As IGB therapy is increasingly incorporated into obesity management for patients with diabetes mellitus, its short-term safety merits careful evaluation given diabetes-associated delayed gastric emptying and autonomic dysfunction that may predispose to early device intolerance. These mechanisms raise the potential for increased nausea, vomiting, and dehydration following placement of space-occupying gastric devices, making evaluation of clinically meaningful short-term outcomes essential.

In this large analysis of the MBSAQIP database, patients with diabetes undergoing IGB placement did not experience higher rates of 30-day healthcare utilization and intolerance-related outcomes compared with patients without diabetes. After propensity score-matching, primary 30-day outcomes, including hospital length of stay, outpatient IV treatments, ED visits, readmissions, reoperations, and procedural interventions were comparable between groups. Similarly, in our sensitivity analysis of years 2020 to 2023, no significant differences in balloon removal were observed. The absence of increased healthcare utilization and intolerance outcomes after adjustment suggests that diabetes itself may not confer excess short-term procedural risk following IGB placement. Within the diabetic cohort, primary outcomes remained comparable between patients with IDDM and NIDDM after matching. Across both analyses, 30-day postoperative SAEs were rare, and rates were comparable. Collectively, these findings suggest that IGB therapy does not appear to be associated with higher observed rates of short-term healthcare utilization, SAEs or intolerability in patients with diabetes, including those with insulin dependence.

Current guidelines from the American Society for Gastrointestinal Endoscopy (ASGE), European Society of Gastrointestinal Endoscopy (ESGE) and the American Gastroenterological Association (AGA) support IGB placement as a primary endoscopic therapy for select patients with BMI ≥ 30 kg/m2 or patients with BMI 27.0–29.9 kg/m2 with at least one obesity-related comorbidity (e.g. T2DM) [18]. Pooled data mentioned in these guidelines from 5 randomized control trials (RCTs) and 18 observational trials found that IGB lowered HbA1c levels more than noninvasive therapy, especially in patients with baseline HbA1c > 6.5% and in patients with a BMI > 40 kg/m2 [19]. While the metabolic efficacy of IGB in patients with diabetes is well-established, high-quality data specifically evaluating the short-term procedural safety in this population have been comparatively limited. Regarding safety, pooled data from seven RCTs cited in the guidelines estimate an absolute risk of 32 SAEs per 1,000 patients undergoing IGB placement compared to controls. However, these analyses were not conducted specifically in patients with diabetes, leaving uncertainty regarding procedural safety in this population [18]. Our findings help address this gap by demonstrating reassuring early safety outcomes in a large, nationally representative cohort.

The guidelines further support IGB as a bridge therapy for high-risk surgical candidates [18]. Individuals with advanced diabetes, severe obesity, or significant cardiometabolic disease may face elevated perioperative risk when undergoing surgical bariatric procedures [20, 21]. Preoperative weight reduction has been shown to improve metabolic parameters [22] and facilitate safer subsequent bariatric surgery by improving operative exposure and reducing surgical complexity [23, 24]. This bridging strategy may be particularly relevant for patients with poorly controlled diabetes, in whom optimization of metabolic status prior to surgery can reduce perioperative morbidity. Prior MBSAQIP data have shown higher 30-day mortality and SAEs after laparoscopic sleeve gastrectomy (LSG), and a higher risk of SAEs after laparoscopic Roux-en-Y Gastric Bypass (L-RYGB) in patients with IDDM compared to those without [13], highlighting the heightened perioperative vulnerability of populations with advanced diabetes. In contrast, our study did not identify statistically significant differences in short-term safety outcomes between patients with and without insulin dependence undergoing IGB placement. This distinction underscores the potential role of IGB therapy as a minimally invasive, anatomy-sparing therapeutic option that may aid in optimizing metabolic control and physiologic status prior to definitive bariatric surgery in appropriately selected patients with diabetes.

Previous large-scale studies and meta-analyses have reported significant improvements in glycemic control following IGB placement, with low overall rates of SAEs, but have not directly compared outcomes between diabetic and non-diabetic cohorts. A systematic review reported significant reductions in HbA1c six months after IGB placement without systematic evaluation of SAEs [22], while a subsequent meta-analysis of randomized and observational studies reported a pooled SAE rate of 1.3% [25]. Additional prospective and retrospective studies have similarly reported improvements in HbA1c following IGB therapy in patients with T2DM [26] and low rates of major and minor complications (0.5% and 5.4%, respectively) [27]. These findings are consistent with the low adverse event rates in diabetic cohorts observed in our study. However, the absence of propensity-matched comparisons and direct diabetic vs. non-diabetic analyses in prior studies limits their ability to isolate the impact of diabetes on procedural safety. Our study addresses this limitation by providing a matched, comparative assessment of short-term outcomes.

