Abstract
Objective:
To determine whether patients with chronic kidney disease (CKD) experience higher rates of perioperative complications after Roux-en-Y gastric bypass (RYGB) compared to sleeve gastrectomy.
Summary of Background Data:
For obese CKD patients who qualify for bariatric surgery, sleeve gastrectomy is often preferred to RYGB based on perceptions of prohibitively-high perioperative risks surrounding RYGB. However, some patients with CKD are not candidates for sleeve gastrectomy and the incremental increased-risk from RYGB has never been rigorously tested in this population.
Methods:
CKD patients who underwent RYGB or sleeve gastrectomy between 2015–2017 were identified from the Metabolic and Bariatric Surgery Accreditation and Quality Improvement Program Participant Use File. RYGB patients were 1:1 propensity-score matched with sleeve gastrectomy patients based on preoperative factors that influence operative choice. Primary outcomes included 30-day readmissions, surgical complications, medical complications, and death. Secondary outcomes included the individual complications used to create the composite surgical/medical complications. Univariate logistic regression was used to compare outcomes. E-value statistic was used to test the strength of outcome point estimates against possible unmeasured confounding.
Results:
Demographics were similar between RYGB (n=673) and sleeve gastrectomy (n=673) cohorts. There were no statistically-significant differences in primary outcomes. Among secondary outcomes, only acute kidney injury was statistically-significantly higher among RYGB patients (4.9% vs 2.7%, P=.035, E-value 1.27).
Conclusions:
Among well-matched cohorts of RYGB and sleeve gastrectomy patients, incidence of primary outcomes were similar. Among secondary outcomes, only acute kidney injury was statistically-significantly higher among RYGB patients; however, the E-value for this difference was small and relatively weak confounder(s) could abrogate the statistical difference. The perception that RYGB has prohibitively-high perioperative risks among CKD patients is disputable and operative selection should be weighed on patient candidacy and anticipated long-term benefit.
MINI-ABSTRACT
In a population-based cohort study of propensity-score matched CKD patients who underwent RYGB or sleeve gastrectomy, there were no statistically-significant differences in 30-day readmissions, surgical complications, medical complications, or death. The common perception that RYGB has prohibitively-high perioperative risks among CKD patients is disputable and operative selection should be weighed on patient candidacy and anticipated long-term benefit.
INTRODUCTION
Obesity with a body mass index (BMI) ≥35 kg/m2 is both common and growing among adult patients with chronic kidney disease (CKD) in the United States. According to the most recent data from the National Health and Nutrition Examination Survey, the proportion of CKD patients with BMI ≥35 kg/m2 has grown from 17.2% between 1999–2002 to 22.2% between 2011–2014, representing over 7.2-million US adults during the later period1,2. Among this population, bariatric surgery has been shown to have substantial and durable long-term benefits including improved glomerular filtration rate, reduced glomerular hyperfiltration, and reversal of albuminuria and/or proteinuria3–12.
For CKD patients with BMI ≥35 kg/m2 being considered for bariatric surgery, the question of whether patients are candidates for Roux-en-Y gastric bypass (RYGB) often arises. One reason sleeve gastrectomy might be preferred over RYGB is due to the common perception of prohibitive perioperative risk for RYGB. This perception is largely informed by observational studies among patients without kidney disease, which have suggested that RYGB is associated with increased 30-day postoperative rates of intraabdominal leak, morbidity, and mortality when compared to sleeve gastrectomy13–15. However, these outcomes have not been rigorously studied among CKD patients and it is unknown whether they are truly generalizable to this population. This is especially important when access to care is considered: CKD patients with BMI ≥35 kg/m2 and relative or absolute contraindications to sleeve gastrectomy might be relegated to intensive medical therapy with poor long-term outcomes instead of being considered for RYGB16.
In this context, we performed a rigorous analysis of perioperative safety between RYGB and sleeve gastrectomy among CKD patients utilizing a robust, national database capturing >95% of all bariatric operations performed in the United States. In particular, we sought to match patients based on preoperative factors known to affect operative selection between RYGB and sleeve gastrectomy to better reflect real-world operative decision-making among surgeons. Finally, as this study is based on observational data, we reported “E-values” for all statistically-significant differences in outcomes to test for the strength of outcome point estimates against possible unmeasured confounding. We hypothesize that perioperative outcomes after RYGB or sleeve gastrectomy in the CKD population will mirror those of patients without kidney disease.
METHODS
Data Source
The Metabolic and Bariatric Surgery Accreditation and Quality Improvement Program (MBSAQIP) is a combination of national bariatric surgery accreditation programs administered by the American College of Surgeons (ACS) and the American Society for Metabolic and Bariatric Surgery (ASMBS). As of April 1, 2012, all institutions that met standards under the ACS and ASMBS programs were extended accreditation to the joint MBSAQIP17. As of June 1, 2019, a total of 819 US centers participate in MBSAQIP, where trained clinical reviewers prospectively extract patient data within 30-days of index procedure from every bariatric operation performed at each participating center18. This subsequently captures >95% of all bariatric operations performed in the United States19. This data is then submitted to MBSAQIP, which regularly audits participating centers to ensure accuracy. Each year, MBSAQIP publishes participant use files that contains de-identified patient-level data that masks center, surgeon, or geographic identifiers.
Patient Population
All adult patients (age ≥18 years) with preoperative kidney insufficiency who underwent primary, laparoscopic RYGB or sleeve gastrectomy between 2015–2017 and completed 30 days of follow-up after their operation were included in the analysis. Preoperative kidney insufficiency is captured as a binary variable in the MBSAQIP dataset and is defined as preoperative serum creatinine level >2 mg/dL without the need for dialysis within 2-weeks before index procedure. As previously described, patients with preoperative kidney insufficiency within the MBSAQIP database are presumed to have CKD given that acute kidney injury is a contraindication to undergoing non-urgent procedures20. Patients who received dialysis within 2 weeks before index procedure were excluded from analysis, as they were assumed to have end-stage kidney disease and represent a unique population with different perioperative risks and clinical management strategies than CKD patients21–25. Operative procedure was defined by Current Procedural Terminology code (sleeve gastrectomy, 43775; RYGB, 43644 and 43645).
