Comparing ART outcomes in women with endometriosis after GnRH agonist versus GnRH antagonist ovarian stimulation: a systematic review

Background: Endometriosis is an oestrogen-dependent disease that can cause subfertility in women who may require assisted reproductive technology (ART) to achieve their pregnancy goals. Objectives: The aim of this study was to compare ART outcomes in women with endometriosis following the long GnRH-agonist controlled ovarian stimulation (COS) protocol with those taking the GnRH-antagonist COS protocol. Data Sources and Methods: MEDLINE, Embase and Web of Science were systematically searched in June 2022. Randomized controlled trials (RCTs) and observational studies comparing the long GnRH-agonist COS protocol and the GnRH-antagonist COS protocol in women with all stages/subtypes of endometriosis were included. Data were synthesized into comprehensive tables for systematic review. The Scottish Intercollegiate Guidelines Network (SIGN) checklists were used for the risk of bias assessment of non-randomized studies and randomized studies, and all the included studies were deemed to have acceptable quality. Main Results: Eight studies (one RCT and seven observational) with 2695 patients (2761 cycles) were included. Most studies generally reported non-significant differences in clinical pregnancy or live birth rates regardless of the COS protocol used. However, the GnRH-agonist protocol may yield a higher total number of oocytes retrieved, especially mature oocytes. Conversely, the GnRH-antagonist protocol required a shorter COS duration and lower gonadotrophin dose. Adverse outcomes, such as rates of cycle cancellation and miscarriage, were similar between both COS protocols. Conclusion: Both the long GnRH-agonist and GnRH-antagonist COS protocols generally yield similar pregnancy outcomes. However, the long GnRH-agonist protocol may be associated with a higher cumulative pregnancy rate due to the higher number of retrieved oocytes available for cryopreservation. The underlying mechanisms of the two COS protocols on the female reproductive tract remain unclear. Clinicians should consider treatment costs, stage/subtype of endometriosis and pregnancy goals of their patients when selecting a GnRH analogue for COS. A well-powered RCT is needed to minimize the risk of bias and compare the risk for ovarian hyperstimulation syndrome. Registration: This review was prospectively registered at PROSPERO under Registration No. CRD42022327604.


Introduction
Endometriosis is an inflammatory oestrogendependent disease characterized by endometrial-like tissue found outside of the uterus. Endometriosis lesions are often located in the peritoneum, ovaries (endometrioma) and uterus, but lesions can also be found in the bowel, urinary tract and vagina. Endometriosis is associated with a wide range of symptoms including visceral syndrome (e.g. pelvic pain, painful urination, dyschezia), dysmenorrhoea and subfertility. Traditionally, endometriosis classification is based on the location of endometrial tissue lesions, and the three most prevalent types are ovarian endometriomas, superficial peritoneal endometriosis or deep endometriosis. 1,2 Endometriosis is commonly graded on the revised American Society for Reproductive Medicine (r-ASRM) classification scale. Depending on the extent of lesions, it is classified according to the four stages: minimal (stage I), mild (stage II), moderate (stage III) and severe (stage IV). 1,2 Endometriosis lesions can alter the pelvic anatomy, lead to excess inflammation and can negatively impact the reproductive cycle resulting in subfertility in 30-50% of affected women. 3,4 In women with endometriosis desiring to become pregnant, around 10-25% require assisted reproductive technology (ART), such as in vitro fertilization (IVF) and intracytoplasmic sperm injection (ICSI). 5 Since the 1980s, the long gonadotrophin-releasing hormone (GnRH) agonist protocol has been the gold standard for controlled ovarian stimulation (COS) to prevent a premature luteinizing hormone (LH) surge and improve ART outcomes. However, this protocol requires an extensive treatment period which is associated with more frequent side effects (such as hot flushes/flashes, bleeding, cyst development and headache) and has a higher risk of ovarian hyperstimulation syndrome (OHSS), which can be life-threatening. 6 The GnRH-antagonist protocol is a promising alternative with a reduced risk of OHSS, shorter treatment time and often requires a reduced gonadotrophin dose as a result of GnRH antagonists being able to rapidly inhibit GnRH receptors within hours of administration. 7 However, previous studies report poorer pregnancy outcomes in infertile couples after the GnRHantagonist protocol. 8,9 Compared to other causes of infertility, little research has focused on patients with endometriosis specifically, and it remains uncertain whether patients with endometriosis respond similarly to the long GnRH-agonist and GnRH-antagonist COS protocols. Furthermore, the fertilization rate is often overlooked, and it has recently been shown that fertilization rate positively correlates with cumulative live birth rate (LBR). 10 In this systematic review, we aim to compare ART outcomes following the long GnRH-agonist COS protocol with the GnRH-antagonist COS protocol specifically for women with endometriosis.

