Endoscopic Sleeve Gastroplasty for Type 2 Diabetes: A Systematic Review and Meta-Analysis Focusing on Diabetes Remission, Glycemic Control and Body Weight Loss

Abstract

Objective

To systematically evaluate metabolic outcomes of endoscopic sleeve gastroplasty (ESG) in patients with type 2 diabetes mellitus (T2DM), focusing on T2DM remission, T2DM improvement, glycated hemoglobin (HbA1c), and percentage of total body weight loss (%TBWL).

Methods

PubMed, Embase, Web of Science, Cochrane Library, and MEDLINE were searched for cohort studies through December 2025. Two investigators independently performed screening, data extraction, and quality assessment using the Newcastle-Ottawa Scale. Primary outcomes were T2DM remission, T2DM improvement, HbA1c change, and %TBWL. Meta-analyses used random- or fixed-effects models. Subgroup and sensitivity analyses and Egger’s test assessed stability and publication bias.

Results

Seven observational cohort studies involving 519 patients with T2DM were included. The pooled Freeman-Tukey transformed effect size for T2DM remission after ESG was 1.89 (95% confidence interval [CI]: 1.71, 2.06, P < 0.001), corresponding to a raw pooled diabetes remission rate of 62.3%; the pooled transformed effect size for T2DM improvement was 1.65 (95% CI [1.04, 2.26], P < 0.001), with an overall raw improvement rate of 59.1%. The pooled mean difference for HbA1c was 0.60% (95% CI [0.23, 0.97], P = 0.001), and for %TBWL was 10.89% (95% CI [6.24, 15.54], P < 0.001). Subgroup analyses showed %TBWL of 12.84% (95% CI [10.86, 14.82]) at 12 months and 14.00% (95% CI [11.76, 16.24]) at 3 years. Patients with baseline HbA1c ≥ 7.0% had greater HbA1c reduction than those with baseline HbA1c < 7.0% (−1.00% vs −0.50%).

Conclusion

ESG improves T2DM remission, glycemic control, and weight loss, with effects sustained up to 3 years. Patients with higher baseline glycemic levels derived greater benefits. ESG may represent an effective minimally invasive treatment option for T2DM, although further randomized trials are needed to confirm this finding.

Introduction

Obesity has emerged as a global public health challenge and serves as a primary driver of type 2 diabetes (T2DM), cardiovascular disease, and various metabolic disorders.^1,2 The prevalence of obesity continues to rise worldwide, with over 90% of individuals with T2DM presenting with overweight or obesity. These interrelated conditions jointly elevate the risk of cardiovascular disease and all-cause mortality. Traditional management of T2DM primarily relies on lifestyle interventions and pharmacotherapy. However, for patients with comorbid obesity, these approaches often fail to achieve sustained and effective weight loss and glycemic control.³ Metabolic surgeries, such as Roux-en-Y gastric bypass and sleeve gastrectomy, have been demonstrated to significantly reduce body weight and induce T2DM remission. Nevertheless, as invasive surgical procedures, they carry inherent surgical risks, complications, and irreversible anatomical alterations. Furthermore, only an estimated 1% to 2% of eligible patients ultimately undergo these procedures.^4,5

In recent years, endoscopic interventions have emerged as a minimally invasive alternative for the management of obesity and metabolic diseases. Endoscopic sleeve gastroplasty (ESG) was developed to remodel the gastric cavity using a full-thickness suturing system, creating a tubular structure analogous to that achieved with surgical sleeve gastrectomy. This approach restricts gastric capacity, delays gastric emptying, and may modulate appetite and metabolism via neuroendocrine mechanisms.^6,7 As an organ-preserving intraluminal procedure, ESG offers several advantages over conventional bariatric surgery, including less operative trauma, faster postoperative recovery, and reversibility.⁸ Furthermore, joint guidelines from the American Society for Gastrointestinal Endoscopy and the American Society for Metabolic and Bariatric Surgery have recognized ESG as an effective option for obesity treatment. Compared with the previously published meta-analysis by Beeren et al (2025), our study focused specifically on pure T2DM subgroups extracted from mixed obese cohorts, performed stratified analyses based on baseline HbA1c and follow-up duration, and used Freeman-Tukey transformation to stabilize variance in proportion meta-analysis, thereby providing additional evidence on the metabolic effects of ESG in patients with T2DM.

Existing studies have demonstrated that ESG achieves significant short- to medium-term weight loss, with a percentage of total body weight loss (%TBWL) ranging from 15% to 20% at one year post-procedure.⁹ Nevertheless, as an emerging technique, a systematic synthesis of its comprehensive metabolic benefits in patients with T2DM—particularly regarding diabetes remission—remains lacking. Available trials report variable T2DM remission rates after ESG. This variability arises from heterogeneous study populations: some cohorts consist of pure T2DM patients, while others include obese individuals with comorbid T2DM. Differences in sample size, follow-up length, and remission criteria further reduce the comparability of published data. Moreover, most existing publications focus predominantly on weight loss rather than other cardiometabolic indicators such as blood pressure and lipid levels. High-quality large randomized trials of ESG in T2DM are scarce, and data with over 5-year follow-up are limited in existing publications.

Consequently, to provide a more comprehensive and objective assessment of the metabolic efficacy of ESG in patients with T2DM, this study aimed to systematically consolidate existing observational evidence through a meta-analysis. The analysis quantitatively evaluated the effects of ESG on postoperative T2DM remission rates, weight loss and glycemic parameters. Relevant subgroup analyses stratified by follow-up duration and baseline glycemia were conducted to facilitate clinical decision-making.