Patients were excluded if they underwent revision or conversion surgery from a previous bariatric procedure. Additionally, patients who did not complete follow-up after 30 days from their operative date were excluded from the analysis. However, if a patient experienced a complication or readmission within 30-days of their operation, they were included in the analysis regardless of their 30-day follow-up status. In all, 70 patients (50 sleeve gastrectomy, 20 RYGB) were excluded from the study due to incomplete 30-day follow up and no recorded complication or readmission.
Outcomes
The primary outcomes were readmissions, surgical complications, medical complications, and death within 30-days of bariatric surgery. Surgical complications included unplanned reoperation, endoscopic intervention, percutaneous drain placement, intraabdominal leak, sepsis, deep or organ-space infection, and superficial surgical site infection. Medical complications included myocardial infarction, unplanned intubation, new venous thromboembolism, pneumonia, stroke, urinary tract infection, acute kidney injury (AKI), need for acute dialysis, unplanned intensive care unit stay, and need for cardiopulmonary resuscitation. AKI was defined as a postoperative rise in serum creatinine >2 mg/dL from preoperative value. Additionally, any patient that needed acute dialysis was considered to have experienced AKI. Secondary outcomes were the individual complications that were used to construct the composite variables.
Propensity Score Matching
In order to account for an inherent selection bias that might exist for patients to undergo RYGB versus sleeve gastrectomy, propensity-score matching was performed based on patient-level factors that influence operative selection. These factors were selected a priori by study investigators based on published data. Previous randomized studies have suggested that RYGB is associated with a modest increase in weight loss and comorbidity resolution when compared to sleeve gastrectomy16,26,27. Patients with severe gastroesophageal reflux disease might also be more likely to undergo RYGB instead of sleeve gastrectomy, as the former may ameliorate the condition and the latter might worsen it27,28. Finally, patients who use chronic steroids/immunosuppressants or are smokers might be less likely to undergo RYGB given the risks of short-term complications and long-term marginal ulcers, respectively29–31.
Therefore, sleeve gastrectomy patients were propensity-score matched with a similar cohort of patients who underwent RYGB in a 1:1 ratio for preoperative body mass index, diabetes, hypertension, hyperlipidemia, smoking status, chronic steroid/immunosuppressant use, and gastroesophageal reflux disease. As described elsewhere, a logistic regression model was used to generate a propensity score for each patient based on confounding factors32. The patients were then matched without replacement based on their propensity scores to the nearest neighbor within 0.005 of the estimated score. Adequate overlap was assessed via common support graphing and covariate balance. All RYGB patients (n=673) were able to obtain a sleeve gastrectomy control (n=673) using this approach.
Statistical Analysis
Student’s t-test was used to compare means between groups, Wilcoxon rank-sum test was used to compare ordinal variables, and chi-square test was used to compare categorical variables. As all of the outcome variables were dichotomous, univariate logistic regression was used to compare outcomes. Unadjusted incidence rates were reported for each complication and 95% confidence intervals were constructed. Due to the rarity of complications and similar frequency of comorbidities between the propensity-score matched cohorts, multivariable logistic regression was not performed.
Given that the MBSAQIP dataset is observational (nonrandomized) and lacks provider, institutional, or geographic identifiers, potential unmeasured confounding might still remain even after propensity-score matching. Therefore, in order to test the robustness of statistically-significant outcome point estimates, E-values were computed based on the lower-95% confidence interval of outcome point estimates. In effect, the E-values estimate the minimum strength of any unmeasured confounder(s) association with both the treatment (surgical choice) and outcome (complications) to nullify the results. Higher E-values correspond to increasing levels of strength needed for any unmeasured confounder(s) to negate the study results. This test makes no assumptions about the source of potential unmeasured confounders. The E-value analysis was done per the methods of Ding and VanderWeele33–35.
Two-sided P<.05 was used to indicate statistical significance. The dataset had minimal missingness for all variables, as noted in Table 1. All analyses were conducted using STATA version 15.1 (StataCorp 2017. Stata Statistical Software: Release 15. College Station, TX: StataCorp LLC). Figures were created using GraphPad Prism 7 for Windows (GraphPad Software, San Diego, CA: www.graphpad.com). This study was deemed exempt by the Michigan Medicine Institutional Review Board.
Table 1.
Patient and operative characteristics.
