Comparing the 5-Year Diabetes Outcomes of Sleeve Gastrectomy and Gastric Bypass: The National Patient-Centered Clinical Research Network (PCORNet) Bariatric Study

Key Points

Question  How do type 2 diabetes (T2DM) outcomes compare across the 2 most common bariatric procedures?

Findings  In this cohort study of 9710 adults with T2DM who underwent bariatric surgery, most patients who had Roux-en-Y gastric bypass or sleeve gastrectomy experienced T2DM remission at some point over 5 years of follow-up. Patients who had Roux-en-Y gastric bypass showed slightly higher T2DM remission rates, better glycemic control, and fewer T2DM relapse events than patients who had sleeve gastrectomy.

Meaning  Understanding diabetes outcomes of different bariatric procedures will help surgeons and patients with diabetes make informed health care choices.

Importance  Bariatric surgery can lead to substantial improvements in type 2 diabetes (T2DM), but outcomes vary across procedures and populations. It is unclear which bariatric procedure has the most benefits for patients with T2DM.

Objective  To evaluate associations of bariatric surgery with T2DM outcomes.

Design, Setting, and Participants  This cohort study was conducted in 34 US health system sites in the National Patient-Centered Clinical Research Network Bariatric Study. Adult patients with T2DM who had bariatric surgery between January 1, 2005, and September 30, 2015, were included. Data analysis was conducted from April 2017 to August 2019.

Interventions  Roux-en-Y gastric bypass (RYGB) or sleeve gastrectomy (SG).

Main Outcome and Measures  Type 2 diabetes remission, T2DM relapse, percentage of total weight lost, and change in glycosylated hemoglobin (hemoglobin A_1c).

Results  A total of 9710 patients were included (median [interquartile range] follow-up time, 2.7 [2.9] years; 7051 female patients [72.6%]; mean [SD] age, 49.8 [10.5] years; mean [SD] BMI, 49.0 [8.4]; 6040 white patients [72.2%]). Weight loss was significantly greater with RYGB than SG at 1 year (mean difference, 6.3 [95% CI, 5.8-6.7] percentage points) and 5 years (mean difference, 8.1 [95% CI, 6.6-9.6] percentage points). The T2DM remission rate was approximately 10% higher in patients who had RYGB (hazard ratio, 1.10 [95% CI, 1.04-1.16]) than those who had SG. Estimated adjusted cumulative T2DM remission rates for patients who had RYGB and SG were 59.2% (95% CI, 57.7%-60.7%) and 55.9% (95% CI, 53.9%-57.9%), respectively, at 1 year and 86.1% (95% CI, 84.7%-87.3%) and 83.5% (95% CI, 81.6%-85.1%) at 5 years postsurgery. Among 6141 patients who experienced T2DM remission, the subsequent T2DM relapse rate was lower for those who had RYGB than those who had SG (hazard ratio, 0.75 [95% CI, 0.67-0.84]). Estimated relapse rates for those who had RYGB and SG were 8.4% (95% CI, 7.4%-9.3%) and 11.0% (95% CI, 9.6%-12.4%) at 1 year and 33.1% (95% CI, 29.6%-36.5%) and 41.6% (95% CI, 36.8%-46.1%) at 5 years after surgery. At 5 years, compared with baseline, hemoglobin A_1c was reduced 0.45 (95% CI, 0.27-0.63) percentage points more for patients who had RYGB vs patients who had SG.

Conclusions and Relevance  In this large multicenter study, patients who had RYGB had greater weight loss, a slightly higher T2DM remission rate, less T2DM relapse, and better long-term glycemic control compared with those who had SG. These findings can help inform patient-centered surgical decision-making.

Bariatric surgery appears more effective than medical care alone for improving diabetes outcomes.^1-3 Remission of type 2 diabetes (T2DM) is common after bariatric surgery^4-7 and may reduce risk for subsequent microvascular and macrovascular disease.^8-11 However, T2DM remission rates after bariatric surgery vary substantially across procedures and populations^4-7 and T2DM relapse has been reported in approximately a quarter to half of patients who have bariatric surgery and achieve remission.^6,7,12

Studies focusing on the 2 most common bariatric procedures, sleeve gastrectomy (SG) and Roux-en-Y gastric bypass (RYGB), show mixed evidence in terms of T2DM outcomes, especially in the longer term.^13-22 It is unclear how the choice between them is likely to affect T2DM. The comparison is particularly salient because SG has begun to supplant RYGB as the dominant bariatric procedure over the past decade, despite limited long-term comparative data.^23-25

The PCORnet Bariatric Study (PBS),^25,26 one of the first scientific initiatives of PCORnet, the National Patient-Centered Clinical Research Network,^27,28 was designed to examine the effectiveness of common bariatric procedures. This article compares T2DM outcomes in PCORnet up to 5 years following surgery for patients who had SG or RYGB. Secondary analyses assess the procedures’ outcomes on body weight and glycemic control independent of diabetes remission.

