Pregnancy outcomes in women with endometriosis: a national record linkage study

Transcription

1 DOI: / Pregnancy outcomes in women with endometriosis: a national record linkage study L Saraswat, a DT Ayansina, b KG Cooper, a S Bhattacharya, c D Miligkos, d AW Horne, e S Bhattacharya f a Ward 315, Department of Gynaecology, Aberdeen Royal Infirmary, Aberdeen, UK b Medical Statistics Team, Division of Applied Health Sciences, University of Aberdeen, Aberdeen, UK c Dugald Baird Centre, Aberdeen Maternity Hospital, Aberdeen, UK d Department of Obstetrics and Gynaecology, University Hospitals of Southampton, Southampton, UK e MRC Centre for Reproductive Health, University of Edinburgh, Edinburgh, UK f Head of Division of Applied Health Sciences, School of Medicine and Dentistry, University of Aberdeen, Aberdeen, UK Correspondence: Dr L Saraswat, Department of Gynaecology, Aberdeen Royal Infirmary, Foresterhill, Aberdeen, AB25 2ZN, UK. lucky. saraswat@nhs.net Accepted 4 January Published Online 16 February Objective To determine pregnancy outcomes in women with endometriosis. Design A national population based cohort study using record linkage. Setting Scotland. Participants A cohort of women followed up over a 30- year period ( ). Methods In a nationwide Scottish study, we compared pregnancy outcomes in 5375 women with surgically confirmed endometriosis with outcomes in 8710 women without endometriosis who were pregnant during the same time period. Data were analysed using univariable and multivariable logistic regression after adjusting for confounding factors. Main outcome measures Outcome measures evaluated included miscarriage, ectopic pregnancy, stillbirths and other pregnancy complications such as hypertensive disorders of pregnancy, antepartum and postpartum haemorrhage, operative delivery and preterm births. The outcomes were presented as adjusted odds ratios (OR) with 95% confidence intervals (CI). Results On multivariable analysis, after adjusting for age, parity, socio-economic status and year of delivery, women with endometriosis when compared to women without endometriosis, had a significantly higher risk of early pregnancy complications with adjusted OR (95% CI) of 1.76 (1.44, 2.15) and 2.70 (1.09, 6.72) for miscarriage and ectopic pregnancy, respectively. A previous diagnosis of endometriosis was associated with a significantly increased risk of [adjusted OR (95% CI)] placenta praevia [2.24 (1.52, 3.31)], unexplained antepartum haemorrhage [1.67 (1.39, 2.00)], postpartum haemorrhage [1.30 (1.61, 1.46)] and preterm births [1.26 (1.07, 1.49)] in pregnancies progressing beyond 24 weeks. Conclusion Endometriosis predisposes women to an increased risk of early pregnancy loss and later pregnancy complications. Keywords Endometriosis, early pregnancy outcomes, late pregnancy complications. Tweetable abstract Endometriosis predisposes women to an increased risk of early pregnancy loss and later pregnancy complications. Please cite this paper as: Saraswat L, Ayansina DT, Cooper KG, Bhattacharya S, Miligkos D, Horne AW, Bhattacharya S. Pregnancy outcomes in women with endometriosis: a national record linkage study. BJOG 2016; DOI: / Introduction Endometriosis is a complex and chronic gynaecological condition that affects women of reproductive age. It is associated with pelvic pain, subfertility, impaired health-related quality of life and reduced work productivity. 1,2 The true prevalence of endometriosis is unknown, although estimated as 2 10% in the general female population and up to 40% in women with subfertility and/or pelvic pain. 3 5 The pathophysiology of endometriosis is poorly understood and its impact on pregnancy relatively unexplored. In a systematic review, Harb et al. 6 concluded that the presence of endometriosis is associated with lower fertilisation, implantation and clinical pregnancy rates in women undergoing in vitro fertilisation (IVF). All of these outcomes were worse with more advanced stages of endometriosis. 7 Endometriosis has been linked with increased inflammation in the peritoneal cavity with higher concentrations of 1

2 Saraswat et al. cytokines, growth factors and angiogenic factors. 8,9 Other factors implicated include poorer oocyte quality, progestogen resistance, molecular and functional abnormalities in eutopic endometrium, 10 and anatomical distortion of fallopian tubes and ovaries in women with endometriosis; all of which may potentially affect fertility and pregnancy outcomes. The results from the few population-based studies examining the link between endometriosis and adverse pregnancy outcomes are inconsistent Population based studies in women with endometriosis are challenging to perform, as the condition can only be diagnosed with certainty by laparoscopy. There is a need for further large well designed studies to explore the influence of endometriosis on pregnancy. We therefore, conducted a study to determine pregnancy