The effect of endometriosis on in vitro fertilisation outcome: a systematic review and meta-analysis

Authors

  • HM Harb,

    1. School of Clinical and Experimental Medicine, University of Birmingham, Birmingham Women's Hospital Foundation Trust, Birmingham, UK
    Search for more papers by this author
  • ID Gallos,

    1. School of Clinical and Experimental Medicine, University of Birmingham, Birmingham Women's Hospital Foundation Trust, Birmingham, UK
    Search for more papers by this author
  • J Chu,

    1. School of Clinical and Experimental Medicine, University of Birmingham, Birmingham Women's Hospital Foundation Trust, Birmingham, UK
    Search for more papers by this author
  • M Harb,

    1. Sandwell and West Birmingham Teaching Hospital, West Bromwich, UK
    Search for more papers by this author
  • A Coomarasamy

    Corresponding author
    1. School of Clinical and Experimental Medicine, University of Birmingham, Birmingham Women's Hospital Foundation Trust, Birmingham, UK
    • Correspondence: Prof A Coomarasamy, School of Clinical and Experimental Medicine, University of Birmingham, Academic Department, 3rd Floor, Birmingham Women's Hospital Foundation Trust, Metchley Park Road, Edgbaston, Birmingham B15 2TG, UK. Email a.coomarasamy@bham.ac.uk

    Search for more papers by this author

Abstract

Background

Endometriosis is found in 0.5–5% of fertile women and 25–40% of infertile women. It is known that endometriosis is associated with infertility, but there is uncertainty whether women with endometriosis have adverse pregnancy outcomes in in vitro fertilisation (IVF) treatment.

Objectives

To explore the association between endometriosis and IVF outcome.

Search strategy

Searches were conducted on MEDLINE, EMBASE, Cochrane Library and Web of Science (inception, December 2012) in all languages, together with reference lists of retrieved papers.

Selection criteria

Studies comparing IVF outcome in women with endometriosis with women without endometriosis. Patients were classified by stage of endometriosis. The outcomes were fertilisation, implantation, clinical pregnancy and live birth rates. Study selection was conducted independently by two reviewers. The Newcastle-Ottawa Quality Assessment Scale was used for quality assessment.

Data collection and analysis

Data extraction was conducted independently by two reviewers. Relative risks from individual studies were meta-analysed.

Main results

Twenty-seven observational studies were included, comprising 8984 women. Meta-analysis of these studies showed that fertilisation rates were reduced in stage I/II of endometriosis (relative risk [RR] = 0.93, 95% confidence interval [95% CI] 0.87–0.99, = 0.03). There was a decrease in the implantation rate (RR = 0.79, 95% CI 0.67–0.93, = 0.006) and clinical pregnancy rate (RR = 0.79, 95% CI 0.69–0.91, = 0.0008) in women with stage III/IV endometriosis undergoing IVF treatment.

Conclusion

The presence of severe endometriosis (stage III/IV) is associated with poor implantation and clinical pregnancy rates in women undergoing IVF treatment.

Introduction

It is estimated that around 0.5–5% of fertile women and 25–40% of infertile women have a diagnosis of endometriosis.[1, 2] Endometriosis is associated with infertility.[3-7] Women with endometriosis often require in vitro fertilisation (IVF); more than a third of women undergoing IVF have endometriosis.[8] There is evidence that the severity of endometriosis may have a direct impact on the success of IVF.[9, 10] However, the association between endometriosis and IVF outcomes remains unclear and is frequently questioned, especially for limited (stage I and II) disease.[11, 12]

'Several mechanisms have been postulated to explain the low fecundity observed with endometriosis, including altered folliculogenesis resulting in reduced quality oocytes,[13, 14] mechanical interference with oocyte pickup and transportation,[15] exposure to a hostile environment of macrophages, cytokines and vasoactive substances in the peritoneal fluid[16-18] and anatomical dysfunction of the fallopian tube and ovary.[19] The aim of our review was to systematically review and meta-analyse the effect of endometriosis on IVF outcomes.

Methods

Identification of literature and study selection

The population for this review was women undergoing IVF. The study cohort consisted of women with endometriosis and the comparison group was women without endometriosis. The outcomes were fertilisation, implantation, clinical pregnancy and live birth rates. The following electronic databases were searched: MEDLINE, EMBASE, Cochrane Central Register of Controlled Trials and Web of Science (inception, December 2012). A search strategy was carried out based on the following key words and/or medical subject heading (MeSH) terminology: endometriosis, endometrioma, pregnancy, live birth, implantation, fertilisation, oocyte, ovocyte, in vitro fertilisation, fertilisation-in vitro and assisted reproductive technologies. The reference lists of all known primary and review articles were examined to identify cited articles not captured by electronic searches. No language restrictions were placed in any of our searches or study selection.