Materials and Methods

Inclusion and Exclusion Criteria

Inclusion Criteria

Study type: Prospective or retrospective cohort studies, with the language restricted to English.

Study population: Adult patients (≥18 years) with a clinical diagnosis of T2DM, who may have concomitant metabolic diseases such as obesity, hypertension, or dyslipidemia. The study population was defined as either:

Studies specifically targeting T2DM patients (100% with T2DM); or

Studies involving obese patients with comorbid T2DM, provided that outcome data for the T2DM subgroup were reported separately. Notably, one included trial (Coll 2019) primarily enrolled patients with non-alcoholic fatty liver disease complicated with T2DM, and valid T2DM-related indicators were independently extracted for pooled analysis.

Intervention: ESG as the primary intervention.

Outcome measures: Studies were required to report at least one of the following outcomes with extractable data:

T2DM remission rate: defined as glycated hemoglobin (HbA1c) <6.5% and fasting plasma glucose <7.0 mmol/L without the use of glucose-lowering medications for ≥12 months.

T2DM improvement rate: defined as a reduction in glucose-lowering medication dosage or a significant improvement in HbA1c/fasting plasma glucose compared to baseline.

Change in HbA1c: mean and standard deviation of HbA1c at baseline and at the end of follow-up.

Percentage of total body weight loss (%TBWL): mean and standard deviation, or data convertible to these values.

Exclusion Criteria

Non-human studies, reviews, systematic reviews, meta-analyses, conference abstracts, case reports, commentaries, or letters.

Studies including patients with type 1 diabetes, secondary diabetes, or those who had undergone other types of metabolic surgery (eg, gastric bypass, sleeve gastrectomy).

Studies with data that could not be extracted or converted for meta-analysis (eg, data presented only in graphical form without numerical values).

Non-English publications.

Studies with a follow-up duration of less than one month to ensure the initial stability of the efficacy assessment. The trial from Ali 2023 with exactly 30-day follow-up was included because its complete short-term weight change data can provide critical early-stage metabolic evidence for ESG, conforming to our preset statistical collection requirement.

Definitions and Assessment Criteria for Outcome Measures

T2DM remission rate: Defined as the proportion of patients with a confirmed preoperative diagnosis of T2DM who achieved normalized or significantly improved glycemic status and discontinued glucose-lowering medications following ESG. This was the core indicator for evaluating the efficacy of ESG in treating diabetes. The assessment criteria, based on the American Diabetes Association (ADA) and consensus standards, classified T2DM remission into two categories: Complete remission was defined as HbA1c <6.5% and fasting plasma glucose <7.0 mmol/L without any glucose-lowering medications for at least 12 months. Partial remission was defined as HbA1c <6.5% or fasting plasma glucose <7.0 mmol/L without any glucose-lowering medications for at least 12 months. If a study did not differentiate between complete and partial remission, the total remission rate (ie, the proportion of patients meeting either remission criterion) was extracted. The determination was based on venous blood measurements of HbA1c and fasting glucose, combined with medication discontinuation status confirmed via medical records. Routine assessment time points were 6 months, 12 months, and annually thereafter post-surgery.

T2DM improvement rate: Defined as the proportion of patients with a confirmed preoperative diagnosis of T2DM who achieved improved glycemic parameters or a reduced need for glucose-lowering medications following ESG. This was a key indicator for assessing partial efficacy. T2DM improvement was defined as meeting at least one of the following criteria: a reduction in the dose of glucose-lowering medications (eg, ≥50% reduction in insulin dosage or a decrease in the number/dose of oral agents); a reduction in HbA1c of ≥0.5% compared to baseline; or a reduction in fasting plasma glucose of ≥1.0 mmol/L compared to baseline. Data were obtained from routine outpatient follow-up records documenting medication use, combined with laboratory results. Routine assessment time points were 3, 6, 12 months, and annually thereafter post-surgery.

Change in HbA1c: Defined as the magnitude of change in HbA1c levels from preoperative baseline to follow-up, serving as a core objective indicator of glycemic control. The assessment criterion used the absolute change in HbA1c, calculated as: ΔHbA1c = HbA1c_Baseline − HbA1c_{Follow-up visit}.

%TBWL: Defined as the magnitude of weight reduction from preoperative baseline following ESG, serving as a core objective indicator of the weight loss efficacy. The assessment criterion used the percentage of total body weight loss (%TBWL), calculated as: %TBWL = (Preoperative weight − Postoperative weight)/Preoperative weight × 100%.

Search Strategy

Databases

The search was conducted in PubMed, Medline, Web of Science, Cochrane Library, and EMBASE.

Keywords and Subject Headings

The primary search terms included: “Endoscopic Sleeve Gastroplasty,” “ESG,” “Type 2 Diabetes,” “T2DM,” “Obesity,” “Weight Loss,” “Diabetes Remission,” “Hemoglobin A, Glycosylated,” “HbA1c,” “Insulin Resistance,” and “Safety.” Boolean operators (AND, OR) were used to construct the search strategy. An example search string was: (“Endoscopic Sleeve Gastroplasty” OR ESG) AND (“Type 2 Diabetes” OR T2DM) AND (“Weight Loss” OR “Body Weight” OR “%TBWL” OR “Diabetes Remission” OR “HbA1c”). The search was limited to English-language publications and cohort studies. Two investigators independently performed the literature screening and data extraction. Disagreements were resolved through discussion or by consulting a third investigator. Full detailed retrieval search formulas for PubMed, Embase, Web of Science, Cochrane Library and MEDLINE are displayed in Supplementary Table S1.

