The effect of G-CSF on infertile women undergoing IVF treatment: A meta-analysis

ABSTRACT

Evidence for the effect of granulocyte colony stimulating factor (G-CSF) on infertile women undergoing in vitro fertilization (IVF) remains inconsistent. This study aimed to evaluate the effectiveness of G-CSF on infertile women undergoing IVF. PubMed and EMBASE databases were searched before August 2016. Comparing the transvaginal perfusion of G-CSF and placebo or no treatment, the available studies were considered. The pooled risk ratio (RR) with 95% confidence intervals (CIs) was used in the analysis and six studies were included. Transvaginal perfusion of G-CSF was significantly associated with a higher clinical pregnancy rate versus the placebo (RR=1.563, 95%CI: 1.122, 2.176), especially for the Asian population. Among patients with a thin endometrium or repeated IVF failure, the implantation and biochemical pregnancy rates were also significantly increased in patients with the use of G-CSF (implantation rate: RR = 1.887, 95% CI: 1.256, 2.833; biochemical pregnancy rate: RR = 2.385, 95% CI: 1.414, 4.023). However, no statistical significance in increasing endometrial thickness was detected. Transvaginal perfusion of G-CSF for infertile women may play a critical role in assisting human reproduction, especially for patients with a thin endometrium or repeated IVF failure in the Asian population.

KEYWORDS:

• G-CSF
• granulocyte colony-stimulating factor
• infertility
• in vitro fertilization
• IVF

Introduction

In vitro fertilization (IVF), through assisted reproductive technology (ART), is used to treat infertility. However, the success rate of IVF is still less than 40%, and it is an expensive, time consuming technology [Zeyneloglu and Onalan 2014]. Many infertile couples who are suffering from physical, financial, and emotional distress, still remain unsuccessful after IVF attempts. IVF success is mainly determined by the age of the woman, embryo quality, anti-Mullerian hormone (AMH) concentrations, and number of embryos implanted [Mutluet al. 2013; Broekmans et al. 2006].

It has also been demonstrated that endometrium thickness below 7 mm is negatively associated with the chance of pregnancy [Revel 2012; Singh et al. 2011]. It is estimated that 0.6–0.8% of IVF patients did not reach a thickness of 7 mm [Al-Ghamdi et al. 2008]. Various therapies have been tested including extended estrogen administration and treatment with low-dose aspirin, vaginal sildenafil citrate, and treatment with pentoxifyllin and tocopherol [Chen et al. 2006; Ledee-Bataille et al. 2002]. However, even using these therapies, some patients with a thin endometrium remain unresponsive. Studies showed that the mechanisms of immunology in the endometrium played an important role in the implantation process [Sharkey 1998]. Some studies demonstrated that the increased implantation and pregnancy rates were due to the enormous release of growth factors, cytokines, and hormones, which were produced by decidual cells [Psychoyos 1986].

As a hematopoietic lineage-specific cytokine, granulocyte colony-stimulating factor (G-CSF) can stimulate the proliferation and differentiation of neutrophils which are produced by stromal cells, endothelial cells, the bone marrow, fibroblasts, monocytes, and macrophages. G-CSF participates in increasing phagocytosis and the oxidative process in mature neutrophils [Thomas et al. 2002]. Recent studies have proposed G-CSF to play an important role in human reproductive success by affecting implantation. The mechanism was supposed to stimulate neutrophilic granulocyte proliferation and differentiation and act on macrophages of decidual cells [Barash et al. 2003]. However, the clear mechanisms involved in affecting pregnancy outcomes are ambiguous. G-CSF administration was reported to be associated with recruiting dendritic cells, promoting Th-2 cytokine secretion, and activating T regulatory cells, which might influence crucial gene expression in the endometrium, including local immune modulation, vascular remodeling of the endometrium, and cellular adhesion pathways [Rahmati et al. 2014]. In addition, studies reported that G-CSF receptor expression also took place in the trophoblastic cells [Salmassi et al. 2004]. As it is known, sexual intercourse could influence the expression of pro-inflammatory cytokines and chemokines, which might facilitate embryo implantation and progression in the reproductive cycle [Schjenken et al. 2015; Robertson 2010]. As one of the cytokines, G-CSF was suggested to be important in improving clinical pregnancy rate [Crawford et al. 2015]. However, in patients with thin endometrium or repeated IVF failures, their capacity to produce G-CSF could be weakened because of the thin endometrium and abnormal function of endometrium. These studies have suggested that G-CSF might play an important role in follicular maturation, ovulation, and implantation. However, whether G-CSF has significant physiological effects or clinical benefit on IVF outcomes remains uncertain.

