For this systematic review and network meta-analysis, we followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [14]. We searched PubMed, Web of Science, Embase, and Cochrane library databases using the keywords and Medical Subject Heading (MeSH) terms “IUA” and “randomized controlled trial” (Supplementary file 1) from database inception to January 17th, 2024, supplemented with manual searches. We also searched Clinicaltrials.gov to secure additional RCTs. The research protocol was registered in PROSPERO (registration code CRD42024502941).

The inclusion criteria were: (1) participants had IUA diagnosis according to hysteroscopy findings; (2) participants had a mean age 25 to 35 years; (3) participants underwent hysteroscopic adhesiolysis followed by the application of at least one antiadhesion or endometrium repair treatment; (4) participants carried out a second-look hysteroscopy to diagnoses whether IUA recurrence or not; (5) use pictorial blood-loss assessment chart (PBAC) score to evaluate menstrual blood score [15]; (6) RCT study. The exclusion criteria were: (1) study that compared the same adjuvant treatment; (2) study with repeated use of the same treatment; (3) animal study.

The primary outcomes were the recurrence of IUA and clinical pregnancy. A second-look hysteroscopy was performed to make clear the recurrence of IUA. Clinical pregnancy was defined as the presence of a gestational sac with or without a fetal heartbeat by ultrasound. Secondary outcomes included menstrual blood score and second-look IUA score in case of adhesion recurrence. Menstrual blood score was assessed by PBAC before the second-look hysteroscopy, second-look IUA score was evaluated according to the classification method of the American Fertility Society (AFS) [16].

The risk of bias for each included trial was evaluated by The Cochrane tool [17]. Potential sources of bias include the aspects of random sequence generation, allocation concealment, blinding of participants and staff, blinding of outcome assessors, incomplete outcome data, selective reporting, and other bias. Each trial received a study level score of low, high, or unclear risk of bias for each aspect. The risk of bias assessment was judged by two authors (LJL and XYQ) independently, and disagreement were resolved by discussion.

Two independent investigators extracted data from the original reports. Specially designed forms were used to collect information on study design, country, patients’ characteristics, sample size, details of adjuvant treatments, hysteroscopic adhesiolysis procedure, menstrual blood volume, second-look IUA score, and pregnancy outcome. Continuous variables represented by medians and quartiles (or interval) were converted to means and standard deviations according to the methods described in Luo D literature [18]. Data that could not be extracted from the original publication were requested from the corresponding author.

Paired meta-analyses were conducted in the random effect model when direct data were available. Heart shape balloon (HSB) placed in uterus for days based on hormone therapy was regarded as the control group. The “mvmeta” and “Network” packages in Stata MP 17.0 (developed by StataCorp, headquartered in College Station, TX, USA) were used for network meta-analyses. For each interested result, the command network element inconsistency $>$ was used to statistically confirm the global consistency hypothesis between networks by fitting the design-by-treatment interaction model, which provided a global Wald test to evaluate the inconsistency in the whole network [19, 20]. The Separating Indirect from Direct Evidence (SIDE)-splitting method using the command network sidesplit all $>$ was used for the local test when closed loop existed [21]. When p $>$ 0.05 indicated that there was no significant inconsistency between the direct comparison and the indirect comparison, and pairwise comparisons were made between different interventions. Network geometry and node connectivity were visualized for all outcomes. For all treatment comparisons, we gave the comprehensive odds ratio (OR) or mean difference and 95% confidence interval (95% CI), which explained the uncertainty of variance estimation in the ranking. We also provided the comprehensive processing effect of 95% confidence intervals of all comparisons in forest plots. To obtain the treatment grade, the effectiveness of each intervention was ranked by the surface under the cumulative ranking curve area (SUCRA), the higher the ranking, the greater the possibility of becoming the best intervention.

To evaluate the presence of small study effects, we visually inspected comparison adjusted funnel plots for each outcome. Funnel plots for all comparisons are also presented.

