Bariatric Surgery and Cardiovascular Outcomes in Patients With Obesity and Cardiovascular Disease:: A Population-Based Retrospective Cohort Study

Obesity is associated with an increase incidence of heart failure (HF), myocardial infarction (MI), stroke, and death.^1–3 Weight loss has become a standard recommendation for all patients with cardiovascular disease (CVD) and coexisting obesity; however, randomized evidence to support this recommendation is limited. The largest trial to date investigating whether intentional weight loss reduces cardiovascular morbidity and mortality in overweight individuals with type 2 diabetes did not show a significant reduction in cardiovascular events.⁴ However, the weight loss observed in this trial was modest, and it is not known whether greater levels of weight loss would lead to clear reductions in recurrent cardiovascular events and deaths. Several studies of overweight and obese patients with CVD suggest an “obesity paradox,” whereby elevated body mass index (BMI) may be associated with lower mortality and cardiovascular events.^5,6 However, other studies have described a U-shaped relationship with respect to weight, with severe obesity being associated with increased risk of cardiovascular outcomes.³ Although the relationship between obesity and cardiovascular outcomes is not well delineated, bariatric surgery, in addition to being the most effective weight loss treatment, is also known to resolve other cardiovascular risk factors, including the burden of recurrent CVD in patients with obesity and previous heart disease.⁷

Randomized clinical trials (RCTs) have revealed that bariatric surgery leads to consistent improvements in diabetes outcomes and cardiovascular risk factors in patients with diabetes and severe obesity.^8–10 However, such trials did not evaluate the effect of bariatric surgery on mortality and other major cardiovascular outcomes. Several observational studies have reported on the promising correlation between bariatric surgery and decreased cardiovascular outcomes.^11–13 Early studies of patients with severe obesity and type 2 diabetes undergoing bariatric surgery showed lower risks of macrovascular events in the surgery group, but confounding bias may have contributed to these results, and patients in these studies received procedures that are no longer common.¹² More recently, retrospective studies of patients with diabetes undergoing sleeve gastrectomy and Roux-en-Y gastric bypass have associated bariatric surgery with significantly lower incidence of incident major cardiovascular events after 8 years.^11,13 Nonetheless, the observed effects of bariatric surgery in patients with severe obesity and a history of heart disease remain a major gap in the literature.

This study aimed to examine the outcomes of patients with both CVD and severe obesity undergoing either bariatric surgery or routine clinical care in Ontario, Canada.

Table 1 presents patient demographics, comorbidity status, and health care use characteristics. There were 1700 surgical patients and 3749 eligible nonsurgical patients from January 2010 to December 2016. From these, 1319 surgical patients were matched to nonsurgical controls in a 1:1 ratio (Figure 1). Among these individuals, 548 (21%; 274 in each group) had a history of HF (with or without a history of IHD), and the remainder had a history of IHD. The overall mean age was 56.0 years at baseline (61% female). The mean BMI on the index date was 47.4 kg/m² (±11.3 kg/m²). Roux-en-Y gastric bypass accounted for 80% of procedures. The nonsurgical group was older and included a greater number of smokers, history of MI, substance abuse, hospitalizations for cardiac reasons, and emergency department visits within 1 year. In contrast, surgical patients were more likely to have been previously hospitalized for any reason, come from urban areas, and reside in higher-income neighborhoods. Regarding the sensitivity analysis based on logit size, using a logit of 0.2 or 0.4 did not substantially change the match results but decreased the number of patients by 10%. The rates of revascularization were not different between groups overall, with 25.7% in the controls and 24.3% in the cases, or within the IHD subgroup, because rates of revascularization within this subgroup were 28.0% of controls and 26.5% of cases, respectively.