| Study Group | P value | ||||
|---|---|---|---|---|---|
| A. Sleeve gastrectomy, unmatched (n=1,442) | B. Sleeve gastrectomy, matched (n=673) | C. Roux-en-Y gastric bypass (n=673) | A vs C | B vs C | |
| Age, mean (SD), y | 54.6 (11.2) | 56.0 (10.8) | 55.6 (10.3) | .062 | .46 |
| Male sex, No. (%) | 658 (45.6) | 327 (48.6) | 295 (44.0) | .44 | .080 |
| Race, No. (%) a | <.001 | .040 | |||
| White | 825 (57.2) | 400 (59.4) | 441 (65.5) | ||
| Black or African American | 421 (29.2) | 186 (27.6) | 147 (21.8) | ||
| Hispanic | 121 (8.4) | 49 (7.3) | 49 (7.3) | ||
| Other b | 17 (1.2) | 12 (1.8) | 19 (2.8) | ||
| Body mass index, mean (SD), kg/m2 c | 46.3 (8.3) | 46.2 (7.9) | 46.9 (8.4) | .14 | .11 |
| Preoperative hematocrit, mean (SD), % d | 38.2 (5.2) | 38.2 (5.1) | 37.9 (5.0) | .25 | .39 |
| Dependent functional status, No. (%) | 85 (5.9) | 45 (6.7) | 33 (4.9) | .36 | .16 |
| Medical comorbidities | |||||
| Hypertension, No. (%) | 1,340 (92.9) | 649 (96.4) | 649 (96.4) | .002 | .99 |
| Hyperlipidemia, No. (%) | 886 (61.4) | 456 (67.8) | 476 (70.7) | <.001 | .24 |
| Diabetes mellitus, No. (%) | 888 (61.6) | 530 (78.8) | 530 (78.8) | <.001 | .99 |
| Gastroesophageal reflux disease, No. (%) | 562 (39.0) | 302 (44.9) | 302 (44.9) | .010 | .99 |
| Obstructive sleep apnea, No. (%) | 855 (59.3) | 397 (59.0) | 441 (65.5) | .006 | .013 |
| Current smoker within 1 year, No. (%) | 108 (7.5) | 48 (7.1) | 35 (5.2) | .051 | .14 |
| Chronic obstructive pulmonary disease, No. (%) | 110 (7.6) | 56 (8.3) | 46 (6.8) | .52 | .30 |
| Chronic steroid or immunosuppressant use, No. (%) | 136 (9.4) | 39 (5.8) | 39 (5.8) | .005 | .99 |
| History of percutaneous coronary intervention, No. (%) | 190 (13.2) | 111 (16.5) | 101 (15.0) | .26 | .45 |
| History of cardiac surgery, No. (%) | 133 (9.2) | 72 (10.7) | 62 (9.2) | .99 | .36 |
| History of venous thromboembolism, No. (%) | 115 (8.0) | 58 (8.6) | 50 (7.4) | .66 | .42 |
| Operative characteristics | |||||
| Operative time, mean (SD), m e | 83 (47) | 87 (49) | 137 (64) | <.001 | <.001 |
| Drain placement during principle operation, No. (%) | 228 (15.8) | 110 (16.3) | 210 (31.2) | <.001 | <.001 |
| Conversion to open procedure, No. (%) | 5 (0.3) | 3 (0.4) | 3 (0.4) | .73 | .99 |
| Index hospital length of stay, mean (SD), d f | 2.3 (3.0) | 2.5 (3.5) | 3.0 (3.8) | <.001 | .011 |
Missingness: sleeve gastrectomy-unmatched 58/1,442 (4.0%), sleeve gastrectomy-matched 26/673 (3.9%), Roux-en-Y gastric bypass 17/673 (2.5%).
Other categories: Asian, American Indian or Alaska Native, Native Hawaiian or Other Pacific Islander.
Missingness: sleeve gastrectomy-unmatched 1/1,442 (<0.1%), sleeve gastrectomy-matched 0/673 (0.0%), Roux-en-Y gastric bypass 0/673 (0.0%).
Missingness: sleeve gastrectomy-unmatched 123/1,442 (8.5%), sleeve gastrectomy-matched 67/673 (10.0%), Roux-en-Y gastric bypass 67/673 (10.0%).
Missingness: sleeve gastrectomy-unmatched 0/1,442 (0.0%), sleeve gastrectomy-matched 0/673 (0.0%), Roux-en-Y gastric bypass 1/673 (0.01%).
Missingness: sleeve gastrectomy-unmatched 2/1,442 (0.1%), sleeve gastrectomy-matched 1/673 (0.1%), Roux-en-Y gastric bypass 3/673 (0.4%)
RESULTS
Between 2015–2017, a total of 2,115 CKD patients underwent primary, laparoscopic sleeve gastrectomy (n=1,442, 68.2%) or RYGB (n=673, 31.8%). Whereas incidence of RYGB remained stable over the study period (range 216 to 229 cases/year), sleeve gastrectomy increased from 437 to 509 cases/year (Figure 1). Before propensity-score matching, patients who underwent RYGB were more likely to be white race; have more medical comorbidities; and have longer operative time and postoperative hospital length of stay (Table 1). RYGB patients were less likely to take chronic steroids or immunosuppressants. After propensity-score matching, overall medical comorbidities and chronic steroid/immunosuppressant use were similar between the cohorts; however, patients who underwent RYGB continued to experience longer operative times and postoperative hospital length of stay (Table 1).
Figure 1.
Frequency of sleeve gastrectomy and Roux-en-Y gastric bypass among chronic kidney disease patients, by year.
Logistic Regression
On logistic regression, there were no statistical differences between RYGB and sleeve gastrectomy in terms of readmissions (OR 1.03, 95%CI 0.73–1.45, P=.86), surgical complications (OR 1.42, 95%CI 0.94–2.14, P=.098), medical complications (OR 1.08, 95%CI 0.74–1.58, P=.70), and death (OR 1.34, 95%CI 0.30–5.99, P=.71) (Figure 2). Among individual complications, only AKI was more likely after RYGB (OR 1.88, 95%CI 1.05–3.37, P=.035). The E value for this association was 1.27. New venous thromboembolism could not be statistically compared between the cohorts because there were no occurrences in the sleeve gastrectomy cohort.
Figure 2.
Univariate logistic regression of 30-day perioperative complications between Roux-en-Y gastric bypass and sleeve gastrectomy among propensity-score matched chronic kidney disease patients. Surgical complications include unplanned reoperation, endoscopic intervention, percutaneous drain placement, intraabdominal leak, sepsis, deep surgical site infection, and superficial surgical site infection. Medical complications include myocardial infarction, unplanned intubation, new venous thromboembolism, pneumonia, stroke, urinary tract infection, acute kidney injury, need for acute dialysis, unplanned intensive care unit stay, and need for cardiopulmonary resuscitation. Dots represent odds ratio and bands represent 95% confidence intervals. Statistically-significant odds ratios are bolded. No instances of new venous thromboembolism among sleeve gastrectomy patients. Abbreviations: ICU, intensive care unit; SSI, surgical site infection; CPR, cardiopulmonary resuscitation; OR, odds ratio; CI, confidence interval.
