Network meta-analysis of four common immunomodulatory therapies for the treatment of patients with thin endometrium

Abstract

Obejective

To compare the effectiveness of endometrial receptivity and pregnancy outcomes of four common immunomodulatory therapies for patients with thin endometrium.

Method

This systematic review and network meta-analysis using a literature search up to January 2024, to identify relevant trials comparing endometrial receptivity and pregnancy outcomes of human chorionic gonadotropin (hCG), platelet-rich plasma (PRP), infusion of granulocyte colony-stimulating factor (IG-CSF), and peripheral blood mononuclear cell (PBMC) for patients with thin endometrium. We used surface under the cumulative ranking (SUCRA) to ranked four common immunomodulatory therapies on endometrium thickness, implantation rate (IR), clinical pregnancy rate (CPR), and live birth rate (LBR). RoB2 and ROBINS-I were used to assess the certainty of evidence.

Results

The pooled results of 22 studies showed that hCG (mean difference [MD]: 3.05, 95% confidence interval [CI]: 1.46–4.64) and PRP (MD: 0.98, 95% CI: 0.20–1.76) significantly increase endometrium thickness. The hCG was the best among the IG-CSF (MD = −2.56, 95% CI = −4.30 to −0.82), PBMC (MD = −2.75, 95% CI = −5.49 to −0.01), and PRP (MD = −2.07, 95% CI = −3.84 to −0.30) in increasing endometrium thickness. However, IG-CSF and PRP significantly improved IR (IG-CSF: risk ratio (RR; IG-CSF: RR = 1.33, 95% CI = 1.06–1.67; PRP: RR = 1.63, 95% CI = 1.19–2.23), and LBR (IG-CSF: RR = 1.53, 95% CI = 1.16–2.02; PRP: RR = 1.59, 95% CI = 1.08–2.36).

Conclusions

Available evidence reveals that hCG and subcutaneous or intrauterine CSF (SG-CSF) may be the best treatment options for current thin endometrium patients. However, future high-quality and large-scale studies are necessary to validate our findings.

Keywords:

• Thin endometrium
• immunomodulatory therapy
• granulocyte colony-stimulating factor
• platelet-rich plasma
• network meta-analysis

Introduction

In vitro fertilization (IVF) and embryo transfer are the most effective assisted reproductive technologies (ART) for treating infertility [1]. Successful embryo implantation depends on perfect synchrony between embryonic development and good endometrial receptivity [2]. The ability of the endometrium to allow normal implantation is known as receptivity, which is the specific period when ‘the trophectoderm of the blastocyst can attach to the endometrial epithelium and subsequently invade the endometrial stroma and vasculature’ [1], known as ‘the implantation window’ opens 4–5 d after ovulation. The success rate of pregnancies was shown to be primarily influenced by endometrial factors (approximately 60%) and embryonic factors (about 40%) [3]. Therefore, increasing endometrial receptivity has become a critical concern of reproduction research [4].

The major indicator of endometrial receptivity is endometrial thickness, and a successful pregnancy greatly depends on an adequate endometrial thickness [5]. Many reasons, such as inflammation and iatrogenic factors, can lead to thinning of the endometrium, which is called thin endometrium and considered an ongoing challenge to female infertility and conception failure [6]. Thin endometrium occurs in 1.5%–9.1% of cases despite attempts at various treatments to maintain endometrial thickness [7–9]. Although the condition is uncommon, clinicians continue to work to find ways to treat thin endometrium and improve pregnancy outcomes for infertile patients [10], such as extending estradiol administration, using low-dose aspirin, administering vitamin E, L-arginine, and sildenafil citrate in intravaginal treatment [11,12]. However, despit the use of these treatments, a small percentage of patients remain unresponsive. Therefore, better treatment strategies are needed [13,14]. Accumulating evidence suggests that active involvement of local immune cells at the implantation site impairs endometrium receptivity [15]. Therefore, various immunomodulatory therapies have been investigated, including intrauterine human chorionic gonadotropin (hCG) [16,17] subcutaneous (SG) or intrauterine (IG) infusion of granulocyte colony-stimulating factor (CSF)) [4,18], peripheral blood mononuclear cells (PBMCs) [19], and autologous platelet-rich plasma (PRP) [20,21]. However, which of these immunomodulatory therapies work best remains controversial. Therefore, this network meta-analysis aims to provide an evidence base for clinical application by comparing the effectiveness of four common immunomodulatory therapies to patients with thin endometrium.