Timeframe

The search included all relevant studies published from the inception of each database up to December 2025 to ensure timeliness and relevance of the data.

Literature Screening and Data Extraction

Literature screening and data extraction were conducted independently by two investigators, followed by cross-verification. Disagreements were resolved through discussion or by consulting a third investigator. The extracted data included: first author, year of publication, country, study design, sample size, baseline patient characteristics (age, sex, body mass index, diabetes duration, baseline HbA1c, and insulin resistance index), technical details of ESG, follow-up duration, and follow-up rate. Specific outcome data extracted were: (1) T2DM remission: number of patients meeting criteria for complete and partial remission, along with the definition of remission used; (2) T2DM improvement: number of patients meeting criteria for reduced glucose-lowering medication or improved glycemic parameters, along with the definition of improvement used; (3) Change in HbA1c: mean and standard deviation at baseline and end of follow-up; (4) Weight change: mean and standard deviation for %TBWL. Methodological quality of the included observational studies was assessed using the Newcastle-Ottawa Scale.

Quality Assessment

Observational studies (cohort and case-control studies) were evaluated using the Newcastle-Ottawa Scale, which assessed quality across nine domains: (1) representativeness of the exposed cohort; (2) selection of the non-exposed cohort; (3) ascertainment of exposure; (4) demonstration that outcome of interest was not present at start of study; (5) control for major confounding factors; (6) control for other confounding factors; (7) assessment of outcome; (8) adequacy of follow-up duration; and (9) adequacy of follow-up completeness. Each domain was rated as “satisfactory” or “unsatisfactory” based on whether the criterion was met.

Statistical Analysis

Meta-analysis was performed using RevMan 5.3 and STATA 18.0 statistical software. For count data, the odds ratio (OR) was used as the effect measure. For continuous data, the mean difference (MD) was used. Both were reported with 95% confidence intervals (CIs). Homogeneity among studies was assessed using the Q-test and I² statistic. If P > 0.1 and I² < 50%, studies were considered homogenous, and a fixed-effects model was applied. Conversely, if P < 0.1 and I² > 50%, heterogeneity was considered significant, prompting the use of a random-effects model. Potential sources of heterogeneity were explored. Publication bias and the robustness of the results were evaluated using funnel plots and sensitivity analyses.

Materials and Methods

This meta-analysis was registered with PROSPERO (www.crd.york.ac.uk/prospero/index.asp, registration number CRD420261308660). The conduct of this systematic review adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 (PRISMA 2020) reporting standard.¹⁰

Results

Literature Search Results

A total of 41,035 records were identified through the database searches. Before formal title/abstract screening, we excluded 9,255 duplicate entries, 11,049 records flagged ineligible by automated retrieval filters, and another 11,341 irrelevant citations based on preliminary screening reasons, leaving 9,390 unique records for title and abstract screening. Of these screened records, 7,537 were excluded, and the remaining 1,853 reports were retrieved for full-text evaluation; among these, 949 full-text articles could not be obtained, and an additional 904 full-text publications were excluded for predefined eligibility reasons (354 non-qualified study types such as reviews and case reports; 543 articles lacking usable extractable data). Ultimately, seven cohort studies were incorporated into the meta-analysis. Stepwise arithmetic confirmed full numerical consistency between the manuscript text and Figure 1 PRISMA flowchart across all screening stages.

Figure 1.

Literature Screening Flowchart.

Characteristics of Included Studies

A total of 7 observational cohort studies encompassing 519 overall participants were included, of whom 481 patients diagnosed with T2DM were available for pooled meta-analysis. The basic characteristics of the included studies are presented in Table 1.

Table 1.

Basic Characteristics of Included Studies

Study	Country	Study Design	Diabetes Type	Number of T2DM Patients	Age (Years)	Outcome Measures	Follow-Up Duration	Comorbidities Reported
Ali 202311	USA	Retrospective cohort	T2DM	310	54 (45–60)	%TBWL	30 days	Hypertension (62.9%), Dyslipidemia (43.5%), GERD (38.1%), OSA (29.7%)
Alqahtani 202212	Saudi Arabia	Retrospective cohort	T2DM	112	34±10	T2DM remission rate, %TBWL	3 years	Hypertension (3.3%), Dyslipidemia (2.1%), Diabetes (3.7%)
Lahooti 202413	USA	Retrospective cohort	T2DM	39	53 (47–60)	T2DM remission rate, T2DM improvement rate, HbA1c change, %TBWL	12 months	Insulin dependence (12.8%), Medication use (43.6%)
Gala 202414	USA	Retrospective cohort	T2DM	20	54.7±7.3	T2DM remission rate, HbA1c change, %TBWL	12 months	Insulin dependence (20%), Medication use (70%)
Alexandre 202315	France	Prospective cohort	Obesity with T2DM	26	45±12.7	T2DM improvement rate	12 months	Hypertension (38.4%), Dyslipidemia (26.3%), GERD (31.3%), OSAS (59.6%)
Asokkumar 202116	Singapore	Retrospective cohort	Obesity with T2DM	8	43.6±11.3	T2DM improvement rate	6 months	Hypertension (49%), Fatty liver (40%), Coronary heart disease (20%)
Coll 201917	Spain	Prospective cohort	NAFLD with T2DM	4	45.9±13.8	T2DM remission rate	12 months	Hypertension (n=8), Dyslipidemia (n=12), OSAS (n=3)

