ABSTRACT
Obesity and T2DM substantially increase postoperative risk, with higher surgical site infections in obesity and up to a 65% increase in overall complications in T2DM. This study assesses the impact of weight loss interventions—metabolic bariatric surgery (MBS) and glucagon‐like peptide‐1 receptor agonists (GLP‐1RA)—on BMI reduction and how that translates to postoperative outcomes in general surgery patients. Patients undergoing general surgery (2016–2024) were identified in the Epic Cosmos database. GLP‐1RA or MBS exposure occurred 1–3 years preoperatively (GLP‐1RA coverage ≥ 80%). Entropy balancing produced weighted cohorts with similar baseline profiles, followed by multivariable regression models assessing the association between weight loss intervention and BMI change, and the impact of BMI on postoperative outcomes. Overall 9470 individuals underwent a general surgery procedure. Median patient age was 64, with mostly females (60.6%). More patients received GLP‐1RA (n = 7823, 82.3%) than MBS (n = 1647, 17.4%). MBS patients had higher initial BMIs (≥ 40: 60.8% vs. 24.5%, p < 0.001). MBS led to greater BMI reduction than GLP‐1RA (Mean difference: −9.89, 95% CI: −9.64, −10.34). Higher BMI at time of a general surgical procedure correlated with increased postoperative complications (OR: 1.01, 95% CI: 1.00–1.01) and extended LOS (OR: 1.01, 95% CI: 1.00–1.01). MBS was associated with lower complication odds (OR: 0.87, 95% CI: 0.78–0.98). MBS improved surgical outcomes in patients with obesity and T2DM through greater BMI reduction compared with GLP‐1RAs. These findings support the role of preoperative weight loss to mitigate surgical risk; however, evaluating outcomes relative to no intervention remains an important future direction.
Keywords: bariatric surgery, GLP‐1 receptor agonists, obesity, surgical outcomes, type 2 diabetes mellitus
1. Introduction
Obesity and diabetes mellitus (DM) constitute growing public health challenges in the United States (US) and globally [1]. The prevalence of obesity has been rising steadily, reaching 41.9% in the US [2]. Concurrently, DM has continued to surge worldwide, with an estimated 38.1 million individuals affected in the US alone [3]. These conditions markedly affect the mental and physical health of an individual, increasing the risk of non‐communicable diseases and overall mortality [1, 4, 5]. Significantly, patients with Type 2 diabetes mellitus (T2DM) and obesity face higher risks of surgical and anaesthetic complications, with T2DM also associated with a higher risk of readmissions and death [6]. These findings highlight the need for effective preoperative risk‐stratification, surgical planning and perioperative management to minimise complications.
Currently, two principal strategies are available to induce significant weight loss that could help achieve improved outcomes among patients with obesity and T2DM: Metabolic bariatric surgery (MBS) and glucagon‐like peptide‐1 receptor agonists (GLP‐1RA), both of which can improve T2DM and obesity [7, 8]. GLP‐1RAs enhance glucose regulation, as well as contribute to non‐glycemic health improvements including weight loss, blood pressure control, dyslipidemia and improved cardiac function [9]. Similarly, across contemporary multicenter data, 5‐year total body weight loss percentage averages between 22% and 27% after MBS procedure [10]. Moreover, MBS has been associated with lowering haemoglobin A1C due to improved glycemic control, while also improving other obesity‐related comorbidities like cardiovascular disease and obesity‐related cancer [11]. Furthermore, cost‐effectiveness analyses suggest that although less invasive, GLP‐1RAs may be less economically favourable compared with MBS when sustained weight loss and long‐term complication avoidance are considered [12]. MBS and GLP‐1RAs represent two unique strategies, sometimes ideally used together, to address obesity, with GLP‐1RAs being less invasive, while MBS provides a more effective, long‐term solution for severe obesity [7, 13].
Weight changes, along with the concomitant metabolic and physiological effects, may influence subsequent non‐MBS surgical outcomes among patients with T2DM and obesity. In particular, weight loss may result in improved in‐hospital outcomes for patients with antecedent Class III obesity [14]. To date, data on previous MBS versus GLP‐1 treatment relative to subsequent perioperative outcomes among patients undergoing non‐bariatric general surgical procedures has not been examined. Therefore, the objective of the current study was to characterise the impact of MBS versus GLP‐1RA therapy on achieving weight loss, as well as define how the associated extent of preoperative weight loss and subsequent body mass index (BMI) impacted postoperative outcomes following non‐MBS surgical procedures.
2. Materials and Methods
2.1. Data Source and Study Population
The Epic Cosmos database was queried via the Epic System Corporation's Cosmos research platform [15]. Epic Cosmos is an application that aggregates electronic health record data submitted voluntarily by health systems across the United States for research purposes. As such, Cosmos data are representative of national census data with the dataset including 296 million unique patients across 41 000 hospitals and clinics [15]. Patients with records at more than one institution were de‐duplicated and anonymised centrally by Epic. However, health systems' inclusion in Epic Cosmos is voluntary in nature and may contribute towards a selection bias.
The Epic Cosmos database was queried using International Classification of Diseases, ninth and tenth editions (ICD‐9/10) and Current Procedural Terminology (CPT) codes to obtain data on patients who underwent a general surgery operative procedure between 1 January 2016 and 31 December 2024. This time period was selected based upon more limited data fields being available prior to 2016. Inclusion criteria consisted of adults (aged ≥ 18 years) with a diagnosis of T2DM and BMI ≥ 30 kg/m2 who underwent a general surgery procedure. General surgery procedures included operations of the lung, liver, gallbladder, pancreas, small bowel, appendix, colon, rectum and hernia operations. Patients undergoing gastric surgery were excluded to avoid including patients undergoing revision or follow‐up surgeries after primary bariatric surgery. The Institutional Review Board at Ohio State University approved this study and waived the requirement for informed consent.
T2DM diagnosis and obesity (BMI ≥ 30 kg/m2) were identified in inpatient and outpatient claims using ICD‐10 diagnosis codes [13]. A BMI cutoff of ≥ 30 kg/m2 was selected to include all obesity classes, acknowledging that although standard BMI‐based indications for MBS require BMI ≥ 35 kg/m2, patients with BMI ≥ 30 kg/m2 and uncontrolled diabetes may also be eligible [16]. Individuals were identified as having T2DM if the individual had at least one diagnosis claim for T2DM in the 2 years before intervention. To identify pre‐treatment obesity, individuals had to have at least one BMI claim in the year before treatment. For individuals with more than one BMI claim within the prior year, the claim closest to the start of intervention was selected.
