Comparative effectiveness of gastric bypass and sleeve gastrectomy on predicted 10-year risk of cardiovascular disease 5 years after surgery

Abstract

Background:

Comparative evidence is needed when deciding on which bariatric operation to have for long-term cardiovascular risk reduction.

Objective:

The Effectiveness of Gastric Bypass vs. Gastric Sleeve for Cardiovascular Disease (ENGAGE CVD) study compared the effectiveness of vertical sleeve gastrectomy (VSG) and Roux-en-Y gastric bypass (RYGB) operations for reduction of the American College of Cardiology (ACA) and the American Heart Association (AHA) predicted 10-year atherosclerotic cardiovascular disease (ASCVD) risk 5 years after surgery.

Setting:

Data for this study came from a large integrated healthcare system in the Southern California region of the U.S. This is one of the most ethnically diverse (64% non-White) bariatric populations in the literature.

Methods:

The ENGAGE CVD cohort consisted of 22,095 patients who underwent VSG or RYGB from 2009 – 2016. The VSG and RYGB were compared using a local instrumental variable (LIV) approach to address observed and unobserved confounding, as well as to conduct heterogeneity of treatment effects for patients of different age groups, baseline predicted 10-year CVD risk using the ASCVD risk score, and those who had T2DM at the time of surgery.

Results:

Patients (2,771 RYGB and 6,256 VVSG) were primarily women (80.6%), Hispanic or non-Hispanic Black (63.7%), were 46±10 years old, with a BMI of 43.40±6.5 kg/m². The predicted 10-year ASCVD risk at surgery was 4.1% for VSG and 5.1% for RYGB, decreasing to 2.6% for VSG and 2.8% for RYGB 1-year postoperatively. By 5 years after surgery, patients remained with relatively low risk levels (3.0% for VSG and 3.3% for RYGB) and there were no significant differences in predicted 10-year ASCVD risk between VSG and RYGB at any time.

Conclusions:

Predicted 10-year ASCVD risk was low in this population and remained low up to 5 years for those with diabetes, Black and Hispanic patients, and older adults. Literature reporting significant differences between VSG and RYGB in 10-year ASCVD risk may be a result of residual confounding.

Introduction

Compared to conventional weight loss strategies, bariatric surgery results in significantly greater weight loss and improvement of cardiovascular disease (CVD) risk factors such as type 2 diabetes mellitus (T2DM) up to 5 years.^1–7 There are fewer studies on the impact of bariatric surgery on predicted 10-year CVD risk, all of which used the Framingham Risk Score.^8–11 This risk score has several limitations including: 1) its development with primarily non-Hispanic White patients, 2) not having contemporary population-based data, and 3) not including stroke as a “hard” CVD outcome.¹² To address these limitations, 10-year atherosclerotic CVD (ASCVD) risk equations were developed by the American College of Cardiology (ACA) and the American Heart Association (AHA), referred to as the Pooled Cohort Equations Risk Calculator or ASCVD risk score and includes age, systolic blood pressure, total cholesterol, high density lipoprotein, treated hypertension, T2DM, and smoking status.¹³ This risk calculator is based upon more contemporary and diverse patient populations and includes all the major “hard” CVD endpoints including stroke.

To our knowledge, only two studies have looked at predicted 10-year CVD risk using the ASCVD risk score in bariatric patients.^14,15 Both studies included vertical sleeve gastrectomy (VSG) and Roux-en-Y gastric bypass (RYGB), the two most common bariatric operations in the U.S.¹⁶ and were conducted between 2003 and 2016. These studies concluded that after 1 year, both operations resulted in significant reductions in predicted 10-year CVD risk using the ASCVD risk score.^14,15 Although there were differences reported between VSG and RYGB, they were difficult to interpret because there was no adjustment for non-random assignment to bariatric operation and sample sizes for VSG were small (< 250 patients). Rigorous comparative effectiveness studies between RYGB and VSG at the population-level, with long-term follow-up, are needed to assist physicians and patients with severe obesity in choosing the right bariatric operation for lowering their predicted 10-year risk of CVD.

