Association Between Bariatric Surgery and Macrovascular Disease Outcomes in Patients With Type 2 Diabetes and Severe Obesity

Key Points

Question  For patients with severe obesity and type 2 diabetes, is there an association between bariatric surgery and incident macrovascular disease (defined as first occurrence of acute myocardial infarction, unstable angina, percutaneous coronary intervention, coronary artery bypass grafting, ischemic stroke, hemorrhagic stroke, carotid stenting, or carotid endarterectomy)?

Findings  In this retrospective cohort study of patients with type 2 diabetes and severe obesity that included 5301 who underwent bariatric surgery and 14 934 control patients without surgery, bariatric surgery was associated with a significantly lower risk of macrovascular events at 5 years’ follow-up (2.1% vs 4.3% at 5 years; hazard ratio, 0.60).

Meaning  Bariatric surgery was associated with a lower risk of incident major macrovascular events.

Importance  Macrovascular disease is a leading cause of morbidity and mortality for patients with type 2 diabetes, and medical management, including lifestyle changes, may not be successful at lowering risk.

Objective  To investigate the relationship between bariatric surgery and incident macrovascular (coronary artery disease and cerebrovascular diseases) events in patients with severe obesity and type 2 diabetes.

Design, Setting, and Participants  In this retrospective, matched cohort study, patients with severe obesity (body mass index ≥35) aged 19 to 79 years with diabetes who underwent bariatric surgery from 2005 to 2011 in 4 integrated health systems in the United States (n = 5301) were matched to 14 934 control patients on site, age, sex, body mass index, hemoglobin A_1c, insulin use, observed diabetes duration, and prior health care utilization, with follow-up through September 2015.

Exposures  Bariatric procedures (76% Roux-en-Y gastric bypass, 17% sleeve gastrectomy, and 7% adjustable gastric banding) were compared with usual care for diabetes.

Main Outcomes and Measures  Multivariable-adjusted Cox regression analysis investigated time to incident macrovascular disease (defined as first occurrence of coronary artery disease [acute myocardial infarction, unstable angina, percutaneous coronary intervention, or coronary artery bypass grafting] or cerebrovascular events [ischemic stroke, hemorrhagic stroke, carotid stenting, or carotid endarterectomy]). Secondary outcomes included coronary artery disease and cerebrovascular outcomes separately.

Results  Among a combined 20 235 surgical and nonsurgical patients, the mean (SD) age was 50 (10) years; 76% of the surgical and 75% of the nonsurgical patients were female; and the baseline mean (SD) body mass index was 44.7 (6.9) and 43.8 (6.7) in the surgical and nonsurgical groups, respectively. At the end of the study period, there were 106 macrovascular events in surgical patients (including 37 cerebrovascular and 78 coronary artery events over a median of 4.7 years; interquartile range, 3.2-6.2 years) and 596 events in the matched control patients (including 227 cerebrovascular and 398 coronary artery events over a median of 4.6 years; interquartile range, 3.1-6.1 years). Bariatric surgery was associated with a lower composite incidence of macrovascular events at 5 years (2.1% in the surgical group vs 4.3% in the nonsurgical group; hazard ratio, 0.60 [95% CI, 0.42-0.86]), as well as a lower incidence of coronary artery disease (1.6% in the surgical group vs 2.8% in the nonsurgical group; hazard ratio, 0.64 [95% CI, 0.42-0.99]). The incidence of cerebrovascular disease was not significantly different between groups at 5 years (0.7% in the surgical group vs 1.7% in the nonsurgical group; hazard ratio, 0.69 [95% CI, 0.38-1.25]).

Conclusions and Relevance  In this observational study of patients with type 2 diabetes and severe obesity who underwent surgery, compared with those who did not undergo surgery, bariatric surgery was associated with a lower risk of macrovascular outcomes. The findings require confirmation in randomized clinical trials. Health care professionals should engage patients with severe obesity and type 2 diabetes in a shared decision making conversation about the potential role of bariatric surgery in the prevention of macrovascular events.