2.2. Exposure
The primary exposure was weight loss intervention, defined as either MBS or treatment with a GLP‐1RA. The GLP1‐RA group included individuals prescribed semaglutide, liraglutide, dulaglutide, exenatide, tirzepatide or lixisenatide identified through National Drug Codes in the database [9]. To ensure sustained exposure, patients were required to have continuous prescriptions for at least one year prior to the index general surgery with medication coverage for at least 80% of days within that period [13, 17]. MBS procedures were identified through inpatient admissions and outpatient service claims using CPT, ICD‐9 and ICD‐10 procedure codes. Eligible MBS procedures included sleeve gastrectomy, Roux‐en‐Y gastric bypass (RYGB) and adjustable gastric banding [17].
To standardise the timing of intervention relative to the index surgery, only patients who underwent MBS or initiated GLP‐1RA treatment within a defined two‐year period were included. Specifically, the intervention (either MBS or the initiation of GLP‐1RA) had to take place within a specific 2‐year window, occurring at least 1 year before the index surgery but no more than 3 years prior. This approach ensured that every patient had at least a 1‐year window between the intervention and the index general surgical procedure while still capturing a comparable timeframe of intervention exposure across both groups. Prior studies have also noted a significant change in BMI among patients using GLP‐1RA for 1–3 years, forming the basis to assess the effect of GLP‐1RA on outcomes after a similar timeframe usage [18, 19, 20]. Patients with overlapping treatment—defined as patients who received both MBS and GLP‐1RA at any point during the study period—were excluded, as the primary objective was to compare outcomes among individuals who underwent either MBS or GLP‐1RA as the sole weight loss intervention (Figure S1) [13].
2.3. Covariates and Outcomes of Interest
Covariates included patient age, sex, marital status, Charlson Comorbidity Index (CCI), race/ethnicity (categorised as White, Black, Asian or Other [with the ‘Other’ category encompassing Hispanic, American Indian and Alaska Native]), area of residence (metropolitan vs. non‐Metropolitan), social vulnerability index (SVI), presence of malignancy, type of surgery, time to surgery and whether surgery occurred during an admission initiated via emergency department (ED) visit. Obesity‐related comorbidities were identified using ICD‐10 codes and reported (Table S1). Data on SVI were obtained from the Center for Disease Control and the Agency for Toxic Substances and Disease Registry [21]. SVI is a validated measure, developed by CDC that uses 18 community factors and evaluates the susceptibility of communities to external stressors.
The primary outcomes were change in BMI (kg/m2) after an intervention, current BMI (defined as post‐intervention BMI at the time of index non‐MBS surgery) and postoperative surgical complications after index non‐MBS procedure. Change in BMI was defined as pre‐intervention BMI subtracted from current BMI. The incidence of postoperative complications was determined using ICD‐10 diagnosis codes [22]. Secondary outcomes included extended length of stay (LOS), mortality and readmission. Extended LOS was defined as an inpatient hospital stay greater than the 75th percentile.
2.4. Statistical Analysis
Descriptive statistics were presented as median values with interquartile range (IQR) for continuous variables and as frequency (%) for categorical variables. Differences in baseline characteristics were assessed using the Wilcoxon Rank‐Sum for continuous variables and either the chi‐square test or Fisher's exact test for categorical variables. To account for baseline differences in sociodemographic and clinical characteristics among patients, entropy balancing (EB) was used to generate balanced groups. Covariates included in the EB were patient age, sex, race, CCI, SVI, type and year of surgery and time from obesity intervention to non‐MBS general surgery procedure. To ensure no residual differences between the two groups, baseline characteristics were compared before and after EB adjustment (Table S2). Subsequently, any covariates that were not balanced between the two cohorts were added to multivariable regression models. Although even with multivariable models and EB for various factors, some unmeasured clinical factors likely remained unaccounted for and could contribute towards outcome.
To address potential heterogeneity by intervention type, additional subgroup analyses were conducted. These included separate comparisons of RYGB and sleeve gastrectomy procedures with GLP‐1RA users limited to semaglutide or tirzepatide. GLP‐1RA was limited to these two newer medication types, as these are the agents approved for weight management and associated with greater efficacy in BMI reduction compared to earlier GLP‐1RAs. These analyses aimed to assess whether surgical subtype or pharmacologic agent materially altered the observed associations with perioperative outcomes and weight loss. EB was applied using the previously described covariates, with the addition of malignancy status and whether the procedure followed an ED visit to ensure balanced comparisons across intervention groups.
Multivariable logistic regression models were utilised to examine the association between changes in BMI and postoperative outcomes, as well as the association between type of weight loss intervention and postoperative outcomes; odds ratios (OR) and 95% confidence intervals (CI) were reported. The models were adjusted for age, sex, ethnicity, region, area, procedure type, CCI, SVI and time to surgery. To facilitate stable estimation in regression models, baseline BMI was modelled as a binary variable (< 35 vs. ≥ 35 kg/m2), which ensured adequate sample sizes [23]. Statistical tests were conducted using a two‐tailed approach at a significance level of p < 0.05. Analyses were performed using the SAS 9.4 (SAS Institute). EB was performed using Stata, version 18 (StataCorp LLC, College Station, TX).