outcomes in women with a previous diagnosis of endometriosis. Our study addresses some of the limitations of previous studies by using a large national cohort of women with a surgical diagnosis of endometriosis and a long duration of follow up. Methods Study design We conducted a population-based cohort study using national Scottish linked data extending over a period of 30 years from 1981 to The Scottish Record Linkage system has developed permanently linked national data sets where all hospital records pertaining to an individual are collated together. These data are stored as Scottish Morbidity Records (SMR) by the Information Services Division (ISD), which is a part of National Health Services. ISD contains information on demographic and clinical details of sequential hospital care (SMR01), pregnancy outcomes (SMR02) and neonatal outcomes [SMR11 and SBR (Scottish Birth record)]. The record linkage is performed using both exact (deterministic) and probabilistic matching to ensure accuracy, minimise discrepancies and identify duplicate entries. 16,17 The database is subjected to regular quality assurance check with completeness estimated to be in excess of 98%. 18,19 Participants The exposed cohort comprised all women in Scotland with a new diagnosis of endometriosis made for the first time at a surgical procedure (laparoscopy or laparotomy) between 1981 and 2009, who then had a subsequent pregnancy recorded until Cases of endometriosis were identified within the data set using ICD codes N80 and 617 from ICD 10 and 9, respectively. The unexposed cohort consisted of a random sample of all women in Scotland with no previous recorded diagnosis of endometriosis, who had a pregnancy during the same time period ( ) as the women with endometriosis. The number of women in the exposed and unexposed cohort were in a 1:1 ratio. Women with multiple births were excluded. Women with suspected diagnosis of endometriosis on the basis of symptoms but absence of surgical confirmation were excluded from both exposed and unexposed cohorts. The follow-up duration in women with endometriosis ranged from a minimum of 1 to a maximum of 30 years. Data extraction Data on socio-demographic characteristics and potential confounding variables [age, parity, socio-economic status and year of delivery or pregnancy event] were extracted for both exposed and unexposed cohorts. Social class was recorded as Carstairs index by ISD. Carstairs quintiles are ordered from I to V, where V represents the most deprived category. 20 Outcomes We compared pregnancy outcomes in women with endometriosis with those without a previous diagnosis of endometriosis. To prevent the effect of clustering, we only evaluated one pregnancy per woman; for the endometriosis cohort this was the first pregnancy after the diagnosis of endometriosis. For the unexposed group, a random pregnancy was chosen for each woman during the same time frame ( ). Matching between exposed and unexposed cohort was not performed, as a truly random sample should be representative of the population it is derived from. A subset analysis was also performed for primigravid women to eliminate the confounding effect of previous pregnancy outcomes. The study spans a long period of 30 years, during which laparoscopic techniques, diagnosis and surgical treatment of endometriosis have evolved considerably. An additional subset analysis was conducted for women who had a laparoscopic diagnosis of endometriosis in the last decade of the study ( ) to explore the effect of changes in practice. For the purpose of this study, pregnancy outcomes were split into early and late pregnancy outcomes. The early pregnancy outcomes included miscarriage (spontaneous pregnancy loss <24 weeks), ectopic pregnancy (pregnancy outside the uterine cavity) and termination of pregnancy (medically or surgically induced abortion before 24 weeks of pregnancy). Although the study did not aim to evaluate the association of termination of pregnancy (TOP) with endometriosis, TOP was included in the early pregnancy outcomes as it remains one of the endpoints for any pregnancy. The late pregnancy outcomes were only evaluated for women whose pregnancy continued beyond 24 weeks. They included any maternal or perinatal complications in the form of hypertensive disorders of pregnancy (preeclampsia, eclampsia and pregnancy-induced hypertension), 2