We excluded articles that did not report the stage of endometriosis or which evaluated outcomes of other assisted reproductive techniques such as GIFT. We also excluded studies of women who had received medical or surgical treatment of their endometriosis before the IVF cycles. Studies were selected in a two-stage process. First, the titles and abstracts from the electronic searches were scrutinised by two reviewers independently (HMH and JC) and full manuscripts of all citations that were likely to meet the predefined selection criteria were obtained. Second, final inclusion or exclusion decisions were made on examination of the full manuscripts. In cases of duplicate publication, the most recent and complete versions were selected. Any disagreements about inclusion were resolved by consensus or arbitration by a third reviewer. Two reviewers (HMH and IDG) completed the quality assessment. The Newcastle-Ottawa Quality Assessment Scales for observational studies were implemented for quality assessment. The quality checklist we used awards one star as maximum for all items except comparability where it can award a maximum of two stars. We used an arbitrary score based on the assumption of equal weight of all items included in the Newcastle-Ottawa Scale. This was used to give a quantitative appraisal of overall quality of the individual studies. The score ranged from 0 to 9, with a score of either 0 or 1 for each item. From each study, outcome data were extracted in 2 × 2 tables by the two reviewers HH and JC.

Statistical analysis

For the analysis of clinical pregnancy and live birth rates we analysed data per total number of women undergoing IVF. For the implantation rate we report successful implantation per total number of embryos transferred and the fertilisation rate per number of oocytes. Relative risks from individual studies were meta-analysed using a fixed effects model or random effects model for analyses with significant heterogeneity (< 0.05).[20] Heterogeneity was evaluated graphically using forest plots and statistically using the I2 statistic to quantify heterogeneity across studies. We performed the Harbord's test to examine funnel-plot asymmetry as this is recommended for binary data.[21] Statistical analyses were performed using RevMan 5.0 (Cochrane Collaboration, Oxford, UK) and Stata 9.0 (Stata Corp, TX, USA).

Results

The search strategy yielded 911 citations (Figure 1) of which 821 publications were excluded because it was clear from the title or abstract that they did not fulfil the selection criteria. We obtained full manuscripts for the remaining 90 articles. Sixty-three publications were excluded because nine did not have a control group without endometriosis, 11 did not report original data, in 16 studies women underwent medical or surgical interventions to treat endometriosis, in 11 studies there were no extractable data, eight studies did not specify the stage of endometriosis and eight studies were duplicates or the same data were used in other included studies. Therefore the total number of studies included in the review was 27.

Figure 1.

Study selection for systematic review on the effect of endometriosis on IVF treatment outcome.

All of the 27 included studies were observational studies whereby the target population (women having IVF treatment), with or without endometriosis were followed up to the outcomes. The characteristics of the 27 studies and the Newcastle-Ottawa Quality Assessment are presented in Tables 1 and 2. The studies scored well on the Newcastle-Ottawa Quality Assessment Scale (Table 2). We performed a stratified analysis splitting studies from 1985 to 1999 and from 2000 to 2012 and found there were no differences between subgroups for the outcome of clinical pregnancy (= 0.853). We have performed the Harbord's test to examine funnel-plot asymmetry as this is recommended for binary data. For the outcome of clinical pregnancy we found no significant P-value for small study effects and plot asymmetry.

