A network meta-analysis of strategies for endoscopic submucosal dissection procedures from the perspectives of efficacy, duration and safety Authors: Guannan Liu, Tianying Chang, Wanqi Wang, Zitong Guo, Zhitao Wei, Yong Yang Published: 2026 ABSTRACT AbstractBackgroundEndoscopic submucosal dissection (ESD) is a well-accepted endoscopic resection modality for early-stage gastrointestinal tumors. Over the past decades, progressive refinements in instruments, traction methods, and closure techniques have enhanced the safety and efficacy of ESD, and ESD further evolves into conventional ESD (C-ESD), traction-assisted ESD (T-ESD), and underwater ESD (U-ESD). The three modalities have their distinct benefits and are complementary to each other in clinical practice. At the same time, they compete with one another, presenting a challenge to operators with respect to which to select in specific scenarios.ObjectiveTo conduct a network meta-analysis to make comparisons of the three approaches of ESD, with a focus on key endpoints including their operative duration, resection rate, R0 resection rate, and adverse event incidence.MethodsA computerized search was conducted across PubMed, Embase, Web of Science, and Cochrane Library databases. The literature was evaluated using the PRISMA 2020 framework. The netmeta package was used for pairwise and network meta-analysis. Continuous outcomes were pooled as mean differences, and dichotomous outcomes were pooled as risk ratios. When continuous data were reported as medians with interquartile ranges or ranges, values were converted to approximate means and standard deviations before pooling.ResultsNine English-language randomized controlled trials involving 919 participants were included. The pooled absolute estimates showed that T-ESD and U-ESD had higher procedure speeds than C-ESD (19.61 and 18.99 vs. 15.20 mm2/min, respectively) and shorter median procedure times (55.02 and 58.51 vs. 70.20 min, respectively). For adverse events, the absolute rates were 7.8% for T-ESD, 6.3% for U-ESD, and 10.8% for C-ESD. For R0 resection, the corresponding rates were 92.0%, 92.0%, and 91.1%, and for en bloc resection they were 97.9%, 97.0%, and 96.0%. According to SUCRA ranking, T-ESD ranked highest for procedure speed (0.76), median procedure time (0.83), en bloc resection (0.87), and R0 resection (0.65), whereas U-ESD ranked highest for adverse events (0.78). These rankings were interpreted together with effect estimates and confidence intervals.ConclusionCompared with C-ESD, U-ESD and T-ESD showed favorable trends for different procedural endpoints, but the certainty of superiority varied across outcomes. T-ESD ranked highest for procedure speed, median procedure time, en bloc resection, and R0 resection, whereas U-ESD ranked highest for adverse events. Because several dichotomous outcomes had confidence intervals crossing the null value, SUCRA-based rankings should be interpreted as comparative probability rankings rather than definitive proof of clinical superiority.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12876-026-04993-1.Keywords: Network meta-analysis, Endoscopic submucosal dissection (ESD), Traction-assisted endoscopic submucosal dissection (T-ESD), Underwater endoscopic submucosal dissection (U-ESD), Efficacy, Resection time Introduction Endoscopic submucosal dissection (ESD) is an endoscopic resection technique specially used to treat various gastrointestinal lesions. It is originally developed in Japan for the treatment of early gastric cancer [1]. Because of its advantages such as small trauma and fast postoperative recovery, it has gradually become the first choice for the treatment of gastrointestinal diseases [2]. ESD enables a high proportion of en bloc resection and R0 resection for large superficial gastrointestinal lesions and early gastrointestinal cancers. However, performing ESD in the colon presents significant challenges as well. The unique anatomical features of the colon-such as its large curvature angles, thin intestinal wall, and continuous peristaltic movements-complicate conventional ESD (C-ESD) procedures. In order to reduce the incidence of adverse events, it is critical to improve C-ESD and to develop more advanced innovative techniques, elevating the success rate and safety of surgery [3]. Traction-assisted endoscopic submucosal dissection (T-ESD) can achieve safe and efficient dissection operation, and has become the “ideal means” for endoscopic submucosal dissection [4]. Inspired by the traction and pushing techniques of surgery, traction can lift the diseased tissue with external force, clearly expose the boundary between blood vessels and muscle layer, make the stripping operation safer, and significantly shorten the operation time without increasing the risk of adverse events. Underwater endoscopic submucosal dissection (U-ESD) is a cutting-edge minimally invasive technique in endoscopy that has emerged in recent years. By performing lesion dissection in a “submerged environment”, it significantly enhances both operational safety and efficiency [5]. The core advantage of U-ESD lies in the “floating effect” generated during water immersion or saline immersion, which naturally provides stable and continuous traction force for submucosal dissection-an extra alternative to the traditional external traction methods. Additionally, sustained water flow perfusion maintains the submucosa in a hydrated state, thereby substantially reducing the frequency and dosage of intraoperative submucosal injections while streamlining the procedure. The theoretical advantages of T-ESD and U-ESD have been evaluated in the literature [6]. To quantify the clinical strength of T-ESD and U-SED, this study conducts a systematic review and network meta-analysis based on surface under the cumulative ranking curve area (SUCRA) of varying endoscopic resection modalities for gastrointestinal lesions, wherein SUCRA, the well-accepted statistical approach is employed for ranking and comparing the relative effectiveness of multiple therapeutic interventions in the management of lesions or early tumors. It is known that the core mechanism of SUCRA lies in quantifying the relative advantages and disadvantages of different treatments by estimating the probability that each intervention likely serves as the optimal treatment option or attains a specific rank in the hierarchy of all compared interventions [7]. Notably, the SUCRA value is bounded within the range of 0 to 1, where a higher SUCRA value denotes a superior ranking performance of the corresponding treatment, indicating a greater probability of being the most effective intervention among the compared alternatives. This study aims to investigate and compare the differences in reducing dissection time and adverse events among various ESD techniques. By ranking these methods, the research is supposed to identify the optimal approach-specifically including C-ESD, T-ESD and U-ESD-thereby providing evidence-based support for selecting the best ESD protocol suitable for dissecting gastrointestinal lesions or tumors. Materials and methods Protocol registration This systematic evaluation and network meta-analysis were conducted in strict compliance with the Cochrane Collaboration’s recommendations, following the “Protocol for Systematic Reviews and Meta-Analysis of Network Meta-Analysis (PRISMA-NMA)” guidelines [8]. The study was based on a pre-registered