While metabolic efficacy is an important consideration in evaluating IGB therapy for patients with diabetes, the present analysis was not designed to assess weight loss or glycemic improvement. The MBSAQIP registry is structured to capture perioperative characteristics and standardized 30-day postoperative outcomes rather than longitudinal metabolic efficacy. Although postoperative BMI values are recorded within the dataset, the timing of these measurements varies considerably across patients and does not represent a uniform follow-up interval. As a result, reliable estimation of short-term weight loss metrics such as percent total body weight loss is difficult to address with the available data. Additionally, postoperative HbA1c values are not captured within the registry, precluding assessment of glycemic improvement or diabetes remission. Consequently, the present study focuses specifically on short-term safety and intolerance-related outcomes, which are systematically and reliably captured within the MBSAQIP database.

A clinically important finding of our study is the comparable rate of outpatient IV treatment for dehydration between cohorts. As mentioned, diabetes is associated with delayed gastric emptying and autonomic dysfunction, raising concern for increased gastrointestinal intolerance symptoms following placement of space-occupying gastric devices. Given that many intolerance symptoms may be managed in outpatient settings without prompting ED visits or hospital readmission, we sought to capture clinically relevant intolerance to the fullest extent possible using available MBSAQIP variables. Outpatient IV treatments for nausea, vomiting or electrolyte depletion represent the most pragmatic surrogate available within the MBSAQIP database for clinically meaningful intolerance that did not lead to hospital-based encounters. The lack of a significant difference in this outcome suggests that diabetes, including insulin dependence, does not appear to confer an increased risk of early intolerance requiring outpatient intervention, supporting the overall tolerability of IGB therapy in this population. Similarly, early balloon removal represents a clinically meaningful indicator of device intolerance, particularly in patients with diabetes mellitus. Although no significant differences were observed in the 2020–2023 cohort, evaluation across the entire study period was limited because this outcome could only be reliably captured in later iterations of the registry.

The absence of detailed anti-hyperglycemic medication data represents a clinically meaningful limitation with direct implications for interpretation of the observed findings of our study. During the study period, the use of glucagon-like peptide-1 receptor agonists (GLP-1 RAs), including liraglutide and semaglutide, expanded substantially among patients with obesity and T2DM [28, 29]. GLP-1 RAs are known to delay gastric emptying and increase gastrointestinal adverse effects, including nausea and vomiting, which may be additive to the space-occupying effects of IGB therapy [30]. Concurrent use of GLP-1 RA could therefore plausibly increase intolerance risk following balloon placement, an effect that cannot be evaluated within the current dataset.

Similarly, sodium-glucose cotransporter-2 (SGLT-2) inhibitors have been associated with an increased risk of perioperative euglycemic diabetic ketoacidosis (DKA), particularly in the setting of reduced oral intake or physiologic stress [31, 32]. The inability to capture medication exposure precludes assessment of medication-related metabolic complications that may influence early postoperative outcomes, including complications not captured by the database such as DKA. Accordingly, the reassuring safety profile observed in this study should be interpreted in the context of unmeasured medications effects, and future prospective studies should strongly consider capture of anti-hyperglycemic medication data to enable comprehensive procedural risk assessment.

IGB systems used during the study period likely differed in fill medium, weight, and gastric distension characteristics, factors that may influence gastrointestinal tolerability and gastric motility [33]. Fluid-filled balloons, which represented the predominant device used during a majority of the study period following United States regulatory approvals in 2015 [6, 34], exert greater intragastric mass and mechanical distension than gas-filled balloons [35]. In contrast, gas-filled systems are substantially lighter and have been associated with improved gastrointestinal tolerability compared with fluid-filled balloons [36], although weight loss outcomes may be modestly lower [35]. Adjustable fluid-filled systems, introduced later in the study period, allow endoscopic modification of balloon volume to mitigate intolerance, but exert similar physiologic effects on gastric emptying and distension [37]. These physiologic differences are relevant in patients with diabetes-related gastric dysmotility. In this context, the greater intragastric mass of fluid-filled balloons may further impair gastric emptying in patients with autonomic neuropathy, whereas lighter gas-filled systems may exert less mechanical influence of gastric motility and therefore be associated with improved gastrointestinal tolerability. However, device-level analyses could not be reliably performed within the MBSAQIP registry across the study period. However, a temporal sensitivity analysis 2016 to 2019 and 2020 to 2023 did not demonstrate significant differences in intolerance-related outcomes, suggesting that differences in device adoption over time were unlikely to meaningfully influence the primary findings.