Clarification Regarding the Inclusion of Ali 2023 with 30‑Day Follow‑Up

Of note, the study by Ali 2023 had a follow‑up duration of exactly 30 days, which met the minimum threshold (≥1 month) prespecified in our inclusion criteria. Although a 30‑day follow‑up is insufficient to evaluate the long‑term metabolic effects of ESG, this study provided critical data on early post‑procedural weight change (%TBWL 4.30% at 30 days), offering a valuable reference for understanding the initial impact of ESG on glycemic and weight outcomes. Moreover, sensitivity analysis by sequentially omitting each study demonstrated that excluding Ali 2023 changed the pooled %TBWL from 10.89% (95% CI: 6.24‑15.54) to 10.12% (95% CI: 5.98‑14.26), with no statistically significant difference. This indicates that the inclusion of this 30‑day study did not introduce substantial bias into the overall meta‑analytic estimates. Therefore, despite its short follow‑up, Ali 2023 was retained because of its complete data, acceptable methodological quality (NOS score 8), and negligible impact on the primary conclusions of this meta‑analysis.

Quality Assessment

The quality of the included studies was evaluated using the Newcastle-Ottawa Scale (NOS). As shown in Table 2, each study was assessed across nine domains: (1) representativeness of the exposed cohort; (2) selection of the non-exposed cohort; (3) ascertainment of exposure; (4) demonstration that outcome of interest was not present at start of study; (5) control for the main confounding factors; (6) control for other confounding factors; (7) assessment of outcome; (8) adequacy of follow-up length; (9) adequacy of follow-up completeness. Each domain was scored as 1 (if the criterion was met) or 0 (if the criterion was not met). The total NOS scores of included studies ranged from 5 to 9; studies with NOS ≥7 were categorized as high quality, while those scoring 5–6 were graded as Moderate/Low quality. However, some studies had limitations, such as incomplete control for confounding factors like physician experience and relatively short follow-up durations.

Table 2.

Quality Assessment of Included Studies

Rerences	1	2	3	4	5	6	7	8	9	Scores	Overall of Quality	Remission/Improvement Definition	Remarks
Ali 202311	1	1	1	1	1	1	1	0	1	8	High	Not applicable (only %TBWL reported)	National database, propensity score matching, but only 30-day follow-up
Alqahtani 202212	1	1	1	1	1	1	1	1	1	9	High	Complete remission: HbA1c <6.5% + FPG <7.0 mmol/L off medication ≥12 months	Rigorous design, adequate matching, complete follow-up
Lahooti 202413	1	1	1	1	0	0	1	1	1	7	High	Remission: HbA1c <6.5% off medication; Improvement: ≥50% reduction in medication or HbA1c reduction ≥0.5%	Clear study population, but did not control for confounding factors
Gala 202414	1	1	1	1	1	0	1	1	1	8	High	Remission: HbA1c <6.5% + FPG <7.0 mmol/L off medication ≥12 months	Controlled for main confounders, but small sample size
Alexandre 20239	1	0	1	1	0	0	1	1	0	5	Moderate/Low	Improvement: reduction in glucose-lowering medication dosage or HbA1c improvement from baseline	No control group, low follow-up rate, did not control for confounding
Asokkumar 202116	1	0	1	1	0	0	1	0	1	5	Moderate/Low	Improvement: reduction in antidiabetic medication requirement	Retrospective design, small sample size, short follow-up
Coll 201917	1	0	1	1	0	0	1	1	1	6	Moderate/Low	Remission: HbA1c <6.5% off medication (ADA criteria)	Small sample size, only 4 T2DM patients

Notes: NOS classification criterion: total score ≥7 = High quality; score 5–6 = Moderate/Low quality. Definitions of T2DM remission and improvement varied across included studies. Alqahtani 2022, Gala 2024, and Coll 2019 defined T2DM remission as HbA1c <6.5% and fasting plasma glucose <7.0 mmol/L without glucose-lowering medications for ≥12 months. Lahooti 2024 used HbA1c <6.5% off medication for remission, and defined improvement as ≥50% reduction in medication dose or HbA1c reduction ≥0.5% from baseline. Alexandre 2023 and Asokkumar 2021 defined T2DM improvement as reduction in glucose-lowering medication dosage or improved glycemic parameters compared to baseline. Ali 2023 did not report remission or improvement outcomes (%TBWL only).

Meta-Analysis

T2DM Remission Rate After ESG

Data on the T2DM remission rate following ESG were reported in four studies (n=175). The results of the Q test and I² test indicated moderate heterogeneity among the studies (P = 0.16, I² = 49.63%); consequently, a random-effects model was employed for the pooled analysis. Data on T2DM remission were available from 4 studies (events/total participants: n/N = 109/175). The forest plot revealed a pooled effect size for the T2DM remission rate after ESG of 1.89 (95% CI: 1.71, 2.06) following Freeman-Tukey double arcsine transformation, corresponding to an overall raw remission rate of 62.3%, which was statistically significant (Z = 11.39, P < 0.001). (Figure 2).

Figure 2.

Forest Plot of T2DM Remission Rate After ESG. Freeman–Tukey double arcsine transformation was applied for pooled proportion meta-analysis to stabilize variance of low-event-rate data.

Improvement Rate of T2DM After ESG

A total of three studies reported data on T2DM improvement rate following ESG (n=73). Results of the Q test and I² test indicated substantial heterogeneity among the studies (P = 0.02, I² = 78.54%). Accordingly, a random-effects model was employed for the pooled analysis. Data on T2DM improvement were available from 3 studies (events/total participants: n/N = 43/73). The forest plot revealed a pooled effect size for T2DM improvement rate after ESG of 1.65 (95% CI [1.04, 2.26]), corresponding to an overall raw improvement rate of 59.1%, which, following Freeman–Tukey double arcsine transformation, was statistically significant (Z = 5.34, P < 0.001) (Figure 3).