Table 1. Characteristics of studies of endometriosis versus control on IVF treatment outcome
Study yearPopulationEndometriosis groupControl groupOutcomesStudy design
Arici 1996[27] (n = 120)Women undergoing IVF at Yale-New Haven Hospital between 1988 and 1994Women with endometriosis (n = 35) classified according to stage (Stage I/II: n = 18, Stage III/IV: n = 17) diagnosed on laparoscopy and/or biopsy of lesions (89 cycles)Women with tubal damage (= 70) or unexplained infertility (n = 15) (195 cycles)Live births, Clinical pregnancies, ImplantationsRetrospective cohort study
Azem 1999[41] (n = 118)Women undergoing IVF at Lis Maternity Hospital, Tel Aviv between 1986 and 1995Women with endometriosis (n = 58) classified as stage III/IV, 12 with endometriomas >3 cm diagnosed sonographically and at time of oocyte retrieval, diagnosed on laparoscopy and biopsy of endometriotic implants (207 cycles)Women with tubal infertility (n = 60) diagnosed on laparoscopy (197 cycles)Live birthsRetrospective cohort study
Bukulmez 2000[42] (n = 834)Women undergoing ICSI at Haceteppe University, Turkey between 1996 and 1999Women with endometriosis (n = 44) classified according to stage (Stage I/II: n = 25, Stage III/IV: n = 19) diagnosed on laparoscopy or laparotomy with biopsy proven endometriosis (78 cycles)Women with no history of endometriosis as documented by laparoscopy or laparotomy (n = 771) (1170 cycles)Clinical pregnancies, ImplantationsRetrospective cohort study
Cahill 1996[13] (n = 35)Women undergoing IVF at St Michael's Hospital, BristolWomen with endometriosis (n = 10) classified as stage I, diagnosed on laparoscopy, who have received no previous treatment (25 cycles)Women with unexplained infertility (n = 9) or with tubal infertility as a result of previous infection (n = 16) (73 cycles)Clinical pregnanciesProspective cohort study
Curtis 1993[43] (n = 331)Women undergoing IVF at the Royal Free Hospital, London over a 3-year periodWomen with endometriosis (n = 53) classified according to stage (Stage I/II: n = 25, Stage III/IV: n = 28) diagnosed on laparoscopy (92 cycles)Women with tubal disease (n = 153) (228 cycles)Live births, Clinical pregnanciesProspective cohort study
Diaz 2000[44] (n = 58)Women undergoing IVF in the Oocyte Donation Program at the Instituto Valenciano de Infertilidad, Spain. Donors were 10 women undergoing IVF for male infertility (n = 3), tubal infertility (n = 6), and unexplained infertility (n = 1); and 15 young fertile women who voluntarily donated oocytesWomen with endometriosis (n = 25) classified as stage III/IV diagnosed on laparoscopy.Women with no history of endometriosis as confirmed on laparoscopy. The indications for oocyte donation were failed IVF (n = 8), low response to conventional IVF ovarian stimulation (n = 10),and ovarian failure (n = 15).Live births, Clinical pregnancies, ImplantationsProspective matched case–control study
Dlugi 1989[45] (n = 15)Women undergoing IVFWomen with endometriosis (n = 7) classified as Stage III/IV with pelvic adhesions and endometriomas, diagnosed on laparoscopy within 6 months prior to IVF (12 cycles)Women with pelvic adhesions and hydrosalpinges (n = 8), with no evidence of endometriosis as confirmed on laparoscopy (27 cycles)Clinical pregnanciesProspective cohort study
Dmowski 1995[17] (n = 193)Women undergoing IVF at Oakbrook Fertility Centre, Chicago between 1991 and 1993Women with endometriosis (n = 84) classified according to stage (Stage I/II: n = 59, Stage III/IV n = 25) subdivided into groups with active or inactive disease (119 cycles)Women with tubal disease (39%), pelvic adhesions (28%), male factor (25%), unexplained infertility (13%), ovarian dysfunction (17%) and other factors (5%) (n = 109) (118 cycles)Clinical pregnanciesRetrospective cohort study
Genovese 2011[46] (n = 60)Women undergoing IVF at S. Bambino Hospital in Catania between 2006 and 2009Women with endometriosis (n = 30) classified according to stage (Stage I/II: n = 15, Stage III/IV n = 15) diagnosed on laparoscopy.‘Healthy’ women (n = 15) undergoing IVF due to male factor infertilityClinical pregnancies, Live birthsRetrospective cohort study
Hickman 2002[47] (n = 141)Women undergoing IVF at Wilford Hall Medical Centre, Texas between 1996 and 1999Women with endometriosis (n = 27) classified according to stage (Stage I/II: n = 18, Stage III/IV: n = 9) diagnosed at laparoscopy or laparotomy. (Stage I/II: 20 cycles, Stage III/IV: 11 cycles)Women with tubal disease and/or occlusion from either previous infection or prior tubal surgery, including tubal ligation (n = 104) diagnosed on laparoscopy or laparotomy. (118 cycles)Clinical pregnancies, ImplantationsProspective cohort study
Hull 1998[28] (n = 1030)Women undergoing IVF at the University of Bristol, Centre for Reproductive Medicine between 1990 and 1996Women with endometriosis (n = 194) classified as stage I/II, diagnosed on laparoscopy within 1 year of IVF.Women with unexplained infertility (n = 327) or with tubal infertility in whom hydrosalpinx was excluded (n = 509)Clinical pregnanciesRetrospective cohort study
Kumbak 2008[48] (n = 168)Women undergoing IVF at Istanbul Memorial Hospital, Turkey between 2003 and 2005Women with unilateral ovarian endometrioma of 10–50 mm (n = 85) diagnosed by direct aspiration during oocyte retrieval (34 women had previous ovarian surgery)Women with basal simple ovarian cysts of 10–35 mm (n = 83), detected on ultrasound. Male factor was the only apparent cause of infertility in the control groupClinical pregnanciesRetrospective cohort study
Kuroda 2009[3] (n = 46)Women undergoing IVF at Juntendo University School of Medicine, Japan between 2006 and 2008Women with endometriosis (n = 25), with no endometrioma (n = 7) diagnosed on laparoscopy or unoperated endometrioma (n = 18) diagnosed on ultrasound or MRI (34 cycles)Women with tubal infertility (n = 21)Pregnancy rateRetrospective cohort study