The interpretation of outcomes within the IDDM subgroup warrants careful qualification. The MBSAQIP variable for insulin-dependent diabetes mellitus does not distinguish between type 1 diabetes mellitus (T1DM) and T2DM, two entities with fundamentally distinct pathophysiology, disease duration, autonomic neuropathy burden, and prevalence of gastroparesis [16, 38]. As a result, the IDDM cohort likely represents a pathophysiologically heterogeneous population, limiting the ability to generalize subgroup findings specifically to either T1DM or insulin-treated T2DM. Differences in gastric neuromuscular dysfunction and autonomic impairment, comparatively more pronounced in long-standing T1DM [39], may differentially influence intolerance risk following space-occupying gastric devices. Consequently, while no excess short-term risk was observed within the aggregated IDDM cohort, these findings should not be interpreted as definitive evidence of equivalent safety across distinct insulin-dependent diabetes phenotypes. Future prospective studies should explicitly differentiate diabetes subtype to permit mechanistically meaningful risk stratification. Additionally, the NIDDM versus IDDM comparison should be interpreted as exploratory. Although propensity score-matching substantially reduced baseline imbalance, complete covariate balance in the sub-analysis was not achieved (post-matching overall SMD = 0.155), likely reflecting the limited sample size of the matched cohorts and inherent clinical differences between diabetes subtypes. Consequently, residual confounding cannot be excluded.

This study has several additional limitations that warrant consideration. First, clinical management decisions, including ED evaluation, readmissions, reoperations, and interventions were made at the discretion of treating providers, which may contribute to inter-institutional variability. Heterogeneity in local protocols, documentation practices, and coding accuracy may also influence reliability of registry-based data, including the MBSAQIP. Standardized symptom reporting is also absent, therefore, milder gastrointestinal intolerance symptoms managed conservatively in the outpatient setting without IV treatments are likely underestimated. Assessment of diabetes severity was limited by frequently missing preoperative HbA1c data, which were unavailable in 88.8% of patients undergoing IGB placement in this cohort, precluding stratification by baseline glycemic control and evaluation of its association with outcomes. Analysis of missingness patterns demonstrated that patients with recorded HbA1c values had higher mean body mass index and a greater prevalence of metabolic comorbidities compared with those without recorded HbA1c values. These findings suggest non-random missingness of HbA1c data within the MBSAQIP registry, likely reflecting institutional documentation practices and patient selection. Given the extent of missing data and the small number of patients with elevated HbA1c values, incorporation of HbA1c into analytic models or stratification by glycemic control was not methodologically appropriate. Consequently, diabetes severity could not be reliably assessed, limiting metabolic contextualization. Insulin dependence was therefore used as a pragmatic surrogate for disease severity, with inherent limitations described earlier.

The retrospective nature of the study limits ability to establish causality and residual confounding cannot be fully excluded despite propensity score-matching. Selection bias remains possible, as factors influencing candidacy for IGB placement are incompletely captured within the dataset. Lastly, interpretation of the findings requires consideration of the low absolute event rates of 30-day healthcare utilization and postoperative SAEs. While the matched cohorts were sufficiently large to evaluate moderate differences between cohorts, the rarity of adverse outcomes limits the ability to discern smaller distinctions related to diabetes and insulin dependence. Detectable-difference analysis demonstrated that, with 424 matched patients per group, absolute differences of approximately 2–4% would be required to detect statistical significance for rare events occurring at rates ≤ 2%. In the NIDDM versus IDDM matched sub-analysis with 106 matched pairs, detectable differences were substantially larger (approximately 7–10%), reflecting the smaller sample size of this exploratory comparison. Therefore, smaller absolute differences in rare serious adverse events occurring at rates below 1% would not be expected to reach statistical significance within the available sample size. Therefore, modest variations in primary outcomes cannot be excluded, and the absence of statistically significant differences in rare SAEs should not be interpreted as evidence of clinical equivalence or non-inferiority between cohorts. Generalizability may also be limited, as MBSAQIP primarily includes accredited bariatric centers, which may differ from community practices where IGB is increasingly offered. Despite these constraints, this remains the largest comparative study of IGB outcomes in patients with and without diabetes.