In recent years, an increasing number of studies were focused on evaluating the effect of G-CSF on infertile women undergoing IVF, especially women with a thin endometrium [Gleicher et al. 2011; Gleicher et al. 2013; Aleyasin et al. 2016; Xu et al. 2015; Barad et al. 2014; Eftekhar et al. 2014]. However, these researches have provided conflicting results. In 2011, Gleicher et al. [2011] first presented a clinical study with four patients regarding the usefulness of granulocyte colony-stimulating factor in thin unresponsive endometrium; all participants were successfully pregnant. In 2013, the same author conducted a prospective observational cohort pilot study of G-CSF in consecutive infertile women with thin endometrium; the results presented an increase in endometrial thickness and clinical pregnancy [Gleicher et al. 2013]. Subsequent to the basic research, a series of randomized controlled trials (RCTs), case–control studies, and self-controlled trials were undertaken on this subject. Aleyasin et al. [2016] and Xu et al. [2015] suggested that the implantation and pregnancy rates in the G-CSF group were significantly higher in IVF patients with thin endometrium or repeated IVF failure compared with the control group, whereas in the study of Barad et al. [2014], IVF patients treated with G-CSF did not have optimum outcomes, including endometrial thickness, implantation, or pregnancy rates. In 2014, Eftekhar et al. [2014] conducted a non-randomized intervention clinical trial for infertile women with a thin endometrium treated with transvaginal perfusion of G-CSF, which demonstrated that G-CSF did not have the potential to improve endometrial thickness and increase chemical and clinical pregnancy rates. On the basis of the current controversy, this meta-analysis was conducted with data published in recent years, aiming to explore the contribution of G-CSF on infertile women undergoing IVF.

Results

Study selection and characteristics

In total, 1,204 articles were collected from the PubMed and EMBASE databases. Then, after screening the titles and abstract, 1,188 studies were excluded for not meeting the criteria, and 16 studies remained to be assessed in detail. After evaluating the full text carefully, ten were removed for the following reasons: five studies with invalid datasets were excluded because they belonged to cohort studies and they were only focused on monitoring endometrial thickness. Further, the raw data for pregnancy outcome could not be extracted; one study did not meet the inclusion criteria; and four were reviews, letters, or case reports . Finally, six studies containing 607 participants were eligible [Aleyasin et al. 2016; Xu et al. 2015; Barad et al. 2014; Eftekhar et al. 2014; Li et al. 2014; Eftekhar et al. 2016]. Among these, three [Xu et al. 2015; Eftekhar et al. 2014; Li et al. 2014] were case-control design and the others [Aleyasin et al. 2016; Barad et al. 2014; Eftekhar et al. 2016] were RCTs. Additionally, five studies [Aleyasin et al. 2016; Xu et al. 2015; Eftekhar et al. 2014; Li et al. 2014; Eftekhar et al. 2016] mainly compared G-CSF with no treatment; these studies were focused on exploring the Asian population. In contrast, the other study [Barad et al. 2014] was a placebo control and focused on the American population. As for the treatment protocol, patients were undergoing IVF or frozen embryo transfer in the six studies [Aleyasin et al. 2016; Barad et al. 2014; Xu et al. 2015; Eftekhar et al. 2014; Li et al. 2014; Eftekhar et al. 2016]. With respect to the time of treatment, transvaginal perfusion G-CSF was administered before embryo transfer in three studies [Aleyasin et al. 2016; Eftekhar et al. 2014; Eftekhar et al. 2016]. In one study [Xu et al. 2015], G-CSF transvaginal perfusion was carried out on the day that one follicle became dominant, while in another study [Barad et al. 2014], it was conducted on the morning of hCG administration before noon. It was used in the remaining study [Li et al. 2014] only on the day of ovulation or the administration of progesterone or hCG. Moreover, 300 micrograms of G-CSF was administered in five studies [Aleyasin et al. 2016; Xu et al. 2015; Barad et al. 2014; Eftekhar et al. 2014; Eftekhar et al. 2016], whereas 100 micrograms of G-CSF was administered in another [Li et al. 2014]. The process flow diagram of selected articles is presented in Figure 1 and the characteristics of the included studies are summarized in Table 1.