Macrovascular disease, including coronary artery disease and cerebrovascular disease, is one of the leading causes of morbidity and mortality for patients with type 2 diabetes. As a result, clinical guidelines for the management of type 2 diabetes emphasize lowering macrovascular disease risk factors by optimizing glycemic control, blood pressure, and serum lipid levels.^1,2 This multifactorial approach has been shown to substantially reduce the incidence of macrovascular complications.^2,3 However, most patients with type 2 diabetes do not achieve these recommended treatment goals,⁴ resulting in continued morbidity and costs.

There is evidence from randomized trials that weight loss interventions—including intensive lifestyle modification, pharmacotherapy, and bariatric surgery—can improve macrovascular disease risk factors in patients with type 2 diabetes. To our knowledge, clinical trials have not yet shown that intensive lifestyle intervention and pharmacotherapy for obesity can reduce macrovascular events.^5-9 In comparison, patients who have bariatric surgery have better glycemic control and greater macrovascular risk factor reduction than those who get both intensive lifestyle and medical treatment combined.^10-12

Several observational studies have provided evidence that bariatric surgery may reduce macrovascular complications of type 2 diabetes when compared with usual medical care,¹³ but limitations of these studies include the inability to study contemporary bariatric procedures because of small numbers of patients¹⁴ and unavailability of body mass index measurements for identifying a cohort of nonsurgical matches.¹⁵

To address these concerns, a study was conducted to test the hypothesis that patients undergoing contemporary bariatric procedures would experience a lower rate of macrovascular events than matched patients with severe obesity and type 2 diabetes who received usual care in 4 integrated health insurance and care delivery systems in the United States.

A retrospective observational matched cohort study was conducted of adults with type 2 diabetes who underwent bariatric surgery between 2005 and 2011 while enrolled in 1 of 4 US integrated health care systems from the Health Care Systems Research Network: Kaiser Permanente Washington in Washington state, HealthPartners in Minnesota, Kaiser Permanente Northern California, and Kaiser Permanente Southern California. All study procedures were reviewed in advance and approved by the institutional review board at each site and permitted to conduct the research without explicit consent from participants.

At each study site, standardized electronic medical records, insurance claims, and other data systems¹⁶ were used to extract information on enrollment; insurance coverage; demographics; blood pressure; height; weight; laboratory values; medications dispensed; deaths; outpatient, inpatient, and emergency department use; and diagnosis and procedure codes of all enrollees. Race/ethnicity data, which were collected as part of routine clinical care, were extracted from electronic medical records, where they had been entered by clinical staff based on patient self-report using fixed categories.

The bariatric surgery population included adults (aged 19-79 years) with severe obesity (body mass index [BMI, calculated as weight in kilograms divided by height in meters squared] ≥35) and type 2 diabetes who had a primary (first observed) bariatric surgical procedure between January 1, 2005, and December 31, 2011. A combination of bariatric registries, medical record reviews, International Classification of Diseases, Ninth Revision (ICD-9) codes (43.89, 44.31, 44.38, 44.39, 44.68, 44.69, and 44.95), and Current Procedural Terminology (CPT) procedure codes (43633, 43644, 43645, 43659, 43770, 43775, 43842, 43843, 43844, 43845, 43846, and 43847) were used to identify bariatric procedures. Patients were classified as having type 2 diabetes if they met 1 of 2 criteria at the time of the procedure: (1) type 2 diabetes, defined as a hemoglobin A_1c (HbA_1c) level greater than or equal to 6.5% (48 mmol/mol) or fasting plasma glucose level greater than or equal to 126 mg/dL (to convert to mmol/L, multiply by 0.0555) at the most recent measurement within 2 years prior to surgery, or (2) medication-treated type 2 diabetes, defined as a current prescription for any oral or injectable diabetes medication at the time of bariatric surgery.

After population selection, the following exclusion criteria were applied based on information in the 2 years before the index date: (1) less than 1 full year of continuous enrollment, (2) a history of cancer (except nonmelanoma skin cancer), (3) pregnancy, (4) gestational diabetes when it was the sole diabetes diagnosis, (5) metformin as the sole indicator of possible type 2 diabetes (ie, no other type 2 diabetes laboratories or diagnoses), (6) pre-existing coronary artery disease or cerebrovascular disease (defined below) or the use of prescription antiplatelet medication, (7) maximum preoperative BMI less than 35, (8) missing preoperative BMI or HbA_1c level in the 2 years before surgery, and (9) unable to match to at least 1 nonsurgical patient (Figure 1).

Quiz Ref IDFor each patient with bariatric surgery, we identified up to 3 matched nonsurgical controls via a multistep process. First, among all patients with diabetes and BMI greater than or equal to 35 who did not undergo bariatric surgery during the study period (n = 320 345), we identified a pool of potential controls who were enrolled at the time of the surgery, satisfied the study inclusion/exclusion criteria, and matched the patient with bariatric surgery on the basis of study site, sex, age (±10 years), BMI (±5), HbA_1c level (±absolute 1.5%), and insulin use. Second, for each control in the pool, we calculated their Mahalanobis distance (a measure of the distance that accounts for correlation between variables) from the bariatric patient on the basis of age, BMI, HbA_1c level, diabetes duration, and the number of days of health care utilization in the 7 to 24 months prior to the date of surgery.¹⁷ Third, up to 3 controls were selected based on the shortest Mahalanobis distance. Throughout, nonsurgical patients could only be used as a control for 1 surgical patient (matching without replacement). Additional details on the process we used to establish the matched cohort are provided in the eAppendix in the Supplement.

Quiz Ref IDThe a priori primary outcome measure was time to incident macrovascular disease (composite indicator of the first occurrence of a coronary artery disease or cerebrovascular event). Using methods from prior studies,^17,18 coronary artery disease events included acute myocardial infarction, unstable angina, percutaneous coronary intervention, or coronary artery bypass grafting identified via ICD-9 diagnosis codes (410.x; 411.x as principal diagnosis or 411.x as secondary diagnosis with 414.x as principal diagnosis), ICD-9 procedure codes (36.01, 36.02, 36.03, 36.05, 36.06, 36.07, 36.10, 36.11, 36.12, 36.13, 36.14, 36.15, 36.16, 36.17, 36.19, 36.31, 36.32, 36.33, and 36.34), and CPT-4 procedure codes (92982, 92984, 92995, 92996, 92980, 92981, 33510, 33511, 33512, 33513, 33514, 33516, 33517, 33518, 33519, 33521, 33522, 33523, 33530, 33533, 33534, 33535, 33536, 93539, and 93540). Cerebrovascular events included ischemic stroke, hemorrhagic stroke, carotid stenting, or endarterectomy procedures identified via ICD-9 diagnosis codes (433.x1, 434.x1, 436.0, 430.x, and 431.x), ICD-9 procedure codes (38.12, 0.61, and 0.63), and CPT-4 procedure codes (37215, 37216, 0075T, 0076T, 35301, 37205, and 37206). Patients were censored at the first incidence of cancer (excluding nonmelanoma skin cancer), disenrollment, death, or study end (eTable 2 in the Supplement).

In secondary analyses, each of the coronary artery disease and cerebrovascular outcomes was considered separately. All-cause mortality was also examined as a post hoc exploratory outcome. As in prior studies,¹⁹ death information was drawn from a combination of electronic medical records, administrative databases, and state death indices.

Cox proportional hazards regression models, fit via the usual partial likelihood, were used to investigate the association between bariatric surgery vs usual care (nonsurgical controls) and incident outcomes. Patients were followed from the index date (the date of bariatric surgery or, for nonsurgical patients, the date of surgery for the patient to whom they had been matched) until the first occurrence of either incident outcome or a censoring event. Based on preliminary analyses, the proportional hazards assumption was assessed to not hold for bariatric surgery vs usual care (P < .001 for an interaction with log-time in the Cox model). We therefore fit a flexible time-varying hazard ratio (HR) association as a function of time since surgery, using restricted cubic splines with knots at the fifth, 35th, 65th, and 95th percentiles of the observed follow-up time scale.²⁰

Cox regression models were used for the primary and secondary outcomes and were adjusted for a priori–identified potential confounders: age, race/ethnicity (Hispanic, non-Hispanic white, non-Hispanic black, or non-Hispanic other), surgical year, BMI, smoking status (current, former, or never), duration of observed diabetes before surgery (defined as first observed diagnosis, laboratory value, or prescription indicating type 2 diabetes), insulin use, oral diabetes medication use, uncontrolled blood pressure (defined as either systolic ≥140 or diastolic ≥90 mm Hg at 2 consecutive measures on different days), use of angiotensin converting enzyme inhibitor, angiotensin receptor blocker medications, use of any other antihypertensive medication, insurance type (commercial, Medicare, Medicaid, or other), estimated glomerular filtration rate, low-density lipoprotein (LDL) cholesterol level greater than or equal to 100 mg/dL (to convert to mmol/L, multiply by 0.0259), serum triglyceride level greater than or equal to 150 mg/dL (to convert to mmol/L, multiply by 0.0113), use of cholesterol-lowering medication (statin or other), and history of peripheral arterial disease, hypertension, dyslipidemia, or microvascular disease (diabetic retinopathy, diabetic neuropathy, or diabetes with renal manifestations; end-stage renal disease; or dialysis) before surgery (defined based on ICD-9 and CPT-4 codes). Because of potential variation in care between health care systems, study site (HealthPartners, Kaiser Permanente Southern California, Kaiser Permanente Northern California, and Kaiser Permanente Washington) was adjusted for via stratification of the baseline hazard function. Race/ethnicity was included in the statistical models because it could confound the outcomes by either influencing a patient’s ability to receive bariatric surgery or by having a direct effect on the macrovascular outcomes by itself.

Missing data were encountered at baseline for race/ethnicity, smoking status, blood pressure, estimated glomerular filtration rate, LDL cholesterol, and triglyceride levels. We performed multiple imputation via chained equations, with unordered categorical variables and multinomial logistic regression for imputation of race/ethnicity and smoking status; logistic regression for elevated blood pressure, LDL cholesterol, and triglyceride levels; and linear regression for estimated glomerular filtration rate.²¹ Ten imputed data sets were generated, with the results combined using Rubin’s rules. Additional detail is provided in the eAppendix in the Supplement.

Four sensitivity analyses were conducted to examine the effect of (1) not adjusting for potential confounders that were not included in the matching process (vs adjusting for those confounders in our main analysis), (2) using 1-to-1 and 10-to-1 matching, (3) incorporating all covariates into our Mahalanobis distance calculation (vs our original matching approach), and (4) an additional sensitivity analysis to assess the robustness of the 5-year results to unmeasured confounding using the E-value methodology of VanderWeele and Ding.²² This estimates what the relative risk would have to be for any unmeasured confounder to overcome the observed association of bariatric surgery with macrovascular disease in this study (see the eAppendix in the Supplement for details).²²

The a priori level of statistical significance was set at a 2-sided P value of .05 for all analyses. Because the significance threshold for the secondary analyses was not adjusted for multiple comparisons, these should be interpreted as exploratory. Throughout, we used SAS version 9.4 (SAS Institute) for data manipulation, Stata version 15.1 (StataCorp) for multiple imputation and analysis, and R (R Foundation) for visualization.