Table 2 and Figure 2 present the association between bariatric surgery and the primary outcome, as well as within subgroups with previous HF or previous IHD. Overall, 410 individuals experienced a primary outcome at a median of 4.5 years of follow-up. Of these, there were 151 events (11.5%) in the surgery group and 259 (19.6%) in the control group. The unadjusted ARD associated with surgery was 8.2% (95% CI, 5.4%–11.0%), and the adjusted HR was 0.58 (95% CI, 0.48–0.71) in favor of the surgery group. In the subgroup of patients with IHD (n=2403), surgery was associated with a 40% lower hazard of primary events. The subgroup of patients with previous HF experienced a 56% lower observed hazard of the primary outcome, with an associated ARD of 19.3%. In terms of the main secondary outcome, 280 patients experienced the secondary outcome, with 107 patients (8.1%) in the surgery group and 173 (13.1%) in the control group. At a median of 4.7 years of follow-up, bariatric surgery was associated with 34% lower hazard of experiencing the secondary outcome (95% CI, 0.52–0.84). In subgroups of patients with ischemic disease and patients with HF, surgery was associated with a 33% (HR, 0.67 [95% CI, 0.52–0.87]) and 60% (HR, 0.40 [95% CI, 0.25–0.63]) lower hazard of the secondary outcome, respectively. The proportionality assumption was met for all outcomes. Censoring for loss of insurance occurred 21 times in our cohort (0.8%), and there was no difference between the groups.

Table 3 illustrates the association between bariatric surgery and the primary outcome according to sex and procedure type. The ARD for the primary outcome was larger in male patients at 10.9% (95% CI, 6.0%–15.8%) after surgery. After adjustment, the HR for the primary outcome was not different in male (HR, 0.55 [95% CI, 0.41–0.73]) and female patients (HR, 0.60 [95% CI, 0.45–0.80]). In terms of procedures, a significant lower hazard of reaching the primary outcome was observed in patients who received a Roux-en-Y gastric bypass (HR, 0.61 [95% CI, 0.49–0.77]) and sleeve gastrectomy (HR, 0.47 [95% CI, 0.30–0.71]). There was no significant interaction in the sex or procedure strata.

Table 3 also demonstrates the association between bariatric surgery and the primary outcome stratified by age and BMI ranges. The lower ARD of surgery was strongest in the group ≥60 years of age with an ARD of 11.5% (95% CI, 6.0%–17.0%). Specifically, patients who were ≥60 years of age experienced a 41% lower hazard of the primary outcome after surgery. In terms of BMI, the association of bariatric surgery with MACE was not different among those with BMI 40 to 50 kg/m² versus BMI >50 kg/m² with ARD of 8.8% (95% CI, 4.8%–12.6%) and 8.3% (95% CI, 3.6%–12.9%), respectively. Surgery was not associated with a significantly lower incidence of experiencing the primary outcome in those with BMI <40 kg/m². However, the direction of the observed incidence was not different from the BMI ≥40 kg/m², and the number of patients in this group was small (n=185 per group). There was no significant interaction within the age or BMI strata.

Table 4 presents the association between bariatric surgery and cause-specific mortality, as well as specific cardiovascular events at a median of 4.5 years in the entire cohort and the IHD and HF subgroups. Bariatric surgery was associated with 65% lower hazards of cardiovascular mortality (HR, 0.35 [95% CI, 0.15–0.80]) and 44% lower hazard for other mortality (HR, 0.56 [95% CI, 0.36–0.88]). These observed lower mortality was seen in both strata of cardiac disease. In the overall cohort, surgery was associated with lower hazards of MI, coronary artery bypass grafting or percutaneous coronary intervention, and HF hospitalization. In the subgroup of HF patients, lower hazards of MI and HF were seen after surgery.

Surgical safety was assessed in this cohort. The overall 30-day complication rate was 7.7% (95% CI, 6.3%–9.3%), and the overall 30-day mortality rate was 0.15% (95% CI, 0.02%–0.50%). The median length of stay after surgery was 3 days (interquartile range, 2–7 days), and 3.5% (95% CI, 2.5%–4.6%) of patients required an intensive care unit stay. There were 6 MIs (0.45% [95% CI, 0.16%–0.98%]) and 5 reoperations (0.38% [95% CI, 0.12%–0.88%]).

In this population-based matched cohort study of 2638 patients with CVD and severe obesity, bariatric surgery was associated with a significantly lower ARD in the primary outcome of all-cause mortality, coronary artery events, coronary revascularization, cerebrovascular events, or HF. In addition, this observed lower incidence was seen in subgroups of patients with ischemic disease and HF, with the latter group showing a larger associated lowered incidence of MACE. Bariatric surgery was also associated with lower incidences of the individual components of the primary outcome, except for cerebrovascular events. This analysis addresses a major gap in the literature by examining the association between bariatric and major adverse CVD outcomes.