3 Pregnancy outcomes in women with endometriosis antepartum haemorrhage (placenta praevia, placental abruption and unexplained antepartum haemorrhage), mode of delivery (spontaneous vaginal, instrumental delivery or caesarean section), postpartum haemorrhage (blood loss >500 ml for vaginal birth or >1000 ml for caesarean delivery), preterm birth (delivery <37 weeks of gestation), low birthweight (birthweight <2500 g), stillbirth (delivery of fetus without any signs of life born after 24 weeks or weighing >500 g) and neonatal death. Each outcome was defined in accordance with International Classification of Diseases (ICD) codes for disease classification as described in Table S1. Statistical analysis We used SPSS version 21.0 for Windows (SPSS Inc., Chicago, IL, USA) for storage, recoding and analysis of data. Descriptive statistics [frequencies and percentages; mean, standard deviation (SD); median interquartile range (IQR) as appropriate] were used to summarise socio-demographic characteristics in the endometriosis and the unexposed groups. Baseline characteristics in the two groups were compared using chi-squared tests for categorical variables and independent samples t-tests or Mann Whitney tests as appropriate for continuous data. All pregnancy outcomes were binary in nature except the mode of delivery, which had three categories (spontaneous vaginal delivery, instrumental delivery and caesarean section). Binary and multinomial logistic regression models were used as appropriate to evaluate association of endometriosis with adverse pregnancy outcomes. Crude and adjusted odds ratios (OR) with 95% confidence intervals (CI) were obtained after adjusting for age, parity, Carstairs quintile and year of pregnancy. A P- value of <0.05 was considered to be statistically significant. Using the figures available for analysis (5375 pregnant women with endometriosis and 8280 unexposed women), we had 87% power to demonstrate a difference of 2% (odds ratio of 1.16) in the risk of miscarriage (assuming a 15% miscarriage rate). In all, 4232 women with endometriosis and 6707 women without endometriosis carried their pregnancy beyond 24 weeks. This gave 96% power to demonstrate a difference of 2% (odds ratio of 1.31) in the risk of preterm birth in women with endometriosis (assuming a 7% preterm birth rate in the unexposed cohort). Appropriate approvals from the Privacy Advisory Committee of ISD, of the National Health Service, Scotland; North of Scotland Research Ethics Committee (NRES) and NHS Research and Development were sought to conduct the study. NRES were of the opinion that formal ethical approval was not required. To maintain privacy and confidentiality, only anonymised linked data sets were provided by ISD and stored in a secure folder in the Grampian Data Safe Haven (DaSH) as per national guidance. Results A total of women had a diagnosis of endometriosis for the first time between 1981 and 2009 in Scotland. Linkage with Scottish Maternity database (SMR02) identified 8710 women, who had a pregnancy subsequent to the diagnosis of endometriosis. An equivalent number (8710) of pregnant women without any recorded diagnosis of endometriosis during the study period were identified randomly from the Scottish Maternity Database. They constituted the unexposed cohort. Women without a surgical diagnosis of endometriosis (n = 2962) were excluded from the exposed cohort. Of the included women, in only 1.3% (n = 71) was endometriosis identified at laparotomy. The remaining 98.7% had a diagnosis confirmed at laparoscopy. Women with multiple births, records with data errors (e.g. >1 or no reproductive outcome recorded) identified during data cleaning and checking that could not be verified by ISD were excluded from both groups. The data extraction and linkage process is illustrated in Figure 1. Thus, 5375 pregnant women with a history of endometriosis and 8280 women without a diagnosis of endometriosis were included in the analysis for the early pregnancy outcomes. Late pregnancy outcomes were reported for women who continued their pregnancy beyond 24 weeks and comprised 4232 women with and 6707 without a diagnosis of endometriosis. The median duration from initial diagnosis of endometriosis to pregnancy (time of birth, miscarriage, ectopic or TOP) was 2 years 4 months (range 2 months to 18 years). The two groups had significantly different socio-demographic characteristics. Pregnant women with endometriosis were significantly older [mean age (SD) of 30.5 (5.2) versus 27.2 (6.1) years], more likely to be nulliparous (72 versus 52%) and from a more affluent social class compared with those without endometriosis (P < 0.001). Smoking was poorly recorded in ISD data with approximately 55 and 68% of data missing in the exposed and the unexposed cohorts, respectively. The analysis of available data demonstrated a significantly lower proportion of current smokers in the endometriosis group. As >50% of data were missing for smoking, neither imputation nor sensitivity analysis was deemed appropriate for dealing with missing data. Smoking was therefore excluded from multivariable analysis. The results of baseline socio-demographic characteristics are summarised in Table 1. Women with endometriosis had significantly higher risks of miscarriage and ectopic pregnancy compared with the non-endometriosis group on univariable analysis. The unadjusted odds of TOP were significantly lower in women with endometriosis. After adjusting for confounders (age, parity, socio-economic status and year of pregnancy), women with endometriosis continued to demonstrate 3

