This cohort study assesses whether an association exists between bariatric surgery and alcohol-related hospitalizations and all-cause mortality in patients at US Veterans Health Administration health centers compared with referral to a weight management program.
Key Points
Question
Is bariatric surgery associated with alcohol use disorder–related hospitalizations compared with referral to a weight management program alone?
Findings
In this cohort study of 7694 patients in the Veterans Health Administration, Roux-en-Y gastric bypass (RYGB) was associated with increased alcohol-related hospitalizations vs sleeve gastrectomy, gastric banding, or referral to a weight management program alone.
Meaning
Results of the study suggest that careful patient selection and alcohol-related counseling are especially critical in patients undergoing RYGB.
Abstract
Importance
Bariatric surgery procedures, in particular Roux-en-Y gastric bypass (RYGB), have been associated with subsequent alcohol-related complications. However, previous studies lack data to account for changes in body mass index (BMI) or alcohol use over time, which are key potential confounders.
Objective
To evaluate the association between RYGB, sleeve gastrectomy, or gastric banding on subsequent alcohol use disorder (AUD)–related hospitalization and all-cause mortality as compared with referral to a weight management program alone.
Design, Setting, and Participants
This cohort study included 127 Veterans Health Administration health centers in the US. Patients who underwent RYGB, sleeve gastrectomy, or gastric banding or who were referred to MOVE!, a weight management program, and had a BMI (calculated as weight in kilograms divided by height in meters squared) of 30 or greater between January 1, 2008, and December 31, 2021, were included in the study.
Exposures
RYGB, sleeve gastrectomy, or gastric banding or referral to the MOVE! program.
Main Outcomes and Measures
The primary outcome was time to AUD-related hospitalization from the time of bariatric surgery or MOVE! referral. The secondary outcome was time to all-cause mortality. Separate propensity scores were created for each pairwise comparison (RYGB vs MOVE! program, RYGB vs sleeve gastrectomy, sleeve gastrectomy vs MOVE!). Sequential Cox regression approaches were used for each pairwise comparison to estimate the relative hazard of the primary outcome in unadjusted, inverse probability treatment weighting (IPTW)–adjusted (generated from the pairwise logistic regression models), and IPTW-adjusted approaches with additional adjustment for time-updating BMI and categorical Alcohol Use Disorders Identification Test-Concise scores.
Results
A total of 1854 patients received RYGB (median [IQR] age, 53 [45-60] years; 1294 men [69.8%]), 4211 received sleeve gastrectomy (median [IQR] age, 52 [44-59] years; 2817 men [66.9%]), 265 received gastric banding (median [IQR] age, 55 [46-61] years; 199 men [75.1%]), and 1364 were referred to MOVE! (median [IQR] age, 59 [49-66] years; 1175 men [86.1%]). In IPTW Cox regression analyses accounting for time-updating alcohol use and BMI, RYGB was associated with an increased hazard of AUD-related hospitalization vs MOVE! (hazard ratio [HR], 1.70; 95% CI, 1.20-2.41; P = .003) and vs sleeve gastrectomy (HR, 1.98; 95% CI, 1.55-2.53; P < .001). There was no significant difference between sleeve gastrectomy and MOVE! (HR, 0.76; 95% CI, 0.56-1.03; P = .08). While RYGB was associated with a reduced mortality risk vs MOVE! (HR, 0.63; 95% CI, 0.49-0.81; P < .001), this association was mitigated by increasing alcohol use over time.
Conclusions and Relevance
This cohort study found that RYGB was associated with an increased risk of AUD-related hospitalizations vs both sleeve gastrectomy and the MOVE! program. The mortality benefit associated with RYGB was diminished by increased alcohol use, highlighting the importance of careful patient selection and alcohol-related counseling for patients undergoing this procedure.
Introduction
Bariatric surgeries are widely used for patients with a body mass index (BMI, calculated as weight in kilograms divided by height in meters squared) of 40 or greater or 35 or greater and associated metabolic comorbidities1,2 to attenuate complications of obesity (eg, diabetes, hypertension, hyperlipidemia,3,4 adverse cardiovascular events,5 and nonalcoholic fatty liver disease–related cirrhosis)6 that are associated with increased health care costs.7 Restrictive procedures, such as sleeve gastrectomy or gastric banding, and mixed restrictive or malabsorptive procedures, such as Roux-en-Y gastric bypass (RYGB), are effective in promoting weight loss and reversing underlying metabolic comorbidities, but they may induce differential alterations in alcohol metabolism. Several national registry and insurance claims database studies identified associations between bariatric surgery and increased risk of alcohol use disorders, alcohol-related hepatitis, and alcohol-related cirrhosis.8,9,10,11,12 However, these studies did not control for longitudinal changes in BMI and alcohol consumption or lacked adequate bariatric or nonsurgical control groups. Recent data suggest alcohol consumption increases during prolonged follow-up of patients who underwent bariatric surgery.13 Thus, it remains unclear if the previously observed increased risk of alcohol-related adverse events is attributable to alterations in metabolism or sensitivity, increased alcohol exposure, or both.
To address this question, we used data from a Veterans Health Administration (VHA) cohort of patients with obesity who received 1 of 3 bariatric surgery procedures or were referred to a weight loss program. The aim of the study was to evaluate the association between bariatric surgery type and key outcomes, including alcohol use disorder (AUD)–related hospitalization and all-cause mortality compared with referral to a weight management program alone.
Methods
Study Design and Data Cohort Creation
This retrospective cohort study used data from the Clinical Data Warehouse, which contains longitudinal patient-level data from 127 US VHA centers. From a source cohort of patients with a BMI of 30 or greater, we identified patients who received 1 of 4 treatment pathways for weight management: (1) RYGB, (2) sleeve gastrectomy, (3) gastric banding, or (4) referral to the VHA weight management program (MOVE!) alone. Patients who received these treatments between January 1, 2008, and December 31, 2021, were included, with the index date noted as the date of bariatric surgery or completion of MOVE! referral. Patients with cirrhosis at the time of MOVE! referral or surgery and those missing baseline BMI and Alcohol Use Disorders Identification Test-Concise (AUDIT-C) scores were excluded. To isolate the impact of specific treatment pathways, patients who moved between pathways (ie, from sleeve gastrectomy to RYGB) were not entered into the cohort. This study received institutional review board approval from the Corporal Michael J. Crescenz Department of Veterans Affairs Medical Center, which waived the requirement for informed consent under the Common Rule. We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.
