A total of 957 records were screened from the PubMed, EMBASE, and CNKI databases. After screening of the titles and abstracts, 921 records were excluded for not meeting the criteria and 36 studies remained for detailed full-text evaluation. 19 articles were excluded for the following reasons: 12 studies lacked valid data or did not compare pretreatment with or without GnRH-agonist before FET and 4 were reviews, reports, or conferences. Finally, 17 studies [16,17,18,19, 23,24,25,26,27,28,29,30,31,32,33,34,35] containing 3646 participants were eligible. The process flow diagram of selected studies is presented in Fig. 1. Comparing the effectiveness of pretreatment with or without GnRH-agonist for PCOS on clinical outcomes before FET, four RCTs [16, 18, 19, 29] and thirteen case–control studies [17, 23,24,25,26,27,28, 30,31,32,33,34,35] were analyzed. In the control group, simple artificial cycle regimen was applied for PCOS patients who received standard treatment for endometrial preparation using estradiol valerate before embryo transfer. The pretreatment group was treated with GnRH-a down-regulated artificial cycle regimen, patients with PCOS received a depot of long-acting GnRH agonist before beginning exogenous hormone supplementation. The characteristics of patients included in the analysis are summarized in Table 1.

Articles identification for the process flow diagram

Ultimately, quality evaluation of articles was carried out including four RCTs and thirteen case–control studies. The quality of one RCT study was for level A, other three studies for level B. The quality assessment for RCT studies was summarized in Tables 2 and 3 for case–control studies.

As for the effect of GnRH-a pretreatment before FET for PCOS patients, nine studies [18, 23, 25, 26, 28, 30, 31, 33, 35] with 1915 participants were included to evaluate endometrial thickness on first day of progesterone supplementation including one RCT [18] and eight case–control studies [23, 25, 26, 28, 30, 31, 33, 35]. The pretreatment of GnRH-agonist before frozen-thawed embryo transfer was effective in increasing the endometrium thickness among infertile women with PCOS. The SMD between patients using GnRH-a pretreatment and controls was 0.56 (95% CI: 0.20, 0.92, p = 0.000). Significant differences were detected in the GnRH-a pretreatment group when compared with the control group for endometrium thickness among PCOS patients (Table 4; Fig. 2A).

The pooled RRs or the SMD with 95% CIs of the relationship between FET cycles with or without GnRH agonist pretreatment for endometrial thickness on first day of progesterone supplementation (A), implantation rates (B), clinical pregnancy rates with random models (C), sensitivity analysis (D) and funnel plots for clinical pregnancy rate (E), and miscarriage rates with fixed model (F)

Eleven studies [19, 23, 25,26,27, 29, 30, 32,33,34,35] were included in the comparison of implantation rates with a total of 2897 events with 5362 embryos transferred, and two RCTs [19, 29] and nine case–control studies [23, 25,26,27, 30, 32,33,34,35]. The implantation rate was 46.96% (1029/2191) in those with receiving GnRH-a pretreatment compared to 43.27% (1372/3171) in those without receiving GnRH-a pretreatment. Significant differences were detected between GnRH-a pretreatment patients and placebo patients with using the random effects model; the RR was 1.12 (95% CI 1.00–1.24, I2 = 63.7%). In the study-design subgroup analysis, a higher implantation rate was also observed in GnRH-a pretreatment group compared with the controls for the case–control study group (RR = 1.16; 95% CI: 1.03, 1.32, I2 = 65.0%). However, no significant difference was observed for the RCT group (RR = 0.89; 95% CI: 0.63, 1.24, I2 = 55.8%). To explore the high heterogeneity among studies, a sensitivity analysis was conducted. After analysis, the study with the highest heterogeneity was shown [27], which was the study only focused on PCOS population with first frozen-thawed embryo transfer transplant failure. After removing the highest heterogeneity study, the statistical difference was also similar to previous results. No significant bias was detected and the funnel plot was estimated to be symmetric using Begg’s test. (Table 5; Fig. 2B).

Seventeen studies [16,17,18,19, 23,24,25,26,27,28,29,30,31,32,33,34,35] involving 3640 participants were analyzed for clinical pregnancy rate including four RCTs [16, 18, 19, 29] and thirteen case–control studies [17, 23,24,25,26,27,28, 30,31,32,33,34,35]. Successful clinical pregnancy occurred in 934 of 1531 (61.01%) patients receiving GnRH-a pretreatment and in 1157of 2109 (54.86%) participants for patients without receiving pretreatment. With using the random-effects model, the results presented that the difference in clinical pregnancy rates between the GnRH-a pretreatment and no pretreatment groups was statistically significant, and the RR was 1.19 (95% CI: 1.08, 1.32, I2 = 59.0%) (Table 4; Fig. 2C). Due to the high heterogeneity for the clinical pregnancy rate, we carried out a subgroup analysis and a sensitivity analysis to explore the sources of heterogeneity. In the subgroup analysis for case–control study, a higher clinical pregnancy rate was observed in PCOS patients receiving GnRH-a pretreatment compared with the control group not receiving GnRH-a pretreatment (RR = 1.24, 95% CI: 1.11, 1.38, I2 = 60.4%). However, for the RCT group, there was no significant difference (RR = 1.04, 95% CI: 0.81, 1.34, I2 = 51.4%) (Table 5; Fig. 2C). In the sensitivity analysis, the study focusing on the GnRH-a pretreatment for PCOS population had the highest heterogeneity [30]. After removing the highest heterogeneity study, the statistical difference was also similar to previous results and no significant difference was observed for RCT group (Table 5; Fig. 2D). No significant bias was detected with using Begg’s test (Fig. 2E).

Seventeen studies [16,17,18,19, 23,24,