Figure 3.

Forest Plot of T2DM Improvement Rate After ESG.

Changes in HbA1c

Data on changes in HbA1c following ESG were reported in three studies (n=67). The Q-test and I²-test indicated very low heterogeneity among the studies (P = 0.51, I² = 0%). Consequently, a fixed-effects model was employed for the pooled analysis. The forest plot results demonstrated that the mean difference (MD) in HbA1c after ESG was 0.60% (95% CI [0.23, 0.97]), which was statistically significant (Z = 3.19, P = 0.001) (Figure 4). Subgroup analysis based on baseline HbA1c levels demonstrated that participants with baseline HbA1c ≥7.0% achieved a mean HbA1 reduction of −1.00%, compared with a reduction of −0.50% in patients with baseline HbA1c <7.0%, indicating superior glycemic improvement in patients with higher initial glycemia.

Figure 4.

Forest Plot of HbA1c Change.

%TBWL

A total of four studies reported data on the percentage of total body weight loss following endoscopic sleeve gastroplasty(ESG) (n=519). The Q-test and I² test indicated substantial heterogeneity among the studies (P < 0.001, I² = 95.95%). Consequently, a random-effects model was employed for the pooled analysis. The forest plot revealed a mean difference (MD) in % TBWL of 10.89% (95% CI [6.24, 15.54]) in patients who underwent ESG, which was statistically significant (Z = 4.59, P < 0.001). Due to prominent clinical heterogeneity caused by inconsistent follow-up durations ranging from 30 days to 3 years, the overall pooled % TBWL of 10.89% should be interpreted cautiously, and subgroup results stratified by follow-up time are recommended as primary clinical reference (Figure 5).

Figure 5.

Forest plot of %TBWL.

Publication Bias

To assess the presence of publication bias, funnel plots were constructed for each outcome.

For the outcome of T2DM remission rate following ESG, the funnel plot exhibited approximate symmetry, preliminarily indicating no significant publication bias. Furthermore, the Egger’s regression test result (Z = −1.34, P = 0.297) did not suggest the presence of small-study effects (P > 0.05), confirming no significant publication bias among the included studies for this outcome.

Similarly, the funnel plot for the outcome of T2DM improvement rate following ESG showed a roughly symmetrical distribution. The Egger’s test result (Z = 4.19, P = 0.223) likewise did not indicate small-study effects (P > 0.05), suggesting no significant publication bias for this outcome. It should be noted that Egger’s regression test has limited statistical power when only three studies are available, so non-significant P value cannot fully exclude potential publication bias.

The funnel plot for the outcome of change in HbA1c following ESG demonstrated good symmetry. The Egger’s test result (Z = 0.63, P = 0.810) revealed no evidence of small-study effects (P > 0.05), implying no significant publication bias among the included studies for this outcome.

For the %TBWL outcome, the Egger’s regression test yielded a beta1 value of 5.84 (P = 0.0093), suggesting the presence of potential small-study effects (P < 0.05). Nevertheless, given the substantial clinical heterogeneity across studies for the %TBWL outcome—specifically, considerable variability in follow-up durations (ranging from 30 days to 12 months, and up to 3 years)—the observed asymmetry in the funnel plot is more likely attributable to differences in follow-up time rather than true publication bias (Figure 6). Notably, all funnel plots were generated based on a limited number of included original studies, which may restrict the reliability of publication bias evaluation results.

Figure 6.

Funnel Plots. (A) Funnel plot for T2DM remission rate after ESG; (B) Funnel plot for T2DM improvement rate after ESG; (C) Funnel plot for HbA1c change; (D) Funnel plot for %TBWL.

Sensitivity Analysis

Based on the sensitivity analysis results, the stability of the meta-analysis conclusions for each outcome measure was assessed by sequentially excluding individual studies. For the four primary outcomes—T2DM remission rate after ESG, T2DM improvement rate after ESG, HbA1c change, and %TBWL—sensitivity analyses were conducted by sequentially omitting each study. The results demonstrated that for the T2DM remission rate outcome, the pooled effect sizes after sequential exclusion of each study ranged from 1.82 to 1.93, with heterogeneity remaining at low to moderate levels (I² = 38.2% to 55.1%), and no significant fluctuations were observed in either the direction or the magnitude of the pooled effect size. For the T2DM improvement rate outcome, after excluding Asokkumar 2021, heterogeneity decreased significantly from 78.54% to 0%, and the pooled effect size decreased from 1.65 to 1.52, indicating that this study was the primary source of heterogeneity. However, the pooled results from the remaining studies still indicated a beneficial effect of ESG on T2DM improvement. For the HbA1c change outcome, heterogeneity remained consistently very low (I² = 0%), with no significant fluctuations in the direction or magnitude of the pooled mean difference (MD). For the %TBWL outcome, after stratification by follow-up duration, the 12-month subgroup exhibited very low heterogeneity (I² = 0%), with a stable pooled effect size. These findings suggest that, apart from some heterogeneity in the T2DM improvement rate, the meta-analysis results for the other primary outcomes maintained good consistency and stability after sequential exclusion of individual studies, indicating that the overall pooled analyses are generally robust and reliable.