Exposures and Covariates
The primary exposure was weight management approach. Bariatric surgery procedures were identified using Current Procedural Terminology codes summarized in eTable 1 in the Supplement.14 The VHA Consults Table was queried to identify all completed referrals to MOVE!, a weight management program focused on increasing physical activity and encouraging healthy eating.15 MOVE! includes a baseline assessment to establish goals, followed by several weeks of educational group sessions with tailored follow-up. Detailed data for each patient were collected, including demographic characteristics (age, sex, self-identified race and ethnicity), BMI, comorbidities (hypertension, hyperlipidemia, diabetes, coronary artery disease), and laboratory data (creatinine, total bilirubin, albumin, alanine aminotransferase, aspartate aminotransferase, and hemoglobin A1c levels, and lipid profile). The most recent data available before the index date were used, up to a maximum of 1 year. Baseline AUDIT-C scores were also collected. The AUDIT-C screening tool measures hazardous alcohol use and ranges from 0 to 12, with alcohol misuse represented by scores of 3 or greater for women and 4 or greater for men.16 Veterans Health Administration clinicians are prompted to screen alcohol use and update AUDIT-C scores annually. Both BMI and AUDIT-C data were obtained in a time-updated fashion throughout follow-up (60 months) and were updated at a maximum frequency of every 30 days for BMI and every year for AUDIT-C scores; for windows without updated data, a last value carried forward approach was used. Time-updated hemoglobin A1c level was also obtained. Finally, the number of previous AUD-related hospitalizations in the year before the index date was recorded for each patient.
Outcomes
The primary outcome was time to AUD-related hospitalization, defined using hospitalization International Classification of Diseases, Ninth Revision and International Statistical Classification of Diseases and Related Health Problems, Tenth Revision codes (summarized in eTable 2 in the Supplement). The outcome was adapted from previous studies, including a study by a Centers for Disease Control and Prevention scientific working group, to capture events 100% attributable to alcohol,17,18,19,20 and included codes for alcohol misuse, alcohol dependence, alcoholic gastritis, alcohol-related hepatitis, and alcohol-induced pancreatitis. The secondary outcome was time to all-cause mortality.
Statistical Analysis
Details regarding descriptive statistics and analysis of patterns are found in the eMethods in the Supplement. Patients who underwent gastric banding were excluded from primary analyses given the small sample size and dwindling number of these procedures performed in the modern era. To identify populations of similar patients at baseline, we used inverse probability treatment weighting (IPTW). Separate propensity scores were created for each pairwise comparison (RYGB vs MOVE!, RYGB vs sleeve gastrectomy, sleeve gastrectomy vs MOVE!) using the following covariates in logistic regression models: age, sex, race and ethnicity, BMI, hypertension, hyperlipidemia, diabetes, coronary artery disease, and baseline AUDIT-C score. These variables were selected a priori based on previous literature and knowledge of factors salient to surgical vs nonsurgical weight management decisions.21,22,23,24 Inverse probability treatment weighting was then computed as 1/propensity score for patients who received bariatric surgery (or RYGB in the RYGB vs sleeve gastrectomy comparison) and as 1/(1 − propensity score) for patients who received the alternative.25 Raw and weighted standardized mean differences were computed for each covariate; adequate balance was reflected by a weighted standardized mean difference within 0.1 or more and 0.1 or less.26 If baseline covariate balance was not achieved through IPTW, unbalanced variables were included as covariates in subsequent regression analysis (for sleeve gastrectomy vs MOVE!, models were adjusted for race and ethnicity, hypertension, hyperlipidemia, and diabetes). To evaluate the association between weight management approach and AUD-related hospitalization, we used a time-to-event approach. Unadjusted (ie, no application of IPTW or covariate adjustment) Kaplan-Meier curves were plotted for a 60-month period, with observations right-censored at death or maximum follow-up. The log-rank test was used for statistical comparison of curves. Sequential Cox regression approaches were used for each pairwise comparison to estimate the relative hazard of the primary outcome in unadjusted, IPTW-adjusted (generated from the pairwise logistic regression models detailed previously), and IPTW-adjusted approaches with additional adjustment for time-updating BMI and categorical AUDIT-C scores (0, 1, 2, ≥3). In preliminary analysis, BMI was observed to have a nonlinear association with the outcome and was therefore transformed using 4-knot restricted cubic splines. Hazard ratios (HRs) and 95% CIs were computed, and adjusted survival curves were plotted for the latter models. Adjusted relative hazards were plotted to visualize the association between BMI and the outcome. In an exploratory analysis, we used stacked bar graphs to follow categorical AUDIT-C scores through each year of study follow-up for each exposure group, with χ2 tests performed for statistical comparisons. In an additional exploratory analysis, we ascertained all AUD hospitalizations for each patient through a maximum 5 years of follow-up and used IPTW-adjusted zero-inflated negative binomial regression (chosen due to excess zeros and overdispersion) to compute the incidence rate ratio of AUD hospitalizations for RYGB vs sleeve gastrectomy.
For secondary analysis, all-cause mortality was compared using Cox regression. A violation of the proportional hazards assumption was noted in an unadjusted model; however, after accounting for time-updated BMI (transformed using restricted cubic splines), time-updated AUDIT-C score categories, and use of IPTW, there was no evidence of proportional hazards violation by Schoenfeld residuals. The HRs and 95% CIs of each pairwise comparison model are presented along with adjusted survival curves. Given that the degree of alcohol use could plausibly modify the association between weight management pathway and mortality, in an exploratory analysis we evaluated an a priori hypothesized interaction between exposure group and time-updated AUDIT-C scores as a continuous variable. To visualize this interaction, the estimated hazard of all-cause mortality was plotted as a function of AUDIT-C score in patients who received RYGB or sleeve gastrectomy, relative to patients referred to MOVE! with an AUDIT-C score of 0. We also evaluated for an interaction between sex and RYGB vs sleeve gastrectomy given the possibility that sex-based differences could be present in the male-predominant VHA cohort; this was performed for both primary and secondary outcomes.
Data management and analyses were performed using structured query language and Stata/BE, version 17.0 (StataCorp LLC). The threshold for statistical significance for all tests was a 2-sided P < .05.
Results
Cohort Characteristics
After application of selection criteria (eFigure 1 in the Supplement), we identified 7694 patients who underwent RYGB (n = 1854; median [IQR] age, 53 [45-60] years; 1294 men [69.8%] and 560 women [30.2%]), sleeve gastrectomy (n = 4211; median [IQR] age, 52 [44-59] years; 2817 men [66.9%] and 1394 women [33.1%]), or gastric banding (n = 265; median [IQR] age, 55 [46-61] years; 199 men [75.1%] and 66 women [24.9%]) or received a MOVE! referral alone (n = 1364; median [IQR] age, 59 [49-66] years; 1175 men [86.1%] and 189 women [13.9%]). Patients who underwent bariatric surgery were more likely to be women (eg, RYGB vs MOVE!: 30.2% vs 13.9%; P < .001), have a higher baseline BMI (eg, median [IQR], 43.1 [38.9-48.0] for RYGB vs 36.8 [33.2-41.2] for MOVE!; P < .001), and have a higher prevalence of diabetes (eg, 51.7% for RYGB vs 41.6% for MOVE!; P < .001) (Table 1). Baseline AUDIT-C scores were lowest among patients who received RYGB (50.7% of patients in the RYGB group vs 39.9% in the MOVE! group among those with AUDIT-C score of 0; P < .001). Previous AUD hospitalizations were rare in the cohort overall, though slightly more common in patients referred to MOVE! (1.0% vs 0.1% in the RYGB group; P < .001).
Table 1. Baseline Cohort Characteristics.
| Characteristic | Patients, No. (%) | P value | |||
|---|---|---|---|---|---|
| Underwent RYGB (n = 1854) | Underwent sleeve gastrectomy (n = 4211) | Underwent gastric banding (n = 265) | Referred to MOVE! (n = 1364)a | ||
| Age, median (IQR) | 53 (45-60) | 52 (44-59) | 55 (46-61) | 59 (49-66) | <.001 |
| Sex | |||||
| Female | 560 (30.2) | 1394 (33.1) | 66 (24.9) | 189 (13.9) | <.001 |
| Male | 1294 (69.8) | 2817 (66.9) | 199 (75.1) | 1175 (86.1) | |
| Race and ethnicity | |||||
| Asian | 23 (1.2) | 72 (1.7) | 2 (0.8) | 12 (0.9) | <.001 |
| Black | 314 (16.9) | 954 (22.7) | 55 (20.8) | 26 (1.9) | |
| Hispanic | 238 (12.8) | 588 (14.0) | 21 (7.9) | 122 (8.9) | |
| White | 1107 (59.7) | 2287 (54.3) | 149 (56.2) | 1054 (77.3) | |
| Otherb | 172 (9.3) | 310 (7.4) | 38 (14.3) | 150 (11.0) | |
| BMI, median (IQR) | 43.1 (38.9-48.0) | 42.6 (38.8-47.0) | 43.0 (39.7-47.4) | 36.8 (33.2-41.2) | <.001 |
| Hypertension | 1420 (76.6) | 3050 (72.4) | 220 (83.0) | 995 (72.9) | <.001 |
| Hyperlipidemia | 1376 (74.2) | 2953 (70.1) | 206 (77.7) | 1095 (80.3) | <.001 |
| Diabetes | 959 (51.7) | 1970 (46.8) | 138 (52.1) | 567 (41.6) | <.001 |
| Coronary artery disease | 364 (19.6) | 695 (16.5) | 65 (24.5) | 333 (24.4) | <.001 |
| AUDIT-C score, median (IQR) | 0 (0-1) | 1 (0-1) | 0 (0-1) | 1 (0-2) | <.001 |
| AUDIT-C score | |||||
| 0 | 940 (50.7) | 1905 (45.2) | 142 (53.6) | 544 (39.9) | <.001 |
| 1 | 528 (28.5) | 1307 (31.0) | 64 (24.2) | 332 (24.3) | |
| 2 | 185 (10.0) | 530 (12.6) | 36 (13.6) | 170 (12.5) | |
| ≥3 | 201 (10.8) | 469 (11.1) | 23 (8.7) | 318 (23.3) | |
| Previous AUD hospitalization | 1 (0.1) | 7 (0.2) | 0 | 14 (1.0) | <.001 |
| Creatinine level, median (IQR), mg/dL | 0.9 (0.8-1.1) | 0.9 (0.8-1.1) | 0.9 (0.9-1.1) | 0.9 (0.8-1.0) | .15 |
| Median (IQR) total bilirubin level, mg/dL | 0.6 (0.4-0.8) | 0.6 (0.4-0.7) | 0.5 (0.4-0.8) | 0.6 (0.5-0.8) | <.001 |
| Median (IQR) albumin level, g/dL | 4 (3.8-4.3) | 4 (3.8-4.3) | 4.05 (3.8-4.2) | 4.3 (4.1-4.5) | <.001 |
| Median (IQR) ALT level, U/L | 28 (20-40) | 29 (21-41) | 26.5 (20-35) | 36 (28-48) | <.001 |
| Median (IQR) AST level, U/L | 22 (18-29) | 23 (18-29) | 25.5 (17-30.5) | 29 (24-37) | <.001 |
| Median (IQR) hemoglobin A1c level, g/dL | 6.0 (5.5-6.9) | 5.9 (5.5-6.5) | 5.8 (5.5-6.0) | 5.9 (5.5-7.0) | <.001 |
| Median (IQR) total cholesterol level, mg/dL | 170 (141-199) | 171 (144-199) | 178 (157-213.5) | 180 (148-205) | .001 |
| Median (IQR) triglyceride level, mg/dL | 140 (98.5-206) | 134 (95-195) | 110 (94.5-150.25) | 162 (108-231) | <.001 |
| Median (IQR) HDL level, mg/dL | 41 (35-50) | 41 (35-50) | 54.5 (46-61) | 39 (33-46) | <.001 |
| Median (IQR) LDL level, mg/dL | 97.4 (74.4-124) | 100 (77.4-125) | 103.5 (80-116.7) | 101.92 (78-125) | .45 |
Abbreviations: ALT, alanine aminotransferase; AST, aspartate aminotransferase; AUD, alcohol use disorder; AUDIT-C, Alcohol Use Disorders Identification Test-Concise; BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); HDL, high-density lipoprotein; LDL, low-density lipoprotein; RYGB, Roux-en-Y gastric bypass.
SI conversion factors: To convert creatinine level to μmol/L, multiply by 88.4; total bilirubin level to μmol/L, multiply by 17.104; albumin level to g/L, multiply by 10; ALT and AST levels to μkat/L, multiply by 0.0167; hemglobin A1c to g/L, multiply by 10; total cholesterol, HDL, and LDL levels to μmol/L, multiply by 0.0259; triglyceride level to μmol/L, multiply by 0.0113.
MOVE! is a weight management program in the Veterans Health Administration that focuses on healthy eating behaviors and increased physical activity.
Other races and ethnicities include American Indian or Alaska Native individuals, other races and ethnicities, or unknown or declined to answer.
Patterns and Weight-Related Outcomes in Bariatric Surgery in the VHA
Before 2010, the most common bariatric procedures were RYGB and gastric banding. After 2010, sleeve gastrectomy became the most frequently performed surgery, whereas gastric banding was rarely pursued (eFigure 2A and B in the Supplement). Procedure volume declined in 2020 corresponding to the onset of the COVID-19 pandemic. Patients who underwent RYGB experienced the largest reduction in BMI both in absolute terms (per median smoothing spline functions, maximum absolute BMI decline, −10.5) and as a percentage change from baseline (−29.0%; eFigure 3A and B in the Supplement), followed by those who underwent sleeve gastrectomy (maximum BMI decline, −7.3; percent BMI change, −21.6%) and those who underwent gastric banding (maximum BMI decline, −5.3 ; percent BMI change, −12.8%). Minimal changes to BMI were noted in patients referred to MOVE! (maximum BMI decline, −0.6; percent BMI change, −0.7%). All bariatric surgery groups experienced a gradual increase in BMI from the nadir during the 60-month follow-up. Hemoglobin A1c levels remained lowest overall in the RYGB group (maximum Δ HbA1c, −0.6%), whereas patients in the MOVE! group experienced a gradual increase in hemoglobin A1c over time (maximum Δ HbA1c, 0.6%; eFigure 3C in the Supplement).
Association Between Bariatric Surgery and AUD-Related Hospitalizations
During a median (IQR) follow-up of 59.6 (33.1-60.0) months, the unadjusted Kaplan-Meier analysis showed an association between weight management approach and AUD-related hospitalization (log-rank P < .001; eFigure 4 in the Supplement; raw events and follow-up data are shown in eTable 3 in the Supplement). With IPTW, excellent covariate balance was achieved in the RYGB and sleeve gastrectomy pseudopopulations; however, several variables remained unbalanced in the RYGB vs MOVE! and sleeve gastrectomy vs MOVE! group comparisons (most notably BMI; ie, weighted standardized main difference, −0.32 for RYGB vs MOVE! and −0.73 for sleeve gastrectomy vs MOVE!; eFigure 5 in the Supplement). In the fully adjusted Cox models, RYGB conferred an increased hazard of AUD-related hospitalization compared with MOVE! (HR, 1.70; 95% CI, 1.20-2.41; P = .003; Table 2, Figure 1A) and compared with sleeve gastrectomy (HR, 1.98; 95% CI, 1.55-2.53; P < .001; Figure 1B). In these models, lower time-updated BMI values were associated with an increased hazard of AUD-related hospitalization (HR, 1.39 for BMI 25 vs BMI 35 in RYGB vs MOVE! model; 95% CI, 1.03-1.89; P = .03; Figure 1D and E). By contrast, in the sleeve gastrectomy vs MOVE! group comparison, there was a nonsignificant difference in hazard of AUD-related hospitalization (HR, 0.76; 95% CI, 0.56-1.03; P = .08; Figure 1C). In an exploratory analysis, the RYGB group had the highest proportion of patients with AUDIT-C scores of 0 both at baseline and through each year of follow-up (eg, baseline: 50.7% in the RYGB group vs 45.2% in the sleeve gastrectomy group vs 39.9% in the MOVE! group; P < .001; Figure 2). Patients in the RYGB group also had significantly smaller proportions of AUDIT-C scores of 1, 2, and 3 or greater compared with those in the MOVE! group at all time points (eg, proportion of AUDIT-C score of 3 or greater in year 1, 6.8% for RYGB vs 7.5% for sleeve gastrectomy vs 22.3% for MOVE!; each P < .001). Finally, in an IPTW-adjusted, zero-inflated negative binomial regression model, RYGB was associated with a higher incidence rate of AUD hospitalizations vs sleeve gastrectomy (incidence rate ratio, 2.12; 95% CI, 1.64-2.75; P < .001); this corresponded to 24.6 AUD hospitalizations per 1000 patient-years for RYGB vs 11.6 per 1000 patient-years for sleeve gastrectomy.
Table 2. IPTW Cox Regression Analyses for AUD-Related Hospitalization.
| Variable | Unadjusted | IPTW adjusted | IPTW with time-updating covariates | |||
|---|---|---|---|---|---|---|
| HR (95% CI) | P value | HR (95% CI) | P value | HR (95% CI) | P value | |
| RYGB vs MOVE! | ||||||
| Weight loss category | ||||||
| MOVE! | 1 [Reference] | Reference | 1 [Reference] | Reference | 1 [Reference] | Reference |
| RYGB | 1.93 (1.23-3.03) | .004 | 2.01 (1.50-2.69) | <.001 | 1.70 (1.20-2.41) | .003 |
| RCS (BMI) (time updating) | NA | NA | NA | NA | Figure 1Da | .02 |
| AUDIT-C score (time updating) | ||||||
| 0 | NA | NA | NA | NA | 1 [Reference] | Reference |
| 1 | NA | NA | NA | NA | 0.81 (0.55-1.18) | .27 |
| 2 | NA | NA | NA | NA | 1.24 (0.78-1.98) | .36 |
| ≥3 | NA | NA | NA | NA | 2.12 (1.50-2.98) | <.001 |
| RYGB vs sleeve gastrectomy | ||||||
| Weight loss category | ||||||
| Sleeve gastrectomy | 1 [Reference] | Reference | 1 [Reference] | Reference | 1 [Reference] | Reference |
| RYGB | 2.17 (1.54-3.05) | <.001 | 2.08 (1.64-2.65) | <.001 | 1.98 (1.55-2.53) | <.001 |
| RCS (BMI) (time updating) | NA | NA | NA | NA | Figure 1Ea | .13 |
| AUDIT-C score (time updating) | ||||||
| 0 | NA | NA | NA | NA | 1 [Reference] | Reference |
| 1 | NA | NA | NA | NA | 0.81 (0.60-1.09) | .17 |
| 2 | NA | NA | NA | NA | 1.11 (0.75-1.64) | .60 |
| ≥3 | NA | NA | NA | NA | 2.80 (2.12-3.69) | <.001 |
| Sleeve gastrectomy vs MOVE! b | ||||||
| Weight loss category | ||||||
| MOVE! | 1 [Reference] | Reference | 1 [Reference] | Reference | 1 [Reference] | Reference |
| Sleeve gastrectomy | 0.89 (0.56-1.39) | .58 | 0.86 (0.66-1.12) | .26 | 0.76 (0.56-1.03) | .08 |
| RCS (BMI) (time updating) | NA | NA | NA | NA | Figure 1Fa | .008 |
| AUDIT-C score (time updating) | ||||||
| 0 | NA | NA | NA | NA | 1 [Reference] | Reference |
| 1 | NA | NA | NA | NA | 0.29 (0.18-0.47) | <.001 |
| 2 | NA | NA | NA | NA | 1.02 (0.67-1.54) | .93 |
| ≥3 | NA | NA | NA | NA | 1.83 (1.32-2.54) | <.001 |
Abbreviations: AUD, alcohol use disorder; AUDIT-C, Alcohol Use Disorders Identification Test-Concise; BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); HR, hazard ratio; IPTW, inverse probability treatment weighting; NA, not applicable; RCS, restricted cubic spline; RYGB, Roux-en-Y gastric bypass.
See Figure 1 for visualization of the adjusted association between BMI and AUD-related hospitalization.
IPTW with time-updating covariate model for sleeve gastrectomy vs MOVE! comparison additionally adjusted for race, baseline hypertension, baseline hyperlipidemia, and baseline diabetes.
Figure 1. Association Between Weight Management Approach and Alcohol Use Disorder (AUD)–Related Hospitalization.
A-C, Results of inverse probability treatment–weighted Cox regression models showing the association between weight management approaches and AUD-related hospitalizations in pairwise comparisons. D-F, The adjusted association between body mass index (calculated as weight in kilograms divided by height in meters squared) and AUD-related hospitalization in each respective Cox model. Shading represents 95% CIs. RYGB indicates Roux-en-Y gastric bypass.
Figure 2. Alcohol Use Disorders Identification Test-Concise (AUDIT-C) Score During Follow-up, Stratified by Weight Management Approach.
The denominator in a given follow-up year excludes patients who experienced a censoring event, such as death or maximum follow-up, before the start of the year. AUDIT-C is a screening tool for hazardous alcohol use and ranges from 0 to 12, with alcohol misuse represented by scores of 3 or greater for women and 4 or greater for men. RYGB indicates Roux-en-Y gastric bypass.
Association Between Bariatric Surgery and All-Cause Mortality
In the unadjusted analysis, there were significant differences in mortality across weight management groups, with the MOVE! group exhibiting the highest mortality and the sleeve gastrectomy group the lowest (log-rank P < .001; Figure 3A). In fully adjusted analyses, sleeve gastrectomy and RYGB were associated with a reduced hazard of mortality (sleeve gastrectomy: HR, 0.49; 95% CI, 0.39-0.61; P < .001; RYGB: HR, 0.63; 95% CI, 0.49-0.81; P < .001) compared with MOVE! (eTable 4, eFigure 6A and C in the Supplement). Roux-en-Y gastric bypass was associated with an increased hazard of mortality vs sleeve gastrectomy (HR, 1.36; 95% CI, 1.09-1.70; P < .001; eFigure 6B in the Supplement). In an exploratory analysis, we identified a significant interaction between weight management approach and time-updated AUDIT-C scores in both the RYGB vs MOVE! (P < .001 for the interaction) and sleeve gastrectomy vs MOVE! models (P = .008 for the interaction). These associations are represented in Figure 3B. Compared with a patient with an AUDIT-C score of 0 referred to MOVE!, the protective association between RYGB and mortality diminished rapidly with increasing AUDIT-C scores, with neutral association observed around an AUDIT-C score of 8 to 9 (AUDIT-C score of 3: HR, 0.63; 95% CI, 0.48-0.84; P = .001 and AUDIT-C score of 8: HR, 0.95; 95% CI, 0.61-1.47; P = .82). By contrast, the sleeve gastrectomy group showed point estimates of mortality benefit across the entire spectrum of AUDIT-C scores (eg, AUDIT-C score of 3: HR, 0.48; 95% CI, 0.36-0.63; P < .001 and AUDIT-C score of 8: HR, 0.60; 95% CI, 0.32-1.13; P = .11). Finally, in an exploratory analysis, we found a significant interaction between sex and RYGB vs sleeve gastrectomy (P = .004 for the interaction). Roux-en-Y gastric bypass was associated with an increased hazard of all-cause mortality vs sleeve gastrectomy in men (HR, 1.49; 95% CI, 1.17-1.91; P = .001) but not in women (HR, 0.65; 95% CI, 0.35-1.19; P = .16). Of note, there was no such interaction in a model studying the primary outcome of AUD hospitalizations (P = .25 for the interaction).
Figure 3. Association Between Weight Management Approach and All-Cause Mortality.
A, The graph shows significant differences in mortality associated with the 3 weight management approaches. B, The hazard was computed relative to patients referred to a MOVE! program with an Alcohol Use Disorders Identification Test-Concise (AUDIT-C) score of 0. AUDIT-C is a screening tool for hazardous alcohol use and ranges from 0 to 12, with alcohol misuse represented by scores of 3 or greater for women and 4 or greater for men. IPTW indicates inverse probability treatment weighting; RYGB, Roux-en-Y gastric bypass. Shaded areas indicate ± 1 SE bands.