The PBS cohort was previously described.²⁵ Patients in the T2DM analyses underwent a primary bariatric procedure at 34 PCORnet-affiliated health systems (eTable 1 in the Supplement) from January 1, 2005, through September 30, 2015. Procedures were identified from more than 59 million patient records using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM), Current Procedure Terminology version 4, and Healthcare Common Procedure Coding System codes. We defined patients with diabetes as having a hemoglobin A_1c (HbA_1c) level of 6.5% or more or a T2DM medication prescription in the year before surgery. Patients taking only metformin, thiazolidinedione, or liraglutide needed an ICD-9-CM or Systematized Nomenclature of Medicine (SNOMED) code for T2DM or an HbA_1c level of 6.5% or more in the year before surgery to be eligible for inclusion. We excluded patients 80 years or older, those without T2DM, and individuals without relevant outcomes data (eFigure 1 and eAppendix 1 in the Supplement).

The Kaiser Permanente Washington Health Research Institute obtained institutional review board approval for oversight of data collection and analyses. Participating sites obtained approval or formal determination that these analyses was not human subjects research.²⁵ A waiver of Health Insurance Portability and Privacy Act privacy authorization (and thus informed consent) was obtained for these analyses of deidentified data.

The PCORnet sites store standardized electronic health record data and sometimes other data (eg, claims data), in PCORnet datamarts.²⁸ Programming queries from the PCORnet Coordinating Center extracted relevant deidentified data on eligible individuals from participating sites’ datamarts. Race/ethnicity, as recorded in electronic health records, was included, reflecting stakeholder input. Data were transmitted to the coordinating site for analysis. Codes from the ICD-9-CM and SNOMED identified diagnoses.

Remission from T2DM was defined as the first postsurgical occurrence of an HbA_1c level less than 6.5% (to convert to proportion of total hemoglobin, multiply by 0.04-0.07) following at least 6 months (presurgical and/or postsurgical time) without T2DM medication prescription orders. This HbA_1c level corresponds to a published, putative partial-remission threshold.²⁹ It was identified by our clinical stakeholders as more clinically meaningful than the affiliated complete remission threshold (a normal hemoglobin A_1c level²⁹ of <5.7%³⁰), since an HbA_1c level less than 6.5% corresponds to a T2DM diagnosis.³⁰ The occurrence of levels of 6.5% or more and/or a prescription for T2DM medication after remission defined relapse. The absolute change in HbA_1c level at 1 year, 3 years, and 5 years after surgery was calculated. The total weight loss percentage was estimated as (weight at surgery − weight at a postoperative point)/weight at surgery × 100).

We compared the associations of RYGB and SG with time to diabetes remission. Pairwise analyses were restricted to sites with at least 1 patient of each procedure type at each point. Possible confounding was addressed with direct adjustment for specific factors and deciles of an estimated propensity score. Analyses examining the adjustable gastric band procedure are provided in eAppendix 2 in the Supplement.

Cox proportional hazards models calculated the adjusted hazard ratio (HR) for remission and estimated the adjusted cumulative proportion of individuals remitting at 1 year, 3 years, and 5 years following surgery. The proportional hazards assumption was tested by including an interaction between time and bariatric surgery group in the model, then inspecting Schoenfeld residuals over time. Models were adjusted for predetermined baseline covariates: age, sex, race, Hispanic ethnicity, body mass index category (BMI; calculated as weight in kilograms divided by height in meters squared), HbA_1c category, Charlson/Elixhauser comorbidity score (range: −2 to 20; a higher score generally indicates worse health),³¹ the health conditions listed in Table 1, the number of diabetes medications, the number of days hospitalized in the year before surgery, the year of surgery, and the site of surgery.