Patient populations
The patient populations consisted of women diagnosed with any form of endometriosis undergoing IVF/ICSI with ovarian stimulation using the long GnRH-agonist protocol compared to the GnRH-antagonist protocol.

Core outcome sets
The primary outcomes were related to pregnancy [clinical pregnancy rate (CPR) and LBR]. Secondary outcomes included the number of oocytes retrieved [total and metaphase II (MII)], fertilization rate, COS parameters (treatment duration and gonadotrophin dose) and adverse ART outcomes (miscarriage rate, cycle cancellation rate and OHSS).
Search strategy, eligibility criteria and study selection. A systematic search of the published literature up to 10 June 2022 was undertaken on MEDLINE, Embase and Web of Science databases using the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines. 11 The following keywords and medical subject headings (MeSH) were queried: endometriosis, endometrioma, infertility, GnRH agonist, GnRH antagonist, in vitro fertilization and ICSI (the full search strategy is detailed in Supplementary Table 1). Database search results were imported into EndNote (X9, Clarivate Analytics) prior to title and abstract screening. The PRISMA flowchart can be found in Figure 1.
Duplicate studies were removed and two authors (K.K.W.K. and S.O.) independently screened titles and abstracts, and excluded obviously irrelevant studies. Equivocal studies were independently screened by the third author (J.A.T.) until a consensus could be reached. Full manuscripts Data synthesis and bias assessment. Data extraction was completed by K.K.W.K. and S.O. All data from randomized controlled trials (RCTs) and observational studies (and their relevant subgroups) comparing the long GnRH-agonist protocol versus the GnRH-antagonist COS protocol for women with endometriosis were included (Supplementary Table 2). Data were synthesized into outcome tables.
The rigour of study methodology and risk of bias was critically appraised using the relevant Scottish Intercollegiate Guidelines Network (SIGN) criteria for cohort studies and RCTs (description of criteria are available on the SIGN website). 12 For cohort studies, this tool aims to assess the internal validity (selection of subjects, assessment of exposure and outcomes, confounding factors, statistical analysis) and overall study quality. Statements 1.3, 1.5, 1.6, 1.11 and 1.12 of the SIGN cohort study tool were excluded as all the studies were retrospective in nature and outcomes were objective in accordance with the SIGN's checklist notes. 12 The RCT tool assesses for a focused research question, patient randomization, blinding methods, interventions, attrition bias, analysis methods and overall study quality.

Study characteristics
Using a systematic searching approach, 744 titles were identified from database searches. After 59 duplicate titles were removed, 608 titles and abstracts were excluded. Ultimately, eight studies were included for the final analysis with a total of  Table 1). [13][14][15][16][17][18][19][20] In these studies, 1721 cycles used the long GnRHagonist protocol and 1040 cycles used the GnRHantagonist protocol. Six retrospective analyses, [13][14][15][16][17]20 one cross-sectional study 18 and one RCT 19 were identified. Each study was undertaken at a single centre, and all inclusion/exclusion criteria were available.