Table 1 presents descriptive statistics for the final analytic sample of 5301 surgical patients and 14 934 matched nonsurgical patients. Surgical patients were primarily middle-aged, female, had commercial insurance, and more than 40% were racial/ethnic minorities. Quiz Ref IDSeventy-six percent of patients with bariatric surgery underwent Roux-en-Y (RYGB) from 2005 to 2011, 17% underwent sleeve gastrectomy (SG), and 7% underwent adjustable gastric band. Among all patients at the index date, 52% of patients had a BMI of 40 to 49.9, the mean HbA_1c level was 7.2%, and 49% had type 2 diabetes for 5 years or more before baseline. Fewer surgical patients had been current smokers in the 2 years before surgery (9% vs 13%). More nonsurgical patients had a diagnosis of diabetic retinopathy, neuropathy, and renal manifestations, including end-stage renal disease or dialysis. Standardized differences indicate imbalance between surgical and matched nonsurgical patients on a number of variables, including insurance type, days of health care usage, use of other antihypertensive medications, and baseline prevalence of peripheral arterial disease.

The 1-, 3-, and 5-year retention rates in the analytic sample were similar for nonsurgical (93%, 79%, and 68%, respectively) and surgical (91%, 77%, and 67%, respectively) patients. Median follow-up durations were 4.7 years (interquartile range, 3.3-6.2) for surgical and 4.6 years (interquartile range, 3.1-6.1) for nonsurgical patients. Missing data were encountered for some characteristics at baseline, including race/ethnicity (12% in the surgical group vs 17% in the nonsurgical group), self-reported smoking status (2% vs 4%, respectively), blood pressure (0% vs 1%, respectively), triglyceride levels (1% vs 4%, respectively), and LDL cholesterol level (67% vs 65%, respectively).

Figure 2 and Table 2 provide estimates of the cumulative probability of incident macrovascular disease over time after bariatric surgery and usual nonsurgical care (for matched controls), as well as the cumulative probability of each of the indicators of macrovascular disease (coronary artery disease and cerebrovascular events). Quiz Ref IDThe 1-, 3-, 5-, and 7-year rates of incident macrovascular disease were 0.5%, 1.1%, 2.1%, and 3.2%, respectively, after bariatric surgery, and 1.1%, 2.6%, 4.3%, and 6.2% for the matched controls. The cumulative number of macrovascular events at each of these time points is shown in eTable 3 in the Supplement. Rates of incident coronary artery disease and cerebrovascular events were also lower for patients with bariatric surgery than for matched nonsurgical patients (Figure 2B and C; Table 2).

Table 3 and Figure 3 show the results from the adjusted multivariable Cox models investigating the association between bariatric surgery vs usual care and incident macrovascular disease (see Table 4 for the fully adjusted model with all covariates). Because the effect of surgery was specified in the model as being time-varying using cubic splines, calculating the HRs shown in Table 3 required using all the components of Table 4 and specifying the time points at which the HRs were estimated. Table 3 shows the HRs estimated at 1, 3, 5, and 7 years and Figure 3 shows the HR estimation for all time points. In Table 4, the most appropriate interpretation of the surgery point estimate in row one is the HR comparing surgery vs no surgery at time zero (the index date). The point estimates for the surgery-by-spline interactions are uninterpretable because of the underlying basis function for the spline. The point estimates for other variables in the model should not be interpreted as unconfounded because this was a matched cohort with respect to surgery and the consideration of confounding was with respect to surgery only, not with respect to these other covariates.

As shown in Table 3, patients who underwent bariatric surgery had a significantly lower risk of incident macrovascular disease at 5 years when compared with matched nonsurgical patients who had not received surgery (2.1% in the surgical group vs 4.3% in the nonsurgical group; HR, 0.60 [95% CI, 0.42-0.86]), as well as a lower 5-year incidence of coronary artery disease (1.6% in the surgical group vs 2.8% in the nonsurgical group; HR, 0.64 [95% CI, 0.42-0.99]). Quiz Ref IDThe incidence of cerebrovascular disease was not significantly different between groups at the 5-year time point (0.7% in the surgical group vs 1.7% in the nonsurgical group; HR, 0.69 [95% CI, 0.38-1.25]).

Results of the sensitivity analyses are provided in the eFigure in the Supplement. A matched, unadjusted analysis of our primary outcome was not significantly different than the fully adjusted model (eFigure, A in the Supplement). Primary outcome results using 3-to-1 matching were qualitatively similar to using 1-to-1 and 10-to-1 matching (eFigure, B in the Supplement). Primary outcome results were similar to results obtained when all covariates were included in the Mahalanobis distance calculation for matching (eFigure, C in the Supplement). The E-values (relative risk) for the point estimate and upper confidence bound for incident macrovascular disease at 5 years were 2.72 and 1.60, respectively (eAppendix in the Supplement).

The rates of mortality at 1, 3, 5, and 7 years were 0.4%, 0.9%, 1.3%, and 2.0%, respectively, for surgical and 0.9%, 2.7%, 4.5%, and 6.4% for nonsurgical patients (Table 2; Figure 2D). The cumulative number of deaths at each of these time points is shown in the eAppendix in the Supplement. The risk of all-cause mortality was significantly lower at 5 years (HR, 0.33 [95% CI, 0.21-0.52]) among patients who underwent bariatric surgery relative to the matched nonsurgical patients (Table 3 and Figure 3D).

In this retrospective cohort study of patients with type 2 diabetes and severe obesity, bariatric surgery was associated with a lower risk of major macrovascular outcomes compared with usual medical care. These findings have strong biological plausibility and are consistent with other research. Randomized trials have demonstrated that bariatric procedures are more effective than the best-available intensive medical and lifestyle interventions in promoting weight loss, improving glycemic control and serum lipid levels, and reducing the need for medications used to control diabetes, hypertension, and dyslipidemia.^10-12,23-25 Improvements in these parameters after bariatric surgery result in lower cardiovascular risk scores, as measured through validated tools such as the Framingham risk equation.^26,27 The effect of bariatric surgery also appeared to be dose-dependent, where patients who underwent RYGB procedures (which is associated with greatest initial weight loss) had larger improvements in glycemic control and components of the metabolic syndrome than patients who underwent SG or underwent the adjustable gastric band procedure,^28,29 although there may also be weight-independent effects of these procedures on glycemic control.¹³

A few other observational studies have also found an association between bariatric surgery and fewer macrovascular events when compared with usual care.^13-15,30 The landmark Swedish Obese Subjects study prospectively recruited 2010 patients who underwent bariatric surgery and 2037 matched patients who received usual care and found that surgery was associated with an adjusted hazard ratio of 0.67 (95% CI, 0.54-0.83) for cardiovascular events (myocardial infarction or stroke).¹⁴ A subanalysis of 343 surgical and 260 usual care patients with type 2 diabetes in the Swedish Obese Subjects study found a similar association on macrovascular event rates (HR, 0.68 [95% CI, 0.54-0.85]; P = .001).¹⁴ These findings are also consistent with another recent single-center study of RYGB, which found fewer cardiovascular events in an RYGB cohort (HR, 0.58 [95% CI, 0.42-0.82]; P = .02) compared with matched controls.³⁰

While accumulating data suggest a causal effect of intentional weight loss on lower cardiovascular events and premature mortality, randomized clinical trials to date have not confirmed these findings. The Look AHEAD study randomized 5145 patients with obesity and type 2 diabetes to intensive lifestyle intervention or usual care, but was stopped early because the intervention had no effect on the primary macrovascular disease end point (HR, 0.95 [95% CI, 0.83-1.09]; P = .51).³¹ The ongoing Alliance of Randomized Trials of Medicine vs Metabolic Surgery in T2DM (ARMMS-T2D) is an observational follow-up of 4 randomized trials comparing bariatric surgery vs intensive medical and lifestyle treatment, but, with approximately 300 randomized participants, that study may not be powered to assess macrovascular end points. Randomized clinical trials with sufficient power to assess these rare outcomes are unlikely to be conducted because such studies are challenging to conduct and prohibitively expensive.³²

The study has several limitations. First, the observational design precluded causal inference, and unmeasured confounding may have persisted despite model adjustment for all major cardiovascular risk factors. However, the sensitivity analysis using E-value methodology indicated that the observed 5-year HR of 0.60 for incident macrovascular disease could only be explained by an unmeasured confounder that was associated with both receipt of bariatric surgery and risk of macrovascular disease by a risk ratio of more than 2.72 above and beyond that of the confounders that were measured in this study (upper confidence bound, 1.60). Given that this risk ratio is much greater than any observed for known macrovascular disease risk factors examined in the current study, such as hypertension, diabetes, or hyperlipidemia, it is implausible that an unmeasured confounder exists that can overcome the effect of bariatric surgery observed in the current analysis study (Table 4).