4 Saraswat et al. Exposed cohort Women with history of endometriosis iden fied from SMR01 and linked with SMR02 to iden fy subsequent pregnancies between 1981 to 2010 n = 8710 Unexposed cohort A random sample of women who had a pregnancy between 1981 to 2010 obtained from SMR02 n = 8710 Women excluded from analysis (n = 3335) no surgical diagnosis of endometriosis (n= = 2962) endometriosis diagnosed at the me of pregnancy event pregnancy (n =1) event (n=1) mul ple births (n=203) women with >one reproduc ve outcome (n=10) women with no reproduc ve outcome recorded (n=159) Women excluded from analysis (n =430) mul ple births (n =50) women with >one reproduc ve outcome (n =18) women with no reproduc ve outcome recorded (n =362) Women with endometriosis included in analysis early pregnancy outcomes (n = 5375) late pregnancy outcomes (n = 4232) Unexposed women included in analysis early pregnancy outcomes (n = 8280) late pregnancy outcomes (n = 6707) Figure 1. Reproductive and pregnancy outcomes: linkage and data extraction. Table 1. Socio-demographic characteristics of study participants Characteristics Participants n = Endometriosis n = 5375 No endometriosis n = 8280 Age at delivery (years) Mean (SD) 28.5 (6.0) 30.5 (5.2) 27.2 (6.1) <0.001 Missing (n) Socio-economic status (quintiles) n (%) I (least deprived) 2621 (19.4) 1144 (21.4) 1477 (18.1) <0.001 II 2486 (18.4) 1019 (19.1) 1467 (18) III 2634 (19.5) 1045 (19.6) 1589 (19.5) IV 2856 (21.2) 1152 (21.6) 1704 (20.9) V (most deprived) 2879 (21.4) 976 (18.3) 1903 (23.4) Missing (n) Smoking status n (%) Non smokers 3388 (66.5) 1632 (66.8) 1756 (66.3) Smokers 1111 (21.8) 496 (20.3) 615 (23.2) Ex-smokers 592 (11.6) 315 (12.9) 277 (10.5) Missing (n) Parity n (%) Nulliparous 6943 (60.1) 3315 (72.3) 3628 (52) <0.001 Parous 4613 (39.9) 1265 (27.6) 3348 (48) Missing (n) significantly higher likelihood of early pregnancy complications with adjusted odds ratios (OR) and 95% confidence intervals (CIs) of 1.76 (1.44, 2.15) and 2.70 (1.09, 6.72) for miscarriage and ectopic pregnancy, respectively. No statistically significant differences were noted in the proportion of women having TOP. 4

5 Pregnancy outcomes in women with endometriosis The results of univariable and multivariable analysis for early pregnancy outcomes are shown in Table 2. As shown in Table 3, compared with women without endometriosis, women with endometriosis were prone to have adverse outcomes in ongoing pregnancies beyond 24 weeks. They had a significantly higher risk [adjusted OR (95% CI)] of placenta praevia [2.24 (1.52, 3.31)], unexplained antepartum haemorrhage [1.67 (1.39, 2.00)], postpartum haemorrhage [1.30 (1.61, 1.46)], caesarean section [1.40 (1.26, 1.55)], instrumental delivery [1.21 (1.08, 1.36)] and preterm birth [1.26 (1.07, 1.49)]. No significant association with endometriosis was observed for hypertensive disorders of pregnancy [1.06 (0.91, 1.24)], placental abruption [0.91 (0.48, 1.74)], low birthweight [1.12 (0.94, 1.32)] and neonatal death [3.83 (0.36, 41.36)]. Additional full logistic regression outputs for miscarriage, ectopic pregnancy, placenta praevia and preterm birth are appended in Table S2. A subgroup analysis for primigravid women showed similar results. There were 3315 women who conceived for the first time after a new diagnosis of endometriosis. The corresponding number of primigravid women in the unexposed cohort was After adjusting for age, socioeconomic status and year of pregnancy, women with endometriosis were significantly more likely to have a miscarriage [OR (95% CI), 1.86 (1.45, 2.40)]. The risks of ectopic pregnancy were significantly higher on univariable analysis [4.39 (1.24, 15.57)] but no longer significant after adjustment for confounding factors [2.99 (0.81, 11.02)], although the overall direction of effect was not altered. The risk of placenta praevia, unexplained APH, PPH and caesarean delivery continued to be significantly higher in women with endometriosis [OR (95% CI) 2.28 (1.30, 3.98), 1.48 (1.18, 1.86), 1.24 (1.08, 1.43) and 1.31 (1.16, 1.50), respectively]. The difference in the odds of preterm births in primigravid women in the two groups was no Table 2. Univariable and multivariable analysis for early pregnancy outcomes in women with and without endometriosis Early pregnancy outcomes Endometriosis n = 5375 (%) No endometriosis n = 8280 (%) Odds ratio (95% confidence interval) Adjusted odds ratio* (95% confidence interval) Miscarriage 662 (12.3) 450 (5.4) 2.44 (2.16, 2.77) < (1.44, 2.15) <0.001 Ectopic pregnancy 86 (1.6) 51 (0.6) 2.62 (1.85, 3.71) < (1.09, 6.72) 0.03 Termination of pregnancy 395 (7.3) 1072 (12.9) 0.53 (0.47, 0.60) < (0.74, 1.69) 0.59 *Adjusted for age, parity, socio-economic status and year of pregnancy. Table 3. Univariable and multivariable analysis for pregnancy outcomes (>24 