Table 1. Characteristics of controlled trials on G-CSF and pregnant outcomes included in the meta-analysis.

		Interventions			Study group		Control group
Authors	Study design/Race	Cases	Controls	Analysis endpoints	Events/Mean ± SD	Total	Events/Mean ± SD	Total	Treatment protocol	Basic information
		300 microgram of G-CSF was administered on the morning of hCG administration before noon	placebo (normal saline)	Endometrium	10.23 ± 2.01	73	10.37 ± 2.22	68	Undergoing IVF and frozen embryo transfer in a first study cycle.	Unselected IVF patients with normal endometrial thickness. Endometrial thickness was assessed 5 days later after G-CSF infusion.
				Implantation	22	176	28	168
Barad et al. (2014)	RCT/Mixed			Biochemical pregnancy	18	73	19	68
				Clinical pregnancy	17	73	16	68
Eftekhar et al. (2014)	Case-control/Asian	300 microgram G-CSF was administered before embryo transfer	without G-CSF	Endometrium	7.91 ± 0.55	34	8.23 ± 0.82	34	Undergoing frozen embryo transfer	Endometrial thickness below 7mm at 12-13 day of frozen embryo transfer cycle. Endometrial thickness was assessed at the first day of progesterone injection.
				Biochemical pregnancy	11	34	4	34
				Clinical pregnancy	9	34	3	34
Li et al. (2014)	Case-control/Asian	100 microgram G-CSF was used only on the day of ovulation or administration of progesterone or hCG	without G-CSF	Implantation	13	82	14	105	Undergoing frozen embryo transfer	Endometrial thickness below 7mm prior to ovulation. Endometrial thickness was assessed 5 days later after G-CSF infusion.
				Clinical pregnancy	10	40	12	80
Xu et al. (2015)	Case-control/Asian	300 microgram G-CSF was carried out on the day that one follicle became dominant	without G-CSF	Implantation	9	28	15	108	Undergoing frozen embryo transfer	Endometrial thickness below 7mm prior to ovulation. Endometrial thickness was assessed 24 or 48 h later after G-CSF infusion.
				Clinical pregnancy	1	14	1	52
Aleyasin et al. (2016)	RCT/Asian	300 microgram G-CSF was administered before embryo transfer	without G-CSF	Endometrium	11.2 ± 2	56	11.5±2.3	56	Undergoing IVF	IVF patients with repeated IVF failure
				Implantation	24	133	10	138
				Biochemical pregnancy	25	56	11	56
				Clinical pregnancy	21	56	8	56
Eftekhar et al. (2016)	RCT/Asian	300 microgram G-CSF was administered at the oocyte retrieval day	without G-CSF	Biochemical pregnancy	9	50	10	50	Undergoing IVF	Unselected IVF patients with normal endometrial thickness who were undergoing IVF
				Clinical pregnancy	9	50	9	50

G-CSF: granulocyte colony-stimulating factor; RCT: randomized controlled trials; Mixed: American population.

Figure 1. The process flow diagram of selected studies.

Implantation rate

Four datasets evaluated the effect of G-CSF on the implantation rate with 938 samples involved [Aleyasin et al. 2016; Xu et al. 2015; Barad et al. 2014; Li et al. 2014]. The implantation rate was 16.2% in those receiving transvaginal perfusion of G-CSF compared to 12.9% in those receiving placebo or no treatment. However, it did not reach statistical significance with the random effects model (RR=1.461; 95% CI: 0.801, 2.664, I²=70.7%) (Table 2). Considering the higher heterogeneity, subgroup and sensitivity analysis was conducted. After analysis, the study with the highest heterogeneity was shown [Barad et al. 2014], which was the only study focused on patients with normal endometrial thickness for the American population. Then, the highest heterogeneity study was removed, leaving just three studies [Aleyasin et al. 2016; Xu et al. 2015; Li et al. 2014] which were focused on patients with thin endometrium or repeated IVF failure in the Asian population. The new results suggested that the use of G-CSF was effective in increasing the implantation rate (RR= 1.887; 95% CI: 1.256, 2.833, I²=23.2%) among patients with thin endometrium or repeated IVF failure in the Asian population, but not for the normal endometrial thickness group (Table 3; Figure 2A). The funnel plot was estimated to be symmetric using Begg’s test and no significant bias was detected.