3. Results
3.1. Baseline Characteristics
A total of 9470 patients with T2DM and obesity underwent a general non‐MBS procedure after a weight loss intervention (lung: 590, 6.2%; liver: n = 135, 1.4%; gallbladder: n = 2522, 26.6%; pancreas: n = 218, 2.3%; small bowel: n = 709, 7.5%; appendix: n = 648, 6.8%; colon: n = 1982, 20.9%; rectum: n = 450, 4.8%; ventral hernia: n = 2216, 23.4%) (Table 1). Median patient age was 64 years (IQR: 55–71) with most patients being female (n = 5736, 60.6%). The majority of patients had a high CCI score (CCI ≥ 2: n = 8846, 93.4%), resided in metropolitan areas (n = 7873, 83.4%) and self‐identified as White (n = 7009, 74.5%). Overall, most patients had BMI ≥ 45 kg/m2 (n = 2916, 30.8%) followed by BMI 35–39 kg/m2 (n = 2367, 25.0%), BMI 30–34 kg/m2 (n = 2255, 23.8%) and BMI 40–44 kg/m2 (n = 1932, 20.4%). Median time to surgery from weight loss intervention to the index non‐MBS surgical procedure was 653 days (IQR: 496–850). Most patients received GLP1‐RA (n = 7823, 82.3%); a subset of patients underwent MBS (n = 1647, 17.4%). Patients who received GLP‐1RA were more likely to be older (GLP1‐RA: 65 years [IQR 56–72] vs. MBS: 57 years [IQR 48–65]) and had more comorbidities (CCI > 2; GLP: 93.8% vs. MBS: 91.4%) (both p < 0.001). Additionally, White patients were more likely to receive GLP1‐RA (GLP: 75.4% vs. MBS: 69.9%), while Black patients were more likely to undergo MBS (GLP: 19.2% vs. MBS: 24.5%) (all p < 0.001) (Table 1). Time from obesity intervention to the index non‐MBS surgical procedure was comparable (GLP: 656 [IQR 497–853] vs. MBS: 637 [IQR 485–845] days; p = 0.052).
TABLE 1.
Baseline characteristics of patients.
| Characteristics | Total (N = 9470) | GLP‐1RA (N = 7823, 82.6%) | Bariatric surgery (N = 1647, 17.4) | p |
|---|---|---|---|---|
| Age | 64 (55–71) | 65 (56–72) | 57 (48–65) | < 0.001 |
| Sex | < 0.001 | |||
| Female | 5736 (60.6) | 4489 (57.4) | 1247 (75.7) | |
| Male | 3734 (39.4) | 3334 (42.6) | 400 (24.3) | |
| Marital status | 0.047 | |||
| Married | 4986 (53.1) | 4155 (53.6) | 831 (50.9) | |
| Other | 4401 (46.9) | 3599 (46.4) | 802 (49.1) | |
| CCI | < 0.001 | |||
| ≤ 2 | 624 (6.6) | 482 (6.2) | 142 (8.6) | |
| > 2 | 8846 (93.4) | 7341 (93.8) | 1505 (91.4) | |
| Race | < 0.001 | |||
| White | 7009 (74.5) | 5863 (75.4) | 1146 (69.9) | |
| Black | 1896 (20.1) | 1495 (19.2) | 401 (24.5) | |
| Asian | 98 (1.0) | 90 (1.2) | 8 (0.5) | |
| Other | 408 (4.3) | 323 (4.2) | 85 (5.2) | |
| SVI | < 0.001 | |||
| Low | 3106 (33.0) | 2462 (34.0) | 464 (28.4) | |
| Moderate | 3202 (34.0) | 2628 (33.8) | 574 (35.1) | |
| High | 3101 (33.0) | 2503 (32.2) | 598 (36.6) | |
| Area | ||||
| Non‐metropolitan | 1561 (16.6) | 1297 (16.7) | 264 (16.1) | |
| Metropolitan | 7863 (83.4) | 6488 (83.3) | 1375 (83.9) | |
| Type of surgery | < 0.001 | |||
| Appendix | 648 (6.8) | 565 (7.2) | 83 (5.0) | |
| Colon | 1982 (20.9) | 1832 (23.4) | 150 (9.1) | |
| Gallbladder | 2522 (26.6) | 2084 (26.6) | 438 (26.6) | |
| Hernia | 2216 (23.4) | 1616 (20.7) | 600 (36.4) | |
| Liver | 135 (1.4) | 131 (1.7) | 4 (0.2) | |
| Lung | 590 (6.2) | 562 (7.2) | 28 (1.7) | |
| Pancreas | 218 (2.3) | 167 (2.1) | 51 (3.1) | |
| Rectum | 450 (4.8) | 362 (4.6) | 88 (5.3) | |
| Small bowel | 709 (7.5) | 504 (6.4) | 205 (12.4) | |
| Presence of Malignancy | 970 (10.2) | 912 (11.7) | 58 (3.5) | < 0.001 |
| ED visit a | 4273 (45.1) | 3594 (45.9) | 679 (41.2) | < 0.001 |
| Time to surgery, d | 653 (496–850) | 656 (497–853) | 637 (485–845) | 0.052 |
Abbreviations: ED, Emergency Department; GLP‐1RA, Glucagon‐like peptide‐1 receptor agonist; SVI, Social Vulnerability Index.
Admission leading to index surgery initiated via ED.
3.2. Weight Loss Intervention and BMI
Patients who underwent MBS were more likely to have an initially higher BMI (BMI ≥ 40: 60.8% vs. 24.5%) pre‐intervention versus individuals who received GLP1‐RA (p < 0.001). Following weight loss intervention, patients experienced a wide range in BMI decrease (ΔBMI 0–2: n = 2860, 30.2%, ΔBMI 3–5: n = 1648, 17.4%, ΔBMI 6–9: n = 2005, 21.2%, ΔBMI ≥ 10: n = 1756, 18.5%); a small subset of patients had an increase in BMI (n = 1201, 12.7%) (Table S3) (Figure 1). Patients who underwent MBS were more likely to experience a decrease in BMI of ≥ 10 units (MBS: 72.4% vs. GLP1‐RA: 7.2%) and were less likely to have an increase in BMI (MBS: 1.0% vs. GLP1‐RA: 15.1%) (both p < 0.001). On EB weighted analysis, MBS remained independently associated with higher BMI loss versus GLP1‐RA treatment (mean difference: −9.89, 95% CI: −9.64, −10.34; p < 0.001). On stratified analysis relative to baseline BMI, MBS was consistently associated with greater weight loss across all groups with the largest difference in pre‐ versus post‐intervention among individuals with an initial BMI ≥ 40 kg/m2 (ΔBMI: −11.61, 95% CI: −12.12, −11.05) (Table 2).
FIGURE 1.
Plot showing change in BMI by weight loss intervention type (GLP‐1RA and MBS), stratified by baseline BMI of patients.
TABLE 2.