weeks gestation) in women with and without endometriosis Pregnancy outcomes (>24 weeks) Endometriosis n = 4232 (%) No endometriosis n = 6707 (%) Odds ratio (95% confidence interval) Adjusted odds ratio* (95% confidence interval) Hypertensive disorders 350 (8.3) 452 (6.7) 1.25 (1.08, 1.44) (0.91, 1.24) 0.57 of pregnancy Placenta praevia 72 (1.7) 54 (0.8) 2.13 (1.50, 3.04) < (1.52, 3.31) <0.001 Placental abruption 18 (0.4) 27 (0.4) 1.05 (0.59, 1.91) (0.48, 1.74) 0.78 Unexplained APH 270 (6.4) 281 (4.2) 1.57 (1.33, 1.86) < (1.39, 2.00) <0.001 Postpartum haemorrhage 844 (19.9) 786 (11.7) 1.88 (1.69, 2.09) < (1.61, 1.46) <0.001 Mode of delivery Spontaneous 2108 (49.8) 4495 (67) Reference Reference Caesarean 1299 (30.7) 1281 (19.1) 2.16 (1.97, 2.38) < (1.26, 1.55) <0.001 Instrumental 822 (19.4) 928 (13.8) 1.89 (1.70, 2.10) < (1.08, 1.36) Preterm birth 321 (7.6) 388 (5.8) 1.33 (1.14, 1.55) < (1.07, 1.49) Low birthweight 285 (6.7) 409 (6.1) 1.11 (0.96, 1.30) (0.94, 1.32) 0.21 Stillbirth 20 (0.5) 28 (0.4) 1.13 (0.64, 2.01) (0.48, 1.66) 0.91 Neonatal death 3 (0.1%) 1 (0.01%) 4.75 (0.50, 45.75) (0.36, 41.36) 0.27 *Adjusted for age, parity, socio-economic status and year of pregnancy. 5

6 Saraswat et al. longer significant on multivariable analysis. The results of subset analysis are presented in Tables S3 and S4. Another subgroup analysis of women who had laparoscopy in the last decade of the study corroborated our earlier findings. There were 1809 women who had a laparoscopic diagnosis of endometriosis between 2000 and 2009 followed by a pregnancy until In the unexposed cohort, 3394 women had a pregnancy recorded between 2000 and On multivariable analysis, after adjusting for age and socio-economic status, women with endometriosis continued to have an increased risk of miscarriage and ectopic pregnancy [OR (95% CI), 1.59 (1.19, 2.10) and 5.78 (1.08, 30.90, respectively]. They were significantly more likely to have unexplained APH, PPH, caesarean delivery and preterm births compared with women without a previous diagnosis of endometriosis. The corresponding ORs (95% CI) for each of these outcomes were 1.73 (1.30, 2.31), 1.41 (1.20, 1.65), 1.24 (1.06, 1.25) and 1.37 (1.05, 1.79). The absolute numbers for certain rare outcomes such as ectopic pregnancy, placenta praevia, placental abruption, stillbirth and neonatal death were very low in the subgroup analysis, resulting in wide confidence intervals. Although the direction of effect remained the same, the association of endometriosis with placenta praevia found in the complete analysis and the subgroup analysis of nulliparous women was no longer noted to be significant. More details of the subgroup analysis are appended in Tables S5 and S6. Discussion Principal findings Pregnant women with endometriosis are at higher risk of miscarriage and ectopic pregnancy; placenta praevia, unexplained antepartum haemorrhage, postpartum haemorrhage, operative delivery and preterm birth. They also tend to be older and are more likely to be nulliparous than women without a similar diagnosis. Strengths and limitations The key strengths of the study lie in its analysis of data from a large population-based national cohort, a long duration of follow up, and a surgical confirmation of endometriosis in the exposed cohort. The high quality of routinely collected data, well maintained national datasets, and record linkage has enabled Scotland to build a track record in research using linked data sets. 17 However, our study is not without limitations. Missing data and limited availability of some of the covariates meant that their effect on pregnancy outcomes could not be explored. Smoking status had a large number of missing values in our data. Smoking is also known to be frequently under-reported by women in pregnancy 21 ; as such, we were unable to use information related to smoking in a meaningful way. Data were not routinely collected regarding ethnicity, body mass index (BMI) or exposure to fertility treatment, the latter is likely to be more common in women with endometriosis. 4,22 Both infertility and medically assisted reproduction are associated with adverse pregnancy outcomes even in singleton pregnancies, with a higher incidence of preterm birth, low birthweight and small-for-gestational-age babies Data on infertility and fertility treatment are not routinely collected by ISD and we therefore were unable to adjust for their effect on pregnancy outcomes. We appreciate that a small proportion of women would voluntarily choose to not conceive. Our data do not capture women s preferences and choices, their intention to conceive, attendance at fertility clinics or use of medically assisted reproduction. As laparoscopy is the gold standard for the diagnosis of endometriosis, there is a small risk of including some women with undiagnosed endometriosis in the unexposed group. We have minimised this risk by exclusion of women with a symptom-based