic diabetes, chronic hypertension, smoking, alcohol use, substance use, parity, obesity, and Elixhauser Comorbidity Index score 7 , 22 , 27 – 34 [ Appendix 1, Supplemental Table 2 ]). We included covariates in the adjusted model based on a standardized difference of greater than 0.1 or a priori considerations, or both. 35 We derived relative risks (RRs) using modified Poisson regression models with robust error variance to account for repeated pregnancies by the same patient during the study period. 36 An odds ratio (OR) will exaggerate the underlying risk ratio when the prevalence of outcomes is greater than 5%. Although the prevalence of defects was generally less than 5%, we selected to directly estimate risk ratios, as that was the parameter of interest. A modified Poisson regression approach does not have convergence issues that arise with log-binomial regression. 37 The sandwich estimation in a robust error variance procedure used in modified Poisson regression addresses the issues of overestimation of error when inappropriately applied to binomial data. 37 We estimated RRs between endometriosis exposure and any infant congenital anomaly, organ system anomalies, and specific anomalies. We used mediation analysis to decompose the association between endometriosis and congenital anomalies (i.e., the total effect) into the association between endometriosis and congenital anomalies in the absence of the mediator (i.e., the natural direct effect) and the association between each mode of fertility and mode of conception and anomalies (the natural indirect effect), 38 , 39 as shown in Appendix 1, Supplemental Figure 1 . We then estimated the proportion of the total effect between endometriosis and congenital anomalies mediated through each of subfertility, ovulation induction, or intrauterine insemination, and in vitro fertilization or intracytoplasmic injection using unassisted conception as the reference. The proportion mediated was calculated by (RR natural indirect effect – 1) / (RR total effect – 1), where RR total effect = (RR natural direct effect × RR natural indirect effect ). All mediation analysis models adjusted for the same covariates as the main model, including patient age at delivery, income quintile, chronic diabetes, chronic hypertension, smoking, alcohol use, substance use, parity, obesity, and comorbidity score. We conducted sensitivity analyses to assess the robustness of our results: restricting to singleton births given the association between multifetal pregnancy and congenital anomalies, 40 and infertility and multifetal pregnancy; 41 conducting multilevel simple bias analysis to remove possible unmeasured confounding by maternal race; 42 – 45 and stratifying based on method of endometriosis diagnosis, either surgical or medical. We analyzed data using SAS version 9.4. We reported the study in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology reporting guideline for cohort studies ( Appendix 1, Supplemental Table 3 ). This study was approved by the Queen’s University Health Sciences and Affiliated Teaching Hospitals Research Ethics Board.

Of the 1 460 564 eligible births in Ontario between Oct. 1, 2006, and Mar. 31, 2021 ( Figure 1 ), 33 619 (2.3%) were infants of patients with endometriosis (exposed) and 1 426 835 (97.7%) were not (unexposed). The mean time from endometriosis diagnosis to first pregnancy was 6.5 (SD 4.5) years. Cohort creation flowchart. Note: IKN = ICES Key Number, OHIP = Ontario Health Insurance Plan. See Related Content tab for accessible version. Compared with patients without endometriosis, those with endometriosis were older at delivery (mean 33.4 [SD 4.8] yr v. 30.7 [SD 5.4] yr), were less likely to be in the lowest income quintile (15.5% v. 20.4%), and were more likely to be subfertile (44.5% v. 15.1%) or conceive via ovulation induction or intrauterine insemination (3.9% v. 1.4%) or in vitro fertilization or intracytoplasmic sperm injection (10.6% v. 1.5%) ( Table 1 ). Characteristics of patients with and without endometriosis who delivered a live birth in Ontario between Oct. 1, 2006, and Mar. 31, 2021 Note: SD = standard deviation. Unless stated otherwise. A total of 2120 (6.3%) infants who had any congenital anomaly were born to a patient with endometriosis, and 77 094 (5.4%) infants with a congenital anomaly were born to a patient without endometriosis ( Table 2 ). In adjusted models, exposed infants had an increased risk of any congenital anomaly (adjusted RR 1.16, 95% CI 1.12 to 1.21). Other covariates independently associated with any congenital anomaly included chronic diabetes, chronic hypertension, higher Elixhauser Comorbidity Index score, substance use, parity, alcohol use in pregnancy, obesity, and smoking ( Appendix 1, Supplemental Table 4 ). Relative risk of congenital anomalies for infants of patients with and without endometriosis Note: CI = confidence interval, RR = relative