Logistic regression models estimating treatment propensity scores included fixed main effects for the prespecified covariates plus baseline variables for automated selection. To allow for differing outcomes of confounding variables by procedure site, propensity score models included subsets of all possible 2-way interactions between the listed variables and site. The subset of interactions and the additional covariates beyond the prespecified set were chosen using the least absolute shrinkage and selection operator method, with cross validation to select the most parsimonious model, with prediction error close to the minimum possible (within 1 SE).³²

Follow-up for T2DM remission was calculated from the index procedure date to the last observable data point following surgery (ie, the last observed visit, weight, blood pressure, HbA_1c laboratory value, or diabetes prescription). Remission analyses’ censoring events included death, conversion to a second bariatric procedure (eg, SG to RYGB), pregnancy (at the delivery date minus 270 days), and an 18-month lapse in diabetes-specific health care at participating sites. The relapse analyses included an additional censoring event, lapse in provision of any care, because patients in remission from T2DM were not necessarily expected to receive HbA_1c measures or T2DM prescriptions but needed to receive care in the system to be observed for relapse. It was defined as more than 18 months without any recorded HbA_1c levels, body weight measurement, blood pressure, diagnosis code, or procedure code. Since inpatient hospitalization can temporarily worsen glycemic control, we excluded HbA_1c measurements from admission date to 90 days after discharge and medication orders from admission dates to the day before discharge.

Exploratory hypothesis-generating analyses examined heterogeneity of treatment outcomes. Following recommendations for use of risk-stratified analyses to detect differences in treatment outcome,³³ subgroups defined by DiaRem score (Table 1) were assessed via interactions with procedure type. The DiaRem score is a widely validated approach to preoperative prognostication of T2DM remission after bariatric surgery; higher scores denote a lower probability of T2DM remission.³⁴ It is calculated based on age, HbA_1c level, insulin use, and use of oral diabetes medications.

Estimates of trends in mean total weight loss percentage were obtained using linear mixed-effects modeling with weight as the outcome and potential confounders (including baseline weight) and deciles of the propensity score as the independent variables. Adjusted total weight loss percentage was computed as the percentage change between the mean weight and the mean baseline weight. Time to T2DM relapse was assessed among patients who experienced diabetes remission, using the same methods as in the remission analyses. Adjusted absolute changes in HbA_1c level at 1 year, 3 years, and 5 years following surgery were estimated by procedure using a linear mixed-modeling framework with random effects for individual (intercept) and follow-up time (slope). A b-spline basis included a smooth function of follow-up time in the model, allowing nonlinearity in the trajectory of percentage change in HbA_1c level following surgery. For HbA_1c level, we considered less than 7% as a goal range, consistent with American Diabetes Association goals for adults who are not pregnant, and more than 8% (well above the goal for many adults, including those with advanced vascular complications) to indicate poor control.³⁵

Sensitivity analyses considered 9-month and 12-month alternative lags from the last observed T2DM medication order to define remission (HbA_1c level <6.5%). To evaluate variability in medication data capture across different health systems, the primary analyses were repeated using only data from 8 integrated health systems, where infrastructure may enable more complete access to medication orders. Additional sensitivity analyses assessed 2 alternate censoring scenarios for inpatient stays: (1) no removal of inpatient medications or HbA_1c values and (2) censoring follow-up at the day of admission. Similar sensitivity analyses were applied to the relapse analyses. Analyses were conducted using R version 3.4.3 (R Foundation for Statistical Computing).

In this unmatched surgical cohort, the analytic sample included 9710 adults, primarily female (7051 female patients [72.6%]) with a mean (SD) age of 49.8 (10.5) years (Table 1). A total of 6233 (64.2%) underwent RYGB, and 3477 (35.8%) had SGs. The mean (SD) preoperative BMI was 49.0 (8.4). Patients were primarily white (6040 [72.2%]). Most (7904 [81.4%]) surgeries occurred between 2010 and 2014.

The mean (SD) preoperative HbA1c was 7.2% (1.3%), and patients took a mean (SD) of 1.66 (1.1) diabetes medications (range, 0-7 medications). The mean (SD) preoperative systolic and diastolic blood pressure were 130.5 (17.2) mm Hg and 73.7 (11.2) mm Hg, respectively. Weight-associated comorbidities were common. Patients who had RYGB had higher prevalence of some comorbidities, such as sleep apnea (RYGB: 3607 patients [57.9%]; SG: 1740 patients [50.0%]), nonalcoholic fatty liver disease (RYGB: 1914 patients [30.7%]; SG: 730 patients [21.0%]), and gastroesophageal reflux disease (RYGB: 2609 patients [41.9%]; SG: 1264 patients [36.4%]). The mean (SD) Charlson/Elixhauser score was negative (−0.089 [0.99]), consistent with the high hypertension prevalence in an otherwise relatively healthy sample.

Patients who had each procedure showed considerable weight loss 1 year after surgery (SG, −22.8% [95% CI, −23.1% to −22.5%]; RYGB, −29.1% [95% CI, −29.3% to −28.8%]); typically, weight regain then occurred. The groups maintained a mean body weight well below the baseline at 5 years (SG, −16.1% [95% CI, −17.3% to −14.8%]; RYGB, −24.1% [95% CI, −25.0% to −23.3%]). Typically, the RYGB group reflected 6.2% to 8.1% more total body weight loss than the SG group at each point (Figure 1; Table 2). This represents a 10.2-kg difference (95% CI, 8.3-12.1 kg; P < .001) in weight loss between RYGB and SG at 5 years.

The cohort was followed up for a median of 2.7 (interquartile range, 1.26-4.19) years. Type 2 diabetes remission occurred primarily in the first 2 years (Figure 2). Patients who underwent RYGB showed slightly (10%) higher T2DM remission rates than those who had SG (hazard ratio, 1.10 [95% CI, 1.04-1.16]; Table 3). We estimated that 59.2% (95% CI, 57.7%-60.7%) of patients who had RYGB vs 55.9% (95% CI, 53.9%-57.9%) of those who had SG experienced remission by 1 year, 84.3% (95% CI, 82.9%-85.5%) vs 81.5% (95% CI, 79.6%-83.2%) at 3 years, and 86.1% (95% CI, 84.7%-87.3%) vs 83.5% (95% CI, 81.6%-85.1%) at 5 years (Table 3).

Sensitivity analyses requiring 9-month and 12-month time frames without a diabetes medication prescription to define remission produced similar results to the primary analysis and its 6-month time frame, although differences between SG and RGB were not always statistically significant (eTable 2 in the Supplement). Analyses restricted to 8 integrated health systems yielded qualitatively similar results to the primary analyses, despite slightly higher cumulative remission rates for SG and RYGB (eTable 3 in the Supplement).

A total of 6141 patients with documented T2DM remission were eligible for the relapse analyses. Preoperation demographic and health features were similar to those of the larger T2DM cohort (eTable 4 in the Supplement). Mean (SD) preoperation HbA_1c levels were slightly lower (7.0% [1.1%]) vs 7.2% [1.3%]) as was the mean (SD) number of diabetes medications (1.5 (1.1) medications vs 1.7 [1.1] medications) and insulin use (2317 of 6141 [37.7%] vs 4692 of 9710 [48.3%]; eTable 4 in the Supplement). They were followed up for relapse for a median of 2.4 (0.003-10.35) years.

The T2DM relapse rate was lower for RYGB than SG (hazard ratio, 0.75 [95% CI, 0.67-0.84]). Estimated proportions of relapse for the RYGB and SG groups, respectively, were 8.4% (95% CI, 7.4%-9.3%) and 11.0% (95% CI, 9.6%-12.4%) 1 year after remission, 21.2% (95% CI, 19.1%-23.2%) and 27.2% (95% CI, 24.1%-30.1%) at 3 years, and 33.1% (95% CI, 29.6%-36.5%) and 41.6% (95% CI, 36.8%-46.1%) at 5 years (Table 3). Sensitivity analyses showed similar findings (eTable 5 and eTable 6 in the Supplement).

Patients who underwent RYGB experienced larger and more-sustained HbA_1c reductions than those using SG (Figure 1). In adjusted comparisons, patients who had RYGB showed a 1.12 percentage point drop in HbA_1c level (95% CI, 1.09-1.14 percentage points) over 1 year. This change was 0.22 (95% CI, 0.18-0.26) percentage points lower than seen for patients who had SG (Table 2). At 5 years, HbA_1c levels remained 0.80 (95% CI, 0.72-0.88) percentage points below baseline among patients who had RYGB and 0.35 (95% CI, 0.19-0.51) percentage points below baseline among patients who had SG, a difference of 0.45 (95% CI, 0.27-0.62) percentage points. The proportion with a poorly controlled HbA_1c level (≥8.0%) declined from baseline through 1 year of follow-up for both groups (patients who had RYGB, 24.6% [95% CI, 23.5%-25.7%] to 6.7% [95% CI, 6.0%-7.7%]; patients who had SG, 17.5% [95% CI, 16.24%-18.88%] to 8.3% [95% CI, 7.05%-9.79%]); it then increased, with 16.2% of patients who had RYGB and 22.4% of patients who had SG having HbA_1c levels greater than 8.0% 5 years after surgery. Following surgery, a well-controlled HbA_1c level (<6.5%) was consistently more common among patients who had RYGB (eFigure 2 in the Supplement).

Analyses for heterogeneity of treatment outcomes indicated that the likelihood of diabetes remission comparing RYGB vs SG varied significantly across DiaRem strata (eTable 7 in the Supplement). Patients with higher DiaRem scores showed greater likelihood of diabetes remission with RYGB compared with SG, with a statistically significant association for scores between 13 and 17. Among individuals with DiaRem scores in the 13-point to 17-point range, 83.4% (95% CI, 77.9%-87.6%) of patients who had RYGB had experienced T2DM remission by 5 years of follow-up vs 76.6% (95% CI, 70.0%-81.8%) of patients who had SG (eTable 8 in the Supplement).

In this sample of US adults with T2DM and bariatric surgery, 56% to 59% of those with RYGB or SG experienced T2DM remission in the year following surgery and 84% to 86% did so within 5 years of follow-up. However, T2DM relapse was common; 33% of patients who had RYGB and 42% of patients who had SG relapsed within 5 years of initial remission. The glycemic control of patients who had RYGB and SG showed sustained improvements from the samples’ baseline mean HbA_1c level of 7.2%, with an estimated mean HbA_1c level 0.80 percentage points below baseline for the RYGB group 5 years after surgery vs 0.35 percentage points below baseline for the SG group. While both groups experienced considerable weight loss, patients who had RYGB lost more weight and maintained weight loss better than did patients who had SG.

Overall, these results indicate that RYGB is associated with better long-term T2DM and weight outcomes than SG in real-world clinical settings. This is at odds with recent randomized clinical trials that compared T2DM outcomes of RYGB and SG and found no significant differences.^19-21 Those trials had longer duration of follow-up but much smaller sample sizes, which may have limited their power to detect differences between the procedures. Also, patients who are willing to undergo randomization between RYGB and SG and surgeons who have equal skill and equipoise for RYGB and SG are likely different from those who choose either RYGB or SG in uncontrolled settings. Thus, while the more rigorous, randomized clinical trial data indicate that RYGB and SG perform similarly in highly controlled environments, in everyday practice, the outcome differences may be larger.

As expected,^1,6,7,22,36 some patient subgroups showed lower rates of T2DM remission. Our findings indicate that patients with lower preoperative probability for T2DM remission (11%-33%) may be more likely to achieve T2DM remission with RYGB compared with SG. Estimating the likelihood of T2DM remission could help inform patients’ and clinicians’ discussions of procedure choice. Preoperative insulin use, older age, higher HbA_1c level, and more complex T2DM medication regimens predispose patients to lower probability of T2DM remission in the DiaRem scoring system.³⁴ Informed decision-making for procedure choice should also consider other factors, such as the potential for adverse events.

A range of T2DM remission rates are found in studies of bariatric surgery,^6,7,12,37-41 reflecting varying follow-up time, remission definitions, and population characteristics (eg, insulin use, HbA_1c level).³⁸ The cumulative remission rates over 80% for SG or RYGB in PBS are consistent with or somewhat higher than estimates from systematic reviews or meta-analyses (54%-78%)^4,37,40 and similar to findings (72%; all procedures) from 3 US health systems.⁶ Literature on T2DM relapse is more limited. Published relapse estimates range from approximately 25% to 53%^7,12,41 and are typically calculated across a mix of procedure types and time frames; those ranges are consistent with PBS’s 5-year cumulative relapse rates.

The large PBS sample and its comparison of remission and relapse rates across procedures, extended follow-up, and evaluation of remission across patient subgroups contribute unique insight to the literature. Findings also contribute to ongoing dialogue about leveraging real-world evidence to understand health and improve care.^42-44 Such data can reflect generalizable populations of patients and clinicians, as well as actual health care practices and settings.⁴⁴ The data standardization and curation processes of PCORnet⁴⁵ help mitigate data quality concerns that have been raised regarding analyses of electronic health record data,^42,44 as do the consistency of our findings with prior literature. Our analyses suggest that, coupled with rigorous attention to study design and analytic methods, PCORnet data can be a valuable resource for health research.