diagnosis of endometriosis from both groups. We believe that the impact of any misclassification on the final results would be offset by the large sample size, such that the overall trends are not affected. Inclusion of undiagnosed women with endometriosis in the unexposed group would bias the association towards a null hypothesis indicating that the risk of adverse outcome in those with endometriosis is likely to be higher than reported in our study. The nature of the data did not allow us to explore the outcomes by severity (minimal, mild, moderate or severe) or site of endometriosis (peritoneal, ovarian, rectovaginal) or influence of any treatment to endometriosis on pregnancy outcomes. This study spans a long time period of 30 years. During this period, changes in diagnosis and treatment of endometriosis, obstetric care, laparoscopic equipment and techniques, and uptake of assisted reproduction could have influenced the outcomes evaluated. In our multivariable analysis, we have adjusted for the year of delivery to account for any confounding effect of the changes in practice with time. One of the important changes in practice over the past decade is the move to a see and treat policy for endometriosis as compared with a more diagnostic approach in 1990s. We therefore also conducted a sub-analysis of women who had a laparoscopic diagnosis of endometriosis in the last decade of our study ( ) to explore the effect of evolution in laparoscopic surgery and changes in practice. Our study is conducted on a Caucasian population and the findings therefore may not be generalisable to other population groups, although they agree with the limited data from a Chinese 15 and a Japanese 26 study. 6

7 Pregnancy outcomes in women with endometriosis Despite all these factors, the nature of our research question does not allow itself to be examined by clinical trials, and prospective cohort studies with large sample sizes are not feasible. Interpretation of findings Relatively few studies have used population-based data to explore pregnancy outcomes in women with endometriosis and only three 11,13,15 of them have had a surgical confirmation of endometriosis in the exposed cohort. The two large studies by Hjordt Hansen et al. 14 and Stephansson et al. 12 in Danish and Swedish populations, respectively, did not confirm the diagnosis of endometriosis surgically. In our data, 34% of women were excluded from the analysis in the endometriosis cohort owing to a lack of surgical diagnosis. Such misclassification of greater than one-third of the exposed population could potentially have a substantial impact on the results. The majority of our results for pregnancy outcomes are in agreement with the Swedish database study 12 and Chinese study, 15 which demonstrated a higher risk of preterm births, antepartum bleeding and caesarean section. The early pregnancy complications in our study are similar to the Danish population, 14 showing increased risk of both miscarriage and ectopic pregnancy. The incidence of miscarriage in our study was 12.4% in women with a previous diagnosis of endometriosis compared with 5.4% in those without a diagnosis of endometriosis. Both these rates are lower than the accepted 15% prevalence rate for miscarriage. The data for our study were obtained from hospital and maternity discharge databases. Only those women who had a contact with the hospital for their miscarriage are recorded in the database. Women who miscarried at home would not be captured by the database and therefore miscarriage rates in both groups are likely to be under-represented. It is difficult to ascertain whether women with a previous diagnosis of endometriosis are more anxious and therefore more likely to attend hospital with miscarriage-related symptoms compared with other pregnant women. Changes in awareness and health-seeking behaviour for miscarriage in pregnant women over the last three decades is also likely to influence the recorded prevalence of miscarriage. The existing data linking endometriosis with hypertensive disorders of pregnancy continues to be controversial. Of the six studies that explored this, the results of two large studies are contradictory. Hadfield et al. 11 did not find any association between endometriosis and either pregnancy-associated hypertension or pre-eclampsia in women with singleton pregnancies, irrespective of the use of medically assisted reproduction. In the same year (2009) another large Swedish database study 12 based on a cohort of 1.4 million women found a significantly increased risk of pre-eclampsia in women with endometriosis. In contrast to this, Brosens et al. 27 reported a reduced risk of pre-eclampsia in a survey of subfertile women with endometriosis. A small sample size and low response rate may have potentially introduced bias in their study. More recently, a Canadian study 13 demonstrated absence of any association between gestational