What is the impact of granulocyte colony-stimulating factor (G-CSF) in subcutaneous injection or intrauterine infusion and during both the fresh and frozen embryo transfer cycles on recurrent implantation failure: a systematic review and meta-analysis?

Background Among recurrent implantation failure (RIF) patients, the rate of successful implantation remains relatively low due to the complex etiology of the condition, including maternal, embryo and immune factors. Effective treatments are urgently needed to improve the outcomes of embryo transfer for RIF patients. In recent years, many researchers have focused on immunotherapy using granulocyte colony-stimulating factor (G-CSF) to regulate the immune environment. However, the study of the G-CSF for RIF patients has reached conflicting conclusions. The aim of this systematic review and meta-analysis was performed to further explore the effects of G-CSF according to embryo transfer cycle (fresh or frozen) and administration route (subcutaneous injection or intrauterine infusion) among RIF patients. Method The PubMed, Embase and Cochrane Central Register of Controlled Trials (CENTRAL) databases were searched for literature published from the initial to October 2020. The meta-analysis, random-effects model and heterogeneity of the studies with I2 index were analyzed. Stata 15 was used for statistical analysis. Results A total of 684 studies were obtained through the databases mentioned above. Nine RCTs included 976 RIF patients were enrolled in this meta-analysis. Subgroup analysis indicated that G-CSF improved the clinical pregnancy rate for both the fresh and frozen embryo transfer cycles (fresh RR: 1.74, 95% CI: 1.27–2.37, I2 = 0.0%, n = 410; frozen RR: 1.44, 95% CI: 1.14–1.81, I2 = 0.0.%, n = 366), and for both subcutaneous injection and intrauterine infusion (subcutaneous RR: 1.73, 95% CI: 1.33–2.23, I2 = 0.0%, n = 497; intrauterine RR: 1.39, 95% CI: 1.09–1.78, I2 = 0.0%, n = 479), but the biochemical pregnancy rate of the RIF group was also higher than that of the control group (RR: 1.85, 95% CI: 1.28–2.68; I2 = 20.1%, n = 469). There were no significant differences in the miscarriage rate (RR: 1.13, 95% CI: 0.25–5.21: I2 = 63.2%, n = 472) and live birth rate (RR: 1.43, 95% CI: 0.86–2.36; I2 = 52.5%; n = 372) when a random-effects model was employed. Conclusion The administration of G-CSF via either subcutaneous injection or intrauterine infusion and during both the fresh and frozen embryo transfer cycles for RIF patients can improve the clinical pregnancy rate. However, whether G-CSF is effective in improving livebirth rates of RIF patients is still uncertain, continued research on the utilization and effectiveness of G-CSF is recommended before G-CSF can be considered mainstream treatment for RIF patients.