The results of the present study are mostly consistent with previous studies reporting the associations of bariatric surgery with major adverse cardiovascular outcomes. Aminian et al¹³ conducted a cohort study of 2287 patients with diabetes undergoing bariatric surgery matched to 11 435 controls. In this study, bariatric surgery was associated with a 38% lower incidence of 3-component MACE.¹³ This is in agreement with the present study, because similarly lower observed incidences of MACE, coronary events, coronary artery bypass grafting, and HF hospitalization are reported. The present study is also similar to the results of the Swedish nationwide cohort study by Näslund et al,²⁴ which matched 509 severely obese patients with history of previous MI undergoing bariatric surgery to 509 control patients with MI. Their results align with the present study, because the hazard of MACE, mortality, and new-onset HF was significantly lower in the bariatric surgery group at 8 years of follow-up. Similar to Näslund et al,²⁴ the present study did not show a protective effect against stroke, although the incidence for stroke was low (2.1%) because of the 5-year follow-up, the largely female study cohort, and younger age. Further follow-up time is needed to evaluate whether an association between bariatric surgery and stroke risk is present. Other cohort studies including the Swedish Obese Subjects study have also suggested that the weight loss caused by bariatric surgery is associated with a lower risk of HF.^25–27 The present study agrees with the previously elucidated associations and demonstrates a pronounced associated lower ARD for those with heart disease, which had not been shown previously.^8,11–13 Conversely, in a study by Pirlet et al,²⁸ there was significantly lower incidence of MACE in the surgery group in CVD patients at median follow-up of 8.9 years, albeit with no significant difference in all-cause death or cardiovascular death. However, the difference in all-cause mortality was only of borderline statistical significance (HR, 0.58 [95% CI, 0.33–1.01]; P=0.056) and may have been underpowered (n=232) to detect a significant difference. The present study also has an early divergence of the cumulative incidence curves between the groups for the primary outcome. However, shapes of survival curves are difficult to interpret given the modest number of events at any time point in any study. In the present study, the mortality curves appeared to diverge early, whereas in the study by Näslund et al²⁴ of 1018 people and 279 events, the curves diverged after 4 to 5 months. The shapes of the curves are likely less reliable than the overall results, and emphasis should be not be placed in either study.

Our patient population was at high risk of recurrent CVD events and had a risk profile that is not different from what would be expected in nonseverely obese cohorts. At a median follow-up of 4.6 years, incidences of the primary and secondary outcomes in the control group were 19.6% and 13.1%, respectively. Similarly, in the REACH registry (Reduction of Atherothrombosis for Continued Health Registry; N=45 227; 28.4% of whom were obese), an observational cohort of patients with previous cardiovascular ischemic events or CVD risk factors, the rate of cardiovascular death, MI, and stroke among patients with previous CVD was 12.2% (95% CI, 11.4%–12.9%), whereas that of the composite of cardiovascular death, MI, stroke, and cardiovascular hospitalization was 31.1% (95% CI, 29.8%–32.4%) at 4 years.²⁹

The associations demonstrated in this and previous studies may act through multiple mechanisms. Well-controlled cardiovascular risk factors, such as hypertension and diabetes, are certainly potential mechanisms, but weight loss in and of itself may also contribute greatly. Individuals with obesity experience a chronic inflammatory state that results in endothelial dysfunction and may increase the risk of atherosclerosis.^2,30,31 Bariatric surgery has been associated with reductions in C-reactive protein concentrations, a marker of atherothrombotic risk, alongside increases in the anti-inflammatory mediator adiponectin.^32–34 Furthermore, obesity can increase the preload, afterload, and stroke volume for the left ventricle, increasing left ventricular wall stress.³⁵ This factor, alongside increases in hormones including adipokines, such as leptin, which drive ventricular hypertrophy and increase circulating volume within the heart, result in ventricular dilation and compensatory left ventricular hypertrophy.^35,36 There is also a beneficial effect of surgery on cardiac structure and function including both regression of left ventricular hypertrophy and improved diastolic function.^7,37,38 Bariatric surgery reduces the left ventricular stroke workload, decreases systolic blood pressure, and reverses metabolic dysfunction, which may result in cardiac reverse remodeling, including improvements in cardiac structure and function.^38–40 Overall, although the obesity paradox exists and is not solved by this study, weight loss may only be part of the mechanism through which bariatric surgery may affect CVD.