hypertension/pre-eclampsia and endometriosis. Similarly, no such association was found in the Chinese population, 15 in keeping with the findings in our study. We are aware that a key determinant of pregnancy complications is the outcome of a previous pregnancy. Our subset analyses of primigravid women and those who had laparoscopic diagnosis of endometriosis in the last decade of the study demonstrated similar trends and direction of effect, confirming that endometriosis has an adverse impact on pregnancy independent of previous pregnancy outcomes or changes in practice. Analysis of certain rare outcomes such as ectopic pregnancy, placenta praevia and stillbirths was compromised by very small numbers in both the subsets, resulting in wide confidence intervals. Another interesting finding was that the odds of preterm birth were similar in primigravid women with and without endometriosis, thereby suggesting that in multipara, an unrecorded previous preterm birth could have been influential in a subsequent pregnancy. Conclusion Endometriosis predisposes women to an increased risk of early pregnancy loss as well as maternal and perinatal complications in ongoing pregnancies. Implications for clinical practice and research The findings from our study indicate that a diagnosis of endometriosis is associated with a poorer pregnancy outcome, but the reasons underlying this phenomenon need further elucidation. Brosens et al. 28 proposed that there are functional and structural abnormalities in both the eutopic endometrium and myometrium in women with endometriosis which may result in defective remodelling of spiral arterioles and placentation, leading to adverse pregnancy outcomes. What is currently unclear is why the risk for certain outcomes such as miscarriage, ectopic pregnancy, placenta praevia, unexplained antepartum haemorrhage and preterm birth is increased but other outcomes such as hypertensive disorders of pregnancy, placental abruption and low birthweight are not influenced by endometriosis. Further research is required to elucidate how endometriosis affects implantation, placentation, fetal growth and birth. There is a paucity of literature on the dose response relation of endometriosis severity with pregnancy outcomes in population-based cohorts. Given the protean nature of endometriosis and lack of correlation of 7

8 Saraswat et al. stage of endometriosis with severity of symptoms, it is difficult to ascertain whether more severe forms of disease or varying sites (peritoneal, ovarian and rectovaginal) of endometriosis have a differential influence on pregnancy. This is an area that warrants further research. A prospective cohort study would be an ideal choice to address most of the limitations as well as conduct subgroup analysis by severity or site of the disease, influence of surgical treatment and collect information on all confounding factors. Prospective studies of this size and duration have significant resource and time implications, which limits their feasibility. We have attempted to explore the impact of endometriosis on pregnancy outcomes a question that is important for both women and clinicians alike. Clinically, women with endometriosis are at risk of a wide range of adverse outcomes in pregnancy. Although these results can be used to counsel women, the limitations and low strength of association implies that a change in practice is not warranted for most of the outcomes. An increased awareness of the adverse impact of endometriosis should expedite prompt medical evaluation in women with symptoms of bleeding or pain during pregnancy to allow timely intervention and minimise complications for outcomes such as ectopic pregnancy, antepartum haemorrhage and preterm births. Acknowledgements We are grateful John Nolan from ISD for data extraction and Katie Wilde from Grampian Data Safe Haven for data management. Disclosure of interests None declared. Completed disclosure of interests form available to view online as supporting information. Contribution to authorship SB conceived the idea. SB, LS, SoB, KC and DA designed the study and interpreted the data. LS drafted the manuscript and performed statistical analysis with DA. LS, SB, KC, SoB, DA, DM and AH provided comments and contributed to the development of the final draft of the manuscript. Details of ethics approval Appropriate approvals from the Privacy Advisory Committee of ISD, of the National Health Service, Scotland; North of Scotland Research Ethics Committee (NRES) and NHS Research and Development were sought to conduct the study. NRES were of the opinion that formal ethical approval was not required. Funding The study was funded by a grant from Chief Scientist Office, Scotland. Supporting Information Additional Supporting Information may be found in the online version of this article: Table S1. Definition of pregnancy outcomes Table S2. Multivariable logistic regression outputs for pregnancy outcomes