risk. Relative risks were adjusted for maternal age at delivery, income quintile, rurality, parity, obesity, chronic diabetes, chronic hypertension, smoking, alcohol, substance use, and Elixhauser Comorbidity Index score. Figure 2 shows the unadjusted and adjusted relative risks between exposure to endometriosis and any anomaly, and anomalies classified by organ system. We found some evidence of confounding for ear, face, and neck anomalies, as the adjusted RR was attenuated compared with the crude RR. Endometriosis was associated with increased risk of cardiovascular anomalies (adjusted RR 1.23, 95% CI 1.12 to 1.35), gastrointestinal anomalies (adjusted RR 1.13, 95% CI 1.04 to 1.23), genital anomalies (adjusted RR 1.27, 95% CI 1.11 to 1.45), musculoskeletal anomalies (adjusted RR 1.17, 95% CI 1.05 to 1.31), and neoplasms and tumours (adjusted RR 1.26, 95% CI 1.01 to 1.58). We found a higher risk of atrial septal anomalies (adjusted RR 1.24, 95% CI 1.09 to 1.41), ventricular septal anomalies (adjusted RR 1.41, 95% CI 1.03 to 1.94), stenosis of the pulmonary arteries (adjusted RR 1.41, 95% CI 1.03 to 1.94), unspecified cleft palates (adjusted RR 1.52, 95% CI 1.03 to 2.25), hypospadias (adjusted RR 1.47, 95% CI 1.04 to 2.09), undescended testes (adjusted RR 1.36, 95% CI 1.10 to 1.69), and unspecified hypospadias (adjusted RR 1.44, 95% CI 1.12 to 1.85) in exposed infants compared with unexposed infants ( Figure 3 ). Forest plot of the adjusted risk of any congenital anomaly and organ system anomalies in infants born to patients with endometriosis. Relative risks were adjusted for maternal age at delivery, income quintile, chronic diabetes, chronic hypertension, smoking, alcohol use, substance use, parity, obesity, and comorbidity score. Note: CI = confidence interval, RR = relative risk. See Related Content tab for accessible version. Forest plot of the adjusted risk of specific anomalies in infants of patients with endometriosis. Relative risks were adjusted for maternal age at delivery, income quintile, chronic diabetes, chronic hypertension, smoking, alcohol use, substance use, parity, obesity, and comorbidity score. CI = confidence interval, RR = relative risk. See Related Content tab for accessible version. In mediation analysis, endometriosis remained independently associated with congenital anomalies. In vitro fertilization or intracytoplasmic injection had a small mediating effect overall, accounting for 11.0% of the association between endometriosis in the patient and any congenital anomaly, but had a stronger mediating effect for a few specific anomalies: 23.6% for stenosis of the pulmonary artery, 47.0% for neoplasms and tumours, and 75.6% for hemangiomas ( Table 3 ). Subfertility or ovulation induction or intrauterine insemination did not mediate the relation between endometriosis and risk of congenital anomalies. Mediation analysis by mode of conception for congenital anomalies showing a significant association with endometriosis Note: CI = confidence interval, NDE = the direct effect, NIE = the indirect effect, RR = relative risk. Relative risks were adjusted for maternal age at delivery, income quintile, rurality, parity, obesity, pre-existing diabetes, chronic hypertension, smoking, alcohol, substance use, and Elixhauser Comorbidity Index score. NA = Results are due to the natural direct effect and the natural indirect effect operating in opposite directions. The proportion mediated is rendered meaningless when the natural direct and indirect effects are approximately the same magnitude, but opposite signs. 38 The direct and indirect effect were not meaningful. In sensitivity analyses restricted to singleton pregnancies ( Appendix 1, Supplemental Table 5 ), endometriosis in the patient remained associated with any congenital anomaly in their infant (adjusted RR 1.15, 95% CI 1.10 to 1.20). Most associations with specific anomalies and organ systems were of similar magnitude to the primary analyses, except that associations with neoplasms and tumours and with stenosis of the pulmonary artery were attenuated and no longer significant. A simple bias analysis to assess for possible unmeasured confounding by race produced similar results to the main analysis for any anomaly (adjusted RR 1.17, 95% CI 1.12 to 1.22) and specific anomalies and organ systems ( Appendix 1, Supplemental Tables 6 and 7 ). Increased risks of any congenital anomaly were observed in association with endometriosis based on a surgical diagnosis (adjusted RR 1.14, 95% CI 1.09 to 1.20) or a medical diagnosis (adjusted RR 1.22, 95% CI 1.13 to 1.32) ( Appendix 1, Supplemental Tables 8 and 9 ). In analyses stratified by method of diagnosis, associations with specific anomalies and organ systems were generally similar to those in the primary analysis, but statistical power was significantly reduced with this stratification and many anomalies could not be included because of small cells.