research protocol (registration number CRD420251040308) published on the International Platform for Systematic Reviews (PROSPERO). Retrieval strategy The review systematically retrieves English-language randomized controlled trials (RCTs) published in PubMed, Embase, Web of Science and Cochrane Library as of May 2025 to compare the efficacy and safety of T-ESD and U-ESD techniques in resection of gastrointestinal lesions. Additionally, a manual search was conducted on references included in the literature and key reviews. The primary sources for this study were PubMed, Embase, Web of Science, and Cochrane Library. The specific search strategy is illustrated in the Provenance of Materials named Specific Search Strategy. Two participants (1 and 2) independently conducted literature searches to minimize bias. After removing duplicates, three reviewers (1, 2, and 3) evaluated the titles and abstracts of all retrieved articles for eligibility. A pre-designed spreadsheet was used to assess the qualifications of each selected article, with any disagreements resolved through discussion and consensus. When multiple articles were available for an individual study, the most recent and complete version is preferred. For missing messages, supplementary information was obtained by contacting the corresponding author via email. The extracted data included: details of research designs, first author and publication years, countries, resection methods, number of participants and outcome data (Table 1). Table 1Characteristics of the RCTs included in the meta-analysisFirst author, YearCountryArmSubjects, nAge, years Median(IQR) Mean ± SDMale sex (%)Lesions size, mmMedian (IQR) Mean ± SDEn - bloc resection (%)R0 resection (%)Procedure speed, mm²/min Median (IQR) Mean ± SDAEs rate (%)Median procedure time(min)1Oh CK, 2024 [9]South KoreaC-ESD2863(50–70)50.0%31.3 ± 7.6100%100%15.2 ± 7.3Delayed bleeding: 3.6%PECS: 14.3%72.5 ± 32.5U-ESD2867(58–72)50.00%31.6 ± 8.5100%100%21.9 ± 6.9Delayed bleeding: 3.6%PECS: 3.6%49.5 ± 20.32Nagata M, 2024 [10]JapanC-ESD6971(63–78)55.1(38/69)26(23–31)10097.117.4(13.3–24.2)4.3 + 2.955.5(33.9–75.4)U-ESD7071(61–79)45.7(32/70)25(20–30)10010019.9(14.4–24.6)2.948.3(33.8–66.3)3Ito M, 2025 [11]JapanT-ESD5363.9 ± 12.964.2(34/53)34(29–45)98.194.314.8(8.9–23.9)7.5(4/53)52(27–87)C-ESD5566.0 ± 13.263.6(35/55)36(30–40)1009113.3(8.9–18.8)9.1(5/55)55(40–80)4Ritsuno H, 2014 [12]JapanT-ESD2766.2 ± 9.666.7(18/27)33.5 ± 12.5100(27/27)NANA037.4 ± 32.6 minC-ESD2366.4 ± 8.952.2(12/23)37.8 ± 13.195.7(22/23)NANA4.3(1/23)67.1 ± 44.1 min5Ichijima R, 2023 [13]JapanC-ESD12870(63–77)60.9(78/128)25(21–35)95.3(122/128)87.5(112/128)14(9–24)13.3(17/128)61(40–100)T-ESD12373(63–79)50.4(62/123)27(20–34)98.4(121/123)92.7(114/123)16(11–25)8.9(11/123)53(40–76)6Harada H, 2019 [14]JapanC-ESD4571(61–80)58.3%(28/48)25(21.8–29.3)100%100%16.3(11.4–19.8)8.9%(4/45)41(31–55)U-ESD4673(64.5–76)63.8%(30/47)25(21–29)100%100%20.1(17.3–28.1)8.7%(4/46)29.5(22.3–44)7Yamasaki Y, 2018 [15]JapanC-ESD4267(43–86)54.76(23/19)30(20–60)10098NA2.38(1 case intraoperative perforation)68(30–180)T-ESD4265(41–84)57.14(24/18)30(20–55)10093NA2.38(1 case delayed perforation)43(11–86)8Tamaru Y, 2022 [16]JapanC-ESD4971.2 ± 10.859.240(34–53)1009823.7 ± 12.2062.3 ± 39.7T-ESD4870.4 ± 10.662.539(32–46)10097.925.6 ± 11.44.2(Perforation 2.1% + Delayed hemorrhage 2.1%)47.3 ± 25.69Mori H, 2017 [17]JapanC-ESD2272 ± 1263.6%(14/22)> 20 mm(resected area: 27.6 cm² median)NANA12.5(10–18)(0.125 (0.1–0.18)cm²/min converted)9.1%(2 cases)130.0(56.0–240.0)T-ESD2174 ± 1076.2%(16/21)> 20 mm(resected area: 27.3 cm² median)NANA23.5(16–36)(0.235 (0.16–0.36)cm²/min converted)0%(0 cases)80(35.0–130.0)Abbreviations: RCTs randomised controlled trials, IQR interquartile range, SD standard deviation, AEs Adverse events, C-ESD conventional endoscopic submucosal dissection, T-ESD traction-assisted endoscopic submucosal dissection, U-ESD underwater endoscopic submucosal dissection, NA no data in the original documentReference [1] Oh CK, Chung HH, Park JK, et al. Comparing underwater endoscopic submucosal dissection and conventional endoscopic submucosal dissection for large laterally spreading tumor: a randomized controlled trial (with video). Gastrointest Endosc. 2024. 100(6): 1079–1087.e1 Characteristics of the RCTs included in the meta-analysis Delayed bleeding: 3.6% PECS: 14.3% Delayed bleeding: 3.6% PECS: 3.6% Abbreviations: RCTs randomised controlled trials, IQR interquartile range, SD standard deviation, AEs Adverse events, C-ESD conventional endoscopic submucosal dissection, T-ESD traction-assisted endoscopic submucosal dissection, U-ESD underwater endoscopic submucosal dissection, NA no data in the original document Reference [1] Oh CK, Chung HH, Park JK, et al. Comparing underwater endoscopic submucosal dissection and conventional endoscopic submucosal dissection for large laterally spreading tumor: a randomized controlled trial (with video). Gastrointest Endosc. 2024. 100(6): 1079–1087.e1 Inclusion and exclusion criteria Inclusion criteria Studies meeting the following criteria are included: (1) randomized controlled trials comparing C-ESD, T-ESD, and U-ESD techniques; (2) those focusing exclusively on endoscopic procedures in the resection of colorectal lesions; (3) studies containing evaluation of relevant outcome metrics; (4) those conducted in adult populations (age ≥ 18 years). Exclusion criteria Non-randomized studies, retrospective studies, studies evaluating upper gastrointestinal endoscopy, animal experiments, and studies lacking essential baseline data are excluded. Data collection and quality assessment The primary outcomes assessed in the study feature resection speed, resection time and incidence of adverse events related to the intervention appraoches, while the secondary outcomes encompass en bloc resection rate and R0 resection rate. En bloc resection is defined as the complete removal of the lesion in one operation, and R0 resection refers to the simultaneous pathological negative of the cutting edge (side edge and base edge) on the basis of the complete removal of the lesion in one operation. Since only 9 studies are included in this network meta-analysis, funnel plots or tests are not feasible to be adopted. The participants (1 and 2) independently extract the result data of those papers included in this review. The study utilized the netmeta software package in R (R Foundation for Statistical Computing, Vienna, Austria) to conduct direct and indirect comparisons of various endoscopic submucosal dissection techniques. A web-based meta-analysis table is created to rank the protocols for colorectal lesion resection using SUCRA. The Cochrane risk assessment tool was used to evaluate the risk of bias in included studies, categorized into low risk, high risk, and unclear. The revised Jadad scale was employed to assess methodological quality, with scores ranging from 0 to 2 for randomization, sequence generation, allocation, concealment, and blinding, and 0–1 for withdrawal and dropout. The total score is out of 7 points, where ≤ 3 points indicate low quality, and ≥ 4 points indicate high quality. The meta-analysis used a random-effects model to calculate the risk ratio (RR) for binary variables and the mean difference (MD) for continuous variables, along with 95% confidence intervals (CIs). For procedure speed and procedure time, all included studies reported values on clinically comparable scales (mm2/min for