Change in BMI after intervention, stratified by baseline BMI, adjusted by EB weighted ANCOVA model.
| Baseline BMI | Bariatric surgery (ref. GLP‐1RA) mean difference | 95% CI | p |
|---|---|---|---|
| Overall | −9.99 | −9.64, −10.34 | < 0.001 |
| 30–34 | −3.91 | −4.70, −3.12 | < 0.001 |
| 35–39 | −7.38 | −7.85, −6.90 | < 0.001 |
| 40–44 | −8.47 | −9.01, −7.94 | < 0.001 |
| ≥ 45 | −11.61 | −12.12, −11.05 | < 0.001 |
Abbreviations: BMI, body mass index; GLP‐1RA, glucagon‐like peptide‐1 receptor agonist.
3.3. Surgical Outcomes and BMI
After the non‐MBS index surgical procedure, 48.2% (n = 4570) of patients experienced at least one complication, with 24.4% (n = 2300) of patients having an extended LOS; overall, the incidence of readmission was 12.4% (n = 1972) and 2.0% (n = 190) of patients died within 30 days of the non‐MBS surgical procedure (Table S4).
On adjusted EB analysis, after controlling for competing risk factors, a higher BMI at the time of non‐MBS general surgical procedure was independently associated with increased odds of a postoperative complication (OR: 1.01, 95% CI: 1.00–1.01) and extended LOS (OR 1.01, 95% CI: 1.00–1.01) (Table 3). Of note, a history of prior MBS versus GLP‐1RA treatment was independently associated with lower odds of postoperative complications (OR: 0.87, 95% CI: 0.78–0.98), lower risk for extended LOS (OR: 0.57, 95% CI: 0.49–0.67) and readmission (OR: 0.81, 95% CI: 0.69–0.95) following a subsequent non‐MBS general surgical procedure (Table S5) (Figure 2). To better understand the impact of prior MBS versus GLP‐1RA treatment on subsequent post‐operative non‐MBS surgical outcomes, change in BMI (∆ BMI) was examined. Of note, on adjusted EB analysis, the decrease in BMI post obesity treatment was associated with the odds of complications after the subsequent non‐MBS surgical procedure among patients with very high initial BMI (Figure 3). Specifically, patients with a BMI of 30–34 prior to obesity intervention had no difference in odds of complications relative to ∆BMI (ref. ∆BMI > 0, −2 ≤ ΔBMI < 0: OR: 0.98, 95% CI: 0.76–1.26; −9 ≤ ΔBMI ≤ −6: OR: 1.15, 95% CI: 0.84–1.59; ∆BMI: ≤ −10: OR 1.13, 95% CI: 0.57–2.29). In contrast, among patients with a higher baseline BMI of ≥ 35, ∆BMI was associated with decreased odds of complications following a subsequent non‐MBS surgical procedure (ref. ∆BMI > 0, −2 ≤ ΔBMI < 0: OR 0.78, 95% CI: 0.66–0.93; −9 ≤ ΔBMI ≤ −6: OR: 0.83, 95% CI: 0.70–0.99; ∆BMI: ≤ −10: OR 0.81, 95% CI: 0.68–0.96) (Table S6).
TABLE 3.
Multivariable analysis examining the association between current BMI and surgical outcomes.
| Outcomes | BMI (continuous) odds ratio | 95% CI | p |
|---|---|---|---|
| Complications | |||
| 30 days | 1.01 | (1.00–1.01) | 0.018 |
| 90 days | 1.01 | (1.00–1.01) | < 0.001 |
| Readmissions | |||
| 30 Days | 1 | (0.99–1.01) | 0.836 |
| 90 Days | 1 | (0.99–1.01) | 0.766 |
| Extended LOS | 1.01 | (1.00–1.01) | 0.014 |
Abbreviation: LOS, length of stay.
FIGURE 2.
Risk‐adjusted odds of post‐operative complications among patients by type of weight loss intervention (bariatric metabolic surgery and glucagon‐like peptide‐1 receptor agonist).
FIGURE 3.
Adjusted odds of post‐operative complications by change in BMI relative to pre‐intervention baseline BMI.
In sensitivity analyses, RYGB (n = 686) and sleeve gastrectomy (n = 860) were compared to GLP‐1RA therapy restricted to semaglutide and tirzepatide. This subset of GLP‐1RA use accounted for 1690 patients (34.4% of GLP‐1RA users). Both surgical subgroups demonstrated outcomes consistent with the overall cohort, with RYGB and sleeve gastrectomy each independently associated with greater reductions in BMI (Table S7). Furthermore, similar trends were observed in postoperative outcomes following non‐bariatric general surgery, as both bariatric surgical groups exhibited lower odds of postoperative complications and extended LOS compared to the GLP‐1RA group (Table S8).
4. Discussion
Patients with T2DM and obesity often experience elevated peri‐operative risks following a general surgical procedure, yet the optimal preoperative weight loss strategy to limit these risks remains unclear. In particular, obesity has been associated with a higher incidence of postoperative complications including wound infections, venous thromboembolism and renal failure [24]. Similarly, diabetes is an independent risk factor associated with increased risk of postoperative complications, prolonged hospitalisation and postoperative mortality [25]. As such, surgeons often have reservations about offering patients with severe obesity and T2DM an elective surgical procedure [6]. Rather, patients with obesity are often recommended weight management prior to an elective non‐MBS procedure in an attempt to improve outcomes. The use of GLP1‐RA has greatly expanded over the last 3–5 years with increasing numbers of patients utilising this treatment strategy to lose weight. To date, whether pharmacological weight loss or bariatric surgery represents the preferred care pathway towards other elective surgical interventions remains poorly defined. The current study was important as we specifically examined post‐operative outcomes among patients undergoing a non‐MBS general surgical procedure relative to prior obesity management with either GLP1‐RA or MBS. Perhaps not surprisingly, a higher BMI at the time of the non‐MBS general surgical procedure was associated with higher odds of complications. While the observed OR of 1.01 may appear modest, it indicates that each one‐unit increase in BMI is associated with a 1% increase in the odds of complications, an effect that becomes more meaningful with larger BMI differences. Of more note, a history of prior MBS versus GLP‐1RA treatment was independently associated with lower odds of postoperative complications, extended LOS and readmission following a subsequent non‐MBS general surgical procedure. Selection bias may have persisted, however, in choice of weight loss intervention arising from patient preference, provider practice and unmeasured patient characteristics. The effect of MBS on peri‐operative outcomes in the current study, however, was largely mediated by the change in BMI. Specifically, MBS was associated with higher BMI loss versus GLP1‐RA treatment, with MBS having the largest weight loss difference among patients with an initial BMI ≥ 45 kg/m2 (Table 2). In turn, ∆BMI was associated with decreased odds of complications following a subsequent non‐MBS surgical procedure. However, odds of complications were generally consistent across weight loss strata, the similarity in effect sizes suggests that even modest weight loss may confer benefit, and a clear dose–response was not observed. These data highlight the importance of a tailored approach to obesity management in the care of patients anticipating elective surgical procedures. While GLP1‐RA treatment may be adequate for patients with lower BMI, higher BMI patients benefit more from MBS to achieve loss in BMI, which was associated with decreased risk of adverse perioperative outcomes following a subsequent elective surgical procedure.
GLP1‐RA have emerged in recent years as a non‐surgical tool in weight management programs. In particular, several drugs in the GLP1‐RA class have received FDA approval for obesity treatment based on data demonstrating high efficacy to induce weight loss [26]. However, GLP‐1RA agents vary in efficacy. Semaglutide and tirzepatide produce substantially more weight loss than liraglutide, as reported in both RCTs and real‐world data. For example, in the STEP 8 trial, semaglutide 2.4 mg resulted in higher mean weight loss versus liraglutide 3.0 mg [27]. A recent real‐world study also noted that tirzepatide users had higher rates of ≥ 15% weight loss than semaglutide users [28]. In the current cohort, heterogeneity among agents may partly explain the wide range of weight‐loss responses among GLP‐1RA users. GLP1‐RA medications not only provide weight loss options but can also optimise glycemic control [26]. While some side effects like delayed gastric emptying have been reported with GLP1‐RA, emerging evidence indicates that perioperative use of GLP1‐RA is safe and does not worsen postoperative outcomes [22, 29]. Recent studies in patients with DM have failed to not an increased risk of postoperative respiratory complications among GLP‐1RA users versus non‐users among patients undergoing emergency surgery; similarly, data from an EHR‐based cohort reported no elevated risk of perioperative aspiration, hypoglycemia or mortality associated with GLP‐1RA exposure [30, 31]. These data suggested that GLP‐1RAs may be useful as a means to manage obesity and DM chronically, as well as potentially help improve a patient's surgical fitness to assist in lowering the risk of postoperative complications for non‐MBS procedures through weight loss. Weight loss interventions can result in a wide variety of BMI changes, however, as noted in the current study (ΔBMI 0–2: n = 2860, 30.2%, ΔBMI 3–5: n = 1648, 17.4%, ΔBMI 6–9: n = 2005, 21.2%, ΔBMI ≥ 10: n = 1756, 18.5%) (Table S3) (Figure 1). Interestingly, patients treated with GLP1‐RA were less likely to experience a decrease in BMI of ≥ 10 units (MBS: 72.4% vs. GLP1‐RA: 7.2%) and were more likely to have an increase in BMI (MBS: 1.0% vs. GLP1‐RA: 15.1%) (both p < 0.001). These data emphasised how GLP1‐RA treatment may not be effective in subsets of patients with obesity.
MBS is a well‐established obesity management modality that has proven effectiveness to treat severe obesity with significant and durable weight loss [32]. In fact, surgical bariatric interventions can be associated with 20%–30% weight loss, along with the profound improvements in metabolic control [32]. These metabolic benefits, including glycemic normalisation and improvement of other comorbidities, have made MBS as an effective treatment option for obesity‐related T2DM. In addition, following MBS, patients may also have improvements in other chronic conditions including hypertension, dyslipidemia and sleep apnea [6]. Current guidelines generally reserve MBS for patients with Class II–III obesity (BMI ≥ 40 or ≥ 35 with significant comorbid conditions), although recent recommendations recommend surgery for patients with BMI 30–34.9 if T2DM is poorly controlled [33]. In the current study, on EB weighted analysis, MBS was independently associated with higher BMI loss versus GLP1‐RA treatment (mean difference: −9.89, 95% CI: −9.64, −10.34; p < 0.001). In addition, on stratified analysis relative to baseline BMI, MBS was consistently associated with greater weight loss across all groups with the largest difference in pre‐ versus post‐intervention among individuals with an initial BMI ≥ 45 kg/m2 (ΔBMI: −11.61, 95% CI −12.12, −11.05) (Table 2). Collectively, these data highlight how MBS should be the preferred approach among patients with higher BMI to ensure a larger ΔBMI, especially among patients anticipating a non‐MBS surgical procedure [26, 29].
The morbidity of an elective surgical procedure can be significant among patients with obesity, as demonstrated by the fact that one in two patients in the current study experienced at least one complication and one in four had an extended LOS; in addition, overall 30‐day mortality was 2%. Perhaps not surprisingly, in the current study, after controlling for competing risk factors, a higher BMI at the time of non‐MBS general surgical procedure was independently associated with increased odds of a postoperative complication and extended LOS (Table 3). These data were consistent with prior reports noting that high BMI was associated with increased perioperative morbidity and mortality [34, 35]. To mitigate obesity‐related risks related to general surgical procedures, some patients are recommended weight management programs in an attempt to decrease their BMI. The current study was novel in that we specifically compared the two predominant weight management approaches (MBS vs. GLP1‐RA) to examine weight loss relative to subsequent non‐MBS surgical outcomes. MBS was associated with greater reductions in BMI versus GLP‐1RA treatment across all baseline BMI strata, with the gap most pronounced in patients with baseline BMI ≥ 40 kg/m2 (Figure 1). Notably, the weight loss associated with MBS translated into improved surgical outcomes. Specifically, after adjustment for patient factors EB, prior MBS was independently associated with lower odds of postoperative complications, shorter hospital stays and reduced readmissions following a subsequent non‐MBS surgical procedure. MBS may exert weight‐independent metabolic effects—such as enhanced insulin sensitivity and reduced systemic inflammation—that contribute to reduced surgical risk [30, 36]. Further analysis revealed each decrease in BMI was associated with incremental lower odds of postoperative complications, underlining that reduction in weight loss was the key modifier of surgical risk. In addition to data in the current study, others have reported that MBS can also be more cost‐effective in the long term compared with the cumulative costs of GLP‐1RA therapy [37, 38]. Given that BMI reduction was independently associated with improved outcomes in our analysis, it is possible that future pharmacologic therapies capable of achieving comparable weight loss to MBS could yield similar reductions in postoperative complications. Currently, real‐world effectiveness of GLP‐1RA therapy is limited by issues of access and adherence, with one large US cohort reporting that nearly half of GLP‐1RA users discontinued use within 6 months [39]. Thus, while MBS and GLP‐1RA both represent effective weight loss strategies, an informed individualised decision‐making approach is needed. The integration of targeted weight loss into preoperative planning, especially among high‐risk patients with obesity, who would benefit from a large ΔBMI may benefit more from MBS. As such, clinicians should consider the patient's baseline BMI and the desired weight loss outcomes when choosing between pharmacological and surgical weight management strategies to optimise subsequent non‐MBS perioperative outcomes. Furthermore, future research should include prospective, head‐to‐head studies that compare structured preoperative weight‐loss strategies—such as MBS, GLP‐1RA therapy or standard care—among high‐risk surgical candidates. Additional work is also needed to evaluate the role of glycemic control (e.g., HbA1c), systemic inflammation and body composition as mediators of surgical outcomes. Cost‐effectiveness analyses linking weight‐loss modality, clinical outcomes and healthcare costs will also be essential to guide evidence‐based surgical optimisation.
The findings of the current study should be interpreted in light of several limitations. Similar to other studies using large national registries, the analyses may have been affected by coding errors, inaccurate data entry and incomplete information. Moreover, data on the complexity of the surgical procedure, intraoperative variables, ASA physical status, provider/institute level factors (experience or volume of procedures), lifestyle recommendations or granular social determinants of health were not available. While included to account for overall socioeconomic context, SVI did capture individual‐level factors such as insurance type or coverage variability. Although CCI was used to account for overall comorbidity burden, severity of specific obesity‐related comorbidities could not be assessed, as conditions were identified using ICD‐10 diagnosis codes. Similarly, postoperative complications were identified using ICD‐10 diagnosis codes without standardised severity grading, which may have under‐represented milder events and precluded detailed severity stratification. While glycemic control may independently influence surgical outcomes, assessment of its impact was not possible due to unavailable or missing HbA1c data in the Epic Cosmos dataset; therefore, the relative contribution of improved glycemic control remains unclear. Similarly, incomplete longitudinal weight records precluded calculation of percent total body weight loss, limiting further characterisation of weight trajectories. Additionally, dosage information for GLP‐1RA therapy was not consistently available and could not be accounted for in the analysis, limiting differentiation between high‐dose and low‐dose regimens. In addition, only patients who had documented care within the Epic Cosmos network were included; if a patient received care at a non‐Epic hospital, those events would not be recorded in the database. Similarly, participation of health systems in Epic Cosmos was voluntary, which may have contributed to potential selection bias, limiting the generalizability to an extent. Direct cost data or detailed resource‐use measures were not available in the Epic Cosmos dataset, which precluded the study from assessing economic outcomes, which could be important in the clinical setting. Lastly, procedures like ventral hernia repair may be more common after bariatric surgery, potentially influencing associations; however, procedure type was included in EB weighting to mitigate this risk. Confounding between patients who received GLP‐1 RA and MBS was mitigated using EB with well‐matched cohorts; however, residual confounding was still possible.
In conclusion, preoperative weight loss is an important modifiable risk factor among patients with obesity and T2DM who have plans to undergo a non‐MBS general surgical procedure. MBS improved surgical outcomes among patients with obesity and T2DM through a more substantial loss in BMI compared with GLP‐1RA. Implementing structured weight loss programs before elective surgery can substantially mitigate surgical morbidity among high‐risk obese populations. To this point, MBS was independently associated with a lower risk of complications, extended LOS as well as readmission following a general surgical procedure. Data in the current study should help inform the choice between GLP‐1RA or MBS treatment in the care pathway of patients undergoing a future general surgical procedure.
Funding
The authors have nothing to report.
Conflicts of Interest
The authors declare no conflicts of interest.
Supporting information
Figure S1: Flowchart displaying derivation of the study cohort, and quantifying excluded patients based on the set criteria.
Appendix S1: cob70075‐sup‐0002‐Appendix.docx.
Table S1: Obesity related comorbidities stratified by intervention group.
Table S2: Baseline characteristics after entropy balancing across intervention groups.
Table S3: cob70075‐sup‐0003‐Tables.docx.
Table S4: cob70075‐sup‐0003‐Tables.docx.
Table S5: cob70075‐sup‐0003‐Tables.docx.
Table S6: cob70075‐sup‐0003‐Tables.docx.
Table S7: cob70075‐sup‐0003‐Tables.docx.
Table S8: cob70075‐sup‐0003‐Tables.docx.
Acknowledgements
The authors have nothing to report.
Data Availability Statement
The data for this study were obtained from the Epic Cosmos database. There are restrictions to the availability of this data, which is used under licence for this study.
References
- 1. Ritchie H. and Roser M., Obesity (Our World Data, 2017), https://ourworldindata.org/obesity. [Google Scholar]
- 2. “Adult Obesity Facts,” (2025), https://www.cdc.gov/obesity/adult‐obesity‐facts/index.html.