Table 2. Comparing G-CSF treatment versus placebo for infertile women undergoing IVF.

	Fixed/Random model			Sensitivity analysis
Outcomes	Pooled relative risk (95% CI)	I^2 (%)	Heterogeneity (P)	Pooled relative risk (95% CI)	I^2 (%)	Heterogeneity (P)
Implantation	1.461 (0.801,2.664)	70.7	0.017	1.887 (1.256,2.833)	23.2	0.272
Biochemical pregnancy	1.426 (0.795,2.559)	61.9	0.049	1.800 (1.172,2.765)	50.2	0.134
Clinical pregnancy	1.563 (1.122,2.176)	26.8	0.234	1.897 (1.267,2.840)	0	0.417
Mean difference (95% CI)
Endometrium	-0.173 (-0.393,0.46)	0	0.411	-0.260 (-0.564,0.043)	5.5	0.304

Table 3. Subgroup analysis of pregnant outcomes.

		Fixed/Random model
Outcomes	Subgroups analysis	Pooled relative risk (95% CI)	I^2 (%)	Heterogeneity (P)
Implantation	Thin endometrium/RIF	1.887 (1.256,2.833)	23.2	0.272
	Normal IVF	0.750 (0.447,1.258)	NA	NA
	Overall	1.461 (0.801,2.664)	70.7	0.017
Biochemical pregnancy	Thin endometrium/RIF	2.385 (1.414,4.023)	0	0.756
	Normal IVF	0.888 (0.562,1.403)	0	0.969
	Overall	1.426 (0.795,2.559)	61.9	0.049
Clinical pregnancy	Thin endometrium/RIF	2.312 (1.444,3.701)	0	0.766
	Normal IVF	0.993 (0.611,1.616)	0	0.984
	Overall	1.563 (1.122,2.176)	26.8	0.234

NA: Not applicable.

Figure 2. The pooled risk ratios (RRs) or the standardized mean difference (SMD) with 95% confidence intervals (CIs) of the relationship between G-CSF/placebo or no treatment and implantation rate with random models (A), clinical pregnancy rate with fixed models (B), endometrial thickness (C), and funnel plots for clinical pregnancy rate (D).

Biochemical pregnancy rate

In the next analysis, four studies were included to evaluate the effect of G-CSF on biochemical pregnancy rate with 421 participants [Aleyasin et al. 2016; Barad et al. 2014; Eftekhar et al. 2014; Eftekhar et al. 2016]. Biochemical pregnancy occurred in 63 of 213 (29.6%) participants utilizing G-CSF and in 44 of 208 (21.2%) patients receiving placebo or no treatment. A significant difference was detected in the subgroup of patients with a thin endometrium or repeated IVF failure, and the RR was 2.385 (95% CI: 1.414, 4.023, I²=0.0%) (Table 3; Figure 2B). However, no statistically significant differences were detected for the normal IVF group (RR=0.888; 95% CI: 0.562, 1.403, I²=0.0%). Compared with the placebo or no treatment, the use of G-CSF was associated with a significantly higher biochemical pregnancy rate for the Asian population after removing the highest heterogeneity study [Barad et al. 2014] (RR=1.800; 95% CI: 1.172, 2.765, I²=50.2%) (Table 2). In addition, the higher heterogeneity (61.9%) was decreased to 50.2% No significant bias was detected.