speed and minutes for time). When a study reported continuous outcomes as median with interquartile range or range, the values were converted to approximate mean and standard deviation before pooling using established conversion methods. Therefore, MD rather than standardized mean difference was adopted to retain direct clinical interpretability. Each study was evaluated and classified by two independent reviewers (1 and 2). There arose minor disagreements between reviewers regarding the collection of qualitative and quantitative data (overall observer variance < 10%), with a third-party arbitrator (the 3rd author) resolving disagreements when necessary. The statistical methods used in this paper were reviewed by experts from Changchun University of Chinese Medicine. Results Literature retrieval results A total of 1,654 records are retrieved from the database. After removing duplicates and irrelevant entries, 94 articles are qualified for a full-text screening, among which 85 are excluded for various reasons. Ultimately, 9 randomized controlled trials involving 919 subjects are included for the network meta-analysis, all of which provide complete outcome data, though potential biases remain uncertain. The PRISMA flowchart is presented in Fig. 1. PRISMA 2020 flow for new systematic reviews which included searches of databases and registers only. Preferred Reporting Items for Systematic Reviewsand Meta-Analyses (PRISMA) flow diagram of the searchprocess Basic characteristics of incorporated texts After the text screening, a total of 9 articles [9–17] meet the criteria and are included in the study, involving 919 participants with their colonic lesions removed with the three types of dissection approaches. The articles are published from 2014 to 2025. Quality evaluation of texts The revised Jadad scale was used to evaluate the methodological quality of the included studies. The Jadad scale is a tool developed by A.R. Jadad et al. in 1996, primarily designed to assess the methodological quality of randomized controlled trials (RCTs) [18]. Through standardized scoring, this scale assists researchers in evaluating the reliability and risk of bias in clinical trial reports, and is widely employed in medical research and systematic reviews. The Jadad scale features three core quality criteria: randomization, double-blinding, and withdrawals or losses to follow-up. The results showed that there were 9 high-quality studies and 0 low-quality studies, with a total score of 38 points and an average score of 4.2 points, indicating that the overall quality of the included studies was relatively high. Network meta-analysis Procedure speed Five studies [9–11, 16, 17] evaluated procedure speed across 443 subjects. The pooled absolute mean procedure speed was 15.20 mm2/min for C-ESD, 19.61 mm2/min for T-ESD, and 18.99 mm2/min for U-ESD. The forest plot (Fig. 2A) showed higher procedure-speed estimates for U-ESD (MD = 3.79, 95% CI = -1.57 to 9.14) and T-ESD (MD = 4.41, 95% CI = -0.53 to 9.35) than for C-ESD (reference, MD = 0.00). Because both confidence intervals crossed the null value, these findings should be interpreted as favorable trends rather than statistically confirmed superiority. According to SUCRA ranking, T-ESD ranked highest (0.76), followed by U-ESD (0.68) and C-ESD (0.06) (Table 2). Figure 2B presents the network structure, and Fig. 2C presents the ranking plot. Forest plot, network and SUCRA score ranking of procedure speed among three ESD treatment approaches. A The forest plot demonstrated that the MD values of U-ESD and T-ESD were both higher than that of C-ESD, indicating that the former two interventions were superior to C-ESD with regard to procedure speed. B The network plot illustrated that C-ESD exhibited a stronger correlation with T-ESD than with U-ESD. C The SUCRA plot showed that T-ESD achieved a higher value (4.41), suggesting that it represented the more favorable intervention Table 2Absolute clinical values and SUCRA score ranking of different interventions in terms of median procedure time, procedure speed, adverse events, en bloc resection rate, and R0 resection rateOutcomeAbsolute value: T-ESDAbsolute value: U-ESDAbsolute value: C-ESDSUCRA ranking (T-ESD/U-ESD/C-ESD)Median procedure time55.02 min58.51 min70.20 min0.83 / 0.66 / 0.01Procedure speed19.61 mm2/min18.99 mm2/min15.20 mm2/min0.76 / 0.68 / 0.06AEs7.8%6.3%10.8%0.60 / 0.78 / 0.12En bloc resection97.9%97.0%96.0%0.87 / 0.34 / 0.30R0 resection92.0%92.0%91.1%0.65 / 0.61 / 0.25Absolute clinical values are presented as pooled estimates for each intervention group and should be interpreted together with MD, RR, 95% CI, and SUCRA rankings Absolute clinical values and SUCRA score ranking of different interventions in terms of median procedure time, procedure speed, adverse events, en bloc resection rate, and R0 resection rate Absolute clinical values are presented as pooled estimates for each intervention group and should be interpreted together with MD, RR, 95% CI, and SUCRA rankings Median procedure time(min) Seven studies evaluated median procedure time [9–12, 14, 16, 17], involving 584 participants. The pooled absolute mean/converted mean procedure time was 70.20 min for C-ESD, 55.02 min for T-ESD, and 58.51 min for U-ESD. The forest plot (Fig. 3A) showed that both U-ESD (MD = -11.69 min, 95% CI = -21.30 to -2.09) and T-ESD (MD = -15.18 min, 95% CI = -27.02 to -3.34) were associated with shorter procedure time than C-ESD (reference, MD = 0.00). In contrast to several dichotomous outcomes, both confidence intervals excluded the null value. According to SUCRA ranking, T-ESD ranked highest for shorter median procedure time (0.83), followed by U-ESD (0.66) and C-ESD (0.01) (Table 2). Figure 3B presents the network structure, and Fig. 3C presents the ranking plot. Forest plot, network and SUCRA score ranking of median procedure time among three ESD treatment approaches. A The forest plot demonstrated that the MD values of U-ESD and T-ESD were both lower than that of C-ESD, indicating that the former two interventions were superior to C-ESD with regard to median procedure time. B The network plot illustrated that C-ESD exhibited a stronger correlation with T-ESD than with U-ESD. C The SUCRA plot showed that T-ESD achieved a lower value (-15.18), suggesting that it represented the more favorable intervention AEs rate(%) A total of 9 studies evaluated adverse events [9–17], e.g. post-ESD coagulation syndrome (PECS), delayed bleeding, intraoperative or delayed perforation, and involving 919 subjects. The pooled absolute adverse-event rates were 10.8% for C-ESD, 7.8% for T-ESD, and 6.3% for U-ESD. Figure 4A and B showed lower point estimates for adverse events with U-ESD (RR = 0.58, 95% CI = 0.25 to 1.36) and T-ESD (RR = 0.72, 95% CI = 0.41 to 1.28) than with C-ESD (reference, RR = 1.00). However, both confidence intervals crossed 1.00, indicating no statistically significant difference. According to SUCRA ranking, U-ESD ranked highest for adverse events (0.78), followed by T-ESD (0.60) and C-ESD (0.12) (Table 2). Figure 4C presents the ranking plot. Forest plot, network and SUCRA score ranking of adverse events among three ESD treatment approaches. A The forest plot demonstrated that U-ESD and T-ESD were associated with lower adverse events compared with C-ESD. However, the 95% CIs for both comparisons crossed the line of no effect, indicating that these differences were not statistically significant. B The network plot illustrated that C-ESD exhibited a stronger correlation with T-ESD than with U-ESD. C The SUCRA plot showed that U-ESD achieved a lower value (0.80), suggesting that it represented the morefavorable intervention R0 resection (%) Seven studies [9–11, 13–16] evaluated R0 resection rates across 826 patients. The pooled absolute R0 resection rates were 91.1% for C-ESD, 92.0% for T-ESD, and 92.0% for U-ESD. As shown in Figs. 5 A and B, the differences in R0 resection rate were not statistically significant for T-ESD (RR = 1.01, 95% CI = 0.97 to 1.05) or U-ESD (RR = 1.01, 95% CI = 0.98 to 1.04) compared with C-ESD (reference, RR = 1.00). SUCRA ranking suggested a modest ranking advantage for T-ESD (0.65), followed by U-ESD (0.61) and C-ESD (0.25) (Table 2). This ranking should be interpreted cautiously because the effect estimates were close to 1.00 and the confidence intervals crossed the null value. Figure 5C presents the ranking plot. Forest plot, network and SUCRA score ranking of RO resection among three ESD treatment approaches. A The forest plot demonstrated that U-ESD and T-ESD were associated with higher RO resection compared with C-ESD. However, the 95% CIs for both comparisons crossed the line of no effect, indicating that these differences were not statistically significant. B The network plot illustrated that C-ESD exhibited a stronger correlation with T-ESD than with U-ESD. C The SUCRA plot showed that T-ESD achieved a lower value (1.01), suggesting that it represented the more favorable intervention En bloc resection (%) A total of 8 studies evaluated the en bloc resection rate [9–16], involving 876 patients. The pooled absolute en bloc resection rates were 96.0% for C-ESD, 97.9% for T-ESD, and 97.0% for U-ESD. As shown in Fig. 6A and B, the RR estimates for T-ESD (RR = 1.02, 95% CI = 0.99 to 1.04) and U-ESD (RR = 1.01, 95% CI = 0.98 to 1.02) were not statistically significant compared with C-ESD (reference, RR = 1.00). According to SUCRA ranking, T-ESD ranked highest for en bloc resection (0.87), followed by U-ESD (0.34) and C-ESD (0.30) (Table 2). Because the confidence intervals crossed the null value, the SUCRA ranking should be read as a probability-based ranking rather than conclusive evidence of superiority. Figure 6C presents the ranking plot. Forest plot, network and SUCRA score ranking of en-bloc resection among three ESD treatment approaches. A The forest plot demonstrated that U-ESD and T-ESD were associated with higher en-bloc resection compared with C-ESD. However, the 95% CIs for both comparisons crossed the line of no effect, indicating that thesedifferences were not statistically significant. B The network plot illustrated that C-ESD exhibited a stronger correlation with T-ESD than with U-ESD. C The SUCRA plot showed that T-ESD achieved a lower value (1.02), suggesting that it represented the more favorable intervention Discussion Endoscopic submucosal dissection (ESD) is an advanced endoscopic therapeutic technique mainly used to resect large early-stage cancers, precancerous lesions, or submucosal tumors in the gastrointestinal tract. As a minimally invasive procedure, it aims to achieve en bloc resection of lesions via an endoscope, which enables more accurate pathological evaluation and reduces the risk of tumor recurrence common in conventional piecemeal resection. At present, ESD has become one of the gold standards in endoscopic therapies [19, 20]. Endoscopic resection has been performed basically as endoscopic mucosal resection (EMR) and endoscopic submucosal dissection (ESD). Deyhle et al. [21] described the basics of EMR in 1973. While in the 1990s, ESD became globally popular, gaining notoriety motivated by the good results in Japan [22]. Nowadays, ESD has showcasing its unique strength and overcoming the limitations of EMR. The cutting-edge technique is well-accepted and gains wide attention due to its advantages of high whole resection rate and low local recurrence rate [23, 24]. However, in clinical practice of colorectal cancer treatment, conventional endoscopic submucosal dissection (C-ESD) encounters challenges such as prolonged procedure time, high risk of perforation, and considerable technical difficulty due to the thin intestinal wall, tortuous lumen, and limited endoscopic working space [25]. These anatomical and physiological constraints make it particularly difficult to maintain a clear view during submucosal dissection and provide sufficient tissue tension, thereby increasing the complexity and risk of the procedure. In addition, lesion size and location, especially lesions in the right colon or those crossing intestinal folds, have been identified as independent risk factors for intraoperative perforation during ESD [26]. Although the therapeutic strategy and instruments of ESD for colorectal cancer have been improved, adverse events still occur from time to time, requiring further refinement of existing techniques to elevate the success rate and safety of the technique [3]. Therefore, various modified techniques have been developed to overcome these technical bottlenecks and improve the safety and efficiency of the procedure. To improve the surgical view and provide adequate tension for the submucosal dissection plane, traction-assisted endoscopic submucosal dissection (T-ESD) is introduced with the aid of traction devices including S-O clips, thread clips, and pulley clips [27]. These traction devices are designed to stabilize the dissection plane by providing effective counter-traction, thus simplifying the procedure and shortening the operation time. In addition, tumor location is also an important factor leading to the prolongation of ESD operation time and the increase of operation difficulty. T-ESD utilizes external mechanical traction to expose the submucosal space, significantly improving surgical visibility and working angles while reducing blind spots caused by tissue retraction. This enhances resection efficiency and minimizes perforation risks. U-ESD uses the natural separation of mucosa and muscle layer by liquid buoyancy to form a protective liquid environment for deep tissue, achieving real-time flushing, and further optimizing the safety of surgery [9]. This study conducts a network meta-analysis to systematically explore the efficacy and safety of three endoscopic procedures (C-ESD, T-ESD, and U-ESD) in treating colorectal lesions. The results demonstrate that both T-ESD and U-ESD significantly outperform C-ESD across multiple key indicators, with each innovative technique exhibiting distinct advantages. These findings provide crucial evidence-based medical guidance for clinicians when selecting appropriate endoscopic procedures for colorectal ESD treatment. Surgical efficiency serves as the core metric for evaluating endoscopic techniques’ clinical value, directly impacting patient outcomes, healthcare resource allocation, and endoscopists’ workload. In this study, both U-ESD and T-ESD showed higher procedure-speed estimates and shorter procedure times than C-ESD, with T-ESD ranking highest according to SUCRA for both speed and time. The absolute values also suggested clinically visible differences across groups. However, the procedure-speed confidence intervals crossed the null value, so the ranking should be interpreted as a favorable trend rather than definitive statistical superiority for this endpoint. These findings are closely tied to the design principles of the two techniques, revealing their distinct impacts on surgical efficiency. The shorter median surgical duration of U-ESD stems from its unique underwater operating environment. The research data shows that the median surgical duration of U-ESD is significantly shorter than that of C-ESD, with this advantage being particularly pronounced in anatomically complex areas like the right half of the colon. This may be attributed to the liquid environment neutralizing the interference caused by intestinal peristalsis during procedures. Additionally, the unique liquid environment of U-ESD not only provides a stable platform for surgical manipulation but, more importantly, utilizes the optical properties of water to improve the clarity of the visual field. The liquid medium can effectively eliminate the smoke interference generated by the conventional air insufflation method; meanwhile, combined with the optical magnification function of the endoscope, it can provide the operator with a clearer and more stable submucosal anatomical field of view [28]. The clear liquid, combined with the optical magnification of the endoscope, can offer an extremely clear and stable surgical field of view, thereby significantly enhancing the recognizability of submucosal blood vessels and fibrous bands [9]. The cutting energy in U-ESD conducts more concentratedly within liquid, reducing damage to surrounding tissues and indirectly shortening both hemostasis and repair time [28]. During prophylactic hemostasis, the underwater environment offers critical advantages. The combination of the optical properties of water and tissue translucency significantly improves the visibility of the submucosal vascular network, which makes it easier and more reliable to precisely coagulate potential bleeding points before resection, thereby reducing the risk of intraoperative bleeding at the source [29]. However, it must be noted that the underwater environment has clear limitations in the management of active bleeding. For minor or oozing hemorrhage, continuous water flow can irrigate the surgical field, and point-to-point coagulation with appropriate instruments is feasible and effective. In contrast, in cases of massive or arterial active bleeding, rapid blood dispersion in water severely deteriorates the underwater visual field, making precise localization of the bleeding source extremely difficult. Under such circumstances, temporary conversion to intraluminal carbon dioxide insufflation has become a standardized procedure [30]. In a gaseous environment, the surgical field remains relatively dry, facilitating rapid identification and control of the bleeding source [31]. Furthermore, the performance of some dedicated hemostatic devices is environment-dependent. Accordingly, the surgical team should preoperatively plan or flexibly adjust the operating environment according to the instruments intended for use. The core of T-ESD lies in the active creation and maintenance of tension in the submucosal dissection plane using various traction devices, thereby improving the surgical field. Currently, the commonly used clinical traction devices mainly include S-O clips, string-clip traction, and pulley clips [27]. The basic operating mechanism of these devices is to mechanically lift or pull the incised mucosal edge, drawing the lesion toward the center of the intestinal lumen or in the opposite direction of dissection [27]. For example, the S-O clip is a specially designed clip that firmly grasps the mucosal edge and provides traction via a connecting thread; string-clip traction typically uses an endoscopic clip combined with a surgical suture, which is pulled by an assistant or fixed externally; the pulley-clip system optimizes traction by changing the direction of force based on a pulley-like principle [27]. Such active traction effectively improves exposure of the surgical field at specific angles, making submucosal blood vessels and fibrous bundles easier to identify and manage, thus creating conditions for safe and precise dissection [27]. The core principle of T-ESD is to apply external or internal traction to fully expand the submucosal layer, improve the surgical field of view, and stabilize the dissection plane, thereby reducing procedural difficulty [27]. Traction techniques are mainly classified into two categories: internal traction and external traction. Internal traction techniques are generally performed through the endoscopic working channel, including clip-thread traction, magnetic anchor traction, and various endoscopic traction devices. The clip-thread traction method is one of the most commonly used internal traction techniques at present, in which one or more clips are applied to the lesion margin or the dissected mucosal flap, and traction force is transmitted externally via a suture thread [32]. Magnetic anchor traction achieves non-invasive traction using attractive forces between internal and external magnets. A magnetic anchor is placed inside the body and interacts with an external magnet to provide adjustable, multi-angular traction force [33]. In addition to these conventional devices, several novel traction devices have been created in recent years, such as the articulating endoscopic traction device TRACMOTION, which enables dynamic adjustment of the traction direction during the procedure [34]. Studies have also reported the use of an Elastic Traction Device (ETD), which can be rotated and reopened to provide more flexible traction [35]. Furthermore, innovative techniques such as Clip and Rubber-band Traction ESD (CRT-ESD) employ three clips and one rubber band to form an elastic triangular structure for multidirectional traction [25]. External traction involves applying traction force from outside the body via percutaneous puncture or with the assistance of dual-channel endoscopes and other auxiliary devices. These approaches aim to overcome the technical challenges in conventional ESD caused by poor visualization and insufficient tension in the submucosal layer, thereby creating favorable conditions for safe and effective dissection [36]. In this study, the T-ESD showed a slightly higher rate of improvement in surgical speed (MD = 4.41) compared to U-ESD, and demonstrated greater stability for large lesions (> 3 cm), likely due to the traction device’s enhanced control over extensive tissue traction. Notably, the efficiency advantage of T-ESD correlates positively with surgeon experience. Skilled operators can achieve precise dissection by adjusting traction angles, while novices may prolong procedures due to improper traction force control. A liquid environment enables buoyancy-assisted dissection, which effectively dissipates and absorbs heat generated during instrument manipulation, significantly reducing thermal injury to the deep muscularis propria and surrounding tissues [37]. Hydrostatic pressure created in the liquid setting provides temporary physical sealing for minor perforations or mucosal defects that may occur during the procedure, gaining valuable time for the operator and potentially lowering the risk of severe complications from intraoperative perforation [38]. Furthermore, perfusion with electrolyte solutions such as normal saline helps maintain effective cutting and coagulation functions of electrocoagulation device due to their favorable conductivity, ensuring smooth progression of the procedure [28]. In clinical practice, the technical details of U-ESD have been continuously refined. For instance, to achieve a clearer surgical view and a more stable operating platform, methods such as the gas-liquid alternating perfusion technique have been developed, allowing flexible switching between gaseous and liquid environments according to the degree of submucosal space exposure during dissection [39]. Studies have adopted continuous low-water-pressure perfusion to reduce bubble interference during the procedure and maintain a clear surgical view [40]. U-ESD can also be combined with internal traction techniques, whereby devices including clips and rubber bands provide counter-traction to further expose the dissection plane, thereby improving surgical efficiency and safety [41, 42]. Collectively, these technical advancements have established the physiological and physical foundations of U-ESD, establishing it as an endoscopic therapeutic modality with distinct advantages. In contrast, the efficiency limitations of C-ESD primarily stem from two key factors: First, uneven mucosal lifting following submucosal injection causes inconsistent depth of dissection planes. Second, the absence of sustained traction necessitates repeated adjustments to endoscopic positioning for optimal angles. These challenges are particularly pronounced in anatomically complex areas such as the hepatic flexure and sigmoid colon, directly contributing to prolonged surgical duration and increased operational complexity. The study’s findings reveal that C-ESD ranked last among all efficiency metrics, indicating that conventional techniques can no longer adequately satisfy the treatment demands of complex colorectal lesions. This study presents a certain degree of clinical heterogeneity. In the comparative analysis of T-ESD and U-ESD, both interventions involved various technical modifications and instrument selections, which may have influenced the pooled results. As a major component of T-ESD, traction-assisted ESD encompasses multiple traction modalities; thus, different traction devices, procedural approaches, and operator experience may all affect surgical outcomes. Meanwhile, the underwater ESD subgroups also differed in several aspects, including the type of irrigation fluid, irrigation pressure, submucosal injection protocol, and procedural steps. The aforementioned clinical heterogeneity may increase between-study bias, indicating that further randomized controlled trials with more unified designs and standardized protocols are warranted in the future to clarify the actual efficacy and safety of different modified ESD techniques. In terms of adverse-event incidence, U-ESD ranked highest according to SUCRA (0.78), and the pooled absolute adverse-event rate was numerically lower for U-ESD than for T-ESD and C-ESD. This finding is biologically plausible because the fluid cushion formed in the underwater environment may buffer tissue damage caused by cutting energy on the muscular layer and may help reduce bleeding or thermal injury in selected cases. Nevertheless, the confidence intervals for adverse events crossed the null value. Therefore, U-ESD should be described as showing a favorable safety ranking rather than proven superiority over the other approaches. T-ESD ranked highest for both en bloc resection (SUCRA = 0.87) and R0 resection (SUCRA = 0.65), and the pooled absolute rates were also numerically favorable. This pattern may be related to the stable operative space created by traction devices, which enables continuous dissection along the submucosal layer and may reduce lesion remnants. U-ESD showed comparable R0 resection performance in the network estimates, whereas the difference between techniques remained small in absolute terms. Because the RR confidence intervals for R0 and en bloc resection crossed the null value, these results should be interpreted as ranking trends rather than definitive proof that one modified technique is clinically superior for curative resection. When comparing the efficacy of T-ESD and U-ESD, this study observed that both techniques achieved excellent outcomes in terms of en bloc resection rate and R0 resection rate, with no statistically significant differences between them. Given the small sample size, such differences are likely influenced by multiple key confounding factors, including the specific location and size of the lesion, operator experience, and technical details, rather than solely by the techniques themselves. In addition, one of the included U-ESD studies involved a combination of U-ESD and pocket technique [14]. Based on these findings, we believe that the inclusion of other combined techniques may also influence outcomes. This component was not clearly described in some of the literature, which may introduce potential bias. This study suggests that both traction-assisted ESD (T-ESD) and underwater ESD (U-ESD), as modifications of conventional ESD (C-ESD), can significantly improve surgical quality (e.g., higher en bloc resection and R0 resection rates) and should therefore be regarded as superior alternatives to C-ESD. U-ESD may offer particular advantages for lesions in specific locations (such as the gastric body and rectum) due to its superior initial visualization and maintenance of the submucosal space. In contrast, T-ESD, by providing continuous exposure of the submucosal plane, may be more suitable for lesions with severe fibrosis or difficult anatomical conditions. Future large-sample, prospective comparative studies are needed to more clearly define the optimal indications for each technique after controlling for variables such as lesion characteristics and operator experience. C-ESD has a steep learning curve due to its technical complexity, which directly limits its widespread adoption in non-specialized centers. Multiple studies have confirmed that operators need to accumulate a large number of cases to achieve a stable and safe level of performance. A large-scale Japanese cohort study indicates that endoscopists require a prolonged learning process to master the essence of colorectal ESD [43]. For endoscopists who are not ESD specialists, mastering derivative techniques such as peroral endoscopic myotomy (POEM) also requires a greater number of cases [44]. This high threshold makes it difficult to perform C-ESD in centers without adequate training and support. Therefore, the steep learning curve of C-ESD is not only a technical challenge but also a major barrier hindering the broader application of this technique in a wider range of medical settings. By providing a stable surgical view and a clear dissection plane, T-ESD has shown the potential to assist non-expert operators in shortening the learning curve. A subgroup analysis of the CONNECT-C trial provided evidence for this, demonstrating that traction devices can effectively help operators grasp the concept of plane dissection more quickly [45]. A dedicated training program study further confirmed that after the introduction of traction devices, the number of cases required for trainees to achieve technical proficiency decreased from 21 to 10, while self-completion rate, dissection speed, and R0 resection rate were all significantly improved [46]. However, T-ESD itself introduces new technical steps, such as the selection, placement, and adjustment of traction devices, which require an additional learning and adaptation process. Using natural physical principles such as water buoyancy, U-ESD provides a more intuitive and simplified approach to maintaining tension in the submucosal plane, which may be particularly helpful for less-experienced endoscopists. This technique streamlines the procedure by offering a clear visual field and reducing smoke interference generated by electrocoagulation. The operator’s level of experience is a key factor in selecting the appropriate ESD technique. Expert operators can flexibly use C-ESD, T-ESD, or U-ESD, and even combine multiple techniques according to the complexity of the lesion to manage the most challenging cases. For non-experts or operators on the learning curve, starting with U-ESD may be easier to master. U-ESD provides more stable passive traction and a clear visual field, lowering the technical barrier. In addition, several simple and user-friendly traction devices, such as traction rings, have proven effective and safe for use in any colorectal segment, with installation time accounting for only a small fraction of the total procedure time, making them suitable for beginners [47]. Therefore, matching operators in the learning phase with U-ESD or simple and reliable T-ESD techniques can help them safely navigate the learning curve and gradually accumulate experience. A study by Kinoshita et al. [48] demonstrated that EndoTrac ESD technology shortened the duration of C-ESD procedures, reduce the incidence of adverse events, and alleviated patients’ physical burden while shortening hospital stays and lowering medical costs. This finding suggested that cost-effectiveness should be considered a key evaluation metric in future clinical research. In a study by XIA et al. [49] evaluating the efficacy of T-ESD versus C-ESD for superficial gastrointestinal tumors, no significant differences were observed in overall resection, complete resection, or delayed bleeding. The T-ESD group demonstrated a marked reduction in surgical time compared to the C-ESD group (MD = -16.19, [95%CI-29.24, -3.13], P = 0.02). Additionally, T-ESD showed a lower perforation rate (OR = 0.32, [95%CI 0.11,0.91], P = 0.03) when compared with C-ESD. Chawin et al. [50] conducted a meta-analysis evaluating the clinical outcomes of T-ESD, incorporating 18 studies. The results demonstrated that compared with conventional C-ESD, T-ESD significantly reduced operative time (MD=-16.63 min [95% CI: -24.65, -8.60], P < 0.0001) and decreased complication rates (OR = 0.56 [95% CI: 0.36–0.87], P = 0.009). These findings are corroborated in our study as well. However, no significant differences were observed between the two groups in terms of overall resection rate or R0 resection rate. Koyama et al. [37] evaluated whether U-ESD could reduce the incidence of post-electrocoagulation syndrome (PECS) compared with C-ESD. Through an analysis of 108 consecutive patients with colorectal lesions treated with ESD, they found that the PECS incidence in the U-ESD group (1.3%) was significantly lower than that in the C-ESD group (9.6%). In addition, the 62 matched patient data also showed that the median stripping speed was faster in the U-ESD group compared to the C-ESD group. Cecinato et al. [51] evaluated the safety and efficacy of hybrid ESD (H-ESD) and U-ESD in consecutive cases of colorectal tumors. The results showed that U-ESD demonstrated higher complete resection rates and faster resection speeds, consistent with the findings of this study. Regarding adverse events, although statistically insignificant, the perforation rate in the U-ESD group (0%) was lower than that in the H-ESD group (10.8%). During an average follow-up period of 9.4 months, no recurrence is observed in the U-ESD group. The findings of this study provide a multidimensional corroboration and supplementation to the existing literature, with the safety advantages of T-ESD and U-ESD being validated in multiple studies. Through a network meta-analysis, this research further justifies that U-ESD exhibits superior safety compared to T-ESD, a conclusion rarely made in previous direct comparative studies, which may offer a fresh perspective for positioning these two emerging technologies. However, the approach of T-ESD discussed in this study exhibits some potential limitations. First, the introduction of additional devices increases the procedural steps and technical complexity of T-ESD, which may result in longer operation time compared with U-ESD [27]. Second, once many traction devices are set, their traction direction is relatively fixed and lacks flexibility, so their efficacy may compromise when dealing with extensive or irregular lesions that require multi-directional dissection [52]. Although novel articulating devices have attempted to address this issue [34], insufficient flexibility remains a common challenge that still needs further exploration in future studies. U-ESD solves the problems of difficult operation and poor vision in traditional ESD in complex parts through the innovation of “underwater environment”, and has become an important choice for minimally invasive treatment of early gastrointestinal cancer and precancerous lesions. Its advantages of “safety, precision and fast recovery” have significantly improved the quality of life of patients [53]. With the maturity of technology, the application scope of U-ESD is still expanding, and it may further replace some traditional surgical operations in the future, becoming one of the mainstream techniques of minimally invasive gastrointestinal treatment. In terms of R0 resection rate, there is no statistically significant difference between T-ESD and U-ESD in this study (RR = 1.01), which may be related to both the operator’s proficiency in techniques and his expertise about lesion characteristics. Targeted subgroup analyses are expected in this respect in the future. Notably, this study quantified the comparative profiles of different surgical approaches based on both absolute endpoint values and SUCRA rankings. This approach helps avoid over-reliance on rank probability alone and provides clinicians with a more transparent reference framework for decision-making. In this study, we explored the association between continuous variables (e.g., lesion diameter, procedure time) and categorical outcomes (e.g., achievement of R0 resection) based on key indicators such as RR and SUCRA, in order to preliminarily identify the factors responsible for these significant correlations. However, given the small sample size and incomplete documentation in some subgroup research referred to in the present study, a robust multivariable regression analysis to control for all confounding factors may be compromised by insufficient statistical power. For instance, larger lesion diameter and lesions located in difficult sites such as the gastric antrum and cardia are usually associated with higher risks of non-en bloc resection, positive resection margins, and perforation. Prolonged procedure time often serves as an indirect indicator of technical difficulty (e.g., severe fibrosis, bleeding) and may be related to an increased risk of complications. Operator experience and variability among medical centers are generally viewed as critical confounders affecting the outcomes of endoscopic procedures. Differences in equipment platforms and team coordination protocols across centers may also contribute to outcome variation. In summary, although this