Mahmood 1991[4] (n = 30)Women undergoing IVF at Aberdeen Maternity Hospital, ScotlandWomen with endometriosis (n = 20) classified as stage I, diagnosed on laparoscopyWomen with pelvic adhesions causing bilateral tubal occlusion (n = 10) diagnosed on laparoscopyClinical pregnancies, FertilisationsProspective cohort study
Matson 1986[29] (n = 124)Women undergoing IVF at King Edward Memorial Hospital, Western AustraliaWomen with endometriosis (n = 96) classified according to stage (Stage I/II: n = 40, Stage III/IV: n = 56) with endometriosis (154 cycles)Women with tubal damage (n = 28) (40 cycles)Clinical pregnancies, FertilisationsProspective cohort study
Meden-Vrtovec 2000[5]Women undergoing IVF at the University Medical Centre, Sloveniabetween 1990 and 1999Women with endometriosis classified as stage I/II diagnosed on laparoscopy (612 cycles)Women with tubal infertility with damaged or removed fallopian tubes (7339 cycles)FertilisationsRetrospective cohort study
Nejad 2009[49] (n = 137)Women undergoing IVF at Royan Institute, Iran between 2005 and 2006Women with endometriosis (n = 80) classified according to stage (stage I/II: n = 32, stage III/IV: n = 48) diagnosed on laparoscopy. No medical treatment for endometriosis received in at least 1 yearWomen with tubal infertility (n = 57) diagnosed on laparoscopyImplantationsRetrospective cohort study
Olivennes 1995[50] (n = 214)Women undergoing IVF at the New York Hospital Cornell University Medical Centre between 1988 and 1992Women with endometriosis (n = 147) classified according to stage (Stage I/II: n = 81, Stage III/IV: n = 66) diagnosed on laparoscopy. All did not receive medical or surgical therapy in the 18 months preceding their IVF cycle (236 cycles)Women with tubal infertility (n = 111) diagnosed on laparoscopy (160 cycles)Live births, Clinical pregnanciesProspective cohort study
Omland 2001[51] (n = 87)Women undergoing IVF at the National Hospital, Norway between 1992 and 2001Women with endometriosis (n = 30) classified as stage I, diagnosed on laparoscopy (77 cycles)Women with unexplained infertility (n = 33) or tubal infertility (n = 24) diagnosed on laparoscopy and/or laparotomy (146 cycles)FertilisationsProspective cohort study
Omland 2006[7] (n = 91)Women undergoing IVF/ICSI at Rikshospitalet-Radiumhospitalet Medical Center, Oslo between 1997 and 2006Women with endometriosis (n = 43) classified as stage I on laparoscopy.Women with unexplained infertility (n = 48) with normal pelvic anatomy on laparoscopyFertilisations, Implantations, Clinical pregnanciesRetrospective cohort study
Pellicer 1998[18] (n = 50)Women undergoing IVF at the Instituto Valenciano de Infertilidad, SpainWomen with endometriosis (n = 24) classified according to stage (Stage I/II: n = 5, Stage III/IV: n = 19) diagnosed at laparoscopy or laparotomy less than one year before IVF.Women with tubal infertility (n = 26) diagnosed on laparoscopy.ImplantationsProspective cohort study
Simon 1994[52] (n = 155)Women undergoing IVF at the Instituto Valenciano de Infertilidad, Spain between 1990 and 1993Women with endometriosis (n = 59) classified according to stage (Stage I/II: n = 9, Stage III/IV: n = 50, 22 of which had an ovarian endometrioma >3 cm sonographically diagnosed and confirmed during oocyte retrieval) diagnosed on laparoscopy (96 cycles)Women with tubal infertility (n = 78). To rule out time variations, only women who underwent IVF simultaneously with the women with endometriosis were included (96 cycles)ImplantationsRetrospective cohort study
Sung 1997[53] (n = 239)Women undergoing IVF at the Mount Sinai Medical Centre, New York between 1991 and 1995Women with endometriosis (n = 55) classified according to stage (Stage I/II: n = 18, Stage III/IV: n = 37) diagnosed on laparoscopy within 2.5 years of IVF treatment (55 cycles)Women with no evidence of endometriosis identified by transvaginal ultrasound (to rule out endometriomas) and/or laparoscopy (n = 184) (184 cycles)Clinical pregnanciesRetrospective cohort study
Wardle 1985[54] (n = 104)Women undergoing IVF at Bristol Maternity Hospital, BristolWomen with endometriosis (n = 17) classified as stage I–III without structural damage, diagnosed on laparoscopy (17 cycles)Women with tubal infertility defined by laparoscopic evidence of bilateral tubal occlusion (n = 47) or failure of sperm or mucus penetration on PCT (n = 19) or unexplained infertility (n = 21) (87 cycles)Clinical pregnanciesProspective cohort study
Lin 2012[57] (n = 4444)Women undergoing IVF in the Assisted Reproductive Unit of Sir Run Run Shaw Hospital, China between 2006 and 2010Women with endometriosis (n = 177) classified according to stage (Stage I/II n = 64, Stage III/IV n = 113) diagnosed on laparoscopy or laparotomyWomen with any factors other than endometriosis (n = 4267) including tubal factor, male factor, or complex factors.Fertilisation, Implantations, Clinical PregnanciesRetrospective cohort study
Yanushpolsky 1998[55] (n = 102)Women undergoing IVF between 1994 and 1995Women with endometriosis (n = 45) classified as Stage III with endometriomas >1 cm diameter diagnosed on transvaginal ultrasound scan at the time of oocyte retrieval.Controls defined as women without any complex cyst at the time of oocyte retrieval (n = 57)Live birthsProspective cohort study
Yovich 1988[56] (n = 58)Women undergoing IVF in the PIVET Medical Centre, Western AustraliaWomen with endometriosis (n = 20) classified as Stage IV diseaseWomen with irreversible tubal disease (n = 28)Clinical pregnanciesProspective cohort study
Table 2. Appraisal of methodological quality (Newcastle-Ottawa Scale) of included studies
StudyCase–cohort representativeSelection of non-exposed controlAscertainment of exposureOutcome negative at startComparability by design§Comparability by analysisOutcome assessmentDuration of follow-upScore
  1. * Indicates that a feature is present; x, that a feature is absent. But for comparability by design this checklist awards a maximum of two stars (**), one (*) or none if the feature is completely absent (x).