in women with and without endometriosis Table S3. Univariable and multivariable analysis for early pregnancy outcomes in primigravid women with and without endometriosis Table S4. Univariable and multivariable analysis for pregnancy outcomes (>24 weeks gestation) in primigravid women with and without endometriosis Table S5. Univariable and multivariable analysis for early pregnancy outcomes in women with and without endometriosis (date of laparoscopy between 2000 and 2009) Table S6. Univariable and multivariable analysis for pregnancy outcomes (>24 weeks gestation) in women with and without endometriosis (date of laparoscopy between 2000 and 2009) & References 1 Nnoaham KE, Hummelshoj L, Webster P, d Hooghe T, de Cicco Nardone F, de Cicco Nardone C, et al. Impact of endometriosis on quality of life and work productivity: a multicenter study across ten countries. Fertil Steril 2011;96: e8. 2 Laux-Biehlmann A, d Hooghe T, Zollner TM. Menstruation pulls the trigger for inflammation and pain in endometriosis. Trends Pharmacol Sci 2015;36: Eskenazi B, Warner ML. Epidemiology of endometriosis. Obstet Gynecol Clin North Am 1997;24: Ozkan S, Murk W, Arici A. Endometriosis and infertility: epidemiology and evidence-based treatments. Ann N Y Acad Sci 2008;1127: Dunselman GA, Vermeulen N, Becker C, Calhaz-Jorge C, D Hooghe T, De Bie B, et al. ESHRE guideline: management of women with endometriosis. Hum Reprod 2014;29: Harb HM, Gallos ID, Chu J, Harb M, Coomarasamy A. The effect of endometriosis on in vitro fertilisation outcome: a systematic review and meta-analysis. BJOG 2013;120: Barnhart K, Dunsmoor-Su R, Coutifaris C. Effect of endometriosis on in vitro fertilization. Fertil Steril 2002;77: Lebovic DI, Mueller MD, Taylor RN. Immunobiology of endometriosis. Fertil Steril 2001;75: Bedaiwy MA, Falcone T. Peritoneal fluid environment in endometriosis. Clinicopathological implications. Minerva Ginecol 2003;55: Giudice LC, Kao LC. Endometriosis. Lancet 2004;364: Hadfield RM, Lain SJ, Raynes-Greenow CH, Morris JM, Roberts CL. Is there an association between endometriosis and the risk of preeclampsia? A population based study Human Reprod 2009;24: Stephansson O, Kieler H, Granath F, Falconer H. Endometriosis, assisted reproduction technology, and risk of adverse pregnancy outcome. Human Reprod 2009;24:

9 Pregnancy outcomes in women with endometriosis 13 Aris A. A 12-year cohort study on adverse pregnancy outcomes in Eastern Townships of Canada: impact of endometriosis. Gynecol Endocrinol 2014;30: Hjordt Hansen MV, Dalsgaard T, Hartwell D, Skovlund CW, Lidegaard O. Reproductive prognosis in endometriosis. A national cohort study. Acta Obstet Gynecol Scand 2014;93: Lin H, Leng JH, Liu JT, Lang JH. Obstetric outcomes in Chinese women with endometriosis: a retrospective cohort study. Chin Med J (Engl) 2015;128: Kendrick SW, Douglas MM, Gardner D, Hucker D. Best-link matching of Scottish health data sets. Methods Inf Med 1998;37: Fleming M, Kirby B, Penny KI. Record linkage in Scotland and its applications to health research. J Clin Nurs 2012;21: Norman JE, Morris C, Chalmers J. The effect of changing patterns of obstetric care in Scotland ( ) on rates of preterm birth and its neonatal consequences: perinatal database study. PLoS Med 2009;6:e Love ER, Bhattacharya S, Smith NC, Bhattacharya S. Effect of interpregnancy interval on outcomes of pregnancy after miscarriage: retrospective analysis of hospital episode statistics in Scotland. BMJ 2010;341:c Carstairs V, Morris R. Deprivation: explaining differences in mortality between Scotland and England and Wales. BMJ 1989;299: Spencer K, Cowans NJ. Accuracy of self-reported smoking status in first trimester aneuploidy screening. Prenat Diagn 2013;33: Buyalos RP, Agarwal SK. Endometriosis-associated infertility. Curr Opin Obstet Gynecol 2000;12: Helmerhorst FM, Perquin DA, Donker D, Keirse MJ. Perinatal outcome of singletons and twins after assisted conception: a systematic review of controlled studies. BMJ 2004;328: Messerlian C, Maclagan L, Basso O. Infertility and the risk of adverse pregnancy outcomes: a systematic review and meta-analysis. Hum Reprod 2013;28: Pandey S, Shetty A, Hamilton M, Bhattacharya S, Maheshwari A. Obstetric and perinatal outcomes in singleton pregnancies resulting from IVF/ICSI: a systematic review and meta-analysis. Hum Reprod Update 2012;18: Takemura Y, Osuga Y, Fujimoto A, Oi N, Tsutsumi R, Koizumi M, et al. Increased risk of placenta previa is associated with endometriosis and tubal factor infertility in assisted reproductive technology pregnancy. Gynecol Endocrinol 2013;29: Brosens IA, De Sutter P, Hamerlynck T, Imeraj L, Yao Z, Cloke B, et al. Endometriosis is associated with a decreased risk of preeclampsia. Human Reprod 2007;22: Brosens I, Pijnenborg R, Benagiano G. Defective myometrial spiral artery remodelling as a cause of major obstetrical syndromes in endometriosis and adenomyosis. Placenta 2013;34: