Systematic Review and Meta-analysis of Medium-Term Outcomes After Banded Roux-en-Y Gastric Bypass

Transcription

1 DOI /s REVIEW ARTICLE Systematic Review and Meta-analysis of Medium-Term Outcomes After Banded Roux-en-Y Gastric Bypass H. Buchwald & J. N. Buchwald & T. W. McGlennon # Springer Science+Business Media New York 2014 Abstract This is the first systematic review and meta-analysis focused exclusively on intermediate-term outcomes for the banded Roux-en-Y gastric bypass (B-RYGB). B-RYGB articles published from 1990 to 2013 were identified through MEDLINE, ScienceDirect, and SpringerLink databases augmented by manual reference review. Articles were assigned an evidence level (Centre for Evidence-Based Medicine [Oxford UK] criteria) and Jadad quality score (randomized controlled trials). Simple and weighted means (95 % confidence interval (CI)) for excess weight loss (EWL) at follow-up (1 10+ years) were calculated. At 5 years, a pooled estimate for BMI (kg/m 2 ) change (weighted mean difference and 95 % CI) for banded bypass patients was computed. Rates for weighted mean complications, non-band- and band-related reoperations, and overall comorbidity resolution were calculated. Three hundred twentyone articles were identified: 286 failed inclusion criteria (i.e., non-english, B-RYGB unrelated, <10 per arm, <3-year followup), leaving 35 articles. Manual review added 10 potentially relevant articles; 30 that failed inclusion criteria were excluded, leaving 15 for analysis. B-RYGB was performed on 8,707 patients: 79.0 % female, mean age 38.7, and BMI 47.6 ( ). Overall BMI weighted mean difference (reduction) at 5 years was 17.8 (95 % CI 12.8, 22.7; p<0.001). Five-year weighted mean EWL of 72.5 % (67.5, 77.4) was sustained at 10+ years (69.4 %; 58.9, 80.0). Weighted mean complication H. Buchwald (*) Department of Surgery, University of Minnesota, 420 Delaware Street SE, MMC 290, Minneapolis, MN 55455, USA buchw001@umn.edu J. N. Buchwald Division of Research Writing, Medwrite Medical Communications, Maiden Rock, WI, USA T. W. McGlennon Quantitative Research, McGlennon MotiMetrics, Maiden Rock, WI, USA rates were as follows: early, 10.9 %, and late, 20.0 %. Non-bandrelated reoperation rate was 15.2 %, and band-specific reoperation rate was 4.1 %. Gastric outlet stenosis, band erosion, and band slippage were 2.8, 2.3, and 1.5 %, respectively. Diabetes remitted in 80/95 (84.2 %). By systematic review and metaanalysis, albeit with limited rates of follow-up, B-RYGB appeared to result in significant, sustained excess weight loss of approximately 70.0 % out to 10 years. Keywords Bariatric surgery. Roux-en-Y. Gastric bypass. Banded gastric bypass. Systematic review. Meta-analysis. Diabetes. Randomized controlled trial. Weight. BMI Introduction Amplified by environmental factors, the obesity epidemic persists despite global educational efforts and treatment with diet, exercise, and pharmaceuticals [1 5]. Although minimally invasive treatments are in development to address unmet needs in obesity management [7], metabolic/bariatric surgery remains the most effective intervention, with a record of >50.0 % excess weight loss (EWL) maintained beyond 10 years [7]. Since the introduction by Mason and Ito in 1966 [8] of the loop gastric bypass, subsequently modified to the Roux-en-Y gastric bypass (RYGB), this procedure has proved the most consistently successful treatment for the greatest number of patients; it is also the most wellresearched bariatric operation, with >7,780 peer-reviewed publications [9]. Yet, RYGB patients may contend with significant complications and intermediate-term weight regain is a serious concern. A growing body of studies of the RYGB variant, vertical banded RYGB (B-RYGB), suggests that its weight loss rivals and may surpass that of RYGB over the longer term while presenting a comparable complication profile.

2 Several surgical modifications to improve outcomes have been suggested over the years. During the late 1970s and early 1980s, to address eventual dilation of the gastric bypass pouch and stoma, the technique of reinforcing the stoma with a mesh, band, or ring was introduced in the vertical banded gastroplasty (VBG) and banded gastric bypass variants of Mason, Eckhout et al., Torres and Oca, Laws and Piantadosi, and Salmon [10 14]. In the 1980s and early 1990s, Linner and Drew [15], Fobi et al. [16, 17], and Howard et al. [18] improved outcomes by completely transecting the stomach to avoid staple-line disruption. In 1992, Fobi and Lee reported a B-RYGB-specific technique that incorporated both gastric transection and interposition of the jejunal Roux limb between the gastric pouch and bypassed remnant, which they believed would reduce the high morbidity and reoperation rate associated with marginal ulceration and outlet stenosis (Fig. 1) [19]. This speculation was confirmed by the consecutive B-RYGB series of Capella and Capella, in which 489 initial patients sustained 64 staple-line disruptions, and the last 134, employing gastric transection and jejunal interposition, had no disruption or fistula [20]. The B-RYGB technique as refined by Fobi et al. and Capella and Capella has been the dominant approach since the mid-1990s. In 1999, Fisher and Barber suggested that adding the band to the gastric bypass results in more weight loss for more patients that is maintained over a longer period of time [21]. In 2013, the third questionnaire-based review of Buchwald et al. of the status of metabolic/bariatric surgery worldwide estimated that 46.6 % of 340,768 operations performed annually were those of RYGB and its variants, although B-RYGB as a unique procedure was absent from this global snapshot [22]. In addition, most studies of this procedure in the surgical literature have reported B-RYGB outcomes as findings for RYGB, often rendering B-RYGB s record of safety and effectiveness indistinguishable from that of RYGB. Yet, while new bariatric procedures come and go, the banded bypass has been performed since the mid-1980s and is one of few bariatric operations for which there are >10- year peer-reviewed safety and effectiveness data to be examined. In 2013, there is as yet no consensus on whether adding a band or ring to RYGB is preferable to RYGB alone [23]. The goal of this study was to review all research reports from 1990 to the present that pertain to the vertical B-RYGB procedure and to analyze the available intermediate- to long-term weight, complications, and comorbidity outcomes. Methods Inclusion Criteria and Search Strategy Studies were included if they were full-length research articles in the English language published after January 1, 1990 that reported weight and complication findings in study arms of 10 adults with follow-up of 3 years who underwent any variant of the vertical B-RYGB. For multiple publications in the same year or kin studies, the most recent study that fulfilled inclusion criteria was incorporated. The National Library of Medicine PubMed /MEDLINE, ScienceDirect, and SpringerLink databases were searched electronically for relevant records. Our systematic review strategy observed the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [24] in the identification, screening, and determination of eligible records. In the initial search stage, articles that met the broad inclusion criteria (i.e., published between 1990 and November 1, 2013; duplicates eliminated) were discovered through Boolean keyword assays of the databases using combinations of 19 terms that denote variants of the studied procedure: banded Roux-en-Y gastric bypass, banded bypass, banded gastric bypass, vertical banded bypass, vertical banded gastroplasty gastric bypass, silastic ring vertical gastric bypass, silastic ring Rouxen-Y gastric bypass, silicone ring bypass, stapled silastic ring vertical gastric bypass, transected silastic ring vertical gastric bypass, and relevant acronyms, including BGBP, BRGB, BGB, B-RYGBP, VBG-RGB, SRVGBP, SSRVGBP, SR-RYGBP, and TSRVGBP. In the second search stage, titles and abstracts of identified records were screened. Articles with any combination of the following were excluded: non-english language, contents unrelated to vertical banded bypass, non-rygb-related, banded bypass study arm N<10, <3-year follow-up, nonadult, comments or letters to the editor, review papers, lengthy non-research commentaries, guidelines, and abstracts or case studies. The third evaluation stage entailed reading full articles and their reference sections to identify RYGB-focused studies with possible vertical banded bypass subgroups not captured in the primary literature search; articles that met inclusion criteria were added to the list of potentially eligible records. In the final review stage, kin studies (i.e., those containing data that overlapped with an included study), reports of nonbanded RYGB procedures, articles without usable weight data, and those with <3-year follow-up were excluded from eligibility for quantitative analysis. Data Extraction and Quality Assessment Two reviewers identified and extracted data that met inclusion criteria according to systematic review of the bariatric/ metabolic surgery literature on B-RYGB. Data were entered into fields comprising a dedicated Excel (Microsoft, Redmond, WA) database. Variables of interest included the following: study characteristics (N [sample size], study design, B-RYGB variant, follow-up time points); patient

3 Fig. 1 a Fobi et al. vertical banded Roux-en-Y gastric bypass (B-RYGB) procedure stapled vertical banded Roux-en-Y gastric bypass [16]; b Capella and Capella variant B-RYGB procedure stapled vertical banded gastroplasty gastric bypass with window [74]; c Fobi and Lee divided B-RYGB procedure transected vertical silastic banded Roux-en-Y gastric bypass with jejunal interposition [19] baseline characteristics (i.e., sex, age, BMI [weight (kg) divided by height (m 2 )]); and weight outcomes, principally, percentage excess weight loss (%EWL), and secondarily, BMI change where possible. Alternative measures of weight loss (e.g., percent total body weight lost) were not recorded with sufficient regularity to be considered for analysis. It should be noted that a recent systematic review has put forward that %EWL, as an outcome measure, may be somewhat biased against higher presurgical BMI patients [25]; however, others have reported that %EWL is a relatively reliable measure across studies, regardless of initial BMI [26]. In light of this information and that %EWL has been historically the most frequently reported weight outcome measure, the present authors believe %EWL to be an appropriate measure for the current analyses. Where studies reported banded bypass data on multiple groups, data were pooled using weighted means and standard deviations to yield summary data. In addition to weight outcomes, data on mortality and complications (e.g., reoperations, gastric outlet stenosis, band erosion, band slippage, solid food intolerance) and comorbidities (type 2 diabetes mellitus [T2DM], dyslipidemia, hypertension, sleep apnea) were extracted and analyzed. The evidence level of articles was assessed according to criteria developed by the Centre for Evidence-Based Medicine, Oxford, UK [27], and the methodological quality (bias assessment) of randomized controlled trials (RCTs) was rated using the Jadad scoringmethod[28]. Data Reporting and Statistical Analysis Data analysis was performed using SPSS software, version 20 (IBM SPSS, Chicago, IL) in conjunction with Comprehensive Meta-Analysis (CMA), version 2.2 (Biostat, Englewood, NJ). Quantitative baseline variables (i.e., age, BMI) were expressed using reported study means and standard deviations (SD). Overall pooled means and ranges were used to summarize variables across studies. Categorical variables were expressed as number and percentage. Quantitative outcome variables (i.e., EWL at time points ranging from 1 to 10+ years postsurgery) were expressed using reported study means and associated sample sizes. In the analysis of EWL, a fixed effects model was implemented using methods adapted from O Brien et al. [7]. Simple pooled EWL means and ranges as well as weighted EWL pooled means and 95 % confidence intervals (CIs) were calculated for each time point. Weighted means were calculated with the following formula using sample size as the weighting factor: x ¼ X n i ¼1 w ix i X n i ¼1 w i At each time point, for the number of studies included in the analysis (ranging from 1 to n, dependent on studies reporting data on EWL), the weighted mean was calculated by summing (Σ) products of all caseweighted study means (w i x i ) and dividing the result by the total weight (w i )(w i, combined N of included studies). Ninety-five percent CIs were calculated as approximations of uncertainty limits surrounding weighted mean estimates. Where follow-up sample size was not specifically reported, a weighting factor estimate was derived, where possible, using properties of the normal distribution or from information provided in the article s tables and figures. Also, where BMI change but no EWL was reported at follow-up, EWL was calculated as ([baseline BMI follow-up BMI] / [baseline BMI 25]) 100. For comparative purposes, a trend line depicting EWL in B-RYGB patients was incorporated within the 1 10-year trend lines that represent EWL findings for B- RYGB, BPD/DS, RYGB, and laparoscopic adjustable gastric banding (LAGB) derived from the integrated 1 10 and 10+year data of O Brien et al.

4 BMI change at the 5-year time point was analyzed by calculating the weighted mean difference and associated 95 % CI under a random effects model; I % was considered indicative of significant heterogeneity and was calculated using the following formula [29]: I 2 ¼ ð½q df Š=QÞx100; where Q is the X 2 statistic and df ¼ degrees of freedom Where included studies provided mean BMI values at baseline and 5-year follow-up but did not report associated SDs or standard errors (SEs), SDs were imputed using methods adapted from Furukawa et al. [30]. Qualitative outcome variables (i.e., complications and comorbidity data) were expressed as number and percentage unless otherwise noted. Weighted mean early and late complication rates, as well as non-band and band-specific reoperation rates, were also calculated. Results Study Characteristics We performed a comprehensive review of the national medical databases (Fig. 2). Three hundred twenty-one records met the broad keyword search with duplicates filtered out. After screening by title and abstract, 286 citations failed inclusion criteria and were disqualified. The remaining 35 articles were screened for RYGB-focused articles that might contain B- RYGB subgroups; 10 such articles were identified, yielding a total of 45 eligible articles. After reading the remaining articles, 30 studies that did not meet all inclusion criteria were excluded, leaving a final set of 15 articles [18, 20, 31 43] for quantitative analysis. In addition, two kin B-RYGB studies that reported more complete weight and complication data than that published in their primary studies in the same years [44, 45] were treated as supplemental material and reported in combination with their related primary studies. Table 1 describes the characteristics of included studies. Articles were published over a 21-year span, between 1992 and The respective country of origin (based on first author s affiliation) and article distribution were as follows: USA 7/15, Brazil 4/15, Mexico 1/15, Chile 1/15, Venezuela 1/15, and New Zealand 1/15. Study designs included 3/15 RCTs (two blinded), 2/15 prospective case control studies, and 10/15 retrospective database reviews. By the Centre for Evidence-Based Medicine (CEBM) scale, there was no level I evidence. Of the three RCTs, according to the 5-point Jadad scale for the three items directly related to the quality of bias reduction measures taken (randomization, blinding, withdrawals), none rated a high score for methodological quality: One study received a score of 2/5 and two studies, 3/5, although it has been suggested that these scores are 1 to 2 points lower than appropriate for surgery trials, which rarely employ double blinding [46]. ThenumberofarticlesthatstudiedthefollowingB-RYGB variants were silastic or polypropylene VBG-RYGB 4/15, silastic ring gastric bypass (stapled [SRGBP]) 3/15, transected silastic ring vertical gastric bypass (TSRVGBP) with jejunal interposition (TSRVGBP+JI) 3/15, mesh-banded RYGB 2/15, fascia-banded RYGB 1/15, SRGBP and transected SRGBP (TSRGBP) 1/15, and TSRVGBP 1/15. All studies reported a minimum of 36 months of follow-up. Patient Characteristics The total number of patients in the included studies that underwent banded bypass of some variant was 8,707. On average, patients were 38.7 years old (range ) and had a mean baseline BMI of 47.6 ( ); 79.0 % were female (Table 2). Weight Loss Overall, banded bypass patients tended to experience immediate and significant weight loss that was sustained over time (1 10+ years). Table 3 details individual study EWL at follow-up and descriptive statistics summarizing all data collected at specific time points. For example, at 1-year postsurgery, nine study groups reported weight loss data; however, only seven provided information on number of patients that were followed to 1 year; thus, the overall pooled 1-year weighted mean EWL of 76.0 % (95 % CI 69.5, 82.5) was based on analysis of results from 1,288 banded bypass patients from seven separate studies. At 2-year postsurgery, the weighted mean EWL for B- RYGB patients was 78.8 % (74.8, 82.8) based on results from 1,587 patients in seven studies. The majority of included studies (80.0 %) reported weight loss data, in some form, at 5-year postsurgery. As a result, the pooled weighted mean EWL estimate may be assumed to be the most accurate accounting of B-RYGB efficacy provided in this report. Based on 1,254 patients in 11 studies, 5-year weighted mean EWL was 72.5 % (67.5, 77.4).

5 Fig. 2 Flow diagram of systematic review of the banded Roux-en-Y gastric bypass (B- RYGB) literature based on the Preferred Reporting Items for Systematic Reviews and Meta- Analyses (PRISMA) statement White et al. [35] provided the longest follow-up of B- RYGB weight loss data beyond 10 years, reporting 58.5 % EWL at 14 years. This value was incorporated into calculation of the 10+-year weighted mean (69.4 %; 58.9, 80.0). B-RYGB weight loss data presented in Table 3 are compared in Fig. 3 to mid- and longer-term weight loss trends for multiple bariatric procedures (RYGB, BPD/DS, LAGB) [7, 47]. At year 1, B- RYGB, RYGB, and BPD/DS achieved similarly high degrees of EWL approximately 70.0±5.0 %, while LAGB achieved a significantly lower EWL of approximately 45.0 %. Between years 1 and 2, each procedure resulted in increased weight loss; however, after 2 years, while B-RYGB, BPD/DS, and LAGB weight loss continued to increase or remain relatively constant, RYGB declined significantly. The weight loss of LAGB continued to increase while that of RYGB continued to decrease until their trend lines intersected at approximately 55.0 % EWL at 7.5 years. At 10+ years, there is virtually no distinction in weight loss efficacy between LAGB and RYGB. In contrast, B-RYGB EWL held relatively constant at approximately 70.0 %, comparable to BPD/DS at each time point up to 10+ years. Although a marked relative difference in 10+-year weight outcomes favoring B-RYGB over RYGB (69.4 vs 53.5 % EWL) seems apparent, a statistically significant difference could not be confirmed. Awide 95 % CI encompassed the current study s B-RYGB EWL (69.4 %; 58.9, 80.0); the somewhat small number of patients (n=202) entered into the variance calculation greatly increases the likelihood of confidence interval overlap. Two studies provided weight outcomes through 5-year follow-up for revisional B-RYGB (bariatric procedure converted to a banded bypass) (i.e., Drew et al. [31]; Capella et al. [32]). Although revisional B-RYGB did not achieve equivalent weight loss to that of primary B-RYGB (Fig. 4), significant differences were not observed at any time point. Also, revisional B-RYGB EWL was comparable to that of primary RYGB at 2 years (70.0 vs 68.0 %) (Figs. 3 and 4) and substantially greater than primary RYGB at 5 years (65.0 vs 58.0 %). Though no firm conclusions can be drawn due to limited sample size in the B-RYGB revision group (n=160), these results suggest that B-RYGB, even as a revisional procedure, does not result in the same degree of weight regain as primary RYGB between 3- and 5-year postsurgery. Finally, as 80.0 % of included studies reported 5-year weight loss data, the effect of B-RYGB on BMI was analyzed

6 Table 1 Characteristics of included studies Study author Country a Number Design B-RYGB variant Follow-up mean b or median b (range) CEBM b level of evidence Jadad score c RCTs Drew et al. [31] USA 461 Retrospective database study Fascia-banded RYGB 51 months (36 72) IV Howard et al. [18] USA 20 RCT, double-blind, single-center Marlex mesh-banded RYGB (12 78) II 2 Capella et al. [20] USA 560 Retrospective, database study Silastic or polypropylene VBG-RYGB (30 66) III Capella et al. [32] USA 60 Retrospective, database study Silastic or polypropylene VBG-RYGB (12 60) IV Capella et al. [33] USA 652 Retrospective, database study Silastic or polypropylene VBG-RYGB (60) IV Fobi et al. [34] USA 576 Retrospective, database study TSRVGBP (3 108) III White et al. [35] New Zealand 342 Retrospective, database study SRGBP and TSRGBP (12 168) IV Pajeckietal.[36] Brazil 130 Retrospective, database study SRGBP 87 (24 96) IV Bessler et al. [37] USA 46 RCT, double-blind, single-center Polypropylene VBG-RYGB (6 36) II 3 Magro et al. [38] Brazil 782 Prospective case control series SRGBP (18 60) IV Salinas et al. [39] Venezuela 160 Retrospective, database study TSRVGBP+JI (60) IV Rasera et al. [40] Brazil 75 Retrospective, database study SRGBP (12 48) IV Awad et al. [41] Chile 260 Retrospective, database study TSRVGBP+JI (12 120) III Zarate et al. [42] Mexico 30 RCT, non-blinded, single-center Mesh-banded RYGB (60) II 3 Valezi et al. [43] Brazil 211 Prospective case control series TSRVGBP+JI (12 120) IV World Health Organization obesity weight class by body mass index (BMI): I (30.0 <35), II ( 35.0 <40), and III ( 40.0) RCT randomized controlled trial, nrct non-randomized controlled trial, B-RYGB banded Roux-en-Y gastric bypass, VBG vertical banded gastroplasty, SRGBP silastic ring gastric bypass (stapled), TSRGBP transected SRGBP, TSRVGBP transected silastic ring vertical gastric bypass, TSRVGBP+JI transected silastic ring vertical gastric bypass with jejunal interposition a Country of first author s affiliation b Level of evidence based on the Centre for Evidence-Based Medicine, Oxford, UK. Level I=high-quality RCT with statistically significant difference or no difference but narrow confidence intervals or systematic review of level 1 RCT. Level 2=lesser quality RCT (<80 % follow-up, no blinding, or improper randomization), or prospective comparative study, or systematic review of level II studies or level I studies with inconsistent results. Level III=case-controlled study, or retrospective comparative study, or systematic review of level III studies. Level IV=case series. Level V=expert opinion [27] c Jadad score for RCTs based on 5-point scale, 5=least biased/highest methodological quality [28]

7 Table 2 Characteristicsofbanded Roux-en-Y gastric (B-RYGB) bypass patients Study author Female N (%) Age mean±sd, years (range) Body mass index mean±sd, kg/m 2 (range) Drew et al. [31] 416 (90.2) 36.0± ±7.6 Howard et al. [18] 15 (75.0) 38.1± ±7.8 Capella et al. [20] 453 (81.0) ±9.0 Capella et al. [32] 43.0± ±9.0 Capella et al. [33] 515 (79.0) 38.0± ±10 Fobi et al. [34] 501 (87.0) ±7.8 White et al. [35] 261 (76.0) ±9.2 Pajecki et al. [36] 99 (76.0) 56.7±10.6 Bessler et al. [37] 26 (56.5) 40.6± ±7.3 Magro et al. [38] 568 (72.6) 37.5± ±4.4 Salinas et al. [39] 121 (76.0) 33.2± ±9.3 Rasera et al. [40] 75 (100.0) 34.8 Awad et al. [41] 41.0±7.8 Zarate et al. [42] 27 (90.0) 37.8± ±5.0 Valezi et al. [43] 101 (75.4) 43.4± ±5.1 Overall N (%) 3,178 (79.0) Mean Range by calculating a pooled estimate (weighted mean difference) of overall BMI reduction for studies reporting baseline and 5- year BMI data (Fig. 5). The weighted mean difference in 5- year BMI for 1,254 B-RYGB patients (11 studies) was 17.8 (95 % CI 12.9, 22.7; significance of overall effect p<0.001, I 2 =88.9 %). In addition, significant heterogeneity was found between independent pooled BMI weighted mean difference estimates for the high BMI group (baseline BMI 50) vs low BMI group (baseline BMI<50). As is typically the case with patient groups with higher baseline BMI (e.g., superobese [BMI 50], super-superobese [BMI 60]), B-RYGB patients (n=156) comprising the three studies reporting a baseline BMI 50 lost significantly more total body weight (weighted mean difference=20.4 kg/m 2 [17.9, 22.9]) at 5 years than those patients (n=1,098) comprising the eight studies reporting a baseline BMI<50 (weighted mean difference=15.4 kg/m 2 [14.0, 16.8]; Q=11.9, Qp<0.001). Complications Complication rates, both early and late, differed markedly across studies (Table 4). For example, of those studies reporting on early complications, individual study rates reported ranged from 0.0 % (Valezi et al. [43] and Zarate et al. [42]) to 26.0 % (Bessler et al. [37]). The overall weighted mean rate of early complications was 10.9 %. For those studies reporting late complications, individual study rates ranged from 0.0 % (Pajecki et al. [36] and Awad et al. [41]) to 79.0 % (Bessler et al. [37]). The overall weighted mean rate of late complications was 20.0 %. The choice of which complications to report seemed unique to each set of researchers rather than dependent on a standardized guideline for inclusion and assessment. The five most frequently reported complications (regardless of intraoperative or postoperative time point) were as follows: mortality (80.0 % of studies), leak (67.0 %), gastric outlet stenosis (53.0 %), pulmonary embolism (47.0 %), and small bowel obstruction (47.0 %). The non-band-related reoperation rate was 15.2 % (1,325/8,707) and the band-related reoperation rate was 4.1 % (353/8,707). Comorbidities A summary of the four studies reporting on comorbidity resolution following B-RYGB is presented in Table 5. Resolution rates for T2DM ranged from 75.0 to 92.0 %, with an overall rate of 84.2 % (80/95). Dyslipidemia resolution rates ranged from 33.6 to 76.7 %, with an overall rate of 39.8 % (129/324). Respective overall resolution rates for obstructive sleep apnea and hypertension were 91.4 % (32/ 35) and 58.0 % (122/212). Due to the relatively few studies included in the current comorbidity analysis, reliable estimates of resolution were not possible; those given must be interpreted with caution.

8 Table 3 Banded Roux-en-Y gastric bypass (B-RYGB) percentage excess weight loss (%EWL) Study author Baseline N Follow-up (years) %(n) Drew et al. [31] N=461 (180) Howard et al. [18] N=20 (20) (6) Capella et al. [20] N=560 (234) (152) (65) Capella et al. [32] N=60 (36) (31) (15) (16) (14) Capella et al. [33] 77.0 N=652 (72) Fobi et al. [34] N=576 (514) (475) (450) (401) (354) (327) (299) (194) (64) White et al. [35] [58.5]* N=342 (265) (203) (166) (115) (72) (55) (35) (37) (24) (26) [26.0]* Pajecki et al. [36] N=130 (75) (19) (19) (19) (19) Bessler et al. [37] N=46 Magro et al. [38] N=782 (594) (363) Salinas et al. [39] 83.0 N=160 (134) Rasera et al. [40] N=75 (75) (75) (75) (49) Awad et al. [41] N=260 (244) (169) (111) (78) (51) (34) Zarate et al. [42] 61.6 N=30 (21) Valezi et al. [43] N=211 (134) (134) (134) (134) (116) Simple mean Range Sum of N 1,288 1,587 1, , No. studies included in meta-analysis Weighted mean % CI 69.5, , , , , , , , , , 80.0 Note: Although total number of patients undergoing bypass variants was 8,707, total baseline n for weight data=4,365 Note: White et al. had 10- and 14-year data: The 14-year data have been integrated into the calculation of the 10+-year values (mean %EWL at 14 years= 58.5; n=26 used in meta-analysis to give equal weight relative to 10-year data of White et al.). Also, White et al. reported mean %EWL data at 1, 2, 5, 10, and 14 years; White et al. 3-, 4-, 6-, 7-, 8-, and 9-year data were derived from O Brien et al. [7] Discussion The banded bypass has been performed for more than a quarter of a century, although its peer-reviewed studies have been overshadowed by the plenitude of reports about its progenitor, RYGB. The current meta-analysis and trend line comparisons suggest that B-RYGB s intermediate-term weight loss outcomes may be superior to those of RYGB. Whereas RYGB and bariatric surgery patients, on the whole, reach their BMI nadir approximately 2 years postsurgery and

9 Fig. 3 Excess weight loss (%) in banded Roux-en-Y gastric bypass (B-RYGB) compared with Rouxen-Y gastric bypass (RYGB), biliopancreatic diversion/ duodenal switch (BPD/DS), and laparoscopic adjustable gastric banding (LAGB). RYGB, BPD/ DS, and LAGB data were from O Brien et al. [7] find that their subsequent weight loss decreases [7], this pattern was not seen in the current analysis of B-RYGB. From the available data, albeit derived from studies with limited follow-up, the band seems to maintain the weight loss achieved by the initial bypass well after the 2-year mark. Weight Loss In O Brien et al. s systematic review and meta-analysis of medium-term weight loss after bariatric surgery in cohorts of Fig. 4 Excess weight loss (%) for revisional versus primary banded Roux-en-Y gastric bypass procedures 100, they observed that B-RYGB appeared to retain effectiveness better than RYGB, with the EWL being significantly greater at 5 years for this group (74.7 vs 58.2; three studies, n=498; p<0.05) [7]. In addition, O Brien et al. urged that more data were needed to further explore the potential superiority of B-RYGB. Our study has attempted to provide such data by incorporating into the analysis all studies reporting B-RYGB weight loss beyond 3 years as long as they were composed of cohorts of 10. The current analysis found a 5-year EWL of 72.5 % (95 % CI 67.5, 77.4; 11 studies, n=1,254) for banded bypass patients, essentially equivalent to that reported by O Brien et al. (74.7 %). Overall BMI reduction at 5 years in B-RYGB patients was 17.8 (12.9, 22.7), consonant with Buchwald et al. s metaanalytic finding for BMI reduction at 2 years of 16.7 (14.9, 18.4) for all bypass patients (n=2,705, 22 studies) [48]. Also, the reviewed B-RYGB patients demonstrated a relatively constant EWL of approximately 70.0 % to 10+ years, rivaling weight outcomes associated with BPD/DS (Fig. 3). In Buchwald et al. s 2004 bariatric surgery meta-analysis, with outcome time point values reflecting at least 50.0 % follow-up, EWL for BPD/DS was 70.1 % (66.3, 73.9) and for RYGB, 61.6 % (56.7, 66.5). It has been observed that typical maximal weight loss following RYGB is obtained at 1.9 years in morbidly obese patients and 2.2 years in the superobese [49]. Thereafter, approximately % of RYGB patients will experience weight regain or fail to lose adequate weight [50 52]. Pories et al. reported a 1-year mean EWL of % for RYGB patients and a subsequent, significantly lower, EWL of 66.0 % at 5 14-year follow-up [53].

10 Group by BMI Level Study name Statistics for each study Difference in means and 95%CI Difference Standard Lower Upper in means error limit limit Z-Value p-value High Capella et al (1) High Capella et al (3) High Pajecki et al High Low Howard et al Low Capella et al (2) Low White et al Low Fobi et al Low Magro et al Low Salinas et al Low Valezi et al Low Zarate et al Low Overall Fig. 5 The table and corresponding forest plot summarize meta-analysis of body mass index (BMI) change from baseline to 5 years in banded Roux-en-Y gastric bypass patients. Each study contributing to the metaanalysis is represented by a single blackened square on the forest plot,its size proportional to the amount of weight the study was given during calculation of the pooled summary estimate (expressed as the weighted mean difference [WMD] represented by the diamond shape at the base of the forest plot). Two additional diamonds represent independent summary estimates for high BMI patients (baseline 50) vs low BMI patients (baseline <50). WMD values >0, with 95 % CIs (i.e., box whiskers) not intersecting the reference line (vertical line above 0 point), are indicative of a significant treatment effect for banded bypass Christou et al. demonstrated that both morbidly obese and superobese RYGB patients experienced significant weight regain from the nadir to 5 years and again from 5 to 10 years [49]. In the current study, although B-RYGB studies analyzed did not have the comparable extensive follow-up rate of Christou et al., the evidence available suggests that B-RYGB patients did not appear to experience the same pattern of weight regain. For example, there was no significant difference between the pooled mean EWL of 74.8 % (representing combined 1 5-year follow-up data) and the pooled mean EWL of 72.3 % (combined 6 10-year follow-up) of the B- RYGB patients analyzed. Rasera et al., Magro et al., and other researchers have observed weight regain primarily between 2 and 5 years for both banded and non-banded bypass patients [38, 40, 49]. However, when weight regain occurs in B- RYGB patients, it appears to do so at a lower rate (about 8.0 % of nadir weight) [38]relativetoweightregainassociated with non-banded RYGB patients (17.0 to 22.0 % of nadir) [49]. Some researchers report that weight regain following RYGBmaybeashighas50.0%ofweightlost[54]. Interestingly, it may be that the most substantial amount of weight regain in B-RYGB patients occurs when the restrictive band must be removed due to complications such as severe food intolerance [55]. Weight regain also contributes to the overall failure rate of a particular bariatric procedure, as defined by percentage of patients that do not achieve or maintain 50.0 % EWL at follow-up. The long-term failure rate for RYGB has been calculated as approximately 20.0 % [49]. This rate can be as high as % for superobese patients undergoing RYGB [56]. In contrast, in Magro et al. s study of banded patients, the failure rate for the superobese was 18.8 vs 11.0 % for the morbidly obese group [38]. Capella et al. and Fobi et al. reported a failure rate of approximately 7.0 % at 5-year follow-up [20, 34], and Awad et al. reported an 8.0 % failure rate at 10-year follow-up [41]. Although there have been few comparative trials of B- RYGB vs RYGB weight outcomes and failure rates, five such studies were identified in our preliminary research, four of which were included in the analysis. One study compared banded vs non-banded bypass outcomes at 3 years and, thus, did not meet our study s inclusion criteria. In 2013, Heneghan et al. reported a matched cohort analysis of patients (n=268, 134/134) who had undergone banded vs non-banded LRYGB with 24-month EWL and BMI reduction as primary endpoints and secondary endpoints including mortality, morbidity, and band-related complications [57]. The authors found that B-RYGB patients achieved significantly greater EWL at 2 years compared to non-banded patients (58.6 vs 51.4 %, p<0.05). Further, and perhaps most interesting, subgroup analysis found the band effect on weight loss localized principally in superobese patients who demonstrated a 10.0 % greater EWL after B-RYGB than after RYGB (57.5 vs 47.6 %, p<0.01; BMI reduction 18.0 vs 15.2, p<0.01). Although the banded bypass group with baseline BMI<50 also achieved a greater EWL than their non-banded counterparts (62.9 vs 57.9 %), the difference was not statistically significant (p=0.41). No significant differences in mortality or morbidity were found; there was a 2.2 % band-related complication rate. Among the four comparative studies in the current analysis, Bessler et al. carried out a prospective randomized trial of

11 Table 4 Banded Roux-en-Y gastric bypass (B-RYGB) complications, by percentage of patients Complication Drew [31] N=461 Howard [18] N=20 Capella [20] N=623 Capella [32] N=60 Capella [33] N=652 White [35] N=342 Fobi [34] N=3,632 Salinas [45] a N=1,383 E/L E/L E/L E/L E/L E/L E/L E/L E/L E/L E/L E/L E/L E/L E/L Pajecki [36] N=130 Bessler [37] N=46 Magro [38] N=782 Rasera [40] N=75 Awad [41] N=260 Zarate [42] N=30 Valezi [43] N=211 Total no. (%) Anemia / / / / / / / 00/17.1 / / / / / / / 237 (17.1) Bleeding 0.2/00 / 0.3/00 1.7/00 0.2/0.9 / / 1.4/00 / / / / / / / 31 (1.0) Deep vein thrombosis / / / / / / 2.5/00 / / / / / / / / 91 (2.5) Dumping syndrome / / / / / / / / / / / / / / / 0 (0.0) Fistula / / / / / / / 00/5.6 / / / / / / 00/ (5.0) Food intolerance / / / / / / 00/1.7 / / 00/79.0 / / / 00/16.7 / 103 (2.8) Gastric outlet stenosis 0.7/00 00/5.0 00/0.3 / 0.5/1.4 00/ /00 0.4/4.8 / / / / 0.4/00 / / 176 (2.4) Incisional hernia 00/6.7 / / / 00/4.0 00/6.7 00/5.8 00/13.7 / / / / / / / 480 (7.4) Leak 0.7/00 00/00 0.2/00 00/00 00/00 00/00 1.6/00 1.7/00 3.8/00 00/00 / / / / / 90 (1.2) Pulmonary embolism 00/00 00/00 / / 0.5/00 0.3/00 0.8/00 0.4/00 / 00/00 / / / / / 39 (0.6) Ring/band erosion / 00/5.0 / / / 00/0.6 00/2.5 / / 00/00 / / / / / 94 (2.3) Ring/band removal / / / / 00/0.2 00/6.7 / / / 00/00 / / / 00/3.3 00/ (2.3) Ring/band slippage / / 00/0.3 / 00/00 / / 00/2.6 / 00/00 / / 1.1/00 / 00/ (1.5) Small bowel obstruction 0.2/1.7 / / 1.7/00 0.5/00 00/0.6 00/ /1.4 / 2.1/00 / / / / / 143 (2.2) Staple line disruption / 00/00 00/ /5.0 / 00/9.6 / / / / / / / / / 100 (9.5) Stoma dilation 0.2/00 / 00/0.3 / / / / / / / / / / / / 3 (0.3) Ulceration 00/0.4 00/ /0.3 / 00/0.5 00/6.0 00/1.6 / / / / / / / / 92 (1.6) Wound related 8.4/00 5.0/00 / / 0.6/00 / 8.8/00 3.4/00 / 15.2/00 / / / / / 418 (6.8) Death 0.2/00 00/00 00/00 00/00 0.3/00 00/00 0.4/00 0.2/00 3.1/00 00/00 / / / 00/00 00/00 26 (0.3) Other 0.4/4.8 00/ / /00 4.0/00 8.8/15.8 / 0.5/0.1 / 8.7/00 / / / 00/6.7 / 159 (4.4) Individual study rates 10.9/ / / / / / / / / /79.0 / / 1.5/00 00/ /5.7 a,b,c Non-band-related reoperation rate=1,325 / 8,707=15.2 %; band-related reoperation rate=353 / 8,707=4.1 % E early complications, L late complications, / not reported a Salinas [45] data is supplemental to Salinas [39] for complication data only b Weighted mean early complication rate=10.9 % c Weighted mean late complication rate=20.0 %

12 Table 5 Banded Roux-en-Y gastric bypass (B-RYGB) comorbidities resolved Study author Type 2 diabetes, n (%) Dyslipidemia, n (%) Obstructive sleep apnea, n (%) Hypertension, n (%) White et al. [35] 53/62 (85.5) 89/265 (33.6) 85/138 (61.6) Pajecki et al. [36] 13/17 (75.6) 23/30 (76.7) 29/31 (93.5) 19/51 (37.3) Bessler et al. [37] 11/12 (92.0) 7/14 (50.0) 18/23 (79.0) Awad et al. [41] 3/4 (75.0) 10/15 (66.0) 3/4 (75.0) Total 80/95 (84.2) 129/324 (39.8) 32/35 (91.4) 122/212 (58.0) superobese patients (n=90, 46/44) [37]. Differences in EWL were analyzed along with complication rates and improvements in comorbidities. Banded patients experienced consistently greater EWL at 6 months and 1 and 2 years (43.1 vs 24.7, 64.0 vs 57.4, and 64.2 vs 57.2 %, respectively), although differences did not reach statistical significance. There was no significant difference in complication rates between banded (26.0 %) vs non-banded (29.5 %) patients and no difference in rates of comorbidity improvement. However, at 3 years, banded patients had achieved a significantly greater EWL than non-banded bypass patients (73.4 vs 57.7 %, p<0.05), consistent with the observation in the current analysis that the advantages of the banded bypass become apparent at 2 3 years postsurgery, when stoma dilation is likely to occur in non-banded patients, potentially causing early emptying of the gastric pouch and increased food consumption [44, 58, 59]. The results of Bessler et al. and Heneghan et al. dovetail, both showing the band being effective in promoting marked weight loss in the superobese (typically, the patients with the highest failure rates following bariatric surgery) [60]. The current study results are in-line with and extend both Heneghan et al. s (2 years) and Bessler et al. s (3 years) findings for B-RYGB in superobese patients, in that the three included studies with mean baseline BMI 50 at 5-year follow-up yielded a weighted mean BMI reduction of 20.4 kg/m 2 (95 % CI 17.9, 22.9) and a corresponding EWL of 72.9 %; this result is significantly better than the 5-year trend (approximately 58.0 %) found for all patients undergoing RYGB (Fig. 3). The only comparative study included in the present review ( 3-year follow-up) that did not find superior weight loss results for banded bypass patients (Zarate et al. [42]) reported a 5-year randomized trial comparing banded vs non-banded LRYGB (n=60, 30/30) [42]. Although EWL was somewhat greater in banded patients (61.6 vs 59.8 %), there were no statistically significant differences between banded and nonbanded bypass patients in EWL, BMI reduction, or failure rate (percentage of patients not reaching a BMI<35) at 5 years. It should be noted that Zarate et al. s61.6%ewlanchoredthe low end of the range for studies reporting 5-year mean EWL for B-RYGB in the current research ( %). In the included comparative study of Rasera et al., a retrospective analysis of female patients was performed (n=143; 75 with silastic ring, 68 without) in which median EWL was calculated at 12, 24, 36, and 48 months [40]. Banded patients lost more weight at each respective time point: 74.0 vs 72.0, 79.0 vs 76.0, 76.0 vs 72.0, and 76.0 vs 71.0 %; however, it was only after 36 months that the between-group difference in EWL became significant. When surgical failure was defined as EWL<50.0 %, at 4 years, a significantly greater percentage of non-banded patients was classified as failures relative to banded patients (30.9 vs 8.2 %, p<0.01). Interestingly, at no time point did the percentage of non-banded patients with % weight regain differ significantly from their banded counterparts. For example, at 4-year follow-up, well beyond the point at which most researchers have noted RYGB stomal dilation, 44.2 % of non-banded patients vs 30.6 % of banded patients (p= 0.22) experienced >10.0 % weight regain and only 9.6 % of non-banded patients vs 10.2 % of banded patients (p=1.0) experienced >20.0 % weight regain. Therefore, Rasera et al. concluded that their results provided evidence corroborating the superiority of the band with respect to overall weight loss but not with respect to improved control of weight recovery following gastric bypass. In a final included 10-year retrospective comparative study by Awad et al. (n=378; 260 banded vs 118 non-banded), longterm EWL, comorbidity resolution, and quality of life were analyzed [41]. From 2 years onward, banded patients experienced significantly greater EWL. For example, at 3 years, banded vs non-banded patients experienced 77.0 vs 72.4 % EWL; although the difference is significant (p<0.05), it is by no means sizeable. However, at 10 years postsurgery, banded (69.4 % follow-up rate) vs non-banded (70.8 % follow-up) patients achieved 82.3 vs 61.3 % EWL, a considerable as well as significant result (p<0.0001). Again, these results are consistent with the trend lines of the current analysis (Fig. 3), wherein banded bypass patients appear to maintain their weight loss to 10+ years while non-banded patients, on average, do not perform as well. In Awad et al., no significant difference between banded and non-banded groups was seen with respect to comorbidity resolution (i.e., 75.0 vs 50.0 % diabetes resolution, 66.0 vs 50 % dyslipidemia resolution, 75.0 vs 60.0 % sleep apnea resolution) and overall complications.

13 Complications As with all surgery, bariatric surgery carries some risk. The complication profile associated with B-RYGB has not been clearly defined; notwithstanding band-related complications, complication rates for RYGB and B-RYGB should be similar. In a case-matched JAMA study of 442 patients by Romy et al., early and long-term complication rates after RYGB were 17.2 and 19.0 %, respectively, with a reoperation rate of 12.7 %. These numbers are comparable with those presented in the current review of 8,707 B-RYGB procedures, in which the weighted mean early and late complication rates were 10.9 and 20.0 %. The non-band-related reoperation rate was 15.2 % (1,325/8,707) and the band-related reoperation rate was 4.1 % (353/8,707). Comparison of the current study s B-RYGB complications with those of RYGB in the literature yielded interesting results. For example, a significant proportion of RYGB patients experience dumping syndrome, whereas no instances of the syndrome were reported in B-RYGB patients in the included studies. Postoperative leak at the anastomotic site following gastric bypass is a potentially life-threatening complication reported in 0.1 to 20.0 % of patients [61 64]. The current B- RYGB analysis found a leak rate of 1.2 %, which is in the low end of the range reported in the literature for RYGB. In a systematic review and meta-analysis of complications after bariatric surgery in RCTs, Chang et al. found a 3.3 % (1.2, 6.0) rate of small bowel obstruction [64]; whereas, the rate in the current B-RYGB study was 2.2 %. With respect to mortality, bariatric surgery has developed to the point of being considered as safe or safer than most major surgical procedures. A mortality-specific meta-analysis by Buchwald et al. found that among patients undergoing RYGB, the mortality rate was 0.44 % for open and 0.16 % for laparoscopic procedures [65]. The mortality rate for B- RYGB patients in the current review of combined open and laparoscopic outcomes was 0.3 %, a rate at the midpoint of the aforementioned rates. Although this study s B-RYGB complication rates, early and late, are comparable to current reported complication rates for RYGB, estimated B-RYGB rates may be somewhat high because they incorporated results across a span of 23 years wherein improvements in the procedure s technique and the surgical learning curve have occurred. For example, post banded bypass studies included in the current review that employed the evolved Fobi technique (complete gastric transection as well as interposition of the jejunal Roux limb) reported markedly fewer staple line disruptions than those of prior studies. Fobi et al. appear to have carried out the most comprehensive B-RYGB research, reporting outcomes for 3,632 TSRVGBP patients [34]. They found that >90.0 % of patients lost and, more importantly, maintained 50.0 % EWL at 5 years. In addition, they reported a perioperative morbidity of 10.0 %, a late complication rate of <6.0 %, a mortality rate of 0.44 %, and a revision rate of <6.0 %. Fobi et al. also provided complication data for laparoscopic B- RYGB alone: In a sample of 415 laparoscopic TSRVBP patients, there was a 9.4 % early complication rate (i.e., 2.2 % outlet stenosis, 4.2 % gastric leak, 0.5 % small bowel obstruction, 1.2 % postoperative bleeding, 0.7 % marginal ulcer, 0.3 % rhabdomyolysis, 0.3 % deep vein thrombosis, and 0.7 % perioperative mortality). The late complication rate in Fobi et al. s study was 5.3 % (i.e., outlet stenosis 3.1 %, marginal ulcer 0.5 %, small bowel obstruction 0.7 %, chronic sinus 0.2 %, ring erosion 0.5 %, death 0.2 %) [44]. Although fear of band slippage and/or erosion stopping or reversing weight loss has been an issue with banded bypass [66], data presented in the current study suggest that bandrelated complications may not be frequent or severe enough to warrant extreme concern: The rates of band erosion, band slippage, and band removal were 2.3, 1.5, and 2.3 %, respectively. Fobi et al. reported band erosion rates for banded gastric bypass ranging from 1.9 to 5.9 %, depending on whether the B-RYGB was a primary or secondary operation or a revision. Primary B-RYGB had a band erosion rate of 1.9 %, while B-RYGB as a secondary operation had a band erosion rate of 5.9 %. The rate of revisional B-RYGB erosion was 4.5 %. Overall incidence of band erosion in 3,632 patients was 2.5 % [44]. Similarly, Salinas et al. reported a migration rate of 1.7 % in 3,800 banded bypasses and suggested that the low rate of slippage is inconsequential relative to the 98.3 % of patients who typically experience better outcomes with a banded RYGB [39]. Finally, it has been shown by some researchers that significantly greater food intolerance is experienced in banded bypass patients [42] and that it might cause quality of life to be diminished; yet, in the longest-running comparative study (10+ years of follow-up) of banded versus non-banded outcomes, included in the current analysis, Awad et al. reported that even with food intolerance factored into the analysis, there was no significant difference between groups in overall quality of life [41]. It should also be noted that in B- RYGB patients, where food intolerance (without stenosis) is severe enough to warrant some form of intervention, reoperation can be averted with aggressive endoscopic dilation in approximately 97.0 % of patients and weight regain due to band removal avoided [23]. Comorbidities Maintenance of weight loss following bariatric surgery is also critical to maintenance of diabetes resolution and prevention. As found in the Nurses Health Study of 89,941 women, the risk of developing T2DM is >38-fold in morbid obesity (BMI 35) and >20-fold in obesity (BMI ); further, as little as 4.2 kg of weight loss sustained for an average of 3.2 years can reduce the progression to T2DM by