The Effectiveness of Gastric Bypass vs. Gastric Sleeve for Cardiovascular Disease (ENGAGE CVD) cohort study was designed to address this gap in the literature and practice by examining the a priori comparative effectiveness of VSG and RYGB for predicted 10-year CVD risk using the ASCVD risk score over 5 years of follow-up. We also examined, a priori, heterogeneity of treatment effects (HTE) for patients of different age groups, baseline predicted 10-year CVD risk using the ASCVD risk score, and those who had T2DM at the time of surgery. These factors were chosen based upon previous work,^1–6,8 risk stratification recommendations for treatment,¹³ and discussions with our stakeholders about which groups of patients were of interest to the healthcare system.¹⁷

The ENGAGE CVD cohort study has several unique features when compared to other comparative effectiveness studies of bariatric outcomes: 1) this is one of the largest (n = 22,095; 60% VSG), most ethnically diverse (64% non-White) bariatric populations in the literature and includes patients from 9 surgical practices with 23 surgeons in a real-world healthcare setting; 2) it is one of the only studies that used an extensive process of bariatric surgeon stakeholder engagement to determine what factors determine the choice between bariatric operations to include in our statistical models,¹⁷ a process critical for addressing the confounding of non-random assignment; and 3) it is one of the only studies in the literature that used an instrumental variable approach to control for both measured and unmeasured confounding of non-random assignment.

Methods

Participants

Data for this study came from a large integrated healthcare system currently serving 4.7 million members living in the Southern California region of the U.S. Most surgical practices contributing data were accredited as Centers of Excellence (COE) through the Metabolic and Bariatric Surgery Accreditation and Quality Improvement (MBSAQIP) program.¹⁸ The study was reviewed and approved by the Institutional Review Board of the target healthcare system.

The details of how patients are prepared in this setting for weight loss surgery¹⁹ and the cohort of patients used for the current study have been published elsewhere.¹⁷ In general, eligibility for weight loss surgery is based upon national recommendations.²⁰ Figure 1 shows the process of selection of the cohort for the study and the rationale for the exclusions are discussed in the Technical Appendix. To be included in the analyses, all patients had to be free of a history of CVD as required by the ASCVD risk score. We excluded 667 patients from the analyses for the following evidence of CVD: any history at any time of a cardiovascular event as defined by ICD-9 diagnosis codes (n = 584), pharmaceutical treatment for CVD within 24 months of surgery (n = 63), and any history of a lower limb amputation at any time identified using ICD-9 diagnosis and procedure codes (n = 20).

Figure 1.

Inclusion and exclusion criteria for the Effectiveness of Gastric Bypass vs. Gastric Sleeve for Cardiovascular Disease (ENGAGE CVD) cohort study.

Data

All data for the study were obtained from patient electronic medical records and electronic billing claims for outside services. The following general categories of data were used: 1) Dates and status of enrollment, and types of insurance coverage; 2) self-reported date of birth, gender, race/ethnicity, and zip code of primary residence; 3) measured height, weight, blood pressure, lipids, and self-reported smoking status; 4) comprehensive prescription data; 5) dates and types of healthcare utilization for inpatient, emergency department, and outpatient settings; 6) previous medical history including diagnoses and procedures; and 7) bariatric operation location and surgeon. Medical records and billing claims were examined for the whole period before a patient’s bariatric operation and after each patient’s bariatric operation until 12/31/2018.

Outcome

The predicted 10-year ASCVD risk score was calculated at the time of surgery and then annually for up to 5 years following each bariatric operation and included: gender (male/female), age, race (White/Black/Other), smoking status (ever/never), pharmaceutical treatment for hypertension, T2DM (defined by pharmaceutical treatment), total cholesterol, HDL-cholesterol, and systolic blood pressure (SBP). Smoking status was self-reported. The SBP, HDL- and total cholesterol values at the time of surgery were obtained using a 5-year look back period (the interval at which lipids were measured in the healthcare system). If a patient was missing variables for the risk equation in a subsequent year of follow-up, values from the previous year were carried forward, unless there were no values in any of the follow-up years in which case they were set to missing.

Predicted 10-year ASCVD risk scores were calculated for all patients in each year of follow-up regardless of whether they experienced a cardiovascular event. While most studies in this area remove patients from the calculation of risk after they experience one of these events,²¹ we chose to retain these patients. We present the data for scores with and without excluding these patients in the Technical Appendix Table A4 and Figure A9.

The adjusted average mean difference between RYGB and VSG operations in predicted 10-year ASCVD risk scores for each year of follow-up was used for the statistical analyses. Average unadjusted rates for 10-year ASCVD risk scores (not the difference) are shown in Figure 2a for each operation for comparison to other studies in the literature.

Figure 2.