For the long GnRH-agonist protocol, four studies administered triptorelin, 14,15,18,19 two studies administered leuprorelin 16,17 and one study administered decapeptyl 13 daily starting from day 20 to 21 of the previous menstrual cycle. One study 20 did not specify which GnRH agonist was used and started treatment after day 21 of the preceding cycle. Seven studies in the GnRHantagonist arm administered subcutaneous cetrorelix or ganirelix. [13][14][15][16][17][18][19] Five of which followed a flexible multiple dosing protocol 13,[15][16][17]19 and two of which followed a fixed protocol from day 5 or day 6. 14,18 One study also gave patients in the GnRH-antagonist arm an oral contraceptive pill pretreatment taken for 14-24 days in the preceding cycle followed by a 3-to 5-day washout period. 16 One study did not specify the antagonist used and started the protocol after at least 6 weeks of oral contraceptives. 20

Study quality and risk of bias assessment
The completed SIGN assessments for observational studies and the RCT can be found in Tables 2 and 3, respectively. Since the study by Hosseini et al. was a cross-sectional study, a SIGN 12 checklist was not required (as described by SIGN's study design algorithm). As mentioned earlier, Statements 1.3, 1.5, 1.6, 1.11 and 1.12 were not applicable for retrospective studies. Statements 1.3, 1.8 and 1.9 did not apply since patients did not have the outcome before starting the intervention (1.4) and the primary outcomes of interest (pregnancy and LBRs) were objective and would not be affected by blinding (1.8 and 1.9). All studies had a clearly focused question, had representative patient characteristics and clearly defined outcomes. All studies had overall acceptable quality and were eligible for review. Two observational studies mentioned that the assignment of the GnRH-agonist protocol or the GnRH-antagonist protocol varied between clinicians. 15, 16 Kolanska et al. 20 were the only observational study to exclusively offer either the GnRH-agonist protocol or the antagonist protocol during specific timelines minimizing selection bias to either protocol. Although Rodriguez-Purata et al. 16 mentioned that poorer responders tended to use the antagonist protocol, a propensity score matching statistical method was used to compare CPRs. This method adjusts for covariates such as disease severity and comorbidities that may affect the probability of patients allocated to a certain treatment. As such, only patients with similar characteristics were compared for this outcome which helped mitigate selection bias. Two studies performed multivariate logistic regression to identify predictive factors affecting pregnancy or birth rates. 13,18 The inclusion of a small number of women with polycystic ovary syndrome, tubal infertility or adenomyosis alongside endometriosis also raised concerns for additional confounding factors. 14,20 Four of the studies only included women undergoing their first IVF/ ICSI cycle, which reduced the risk of confounders from women who require multiple IVF cycles due to poorer ART outcomes. 14,15,18,19 Since the primary outcomes of interest were objective, the studies were at lower risk of measurement bias. For the RCT, randomization methods were adequate, although there was a lack of blinding. An adequate sample size for pretest power estimation could not be calculated since there was a lack of studies comparing the long GnRH-agonist versus the GnRH-antagonist protocol prior to this RCT. 19 ART outcomes Clinical pregnancy rate. CPR was reported by all eight studies and was calculated by CPR per embryo transfer (ET) in three studies 13,17,18 or CPR per patient/cycle in four studies, [14][15][16]19 (see Table 4). Kolanska et al. 20 were the only study to report both CPR per cycle with ET and CPR per patient and analysed fresh/frozen ETs separately. Most studies found no significant difference in CPR 13-20 between the long GnRH-agonist and GnRH-antagonist protocols except for two subgroup analyses. 