The present study provides strong evidence regarding the association between bariatric surgery and lower MACE among patients with heart disease and obesity. While not an RCT, it suggests a potential benefit from surgery in this high-risk population, although the potential benefit reported in this study is likely overestimated because of the observational design of this study. To this point, the use of surgery has been limited in patients with CVD out of concern for perioperative risk, but previous studies have shown excellent safety, even in those awaiting heart transplant.⁴¹ The cohort of the present study also compared favorably to previously published data for Ontario.⁴² Accordingly, consideration should be given to surgery as a treatment option in patients with CVD and obesity, especially when no further treatment options exist. In addition, in certain groups with very high event rates despite optimal care (eg, those with HF, older populations, and males), the potential for benefit is high, although this benefit requires confirmation with an RCT. It is important to note that even for groups with no significant difference in outcomes in this study, these effects should be properly contextualized. Although the association within BMI <40 kg/m² was not statistically significant, this may be because of the small sample size and lower event rate. Compared with the other BMI subgroups, the apparent effect size in this subgroup was not statistically heterogeneous relative to the overall results, and so care should be taken in interpreting underpowered subgroups on their own without reference to the overall results. Some caution should be noted as safety outcomes in this study are in the context of a high-volume Bariatric Centers of Excellence surgical care system, and these complex CVD patients should be referred to experienced centers for best results. Although this study does show tremendous promise for bariatric surgery as an important adjunct treatment, future research should be aimed at a large RCT to confirm the results of this study and establish the benefits and risks of surgery in this important patient population.

The present study should be interpreted with the understanding of several important limitations. There was no access to patient medication records, diabetes laboratory test results, or assessments of their left ventricular function, which would have been helpful to include in the adjustments, although all relevant hard clinical outcomes for adjustment are included. We also had no reliable follow-up data on weight loss to establish a dose-response relationship. Because EMRALD is EMR-based, weights are recorded unsystematically during the course of routine primary care, making such analyses difficult. Similarly, for the bariatric patients, many do not return to clinic but rather complete follow-up with their family physician. Thus, subsequent weight data are limited in the bariatric registry. There are multiple randomized studies that show weight differences over time in similar clinical cohorts with expected weight loss of 20% to 30% total body weight in the surgery group and 4% to 5% in the nonsurgical group. We would expect to have similar outcomes in this cohort. Secondly, although every effort was made to extensively match and adjust for potential confounders, there were some differences in the baseline patient characteristics between the surgical and nonsurgical patients, and unmeasured confounding bias could not be ruled out entirely. Furthermore, patients who received surgery may be more likely to comply with challenging postoperative lifestyle changes, and this could lead to a potential healthy user bias.⁴³ To account for potential confounding on the basis of healthy user bias, multiple preventative health behaviors (eg, cancer screening and flu shot) were included, and these were relatively balanced between the groups. In addition, nonreferral for surgery is multifactorial, and only a small fraction of patients who are eligible actually are referred. These may be not related to patients healthy-user bias but rather a lack of knowledge and negative perception by either the patient or their primary care physician.^44,45 Ultimately, the present study does not provide data on whether patients were counseled for bariatric surgery, although all patients within the present study had access to surgery and sought other preventative health measures at rates that are not different. Immortal time bias is another consideration for comparative observational studies. However, by design, no follow-up time was misattributed as treatment time and there was no immortal time. Last, whereas all surgery patients in this study would be clinically eligible to be placed in the primary care database used to determine the control population, this database is not population-based but rather was selected to represent a reasonable extent of practices in the province of Ontario. To help further maximize the comparability of the surgical and nonsurgical cohorts, individual and neighborhood demographic characteristics, such as immigration status, income quintile, and marginalization index, and prior health service use were used in the matching process. Despite this, some residual confounding by geography or service access could remain.

This study was supported by the Ontario Bariatric Registry and ICES, which is funded by an annual grant from the MOHLTC (Ontario Ministry of Health and Long-Term Care). Parts of this material are based on data and information compiled and provided by the Ontario MOHLTC; Cancer Care Ontario; Immigration, Refugees and Citizenship Canada; the Ontario Registrar General (the original source of which is Service Ontario), and the Canadian Institute for Health Information. The analyses, conclusions, opinions and statements expressed herein are solely those of the authors and do not reflect those of the funding or data sources; no endorsement is intended or should be inferred.