Incidence Rates
Incidence of readmission was 11.1% (95%CI 8.8 to 13.5%) after RYGB and 10.8% (95%CI 8.5 to 13.2%) after sleeve gastrectomy (absolute rate difference +0.3%, 95%CI −3.0 to 3.6%) (Figure 3A). Incidence of death was 0.6% (95%CI 0.0 to 1.2%) after RYGB and 0.4% (95%CI 0.0 to 0.9%) after sleeve gastrectomy (absolute rate difference +0.1%, 95%CI −0.6 to 0.9%).
Figure 3A.
Incidence of 30-day composite perioperative complications between Roux-en-Y gastric bypass and sleeve gastrectomy among propensity-score matched chronic kidney disease patients. Surgical complications include unplanned reoperation, endoscopic intervention, percutaneous drain placement, intraabdominal leak, sepsis, deep surgical site infection, and superficial surgical site infection. Medical complications include myocardial infarction, unplanned intubation, new venous thromboembolism, pneumonia, stroke, urinary tract infection, acute kidney injury, need for acute dialysis, unplanned intensive care unit stay, and need for cardiopulmonary resuscitation. Bands indicate upper-95% confidence interval. *Indicates P<.05. Abbreviations: ICU, intensive care unit; SSI, surgical site infection; CPR, cardiopulmonary resuscitation.
Incidence of surgical complications was 8.6% (95%CI 6.5 to 10.7%) after RYGB and 6.2% (95%CI 4.4 to 8.1%) after sleeve gastrectomy (absolute rate difference +2.4%, 95%CI −0.4 to 5.2%) (Figure 3B). Among individual surgical complications, none were statistically-significantly different between RYGB or sleeve gastrectomy cohorts (Table 2).
Figure 3B.
Incidence of individual 30-day surgical complications between Roux-en-Y gastric bypass and sleeve gastrectomy among propensity-score matched chronic kidney disease patients. Bands indicate upper-95% confidence interval. *Indicates P<.05. Abbreviations: SSI, surgical site infection.
Table 2.
Incidence of 30-day perioperative complications between Roux-en-Y gastric bypass and sleeve gastrectomy among propensity-score matched chronic kidney disease patients. Surgical complications include unplanned reoperation, endoscopic intervention, percutaneous drain placement, intraabdominal leak, sepsis, deep surgical site infection, and superficial surgical site infection. Medical complications include myocardial infarction, unplanned intubation, new venous thromboembolism, pneumonia, stroke, urinary tract infection, acute kidney injury, need for acute dialysis, unplanned intensive care unit stay, and need for cardiopulmonary resuscitation.
| Complication | Incidence rate, % (95%CI) | Absolute rate difference, % (95%CI) | |
|---|---|---|---|
| Sleeve gastrectomy | Roux-en-Y gastric bypass | ||
| Readmission | 10.8 (8.5 to 13.2) | 11.1 (8.8 to 13.5) | +0.3 (−3.0 to 3.6) |
| Surgical complication | 6.2 (4.4 to 8.1) | 8.6 (6.5 to 10.7) | +2.4 (−0.4 to 5.2) |
| Unplanned reoperation | 1.9 (0.9 to 3.0) | 3.3 (1.9 to 4.6) | +1.3 (−0.4 to 3.0) |
| Endoscopic intervention | 1.8 (0.8 to 2.8) | 2.8 (1.6 to 4.1) | +1.0 (−0.6 to 2.6) |
| Transfusion | 2.2 (1.1 to 3.3) | 2.4 (1.2 to 3.5) | +0.1 (−1.5 to 1.8) |
| Deep surgical site infection | 0.4 (0.0 to 1.0) | 1.2 (0.4 to 2.0) | +0.7 (−0.2 to 1.7) |
| Intraabdominal leak | 0.7 (0.1 to 1.4) | 1.0 (0.3 to 1.8) | +0.3 (−0.7 to 1.3) |
| Percutaneous drain placement | 0.1 (0.0 to 0.4) | 0.9 (0.2 to 1.6) | +0.7 (0.0 to 1.5) |
| Sepsis | 0.1 (0.0 to 0.4) | 0.6 (0.0 to 1.2) | +0.4 (−0.2 to 1.1) |
| Superficial surgical site infection | 0.3 (0.0 to 0.7) | 0.6 (0.0 to 1.2) | +0.3 (−0.4 to 1.0) |
| Medical complication | 8.3 (6.2 to 10.4) | 8.9 (6.8 to 11.1) | +0.6 (−2.4 to 3.6) |
| Acute kidney injury | 2.7 (1.5 to 3.9) | 4.9 (3.3 to 6.5) | +2.2 (0.2 to 4.3) |
| Unplanned intensive care unit stay | 4.0 (2.5 to 5.5) | 4.3 (2.8 to 5.8) | +0.3 (−1.8 to 2.4) |
| Need for acute dialysis | 1.8 (0.8 to 2.8) | 2.7 (1.5 to 3.9) | +0.9 (−0.7 to 2.5) |
| Unplanned intubation | 1.3 (0.5 to 2.2) | 1.2 (0.4 to 2.0) | −0.1 (−1.3 to 1.0) |
| Pneumonia | 1.0 (0.3 to 1.8) | 0.9 (0.2 to 1.6) | −0.1 (−1.2 to 0.9) |
| Myocardial infarction | 1.0 (0.3 to 1.8) | 0.4 (0.0 to 1.0) | −0.6 (−1.5 to 0.3) |
| Urinary tract infection | 1.3 (0.5 to 2.2) | 0.3 (0.0 to 0.7) | −1.0 (−2.0 to 0.1) |
| New venous thromboembolism a | No observations | 0.4 (0.0 to 0.8) | |
| Stroke | 0.4 (0.0 to 0.9) | 0.1 (0.0 to 0.4) | −0.3 (−0.9 to 0.3) |
| Need for cardiopulmonary resuscitation | 0.6 (0.0 to 1.2) | 0.1 (0.0 to 0.4) | −0.4 (−1.1 to 0.2) |
| Death | 0.4 (0.0 to 0.9) | 0.6 (0.0 to 1.2) | +0.1 (−0.6 to 0.9) |
No incidence of new venous thromboembolism among sleeve gastrectomy patients.