Arici et al.[27]*********9
Azem et al.[41]********8
Bukulmez et al.[42]********8
Cahill et al.[13]****x***7
Curtis et al.[43]*********9
Diaz et al.[44]*********9
Dlugi et al.[45]********8
Dmowski et al.[17]*********9
Genovese et al.[46]********8
Hickman et al.[47]*********9
Hull et al.[28]xx*******7
Kumbak et al.[48]*********9
Kuroda et al.[3]********8
Mahmood et al.[4]********8
Matson et al.[29]****x***7
Meden-Vrtovec et al.[5]********x8
Nejad et al.[49]********8
Olivennes et al.[50]xx*******7
Omland et al.[51]*********9
Omland et al.[7]*********9
Pellicer et al.[18]********8
Simon et al.[52]*********9
Sung et al.[53]****x***7
Wardle et al.[54]*********9
Lin et al.[57]********8
Yanushpolsky et al.[55]*********9
Yovich et al.[56]********8

Fertilisation rate

Pooling of results from seven studies that reported fertilisation rate as an outcome for stage I/II endometriosis (Figure 2) found a 7% reduction in fertilisation rate (relative risk [RR] = 0.93, 95% confidence interval [95% CI] 0.87–0.99, = 0.03). There was significant variation across studies as indicated by an I2 value of 66% (= 0.008). Pooling of results from three studies that reported fertilisation rate as an outcome for stage III/IV endometriosis did not show a difference in fertilisation rate (RR = 1.01, 95% CI 0.93–1.10, = 0.84). There was significant heterogeneity between the studies, as indicated by an I2 value of 70% (= 0.03).