American College of Cardiology and the American Heart Association Pooled Cohort Equations Risk Calculator for 10-year Atherosclerotic Cardiovascular Disease (predicted 10-Year ASCVD risk)¹³ for patients having vertical sleeve gastrectomy (VSG) and Roux-en-Y gastric bypass (RYGB) in the Effectiveness of Gastric Bypass vs. Gastric Sleeve for Cardiovascular Disease (ENGAGE CVD) cohort study. Data are presented at baseline and in each year of follow-up for a) unadjusted rates (to compare to the literature) and b) the adjusted mean absolute difference between VSG and RYGB in 10-year ASCVD risk scores (used for statistical analyses).

Cardiovascular event rates were calculated using methods previously published^4,22,23 in each year of follow-up separately for VSG and RYGB. Details for these calculations are presented in the Technical Appendix. Data are presented as descriptive statistics only due to the low event rates. Events were coronary artery disease-related (CAD; acute myocardial infarction, unstable angina, percutaneous coronary intervention, or coronary artery bypass grafting) and cerebrovascular disease-related (ischemic stroke, hemorrhagic stroke). The international statistical classification of diseases 9^th and 10^th revisions (ICD-9; ICD-10), and current procedural terminology (CPT) codes used for these events are available upon request.

Confounders and Covariates

Table 1 provides a list of all the confounders and covariates we used in our analyses. An advisory board of bariatric surgeons in addition to a bariatric patient and bariatric medicine specialist who served as co-investigators for the study assisted the study team in choosing the key determinants of why patients would choose/undergo VSG or RYGB, which were then used for covariate adjustment.¹⁷ In addition, we adjusted for other variables that have been shown to be related to bariatric surgery outcomes.^1–6,8

Table 1.

Differences between vertical sleeve gastrectomy (VSG) and Roux-en-Y gastric bypass (RYGB) in confounders and covariates before surgery used for the Effectiveness of Gastric Bypass vs. Gastric Sleeve for Cardiovascular Disease (ENGAGE CVD) cohort study before (unadjusted) and after Local Instrumental Variable (LIV) adjustment (across IV median).

	Unadjusted					Adjusted
	VSG		RYGB			< IV Median	≥ IV Median
	(n = 6,256)		(n = 2,771)			(n = 4,514)	(n = 4,513)
	mean	SD	mean	SD	p	mean	SD	p
Age (years)	46.1	9.7	46.95	9.5	<.001	46.2	46.6	.10
Women	5,069	81.0%	2,205	79.6%	.11	80.6%	80.5%	.87
Non-Hispanic White	2,022	32.3%	1,002	36.2%	<.001	32.5%	34.4%	.09
Non-Hispanic Black	1,374	22.0%	431	15.6%	<.001	21.6%	20.2%	.13
Hispanic	2,690	43.0%	1,253	45.2%	.05	42.9%	44.4%	.24
Ever Smoker	2,034	32.5%	964	34.8%	.03	33.9%	32.5%	.22
BMI (kg/m2)	43.30	6.40	43.70	6.62	.009	43.10	43.30	.07
BMI ≥ 50 kg/m2	877	14.0%	434	15.7%	.04	14.2%	14.9%	.40
Weight Change (lbs) in 12 Mo Before Surgery	−17.2	13.8	−17.4	14.3	0.888	−17.4	−17.1	.25
Predicted 10-Year ASCVD Risk	4.09%	5.80%	5.10%	6.53%	<.001	4.40	4.40	.95
HDL <	4429	70.8%	2,226	80.3%	<.001	73.8%	73.7%	.93
LDL >	2152	34.4%	1,416	51.1%	<.001	39.2%	39.9%	.58
Triglycerides >	2538	40.6%	1,592	57.5%	<.001	44.9%	46.6%	.18
Total Cholesterol >	2,159	34.5%	1,418	51.2%	<.001	39.7%	39.5%	.88
Dyslipidemia	4,792	76.6%	2,330	84.1%	<.001	79.4%	78.4%	.31
Gastro-esophageal Reflux Disease (GERD)	2,128	34.0%	1,158	41.8%	<.001	34.6%	38.2%	<.001
Esophagitis	100	1.6%	59	2.1%	.08	1.9%	1.6%	.38
Duodenal Ulcer	371	5.9%	187	6.7%	.36	5.8%	6.5%	.23
Peptic Ulcer	93	1.5%	43	1.6%	.795	1.3%	1.7%	.16
Gastritis Duodenitis	738	11.8%	368	13.3%	.05	12.4%	12.1%	.78
Dyspepsia	779	12.5%	376	13.6%	.14	12.5%	13.1%	.48
Hiatal Hernia	167	2.7%	116	4.2%	<.001	3.0%	3.3%	.44
Severe Mental Illness	318	5.1%	171	6.2%	.035	5.5%	5.3%	.76
Severe Depression and/or Anxiety	371	5.9%	195	7.0%	.05	6.5%	6.0%	.37
Mild-to-Moderate Anxiety/Depression	2,666	42.6%	1,186	42.8%	.87	42.0%	43.3%	.30
Sleep Apnea	1019	16.3%	501	18.1%	.04	17.7%	16.0%	.06
Type 2 Diabetes Mellitus (T2DM)	1869	29.9%	1,553	56.0%	<.001	37.3%	38.5%	.32
Hypertension	3239	51.8%	1,687	60.9%	<.001	54.2%	54.9%	.58
Chronic Kidney Disease	610	9.8%	346	12.5%	<.001	10.5%	10.7%	.81
Aspirin Use 36 Mo Before Surgery	864	13.8%	709	25.6%	<.001	17.8%	17.0%	.41
Aspirin Use 3 Mo Before Surgery	550	8.8%	466	16.8%	<.001	12.4%	10.1%	<.001
NSAID Use 36 Mo Before Surgery	2872	45.9%	1,271	45.9%	.97	45.5%	46.3%	.53
NSAID Use 3 Mo Before Surgery	951	15.2%	407	14.7%	.53	15.3%	14.8%	.51
Scheduled Visit Attendance Rate in 12 Mo Before Surgery	.77	0.12	.77	0.12	.865	0.77	0.77	.32
Hospitalization 12 Mo Before Surgery	294	4.7%	135	4.9%	.72	4.7%	4.8%	.72
Emergency Department Use 12 Mo Before Surgery	1,225	19.6%	578	20.9%	.16	20.2%	19.7%	.60