Incidence of medical complications was 8.9% (95%CI 6.8 to 11.1%) after RYGB and 8.3% (95%CI 6.2 to 10.4%) after sleeve gastrectomy (absolute rate difference +0.6%, 95%CI −2.4 to 3.6%) (Figure 3C). Among individual medical complications, perioperative AKI was significantly higher after RYGB: 4.9% (95%CI 3.3 to 6.5%) vs 2.7% (1.5 to 3.9%), absolute rate difference +2.2% (95%CI 0.2 to 4.3%) (Table 2). No other individual medical complication was significantly different between RYGB or sleeve gastrectomy cohorts.
Figure 3C.
Incidence of individual 30-day medical complications between Roux-en-Y gastric bypass and sleeve gastrectomy among propensity-score matched chronic kidney disease patients. Bands indicate upper-95% confidence interval. Abbreviations: CPR, cardiopulmonary resuscitation; ICU, intensive care unit; VTE, venous thromboembolism.
* Indicates P<.05.
a No instances of new venous thromboembolism among sleeve gastrectomy patients.
DISCUSSION
In this contemporary analysis of 30-day bariatric surgery outcomes among CKD patients, RYGB and sleeve gastrectomy had statistically-equivalent rates of primary outcomes including readmissions, surgical complications, medical complications, and death. Of the 17 secondary outcomes that were analyzed, only AKI was statistically-significantly different between the cohorts, with RYGB patients experiencing a 2.2% absolute increased incidence of this outcome (4.9% vs 2.7%).
No statistical difference in primary outcomes between RYGB and sleeve gastrectomy might appear surprising, especially given that RYGB has previously been shown in observational studies to have worse perioperative outcomes among patients with normal kidney function13–15. However, the nuance is in the matching. Previous studies of patients with normal kidney function have utilized multivariable logistic regression to compare outcomes between operative approaches while controlling for up to 20+ different covariates14. In essence, this approach asks whether RYGB has inferior outcomes after controlling for the influence of those covariates on the outcome. Yet we know from clinical experience that many of these factors are not only related to outcome, but treatment choice as well. For instance, gastroesophageal reflux disease is seen by many surgeons as a relative contraindication to sleeve gastrectomy due to worsening reflux in 32% of patients and up to a 9% reoperation rate for medically-refractory disease in randomized trials16,26,27. Such multivariable logistic regression models fail to account for these baseline differences in study populations and thereby compare populations that would not be clinically eligible for both operations. As such, they are limited by an inherent selection bias that confounds the results.
Conversely, our analysis matches on covariates related to treatment choice in order to limit the potential confounding found in observational (nonrandomized) data. In effect, this mimics some of the particular characteristics of a randomized controlled trial, where subjects have similar probability of either treatment36. This technique has been well-described and utilized elsewhere32,37. Therefore, our study answers not which bariatric operation was safer for every CKD patient who underwent bariatric surgery, but rather which is safer for CKD patients with similar propensity to undergo either operation.
Similar 30-day incidences of readmissions, surgical complications, medical complications, and death between well-matched CKD patients who underwent RYGB or sleeve gastrectomy should help to assuage concerns among bariatric surgeons that RYGB is prohibitively dangerous in the perioperative period for CKD patients who would qualify for either procedure. Patients with CKD who might benefit more from RYGB should not be steered towards sleeve gastrectomy based on the perceived short-term risks of RYGB. Even more importantly, our findings should impact access to care among CKD patients with relative or absolute contraindications to sleeve gastrectomy. Given the perception that RYGB is prohibitively risky for a medically-vulnerable CKD population, patients with relative or absolute contraindications to sleeve gastrectomy (e.g. Barrett’s esophagus or severe gastroesophageal reflux disease) might be relegated to intensive medical therapy instead. Intensive medical therapy is rarely successful for long-term weight loss among any adult population, and persistent obesity with BMI >35 kg/m2 has been prospectively shown to predict worsening kidney disease over time16,38,39. Should these patients’ kidney disease later progress to end-stage kidney disease, persistent obesity with BMI >35 kg/m2 can then act as a barrier to kidney transplantation, as the majority of centers in the United States use BMI 35–40 kg/m2 as a cutoff to kidney transplantation40. Our findings suggest that CKD patients with relative or absolute contraindications for sleeve gastrectomy should not be precluded from RYGB consideration based on concerns of perioperative safety alone.
Among the 17 secondary outcomes in this study, the only statistically-significant difference between RYGB and sleeve gastrectomy was a higher incidence of perioperative AKI among the RYGB cohort. Nearly 1 in 20 (4.9%) CKD patients who underwent RYGB experienced this complication compared to nearly 1 in 40 (2.7%) sleeve gastrectomy patients. Each of these incidence rates are significantly higher than rates of perioperative AKI among patients with normal kidney function, which range from 0.2–0.4%14,41. This difference between patients with CKD and normal kidney function underscores the vulnerability of CKD patients to perioperative AKI.