This study has limitations. Because of the observational study design, procedure choice may have been influenced by unmeasured factors that impact the surgical effect on diabetes. Despite direct adjustment and the use of propensity scores, confounding may persist. Using ICD-9-CM codes to assess baseline health may underestimate comorbidity prevalence. The PBS definitions for T2DM relapse and remission rely on medication-prescribing data. To the extent that prescriptions were not filled, medication use may be overestimated. Some patients may have had T2DM medications ordered outside of the health systems in the study. All dates were normalized to the date of surgery, so within a calendar year, we cannot differentiate patients with loss to follow-up from those for whom the study end date had been reached. Future work should address the potential role of weight loss in mediating diabetes remission and relapse.

Similar to prior research,⁷ 19% of the cohort was not prescribed diabetes medication preoperatively. Some people may have used lifestyle alone to treat diabetes.⁴⁶ Undiagnosed diabetes is common,⁴⁷ and others may have been diagnosed during the preoperative evaluation—emphasizing the importance of care coordination between medical and surgical health professions among patients considering bariatric surgery.

In conclusion, among patients with T2DM who underwent RYGB or SG, most experienced T2DM remission at some point over 5 years of follow-up. While SG and RYGB resulted in similar rates of initial T2DM remission, RYGB was associated with larger and more persistent improvements in glycemic control and 25% lower rates of T2DM relapse compared with SG. Patients with more advanced T2DM at the time of surgery for whom remission is more difficult to achieve (eg, those with older age, insulin use, more complex T2DM medications, and/or poor glycemic control) may expect larger improvements in T2DM with RYGB compared with SG. On the other hand, for patients with higher likelihood of T2DM remission, RYGB and SG are likely to yield similar 5-year T2DM outcomes. For patients, clinicians and policy makers to make informed decisions about which procedure is best suited to patients’ personal situations, additional data are needed to understand the adverse event profile of the procedures as well as patient values regarding procedure choice and the role of surgery relative to other aspects of lifelong weight management.

Accepted for Publication: January 15, 2020.

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2020 McTigue KM et al. JAMA Surgery.

Corresponding Author: Kathleen McTigue, MD, Department of Medicine, University of Pittsburgh, 230 McKee Pl, Ste 600, Pittsburgh, PA 15213 (kmm34@pitt.edu).

Published Online: March 4, 2020. doi:10.1001/jamasurg.2020.0087

Correction: This article was corrected on March 25, 2020, to fix an error in the name of a healthcare organization. The name was rendered as “the National Patient-Centered Clinical Research Network (PCORnet),” but it should have been “PCORnet, the National Patient-Centered Clinical Research Network.” This occurred once in the Introduction section and once in the Funding/Support section of the Article Information section. Both have been fixed online.

Author Contributions: Drs McTigue and Arterburn had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: McTigue, Wellman, Coley, Toh, Janning, Williams, Arterburn.

Acquisition, analysis, or interpretation of data: McTigue, Wellman, Nauman, Anau, Coley, Tice, Coleman, Courcoulas, Pardee, Toh, Cook, Sturtevant, Horgan, Arterburn.

Drafting of the manuscript: McTigue, Wellman, Anau, Coley, Coleman, Janning, Arterburn.

Critical revision of the manuscript for important intellectual content: McTigue, Wellman, Nauman, Anau, Coley, Tice, Courcoulas, Pardee, Toh, Williams, Cook, Sturtevant, Horgan, Arterburn.

Statistical analysis: Wellman, Coley, Toh, Cook.

Obtained funding: McTigue, Anau, Arterburn.

Administrative, technical, or material support: McTigue, Nauman, Anau, Coleman, Courcoulas, Pardee, Sturtevant, Horgan.

Supervision: Coleman, Arterburn.

Other—patient perspective: Janning.

Conflict of Interest Disclosures: Dr Courcoulas reports grants from Covidien/Ethicon Johnson & Johnson, during the conduct of the study. Dr Tavakkoli reports personal fees from Medtronic and AMAG pharmaceuticals. Dr Jones reports personal fees from Allurion. Mr Nadglowski reports other support from the Obesity Action Coalition outside the submitted work.

Funding/Support: The PCORnet Study reported in this article was conducted using PCORnet, the National Patient-Centered Clinical Research Network, an initiative funded by the Patient-Centered Outcomes Research Institute (grant OBS-1505-30683).

Role of the Funder/Sponsor: The funder did not have a role in the study design; in the collection, management, analysis, and interpretation of data; in the preparation, review, or approval of the manuscript; and in the decision to submit the manuscript for publication.

PCORnet Bariatric Study Collaborative: Corrigan L. McBride, MD, and James McClay, MD, University of Nebraska Medical Center, Omaha; Jeanne M. Clark, MD, Johns Hopkins University and Health Plan, Baltimore, Maryland; Thomas H. Inge, MD, Children’s Hospital Colorado and University of Colorado, Denver; Michelle R. Lent, PhD, Geisinger Health System, Danville, Pennsylvania; David G. Schlundt, PhD, Vanderbilt University, Nashville, Tennessee; Meredith Duke, MD, University of North Carolina–Chapel Hill; Steven R. Smith, MD, Florida Hospital–Translational Research Institute, Orlando; Andrew O. Odegaard, PhD, University of California, Irvine; Nirav K. Desai, MD, Boston Children’s Hospital, Boston, Massachusetts; Ali Tavakkoli, MD, and Elizabeth Cirelli, MS, Brigham and Women’s Hospital, Boston, Massachusetts; Stavra A. Xanthakos, MD, Cincinnati Children's Medical Center, Cincinnati, Ohio; Laura J. Rasmussen-Torvik, PhD, Northwestern University Feinberg School of Medicine, Chicago, Illinois; Marc P. Michalsky, MD, Nationwide Children’s Hospital, Columbus, Ohio; Matthew F. Daley, MD, Institute for Health Research, Kaiser Permanente Colorado, Aurora; Gabrielle Purcell, MPH. University of California; San Francisco; Sameer Murali, MD, Southern California Permanente Medical Group, Fontana; Ana Emiliano, MD, and Rhonda G. Kost, MD, The Rockefeller University, New York, New York; Caroline M. Apovian, MD, and Donald Hess, MD, Boston Medical Center, Boston, Massachusetts; Cynthia A. Blalock, APRN, Vanderbilt University Medical Center, Nashville, Tennessee; Elisha Malanga, BS, COPD Foundation, Miami, Florida; Jay R. Desai, MD, HealthPartners Institute, Bloomington, Minnesota; Joe Nadglowski, BS, Obesity Action Coalition, Tampa, Florida; John H. Holmes, PhD, University of Pennsylvania Perelman School of Medicine, Philadelphia; Joseph Vitello, MD, Jesse Brown VA Medical Center, Chicago, Illinois; Michael A. Horberg, MD, Kaiser Permanente Mid-Atlantic Permanente Medical Group, Rockville, Maryland; Robert T. Greenlee, PhD, Marshfield Clinic Research Institute, Marshfield, Wisconsin; Stephanie L. Fitzpatrick, PhD, Kaiser Permanente Center for Health Research, Portland, Oregon; Roni Zeiger, MD, Smart Patients, Inc, Mountain View, California; Molly B. Conroy, MD, University of Utah, Salt Lake City; Douglas S. Bell, MD, David Geffen School of Medicine at UCLA, Los Angeles, California; Jamy Ard, MD, Wake Forest School of Medicine, Salem, North Carolina; Jing Bian, PhD, University of Florida, Gainesville; Bipan Chan, MD, Loyola University Medical Center, Maywood, Illinois; Michael A. Edwards, MD, Temple University, Philadelphia, Pennsylvania; Christina Wee, MD, and Daniel B. Jones, Beth Israel Deaconess Medical Center, Boston, Massachusetts; Jennifer L. Kraschnewski, MD, Penn State University, College of Medicine, Hershey, Pennsylvania; Kirk Reichard, MD, Nemours AI DuPont Hospital for Children, Wilmington, Delaware; Howard S. Gordon, MD, and David O. Meltzer MD, University of Illinois, Chicago; Erin D. Roe, MD, Baylor Scott & White, Dallas, Texas; William Richardson, MD, Ochsner Clinic, New Orleans, Louisiana; Sameer Malhotra, MD, Weill Cornell Medicine, New York, New York; Lindsay G. Cowell, PhD, University of Texas Southwestern Medical Center, Dallas; Lydia A. Bazzano, MD, PhD, Tulane University, New Orleans, Louisiana; Jefferey S. Brown, Sengwee Toh, ScD, Jessica L. Sturtevant, MS, and Casie Horgan, MPH, Department of Population Medicine, Harvard Pilgrim Health Care Institute and Harvard Medical School, Boston, Massachusetts; Anita Courcoulas, MD, Department of Surgery, University of Pittsburgh, Pittsburgh, Pennsylvania, and Kathleen McTigue, MD, Departments of Medicine and Epidemiology, University of Pittsburgh, Pittsburgh, Pennsylvania; R. Yates Coley, PhD, David Arterburn, MD, Robert Wellman, MS, Jane Anau, BS, Roy E. Pardee, JD, and Andrea J. Cook, PhD, Kaiser Permanente Washington Health Research Institute, Seattle; Karen J. Coleman, PhD, Kaiser Permanente Southern California, Department of Research and Evaluation, Pasadena; Cheri D. Janning, MS, Duke Clinical & Translational Science Institute, Durham, North Carolina; Neely Williams, MDiv, Mid-South Clinical Data Research Network and Meharry–Vanderbilt Alliance Community Partner, Nashville, Tennessee.