Material and methods

Study design

We conducted this study followed Cochrane Handbook [22] and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) for network meta-analysis (PRISMA-NMA) [23]. Ethics approval was not required, since data was not collected directly from patients.

Selection criteria

Studies were selected followed criteria as follows:

1. Studies targeted on participants clinically diagnosed with thin endometrium (<7 mm);

2. Intervention groups received immunomodulatory monotherapy or combination therapy;

3. Control groups received any other anti-thin endometrium treatment;

4. Study endpoints included post-treatment endometrial thickness or any pregnancy outcomes (IR, CPR, and LBR).

5. Studies with any comparative designs (i.e. randomized controlled trials [RCTs], prospective cohort studies [PC], or retrospective cohort studies [RPC]);

The exclusion criteria were

1. Animal/cell researches were excluded;

2. Studies not published in an international peer-reviewed journal;

3. Excluded studies with faulty randomized design or study methodology.

Literature retrieval

We retrieved studies published before January, 2024 by searching PubMed, Web of Science, China National Knowledge Infrastructure (CNKI), Embase, and Cochrane Library. Search strategies were summarized in Supplementary Table 1. We used the following terms and other relevant key words to develop search strategy, including ‘thin endometrium,’ ‘human chorionic gonadotropin,’ ‘granulocyte colony-stimulating factor,’ ‘peripheral blood mononuclear cells,’ and ‘ platelet-rich plasma.’ We manually examined references of eligible articles and meta-analyses to identify additional studies. The literature retrieval and selection process was performed independently by two investigators. Any discrepancy was solved by discussion.

Data extraction

Two authors independently selected eligible studies by screening titles, abstracts and full texts. Then, one author (Lifei Li) extracted data, and another (Zhijian Kou) examined the data. The following information was obtained from eligible studies: (1) the first author’s name, published year, country, and study design; (2) sample size, average age, body mass index (BMI), infertility duration, number of embryos transfer, endometrium thickness before treatment, and details of interventions; (3) outcomes of interest; (4) details of risk of bias. For outcomes reported in median and standardized error or interquartile range (IQR), we used recognized formulas to calculate the required data [24].

Outcome measures

Post-treatment endometrial thickness was measured as the maximum distance between the 2 interfaces of the endometrial-myometrial junction, in the midsagittal plane of the uterus by B ultrasound radiography. Pregnancy outcomes included IR, CPR, and LBR. IR was determined by the number of live embryos, CPR was considered with the presence of a gestational sac containing a yolk sac at transvaginal ultrasonography, and LBR was classified as those cycles resulting in the delivery of a live infant after 24 weeks gestation.

Evidence structure

An evidence map for each outcome was drawn to display evidence pattern. Direct comparison between two immunomodulatory therapies was represented by a solid line, and the thickness of line was weighted by the number of direct comparisons. Solid circle represented an immunomodulatory therapy, and size of the circle was weighted by the number of accumulated participants.

Risk of bias assessment

Two authors (Lifei Li and Fei Zhao) used the revised Cochrane risk-of-bias tool (RoB2) [25] to evaluate the risk of bias of RCTs and the Cochrane risk of bias in non-randomized studies of interventions (ROBINS-I) tool [26] to evaluate the risk of bias of PC and RPC studies. In the RoB2 tool, five domains were involved: (a) randomization, (b) incomplete data, (c) deviation from expected intervention, (d) selected reported results, and (e) outcome measurement bias, and each entry was divided into high risk, low risk, and other concerns. In the ROBINS-I tool, seven domains were involved, including confounding, selection of participants, classification of intervention, deviations from intended interventions, missing data, outcomes measurements, and selection of the reported results. Individual domain was classified as ‘low,’ ‘moderate,’ ‘serious,’ or ‘critical.’ Results of risk of bias assessment were graphically presented using ‘robvis’ package [27].