A total of 1815 studies were identified through the database search. Among them, 1024 studies were duplicated and were deleted. After screening the titles and abstracts, 751 studies were deleted. Of the remaining 40 studies that were evaluated in full text, 10 were not RCTs so were removed, 11 were deleted because the intervention group and the control group compared the same adjuvant treatments, and 4 trials were excluded because they repeated used the same adjuvant treatment which were different from other included trials obviously. Consequently, 15 studies were included for network meta-analysis [5, 6, 7, 8, 9, 10, 11, 12, 13, 22, 23, 24, 25, 26, 27]. The literature search process is described in a flowchart (Fig. 1). Supplementary file 2 provides the full citations of the included trials, and Supplementary file 3 provides details of the excluded trials after evaluating the full text. Supplementary Fig. 1A,B presents the results for the assessment of risk of bias. Briefly, all of the included trials were considered low risk of bias, with none classified as moderate or high risk. The included studies comprised 1827 participants diagnosed with IUA and who received one of the ten different antiadhesion based on the hormone treatment, including intrauterine device (IUD), auto-cross-linked hyaluronic acid (ACP), medical chitosan (MC) + IUD, Foley (Foley balloon catheter 3–7 day) + IUD, heart shape balloon (HSB), dried amnion graft or bone marrow stem cells-scaffold or G-CSF (DBG) + Foley/HSB, autologous platelet gel or platelet-rich plasma (APG/PRP) + IUD/HSB, ACP + Foley/IUD, Foley, heart shape balloon or Foley 1 month (H/F) + IUD.

Fig. 1.

Flow chart showing study selection. RCT, randomized clinical trial.

The mean age of participants ranged from 25 to 33 years in the 15 studies included in this research. The basic characteristics of the studies included in the network meta-analysis are summarized in Supplementary Table 1. Hysteroscopic adhesiolysis was performed with a 4.5–5-mm office hysteroscope in five studies [5, 10, 12, 23, 25], whereas an 8.5-mm rigid hysteroscope was used in four studies [7, 8, 9, 11], and one study employed a 5-mm office hysteroscope initially, but switched to an 8.5-mm rigid hysteroscope when met with thick fibrosis [22]. The remaining five trials did not state the type of hysteroscope used [6, 13, 24, 26, 27]. Microscissors without energy were used for adhesiolysis in five studies [12, 13, 23, 25, 26], monopolar energy was used in two studies [5, 6], five trials opted for bipolar energy resection [7, 8, 9, 10, 11], one study utilized a plasma electric cutting needle for adhesiolysis [27], one study employed 5F pointed tips or semirigid scissors (except in patients with thick fibrous adhesions, where lysis was performed with a hysteroscopic monopolar knife) [22], and one study did not state the type of hysteroscopic surgical instruments used [24]. Postoperative hormonal treatment was administered in all the included studies. The follow-up period varied from 1 to 60 months.

Supplementary Fig. 2A,B presents the pairwise meta-analysis results for IUA recurrence and clinical pregnancy, respectively. Briefly, dried amnion graft or bone marrow stem cells scaffold or granulocyte colony-stimulating factor (DBG) + Foley/HSB showed no advantages in reducing the recurrence of IUA compared with the control (odds ratio (OR) = 0.84, 95% confidence interval (95% CI) 0.57, 1.24). For clinical pregnancy, DBG + Foley/HSB was advantageous compared with the control (OR = 2.11, 95% CI 1.34, 3.33)

The network plots of primary outcomes are presented in Supplementary Fig. 3A,B. For primary outcomes of IUA recurrence and clinical pregnancy, 13 and 8 studies, respectively, were available for analysis across all treatments. As shown in Supplementary Fig. 3A,B, the comparisons of recurrence of IUA in Foley + IUD, dried amnion graft or bone marrow stem cells-scaffold or granulocyte colony-stimulating factor (DBG) + Foley/HSB, medical chitosan (MC) + IUD, and heart shape balloon or Foley 1 month (H/F) + IUD were compared in open loops, and the comparisons of clinical pregnancy in DBG + Foley/HSB and autologous platelet gel or platelet-rich plasma (APG/PRP) + IUD/HSB were compared in open loops.