^*After controlling for year of surgery and 3-digit zip code indicators, zip code-level surgery volume, and surgeon-specific caseload in previous year.

Analyses

We chose a local instrumental variable (LIV) approach to address observed and unobserved confounding, as well as to conduct the HTE.^24–29 Further details about the conceptual foundations of LIV methods and their implementation is provided in the Technical Appendix. The LIV approach was carried out in two steps. In the first step, the choice of RYGB versus VSG was modelled as a function of the RYGBrate, after controlling for all baseline risk-factors (please see Table 1), 3-digit zip-code, year of surgery, and the denominator of the RYGBrate using a probit regression model. The RYGBrate was found to be significantly predictive of operation choice with an F-statistic of 200. A propensity for RYGB selection was estimated from this first-step regression.

In the second step, the predicted 10-year predicted ASCVD risk score was converted to a proportion estimate and modeled with quasi-likelihood-based generalized estimating equations using a logit link function and exchangeable correlation structure.^30,31 In addition to the risk factors from step 1 (see Table 1), clinical factors could interact with the propensity of RYGB selection from step 1 and were tested for an appropriate polynomial of the propensity score itself. The partial derivative of the predicted 10-year ASCVD risk with respect to the propensity score was used as an estimator of the marginal treatment effects.²⁶ These effects were then aggregated to form the population average and subgroup-specific average treatment effects for the HTE analyses.²⁷

HTE analyses.

All standard errors were calculated using non-parametric bootstrapping and allowed for clustering of individual outcomes over time. Sub-groups tested for the HTE analyses were categories of age (30 – 39, 40 – 47, 48 – 54, and 55 –74 years; selected to have equal sample sizes), baseline predicted 10-year ASCVD risk (0 – 4.9%, 5.0% – 7.4%, 7.5% – 14.9%, ≥ 15%; selected based upon recommendations from the American College of Cardiology),³² and presence of T2DM at baseline (yes, no). As secondary sensitivity analyses, we conducted non-IV based inverse-probability weighted propensity score regression. The details of these methods are provided in the Technical Appendix. Data analyses were performed in Stata (StataCorp), version 15.1.

Results

Patients

In general, patients were primarily women (80.6%), Hispanic or non-Hispanic Black (63.7%), were 46 ± 10 years old, with BMI at the time of surgery of 43.40 ± 6.5 kg/m², and low predicted 10-year ASCVD risk at the time of surgery (4.6% ± 6.2%). One of the critical steps in the LIV approach was to use the patient sample from the central 80% of the rate of RYGB operations (RYGBrate; our IV) where differences between VSG and RYGB patients were balanced. Table 1 provides the results of the reduction in mean differences for covariates and confounders between VSG (n = 6,256) and RYGB (n = 2,771). All standardized mean differences were < 0.1 across the IV median, indicating satisfactory balance in confounder levels.³³

However, this resulted in a loss of 2,305 patients, potentially biasing our results. Differences between patients who were treated by surgeons within the central 80% of RYGBrate (n = 9,096) and patients treated by surgeons in the peripheral 20^th percentile of RYGBrate (n = 2,305) are shown in the Technical Appendix Table A2. Patients treated by surgeons in the peripheral 20^th percentile had higher weight loss before surgery (p<.001), were more likely to be non-Hispanic White (p=.01), more likely to have GERD or dyslipidemia (p<.001), and less likely to have sleep apnea (p<.001).