Statistical analysis

Mean difference (MD) and risk ratio (RR), with 95% confidence interval (CI), were used to report pooled results. First, we conducted transitivity assessment. Second, we used the design-by-treatment interaction method [28] and the node-splitting method [29] to examine global and local inconsistencies. Third, random network meta-analysis was conducted to assess the relevant efficacy of various immunomodulatory therapies. Fourth, we examined loop inconsistency to evaluate the validity of pooled results. Fifth, we ranked four common immunomodulatory therapies based on the surface under the cumulative ranking (SUCRA) [30]. Finally, we checked publication bias with Begg’s [31] and Egger’s [32] tests. The present network meta-analysis was done with STATA version 14.0 (StataCorp LP, College Station, TX).

Results

Literature selection

Initial search yielded 782 records. After removed 214 duplicate records, and 546 ineligible studies, 22 studies were rated for the final evaluation. Finally, 22 studies [1,4,10,13,14,16–21,33–43] were all eligible for inclusion. The screening process and the reasons for exlucidng studies are depicted in Figure 1.

Figure 1. PRISMA flowchart of study selection. WoS: web of science; CNKI: China National Knowledge Infrastructure.

Study characteristics

The baseline characteristics of included studies are presented in Table 1. The publication time was between 2014 and 2022. 63.6% of included studies [1,4,10,13,14,16,17,19,33,34,39–41,43] were performed in China, four studies [21,36,37,42] in Iran, two studies [20,35] in Russia, and other two studies in Poland [38] and India [18]. Eight studies [1,16,21,34,36,40,42,43] were RCTs, and other 14 studies [4,10,13,14,17–20,33,35,37–39,41] were PC and RPC studies. A total of 2883 participants were accumulated, with five therapeutic modalities (i.e. hCG, IG-CSF, SG-CSF, PBMC, and PRP). The evidence structure of each outcome is shown in Supplementary Figure 1.

Table 1. Basic characteristics of included studies (n = 22).