Discussion
In this study of veterans with obesity treated with different weight management approaches, RYGB was associated with an increased risk of AUD-related hospitalizations vs sleeve gastrectomy or MOVE! program. Sleeve gastrectomy and RYGB were both associated with reduced mortality vs MOVE!; however, the mortality benefit associated with RYGB diminished with increased alcohol exposure during follow-up.
The volume of bariatric surgeries in VHA has increased steadily over time, with a break in the pattern in 2020 corresponding to the onset of the COVID-19 pandemic. Sleeve gastrectomy has become the preferred bariatric surgery approach, whereas the relative proportion of RYGB has reduced gradually over time. These patterns mirror data from national administrative studies.27,28 Similar to previous literature, RYGB was the most effective procedure in reducing BMI, with nearly 30% reduction at 1 year compared with approximately 22% for sleeve gastrectomy and approximately 12% for gastric banding.29,30 Patients referred to the MOVE! program did not achieve consistent weight loss and had an increasing hemoglobin A1c level over time, suggesting this strategy alone is likely ineffective for most patients.
The primary novel finding in this study is that patients who underwent RYGB had significantly increased risk of AUD-related hospitalizations vs sleeve gastrectomy and the MOVE! program. Although previous studies have identified an association between RYGB and potentially increased AUD-related complications, none contained granular, longitudinal patient-level data to account for changes in BMI or alcohol consumption over time or to comprehensively account for confounders. We found that patients who underwent RYGB had a higher risk of AUD-related hospitalization despite generally consuming the least amount of alcohol at baseline or during follow-up, suggesting that alterations in alcohol metabolism more than changes in alcohol consumption may be associated with increased risks of alcohol-related complications in patients who underwent RYGB. This hypothesis is biologically plausible; alcohol dehydrogenase expressed in gastric mucosa initiates metabolism of alcohol in the proximal gut, before rapid absorption in the proximal small bowel and arrival in the liver via the portal venous circulation.31,32 Roux-en-Y gastric bypass may reduce initial luminal metabolism by reducing gastric surface area and causing rapid gastric transit of contents into the small bowel. Thus, patients who undergo RYGB may be prone to highly toxic levels of blood alcohol reaching the liver even with modest alcohol consumption.33,34,35 Through mechanisms yet to be elucidated, patients who undergo RYGB may also develop reduced liver function that impairs alcohol metabolism, further increasing toxic effects.36 In the present study, only RYGB (and not sleeve gastrectomy, which preserves pyloric function) was associated with an increased risk of AUD-related hospitalizations. This finding differs from previous studies that reported an increased risk of AUD-related complications in both patients who underwent sleeve gastrectomy and those who underwent RYGB.10,11 By contrast, the present study implies a uniquely increased risk of AUD-related complications with combined malabsorptive and restrictive compared with restrictive procedures alone, an observation in agreement with a recently published study in which bariatric surgery procedures were compared with a cholecystectomy control group.9
Roux-en-Y gastric bypass and sleeve gastrectomy also conferred reduced risk of all-cause mortality vs the MOVE! program. This is consistent with previous literature, though data conflict as to whether long-term survival is superior with RYGB vs sleeve gastrectomy.37,38 In this study, the association with reduced mortality was more modest with RYGB vs sleeve gastrectomy; underlying cohort differences, including male predominance in the VHA cohort or residual confounding, may in part explain this discordance. Because patients who underwent RYGB experienced greater reductions in BMI and hemoglobin A1c level, a mechanism apart from reduction in metabolic comorbidities may be attenuating mortality benefits in these patients. In an exploratory analysis, we found that the survival benefit associated with RYGB diminished rapidly with increasing alcohol use and was nullified at an AUDIT-C score of 8; the mortality benefit persisted across AUDIT-C scores in patients who underwent sleeve gastrectomy. These findings underscore the importance of careful patient selection for RYGB, counseling regarding postoperative alcohol use and ideally sobriety, and active monitoring for hazardous alcohol use.
There are important study limitations to acknowledge. First, there is possible misclassification of exposures and outcomes. Some patients receiving bariatric surgery outside the VHA may have been misclassified. However, in our initial queries to identify this exposure, we included fee basis tables, which capture reimbursements for procedures performed outside the VHA. Second, due to small event numbers, we could not study subclasses of AUD-related hospitalizations (eg, alcohol intoxication, alcohol withdrawal, alcoholic hepatitis). Given the hypothesized mechanism of increased alcohol sensitivity in patients who underwent RYGB, however, the risk for all subclasses would likely be increased. Third, there are external validity limitations to VHA studies, as this cohort is male predominant and enriched in psychosocial comorbidities. The primary findings are in general agreement with large multicenter administrative data sets, though mortality differences in RYGB vs sleeve gastrectomy groups may in part be attributable to a male-predominant cohort. Fourth, residual confounding is possible, especially as pertains to center-specific or surgeon-specific factors and variables related to patient preference to pursue bariatric surgery. Fifth, some degree of selection bias related to exclusion of patients with baseline missing AUDIT-C scores is possible. Finally, we were unable to explore causes of death in detail in this data set.
Conclusions
Patients who underwent RYGB had a significantly increased risk of AUD-related hospitalizations vs those who underwent sleeve gastrectomy or those referred to the MOVE! program. Patients who undergo RYGB may be uniquely sensitive to alcohol exposure, and increased alcohol use may nullify potential gains in overall survival. These findings highlight the importance of careful patient selection for RYGB, strict counseling regarding alcohol use, and long-term monitoring for AUD-related complications.
eTable 1. Classification of Bariatric Surgery Procedures
eTable 2. Classification of Alcohol Use Disorder (AUD)-Related Hospitalizations
eMethods. Descriptive and Trend Analysis
eTable 3. Summary of Follow-up Duration and Raw Outcomes in Survival Analysis, Stratified by Cohort
eTable 4. IPTW Cox Regression Analyses for All-Cause Mortality
eFigure 1. Cohort Selection Criteria
eFigure 2. VHA Trends in Bariatric Surgery Approach Over Time
eFigure 3. Change in BMI and HbA1c Over Time, Stratified by Weight Management Approach
eFigure 4. Unadjusted Kaplan-Meier Analysis for AUD-Related Hospitalizations, Stratified by Weight Management Approach
eFigure 5. Covariate Balance Achieved Through Inverse Probability Treatment Weighting for Each Pairwise Group Comparison
eFigure 6. IPTW-Adjusted Cox Regression Models for All-Cause Mortality With Time Updating Covariates for Each Pairwise Group Comparison
References
- 1.Mechanick JI, Youdim A, Jones DB, et al. ; American Association of Clinical Endocrinologists; Obesity Society; American Society for Metabolic & Bariatric Surgery . Clinical practice guidelines for the perioperative nutritional, metabolic, and nonsurgical support of the bariatric surgery patient—2013 update: cosponsored by American Association of Clinical Endocrinologists, The Obesity Society, and American Society for Metabolic & Bariatric Surgery. Obesity (Silver Spring). 2013;21(suppl 1):S1-S27. doi: 10.1002/oby.20461 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Brito JP, Montori VM, Davis AM. Metabolic surgery in the treatment algorithm for type 2 diabetes: a joint statement by international diabetes organizations. JAMA. 2017;317(6):635-636. doi: 10.1001/jama.2016.20563 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Flegal KM, Kruszon-Moran D, Carroll MD, Fryar CD, Ogden CL. Trends in obesity among adults in the United States, 2005 to 2014. JAMA. 2016;315(21):2284-2291. doi: 10.1001/jama.2016.6458 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Pi-Sunyer FX. The obesity epidemic: pathophysiology and consequences of obesity. Obes Res. 2002;10(suppl 2):97S-104S. doi: 10.1038/oby.2002.202 [DOI] [PubMed] [Google Scholar]
- 5.Bastien M, Poirier P, Lemieux I, Després JP. Overview of epidemiology and contribution of obesity to cardiovascular disease. Prog Cardiovasc Dis. 2014;56(4):369-381. doi: 10.1016/j.pcad.2013.10.016 [DOI] [PubMed] [Google Scholar]
- 6.Baki JA, Tapper EB. Contemporary epidemiology of cirrhosis. Curr Treat Options Gastroenterol. 2019;17(2):244-253. doi: 10.1007/s11938-019-00228-3 [DOI] [PubMed] [Google Scholar]
- 7.Li Q, Blume SW, Huang JC, Hammer M, Ganz ML. Prevalence and healthcare costs of obesity-related comorbidities: evidence from an electronic medical records system in the United States. J Med Econ. 2015;18(12):1020-1028. doi: 10.3111/13696998.2015.1067623 [DOI] [PubMed] [Google Scholar]
- 8.Backman O, Stockeld D, Rasmussen F, Näslund E, Marsk R. Alcohol and substance abuse, depression and suicide attempts after Roux-en-Y gastric bypass surgery. Br J Surg. 2016;103(10):1336-1342. doi: 10.1002/bjs.10258 [DOI] [PubMed] [Google Scholar]
- 9.Kim HP, Jiang Y, Farrell TM, Peat CM, Hayashi PH, Barritt AS IV. Roux-en-Y gastric bypass is associated with increased hazard for de novo alcohol-related complications and liver disease. J Clin Gastroenterol. 2022;56(2):181-185. doi: 10.1097/MCG.0000000000001506 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Mellinger JL, Shedden K, Winder GS, et al. Bariatric surgery and the risk of alcohol-related cirrhosis and alcohol misuse. Liver Int. 2021;41(5):1012-1019. doi: 10.1111/liv.14805 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Svensson PA, Anveden Å, Romeo S, et al. Alcohol consumption and alcohol problems after bariatric surgery in the Swedish obese subjects study. Obesity (Silver Spring). 2013;21(12):2444-2451. doi: 10.1002/oby.20397 [DOI] [PubMed] [Google Scholar]
- 12.Östlund MP, Backman O, Marsk R, et al. Increased admission for alcohol dependence after gastric bypass surgery compared with restrictive bariatric surgery. JAMA Surg. 2013;148(4):374-377. doi: 10.1001/jamasurg.2013.700 [DOI] [PubMed] [Google Scholar]
- 13.Maciejewski ML, Smith VA, Berkowitz TSZ, et al. Association of bariatric surgical procedures with changes in unhealthy alcohol use among US veterans. JAMA Netw Open. 2020;3(12):e2028117-e2028117. doi: 10.1001/jamanetworkopen.2020.28117 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Abraham CR, Werter CR, Ata A, et al. Predictors of hospital readmission after bariatric surgery. J Am Coll Surg. 2015;221(1):220-227. doi: 10.1016/j.jamcollsurg.2015.02.018 [DOI] [PubMed] [Google Scholar]
- 15.Kinsinger LS, Jones KR, Kahwati L, et al. Peer reviewed: design and dissemination of the MOVE! weight-management program for veterans. Prev Chronic Dis. 2009;6(3):A98. [PMC free article] [PubMed] [Google Scholar]
- 16.Bush K, Kivlahan DR, McDonell MB, Fihn SD, Bradley KA. The AUDIT alcohol consumption questions (AUDIT-C): an effective brief screening test for problem drinking. Arch Intern Med. 1998;158(16):1789-1795. doi: 10.1001/archinte.158.16.1789 [DOI] [PubMed] [Google Scholar]
- 17.Heikkinen M, Taipale H, Tanskanen A, Mittendorfer-Rutz E, Lähteenvuo M, Tiihonen J. Real-world effectiveness of pharmacological treatments of alcohol use disorders in a Swedish nation-wide cohort of 125 556 patients. Addiction. 2021;116(8):1990-1998. doi: 10.1111/add.15384 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Centers for Disease Control and Prevention . Alcohol-related disease impact (ARDI) International Classification of Diseases (ICD) codes and alcohol-attributable fraction (AAF). Reviewed July 6, 2021. Accessed August 25, 2022. https://www.cdc.gov/alcohol/ardi/alcohol-related-icd-codes.html
