P values reflect significance from log-rank test. The median follow-up time for the study cohort was 3.0 years (interquartile range, 1.7 to 4.3 years).
eTable 1. Selection of patients undergoing bariatric surgery using Medicare claims
eFigure 1. USRDS cause of death classification (derived from CMS Form 2746)
eFigure 2. Selection of study cohort
eTable 2. Patient characteristics for study cohort and all incident patients included in USRDS
eTable 3. Cumulative incidence rates and adjusted Cox proportional hazard ratios for study outcomes for patients receiving bariatric surgery compared to non-surgical controls
eFigure 3. Kaplan-Meier estimates for the cumulative incidence of secondary study outcomes. P-values reflect significance from log-rank test. The median follow up time for the study cohort was 3.0 years (interquartile range 1.7 to 4.3 years).
eTable 4. Primary analysis censoring for receipt of renal transplant
eFigure 4. Kaplan-Meier estimates for the cumulative incidence of all-cause mortality after censoring for loss to follow up and receipt of renal transplant.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Identify all potential conflicts of interest that might be relevant to your comment.
Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.
Err on the side of full disclosure.
If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.
Not all submitted comments are published. Please see our commenting policy for details.
Sheetz KH, Gerhardinger L, Dimick JB, Waits SA. Bariatric Surgery and Long-term Survival in Patients With Obesity and End-stage Kidney Disease. JAMA Surg. 2020;155(7):581–588. doi:10.1001/jamasurg.2020.0829
Is bariatric surgery associated with improvement in long-term survival for patients with obesity and end-stage kidney disease?
In this cohort study, patients receiving bariatric surgery had lower all-cause mortality at 5 years compared with matched nonsurgical control patients. Bariatric surgery was also associated with a higher incidence of kidney transplant during the study period.
Bariatric surgery may warrant further consideration in the treatment of patients with obesity and end-stage kidney disease.
Obesity rates in patients with end-stage kidney disease are rising, contribute to excess morbidity, and limit access to kidney transplant. Despite this, there continues to be controversy around the use of bariatric surgery in this patient population.
To determine whether bariatric surgery is associated with improvement in long-term survival among patients with obesity and end-stage kidney disease.
Design, Setting, and Participants
Retrospective cohort study and secondary analysis of previously collected data from the United States Renal Data System registry (2006-2015). We used Cox proportional hazards analysis to evaluate differences in outcomes for patients receiving bariatric surgery (n = 1597) compared with a matched cohort of nonsurgical patients (n = 4750) receiving usual care. Data were analyzed between September 3, 2019, and November 13, 2019.
Receipt of bariatric surgery.
Main Outcomes and Measures
All-cause mortality at 5 years. Secondary outcomes included disease-specific mortality and incidence of kidney transplant.
Surgical and nonsurgical control patients had similar age, demographics, and comorbid disease burden. The mean (SD) age was 49.8 (11.2) years for surgical patients vs 51.7 (11.1) years for nonsurgical patients. Six hundred fifteen surgical patients (38.5%) were black vs 1833 nonsurgical patients (38.6%). Surgery was associated with lower all-cause mortality at 5 years compared with usual care (cumulative incidence, 25.6% vs 39.8%; hazard ratio, 0.69, 95% CI, 0.60-0.78). This was driven by lower mortality from cardiovascular causes at 5 years for patients undergoing bariatric surgery compared with nonsurgical control patients (cumulative incidence, 8.4% vs 17.2%; hazard ratio, 0.51; 95% CI, 0.41-0.65). Bariatric surgery was also associated with an increase in kidney transplant at 5 years (cumulative incidence, 33.0% vs 20.4%; hazard ratio, 1.82; 95% CI, 1.58-2.09). However, at 1 year, bariatric surgery was associated with higher all-cause mortality compared with usual care (cumulative incidence, 8.6% vs 7.7%; hazard ratio, 1.45; 95% CI, 1.13-1.85).
Conclusions and Relevance
Among patients with obesity and end-stage kidney disease, bariatric surgery was associated with lower all-cause mortality compared with usual care. Bariatric surgery was also associated with an increase in kidney transplant. Bariatric surgery may warrant further consideration in the treatment of patients with obesity and end-stage kidney disease.