Follow-up Rates

Five-year retention rates did not vary by operation and were 66.5% for RYGB and 63.6% for VSG. Across both operations, patients were lost to follow-up because they were no longer members of the healthcare system (n = 1,375; 73%), had no information in their electronic medical record at 5 years (n = 456; 24%), they died (n = 3; 0%), or they did not have any of the indicators necessary to calculate the ASCVD risk score throughout all years of the follow-up period (n = 58; 3%). A more detailed breakdown of how retention rates were calculated and what sample was available for analyses at each year of follow-up are presented in the Technical Appendix Figure A7.

Outcome and HTE

After LIV adjustment, there were no statistically significant differences between RYGB and VSG in predicted 10-year ASCVD risk scores at any time during the follow-up. The average adjusted difference in 10-year ASCVD risk scores between RYGB and VSG at year 1 was −1.62% (95% CI: −4.35,1.39; p=.25) (Figure 2b) and decreased to --.77% (95% CI: −3.75,1.52; p=.61) by 5 years. Inverse-probability weighted propensity score regression findings were similar and are shown in the Technical Appendix Figure A8. The HTE results for the average adjusted difference in 10-year ASCVD risk scores between RYGB and VSG are presented in Table 2 and Figure 3a–3d. There were no differences by operation for any subgroup at any year of follow-up.

Table 2.

Average difference (95% Confidence Intervals) in adjusted mean predicted 10-Year Atherosclerotic Cardiovascular Disease (ASCVD) risk¹³ between vertical sleeve gastrectomy (VSG) and Roux-en-Y Gastric Bypass (RYGB) in the Effectiveness of Gastric Bypass vs. Gastric Sleeve for Cardiovascular Disease (ENGAGE CVD) cohort study. Data are presented to accompany heterogeneity of treatment effects shown in Figure 3.

Effect	Baseline	Year 1	Year 2	Year 3	Year 4	Year 5
Age (years)
30 – 39	0	−0.1 (−1.6,1.5)	0 (−1.6,1.5)	0 (−1.6,1.6)	0.2 (−1.4,1.7)	0.3 (−1.3,1.8)
40 – 47	0	−0.4 (−2.2,1.4)	−0.3 (−2.1,1.5)	−0.3 (−2.1,1.5)	−0.1 (−1.9,1.7)	0 (−1.8,1.8)
48 – 54	0	−0.6 (−2.9,1.8)	−0.5 (−2.9,1.8)	−0.5 (−2.9,1.8)	−0.3 (−2.7,2.1)	−0.1 (−2.5,3.3)
55 – 74	0	−0.4 (−3.8,3.1)	−0.2 (−3.7,3.2)	−0.2(−3.6,3.2)	0.2 (−3.3,3.6)	0.4 (−3.0,3.9)
Type 2 Diabetes Mellitus (T2DM)
% yes	0	−0.5 (−2.8,1.8)	−0.4 (−2.7,1.9)	−0.4 (−2.7,1.9)	−0.1 (−2.5,2.2)	0.1 (−2.3,2.4)
% no	0	−0.2 (−2.4,1.9)	−0.2 (−2.4,2.0)	−0.2 (−2.4,2.1)	0.1 (−2.2,2.3)	0.2 (−2.0,2.4)
ASCVD Risk
0 – 4.9%	0	−0.2 (−2.0,1.7)	−0.1 (−1.9,1.7)	−0.1 (−1.9,1.7)	0.1 (−1.7,1.9)	0.2 (−1.6,2.1)
5.0 – 7.4%	0	−0.4 (−3.2,2.5)	−0.3 (−3.1,2.6)	−0.3 (−3.1,2.6)	0.1 (−2.8,3.0)	0.3 (−2.6,3.2)
7.5 – 14.9%	0	−0.6 (−4.0,2.8)	−0.5 (−3.9,3.0)	−0.5 (−3.9,2.9)	−0.1 (−3.5,3.4)	0.2 (−3.2,3.6)
≥ 15%	0	−2.0 (−7.0,3.0)	−1.8 (−6.8,3.1)	−1.9 (−6.9,3.1)	−1.4 (−6.4,3.7)	−1.0 (−6.0,4.0)

Figure 3.

Heterogeneity of treatment effects (HTE) for the American College of Cardiology and the American Heart Association Pooled Cohort Equations Risk Calculator for 10-year Atherosclerotic Cardiovascular Disease (predicted 10-Year ASCVD risk)¹³ for patients having sleeve gastrectomy (VSG) and Roux-en-Y Gastric Bypass (RYGB) in the Effectiveness of Gastric Bypass vs. Gastric Sleeve for Cardiovascular Disease (ENGAGE CVD) cohort study. Data are presented at baseline and in each year of follow-up for the following subgroups at the time of surgery: a) age categories (30 – 39, 40 – 47, 48 – 54, and 55 – 74 years; selected to have equal sample sizes), b) having type 2 diabetes mellitus (T2DM) (yes/no), c) predicted 10-year ASCVD risk category (0 – 4.9%, 5.0% – 7.4%, 7.5% – 14.9%, ≥ 15%; selected based upon recommendations from the American College of Cardiology),³² and d) the interaction of age and predicted ACSVD 10-year CVD risk categories. Point estimates and 95% confidence intervals (CI) are shown in Table 2.

Table 3 presents the unadjusted mean values for the individual components used to calculate the predicted 10-year ASCVD risk score. For both operations, the prevalence of self-reported smoking began to increase immediately following surgery (p<.001). However, all other indicators used for the predicted 10-year ASCVD risk score significantly improved over time (p<.001). Although SBP (p<.001) and total cholesterol (p=.01) were significantly lower than baseline at 5 years, their change was not as pronounced as that found for the other health indicators. Table 3 also presents the CVD event rates by operation. In general, total CVD event rates were low for both VSG (1.9%) and RYGB (3.6%).

Table 3.

Unadjusted components of the American College of Cardiology and the American Heart Association Pooled Cohort Equations Risk Calculator for 10-year Atherosclerotic Cardiovascular Disease (ASCVD) risk over time¹³ in 9,027 bariatric surgery patients. Change at year–1 and year-5 after surgery is relative to values at the time of surgery.

	Baseline	Year 1	Year 2	Year 3	Year 4	Year 5	Change at Year 1	Change at Year 5
Vertical Sleeve Gastrectomy (VSG)	n = 6,256	n = 6,079	n = 5,167	n = 3,994	n = 3,130	n = 2,487
Predicted 10-Year ASCVD Risk	4.1%	2.6%	2.7%	2.8%	2.9%	3.0%	−36.4%ǂ	−26.4%ǂ
Age (years)	46.1	47.1	48.1	49.1	50.1	51.1	2.2%ǂ	10.9%ǂ
Systolic BP (mmHg)	130.0	120.3	122.2	123.3	123.9	124.4	−7.5%ǂ	−4.3%ǂ
Total Cholesterol (mg/dL)	188.3	187.0	186.9	187.1	187.4	187.4	−0.7%ǂ	−0.5%ǂ
HDL (mg/dL)	48.2	48.8	51.0	52.7	53.7	54.3	1.2%	12.7%ǂ
Treated Hypertension	43.8%	22.4%	23.0%	22.2%	19.6%	18.3%	−48.9%ǂ	−58.2%ǂ
Type 2 Diabetes Mellitus (T2DM)	23.8%	4.5%	4.6%	4.6%	4.4%	3.9%	−81.1%ǂ	−83.6%ǂ
Former Smoker	31.5%	30.8%	30.2%	30.0%	29.9%	30.8%	−2.2%ǂ	−2.2%ǂ
Current Smoker	0.80%	1.7%	2.6%	3.0%	3.2%	3.3%	112.5%ǂ	312.5%ǂ
Cardiovascular Events: n (%)	-	17	23	34	25	22	121 (1.9%)
Cerebrovascular Disease	-	13	15	13	8	8	57 (0.9%)
Coronary Artery Disease (CAD)	-	4	8	21	17	14	64 (1.0%)
Roux-en-Y Gastric Bypass (RYGB)	n = 2,771	n = 2,681	n = 2,299	n = 1,765	n = 1,420	n = 1,111
Predicted 10-Year ASCVD Risk	5.1%	2.8%	2.9%	3.1%	3.1%	3.3%	−45.1%ǂ	−35.3%ǂ
Age (years)	46.9	47.9	48.9	49.9	50.9	51.9	2.1%ǂ	10.7%ǂ
Systolic BP (mmHg)	131.1	119.6	121.6	122.6	123.5	124.2	−8.8%ǂ	−5.3%ǂ
Total Cholesterol (mg/dL)	183.6	178.2	175.6	174.8	174.6	173.6	−2.9%ǂ	−5.5%ǂ
HDL (mg/dL)	46.6	46.5	48.8	50.8	51.9	52.6	−0.2%	12.9%ǂ
Treated Hypertension	56.4%	26.1%	24.4%	23.5%	20.9%	18.3%	−53.7%ǂ	−67.6%ǂ
Type 2 Diabetes Mellitus (T2DM)	50.2%	10.2%	10.0%	10.4%	9.6%	8.1%	−79.7%ǂ	−83.9%ǂ
Former Self-Reported Smoker	33.7%	33.0%	32.1%	31.8%	31.1%	30.8%	−2.1%ǂ	−8.6%ǂ
Current Self-Reported Smoker	1.0%	1.7%	2.7%	3.0%	3.7%	4.1%	70.0%ǂ	310.0%ǂ
Cardiovascular Events: n (%)	-	9	20	25	20	27	101 (3.6%)
Cerebrovascular Disease	-	3	10	8	7	7	35 (1.3%)
Coronary Artery Disease (CAD)	-	6	10	17	13	20	66 (2.4%)