Figure 2.

Forest plot of studies of endometriosis versus control in women undergoing IVF treatment for the outcome of fertilisation rate.

Implantations

Pooling of results from eight studies that reported implantation rate as an outcome for stage I/II endometriosis (Figure 3) did not show a difference in implantation rate compared with controls (RR = 0.83, 95% CI 0.68–1.01, = 0.07). The I2 value was 58% indicating moderate variation among the studies (P = 0.02). Pooling of results from eight studies that reported implantation rate as an outcome for stage III/IV endometriosis found a 21% relative reduction in implantation rate (RR = 0.79, 95% CI 0.67–0.93, = 0.006). There was limited heterogeneity among the studies (I2 = 20%, = 0.27).

Figure 3.

Forest plot of studies of endometriosis versus control in women undergoing IVF treatment for the outcome of implantation rate.

Clinical pregnancies

Pooling of results from 14 studies that reported clinical pregnancy as an outcome for stage I/II endometriosis (Figure 4) did not show a difference in clinical pregnancies (RR = 0.94, 95% CI 0.83–1.07, = 0.35). There was little variation across studies as indicated by an I2 value of 10% (= 0.34). Pooling of results from 14 studies that reported clinical pregnancy as an outcome for stage III/IV endometriosis showed a relative risk reduction of 21% in clinical pregnancies (RR = 0.79, 95% CI 0.69–0.91, = 0.0008). There was limited heterogeneity among the studies, as indicated by an I2 value of 28% (= 0.15).

Figure 4.

Forest plot of studies of endometriosis versus control in women undergoing IVF treatment for the outcome of clinical pregnancy.

Live births

Pooling of results from six studies that reported live birth as an outcome for stage I/II endometriosis (Figure 5) did not show a difference in live birth rate compared with controls (RR = 0.92, 95% CI 0.83–1.02, = 0.10). The I2 value was 19%, indicating little variation among the studies (P = 0.29). Pooling of results from nine studies that reported live birth as an outcome for stage III/IV endometriosis did not show a statistically significant difference in this outcome (RR = 0.86, 95% CI 0.68–1.08, = 0.19), although the direction of effect suggested a reduction in live births with severe endometriosis. The I2 value was 0%, indicating that heterogeneity was unlikely among the studies (P = 0.52).

Figure 5.

Forest plot of studies of endometriosis versus control in women undergoing IVF treatment for the outcome of live birth rate.

Discussion

Main findings

This systematic review, which included 27 studies, found that the presence of severe endometriosis (stages III/IV) reduces implantation and clinical pregnancy rates. There was also a trend towards reduction of live births in the presence of severe endometriosis, which did not reach statistical significance. Our study demonstrated that the presence of mild endometriosis (stage I/II) does not seem to affect implantation, clinical pregnancy or live birth with rates comparable to those of women with other causes of infertility, contrary to the findings reported in the meta-analysis by Barnhart et al.[12]

Strengths and weaknesses

There are several factors that strengthen our analysis. We performed an extensive search strategy and used valid data synthesis methods. We used the Newcastle-Ottawa Quality Assessment Scale to rate the quality of the included studies and the included studies scored well on this scale, suggesting low risk of bias. Furthermore, no language restrictions were applied. Although we aimed to include studies that are comparable in patient characteristics and methods, there will always be some clinical heterogeneity between studies included in a meta-analysis. However, a degree of clinical heterogeneity is to be expected, as in almost all meta-analyses, and is not necessarily a drawback. It can increase the generalisability of a meta-analysis over that of single studies. In this study, the stage of endometriosis, which represents a great range of disease, is likely to have contributed significantly to the variability of our findings.[25] However, there were also other sources of clinical heterogeneity, such as between studies reporting IVF outcome after a number of cycles and others reporting outcomes after the first IVF cycle. Studies also differed in the diagnostic methods used to confirm the presence of endometriosis, with most studies using diagnostic laparoscopy, whereas some reported diagnosis based on ultrasonographic findings of ovarian endometrioma during oocyte retrieval. Finally, we cannot rule out the possibility that there may have been a degree of unreported concurrent treatment in some studies. Our intention was to isolate as clean a cohort as possible; however we accept a degree of uncertainty, and a potential impact of unreported surgical procedures on the estimate of the effect of advanced endometriosis. The recent ASRM guidelines suggest that surgery for stage III/IV endometriosis can be useful to treat pelvic adhesions that may impact reproductive function.[58] Furthermore, it has been demonstrated that in 216 infertile women with severe endometriosis, the cumulative pregnancy rates after laparoscopy or laparotomy were 45% and 63% respectively.[59] Moreover, laparoscopic cystectomy for ovarian endometriosis >4 cm was found to improve fertility compared with cyst drainage and coagulation.[60] It is therefore possible that IVF outcomes, such as live birth rate, for women with untreated endometriosis may be under-represented in our analysis.