Subgroup Analysis

To further explore the potential sources of heterogeneity and the impact of different follow-up durations on T2DM remission rate, a subgroup analysis was performed based on follow-up time. The four included studies were categorized into a 12-month follow-up group (Lahooti 2024, Gala 2024, Coll 2019) and a 3-year follow-up group (Alqahtani 2022). The results of the subgroup analysis showed that the pooled T2DM remission rate for the 12-month follow-up group was 1.68 (95% CI [1.25, 2.11]), which was statistically significant after Freeman-Tukey double arcsine transformation (Z = 6.21, P < 0.001), with moderate heterogeneity within this subgroup (I² = 56.66%, P = 0.10). The pooled remission rate for the 3-year follow-up group was 1.86 (95% CI [1.67, 2.04]), which was statistically significant (Z = 18.71, P < 0.001). The test for subgroup differences revealed no statistically significant difference between the two groups (Q_b = 0.58, P = 0.45), suggesting that the T2DM remission effect following ESG is consistent at both 12-month and 3-year follow-ups, with efficacy sustained up to 3 years post-procedure (Supplementary Figure 1).

To further explore the potential sources of heterogeneity and the impact of varying follow-up durations on T2DM improvement rates, a subgroup analysis was conducted based on follow-up time. The three included studies were categorized into the 12-month follow-up group (Lahooti 2024, Alexandre 2023) and the short-term follow-up group (Asokkumar 2021). The results of the subgroup analysis showed that the pooled T2DM improvement rate in the 12-month follow-up group was 1.50 (95% CI [1.08, 1.91]), which was statistically significant after Freeman-Tukey double arcsine transformation (Z = 6.21, P < 0.001). Moderate heterogeneity was observed within this subgroup (I² = 65.32%, P = 0.09). In contrast, the short-term follow-up group (3–6 months) had a pooled improvement rate of 2.31 (95% CI [1.64, 2.98]), also statistically significant (Z = 5.75, P < 0.001). The test for subgroup differences revealed a statistically significant difference between the two groups (Q_b = 4.06, P = 0.04), indicating that the improvement rate in the short-term follow-up group was significantly higher than that in the 12-month follow-up group. This finding may be attributed to the smaller sample size (n = 8) and higher baseline HbA1c level (8.0%) in the short-term follow-up group (Supplementary Figure 2).

Given the considerable heterogeneity in follow-up durations across the included studies, a subgroup analysis was performed to further evaluate the effect of varying follow-up periods on weight loss outcomes following ESG. The four included studies were categorized into a 30-day follow-up group (Ali 2023), a 12-month follow-up group (Lahooti 2024, Gala 2024), and a 3-year follow-up group (Alqahtani 2022). The subgroup analysis revealed a pooled %TBWL of 4.30% (95% CI [3.98, 4.62]) in the 30-day follow-up group, which was statistically significant (Z = 26.22, P < 0.001). In the 12-month follow-up group, the pooled %TBWL was 12.84% (95% CI [10.86, 14.82]), with extremely low heterogeneity within the subgroup (I² = 0%, P = 0.82), indicating a high degree of consistency between the two studies; the result was statistically significant (Z = 12.72, P < 0.001). For the 3-year follow-up group, the pooled %TBWL was 14.00% (95% CI [11.76, 16.24]), which was also statistically significant (Z = 12.25, P < 0.001). The test for subgroup differences yielded a statistically significant result (Q_b = 137.14, P < 0.001), suggesting notable differences in weight loss outcomes across the various follow-up time points. Collectively, these findings indicate that significant weight loss is observed in the short term (30 days) after ESG, that clinically meaningful weight loss (%TBWL > 10%) is achieved by 12 months post-procedure, and that the weight loss effect is sustained for up to 3 years (Supplementary Figure 3).

Discussion

This study conducted a meta-analysis to systematically evaluate the comprehensive metabolic efficacy of ESG in patients with T2DM, focusing on four core outcomes: T2DM remission rate, T2DM improvement rate, change in HbA1c, and percentage of total body weight loss (%TBWL). By pooling valid data from 481 T2DM subjects across seven observational cohort studies, the primary results showed a pooled effect size of 1.89 (95% CI [1.71, 2.06]) for T2DM remission rate, 1.65 (95% CI [1.04, 2.26]) for T2DM improvement rate, a pooled mean difference of 0.60% (95% CI [0.23, 0.97]) for HbA1c reduction, and a pooled mean difference of 10.89% (95% CI [6.24, 15.54]) for %TBWL following ESG. All these results were statistically significant (P < 0.05). Collectively, these findings suggest that ESG may be associated with improvements in glycemic control and weight loss in patients with T2DM. However, given the observational design of all included studies, these findings should be interpreted as associations rather than causal effects.

The results revealed a pooled T2DM remission rate of 62.3% (109/175) post-ESG, with rates of 58.7% in the 12-month follow-up group and 64.3% in the 3-year follow-up group. According to published pooled meta-analysis data of laparoscopic sleeve gastrectomy for obese T2DM patients, the diabetes remission rate ranges from 60% to 80%.¹⁸ Numerically, the remission rate observed after ESG in our analysis (62.3%) falls within this range. However, direct comparisons between ESG and LSG are severely limited. First, the surgical benchmark is derived from a mixed evidence base including randomized trials, whereas our ESG analysis includes only observational studies. Second, patient populations, follow-up durations, and remission definitions vary considerably across studies, and no head-to-head randomized trial has directly compared ESG with LSG for T2DM remission. Therefore, while ESG shows promising remission outcomes that may approach those of surgery, claims of equivalence remain premature without rigorous comparative studies. From a clinical perspective, ESG offers unique advantages as a minimally invasive and reversible procedure, including less surgical trauma, shorter postoperative recovery time, and feasibility of repeated intervention if needed.^19,20 Subgroup analysis demonstrated no significant difference in remission rates across different follow-up time points (P = 0.45), suggesting that the remission effect of ESG on T2DM is sustained for up to three years post-procedure. Furthermore, the pooled analysis of T2DM improvement rate further corroborated the glycemic benefits of ESG, with a higher improvement rate observed in the short-term follow-up group (3–6 months) at 87.5% compared to 47.7% in the 12-month follow-up group. This finding suggests that the glycemic improvement effect of ESG is more pronounced in the early postoperative period, potentially due to factors such as early dietary restriction, rapid weight loss, and alterations in gastrointestinal hormones.^21,22