Study	Country	Study design	Group	Sample size	Average age, years	BMI, kg/m^2	Years of infertility	No. of transferred embryos	ET before, mm	Details of intervention
Wang [16]	China	RCT	Control	39	31.3 ± 3.5	21.7 ± 0.5	4.85 ± 0.83	n.a.	4.24 ± 0.39	No intrauterine perfusion
			hCG	39	31.5 ± 3.4	21.7 ± 0.5	4.81 ± 0.75	n.a.	4.20 ± 0.27	Intrauterine perfusion with hCG 500 U
Zhang et al. [17]	China	RPC	Control	50	31.7 ± 3.6	23.1 ± 4.3	3.00 ± 1.18	1.76 ± 0.48	6.37 ± 0.65	No intrauterine perfusion
			hCG	55	32.2 ± 3.2	22.8 ± 3.2	2.80 ± 1.06	1.84 ± 0.42	6.41 ± 0.54	Intrauterine perfusion with hCG 500 U
Dzhincharadze et al. [35]	Russia	RPC	Control	17	36.8 ± 6.8	23.7 ± 3.5	2.5 ± 1.66	n.a.	<7	No intrauterine injection
			IG-CSF	15	37.7 ± 6.6	26.1 ± 6.5	4.3 ± 1.8	n.a.	<7	Intrauterine injection of G-CSF 300 mcg/mL
Eftekhar et al. [37]	Iran	non-RCT	Control	34	28.6 ± 5.2	24.9 ± 1.6	n.a.	2.39 ± 0.49	5.76 ± 0.86	No intrauterine perfusion
			IG-CSF	34	30.8 ± 4.6	25.3 ± 1.5	n.a.	2.46 ± 0.63	5.63 ± 0.78	Intrauterine injectionl with G-CSF 300 mcg/ml
Jiang et al. [4]	China	RPC	Control	41	31.5 ± 4.3	22.7 ± 3.2	4.3 ± 3.4	1.89 ± 0.31	4.35 ± 1.13	No intrauterine injection
			IG-CSF	25	32.7 ± 3.5	23.2 ± 3.0	5.2 ± 2.5	1.84 ± 0.37	4.16 ± 0.80	Intrauterine injection with G-CSF 300 mcg/mL
Kunicki et al. [38]	Poland	RPC	Control	33	37.3 ± 3.7	20.9 ± 0.9	n.a.	n.a.	6.4 ± 1.1	No intrauterine injection
			IG-CSF	29	38.6 ± 6.8	23.5 ± 0.39	n.a.	n.a.	6.5 ± 0.96	Intrauterine injection with G-CSF 300 mcg/mL
Li et al. [39]	China	RPC	Control	25	3.19 ± 3.7	20.9 ± 2.9	5.7 ± 4.0	2.56 ± 0.55	6.42 ± 0.91	No intrauterine infusion
			IG-CSF	34	30.9 ± 3.5	20.8 ± 2.9	4.3 ± 2.9	2.48 ± 0.57	6.51 ± 0.69	Intrauterine infusion with G-CSF 100 μg/0.6 mL
Lian et al. [10]	China	RPC	Control	154	33.8 ± 3.9	21.0 ± 2.3	4.0 ± 2.6	1.90 ± 0.63	6.72 ± 0.61	No intrauterine infusion
			IG-CSF	117	34.5 ± 3.9	21.2 ± 2.6	3.4 ± 2.4	1.86 ± 0.61	6.13 ± 0.94	Intrauterine infusion with G-CSF 100 μg/0.4 mL
Mao et al. [1]	China	RCT	Control	143	37.0 ± 5.0	21.7 ± 2.8	5.4 ± 2.6	n.a.	6.30 ± 0.64	No intrauterine infusion
			IG-CSF	161	37.8 ± 5.4	22.0 ± 2.8	5.0 ± 3.3	n.a.	6.14 ± 0.92	Intrauterine infusion with G-CSF
Sarvi et al. [42]	Iran	RCT	Control	15	31.2 ± 3.2	26.6 ± 3.1	n.a.	2.3 ± 0.33	4.1 ± 1.8	Intrauterine infusion with saline 1 mL
			IG-CSF	13	31.6 ± 3.8	26.2 ± 3.5	n.a.	3.03 ± 0.5	4.2 ± 0.6	Intrauterine infusion with G-CSF 300 mcg/mL
Xu et al. [13]	China	RPC	Control	52	32.0 ± 3.9	21.5 ± 3.0	5.5 ± 3.8	2.08 ± 0.33	6.5 ± 0.5	No intrauterine perfusion
			IG-CSF	27	31.4 ± 4.0	21.8 ± 2.5	4.6 ± 2.9	2.00 ± 0.00	5.7 ± 0.7	Intrauterine infusion with G-CSF 100 μg/0.6 mL
Zhang et al. [17]	China	RCT	Control	115	32.2 ± 4.1	n.a.	n.a.	n.a.	5.68 ± 2.22	No intrauterine perfusion
			IG-CSF	114	31.7 ± 4.3	n.a.	n.a.	n.a.	5.50 ± 1.96	Intrauterine injection with G-CSF 300 mcg/1.8 mL
Banerjee et al. [18]	India	RPC	Control	44	31.8 ± 0.60	21 ± 0.8	5 ± 4.5	n.a.	<7	No subcutaneous infusion
			SG-CSF	44	31.6 ± 0.60	20.8 ± 0.8	4.3 ± 4.5	n.a.	<7	Subcutaneous infusion with G-CSF 300 mcg/mL
Li et al. [19]	China	RPC	Control	339	30.5 ± 4.1	n.a.	4.6 ± 3.3	2.18 ± 0.38	<7	No subcutaneous infusion
			PBMC	294	30.8 ± 4.1	n.a.	4.7 ± 3.3	2.07 ± 0.47	<7	Subcutaneous infusion with G-CSF 300 mcg/mL
Chang et al. [33]	China	RPC	Control	41	32.8 ± 1.7	22.4 ± 0.4	3.6 ± 1.6	2	<7	No intrauterine infusion
			PRP	53	33.5 ± 1.5	22.4 ± 0.6	3.6 ± 1.8	2	<7	Intrauterine infusion with PRP 0.5 mL
Chang et al. [14]	China	PC	Control	30	32.6 ± 1.7	22.4 ± 0.8	3.7 ± 1.7	2	6.39 ± 0.72	No intrauterine infusion
			PRP	34	34.8 ± 0.8	22.4 ± 0.4	3.6 ± 1.8	2	6.32 ± 0.54	Intrauterine infusion with PRP 0.5 mL
Cheng et al. [34]	China	RCT	Control	46	32.2 ± 2.1	n.a.	4.3 ± 3.5	2.1 ± 0.62	61.2 ± 0.39	No intrauterine infusion
			PRP	46	31.4 ± 2.4	n.a.	4.3 ± 3.4	2.0 ± 0.73	6.06 ± 0.49	Intrauterine infusion with PRP 0.5 mL
Dzhincharadze et al. [20]	Russia	RPC	Control	17	36.8 ± 6.8	23.7 ± 3.5	6.7 ± 5.7	1	<7	No intrauterine infusion
			PRP	37	36.0 ± 6.0	23.1 ± 3.4	5.2 ± 3.8	1	<7	Intrauterine infusion with PRP 5-7 mL
Eftekhar et al. [36]	Iran	RCT	Control	43	32.4 ± 2.6	n.a.	n.a.	1.9 ± 0.6	6.15 ± 0.37	No intrauterine infusion
			PRP	40	32.0 ± 2.3	n.a.	n.a.	2 ± 0.4	6.09 ± 0.47	Intrauterine infusion with PRP 1.5 mL
Nazari et al. [21]	Iran	RCT	Control	30	32.3 ± 4.8	25.5 ± 2.7	n.a.	n.a.	5.06 ± 0.82	No intrauterine infusion
			PRP	30	33.9 ± 2.8	24.3 ± 2.2	n.a.	n.a.	4.92 ± 0.67	Intrauterine infusion with PRP 0.5 mL
Qi et al. [41]	China	RPC	Control	65	31.4 ± 1.4	21.5 ± 0.6	3.2 ± 1.8	1.8 ± 0.5	5.0 ± 0.3	No intrauterine infusion
			PRP	40	32.7 ± 1.4	22.4 ± 0.4	3.1 ± 1.5	1.8 ± 0.7	4.8 ± 0.4	Intrauterine infusion with PRP 0.5 mL
Long et al. [40]	China	RCT	Control	68	33.4 ± 4.4	21.2 ± 2.3	4.3 ± 3.1	1.74 ± 0.45	6.62 ± 0.52	Intrauterine infusion with saline 0.7 mL
			IG-CSF	81	34.0 ± 4.4	22.2 ± 2.7	3.5 ± 2.4	1.80 ± 0.40	6.35 ± 0.63	Intrauterine infusion with G-CSF 300 mcg/mL
			PRP	80	35.4 ± 3.8	21.3 ± 2.5	3.5 ± 2.3	1.88 ± 0.33	6.08 ± 0.89	Intrauterine infusion with PRP 0.8 mL