study confirms the overall advantages of modified ESD techniques (T-ESD/U-ESD) over conventional ESD, the actual outcomes are indeed jointly influenced by lesion characteristics, procedural parameters, and important non-technical factors (operator and center). Future study designs should aim for prospective, multi-center collaborations, with pre-specified stratification or randomization to balance operator experience. Meanwhile, more detailed baseline variables should be incorporated in future research to allow for multivariable models, in order to more clearly elucidate the net benefits of each modified technique. This future research paradigm is expected to provide higher-level evidence for the development of individualized and precise therapeutic strategies. Limitations and future research prospects The limitations of this study are as follows: (1) The number of included studies was small (9 RCTs), and the statistical power of some indicators was insufficient, which may affect the stability of ranking estimates. (2) Important elements of network meta-analysis interpretation, including heterogeneity and consistency/inconsistency assessment, remain limited in the current study; this does not invalidate the analysis, but it reduces the strength of causal interpretation. (3) T-ESD and U-ESD were not uniform interventions. T-ESD included different traction devices and traction directions, while U-ESD differed in irrigation medium, irrigation pressure, submucosal injection strategy, and whether it was combined with other techniques. This clinical heterogeneity may affect the generalizability and interpretation of treatment rankings. (4) Some continuous outcomes were reported as medians with interquartile ranges or ranges and required conversion to approximate means and standard deviations, which may introduce additional uncertainty. (5) The number of direct head-to-head comparisons between U-ESD and T-ESD was limited, and most conclusions relied partly on indirect evidence. (6) Several technical statements in the Discussion depend on currently available mechanistic or observational evidence and should be interpreted as explanatory hypotheses rather than direct proof from this network meta-analysis. Given the limitations and findings of this study, future research should focus on the following directions: (1) Conduct multicenter, large-sample randomized controlled trials (RCTs) with subgroup analyses based on different lesion characteristics (size, location, pathological type), providing more precise evidence for personalized surgical approach selection. Particularly, give priority to comparative studies targeting large lesions (> 3 cm in diameter) and special anatomical sites (e.g., ileocecal region) as they remain under-researched due to current data scarcity. (2) Compare the learning curve differences of the two techniques and evaluate the efficacy differences between the two techniques used by operators with different experience levels, providing reference for operator training, which is of great significance for the promotion of innovative techniques in primary hospitals. (3) Evaluate the effects on recurrence rate and survival rate of C-ESD, T-ESD and U-ESD through long-term follow-up (≥ 3 years), and collect the data of long-term oncology outcomes to improve the surgical evaluation system. (4) Explore the application prospects of joint technologies, such as the combination of magnetic control traction device in U-ESD, or the use of underwater operation environment in T-ESD, in order to give consideration to efficiency and safety, and break through the limitations of a single technology. (5) Conduct studies with larger sample sizes and more balanced designs to construct more robust evidence networks for methodological validation of those advanced techniques. (6) Conduct cost-effectiveness analyses to compare the economic burden and clinical benefits of the three procedures, and provide an evidence basis for health policy making and allocation of medical resources, in particular highlighting the cost-effectiveness ratio of the special equipment required by U-ESD. Conclusion This network meta-analysis makes comparisons of the efficacy and safety between C-ESD, T-ESD and ESD in the resection of colorectal lesions, with a focus on surgical speed and safety. In addition, the study tentatively corroborates its findings with existing literature, validating the safety and efficiency of T-ESD and U-ESD and highlighting U-ESD’s superior safety. T-ESD and U-ESD turn out to be superior to C-ESD in the resection of colorectal lesions, pinpointing precisely their clinical scenarios: T-ESD is more suitable for cases requiring rapid and efficient procedures, such as patients intolerant to prolonged surgery, as well as cases demanding high en bloc resection rates and R0 resection rates, including fibrotic lesions and large-area lesions., while U-ESD is preferred for optimizing surgical safety. Future research in this regard is supposed to prioritize conducting large-sample, multicenter RCTs with subgroup analyses, evaluating learning curve differences between the two techniques for operators of varying proficiency, carrying out long-term follow-ups (≥ 3 years) to assess their oncological outcomes, exploring combined applications of T-ESD and U-ESD to cope with the existing single-technique limitations, and conducting cost-effectiveness analyses to inform medical resource allocation and health policy-making. It is hoped that this research provides evidence-based medical guidance for individualizing ESD technique selection in clinical practice for colorectal lesion resection. Supplementary Information Supplementary Material 1. Supplementary Material 2. Supplementary Material 3. Acknowledgements The authors would like to thank the teachers of Changchun University of Chinese Medicine and all the authors for their help in writing this article. Conflict of interest All authors declare no conflicts of interest during the research and article preparation. Use of artificial intelligence tools In the process of manuscript preparation we utilized artificial intelligence tools for language polishing of the manuscript. Publication standards This article adheres to the PRISMA Guidelines for Systematic Reviews and Meta-Analysis. Prior to publication, the manuscript underwent three rounds of plagiarism detection using professional anti-plagiarism systems for text and figure verification. Authors’ contributions The design and writing of this article were completed by Guannan Liu and Tianying Chang, the literature retrieval, screening and data analysis were completed by Wanqi Wang and Zitong Guo, and the review was done by Zhitao Wei and Yong Yang. Funding No specific funding received. Data availability The specific search strategy of all included studies in this Meta-analysis has been uploaded as provenance of materials. Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Zhitao Wei and Yong Yang contributed equally to this work and should be considered co-corresponding authors of this paper. Contributor Information Zhitao Wei, Email: dr_sirius@163.com. Yong Yang, Email: yongyang301@163.com. References Associated Data Supplementary Materials Supplementary Material 1. Supplementary Material 2. Supplementary Material 3. Data Availability Statement The specific search strategy of all included studies in this Meta-analysis has been uploaded as provenance of materials.