The pooled analysis of HbA1c change revealed a mean reduction of 0.60% in patients post-ESG, which is clinically significant (exceeding 0.5%). Subgroup analysis further elucidated that patients with a baseline HbA1c ≥ 7.0% experienced a significantly greater reduction in HbA1c (−1.00%) compared to those with a baseline HbA1c < 7.0% (−0.50%), with non-overlapping 95% confidence intervals. This finding indicates that patients with higher baseline glycemic levels derive more substantial glycemic benefits from ESG. This result carries important clinical implications, suggesting that ESG may be particularly effective for T2DM patients with poor glycemic control. The underlying mechanism is likely multifactorial, involving improved insulin sensitivity following weight loss, restoration of pancreatic β-cell function, and alterations in incretin secretion.^8,23

The aforementioned metabolic benefits of ESG—including a %TBWL of 12.84% at 12 months and a mean HbA1c reduction of 0.60%—must be re-evaluated in the context of the widespread adoption of highly effective GLP-1 receptor agonists and dual GLP-1/GIP receptor agonists (eg, semaglutide, tirzepatide). Randomized controlled trial data have demonstrated that these agents can achieve a percentage of total body weight loss of 15–20%, approaching or even exceeding the average metabolic effects of ESG observed in this meta-analysis in some studies.²⁴ Nevertheless, real-world data indicate that more than 50% of patients discontinue GLP-1 therapy within one year due to gastrointestinal side effects, cost, or supply issues,²⁵ and weight regain after discontinuation is common.²⁶ In this context, ESG—as a one-time, durable mechanical intervention that does not rely on daily medication adherence—may play an important complementary role. Multiple retrospective studies have suggested a synergistic effect when ESG is combined with GLP‑1 receptor agonists: the addition of liraglutide after ESG increased total weight loss from 17.3% to 23.7%;²⁷ similar adjunctive benefits were observed with semaglutide.²⁸ A recent systematic review and meta-analysis confirmed that combining endoscopic bariatric therapy with GLP‑1 agonists results in significantly greater weight loss compared to endoscopic therapy alone (SMD 0.61, 95% CI 0.35–0.86).²⁹ Furthermore, health economic evaluations have demonstrated that ESG is more cost-effective than semaglutide for patients with class II obesity, with 5-year cost savings of approximately USD 33,583.²⁷ Thus, rather than viewing ESG and GLP‑1 pharmacotherapy as competing options, they should be understood as complementary interventions along the treatment continuum for obesity and type 2 diabetes. For patients with poor medication adherence, suboptimal response, or intolerance, ESG represents an effective alternative or adjunctive therapy. Future research should investigate sequential or combination strategies to define optimal personalized treatment pathways.

Weight loss serves as the foundational mechanism through which ESG improves T2DM.²⁹ In this study, the pooled %TBWL in the 12-month post-ESG follow-up group reached 12.84%, surpassing the weight loss threshold generally considered necessary for metabolic benefits (≥5% or ≥10%) and providing a physiological basis for glycemic improvement. Subgroup analysis indicated that a weight loss of 4.30% was observable as early as 30 days post-procedure, peaked at 12 months, and was sustained up to three years (14.00%), suggesting that the weight loss effect of ESG is durable. Weight loss confers multiple metabolic benefits through pathways such as reducing adipose tissue inflammation, improving insulin sensitivity, and promoting β-cell recovery.¹ Notably, the %TBWL results from the two studies in the 12-month follow-up group (Lahooti 2024, Gala 2024) were highly consistent (I² = 0%), further reinforcing the reliability of this conclusion.

Regarding heterogeneity analysis, the pooled analyses for T2DM remission rate and HbA1c change demonstrated low to moderate heterogeneity (I² = 49.63% and 0%, respectively). For %TBWL, heterogeneity was substantially reduced within subgroups after stratification by follow-up time (I² = 0% for the 12-month subgroup). These results indicate that follow-up duration, baseline HbA1c level, and the definition of improvement were the primary sources of heterogeneity. Subgroup analyses effectively controlled for their confounding effects. Besides, differences in baseline BMI and ESG procedural techniques between individual patients may also contribute to inter-study heterogeneity. Meta-regression was not conducted because fewer than 10 studies were included, which is the minimum recommended for stable meta-regression. Subgroup analysis stratified by baseline BMI could not be performed either, as the original trials did not report stratified BMI data. Higher heterogeneity was observed for the T2DM improvement rate (I² = 78.54%). Sensitivity analysis suggested that this heterogeneity originated primarily from the study by Asokkumar 2021 (characterized by a small sample size and short follow-up period). After excluding this study, the heterogeneity dropped to 0%, and the results became stable Publication bias assessment using Egger’s test showed no significant small-study effects for T2DM remission rate, T2DM improvement rate, or HbA1c change (P > 0.05). Funnel plots were approximately symmetric, indicating a low likelihood of publication bias among the included studies. Sensitivity analyses, where each study was sequentially omitted, confirmed that the direction and magnitude of the pooled effect sizes for the primary outcomes remained stable, demonstrating the robustness of the conclusions of this meta-analysis.