Clinical pregnancy rate

As for the effect of G-CSF on the clinical pregnancy rate, six studies involving 607 participants were included [Aleyasin et al. 2016; Xu et al. 2015; Barad et al. 2014; Eftekhar et al. 2014; Li et al. 2014; Eftekhar et al. 2016]. Successful clinical pregnancy occurred in 67 of 267 (25.1%) patients utilizing G-CSF and in 49 of 340 (14.4%) participants for patients receiving placebo or no treatment. Compared with placebo or no treatment, the use of G-CSF was associated with a higher clinical pregnancy rate, and the RR was 1.563 (95% CI: 1.122, 2.176, I²=26.8%) with the fixed effects model (Table 2). Moreover, more significant differences were detected in the subgroup of patients with thin endometrium or repeated IVF failure, and the RR was 2.312 (95% CI: 1.444, 3.701, I²=0.0%) (Table 3; Figure 2B). However, no statistically significant differences were observed for the normal IVF group (RR=0.993; 95% CI: 0.611, 1.616, I²=0.0%). In the sensitivity analysis, the study focusing on the effect of G-CSF for patients with normal endometrial thickness also had the highest heterogeneity [Barad et al. 2014]. After removing this article, five studies for the Asian population [Aleyasin et al. 2016; Xu et al. 2015; Eftekhar et al. 2014; Liet al. 2014; Eftekhar et al. 2016] remained and the statistical difference was also significant. Using Begg’s test, we did not detect any significant bias (p=0.326) (Figure 2D).

Endometrium thickness

Only three studies with 321 participants were included to assess the endometrium thickness [Aleyasin et al. 2016; Barad et al. 2014; Eftekhar et al. 2014]. The use of G-CSF was ineffective in increasing the endometrium thickness among infertile women undergoing IVF. The SMD between patients using G-CSF and controls was −0.173 (95% CI: -0.393, 0.46, p=0.411). However, there was also no statistically significant difference compared with the placebo or no treatment (SMD: -0.260, 95% CI: -0.564, 0.043, p=0.304) (Table 2, Figure 2C).

Discussion

In this meta-analysis, the effect of G-CSF on pregnancy outcomes was assessed in patients undergoing IVF for infertility, especially for women with a thin endometrium or repeated IVF failure. Six studies with 607 participants were analyzed and the results indicated that G-CSF might play a critical role in assisting human reproduction, especially for patients with a thin endometrium or repeated IVF failures in the Asian population. The increase in clinical pregnancy rate associated with the use of G-CSF was statistically significant. However, compared with the placebo or no treatment, no statistical significance was found in increasing endometrial thickness, implantation rate, and biochemical pregnancy rate. It is worthy of attention that statistical significance was found in implantation and biochemical pregnancy rates after removing the study with the highest heterogeneity [Barad et al. 2014] and the clinical pregnancy rate was more statistically significant for the Asian population.

For infertile women with a thin endometrium or repeated IVF failure, G-CSF could be expected to improve pregnancy outcomes. In the four included studies, thin endometrium was diagnosed when the thickness was below 7mm on the day of hCG administration. However, Barad et al. [2014] evaluated the effect of G-CSF in women with normal endometrium thickness undergoing IVF treatment; this was the study with the highest heterogeneity in our analyses.

The first time G-CSF was used to improve the endometrium in women with a thin endometrium undergoing IVF was in 2011 by Gleicher et al. [2011]. Four older patients were treated with intrauterine infusions of G-CSF and all became pregnant. Having shown promising initial results, Gleicher et al. [2013] went on to a study with 21 participants. In this small cohort study, G-CSF was administered by an intrauterine route and the effect of G-CSF was detected by increasing the endometrial thickness. Following Gleicher et al. [2011, 2013], and with a similar protocol, Kunicki et al. [2014] assessed the G-CSF effects on unresponsive thin endometrium in women undergoing IVF. This included thirty-seven patients who had a thin unresponsive endometrium. However, there was no difference in endometrial thickness between women who became pregnant and those who did not. Because of the limited number of patients and the absence of a control group, the inconsistent conclusions require additional research. In cycles of frozen embryo transfer, Li et al. [2014] conducted a retrospective study on the use of G-CSF with non-responsive (<7mm) endometrium. During endometrial preparation, cases were treated with uterine infusions of 100 μg G-CSF, whereas controls did not use G-CSF. Compared with that after G-CSF infusion, no increase in endometrial thickness was observed before uterine infusion of G-CSF in the case group. In this meta-analysis, there were three studies providing clear data to explore the effect of G-CSF on endometrial thickness. Our results suggested that the use of G-CSF was ineffective in increasing the endometrium thickness among infertile women undergoing IVF. However, compared with placebo or no treatment, the implantation rate, biochemical pregnancy rate, and clinical pregnancy rate were significantly higher in patients with a thin endometrium or repeated IVF failure using G-CSF. In other words, uterine infusion of G-CSF is a novel proposal for immune therapy in patients with a thin endometrium or cases of repeated IVF failure [Cavalcante et al. 2015], and it might not be simply for the increased endometrial thickness. In addition, the inconsistent measurement time regarding endometrial thickness may also affect the results.