^ǂp ≤ .01

Discussion

This was the largest (n = 9,027), most racially/ethnically diverse (64% Hispanic or non-Hispanic Black), and one of the first studies to examine the comparative predicted 10-year CVD risk using the ASCVD risk score for VSG and RYGB up to 5 years after bariatric surgery. As with other studies that published ASCVD risk score results for RYGB and VSG operations,^14,15 we found that both RYGB and VSG patients had a marked decrease in predicted 10-year ASCVD risk scores 1 year after surgery which remained ow up to 5 years after surgery. However, unlike these studies,^14,15 we did not find any differences between operations when we used an LIV approach to adjust for confounding (see Figure 2b). If we had not adjusted for confounding, we too would have concluded that after 1-year RYGB had a greater reduction in predicted 10-year ASCVD risk score than VSG (unadjusted rates for VSG 4.09% to 2.80% and RYGB 5.1% to 3.0%; see Figure 2a). We did not find a difference in predicted 10-year ASCVD risk scores between bariatric operations by age, baseline CVD risk level, or T2DM.

It is difficult to compare the rates of CVD events to other studies in the literature. Most of the papers published on CVD risk for bariatric patients have only examined 1 year of follow-up. One paper from the Longitudinal Assessment of Bariatric Surgery (LABS) study published CVD event rates after 7 years of follow-up in RYGB and found a total of 250 events for 1,770 patients after 5 years using similar classification methods we had in our study; 157 of which were angina-related.²¹ We reported a lower 5-year rate of 101 events for 2,771 RYGB patients. In our own previous research in patients with T2DM, with a larger sample that included other health systems, we found an overall event rate of 2.1%,⁴ which is similar to the current study (2.4%).

It would be hard to determine why our rates are different than the LABS study. Part of this could be due to the way data was collected. For example, the LABS study outcomes were collected prospectively based upon patient self-reports which could have resulted in over-reported rates of angina. Our data was abstracted from the medical record and used methods from previous studies to ensure that the diagnosis codes we used were as accurate as possible.^4,22,23 In addition, the practices of the health system in which the patients received their care could also be responsible for low event rates. The health system in which our study took place is very proactive about management of CVD risk factors, including hypertension and dyslipidemia which could lead to prevention of CVD events.^34,35

Any increase in CVD risk over time was primarily due to smoking (see Table 3). Although self-reported smoking for both operations was only 3–4% 5 years after surgery, this represented an increase of over 300%. Smoking is a strong contra-indication to bariatric surgery and in the targeted health system, patients are screened using urine cotinine tests before surgery. Any patient in the targeted health system who tests positive for nicotine cannot undergo any bariatric operation. Consequently, the increase we saw from baseline could be due to inaccurate self-report at surgical consult (which would be artificially low because smoking is a contraindication to surgery). There is evidence from a small study that used an objective measure of nicotine levels that patients may significantly underreport self-reported smoking on the day of surgery (6% via self-report versus 18% via urine cotinine measures).³⁶

