Is a GnRH Antagonist Protocol Better in PCOS Patients? A Meta-Analysis of RCTs

Objective To review published randomized controlled trials (RCTs) evaluating the outcomes of in vitro fertilization/intra-cytoplasmic sperm injection (IVF/ICSI) utilization of gonadotropin-releasing hormone (GnRH) antagonists for ovarian stimulation in polycystic ovarian syndrome (PCOS) patients compared with classic luteal long agonist protocols. Design A meta-analysis of prospective randomized trials published in English between 2002 and 2013. Patient(s) and Interventions Nine RCTs examining PCOS patients undergoing IVF/ICSI including 588 women who underwent long agonist protocols and 554 women who underwent GnRH antagonist protocols. Main Outcome Measure(s) Clinical pregnancy rate (CPR), ongoing pregnancy rate (OPR) and ovarian hyperstimulation syndrome (OHSS) rate. Result(s) Nine RCTs were included in this analysis. The CPR-per-embryo transferred was similar in the two groups (relative risk (RR): 0.97, 95% confidence interval (CI): 0.85–1.10). Non-significant estimates comparing the two protocols were found for age, BMI, total dose of gonadotropin administered, number of days of stimulation and number of oocytes retrieved. After meta-analysis of 4 of the RCTs, it was concluded that a GnRH antagonist protocol is better than an agonist long protocol to reduce the rate of severe OHSS (odds ratio (OR): 1.56, 95% CI: 0.29–8.51). Conclusion(s) With respect to CPR, a GnRH antagonist protocol is similar to a GnRH agonist long protocol. However, for severe OHSS, a GnRH antagonist protocol is significantly better in PCOS patients.


Introduction
The first reports of gonadotropin-releasing hormone (GnRH) agonists for in vitro fertilization (IVF) were published in the 1980s. The function of GnRH agonists to suppress luteinizing hormone (LH) and prevent premature LH surges allowed optimal timing of human chorionic gonadotropin (hCG) administration and ovum collection, which improved IVF outcomes with respect to pregnancy rates [1]. Since that time, GnRH agonist long protocols have been the standard and mostly commonly used protocols.
Polycystic ovarian syndrome (PCOS) is the most common endocrinopathy of women of reproductive age and is characterized by oligomenorrhea, hyperandrogenism and a cystic appearance of the ovaries. Elevated LH levels are thought to be responsible for the high androgen levels, which adversely affect the development of follicles. Theoretically, blockade of endogenous LH secretion by antagonists combined with ovulation induction could result in improved follicular development. Women with PCOS undergoing IVF are at risk for a higher rate of ovarian hyperstimulation syndrome (OHSS). Triggering ovum final maturation with hCG is an important mechanism in OHSS. To overcome this barrier, GnRH antagonist protocols that use GnRH agonist triggering emerged. Since the publication of Hesham's [2] meta-analysis of 5 randomized controlled trials (RCTs), another series of RCTs [3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19] has been published.
These studies all used the OHSS rate as the primary outcome measurement, but there was no consensus on the classification system for OHSS between the different study sites, which limited the validity of meta-analysis. Theoretically, a GnRH antagonist protocol could reduce the OHSS rate. In addition, perhaps it is not necessary to compare OHSS rates as the primary outcome measurement. Thus, the clinical pregnancy rate (CPR) was used as the primary outcome measurement in this analysis, and the aim of this meta-analysis was to compare IVF outcomes for GnRH agonist long protocols and GnRH antagonist protocols in women with PCOS using the highest quality and most recent of the available data.

Criteria for considering studies for this review
Studies that compared long agonist protocols with GnRH antagonist protocols in PCOS patients undergoing IVF were considered for this review. The PCOS diagnosis had to fulfill the Rotterdam consensus criteria (Rotterdam ESHRE/ASRM-sponsored PCOS consensus workshop group, 2004). Information regarding patients and cycle characteristics such as age, number of oocytes retrieved and pregnancy outcomes was also required.

Search strategy to identify studies
Studies were identified by searching the electronic literature through PubMed for relevant reports published between 2002 and 2013. A search strategy was employed based on the following medical subject headings (MeSH): 'polycystic ovary syndrome' AND 'fertilization in vitro' OR 'reproductive medicine' OR 'reproductive techniques, assisted' AND 'GnRH agonist' OR 'GnRH antagonist' with the time limitation of 2002-2013. In addition, the Google Scholar database was similarly searched for studies related to this topic from 2002-2013. Finally, the bibliographies of the identified studies were hand searched. Only RCTs were included in this systematic review. The databases were electronically searched in May 2013. In addition, the references of all of the selected studies were searched manually. Only abstracts written in English were considered. Excluded studies included retrospective and uncontrolled studies, editorials, reviews and meta-analyses.