- 3. CDC , Centers for Disease Control National Diabetes Statistics Report (Diabetes, 2024), https://www.cdc.gov/diabetes/php/data‐research/index.html. [Google Scholar]
- 4. Leutner M., Dervic E., Bellach L., Klimek P., Thurner S., and Kautzky A., “Obesity as Pleiotropic Risk State for Metabolic and Mental Health Throughout Life,” Translational Psychiatry 13, no. 1 (2023): 175, 10.1038/s41398-023-02447-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Brown O. I., Drozd M., McGowan H., et al., “Relationship Among Diabetes, Obesity, and Cardiovascular Disease Phenotypes: A UK Biobank Cohort Study,” Diabetes Care 46, no. 8 (2023): 1531–1540, 10.2337/dc23-0294. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Zhang X., Hou A., Cao J., et al., “Association of Diabetes Mellitus With Postoperative Complications and Mortality After Non‐Cardiac Surgery: A Meta‐Analysis and Systematic Review,” Frontiers in Endocrinology 13 (2022): 841256, 10.3389/fendo.2022.841256. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Maciejewski M. L., Arterburn D. E., Van Scoyoc L., et al., “Bariatric Surgery and Long‐Term Durability of Weight Loss,” JAMA Surgery 151, no. 11 (2016): 1046–1055, 10.1001/jamasurg.2016.2317. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Webb D. L., Abrahamsson N., Sundbom M., and Hellström P. M., “Bariatric Surgery – Time to Replace With GLP‐1?,” Scandinavian Journal of Gastroenterology 52, no. 6–7 (2017): 635–640, 10.1080/00365521.2017.1293154. [DOI] [PubMed] [Google Scholar]
- 9. Rashid Z., Woldesenbet S., Khalil M., et al., “Impact of Preoperative Glucagon‐Like Peptide‐1 Receptor Agonist on Outcomes Following Major Surgery,” World Journal of Surgery 49, no. 3 (2025): 698–707, 10.1002/wjs.12484. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Biter L. U., ‘T Hart J. W., Noordman B. J., et al., “Long‐Term Effect of Sleeve Gastrectomy vs Roux‐En‐Y Gastric Bypass in People Living With Severe Obesity: A Phase III Multicentre Randomised Controlled Trial (SleeveBypass),” Lancet Regional Health ‐ Europe 38 (2024): 100836, 10.1016/j.lanepe.2024.100836. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Docimo S., Shah J., Warren G., Ganam S., Sujka J., and DuCoin C., “A Cost Comparison of GLP‐1 Receptor Agonists and Bariatric Surgery: What Is the Break Even Point?,” Surgical Endoscopy 38, no. 11 (2024): 6560–6565, 10.1007/s00464-024-11191-1. [DOI] [PubMed] [Google Scholar]
- 12. Huh Y. J., “Cost‐Effectiveness of Obesity Treatments: Glucagon‐Like Peptide‐1 Receptor Agonists, Endoscopic Sleeve Gastroplasty, and Metabolic/Bariatric Surgery,” Journal of Metabolic and Bariatric Surgery 14, no. 2 (2025): 97–105, 10.17476/jmbs.2025.14.2.97. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Edwards M. A., Wall‐Wieler E., Liu Y., Zheng F., and Coviello A., “Out‐Of‐Pocket Costs Among Commercially Insured Individuals With Type 2 Diabetes and Obesity: Comparison Between Ozempic and Sleeve Gastrectomy,” Annals of Surgery 281, no. 2 (2025): 196–200, 10.1097/SLA.0000000000006353. [DOI] [PubMed] [Google Scholar]
- 14. Kassahun W. T., Mehdorn M., and Babel J., “The Impact of Obesity on Surgical Outcomes in Patients Undergoing Emergency Laparotomy for High‐Risk Abdominal Emergencies,” BMC Surgery 22, no. 1 (2022): 15, 10.1186/s12893-022-01466-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. “About|Epic Cosmos,” (2025), https://cosmos.epic.com/about/.
- 16. Eisenberg D., Shikora S. A., Aarts E., et al., “2022 American Society for Metabolic and Bariatric Surgery (ASMBS) and International Federation for the Surgery of Obesity and Metabolic Disorders (IFSO): Indications for Metabolic and Bariatric Surgery,” Surgery for Obesity and Related Diseases 18, no. 12 (2022): 1345–1356, 10.1016/j.soard.2022.08.013. [DOI] [PubMed] [Google Scholar]
- 17. Cotugno M., Nosso G., Saldalamacchia G., et al., “Clinical Efficacy of Bariatric Surgery Versus Liraglutide in Patients With Type 2 Diabetes and Severe Obesity: A 12‐Month Retrospective Evaluation,” Acta Diabetologica 52, no. 2 (2015): 331–336, 10.1007/s00592-014-0644-5. [DOI] [PubMed] [Google Scholar]
- 18. Hemmer A., Maiter D., Buysschaert M., and Preumont V., “Long‐Term Effects of GLP‐1 Receptor Agonists in Type 2 Diabetic Patients: A Retrospective Real‐Life Study in 131 Patients,” Diabetes and Metabolic Syndrome: Clinical Research and Reviews 13, no. 1 (2019): 332–336, 10.1016/j.dsx.2018.09.007. [DOI] [PubMed] [Google Scholar]
- 19. Piccini S., Favacchio G., Panico C., et al., “Time‐Dependent Effect of GLP‐1 Receptor Agonists on Cardiovascular Benefits: A Real‐World Study,” Cardiovascular Diabetology 22, no. 1 (2023): 69, 10.1186/s12933-023-01800-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Siriyotha S., Anothaisintawee T., Looareesuwan P., et al., “Effectiveness of Glucagon‐Like Peptide‐1 Receptor Agonists for Reduction of Body Mass Index and Blood Glucose Control in Patients With Type 2 Diabetes Mellitus and.222 Obesity: A Retrospective Cohort Study and Difference‐In‐Difference Analysis,” BMJ Open 14, no. 11 (2024): e086424, 10.1136/bmjopen-2024-086424. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. “CDC/ATSDR Social Vulnerability Index U.S.,” (2024), https://www.atsdr.cdc.gov/placeandhealth/svi/.
- 22. Khalil M., Woldesenbet S., Munir M. M., et al., “Surgical Outcomes and Healthcare Expenditures Among Patients With Dementia Undergoing Major Surgery,” World Journal of Surgery 48, no. 5 (2024): 1075–1083, 10.1002/wjs.12106. [DOI] [PubMed] [Google Scholar]
- 23. Safaee M. M., Tenorio A., Osorio J. A., et al., “The Impact of Obesity on Perioperative Complications in Patients Undergoing Anterior Lumbar Interbody Fusion,” Journal of Neurosurgery. Spine 33, no. 3 (2020): 332–341, 10.3171/2020.2.SPINE191418. [DOI] [PubMed] [Google Scholar]
- 24. “Study Clarifies Link Between Obesity and Surgical Complications|ACS,” (2025), https://www.facs.org/for‐medical‐professionals/news‐publications/news‐and‐articles/bulletin/2023/october‐2023‐volume‐108‐issue‐10/study‐clarifies‐link‐between‐obesity‐and‐surgical‐complications/.