ET: endometrium thickness; hCG: human chorionic gonadotropin; IG-CSF: intrauterine infusion of granulocyte colony-stimulating factor; SG-GSF: subcutaneous intrauterine; PBMC: peripheralblood mononuclear cell; PRP: platelet-rich plasma; BMI: bodymassindex; RCT: randomizedcontrolledtrial; RPC: retrospectivecohort; PR: prospectivecohort; n.a.: not applicable

Risk of bias of eligible studies

Supplementary Figure 2 shows the results of risk of bias assessment. Overall, all RCTs were rated as low risk or some concerns in overall methodological quality, while almost all PC and RPC studies were rated as a moderate risk in overall methodological quality except for one study, which was rated as a serious risk.

Transitivity assessment

We performed transitivity assessment according to publication year, sample size, average age and BMI of participants, number of embryos transferred, and endometrial thickness before treatment. As summarized in Supplementary Table 2, transitivity assessment confirmed that six key characteristics were evenly distributed between different comparisons (p > .05 for all).

Global and local consistency examination

There is no statistical significance shown in the global consistency test: endometrium thickness after treatment (p = .975), IR (p = .229), CPR (p = .289), and LBR (p = .268) (see Supplementary Figure 3). Comparisons of all outcomes also did not show local inconsistency (see Supplementary Table 3).

Meta-analysis of endometrium thickness

Twenty studies [1,4,10,14,16,17,19–21,33–43] reported endometrium thickness after treatment. As shown in Figures 2(a) and 3(a), compared with control, hCG (MD = 3.05, 95% CI = 1.46–4.64) and PRP (MD = 0.98, 95% CI = 0.20–1.76) significantly increased post-intervention endometrium thickness. hCG was better than IG-CSF (MD = −2.56, 95% CI = −4.30 to −0.82), PBMC (MD = −2.75, 95% CI = −5.49 to −0.01), and PRP (MD = −2.07, 95% CI = −3.84 to −0.30) in increasing endometrium thickness. However, except for the five comparisons mentioned earlier, differences between other comparisons did not get statistical significance.