- 19.Trefan L, Akbari A, Morgan JS, et al. Visualisation and optimisation of alcohol-related hospital admissions ICD-10 codes in Welsh e-cohort data. Int J Popul Data Sci. 2021;6(1):1373. doi: 10.23889/ijpds.v6i1.1373 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Fone D, Dunstan F, White J, et al. Change in alcohol outlet density and alcohol-related harm to population health (CHALICE). BMC Public Health. 2012;12:428. doi: 10.1186/1471-2458-12-428 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Benotti P, Wood GC, Winegar DA, et al. Risk factors associated with mortality after Roux-en-Y gastric bypass surgery. Ann Surg. 2014;259(1):123-130. doi: 10.1097/SLA.0b013e31828a0ee4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Still CD, Wood GC, Chu X, et al. Clinical factors associated with weight loss outcomes after Roux-en-Y gastric bypass surgery. Obesity (Silver Spring). 2014;22(3):888-894. doi: 10.1002/oby.20529 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Wood GC, Benotti PN, Lee CJ, et al. Evaluation of the association between preoperative clinical factors and long-term weight loss after Roux-en-Y gastric bypass. JAMA Surg. 2016;151(11):1056-1062. doi: 10.1001/jamasurg.2016.2334 [DOI] [PubMed] [Google Scholar]
- 24.Santry HP, Lauderdale DS, Cagney KA, Rathouz PJ, Alverdy JC, Chin MH. Predictors of patient selection in bariatric surgery. Ann Surg. 2007;245(1):59-67. doi: 10.1097/01.sla.0000232551.55712.b3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Austin PC, Stuart EA. Moving towards best practice when using inverse probability of treatment weighting (IPTW) using the propensity score to estimate causal treatment effects in observational studies. Stat Med. 2015;34(28):3661-3679. doi: 10.1002/sim.6607 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Austin PC. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Stat Med. 2009;28(25):3083-3107. doi: 10.1002/sim.3697 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Alalwan AA, Friedman J, Park H, Segal R, Brumback BA, Hartzema AG. US national trends in bariatric surgery: a decade of study. Surgery. 2021;170(1):13-17. doi: 10.1016/j.surg.2021.02.002 [DOI] [PubMed] [Google Scholar]
- 28.Campos GM, Khoraki J, Browning MG, Pessoa BM, Mazzini GS, Wolfe L. Changes in utilization of bariatric surgery in the United States from 1993 to 2016. Ann Surg. 2020;271(2):201-209. doi: 10.1097/SLA.0000000000003554 [DOI] [PubMed] [Google Scholar]
- 29.Celio AC, Wu Q, Kasten KR, Manwaring ML, Pories WJ, Spaniolas K. Comparative effectiveness of Roux-en-Y gastric bypass and sleeve gastrectomy in super obese patients. Surg Endosc. 2017;31(1):317-323. doi: 10.1007/s00464-016-4974-y [DOI] [PubMed] [Google Scholar]
- 30.Carlin AM, Zeni TM, English WJ, et al. ; Michigan Bariatric Surgery Collaborative . The comparative effectiveness of sleeve gastrectomy, gastric bypass, and adjustable gastric banding procedures for the treatment of morbid obesity. Ann Surg. 2013;257(5):791-797. doi: 10.1097/SLA.0b013e3182879ded [DOI] [PubMed] [Google Scholar]
- 31.Seitz HK, Egerer G, Simanowski UA, et al. Human gastric alcohol dehydrogenase activity: effect of age, sex, and alcoholism. Gut. 1993;34(10):1433-1437. doi: 10.1136/gut.34.10.1433 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Crabb DW, Matsumoto M, Chang D, You M. Overview of the role of alcohol dehydrogenase and aldehyde dehydrogenase and their variants in the genesis of alcohol-related pathology. Proc Nutr Soc. 2004;63(1):49-63. doi: 10.1079/PNS2003327 [DOI] [PubMed] [Google Scholar]
- 33.Klockhoff H, Näslund I, Jones AW. Faster absorption of ethanol and higher peak concentration in women after gastric bypass surgery. Br J Clin Pharmacol. 2002;54(6):587-591. doi: 10.1046/j.1365-2125.2002.01698.x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Woodard GA, Downey J, Hernandez-Boussard T, Morton JM. Impaired alcohol metabolism after gastric bypass surgery: a case-crossover trial. J Am Coll Surg. 2011;212(2):209-214. doi: 10.1016/j.jamcollsurg.2010.09.020 [DOI] [PubMed] [Google Scholar]
- 35.Pepino MY, Okunade AL, Eagon JC, Bartholow BD, Bucholz K, Klein S. Effect of Roux-en-Y gastric bypass surgery: converting 2 alcoholic drinks to 4. JAMA Surg. 2015;150(11):1096-1098. doi: 10.1001/jamasurg.2015.1884 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Eilenberg M, Langer FB, Beer A, Trauner M, Prager G, Staufer K. Significant liver-related morbidity after bariatric surgery and its reversal—a case series. Obes Surg. 2018;28(3):812-819. doi: 10.1007/s11695-017-2925-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Tarride J-E, Doumouras AG, Hong D, et al. Comparison of 4-year health care expenditures associated with Roux-en-Y gastric bypass vs sleeve gastrectomy. JAMA Netw Open. 2021;4(9):e2122079-e2122079. doi: 10.1001/jamanetworkopen.2021.22079 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Reges O, Greenland P, Dicker D, et al. Association of bariatric surgery using laparoscopic banding, Roux-en-Y gastric bypass, or laparoscopic sleeve gastrectomy vs usual care obesity management with all-cause mortality. JAMA. 2018;319(3):279-290. doi: 10.1001/jama.2017.20513 [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
eTable 1. Classification of Bariatric Surgery Procedures
eTable 2. Classification of Alcohol Use Disorder (AUD)-Related Hospitalizations
eMethods. Descriptive and Trend Analysis
eTable 3. Summary of Follow-up Duration and Raw Outcomes in Survival Analysis, Stratified by Cohort
eTable 4. IPTW Cox Regression Analyses for All-Cause Mortality
eFigure 1. Cohort Selection Criteria
eFigure 2. VHA Trends in Bariatric Surgery Approach Over Time
eFigure 3. Change in BMI and HbA1c Over Time, Stratified by Weight Management Approach
eFigure 4. Unadjusted Kaplan-Meier Analysis for AUD-Related Hospitalizations, Stratified by Weight Management Approach
eFigure 5. Covariate Balance Achieved Through Inverse Probability Treatment Weighting for Each Pairwise Group Comparison
eFigure 6. IPTW-Adjusted Cox Regression Models for All-Cause Mortality With Time Updating Covariates for Each Pairwise Group Comparison