More than one-third of the 750 000 adults living with end-stage kidney disease in the United States also experience obesity.1 Use of bariatric surgery in patients without end-stage kidney disease has increased over the past decade owing to numerous studies demonstrating the health benefits of surgical weight loss.2 For example, bariatric surgery is associated with improvement in 5-year and 10-year survival compared with patients receiving medical management alone.3,4 Bariatric surgery is also associated with a lower risk of hyperglycemia, hypertension, and adverse cardiovascular events in patients with obesity and type 2 diabetes.5-7
Despite these benefits, use of bariatric surgery in patients with end-stage kidney disease, a population with baseline high rates of diabetes and hypertension, remains low. This practice was previously driven by short-term concerns that these patients are of prohibitive surgical risk.8 However, work9,10 suggests that surgical complication rates are low and comparable with patients without end-stage kidney disease.9,10 Long-term concerns remain, arising from disagreement between the potential benefits of bariatric surgery and the obesity paradox, an observation that patients undergoing dialysis with higher body mass index actually have lower all-cause mortality.11,12 As a result, it remains unclear whether bariatric surgery is associated with better or worse long-term survival in patients with end-stage kidney disease.
To address this question, we used population-based data from the United States Renal Data System (USRDS) from 2006 through 2015 to study the association between bariatric surgery and all-cause mortality in patients with end-stage kidney disease. We compared study outcomes between patients receiving bariatric surgery and a cohort of nonsurgical matched control patients receiving usual care alone.
We used patient-level data from the USRDS for the years 2006 through 2015 to identify all incident patients receiving dialysis in the United States. The USRDS is composed of several files that provide detailed demographic, clinical, and claims-based information for Medicare beneficiaries receiving dialysis. Beneficiaries are identified by unique identifiers that permit linking of information across the various files included the USRDS.
We used the Medical Evidence files to obtain information on demographics including age, race/ethnicity, and geographic location. These files also contain information on patient comorbidities (eg, presence of hypertension, diabetes, or congestive heart failure). We used the Diagnosis and Procedure Details and Physician/Supplier Claims files to identify claims for bariatric surgery. Details regarding the use of claims, including specific International Classification of Diseases, Ninth Revision and International Classification of Diseases, Tenth Revision codes used to identify procedures is included in eTable 1 in the Supplement. The USRDS is unique to other claims-based or even national clinical registries because it includes data on patients’ body mass index (BMI, calculated as weight in kilograms divided by height in meters squared). Body mass index data were derived from the Medical Evidence file, Transplant files, and the CROWNWeb files, which are collected by Medicare from dialysis and other facilities where beneficiaries receive care. We obtained information on kidney transplants from the linked United Network for Organ Sharing (UNOS) files also included in the USRDS registry. Finally, the USRDS registry maintains a Death file, which is specifically designed to include detailed information on patients’ causes of death (eFigure 1 in the Supplement). Universal reporting of patient deaths is required via the US Centers for Medicare and Medicaid form 2746 as a condition of coverage for dialysis units and transplant centers. Death data are further substantiated by the Social Security Death Master File data. We followed the Strengthening of Reporting of Observational Studies in Epidemiology (STROBE) criteria for observational studies to construct our study cohort.
This study was deemed exempt by the institutional review board at the University of Michigan owing to the use of deidentified data with no risk to participants because of the retrospective nature of the study. For this reason, patient consent was not obtained.
We identified all patients undergoing bariatric surgery (sleeve gastrectomy, Roux-en-Y gastric bypass, gastric banding, or duodenal switch) during the study period (eFigure 2 in the Supplement). We excluded patients younger than 18 years and those who underwent a similarly coded surgery for a diagnosis of malignancy. We also excluded patients with a BMI less than 35 or without a recorded BMI because this would preclude accurate matching of control individuals.