In 2014, the effect of G-CSF on pregnancy outcome following IVF treatment in 141 women with normal endometrium thickness was reported by Barad et al. [2014]. Cases received an intrauterine infusion of 300 μg G-CSF, whereas controls received an intrauterine infusion of saline. In the group of patients with normal endometrial thickness, Barad et al. [2014] concluded that the use of G-CSF was ineffective in improving the implantation and pregnancy rates. This, the most heterogeneous study in our analyses, was from an American population. These subjects had no history of previous unsuccessful IVF attempts and no history of previous cancellation of attempts because of a thin endometrium. Therefore, as an immunogenic therapy, by stimulating neutrophilic granulocyte proliferation and differentiation and acting on macrophages of decidual cells, uterine infusion of G-CSF might not play a role in pregnancy outcome for patients with normal endometrium thickness.

Recent studies suggested that pro-inflammatory cytokines and chemokines could appear in the female reproductive tract through the interaction of seminal fluid and epithelial cells [Schjenken et al. 2015; Sharkey et al. 2012]. Moreover, by balancing cytokines and immune cells, the female immune response to seminal fluid might facilitate or prevent embryo implantation and progression [Robertson 2010]. In this process, as one of the important cytokines and chemokines G-CSF was identified, which suggested that the levels of G-CSF increased during coitus and seminal fluid exposure, further contributing to conception [Schjenken et al. 2015; Sharkey et al. 2012]. Additionally, in 2015, a meta-analysis demonstrated that significantly improved pregnancy outcomes presented when women were exposed to seminal fluid during the time of ovum pick-up or embryo transfer [Crawford et al. 2015]. Crawford et al. [2015] also emphasized seminal fluid played a vital role in the regulation of implantation by regulating endometrial epithelial cells and the release of pro-inflammatory cytokines including G-CSF. As for the patients undergoing ART treatment, adding G-CSF to the embryo culture medium could increase clinical pregnancy outcomes to some extent [Ziebe et al. 2013]. Similar data was also shown for IVF treatment, which not only identified the association between G-CSF and seminal plasma, but also the benefit of G-CSF administration. However, on the contrary, as for patients with thin endometrium or repeated IVF failures, the production of G-CSF was influenced by abnormal function of endometrium, which might also further affect implantation.

To explore the source of heterogeneity, sensitivity analysis was conducted in IVF outcomes by removing the highest heterogeneity study [Barad et al. 2014]. Statistically significant changes were detected, and the increase in implantation and biochemical pregnancy rates was associated with the use of G-CSF. In addition, the clinical pregnancy rate was more statistically significant in the subgroup of patients with thin endometrium or repeated IVF failure for the Asian population.

Although large sample sizes were included in the present analysis, some limitations are clear. Firstly, a series of trials was undertaken on this subject and the studies included were published in a three year period. While independent verification with time is required. In addition, the moderate heterogeneity was detected using the I² statistic among the included studies compared with the results of the meta-analysis overall. The different statistical results were presented for the rates of implantation and biochemical pregnancy after sensitivity analysis. This was due to the different inclusion and exclusion criteria for the study published by Barad et al. [2014]. Furthermore, the other studies mainly focused on exploring the effect of G-CSF on infertile women with a thin endometrium or repeated IVF failure who were undergoing IVF treatment within an Asian population, which can highlight our research focus. However, other ethnic groups cannot be further analyzed on the basis of the limited studies. Second, due to the limitation of a lack of data, a more thorough exploration of the possible effects G-CSF on other pregnancy outcomes was not possible. Lastly, some confounding factors were observed in the studies included in this analysis, such as the high and different age of participants, the different duration and the period of use of G-CSF therapy, statistical methods of sample sources, and the different types of experimental design. Thus, to determine whether uterine infusion of G-CSF can improve pregnancy outcomes, randomized controlled trials with larger sample sizes are needed.