Keywords: Granulocyte colony-stimulating factor, Repeated implantation failure, Intrauterine infusion, Subcutaneous injection, Fresh embryo transfer, Frozen embryo transfer, Clinical pregnancy rate, meta-analysis Background After more than 4 decades of development, assisted reproductive technology (ART) has become one of the main ways to treat infertility. However, recurrent implantation failure has consistently been a bottleneck preventing infertile patients from achieving a better pregnancy during in vitro fertilization and embryo transplantation (IVF-ET). RIF usually refers to a history of embryo transplantation failure following several ARTbased procedures. A positive blood HCG test indicates the initial implantation of the embryo [1], and the absence of an intrauterine gestational sac on ultrasound examination is defined as embryo implantation failure [2]. An international consensus on the definition of embryo implantation failure has yet to be reached, however. Many researchers consider RIF patients to be women under the age of 40 years who have failed to achieve a clinical pregnancy after the transfer of at least four good-quality embryos in a minimum of three IVF fresh or frozen cycles [3]. One cause of RIF is maternal factors, such as immune rejection and problems with endometrial receptivity and the intrauterine environment. Another cause of RIF is embryonic factors, such as poor developmental potential and chromosome aneuploidy [4]. Many treatments for these problems have been found, such as hysteroscopy and preimplantation genetic diagnosis (PGT). However, the way to address the immune problem remains unclear. In recent years, some research has focused on improving endometrial receptivity by using G-CSF to address the self-immune problem.
Granulocyte colony stimulating factor (G-CSF) is a glycoprotein secreted by endothelial cells, macrophages and other immune cells that functions as a growth factor and cytokine. The functional sites of G-CSF are widespread in the human body; some studies have reported [5][6][7] that the receptor for G-CSF can be found not only on stromal cells, endothelial cells, bone marrow cells, fibroblasts, monocytes and macrophages but also on cells in the reproductive system, such as placental cytotrophoblasts, syncytiotrophoblasts, decidual stromal cells, endometrial glandular cells and follicular cells. G-CSF is known for its specific effects on the activation of intracellular signaling pathways that are associated with cell proliferation and differentiation and was first used for patients with myelosuppression and severe neutropenia [8,9]. Recently, researchers found that G-CSF induces trophoblast proliferation, invasion and maintenance during pregnancy. Other researchers found that it improved endometrial receptivity for patients with RIF by promoting endometrial vascular remodeling, embryo adhesion and invasion and regulating endometrial immunity. It can also inhibit apoptosis to maintain endometrial growth [10]. G-CSF also plays an essential role in embryo implantation by regulating the expression of genes associated with embryo adhesion, cell migration, tissue remodeling and angiogenesis, which are essential for endometrial growth, successful embryo implantation and further formation of the placenta [11]. In addition, G-CSF might be involved in the induction of adaptive changes that favor immune tolerance in pregnancy; due to the semi allogenic nature of the fetus, pregnancy represents an immune challenge to the mother. G-CSF switches the T-cell cytokine secretion profile to Th2 responses and promotes IL-10-producing regulatory T-cell and tolerogenic dendritic cell differentiation [12], which are important parts of the immunoregulatory events during the implantation period [6].
The first use of G-CSF in ART was reported by Gleicher et al. [13], who revealed that four patients with thin endometrium achieved clinical pregnancy after G-CSF treatment in their study. Following verification of the safety of G-CSF administration in infertile patients [10], an increasing number of researchers began to explore the efficacy of G-CSF on embryo transfer outcomes for RIF patients; however, the conclusions among those studies are inconsistent. Many RCTs report an improvement in the clinical pregnancy rate of RIF patients by using G-CSF over that of the control group [14][15][16]. However, neither the clinical pregnancy rate nor the live birth rate increased in an RCT including 157 RIF patients reported by Kalem [17]. Recently, in a metaanalysis that included 1253 infertile women, Kamath et al. reported no difference between the G-CSF group and the control group [18]. However, their target population was all infertile women undergoing ART instead of only RIF patients. Another outstanding issue is the kind of embryo transfer cycle that better fits G-CSF treatment. Some studies treated RIF patients with G-CSF in the fresh embryo transfer cycle, while others treated them in the frozen cycle. Furthermore, the best administration route for G-CSF remains uncertain. In this systematic review and meta-analysis, in addition to the clinical pregnancy rate, biochemical pregnancy rate, miscarriage rate and live birth rate, subgroup analysis was conducted on patients treated during different transplantation cycles (fresh and frozen) and with administration routes (subcutaneous injection and intrauterine perfusion) to obtain further information on the influence of G-CSF on RIF patients.

Methods
The PubMed, Embase and Cochrane Central Register of Controlled Trials (CENTRAL) databases were searched for literature published before October 2020. The protocol of this systematic review was registered with INPL ASY (registration number INPLASY202170040). The obtained studies were screened in accordance with the PICOS rule (participants, intervention, control, outcome and study type). Participants: Patients with RIF; intervention group: G-CSF; control group: placebo or no treatment; main outcome: clinical pregnancy rate; study type: RCT. The search strategy for PubMed is as follows:

Selection of articles
A study was included when it met all of the following criteria: 1. Patients under the age of 40 years who failed to achieve clinical pregnancy after the transfer of at least four good-quality embryos in a minimum of two IVF fresh or frozen cycles. 2. Subcutaneous injection or intrauterine infusion of G-CSF before a fresh or frozen embryo transfer cycle. 3. RCT with a control group that received placebo or no treatment. 4. Reporting of one of the following outcomes: clinical pregnancy rate, live birth rate, miscarriage rate, biochemical pregnancy rate. 5. Patients in RCT with uterine factors such as uterine anomalies, myoma, endometrial polyps were excluded. 6. Patients in RCT with systemic diseases and contraindications of G-CSF were excluded. Studies that did not meet all the criteria above were excluded.
Data extraction was performed by two review authors (Hou and Jiang) independently, and disagreement was resolved by discussion or decided by a third review author (Meng). Randomized controlled trials were assessed for risk of bias using the Cochrane 'Risk of bias' tool according to the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Version 5.1.0). The following domains were assessed by two researchers (Hou and Jiang) independently: 1.Selection bias (random sequence generation and allocation concealment);2. Performance bias (blinding of participants and personnel);3. Detection bias (blinding of outcome assessors);4. Attrition bias (incomplete outcome data);5. Reporting bias (selective reporting);6. Other bias (including unplanned interim analysis).
The evaluations were categorized as having 'low risk', 'unclear risk' or 'high risk' of bias; a detailed list is shown in Table 1.

Data analysis
Stata software (Version 15.0; Stata Corporation, College Station, TX, USA) was employed to perform all data analysis. The influence of G-CSF treatment on the outcomes of IVF-ET for RIF patients was assessed with pooled risk ratios (RRs) and their 95% confidence intervals (CIs). The pooled RRs were calculated through a Mantel-Haenszel fixed-effects model if there was no heterogeneity; otherwise, a random-effects model was adopted. Statistical heterogeneity across studies was formally tested using Cochran's Q test. The I 2 statistic was examined, and I 2 > 50% was considered to represent significant heterogeneity between studies. Subgroup analysis was performed in terms of embryo transplantation cycle and administration route. We did not assess publication bias across studies since the number of included studies was under 10.

Study selection
A total of 684 studies were obtained by primary search: 230 from PubMed, 393 from Embase, and 61 from CENTRAL. After a thorough screening, 10 RCTs [15][16][17][19][20][21][22][23][24] published between 2015 and 2020 were finally included. The specific steps of the study selection are shown in Fig. 1. A total of 976 patients with RIF were included, including 462 in the study group with G-CSF treatment and 524 in the control group with placebo or no treatment. There were four valid outcome indicators in the extracted data, including the clinical pregnancy rate, live birth rate, biochemical pregnancy rate, and miscarriage rate. Basic characteristics of all included studies were extracted, including author, year of publication, country of study, study period, administration route, transplantation cycle, method of intervention, definition of RIF and outcome indicators. Details of the extracted characteristics are shown in Table 2.

Biochemical pregnancy rate
Four studies reported the biochemical pregnancy rate and included 469 RIF patients, 228 of whom were in the study group and 241 of whom were in the control group. A fixed-effects model was used for data synthesis, and the results indicated that G-CSF treatment improved the biochemical pregnancy rate for RIF patients (RR: 1.85, 95% CI: 1.28-2.68; I 2 = 20.1%); see Fig. 4 for details.

Miscarriage rate
Four studies reported miscarriage rate and included 472 RIF patients, 206 of whom were in the study group and 266 of whom were in the control group. A random-effects model was used for data synthesis, and no significant difference was found between the study and control groups (RR: 1.13, 95% CI: 0.25-5.21; I2 = 63.2%); see Fig. 5 for details. An I2 of 63.2% was observed in this synthesis, indicating considerable heterogeneity, which might have been caused by the small number of included patients.

Live birth rate
Three studies reported the live birth rate and included 372 RIF patients, 166 of whom were in the study group Table 1 Risk of bias assessment. *①Random sequence generation (selection bias); ②Allocation concealment (selection bias); ③Blinding of participants and personal (performance bias); ④Blinding of outcome assessment (detection bias); ⑤Incomplete outcome data (attrition bias);⑥Selection reporting (reporting bias); ⑦Other bias