RYGB has previously been shown to have a higher risk of perioperative AKI when compared to sleeve gastrectomy; however, the etiology of this difference remains uncertain14. Retrospective studies have shown diabetes, postoperative dehydration, surgical complications, and perioperative use of angiotensin-converting enzyme (ACE) inhibitors or angiotensin II receptor blockers (ARB) to be associated with AKI after bariatric surgery42,43. While incidence of diabetes and surgical complications were similar between our cohorts, we were unable to test for postoperative dehydration or perioperative ACE inhibitor/ARB use. Of those CKD patients who experienced perioperative AKI (RYGB 4.9%, sleeve gastrectomy 2.7%), over half of each cohort also required dialysis (RYGB 2.7%, sleeve gastrectomy 1.8%). As the MBSAQIP database only follows patients to 30-days after their index operation, it is unknown whether these patients required long-term dialysis as a result of their perioperative AKI. However, previous prospective studies show that as many as 60% of CKD patients who require short-term dialysis after bariatric surgery will progress to long-term dialysis as a result41. This underscores the need for aggressive measures to prevent perioperative AKI in this population and further investigation is needed to identify AKI risk factors among CKD patients after bariatric surgery.
However, as this study’s finding of increased rates of perioperative AKI is based on observational data, consideration must be given to the strength of this difference against potential unmeasured confounding. The lower-95% confidence interval for this point estimate is 1.05, which corresponds to an E value of 1.27. This E value is quite low, suggesting that presence of relatively weak unmeasured confounder(s) would be sufficient to abrogate the statistical difference between cohorts. For instance, previous studies have assessed the influence of provider or institutional factors on perioperative AKI after bariatric surgery. Haskins et al. recently showed on multivariable analysis that senior-level resident versus no resident was associated with a 1.96 (95%CI 1.13–3.42) odds of perioperative AKI, whereas junior-level resident versus no resident was associated with an even higher 2.08 (95%CI 1.12–3.86) odds of perioperative AKI44. It might also be plausible that the routine postoperative care is dissimilar between RYGB and sleeve gastrectomy patients (e.g. fluid management strategies), which may contribute to kidney outcomes. The relative weakness of our finding and inability to adjust for provider, institutional, or geographic factors underscores that a circumspect approach is needed when attempting to generalize observational data where association does not necessarily imply causation.
This study has several limitations. First, CKD and associated comorbidities (e.g. diabetes, gastroesophageal reflux disease) are classified as binary variables in the MBSAQIP database. In effect, this ignores the varying clinical severity of each patient’s disease(s) and makes our analysis prone to misclassification bias (e.g. patients with more advanced CKD stages or medical comorbidities being concentrated in one of the surgical cohorts). We attempted to control for disease characteristics as much as possible through propensity-score matching, but the aptitude our model is only as strong as the granularity of its components. Second, as the MBSAQIP database uses a preoperative serum creatinine ≥2 mg/dL to define CKD, this definition does not reflect international classifications of CKD that are based on glomerular filtration rate, presence of albuminuria, and time interval of kidney function impairment (generally >3 months)45. Similarly, the MBSAQIP definition of acute kidney injury (postoperative rise in serum creatinine >2 mg/dL from preoperative value) is not equivalent to internationally-accepted criteria that incorporate urine output per kilogram, percentage increase in serum creatinine above baseline level, increase in serum creatinine above a particular threshold, time interval, and need for dialysis46–48. The MBSAQIP definitions may further contribute to misclassification bias in our study. Third, the MBSAQIP dataset is blinded to provider, institutional, and geographic factors. As such, we were unable to test or adjusted for these factors. To address this, we performed E value calculations for any statistically-significant differences in order to test the strength of any point estimates against unmeasured confounding. Finally, MBSAQIP outcomes are limited to only 30-days after index operation. A true recommendation on operative selection must consider postoperative outcomes well-beyond this short timeframe.
CONCLUSIONS
Among a carefully-selected cohort of CKD patients who underwent RYGB or sleeve gastrectomy, neither operation was conclusively safer in the perioperative period. The perception that RYGB has prohibitively-high perioperative risks among CKD patients is disputable and operative selection among surgeons should be weighed on the anticipated long-term benefit to the individual patient. Patients precluded from sleeve gastrectomy should be considered for RYGB instead of being relegated to intensive medical therapy. Further randomized trials that assess both short and long-term outcomes of bariatric surgery among patients with CKD are warranted.
ACKNOWLEDGEMENTS
John R. Montgomery is supported by the Obesity Surgery Scientist Fellowship Award administered from the National Institute of Diabetes and Digestive and Kidney Diseases T32-DK108740. Justin B. Dimick is supported by the Long-Term Comparative Safety and Economic Consequences of Sleeve Gastrectomy grant administered by the National Institute of Diabetes and Digestive and Kidney Diseases R01-DK115401. Dana A. Telem is supported by the Developing and Implementing Evidence-Based Hernia Care grant administered by the Agency for Healthcare Research & Quality K08-HS025778.
Sources of support:
T32-DK108740 (John R. Montgomery)
R01-DK115401 (Justin B. Dimick)
K08-HS025778 (Dana A. Telem)
Disclosure of funding:
T32-DK108740 (John R. Montgomery)
R01-DK115401 (Justin B. Dimick)
K08-HS025778 (Dana A. Telem)
REFERENCES
- 1.2018 USRDS Annual Data Report Reference Tables. Available at: https://www.usrds.org/reference.aspx. Accessed on June 14, 2019.