Interpretation

Barnhart et al.[12] published a meta-analysis on the effect of endometriosis on IVF over 10 years ago. Their meta-analysis did not distinguish between women who had received previous medical and surgical interventions, potentially weakening the associations that they identified. Furthermore, their meta-analysis included publications from 1983 to 1998. Techniques and outcomes of IVF have dramatically changed over recent years, therefore limiting the applicability of their findings.

Our review demonstrated that the implantation and clinical pregnancy rate are reduced in women with severe endometriosis. There are likely to be multiple causal relationships between endometriosis and clinical pregnancy as an outcome. A reduction in implantation rate appears to partially explain this relationship. Successful implantation requires complex interactions between a receptive endometrium and a synchronously developed embryo at the blastocyst stage.[26] Defective implantation capacity of the embryos, especially in severe endometriosis, has been reported in previous studies.[5, 17, 27-29] Moreover, eutopic endometrium from women with endometriosis has been shown to have significant biochemical and ultrastructural differences from normal control tissue,[30] particularly in the receptive cycle[31, 32] which may be associated with alterations in gene expression.[33, 34] Aberrant integrin expression[22] including integrin ανβ3,[35] as well as other relevant molecular markers of endometrial receptivity, such as HOXA10 and HOXA11, were found to be altered in luteal-phase endometrium of women with endometriosis.[36] Other studies have linked abnormal endometrial pinopod formation,[37] an embryotoxic milieu[38] and progesterone resistance in women with endometriosis[39, 40] to defective implantation in endometriosis. These factors would potentially result in dysfunctional embryo attachment and defective signalling, contributing to the diminished fertility associated with this disease.[31, 32]

Interestingly, our review found a reduction in fertilisation rates in stage I/II endometriosis compared with controls but not in stage III/IV disease and this finding is consistent with Barnhart et al.'s report.[12] This finding has also biological plausibility and may be explained by an increased chemotactic activity due to active endometriotic lesions in stage I/II disease, resulting in poorer oocyte quality and so affecting fertilisation.[22-24] However, the effect is small and the association is weak. As a result, we would advise that this finding is interpreted with caution.

The most important clinical outcome is live birth rate, and although a reduction of 14% in live births (RR = 0.86, 95% CI 0.68–1.08) was observed with stage III/IV endometriosis, this did not reach statistical significance. There were fewer reports for this outcome in the primary literature and the power of this review to detect this difference was therefore weakened.

A demonstration of reduction in IVF clinical pregnancies in women with stage III/IV endometriosis does not necessarily mean that treatment of endometriosis will restore the clinical pregnancy rates to the level expected in women without endometriosis. Therefore this evidence does not justify advocating medical or surgical treatment of endometriosis for these women, as a favourable risk benefit analysis of intervention in this clinical context is currently absent.

Disclosure of interests

The authors declared that there are no conflicts of interest.

Contribution to authorship

HMH developed the search strategy, identified studies for inclusion, performed data extraction, analysis and interpretation and drafted the manuscript. IDG contributed to data analysis and interpretation as well as drafting of the manuscript. JC contributed to identifying studies for inclusion, acquisition of the data and to revision of the manuscript. MH contributed to identifying studies for inclusion, data interpretation and revised the manuscript critically. AC contributed to data analysis, interpretation of the data and revised the manuscript critically.

Details of ethics approval

Not required.

Funding

No funding was received.

Acknowledgement

We thank Dr Kiren Shabir for her support with developing the search strategy.

Ancillary