Identification
Eighty-two potentially relevant studies were identified. Of these, 61 studies were excluded because it was clear from the title that they did not fulfill the selection criteria. Four were excluded based on the abstract. The remaining 17 studies were read in full by 2 independent researchers (Haiyan Lin and Yu Li) to evaluate whether they fulfilled the inclusion criteria. Eight of these manuscripts were excluded, and nine [4][5][6][7][8]10,12,14,17] were considered potentially appropriate for the meta-analysis ( Figure 1). All of the bibliographies of the included publications were reviewed, and no additional articles were identified. The supporting PRISMA checklist are available as supporting information; see Checklist S1.

Review methods
Nine prospective RCTs were identified that addressed our research question. The study characteristics and participant features were extracted from each study. The main characteristics and level of quality of the included studies are summarized in Table 1

Statistical analysis
The meta-analysis was conducted using Review Manager software (RevMan) (Version 5.2, The Nordic Cochrane Centre, The Cochrane Collaboration, 2012) to calculate odds ratios (ORs), 95% confidence intervals (CIs) and heterogeneity using the Mantel-Haenszel method, assuming random effects. Statistical pooling was performed for the following outcome parameters: age, BMI, dose of gonadotropin administered, duration of stimulation, number of oocytes retrieved, number of clinical pregnancies and rate of severe OHSS. Considering an I 2 .50% as evidence of substantial heterogeneity, a random-effect model was used for the analysis where appropriate.

Results
In total, 588 women were randomized to treatment with a GnRH agonist long protocol, and 554 women were randomized to treatment with a GnRH antagonist protocol. Seven studies compared the mean age and BMI between the two protocols. However, two studies analyzed the medians only. Pooling the results of these nine RCTs showed no significant difference between patients treated with the long agonist protocol compared with the antagonist protocol in age (weighted mean difference (WMD): 0.18, 95% CI: 20.47-0.83) and BMI (WMD: 20.31, 95% CI: 20.90-0.27). No significant statistical heterogeneity was detected (I 2 : 23% and 31%, respectively; Figure 2).
Two studies showed the median gonadotropin dose administered. One study did not supply the BMI for comparison. The remaining 6 studies were pooled for meta-analysis. A significant difference was observed in the amount of gonadotropin administered in PCOS patients between the two protocols (WMD: 367.90, 95% CI: 87.62-648.18) in the random-effect model. Significant heterogeneity was detected between studies (I 2 : 92%) because some studies used measurements such as ampoules of gonadotropin such as human menopausal gonadotropin (hMG), while others reported the total number of units of gonadotropin.
Three studies reported data on the median duration of gonadotropin stimulation, and these data were not used in the meta-analysis calculations. The mean duration of stimulation was not significantly between the two groups. The WMD was 1.06 days (95% CI: 0.39-1.74). Significant statistical heterogeneity was detected between studies (I 2 : 79%; Figure 2). Significantly more oocytes per retrieval cycle were obtained in patients treated with GnRH agonists compared with antagonists (WMD: 21.61, 95% CI: 23.12-20.10). No significant heterogeneity was detected between studies.
The CPR per embryo transfer was reported in all nine studies. Pooling the results of all nine RCTs showed no significant difference between patients treated with the long agonist protocol compared with the antagonist protocol (OR: 0.91, 95% CI: 0.68-1.23;). No significant statistical heterogeneity was detected (I 2 : 25%; Figures 3 and 4). Six studies also reported the ongoing pregnancy rate (OPR). Pooling the results of these studies showed no significant difference between patients treated with the long agonist protocol compared with the antagonist protocol (OR: 1.05, 95% CI: 0.81-1.37); two studies reported the exact live birth rate (LBR) or delivery rate.
Data regarding fertilization rates and implantation rates were reported in different statistical measures without standard errors but with significant heterogeneity and therefore could not be pooled. All nine RCTs reported the fertilization rate, but some studies expressed it as a percentage (low and maximum interval), while others reported the median rate. Only 2 RCTs expressed the result as the mean +/2 standard error. Four RCTs reported the implantation rate, while 2 RCTs reported the rate as median (range).
In total, 8 studies recorded the rates of moderate and severe OHSS. One study simply compared the number of patients at high risk for OHSS. No cases of OHSS occurred in one study. The data on the occurrence of severe OHSS were pooled from 4 studies (OR: 1.56, 95% CI: 0.29-8.51; Figure 5).