In this population-based cohort study, endometriosis in the birthing parent was associated with an increased risk of any anomaly, cardiovascular anomalies, gastrointestinal anomalies, genital anomalies, musculoskeletal anomalies, and neoplasms and tumours in offspring. Most other organ systems had associations of similar magnitudes for which analyses may have been underpowered. Specific anomalies with increased risk included atrial septal anomalies, ventricular septal anomalies, stenosis of the pulmonary arteries, cleft palate, hypospadias, and undescended testes. Although we confirmed endometriosis to be an independent risk factor for congenital anomalies in the mediation analysis, in vitro fertilization or intracytoplasmic injection had a small but significant mediating effect between endometriosis and the risk of congenital anomalies, whereas subfertility alone or fertility treatment with ovulation induction or intrauterine insemination did not. Whether endometriosis was diagnosed surgically or medically in patients did not appear to modify the risk of congenital anomalies in their infants. Although we observed modest relative increases in risk, the absolute risk of congenital anomalies for infants born to patients with endometriosis remained low, because congenital anomalies are uncommon. Fertility treatments, specifically in vitro fertilization and intracytoplasmic injection, have been associated with congenital anomalies. 11 , 12 , 46 A recent study reported an increased risk of any congenital anomaly (mainly cardiovascular and musculoskeletal) in infants born to women with endometriosis who conceived by in vitro fertilization compared with those without endometriosis who conceived via in vitro fertilization. 47 Our findings show a similar increased risk of any anomaly, cardiovascular anomalies, and musculoskeletal anomalies in patients with endometriosis and further suggest that most of this increased risk was not due to in vitro fertilization. Two prior studies have assessed the association between endometriosis and congenital anomalies, specifically the risk of genital defects among male infants. 7 , 8 A population-based cohort study found a small increased risk between endometriosis and cryptorchidism (adjusted hazard ratio 1.27, 95% CI 0.96 to 1.88). 8 A case–control study found a similar association with cryptorchidism (OR 2.43, 95% CI 1.71 to 3.42). 7 We had similar findings of an increased risk of hypospadias and undescended testes. The increased risk of anomalies persisted after exclusion of multifetal births, except for cleft anomalies and neoplasms and tumours. Thus, it is possible that the increased risk for cleft anomalies and neoplasms and tumours is partially explained by multifetal births, which are more common in pregnancies conceived by fertility treatment. 41 As would be expected, the indirect effect of subfertility was attenuated compared with the direct effect of endometriosis: Those with subfertility who did not require fertility treatment to conceive may have had less severe endometriosis. A dose-dependent relation between severity of endometriosis and fertility has been observed; approximately 50% of those with mild endometriosis can conceive without fertility treatment, whereas 25% of those with moderate disease can conceive without treatment, and very few with severe disease can conceive. 48 Additionally, those with more severe stages of endometriosis have a worse prognosis for in vitro fertilization or intracytoplasmic injection treatments than those with less severe stages, and could require more infertility treatment cycles to achieve pregnancy. 49 Thus, it is possible that stage of endometriosis is important to consider in estimating causal effect on risk of congenital anomalies and could potentially explain the differences in the proportions mediated by in vitro fertilization or intracytoplasmic injection compared with ovulation induction or intrauterine insemination or subfertility. Although this study contributes to accumulating data on a potential increased risk of birth defects for infants born to patients with endometriosis, the specific mechanistic pathways remain largely unknown. A theorized mechanism by which endometriosis might increase risk of any congenital anomaly is through inflammatory pathways. Genome sequencing studies found that DNA methylation, a change linked to chronic inflammation, 50 was associated with repression of developmental programming of cells, as well as the repression of the activation of genes involved in tissue-specific processes. 