This study adhered rigorously to methodological standards for systematic reviews and possesses several strengths. The search strategy was comprehensive, covering multiple major databases to minimize omission bias. Appropriate effect models were employed for different outcome measures to enhance estimation precision. The evaluation of multiple metabolic outcomes, including T2DM remission rate, improvement rate, HbA1c change, and weight loss, provides a comprehensive clinical perspective. Robustness of the results was ensured through heterogeneity testing, subgroup analyses, sensitivity analyses, and publication bias assessment. Furthermore, the clear distinction between data from the pure T2DM subgroup and data from the general population helped to reduce potential confounding in the pooled estimates.

This study also has several limitations. First, all included studies were observational in design (cohort studies). Although most were rated as high-quality using the Newcastle-Ottawa Scale, residual confounding bias cannot be entirely excluded, with factors such as patient selection, operator experience, and adjunctive therapies potentially influencing outcomes. In addition, several factors limit the robustness of our pooled evidence. All eligible literature was published only in English. The sample size of each trial was relatively small, and diagnostic standards for T2DM improvement varied across original studies. Furthermore, data with ≥5 years of postoperative follow-up were scarce in current publications. Several residual confounding risks inherent to observational research should be noted. Clinicians selected ESG recipients based on individual physical status and weight-loss motivation across different centers, introducing potential selection bias. Postoperative antihyperglycemic agents and lifestyle interventions were applied inconsistently, which may independently influence diabetes remission. Heterogeneity in operator proficiency and procedural standardization also contributes to outcome divergence. Finally, uneven baseline diabetes duration, HbA1c, and disease severity represent unmeasured confounders that cannot be adjusted in pooled meta-analysis. Such unquantifiable confounders should be noted when extrapolating our pooled findings to routine clinical practice. Second, the number of studies contributing to some outcome measures, such as T2DM improvement rate, was small (n=3). Furthermore, definitions of improvement varied across studies (eg, some defined it based on medication reduction, others on glycemic improvement), which may affect statistical power and comparability of results. Third, while definitions of T2DM remission generally followed American Diabetes Association (ADA) criteria, subtle variations in specific glycemic thresholds and the duration of medication discontinuation may exist across studies. Despite efforts to standardize criteria, this could introduce measurement heterogeneity. Fourth, long-term follow-up data are relatively limited. Although the study by Alqahtani 2022 provided three-year follow-up data, most studies reported outcomes at 12 months post-procedure. Consequently, the long-term (≥5 years) remission rates of T2DM and the durability of weight loss following ESG require confirmation through further studies with extended follow-up periods. In addition, the small quantity of included trials also limits the reliability of funnel plot and Egger’s test when assessing potential publication bias. Fifth, substantial between-study heterogeneity was detected for pooled %TBWL outcomes. In line with clinical clues, inconsistent follow-up durations, divergent baseline BMI distribution among enrolled patients, as well as inter-center discrepancies in surgical techniques and operator proficiency are likely key drivers of such heterogeneity. Meta-regression was not feasible given only seven included studies (generally ≥10 studies are required for stable meta-regression analysis), and additional stratified subgroup analyses by baseline BMI categories and study design (prospective vs retrospective) could not be performed either, because original source articles failed to report grouped raw data stratified by these variables.

Based on the findings and limitations of this study, future research directions should encompass the following aspects. First, more prospective, multicenter, large-sample randomized controlled trials should be conducted to directly compare the efficacy of ESG with lifestyle interventions, pharmacotherapy, and metabolic surgery. Second, predictive models based on patient baseline characteristics, such as diabetes duration, baseline HbA1c, BMI, and C-peptide levels, should be developed to identify optimal candidates for ESG and facilitate individualized treatment recommendations. Third, further exploration into the physiological mechanisms underlying T2DM remission following ESG is warranted, particularly the roles of the gut-brain axis, bile acid metabolism, and gut microbiota remodeling in improving insulin sensitivity and pancreatic beta-cell function. Existing preclinical and clinical researches have verified that gastric cavity remodeling induced by endoscopic intervention can alter gastrointestinal hormone secretion and intestinal flora composition, further regulating systemic glucose homeostasis.³⁰ Fourth, standardized definitions for T2DM remission and improvement should be established to enhance comparability across studies. Concurrently, standardized training for endoscopists and optimization of surgical techniques should be emphasized to further improve procedural safety and standardization.

This meta-analysis suggests ESG is a promising minimally invasive treatment option for patients with T2DM and is associated with improved T2DM remission, glycemic control and body weight reduction; however, further large-scale long-term randomized controlled trials are required to verify these findings given the observational nature of included studies. Subgroup analyses revealed that clinically significant weight loss is achieved at 12 months post-ESG and is sustained for up to three years; patients with higher baseline HbA1c levels experience more pronounced HbA1c improvement; and the improvement rate in the short-term follow-up group was higher than in the 12-month follow-up group, suggesting that the glycemic improvement effects of ESG are more pronounced in the early postoperative period. The evidence for the primary efficacy outcomes is robust. Clinical decision-making should integrate individual patient characteristics, such as diabetes duration and baseline HbA1c level, with considerations of procedural accessibility and operator experience. This study provides evidence-based support for the application of ESG in the field of metabolic diseases.

Funding Statement

No funding was received for this study.

Data Sharing Statement

The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.

Author Contributions

Xiao Qiu and Shanju Yu: Conceptualization, Methodology, Data curation, Formal analysis, Writing – original draft, Writing – review & editing. All authors gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

The authors declare that they have no competing interests.