Discussion
In addition to the proliferation and differentiation of granulocytes, G-CSF has a clear immunomodulatory effect. G-CSF receptors can be found throughout the reproductive system, especially in placental cytotrophoblasts, syncytiotrophoblasts, decidual stromal cells, endometrial glandular cells and follicular cells, which provide the foundation for the establishment and maintenance of pregnancy. Most studies so far have reported an improvement in the clinical pregnancy rate of RIF patients by G-CSF treatment, which coincides with our results. The reasons for this improvement may include induction of local immune regulation of the endometrium, embryo adhesion and implantation, proliferation of trophoblasts and endometrial vascular remodeling by G-CSF. As we mentioned above, the using of G-CSF improved clinical pregnancy rate of RIF patients, other studies proved that G-CSF may play an essential role before and after the establishment of pregnancy. However, few studies have compared the effect of G-CSF on different embryo transfer cycles. Both fresh and frozen embryo transfer cycles are widely used in IVF-ET. However, the endometrial environment is not identical between these two cycles. Previous studies found that 15% of women received autoantibodies after controlled ovarian hyperstimulation (COH), which can lead to reproductive disorders such as endometriosis, recurrent abortion, and premature ovarian failure [25]. It is known that G-CSF might be involved in the induction of adaptive changes that favor immune tolerance in pregnancy, which might be an important part of the immunoregulatory events during the implantation period. In this review and metaanalysis, we performed a subgroup analysis in terms of whether the fresh or frozen cycle was used to compare the effect of G-CSF as an immune factor on patients with RIF between the two subgroups. The results indicate an increase in the pregnancy rates for both cycles. It suggested that G-CSF can be used in the fresh or frozen cycle without distinction. However, more high-quality studies with larger populations are still needed to form clinical guidelines. Biochemical pregnancy tends to occur when the endometrial environment cannot maintain the continuation of the established pregnancy. Many previous studies [16,23] reported an increased biochemical pregnancy rate of RIF patients after G-CSF treatment, which coincides with our results. These authors suggest that this increase in the biochemical pregnancy rate might be related to a decrease in the sustaining effect of G-CSF. G-CSF was administered at a single dose before embryo transfer in most studies included in this review and meta-analysis. This protein improves endometrial receptivity and thus the embryo implantation rate. However, with the decrease in the sustaining effects of G-CSF, some RIF patients with poor endometrial receptivity are unable to maintain pregnancy, leading to the occurrence of biochemical pregnancy. If this is true, does continuous administration of G-CSF help RIF patients maintain their hard-won pregnancy? More studies are needed to address this hypothesis.
The routine G-CSF administration routes in ART include subcutaneous injection and intrauterine infusion.  However, whether one route is better than the other remains obscure. Zeyneloglu et al. reported a better embryo transfer outcome by combining subcutaneous and intrauterine G-CSF treatment for RIF patients than by subcutaneous injection alone [14]. In this review and meta-analysis, we performed a subgroup analysis by subcutaneous injection and intrauterine infusion to compare the influence of G-CSF on RIF patients. The result indicates an increase in the pregnancy rates for both routes. The G-CSF receptor can be found throughout the human body, and the systematic use of G-CSF assuredly induces more side effects than its topical use. Whether these side effects harm the pregnancy is unknown. Intrauterine infusion of G-CSF will have a direct effect on the endometrium, which might be easier and safer for the patient.
As mentioned above, G-CSF might be beneficial for the establishment and maintenance of pregnancy. The results of this review and meta-analysis show no significant effect of G-CSF treatment on the miscarriage or live birth rate. Due to the limited number of included patients, the exact effect of G-CSF on these rates. Miscarriage for RIF patients remains unclear, and more highquality, larger and multicenter RCTs are needed to provide more authoritative evidence for clinical practice.

Conclusion
In this review and meta-analysis, the results of data synthesis for 976 RIF patients indicate that G-CSF treatment improved the clinical pregnancy rate; however, the biochemical pregnancy rate increased as well, which is coincident with previous studies. In addition, subgroup analysis indicated that the administration of G-CSF by either subcutaneous injection or intrauterine infusion during both fresh and frozen embryo transfer cycles for RIF patients improved the clinical pregnancy rate. The influence of G-CSF on the miscarriage and live birth rates of RIF patients is unclear, and more studies are needed before G-CSF become mainstream treatment for RIF patients.