We then used a multistep process to construct a cohort of matched control patients receiving usual care alone. This overall approach has been used multiple times for similar studies and interventions.5,6 We limited controls to patients with a BMI greater than 35. We then assembled a pool of potential control patients for each patient undergoing bariatric surgery based on several criteria. These included the same census region, sex, race/ethnicity, year of enrollment in USRDS (specifically, the initiation of dialysis), age ±10 years and BMI ±5. From these pools of potential control patients, we then calculated the Mahalanobis distance between surgical patients and potential control patients based on age and the presence of comorbidities (type 2 diabetes, coronary artery disease, peripheral vascular disease, chronic obstructive pulmonary disease, history of stroke/transient ischemic attack, hypertension, smoking, or congestive heart failure).13 We selected up to 3 unique control patients for each surgical patient to create the final study cohort (eFigure 2 in the Supplement).
The a priori primary outcome for this study was 5-year all-cause mortality. We included secondary all-cause mortality at different intervals and outcomes associated with death from specific causes. These included the following categories based on the most common causes of death in the study cohort overall: cardiovascular (cardiac and vascular combined), infectious, withdrawal from dialysis, or other (all remaining causes). Finally, we compared the incidence of kidney transplant based on documented receipt of an organ from the transplant files.
The purpose of this analysis was to evaluate the association between bariatric surgery vs usual care and each study outcome. To accomplish this, we fit Cox proportional hazard regression models. The index date was the surgery date for surgical patients. Nonsurgical control patients were assigned a mock index date based on the date of surgery for their matched surgical patient. All models adjusted for potential confounders including age, race/ethnicity (white, black, Hispanic, or other), BMI, history of myocardial infarction, congestive heart failure, chronic obstructive pulmonary disease, history of stroke/transient ischemic attack, hypertension, smoking status, census region, and incident year in the USRDS registry to reflect time receiving dialysis. All analyses were censored for loss to follow-up, determined by the last date of any claim in the USRDS registry for each patient. We replicated our primary analysis censoring for kidney transplant to further isolate the association between bariatric surgery and all-cause mortality.
We conducted several sensitivity analyses to test the robustness of our findings. We replicated the primary analysis, stratifying patients by race/ethnicity (white vs black), patients who had less than and greater than the median number of comorbid conditions (median, 2), and across several strata of BMI (35-39.9, 40-45, and >45). We also tested whether our findings differed based on the type of surgery (sleeve gastrectomy vs Roux-en-Y gastric bypass) or based on whether the patient was on hemodialysis or peritoneal dialysis.
Part of our analysis was designed to evaluate the presence and extent of residual confounding from unmeasured differences in our surgical and nonsurgical cohorts. We evaluated the intervention against a falsification outcome, death from infectious causes, which tested whether the surgical cohort was healthier than matched nonsurgical control patients at baseline. A falsification outcome acts as a negative control that should not be directly influenced by the treatment in question.14,15 Because a reduction in infectious complications has not been ascribed to bariatric surgery, differences in baseline patient characteristics should have a greater influence on that outcome. Finally, we performed an E-value analysis, which assesses how large the effect from unmeasured confounding would be to negate the study results.16
All statistical analyses were performed using Stata statistical software, version 14.2 (StataCorp). All estimates were derived using robust confidence intervals with 100 replications. We used a 2-sided approach at the .05 significance level for all hypothesis testing. To detect a 30% relative difference in 5-year all-cause mortality with 95% confidence, we estimated a minimum sample size of at least 350 patients.
Patient characteristics for those receiving bariatric surgery (n = 1597) and matched nonsurgical control patients (n = 4750) are presented in Table 1. The mean age of patients receiving bariatric surgery was similar to that of nonsurgical control patients (49.8 vs 51.7 years; standardized difference, −16.9). A similar proportion of surgical patients and nonsurgical control patients were women (898 [56.2%] vs 2665 [56.1%]; standardized difference, 0.3). There was a uniform geographic distribution of surgical patients and nonsurgical control patients. Among patients receiving bariatric surgery, 720 (45.1%) underwent sleeve gastrectomy, 665 (41.6%) underwent Roux-en-Y gastric bypass, 205 (12.8%) underwent laparoscopic gastric banding, and 7 (0.4%) underwent duodenal switch. The median follow-up period was 3.0 years (interquartile range, 1.7-4.3 years). A comparison of the study cohort with all incident patients included in the USRDS registry is included in eTable 2 in the Supplement.