The netleague analyses of the reduction of IUA recurrence are presented in Fig. 2. Auto-cross-linked hyaluronic acid (ACP) was associated with lower IUA recurrence when compared with DBG (dried amnion graft or bone marrow stem cells-scaffold or granulocyte colony-stimulating factor) + Foley/HSB (OR = 0.22, 95% CI 0.06, 0.87), HSB (OR = 0.18, 95% CI 0.05, 0.67), Foley (OR = 0.14, 95% CI 0.03, 0.59), IUD (OR = 0.14, 95% CI 0.04, 0.48), and heart shape balloon or Foley 1 month (H/F) + IUD (OR = 0.10, 95% CI 0.02, 0.47), medical chitosan (MC) + IUD (OR = 0.08, 95% CI 0.01, 0.56); Foley + IUD was associated with lower IUA recurrence when compared with HSB (OR = 0.39, 95% CI 0.16, 0.95), Foley (OR = 0.30, 95% CI 0.10, 0.92), IUD (OR = 0.30, 95% CI 0.10, 0.89), and H/F + IUD (OR = 0.22, 95% CI 0.05, 0.90), MC + IUD (OR = 0.18, 95% CI 0.03, 0.93); ACP + Foley/IUD was associated with lower IUA recurrence when compared with HSB (OR = 0.53, 95% CI 0.31, 0.93), Foley (OR = 0.42, 95% CI 0.20, 0.87), IUD (OR = 0.41, 95% CI 0.19, 0.86), and H/F + IUD (OR = 0.30, 95% CI 0.09, 0.96), autologous platelet gel or platelet-rich plasma (APG/PRP) + IUD/HSB was associated with decreased IUA recurrence when compared with medical chitosan (MC) + IUD (OR = 0.28, 95% CI 0.10, 0.82). The SUCRA analysis showed comparable probability between the ten treatments (Supplementary Fig. 4A). Among the ten treatments according to the SUCRA rank, ACP and Foley + IUD (96.4% and 83.5%, respectively) had the highest mean ranks, whereas H/F + IUD and MC + IUD had the lowest mean ranks (16.6% and 13.5%, respectively).

Fig. 2.

The netleague analyses of IUA recurrence and clinical pregnancy. The comparisons of data are odds ratios (95% confidence interval), which should be read from left to right. ORs higher than 1 favor the left treatments, lower than 1 favor the right treatments. Significant results are in bold and underline. IUA, Intrauterine Adhesion; ACP, Auto-cross-linked Hyaluronic Acid; Foley, Foley balloon catheter; IUD, Intrauterine Device; APG, Autologous Platelet Gel; PRP, Platelet-Rich Plasma; DBG, Dried Amnion Graft or Bone Marrow Stem Cells-scaffold or Granulocyte Colony-Stimulating Factor; HSB, Heart Shape Balloon; H/F, Heart Shape Balloon or Foley; MC, Medical Chitosan; NA, Not available.

The netleague analyses of clinical pregnancy are presented in Fig. 2. Dried amnion graft or bone marrow stem cells-scaffold or granulocyte colony-stimulating factor (DBG) + Foley/HSB was associated with increased clinical pregnancy when compared with HSB (OR = 2.11, 95% CI 1.34, 3.32). Foley + IUD, auto-cross-linked hyaluronic acid (ACP) + Foley/IUD, and IUD was associated with decreased clinical pregnancy when compared with heart shape balloon or Foley 1 month (H/F) + IUD (OR = 0.31, 95% CI 0.11, 0.87; OR = 0.14, 95% CI 0.03, 0.69; OR = 0.21, 95% CI 0.07, 0.59; respectively). The SUCRA analysis showed comparable probability between the eight treatments (Supplementary Fig. 4B). In addition, the SUCRA rank indicated that H/F + IUD and DBG + Foley/HSB (89.7% and 82.1%, respectively) had the highest mean ranks, whereas IUD and ACP + Foley/IUD had the lowest mean ranks (20.4% and 10.5%, respectively).

Fitting the design-by-treatment interaction model did not provide evidence for inconsistent of the recurrence of IUA and clinical pregnancy (Global Wald test: p = 0.229; 0.476, respectively). Supplementary Fig. 5 presents inconsistency plots for the recurrence of IUA, and we did not find any evidence of inconsistency. Owing to the small number of studies on clinical pregnancy, we could not perform inconsistency plots for this outcome. In addition, we cannot exclude the possibility that the actual number of inconsistent loops is higher than the number we reported.

All our secondary outcomes were infrequently reported. Two RCTs reported on menstrual blood volume, and three RCTs reported on second-look IUA score. Consequently, network meta-analysis was not performed for secondary outcomes. The results of pairwise meta analyses were presented in Supplementary Fig. 6 and showed that dried amnion graft or bone marrow stem cells scaffold or granulocyte colony-stimulating factor (DBG) + Foley/HSB may increase menstrual blood volume (mean difference (MD) = 1.06, 95% CI 0.70, 1.41). There are insufficient studies on the second-look IUA score to perform relevant pairwise meta-analyses.

Overall, we found no strong evidence of the influence of small study effects across results, and the symmetrical funnel diagram (Supplementary Fig. 7A,B) showed that there was no significant publication bias about the recurrence of IUA and clinical pregnancy in our meta-analysis.