The above analyses of the effect of G-CSF on IVF outcomes in infertile women suggest that G-CSF may play a role in assisting human reproduction, especially among patients with a thin endometrium or repeated IVF failure in the Asian population. This reflects increasing implantation rate, biochemical pregnancy rate, and clinical pregnancy rate. However, no statistical significance was detected in increasing endometrial thickness when compared to the placebo or no treatment groups. Thus, larger randomized controlled trials are needed to confirm these findings with particular attention paid to other pregnancy outcomes and other ethnic groups.

Materials and methods

Search strategy and selection criteria

Two databases (PubMed and Embase) were searched for studies that explored the effect of transvaginal perfusion of G-CSF for infertile women undergoing IVF, with the following keywords combined: “granulocyte colony-stimulating factor”, “G-CSF”, “endometrium”, “recurrent implantation failure”, “RIF”, “in vitro fertilization”, “IVF”, “embryo transfer”, “ET”, “intracytoplasmic sperm injection”, and “ICSI”. The last retrieval was carried out in August 2016, with no language restrictions.

RCTs and case–control studies were eligible for inclusion. All of the available studies were comparisons between the transvaginal perfusion of G-CSF and placebo or no treatment. In the process of IVF treatment, 100 or 300 micrograms of G-CSF was given to patients before embryo transfer. Transvaginal perfusion of G-CSF was conducted at different stages of the IVF cycle: i.e., i) on the day that one follicle became dominant; ii) on the day of ovulation or administration of progesterone or hCG; iii) on the days before embryo transfer; or iv) in the morning of hCG administration before noon. Finally, the following outcomes were assessed: endometrial thickness, implantation rate, biochemical pregnancy rate, and clinical pregnancy rate. Meanwhile, duplicate articles, case reports, review articles, and letters were excluded. As a meta-analysis which mainly tried to combine all the published data, a statement of institutional ethics committee approval was not required for this study.

Two investigators (JL and YC) mainly performed the process of data extraction. If there were any disagreements, the original articles were referred back to reach a consensus. The characteristics were collected as follows: names of authors, year of publication, study design and race, interventions, IVF outcomes, sample size, treatment protocol, and other basic information.

Types of outcome measures

In this study, the endometrial thickness, implantation rate, biochemical pregnancy rate, and clinical pregnancy rate were treated as the primary outcomes analyzed. Other outcomes were not assessed because of missing related data on these measures. In order to further explore the effect of G-CSF on infertile women with thin endometrium or repeated IVF failure or normal endometrium who were undergoing IVF treatment, subgroups regarding implantation rate, biochemical pregnancy rate, and clinical pregnancy rate were summarized.

Statistical analysis

In the study, the pooled risk ratios (RRs) and standardized mean differences (SMDs) with 95% confidence intervals (CIs) were calculated for IVF outcomes. Forest plots were used to present the data graphically. In addition, to estimate heterogeneity, the Cochrane’s Q and I² statistic were applied (p < 0.10 as the standard) [Lau et al. 1997]. Values of 25%, 50%, and 75% for I² indicate low, moderate, and high heterogeneity, respectively. Meanwhile, in selecting the effect models, I² was used as the standard (fixed model: I² < 50%; random effect: I² ≥ 50%). In order to explore the sources of heterogeneity, subgroup analysis especially for infertile women with a thin endometrium or repeated IVF failure treated with G-CSF was performed in the Asian population. Then, a sensitivity analysis was also carried out.

The potential publication bias was assessed by Begg’s unweighted regression test and funnel plots [Egger et al. 1997]. All of the analyses in this study were conducted with Stata version 9.0 (StataCorp, College Station, TX, USA). All p-values were two-sided.

Declaration of interest

The authors have no conflicts of interest.

Additional information

Notes on contributors

Jie Li

Took part in the formation of preliminary idea and selection of this topic and wrote the “Introduction” and “Discussion” sections: JL, SM; Wrote the “Materials and methods” and “Results” sections and participated in summarizing and systematically assessing the effect of G-CSF on IVF outcomes in infertile women: JL, YC. All authors were responsible for writing and modify the draft and final paper.