Figure 2. Relative therapeutic effectiveness of different immunomodulatory therapies in terms of post-intervention endometrium thickness (a), IR (b), CPR (c), and LBR (d). IR: implantation rate; CPR: clinical pregnancy rate; LBR: live birth rate; hCG: human chorionic gonadotropin; IG-CSF: intrauterine infusion of granulocyte colony-stimulating factor; SG-GSF: subcutaneous intrauterine; PBMC: peripheral blood mononuclear cell; PRP: platelet-rich plasma.

Figure 3. League tables of network meta-analysis results (a), IR (b), CPR (c), and LBR (d). IR: implantation rate; CPR: clinical pregnancy rate; LBR: live birth rate; hCG: human chorionic gonadotropin; IG-CSF: intrauterine infusion of granulocyte colony-stimulating factor; SG-GSF: subcutaneous intrauterine; PBMC: peripheral blood mononuclear cell; PRP: platelet-rich plasma.

Meta-analysis of pregnancy outcomes

Twenty studies [1,4,10,13,14,16,17,19,20,33–43] reported IR after treatment. Results showed that, as shown in Figures 2(b) and 3(b), IG-CSF (RR = 1.34, 95% CI = 1.06–1.70) and PRP (RR = 1.74, 95% CI = 1.25–2.42) significantly improved IR as compared with control. However, except for the two comparisons mentioned earlier, there were no significant differences for other comparisons.

All studies [1,4,10,13,14,16–21,33–43] reported CPR after treatment. Results indicated that, as shown in Figures 2(c) and 3(c), the administration of IG-CSF (RR = 1.33, 95% CI = 1.06–1.67) and PRP (RR = 1.63, 95% CI = 1.19–2.23) significantly improved CPR when compared to control, while differences of other comparisons did not reach statistical significance.

LBR was available in 14 studies [1,4,13,14,17,19,20,35,36,38–41,43]. As shown in Figures 2(d) and 3(d), compared with control, only IG-CSF (RR = 1.53, 95% CI = 1.16–2.02) and PRP (RR = 1.59, 95% CI = 1.08–2.36) led to a significant higher LBR. However, no significant differences were detected for other comparisons.

The ranking probability of SUCRA

We drew Figure 4 to depict probability rankings for each therapeutic modality. Regarding the endometrium thickness after treatment, hCG ranked first (99.0%), followed by PRP (63.7%), IG-CSF (41.2%), and PBMC (33.7%). For IR, SG-CSF obtained the highest probability, followed by PRP (76.9%), hCG (67.5%), IG-CSF (47.0%), and PBMC (18.9%). As far as the CPR was concerned, SG-CSF was the most effective therapeutic modality (81.8%), followed by PRP (72.9%), hCG (56.2%), IG-CSF (45.8%), and PBMC (36.0%). Finally, regarding the LBR, the most effective therapeutic modality was hCG (79.8%), followed by PRP (71.9%), IG-CSF (66.6%), and PBMC (19.9%).

Figure 4. SUCRA results of different immunomodulatory therapies in terms of post-intervention endometrium thickness (a), IR (b), CPR (c), and LBR (d). IR, implantation rate; CPR, clinical pregnancy rate; LBR, live birth rate; hCG, human chorionic gonadotropin; IG-CSF, intrauterine infusion of granulocyte colony-stimulating factor; SG-GSF, subcutaneous intrauterine; PBMC, peripheral blood mononuclear cell; PRP, platelet-rich plasma.

Heterogeneity assessment

I² value ranged from 0.0% to 98.9% for the endometrium thickness after treatment, I² value ranged from 0.0% to 59.7% for IR, I² value ranged from 0.0% to 43.4% for CPR, and I² value ranged from 0.0% to 8.3% for LBR. Details of heterogeneity examination are depicted in Supplementary Figure 4.

Loop inconsistency

As depicted in Supplementary Figure 5, the lower limits of all 95% CIs were zero, suggesting no inconsistency across all available closed loops for these results.

Publication bias

Results of Begg’s and Egger’s tests showed a risk of publication bias in endometrium thickness after treatment (Begg: p = .007, Egger: p = .00), IR (Begg: p = .042, Egger: p = .001), CPR (Begg: p = .008, Egger: p = .002), and LBR (Begg: p = .392, Egger: p = .029).