The cumulative incidence of all-cause and cause-specific mortality is presented in the Figure and Table 2. Bariatric surgery was associated with lower cumulative incidence of all-cause mortality at 5 years compared with usual care (cumulative incidence, 25.6% vs 39.8%; adjusted hazard ratio, 0.69; 95% CI, 0.60-0.78). (Table 3) The adjusted risk of all-cause mortality associated with bariatric surgery was higher at 1 year (adjusted hazard ratio, 1.45; 95% CI, 1.13-1.85). However, bariatric surgery was associated with a lower adjusted risk of all-cause mortality at all other times. For example, bariatric surgery was associated with a lower cumulative incidence of death from cardiovascular causes at 5 years compared with nonsurgical control patients (cumulative incidence, 8.4% vs 17.2%; adjusted hazard ratio, 0.51; 95% CI, 0.41-0.65). Bariatric surgery was also associated with an increase in kidney transplant at 5 years compared with nonsurgical control patients (cumulative incidence, 33.0% vs 20.4%; adjusted hazard ratio, 1.82; 95% CI, 1.58-2.09). In contrast, bariatric surgery was not associated with an adjusted risk of death from infectious causes, our falsification outcome, at 5 years (hazard ratio, 0.82; 95% CI, 0.52-1.30). eTable 3 in the Supplement includes data on death from additional causes, including withdrawal from dialysis.
Results from the sensitivity analyses are presented in Table 4. The association between bariatric surgery and lower all-cause mortality was consistent across various patient demographics. For example, the adjusted risk of death at 5 years was similar for patients with fewer than the median number of comorbidities (adjusted hazard ratio, 0.69; 95% CI, 0.52-0.90) and those at or greater than the median (adjusted hazard ratio, 0.68; 95% CI, 0.60-0.78). Results at 5 years were also similar for patients of white race/ethnicity (adjusted hazard ratio, 0.69; 95% CI, 0.59-0.80) compared with patients of black race/ethnicity (adjusted hazard ratio, 0.72; 95% CI, 0.58-0.90).
We performed additional analyses to confirm the accuracy of our primary findings. To isolate the potential effects of bariatric surgery from receipt of kidney transplant, we replicated our primary analysis censoring patients at the time of transplant (eFigure 4 in the Supplement). In this analysis, bariatric surgery was still associated with lower cumulative incidence of all-cause mortality at 5 years compared with usual care (cumulative incidence 31.6% v. 44.0%; adjusted hazard ratio, 0.75; 95% CI, 0.66-0.85) (eTable 4 in the Supplement), The E-value (hazard ratio) for 5-year all-cause mortality was 2.26, with a confidence interval of 1.88.
In this contemporary, population-based study of patients with obesity and end-stage kidney disease, patients receiving bariatric surgery had lower all-cause mortality compared with a matched cohort of patients receiving usual care. These results were consistent across a range of patient demographics and persisted for up to 7 years following surgery. Differences in all-cause mortality were associated with lower rates of death from cardiovascular causes in patients receiving bariatric surgery. Surgical patients were also more likely to undergo kidney transplant during the follow-up period.
Most prior work on the health effects of obesity in patients with end-stage kidney disease centers on the obesity paradox. Numerous studies have demonstrated an association between higher BMI and lower all-cause mortality in patients receiving dialysis.11,12,17-20 For example, compared with patients with a BMI of 23 to 25, those with a BMI greater than 35 had a lower risk of death overall (hazard ratio, 0.62) and death from cardiovascular causes (hazard ratio, 0.60).19 As a result of these data, some have even suggested that increasing dialysis patients’ dry weight (a measure that accounts for the incremental accumulation of body fluid between dialysis runs) may be a strategy to improve long-term outcomes.