However, there is also evidence that this large increase in self-reported smoking may be real in that bariatric patients may begin smoking after surgery to prevent weight regain.³⁷ In addition, there have been reports of addiction transference in bariatric patients who may substitute smoking, alcohol, and other drugs, for food after surgery.^38–40 Smoking is one of the strongest modifiable predictors of CVD.⁴¹ This underscores the importance of post-operative monitoring and support for bariatric patients who might experience a number of unintended negative consequences following surgery including substance use disorder,³⁸ suicide,³⁹ and disordered eating.⁴⁰

There were several limitations with the current study, most notably that it was a retrospective observational design with non-random assignment to operation. To mitigate confounding we worked closely with our advisory board of bariatric surgeons and our patient and provider co-investigators to identify the key determinants of why patients would choose/undergo VSG or RYGB.¹⁷ There were several factors that were not available from electronic sources and thus could not be measured.¹⁷ The LIV approach was specifically designed to account for these unobserved confounders instead of more traditional non-IV based inverse-probability weighted propensity score regression which is typically used in bariatric comparative effectiveness research and only includes adjustment for factors that can be observed. Although rigorous, the LIV approach also resulted in an additional loss of several thousand patients who were different than those upon which the findings were based (see Technical Appendix Table A2), and thus could limit the generalizability of the LIV results. To address this concern, we conducted traditional propensity score analysis, which contained these missing patients, and the results were very similar to the LIV findings (see Technical Appendix Figure A8).

Another limitation of the study was that follow-up rates at 5 years were only 66%. These rates are comparable to one of the largest, longest studies of bariatric patients, the Longitudinal Assessment of Bariatric Surgery (LABS) study that had 63% follow-up rate at 7 years for women and 55% for men.²¹ Loss to follow-up in our study was primarily because we had to rely on the electronic medical record as our only data source; the majority of those patients lost to follow-up had lost their membership in the health plan at 5 years after surgery and could not contribute any data as a result.

Finally, our patients had very low predicted 10-year ASCVD risk scores before surgery and very low rates of CVD events after surgery. Part of the reason for this was that patients with a history of CVD were excluded from the ASCVD calculations as required by the score. Our findings might have been different if we had patients with high risk before surgery.¹⁴ This would have been difficult to do with a retrospective observational design as the targeted healthcare system follows U.S. national recommendations⁴² for the optimization of patients for surgery, including smoking cessation, some weight loss, and the control of CVD risk factors such as hypertension and T2DM.

Conclusions

Despite these limitations, our study is one of the largest, most racially and ethnically diverse and methodologically rigorous to date on the comparative effectiveness of VSG and RYGB for predicted 10-year CVD risk reduction over 5 years of follow-up using the ASCVD risk score. Future research should focus on the comparative effectiveness of VSG and RYGB for hard CVD outcomes such as stroke and myocardial infarction and look at the lifetime risk for CVD using even longer post-operative follow-up (≥ 10 years). The reliance on bariatric surgery alone to address major chronic conditions related to predicted 10-year CVD risk, without addressing post-operative behaviors such as smoking which substantially increase risk, may undermine the effect of surgery for long-term protection against CVD. Our findings, and others in this field,^43,44 suggest that healthcare systems should concentrate on post-operative patient management to ensure that patients do not start smoking, maintain a healthy diet, and exercise to retain the reduction in their comorbidity burden.

Highlights.

• The Effectiveness of Gastric Bypass vs. Gastric Sleeve for Cardiovascular Disease (ENGAGE CVD) study compared the effectiveness of vertical sleeve gastrectomy (VSG) and Roux-en-Y gastric bypass (RYGB) operations for reduction of the American College of Cardiology (ACA) and the American Heart Association (AHA) predicted 10-year atherosclerotic cardiovascular disease (ASCVD) risk 5 years after surgery.

• Patients (2,771 RYGB and 6,256 VSG) were primarily women (80.6%), Hispanic or non-Hispanic Black (63.7%), were 46±10 years old, with a BMI of 43.40±6.5 kg/m². The predicted 10-year ASCVD risk at surgery was 4.1% for VSG and 5.1% for RYGB, decreasing to 2.6% for VSG and 2.8% for RYGB 1-year postoperatively. By 5 years after surgery, patients remained with relatively low risk levels (3.0% for VSG and 3.3% for RYGB) and there were no significant differences in predicted 10-year ASCVD risk between VSG and RYGB at any time.

• For both operations, the prevalence of smoking began to increase immediately following surgery (p<.001). However, all other indicators used for the predicted 10-year ASCVD risk score significantly improved over time (p<.001).

• Our findings suggest RYGB and VSG provide similar benefits for 10-year risk of cardiovascular disease. Literature reporting significant differences between VSG and RYGB in 10-year ASCVD risk may be a result of residual confounding.