- 2.Chang AR, Grams ME, Navaneethan SD. Bariatric Surgery and Kidney-Related Outcomes. Kidney international reports. 2017;2(2):261–270. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Navaneethan SD, Yehnert H. Bariatric surgery and progression of chronic kidney disease. Surgery for obesity and related diseases : official journal of the American Society for Bariatric Surgery. 2009;5(6):662–665. [DOI] [PubMed] [Google Scholar]
- 4.Schuster DP, Teodorescu M, Mikami D, et al. Effect of bariatric surgery on normal and abnormal renal function. Surgery for obesity and related diseases : official journal of the American Society for Bariatric Surgery. 2011;7(4):459–464. [DOI] [PubMed] [Google Scholar]
- 5.Luaces M, Martinez-Martinez E, Medina M, et al. The impact of bariatric surgery on renal and cardiac functions in morbidly obese patients. Nephrol Dial Transplant. 2012;27Suppl 4:iv53–57. [DOI] [PubMed] [Google Scholar]
- 6.Hou CC, Shyu RS, Lee WJ, et al. Improved renal function 12 months after bariatric surgery. Surgery for obesity and related diseases : official journal of the American Society for Bariatric Surgery. 2013;9(2):202–206. [DOI] [PubMed] [Google Scholar]
- 7.Fenske WK, Dubb S, Bueter M, et al. Effect of bariatric surgery-induced weight loss on renal and systemic inflammation and blood pressure: a 12-month prospective study. Surgery for obesity and related diseases : official journal of the American Society for Bariatric Surgery. 2013;9(4):559–568. [DOI] [PubMed] [Google Scholar]
- 8.Imam TH, Fischer H, Jing B, et al. Estimated GFR Before and After Bariatric Surgery in CKD. American journal of kidney diseases : the official journal of the National Kidney Foundation. 2017;69(3):380–388. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Holcomb CN, Goss LE, Almehmi A, et al. Bariatric surgery is associated with renal function improvement. Surgical endoscopy. 2018;32(1):276–281. [DOI] [PubMed] [Google Scholar]
- 10.Bilha SC, Nistor I, Nedelcu A, et al. The Effects of Bariatric Surgery on Renal Outcomes: a Systematic Review and Meta-analysis. Obesity surgery. 2018;28(12):3815–3833. [DOI] [PubMed] [Google Scholar]
- 11.Friedman AN, Wahed AS, Wang J, et al. Effect of Bariatric Surgery on CKD Risk. J Am Soc Nephrol. 2018;29(4):1289–1300. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Magalhaes DSC, Pedro JMP, Souteiro PEB, et al. Analyzing the Impact of Bariatric Surgery in Kidney Function: a 2-Year Observational Study. Obesity surgery. 2019;29(1):197–206. [DOI] [PubMed] [Google Scholar]
- 13.Alizadeh RF, Li S, Inaba C, et al. Risk Factors for Gastrointestinal Leak after Bariatric Surgery: MBASQIP Analysis. Journal of the American College of Surgeons. 2018;227(1):135–141. [DOI] [PubMed] [Google Scholar]
- 14.Kumar SB, Hamilton BC, Wood SG, et al. Is laparoscopic sleeve gastrectomy safer than laparoscopic gastric bypass? a comparison of 30-day complications using the MBSAQIP data registry. Surgery for obesity and related diseases: official journal of the American Society for Bariatric Surgery. 2018;14(3):264–269. [DOI] [PubMed] [Google Scholar]
- 15.Ladak F, Dang JT, Switzer NJ, et al. Rates of reoperation and nonoperative intervention within 30 days of bariatric surgery. Surgery for obesity and related diseases: official journal of the American Society for Bariatric Surgery. 2019;15(3):431–440. [DOI] [PubMed] [Google Scholar]
- 16.Schauer PR, Bhatt DL, Kirwan JP, et al. Bariatric Surgery versus Intensive Medical Therapy for Diabetes - 5-Year Outcomes. N Engl J Med. 2017;376(7):641–651. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.MBSAQIP website. Available at: https://asmbs.org/integrated-health/mbsaqip. Accessed on June 14, 2019.
- 18.MBSAQIP Accredited Bariatric Surgery Centers. Available at: https://www.facs.org/search/bariatric-surgery-centers?page=1&n=25&country=United%20States. Accessed on June 14, 2019.
- 19.Hutter MM. MBSAQIP Data Collection Program. 2016; http://mdepinet.org/wp-content/uploads/S2_7_Hutter-MBSAQIP-Data-PROs-Postmarket-surveillance-and-Performance-Metrics-for-MDEpinet-FDA-7-29-16.pdf. Accessed on June 14, 2019.