There is a large and growing body of data indicating that bariatric surgery is associated with a multitude of health benefits, including improvements in long-term survival and lower incidence of adverse cardiovascular events.3-6 But there continues to be debate regarding the potential benefits of bariatric surgery in patients with end-stage kidney disease. On one hand, if the principles of the obesity paradox hold true, surgical weight loss could be harmful and even lead to worse long-term survival. On the other hand, patients with end-stage kidney disease continue to have poor long-term prognosis. This is driven by high rates of death from cardiovascular events, such as myocardial infarction or stroke, whose risk may be specifically attenuated by bariatric surgery.21-23
This study addresses that gap in evidence to determine whether the long-term health effects of bariatric surgery are transferrable to patients with end-stage kidney disease. Bariatric surgery is not only associated with lower all-cause mortality and death from cardiovascular causes, but also higher rates of kidney transplant. These findings indicate that bariatric surgery has direct long-term health benefits for patients with end-stage kidney disease. These findings also suggest that there may be secondary benefits associated with receipt of kidney transplant. Patients with obesity are less likely to receive a kidney transplant compared with those with normal BMI.1 These data substantiate practice changes that others have proposed, potentially using bariatric surgery as a means of increasing transplant rates.8 Finally, higher mortality in the first year following surgery suggests some upfront risks associated with treatment or potentially imperfect selection of patients.
The results from this study should be interpreted within the context of certain limitations. While our analysis may lack generalizability because it is limited to Medicare beneficiaries, Medicare provides coverage to a large proportion of US patients with end-stage kidney disease, not limited to those qualifying for benefits based on age alone. Despite our best attempts to construct a cohort of matched control patients, it is possible that our results are biased by residual confounding from important and unmeasured differences between surgical and nonsurgical patients. We performed an analysis using E-value methods, which suggests that the effect of an unmeasured covariate would have to be larger than all other measured factors, including diabetes (adjusted hazard ratio, 1.46; 95% CI, 1.30-1.64) or congestive heart failure (adjusted hazard ratio, 1.40; 95% CI, 1.25-1.57), which had the largest effect on survival among all covariates. We also found that while all-cause and cardiovascular mortality were lower for surgical patients, mortality rates from infectious causes were similar between surgical patients and nonsurgical control patients. If surgical patients were meaningfully healthier than nonsurgical control patients, we would anticipate their mortality rates to be uniformly lower. Instead, we see differences in mortality only among outcomes that have been previously ascribed to bariatric surgery. We also observed higher adjusted mortality in surgical patients in the year following their operation, further suggesting that surgical patients were not markedly healthier at baseline. While we report on the cumulative incidence of transplant, we were not able to determine whether or how bariatric surgery was used to potentially facilitate transplant. As a result, we cannot make any determinations about direct causality between bariatric surgery and eventual receipt of a kidney transplant. It is also plausible that the survival benefit of bariatric surgery is derived through access to transplant alone. However, our analysis censoring for transplant suggests that the benefits of bariatric surgery are at least partially independent of transplant.
This population-based study of patients with obesity and end-stage kidney disease found that bariatric surgery was associated with lower all-cause mortality compared with a matched cohort of nonsurgical control patients who received usual care. Receipt of bariatric surgery was also associated with an increase in the cumulative incidence of kidney transplantation. These findings suggest that bariatric surgery may warrant further consideration in the longitudinal management of patients with obesity and end-stage kidney disease.
Accepted for Publication: February 26, 2020.
Corresponding Author: Kyle H. Sheetz, MD, MSc, Center for Healthcare Outcomes and Policy, University of Michigan School of Medicine, 2800 Plymouth Rd, NCRC B016, Room 100N-11, Ann Arbor, MI 48109 (firstname.lastname@example.org).
Published Online: May 27, 2020. doi:10.1001/jamasurg.2020.0829
Author Contributions: Drs Sheetz and Waits 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: Sheetz, Dimick, Waits.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Sheetz, Gerhardinger, Waits.
Critical revision of the manuscript for important intellectual content: Sheetz, Dimick, Waits.
Statistical analysis: All authors.
Obtained funding: Dimick, Waits.
Administrative, technical, or material support: Dimick, Waits.
Supervision: Dimick, Waits.
Conflict of Interest Disclosures: Dr Dimick reported personal fees from ArborMetrix Inc outside the submitted work. No other disclosures were reported.