51 Additionally, endometriosis has been associated with increased risk of placental abnormalities through inflammatory pathways. 52 The existing literature suggests a need for more comprehensive models that integrate genetic, epigenetic, and environmental factors to better understand this relation. 53 Additionally, there are limited interventions targeted to females with endometriosis or those pregnant after fertility treatment, 54 and these should be considered for future research studies. The prevalence of endometriosis in this cohort was 2.3%, whereas we would expect approximately 10% of reproductive-aged females to have a diagnosis of endometriosis. The lower prevalence might be explained by a higher rate of infertility in females with endometriosis and our cohort of deliveries. 3 Although the positive predictive value for endometriosis coding is relatively high, the low prevalence of endometriosis in our cohort means that even modest false-positive fractions could result in a meaningful proportion of patients who do not have the disease being coded as having endometriosis. This misclassification would tend to bias the estimated association between endometriosis and anomalies toward the null. Regarding the validation of outcome diagnosis codes, a study from Alberta that compared hospital data (ICD-9 and ICD-10 codes) with data from the Alberta Congenital Anomalies Surveillance System estimated a sensitivity of 87.8 (95% CI 73.8 to 95.9) and positive predictive value of 78.3 (95% CI 63.6 to 89.1). 55 Thus, some defects would be missed and other conditions would be falsely misclassified as defects, which would likely attenuate the results toward the null. Additionally, exposure to fertility treatments is captured by trained data abstractors and is likely to be highly reliable. 56 However, it is possible that some pregnancies achieved via fertility treatment were not captured, which could result in nondifferential misclassification and underestimate the effect on congenital anomalies. Although selection bias from miscarriage of pregnancies with fetal anomalies is a potential concern in this retrospective cohort study, the observed risk of miscarriage is comparable in those with endometriosis (18.9%) and without endometriosis (17.3%), and thus selection bias is unlikely to have attenuated our observed associations. 57 Limited data were available that estimated the associations between race and congenital anomalies to inform a bias analysis for all congenital anomalies in our study. 43 , 44 Thus, the bias analysis was limited to organ-system anomalies and few specific anomalies. Pregnancies delivered at less than 20 weeks’ gestation are not recorded in the study data sets. We examined possible selection bias resulting from early pregnancy loss, and the results were similar, suggesting little effect on the observed findings. The results persisted after adjustment for covariates that might confound the association between endometriosis and risk of congenital anomalies, showing no evidence of confounding. The study is limited by the availability of the variables captured in administrative data; thus, there is potential for unmeasured confounding. For example, the stage of endometriosis and number of in vitro cycles may be important for parameterization, 58 but they were unavailable in our health administrative data set. Additionally, nonsteroidal anti-inflammatory drugs are often used to manage pain associated with endometriosis, 59 and patients with endometriosis are more likely to use opioid analgesics (adjusted OR 2.76, 95% CI 3.57 to 4.06). 60 Our study did not assess the mediating role of pain management on risk of congenital anomalies. Infants born to patients with endometriosis have a small but significantly increased risk of congenital anomalies that is partially mediated by the use of in vitro fertilization or intracytoplasmic sperm injection. Future studies are needed to elucidate potential mechanisms.