- 20.Cohen JB, Tewksbury CM, Torres Landa S, et al. National Postoperative Bariatric Surgery Outcomes in Patients with Chronic Kidney Disease and End-Stage Kidney Disease. Obesity surgery. 2019;29(3):975–982. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Meersch M, Schmidt C, Zarbock A. Patient with chronic renal failure undergoing surgery. Current opinion in anaesthesiology. 2016;29(3):413–420. [DOI] [PubMed] [Google Scholar]
- 22.Trainor D, Borthwick E, Ferguson A. Perioperative management of the hemodialysis patient. Seminars in dialysis. 2011;24(3):314–326. [DOI] [PubMed] [Google Scholar]
- 23.Craig RG, Hunter JM. Recent developments in the perioperative management of adult patients with chronic kidney disease. British journal of anaesthesia. 2008;101(3):296–310. [DOI] [PubMed] [Google Scholar]
- 24.Kanda H, Hirasaki Y, Iida T, et al. Perioperative Management of Patients With End-Stage Renal Disease. Journal of cardiothoracic and vascular anesthesia. 2017;31(6):2251–2267. [DOI] [PubMed] [Google Scholar]
- 25.Karambelkar A, Kasekar R, Palevsky PM. Perioperative Pharmacologic Management of Patients with End Stage Renal Disease. Seminars in dialysis. 2015;28(4):392–396. [DOI] [PubMed] [Google Scholar]
- 26.Salminen P, Helmio M, Ovaska J, et al. Effect of Laparoscopic Sleeve Gastrectomy vs Laparoscopic Roux-en-Y Gastric Bypass on Weight Loss at 5 Years Among Patients With Morbid Obesity: The SLEEVEPASS Randomized Clinical Trial. Jama. 2018;319(3):241–254. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Peterli R, Wolnerhanssen BK, Peters T, et al. Effect of Laparoscopic Sleeve Gastrectomy vs Laparoscopic Roux-en-Y Gastric Bypass on Weight Loss in Patients With Morbid Obesity: The SM-BOSS Randomized Clinical Trial. Jama. 2018;319(3):255–265. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Catheline JM, Fysekidis M, Bendacha Y, et al. Prospective, multicentric, comparative study between sleeve gastrectomy and Roux-en-Y gastric bypass, 277 patients, 3 years follow-up. Journal of visceral surgery. 2019. [DOI] [PubMed] [Google Scholar]
- 29.Inadomi M, Iyengar R, Fischer I, et al. Effect of patient-reported smoking status on short-term bariatric surgery outcomes. Surgical endoscopy. 2018;32(2):720–726. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Spaniolas K, Yang J, Crowley S, et al. Association of Long-term Anastomotic Ulceration After Roux-en-Y Gastric Bypass With Tobacco Smoking. JAMA surgery. 2018;153(9):862–864. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Hefler J, Dang J, Modasi A, et al. Effects of Chronic Corticosteroid and Immunosuppressant Use in Patients Undergoing Bariatric Surgery. Obesity surgery. 2019. Epub ahead of print. [DOI] [PubMed] [Google Scholar]
- 32.Joseph B, Zeeshan M, Sakran JV, et al. Nationwide Analysis of Resuscitative Endovascular Balloon Occlusion of the Aorta in Civilian Trauma. JAMA surgery. 2019;154(6):500–508. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Ding P, VanderWeele TJ. Sensitivity Analysis Without Assumptions. Epidemiology (Cambridge, Mass). 2016;27(3):368–377. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Mathur MB, Ding P, Riddell CA, et al. Web Site and R Package for Computing E-values. Epidemiology (Cambridge, Mass). 2018;29(5):e45–e47. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Haneuse S, VanderWeele TJ, Arterburn D. Using the E-Value to Assess the Potential Effect of Unmeasured Confounding in Observational Studies. Jama. 2019;321(6):602–603. [DOI] [PubMed] [Google Scholar]
- 36.Austin PC. An Introduction to Propensity Score Methods for Reducing the Effects of Confounding in Observational Studies. Multivariate behavioral research. 2011;46(3):399–424. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Yu EW, Kim SC, Sturgeon DJ, et al. Fracture Risk After Roux-en-Y Gastric Bypass vs Adjustable Gastric Banding Among Medicare Beneficiaries. JAMA surgery. 2019. Epub ahead of print. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Herrington WG, Smith M, Bankhead C, et al. Body-mass index and risk of advanced chronic kidney disease: Prospective analyses from a primary care cohort of 1.4 million adults in England. PLoS One. 2017;12(3):e0173515. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Russo D, Morrone LF, Errichiello C, et al. Impact of BMI on cardiovascular events, renal function, and coronary artery calcification. Blood Purif. 2014;38(1):1–6. [DOI] [PubMed] [Google Scholar]
- 40.Bunnapradist S, Danovitch GM (2007). Evaluation of adult kidney transplant candidates. Am J Kidney Dis. 2007;50(5):890–898. [DOI] [PubMed] [Google Scholar]
- 41.Nor Hanipah Z, Punchai S, Augustin T, et al. Impact of Early Postbariatric Surgery Acute Kidney Injury on Long-Term Renal Function. Obesity surgery. 2018;28(11):3580–3585. [DOI] [PubMed] [Google Scholar]
- 42.Abdullah HR, Tan TP, Vaez M, et al. Predictors of Perioperative Acute Kidney Injury in Obese Patients Undergoing Laparoscopic Bariatric Surgery: a Single-Centre Retrospective Cohort Study. Obes Surg 2016;26(7):1493–1499. [DOI] [PubMed] [Google Scholar]
- 43.Weingarten TN, Gurrieri C, McCaffrey JM, et al. Acute kidney injury following bariatric surgery. Obes Surg 2013;23(1):64–70. [DOI] [PubMed] [Google Scholar]
- 44.Haskins IN, Kudsi J, Hayes K, et al. The effect of resident involvement on bariatric surgical outcomes: an ACS-NSQIP analysis. The Journal of surgical research. 2018;223:224–229. [DOI] [PubMed] [Google Scholar]
- 45.Inker LA, Astor BC, Fox CH, et al. KDOQI US commentary on the 2012 KDIGO clinical practice guideline for the evaluation and management of CKD. Am J Kidney Dis 2014;63(5):713–735. [DOI] [PubMed] [Google Scholar]
- 46.Kidney Disease: Improving Global Outcomes (KDIGO) Acute Kidney Injury Work Group. KDIGO Clinical Practice Guideline for Acute Kidney Injury. Kidney inter, Suppl.2012;2:1–138. [Google Scholar]
- 47.Bellomo R, Ronco C, Kellum JA, et al. Acute renal failure - definition, outcome measures, animal models, fluid therapy and information technology needs: the Second International Consensus Conference of the Acute Dialysis Quality Initiative (ADQI) Group. Crit Care. 2004;8(4):R204–212. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Mehta RL, Kellum JA, Shah SV, et al. Acute Kidney Injury Network: report of an initiative to improve outcomes in acute kidney injury. Crit Care. 2007;11(2):R31. [DOI] [PMC free article] [PubMed] [Google Scholar]