Second, baseline health characteristics and outcomes were established using data collected during routine medical care and billing, which meant that some information was missing and some comorbid conditions could be misclassified (eg, ICD-9 diagnosis codes could be misapplied); however, major cardiac and cerebrovascular outcomes were more likely to be accurately captured claims for all diagnoses and procedures associated with emergency department and hospital admissions.

Third, cause-specific mortality was not examined because cause of death data were not extracted a priori.

Fourth, loss to follow-up could bias the result if patients who underwent bariatric surgery and left the integrated health care systems in this study had very different macrovascular outcomes than the nonsurgical patients who left these systems.

Fifth, the sample size was insufficient to compare the effectiveness of alternative bariatric procedures for these outcomes. There has been a shift toward increased use of the SG procedure in recent years in the United States, and although this study included 17% SG, it is unclear whether the benefits observed in a primarily RYGB population will be seen with SG.

Sixth, given sample size and statistical constraints related to the number of variables that could be accommodated in the matching process, we could not match on every available characteristic. This left some imbalances in other variables (Table 1; standardized differences >10%) that were not part of our matching algorithm. To further address confounding, all variables in Table 1 were adjusted for in our multivariable Cox models.

In this observational study of patients with type 2 diabetes and severe obesity who underwent surgery, compared with those who did not undergo surgery, bariatric surgery was associated with a lower risk of macrovascular outcomes. The findings require confirmation in randomized clinical trials. Health care professionals should engage patients with severe obesity and type 2 diabetes in a shared decision making conversation about the potential role of bariatric surgery in the prevention of macrovascular events.

Corresponding Author: David Arterburn, MD, MPH, Kaiser Permanente Washington Health Research Institute, 1730 Minor Ave, Ste 1600, Seattle, WA 98101 (david.e.arterburn@kp.org).

Accepted for Publication: September 11, 2018.

Correction: This article was corrected on December 11, 2018, to fix an inccorect grant number in the acknowledgment section.

Author Contributions: Dr Arterburn and Mr Johnson 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. Drs Fisher, Sidney, and Arterburn take full responsibility for the contents of the manuscript.

Concept and design: Fisher, Johnson, Haneuse, Arterburn, Coleman, O’Connor, O’Brien, Sidney.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Fisher, Johnson, Haneuse, Arterburn, Sidney.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Johnson, Haneuse, Bogart.

Obtained funding: Haneuse, Arterburn, Coleman, Sidney.

Administrative, technical, or material support: Arterburn, Theis, Anau, Sidney.

Supervision: Arterburn, Sidney.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Mr Johnson and Dr O’Connor reported receiving grants from the National Institutes of Health (NIH). Dr Arterburn reported receiving grants from the National Institutes of Health and the Patient-Centered Outcomes Research Institute. Drs O’Brien and Sidney, Mr Bogart, and Ms Anau reported receiving grants from the National Institute of Diabetes and Digestive and Kidney Diseases. No other disclosures were reported.

Funding/Support: This study was funded by a National Institutes of Diabetes and Digestive and Kidney Diseases award 5R01DK092317-04. Dr Schroeder was supported by NIH award 1K23DK099237-01, and Dr O’Connor was supported in part by NIH award 5P30DK092924-08.

Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: We acknowledge the following individuals for their contributions in creating the study database: Nancy Sherwood, PhD, HealthPartners; Mike Sorel, MPH, Division of Research, Kaiser Permanente Northern California; Terese Defor, Research Informatics, HealthPartners; and Jessica Liu, Department of Research & Evaluation, Kaiser Permanente Research. They all received compensation.