Discussion
After prospective randomized phase III studies comparing GnRH antagonist protocols with GnRH agonist long protocols were published in 2000-2001 [20], GnRH antagonist protocols came to be used more frequently in clinics in a wide range of patients, including patients with normal ovarian responses as well as patients with poor or high responses; this last group includes patients with PCOS.
This systematic review and meta-analysis compares two of the most commonly used ovarian stimulation protocols in women with PCOS. In GnRH antagonist protocols, a GnRH agonist can be used as an alternative to hCG as the ovulation trigger in cases at high risk for OHSS, which can be a secondary confounding factor for meta-analysis. According to the most recent systematic review [21], there was no significant difference in the incidence of severe OHSS in the antagonist group compared with the long agonist group. Thus, because it was not appropriate to perform metaanalysis of RCTs from multiple centers using the OHSS rate as a primary outcome, we used the CPR as the primary outcome in this meta-analysis.
The CPR, OPR and LBR achieved after ovarian stimulation and IVF represent very important outcome measures for any treatment protocol. There has been conflicting evidence in the literature regarding whether an antagonist protocol is as effective as a long agonist protocol with respect to CPR and OPR. Significantly lower OPRs are observed in patients on fixed 6-day GnRH antagonist cycles [22]. Furthermore, one study indicated that the stability of LH levels rather than absolute LH levels were associated with the CPR because no pregnancies occurred if the LH levels changed too markedly during GnRH antagonist administration [23]. To further test this suggestion, more largescale RCTs comparing these regimens are required. However, the most recent Cochrane review in 2011 reported no significant difference in pregnancy rate (OR: 0.80, 95% CI: 0.57-1.11) or abortion rate (OR: 1.29, 95% CI: 0.49-3.36) between the two protocols in PCOS patients [24]. To identify a 5% difference in CPR using a b of 0.1 and an a of 0.05 with a two-tailed hypothesis test, over 1364 patients would be required in each treatment arm. Fortunately, the results of 2 large-scale studies were published in 2012.
Previous studies such the report by Xiao et al. [25] performed meta-analysis on seven studies published through 2011. We have extended that time period to include through May 2013. The present meta-analysis included nine RCTs with a total of 1142 women with PCOS and showed no significant difference between the antagonist and long agonist protocols with regard to CPR and OPR. No statistical heterogeneity was detected in the results of the included studies regarding the CPR and OPR outcomes. Thus, these results, based on a large patient population, provide further evidence that using either protocol yields a similar CPR for the treated cycle.
Meta-analysis in general has several drawbacks, such as a dependence on the quality of the reporting of the primary analysis findings and on a sufficient number of eligible studies to justify the statistical analysis. This meta-analysis has an additional disadvantage because of the different statistics regarding gonadotropin dose  and duration and the lack of data on the miscarriage rates and the LBRs. Nevertheless, the findings of this meta-analysis contribute to systematizing our knowledge about outcomes of conventional IVF in women with PCOS. In our meta-analysis, we could not use the same statistics to study the total doses of gonadotropin used or the duration of stimulation due to the high statistical heterogeneity. Previous studies have all agreed that the antagonist protocol is more patient friendly and convenient [3,12]. However, the number of scans, blood tests and clinic visits were not recorded in each study, making a full economic evaluation impossible. This convenience, added to the equal CPR compared to the agonist protocol, suggests that the antagonist protocol may be the most suitable for PCOS patients. Still, because the antagonist can be started alternatively on the first day of stimulation, on day 6 or according to the follicles' diameter, to date, no comparative studies have been performed in PCOS patients to determine which protocol is best for this population.
In conclusion, this meta-analysis did not find a statistically significant difference between the use of a GnRH antagonist protocol compared with the standard long protocol with respect to CPR and OPR in patients with PCOS undergoing IVF. Although different classifications for OHSS were used in different studies, a GnRH antagonist protocol also appears to significantly reduce the rate of severe OHSS, but further RCTs are necessary to allow more solid conclusions to be drawn. In addition, these larger RCTs should include adequate sample sizes and a standardized classification for OHSS along with economic evaluations and LBR tracking.