18,20 For advanced endometriosis, Hosseini et al. 18 reported a significantly higher pregnancy rate with the GnRH agonist when anti-Müllerian hormone (AMH) levels were between 1.1 and 2.7 ng/ml (p = 0.04). Kolanska et al. 20 found significantly higher CPR per started cycle with the GnRH agonist when analysing fresh ETs from women with all forms of endometriosis combined (p = 0.02) but no significant   15 Drakopoulos et al. 14 Kolanska et al. 20 Rodriguez-Purata et al. 16 Ruggiero et al. 17 Zhao et al. 13 Section 1: Internal validity  Live birth rate. LBR was included in three studies 13,14,20 (see Table 5). The LBR was calculated as either births per ET cycles 13,20 or births per started cycle regardless of the number of embryos transferred. 14,20 Two studies found no significant difference in LBR between protocols. 13,14 Kolanska et al. performed subgroup analyses by endometriosis subtype and fresh/freeze-thaw embryos and found no significant difference between protocols in LBR for patients with DE or endometriomas in isolation regardless of ET methods. However, the LBR per started cycle (regardless of whether embryos were transferred) was significantly higher (p = 0.02) in the long GnRH-agonist group. 20 Zhao et al. 13 were the only study to perform regression analysis and found maternal age to be the strongest predictive factor for women with diminished ovarian reserve (DOR) following ovarian cystectomy.  Number of oocytes retrieved. Six studies assessed the total number of oocytes retrieved 13,14,[16][17][18][19] (see Table 6). Two studies included women with resected endometrioma and found no difference in the number of oocytes retrieved between COS protocols. 13,19 However, in women with active endometriomas, Pabuccu et al. reported a higher number of oocytes (p = 0.002) retrieved using the GnRH-agonist protocol. The number of oocytes retrieved from patients with stage I-II endometriosis was reported by two studies, and both found no significant difference between protocols. 14,19 Three studies included women with stage III-IV endometriosis and two found no significant difference. 14,17 In a subgroup analysis of advanced endometriosis grouped by AMH levels, women with AMH levels between 1.1 and 2.7 ng/ ml did not differ in the number of oocytes retrieved between the two COS protocols. However, in women with AMH less than 1.1 ng/ml, the long GnRH-agonist protocol yielded more oocytes, while in women with AMH greater than 2.7 ng/ml, the GnRH-antagonist protocol led to an increased number of oocytes retrieved. 18 Rodriguez-Purata et al. 16 included all stages of endometriosis and found a significantly higher number of oocytes retrieved using the long GnRH-agonist protocol (p = 0.001). However, the propensity score was not applied to this outcome.   Four studies included the number of MII oocytes retrieved [15][16][17]19 (see Table 6). Pabuccu et al. were the only study to analyse patients with stage I-II endometriosis and found no difference between the two protocols. In severe stages of endometriosis, Ruggiero et al. 17 reported a significantly higher number of MII oocytes retrieved when the GnRH-antagonist protocol was used. Two studies included patients with active/resected endometrioma and found a significantly higher number of MII oocytes retrieved when GnRH-agonist COS was used (p = 0.0001-0.01). 15,19 Rodriguez-Purata et al. 16 did not apply the propensity score matching for this outcome but also found a significantly higher number of MII oocyte yield using the GnRH-agonist protocol.
Fertilization rate. Fertilization rate (FR) was reported by four studies 13,15,17,19 (see Table 7). Pabuccu et al. 19 were the only study to compare FR in women with stage I-II endometriosis and found no difference between COS protocols. Also, no significant difference was found in women with severe endometriosis. 17 Two observational studies of women with endometrioma resection 13,15 reported no significant difference in FR although the RCT found a significantly higher FR when the long GnRH-agonist was used in resected endometrioma (p = 0.001) but not in active endometrioma. 19 COS parameters COS duration. Among the seven studies that reported COS duration [13][14][15][16][17]19,20 (see Table 8), only one found a significant difference in the COS duration 14 with the agonist protocol having a longer duration compared to the antagonist protocol (p = 0.001). Drakopoulos et al. 14 found a significant difference between the GnRH-agonist and GnRH-antagonist duration in women with stage III-IV endometriosis (p < 0.001) but no difference in women with stage I-II endometriosis.