Disclaimer: The views expressed in this article are solely those of the authors and do not reflect the views of PCORnet or PCORI. Dr McTigue attests that all listed authors meet authorship criteria and nobody meeting authorship criteria has been omitted.

Additional Contributions: The study team also wishes to acknowledge the clinicians, analysts, and staff at the 34 health systems which contributed to the study: Stephen R. Perry, Kin Lam, David Hawkes, Thomas Dundon, and Kelli Kinsman, Kaiser Permanente Washington Health Research Institute, Shelly Sital, The Chicago Community Trust, Elizabeth Tarlov, University of Illinois at Chicago, Jasmin Phua, Medical Research Analytics and Informatics Alliance, Mia Gallagher, Lindsey Petro, Beth Syat, Harvard Pilgrim Health Care Institute and Harvard Medical School, Prakash Nadkarni, and Elizabeth Chrischilles, University of Iowa, Steffani Roush, and Laurel Verhagen, Marshfield Clinic Research Institute, Umberto Tachincardi, and Lawrence P. Hanrahan, University of Wisconsin, Phillip Reeder, Shiby Antony, Rania AlShahrouri, University of Texas–Southwestern Medical Center, Bret Gardner, James Campbell, Russell Buzalko, and Jay Pedersen, University of Nebraska Medical Center, Dan Connolly, and Russel Waitman, University of Kansas Medical Center, Russel Rothman, David Crenshaw, and Katie Worley, Vanderbilt University Medical Center, Emily Pfaff, Robert Bradford, Kellie Walters, Tim Carey, Timothy Farrell, and D. Wayne Overby, University of North Carolina, Maija Neville-Williams, The Rockefeller University, Elizabeth Shenkman, William Hogan, Kathryn McAuliffe, and Gigi Lipori, University of Florida, Rebecca Zuvich Essner, Florida Hospital, Howard Su, Michael George, Michael J. Becich, Barbara Postol, Giselle G. Hamad, Ramesh C. Ramanathan, Bestoun H. Ahmed, William F. Gourash, Bill Shirey, Chuck Borromeo, John Milnes, Nickie Cappella, and Desheng Li, University of Pittsburgh, Anthony T. Petrick, H. Lester Kirchner, Geisinger Health System, Daniel E. Ford, Michael A. Schweitzer, Karl Burke, Harold Lehmann, Megan E. Gauvey-Kern, and Diana Gumas. Johns Hopkins, Rachel Hess, Meghan Lynch, and Reid Holbrook, University of Utah, Jody McCullough, Matt Bolton, Wenke Hwang, Ann Rogers, Alison Bower, and Cynthia Chuang, Penn State, Cecilia Dobi, Mark Weiner, Anuradha Paranjape, Sharon J. Herring, and Patricia Bernard, Temple University, Janet Zahner, Parth Divekar, Keith Marsolo, and Lisa Boerger, Cincinnati Children’s Hospital, Kimberly J. Holmquist, Kaiser Permanente Southern California, Ray Pablo and Robynn Zender, University of California at Irvine, Lucila Ohno-Machado, Paulina Paul, and Michele Day, University of California at San Diego, Thomas Carton, Elizabeth Crull, and Iben McCormick-Ricket, Louisiana Public Health Institute, Ashley Vernon, Malcom Robinson, Scott Shikora, David Spector, Eric Sheu, Edward Mun, Matthew Hutter, Shawn Murphy, Jeffrey Klann, and Denise Gee, Partners Healthcare, Daniel Jones, Benjamin Schneider, Griffin Weber, and Robert Andrews, Beth Israel Deaconess Medical Center, Brian Carmine, Miguel Burch, and Galina Lozinski, Boston Medical Center, Ken Mandl, Jessica Lyons, and Margaret Vella, Harvard Medical School, and Joseph Skelton and Kun Wei, Wake Forest Integrated Health System. Some of these individuals were compensated for their contributions.