- 25. Jämsen E., Nevalainen P., Eskelinen A., Huotari K., Kalliovalkama J., and Moilanen T., “Obesity, Diabetes, and Preoperative Hyperglycemia as Predictors of Periprosthetic Joint Infection: A Single‐Center Analysis of 7181 Primary Hip and Knee Replacements for Osteoarthritis,” Journal of Bone and Joint Surgery 94, no. 14 (2012): e101, 10.2106/JBJS.J.01935. [DOI] [PubMed] [Google Scholar]
- 26. Schauer P. R., Bhatt D. L., Kirwan J. P., et al., “Bariatric Surgery Versus Intensive Medical Therapy for Diabetes — 5‐Year Outcomes,” New England Journal of Medicine 376, no. 7 (2017): 641–651, 10.1056/NEJMoa1600869. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Rubino D. M., Greenway F. L., Khalid U., et al., “Effect of Weekly Subcutaneous Semaglutide vs Daily Liraglutide on Body Weight in Adults With Overweight or Obesity Without Diabetes: The STEP 8 Randomized Clinical Trial,” JAMA 327, no. 2 (2022): 138, 10.1001/jama.2021.23619. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Rodriguez P. J., Goodwin Cartwright B. M., Gratzl S., et al., “Semaglutide vs Tirzepatide for Weight Loss in Adults With Overweight or Obesity,” JAMA Internal Medicine 184, no. 9 (2024): 1056–1064, 10.1001/jamainternmed.2024.2525. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Villarreal‐Calderon J. R., Cuellar‐Tamez R., Castillo E. C., Luna‐Ceron E., García‐Rivas G., and Elizondo‐Montemayor L., “Metabolic Shift Precedes the Resolution of Inflammation in a Cohort of Patients Undergoing Bariatric and Metabolic Surgery,” Scientific Reports 11, no. 1 (2021): 12127, 10.1038/s41598-021-91393-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Klonoff D. C., Kim S. H., Galindo R. J., et al., “Risks of Peri‐ and postoperative Complications With Glucagon‐Like Peptide‐1 Receptor Agonists,” Diabetes, Obesity and Metabolism 26, no. 8 (2024): 3128–3136, 10.1111/dom.15636. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Dixit A. A., Bateman B. T., Hawn M. T., Odden M. C., and Sun E. C., “Preoperative GLP‐1 Receptor Agonist Use and Risk of Postoperative Respiratory Complications,” Journal of the American Medical Association 331, no. 19 (2024): 1672–1673, 10.1001/jama.2024.5003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Sarma S. and Palcu P., “Weight Loss Between Glucagon‐Like Peptide‐1 Receptor Agonists and Bariatric Surgery in Adults With Obesity: A Systematic Review and Meta‐Analysis,” Obesity 30, no. 11 (2022): 2111–2121, 10.1002/oby.23563. [DOI] [PubMed] [Google Scholar]
- 33. Griffin S. B., Palmer M. A., Strodl E., Lai R., Burstow M. J., and Ross L. J., “Elective Surgery in Adult Patients With Excess Weight: Can Preoperative Dietary Interventions Improve Surgical Outcomes? A Systematic Review,” Nutrients 13, no. 11 (2021): 3775, 10.3390/nu13113775. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Yeh T. L., Tsai M. C., Tsai W. H., Tu Y. K., and Chien K. L., “Effect of Glucagon‐Like Peptide‐1 Receptor Agonists on Glycemic Control, and Weight Reduction in Adults: A Multivariate Meta‐Analysis,” PLoS One 18, no. 1 (2023): e0278685, 10.1371/journal.pone.0278685. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Wilding J. P. H., Batterham R. L., Calanna S., et al., “Once‐Weekly Semaglutide in Adults With Overweight or Obesity,” New England Journal of Medicine 384, no. 11 (2021): 989–1002, 10.1056/NEJMoa2032183. [DOI] [PubMed] [Google Scholar]
- 36. Ihnat J. M. H., De Baun H., Carrillo G., Dony A., Mukherjee T. J., and Ayyala H. S., “A Systematic Review of the Use of GLP‐1 Receptor Agonists in Surgery,” American Journal of Surgery 240 (2025): 116119, 10.1016/j.amjsurg.2024.116119. [DOI] [PubMed] [Google Scholar]
- 37. Buchwald H., Avidor Y., Braunwald E., et al., “Bariatric Surgery: A Systematic Review and Meta‐Analysis,” JAMA 292, no. 14 (2004): 1724, 10.1001/jama.292.14.1724. [DOI] [PubMed] [Google Scholar]
- 38. Yashi K. and Daley S. F., “Obesity and Type 2 Diabetes,” in StatPearls (StatPearls Publishing, 2025), http://www.ncbi.nlm.nih.gov/books/NBK592412/. [Google Scholar]
- 39. Gleason P. P., Urick B. Y., Marshall L. Z., Friedlander N., Qiu Y., and Leslie R. S., “Real‐World Persistence and Adherence to Glucagon‐Like Peptide‐1 Receptor Agonists Among Obese Commercially Insured Adults Without Diabetes,” Journal of Managed Care and Specialty Pharmacy 30, no. 8 (2024): 860–867, 10.18553/jmcp.2024.23332. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Figure S1: Flowchart displaying derivation of the study cohort, and quantifying excluded patients based on the set criteria.
Appendix S1: cob70075‐sup‐0002‐Appendix.docx.
Table S1: Obesity related comorbidities stratified by intervention group.
Table S2: Baseline characteristics after entropy balancing across intervention groups.
Table S3: cob70075‐sup‐0003‐Tables.docx.
Table S4: cob70075‐sup‐0003‐Tables.docx.
Table S5: cob70075‐sup‐0003‐Tables.docx.
Table S6: cob70075‐sup‐0003‐Tables.docx.
Table S7: cob70075‐sup‐0003‐Tables.docx.
Table S8: cob70075‐sup‐0003‐Tables.docx.
Data Availability Statement
The data for this study were obtained from the Epic Cosmos database. There are restrictions to the availability of this data, which is used under licence for this study.