Gonadotrophin dose. In the majority of the papers reviewed, there were no significant differences in the total gonadotrophin dose (IU) required for COS treatment between the two protocols (see Table 8). Drakopoulos et al. 14 reported that both women with stage I-II and stage III-IV endometriosis required a greater gonadotrophin dose when using the long GnRH-agonist protocol (p < 0.001) as opposed to the GnRH-antagonist protocol. Ruggerio et al. 17 also found that the gonadotrophin dose between the two protocols was greater in the agonist arm (p = 0.05) when observing women with stage III-IV endometriosis. Whereas two studies that only included women with resected endometriomas found no difference in gonadotrophin dose between the protocols. 13,15  Pabuccu et al. 19 reported the amount of gonadotrophin used by the number of recombinant follicle-stimulating hormone (FSH) ampoules and no significant differences were observed in women with stage I-II endometriosis, resected endometriomas or active endometriomas.

Adverse ART cycle outcomes
The risk of developing OHSS was not explicitly reported as an outcome in any of the studies. The miscarriage rate was reported by three studies but there was no significant difference between the outcomes of the GnRH-agonist or antagonist protocols 17,19,20 (see Table 9). Pabuccu et al. observed no significant difference between the two protocols in the miscarriage rate in women with stage I-II endometriosis, resected endometrioma or active endometrioma. Interestingly, this study included the total number of cycle cancellations due to the risk of developing OHSS or insufficient ovarian response but did not specify how the cancellations were distributed between the two protocols nor were p-values specified. 19 Three papers measured the cycle cancellation rate in patients taking these two protocols and found that cycles were cancelled due to a variety of reasons, including insufficient ovarian response, risk of ovarian hyperstimulation, elevated progesterone levels and a low number of oocytes or embryos. 17,18,20 Of these three studies, all found that the cancellation rate did not differ significantly between the two protocols.

Main findings
Most studies found comparable clinical pregnancy and live birth rates between the long GnRH-agonist and GnRH-antagonist ovarian stimulation protocols. This is similar to that of women in the general IVF population and poor ovarian responders. [21][22][23] In addition, fertilization rates were similar although the long GnRHagonist protocol might be beneficial for some women with specific endometriosis subtypes and those with low ovarian reserve.
When comparing COS parameters, the GnRHagonist protocol generally required greater gonadotrophin dose and longer treatment duration although this did not always reach significance.
Adverse ART outcomes such as cycle cancellation rate and miscarriage rate were similar between the two protocols. The direct risk of developing OHSS could not be assessed because data regarding OHSS were not reported in these studies.
Interpreting pregnancy and LBRs. How pregnancy and LBRs are reported in studies is important to consider when discussing ART outcomes with patients. Since the number of embryos retrieved could be a predictive factor for pregnancy rate, 24 20 were the only study to analyse both fresh and freeze-thaw cycles and found that the long GnRH-agonist protocol led to a significantly higher pregnancy rate in patients with endometriosis regardless of subtype. The authors suggested that this difference may be explained by the action of GnRH antagonists on the endometrium rather than ovaries which is in line with previous studies. In 2006, Ruan et al. using an IVF mice model compared GnRH-agonist versus GnRH-antagonist COS protocols and found that the expression of two uterine receptivity biomarkers (integrin β3 and leukaemia-inhibitory factor) during the implantation window was significantly lower in the GnRH-antagonist group. This correlated with a significantly lower implantation rate. 26 A later case-control study in 2008 evaluated another receptivity marker, homeobox A10 (HOXA10) expression, from human endometrial biopsies and found decreased stromal and glandular cell HOXA10 expression in the GnRHantagonist group. 27 Although laboratory studies have found that the GnRH-antagonist protocol can reduce endometrial receptivity, in the clinic, pregnancy and birth rates in endometriosis patients are similar between both protocols 13-20 Table 9. Summary of adverse ART outcome data for GnRH-agonist and GnRH-antagonist COS protocols in women with endometriosis.