Discussion

Main findings

This network meta-analysis included 22 eligible studies with 2883 participants. Results revealed that hCG and PRP significantly increased the endometrium thickness, and hCG was superior to IG-CSF, PBMC, and PRP in terms of endometrium thickness. IG-CSF and PRP led to significantly higher IR, CPR, and LBR than the control group. SUCRA results revealed that hCG was probably the optimal modality in increasing endometrium thickness and LBR, SG-CSF might be the most effective option in improving IR and CPR, followed by PRP and hCG.

Interpretations of findings

Previous studies revealed immunomodulatory therapy benefited ART patients [44–46]. Xie et al. [44] reported that compared with the control group, IR could be greatly enhanced by IG-CSF, and Liu et al. [45] showed that IG-CSF could significantly increase IR, but SG-CSF was most effective for CPR. Zhao et al. [46] found that a positive clinical effect may be found in pregnancy outcomes after the administration of G-CSF, and a SG injection may be the best way to administer the G-CSF. G-CSF is a glycoprotein that belongs to the growth factor family. The results of a meta-analysis published in 2021 showed that G-CSF, as a common endogenous hematopoietic growth factor, can improve clinical pregnancy rates (CPRs) in patients with RIF, which supports our findings [47]. G-CSF secretion and receptor expression are present in many reproduction-related cells (endometrial epithelial cells, trophoblast cells, and placental cells, etc.) [48]. The positive effects of G-CSF on immune environment and trophoblast proliferation may mediate the improvement of CPR and implantation rate (IR). In vitro and in vivo experiments have shown that G-CSFR expression by CD4 + and CD8 + T cells induced by G-CSF can promote the production of interleukin-10 (IL-10) and transforming growth factor-β(TGF-β) regulatory T cells (Tregs). While inhibiting Th1 differentiation, they polarize toward Th2 differentiation, thereby reducing Th1/Th2 and Th17/Treg ratios and promoting embryo implantation [49–51]). Meanwhile, G-CSF can regulate the increase of endometrium and affected the growth of early endometriosis [52], and can also improve endometrial stem cells, activate bone marrow stem cells, and promote the development of endometrial cells [13]. However, there remains controversy regarding the ideal route of G-CSF administration, and the reasons for the different effects of the routes have not yet been fully clarified. So, performing more high-quality studies to clarify these issues is imperative. In addition, studies have demonstrated that endometrial thickness is very important for successful embryo implantation. Based on the effects of SG-CSF on IR and CPR, it is reasonable to speculate that SG-CSF may be also best regimen for increasing endometrium thickness and LBR. However, currently available studies have not evaluated the role of SG-CSF in both two outcomes; therefore, this open question should be further investigated.

The current network meta-analysis also found that hCG may be the best therapeutic option for increasing endometrium thickness and improving LBR. As a primary embryonic signal, hCG is crucial to initiating and maintaining a pregnancy [53]. To be specific, several endometrial paracrine parameters, which correlate to endometrial differentiation, angiogenesis, implantation, and tissue remodeling, can be modulated by IG perfusion of hCG during the secretory phase [54], thus promoting endometrial repairs and thickening [55]. Previous network meta-analyses and meta-analysis results have also shown that hCG infusion significantly improves CPRs in patients with recurrent implantation failure (RIF) [56] and significantly increases live birth rates (LBRs) [57]. However, the majority of embryos transferred into a woman’s uterine cavity through IVF are eight cell stage or longer, and such embryos are unable to communicate with the mother before implantation, which is compensated for by hCG infusion before embryo transfer. Some studies have also shown that hCG infusion can inhibit the expression of insulin-like growth factor binding protein-1 (IGFBP-1) and macrophage colony-stimulating factor (M-CSF) by activating LIF, VEGF, MMP-9, and other factors to further improving endometrial, receptivity [58]. However, it has also been suggested that the duration of pre-implantation hCG infusion needs to be controlled to prevent prolonged exposure of the endometrium to HCG, resulting in loss of sensitivity to the blastocyst secretion of hCG [59,60]. Additionally, hCG works on the endometrium and endometrial receptors (such as hCG/luteinizing hormone [hCG/LH] receptors and IL-1 receptor) to increase local angiogenesis at the embryo implantation site on the endometrium, promote embryo implantation, and reduce early embryo loss, thereby increasing LBR [55]. Nevertheless, owing to the lack of eligible studies, we cannot analyze the effectiveness of SG-CFS on the endometrium thickness and LBR, so this study does not answer whether hCG is also better than SG-CSF, additional high-quality studies are required to verify the benefits of SG-CSF on these two outcomes.

Heterogeneity explanation

Significant statistical heterogeneity was detected for the comparison of hCG vs. control (98.9%), IG-CSF vs. control (78.7%), and PRP vs. control (91.3%) in endometrium thickness. Moreover, we also detected statistical heterogeneity in the comparison of IG-CSF vs. control (59.7%) in the meta-analysis of IR. We are not surprised at these results due to the variations in the study design, years of infertility, and the endometrium thickness before treatment across studies included in each comparison. For example, the RCT by Wang [16] reported a significant increase in the endometrium thickness of hCG (MD = 5.80, 95% CI = 4.77–6.82), but the retrospective cohort (RPC) by Zhang et al. [17] reported that hCG did not significantly increase the endometrium thickness (MD = 0.38, 95% CI = −0.01–0.77). In addition, we also recognized that most eligible studies included extremely limited sample sizes except for four studies which included more than 100 patients in each group. Studies with small sample sizes will yield more biased results than those with adequate sample sizes, inevitably introducing heterogeneity to the pooled results. In addition, as a primary outcome of interest, the measurement of endometrial thickness will be affected by operator and technology. Nevertheless, we cannot perform subgroup analysis to definitively explore which factors may be the major contributor to the significant statistical heterogeneity due to the limited number of eligible studies. So, our findings should be interpreted with caution and future studies should be performed to address these questions.

Limitations of the current network meta-analysis

Undeniably, there are some limitations in this network meta-analysis. First, this study did not register formal protocol, but we strictly followed the PRISMA-NMA statement. Second, details of randomization, assignment hiding method, and blind method were not reported in some eligible RCTs, which may introduce bias in selection and measurement. Third, over half of the studies in this systematic review are from China, so our findings inevitably have a geographical bias. Fourth, the meta-analysis eliminated conference abstracts and studies that were not in English and Chinese, with a relatively small number of studies included and only two on hCG, one on PBMC, and one on SG-CSF. Fifth, although we found that the dose of the same drug varied in different studies, we did not classify the included studies into subgroups to conduct further analysis due to the limited number of eligible studies and insufficient sample size. Sixth, safety assessments for different drugs were not ensured since the adverse events in tested treatments were reported in very few studies. Seventh, we combined studies with different designs to estimate the comparative efficacy of different immunomodulatory therapies. Therefore, our findings will inevitably be biased because non-RCTs are more susceptible to confounding factors than RCTs. Finally, we need to clearly know that SUCRA cannot show whether the difference between treatments is clinically meaningful, and ranking of treatments can distract attention from the evidence and issues underlying individual studies.

Conclusions

In summary, among four kinds of immunomodulatory therapies for the treatment of patients with thin endometrium, SG-CSF, and hCG may increase endometrium thickness and improve IR rate through the regulation of IG immune environment, which is the optimal immunotherapy based on current evidence. Given the potential limitations, our findings need to be treated with caution in clinical practice.

Patient consent

Not required.

Author contributions

Substantially contributed to conception or design: Lifei Li and Yan Wang. Contributed to data acquisition, analysis, or interpretation: Lifei Li, Zhijian Kou, and Fei Zhao. Drafted the manuscript for important content: Lifei Li. Critically revised the manuscript for important intellectual content: Xuehong Zhang. Gave final approval: All authors.

Acknowledgments

We would like to deeply appreciate all authors who performed all eligible studies which have been included in the present network meta-analysis.

Availability of data and materials

All data generated or analyzed during this study are included in this published article/as Supplementary Information Files.

Disclosure statement

None.

Additional information

Funding

The study was supported by Natural Science Foundation of Gansu Province (No. 21JR7RA366), Lanzhou Chengguan District Science and Technology Project (No. 2021-9-8), Gansu Administration of Traditional Chinese Medicine (No. GZKP-2021-29), and the Gansu Natural Science Foundation [21JR1R097].