Bariatric surgery: a systematic review and network meta-analysis of randomized trials


  • No reprints will be available.

  • Marcello Tonelli, the Corresponding Author and Guarantor, has the right to grant on behalf of all authors and does grant on behalf of all authors, a worldwide licence to the Publishers and its licensees in perpetuity, in all forms, formats and media (whether known now or created in the future), to (i) publish, reproduce, distribute, display and store the Contribution, (ii) translate the Contribution into other languages, create adaptations, reprints, include within collections and create summaries, extracts and/or, abstracts of the Contribution, (iii) create any other derivative work(s) based on the Contribution, (iv) to exploit all subsidiary rights in the Contribution, (v) the inclusion of electronic links from the Contribution to third party material where-ever it may be located; and, (vi) licence any third party to do any or all of the above.

  • Sources of funding: This project was primarily funded by the Canadian Agency for Drugs and Technology in Health. Additional support for this work was provided through an Interdisciplinary Team Grant from the Alberta Heritage Foundation for Medical Research. Dr Tonelli was supported by a New Investigator Award from the Canadian Institutes of Health Research. Dr Tonelli and Dr Klarenbach were supported by salary awards from the Alberta Heritage Foundation for Medical Research. Dr Tonelli, Dr Padwal, and Dr Klarenbach were supported through an alternative funding plan from the Government of Alberta and the University of Alberta.

M Tonelli, 7-129 CSB, University of Alberta, 8440 112 St NW, Edmonton, Alberta, Canada, T6B 2B7. E-mail:


The clinical efficacy and safety of bariatric surgery trials were systematically reviewed. MEDLINE, EMBASE, CENTRAL were searched to February 2009. A basic PubCrawler alert was run until March 2010. Trial registries, HTA websites and systematic reviews were searched. Manufacturers were contacted. Randomized trials comparing bariatric surgeries and/or standard care were selected. Evidence-based items potentially indicating risk of bias were assessed. Network meta-analysis was performed using Bayesian techniques. Of 1838 citations, 31 RCTs involving 2619 patients (mean age 30–48 y; mean BMI levels 42–58 kg/m2) met eligibility criteria. As compared with standard care, differences in BMI levels from baseline at year 1 (15 trials; 1103 participants) were as follows: jejunoileal bypass [MD: −11.4 kg/m2], mini-gastric bypass [−11.3 kg/m2], biliopancreatic diversion [−11.2 kg/m2], sleeve gastrectomy [−10.1 kg/m2], Roux-en-Y gastric bypass [−9.0 kg/m2], horizontal gastroplasty [−5.0 kg/m2], vertical banded gastroplasty [−6.4 kg/m2], and adjustable gastric banding [−2.4 kg/m2]. Bariatric surgery appears efficacious compared to standard care in reducing BMI. Weight losses are greatest with diversionary procedures, intermediate with diversionary/restrictive procedures, and lowest with those that are purely restrictive. Compared with Roux-en-Y gastric bypass, adjustable gastric banding has lower weight loss efficacy, but also leads to fewer serious adverse effects.


Obesity is a chronic disease that currently affects over 300 million individuals worldwide (1) and leads to substantial morbidity (2), premature mortality (3) and impaired quality of life (4). Obesity is defined by body mass index (BMI), with BMI of 30–34.9, 35–39.9 and 40 kg m−2 or greater corresponding to mild, moderate and severe obesity, respectively (BMI between 18.5–24.9 kg m−2 is considered normal). Severe obesity (the fastest growing obesity subgroup) affecting 2–6% of individuals in Canada, the USA and the UK (5–7), has increased in prevalence in these countries by twofold to fourfold in less than two decades (6,8,9), and is associated with a 13- to 18-fold increased risk of type 2 diabetes compared with individuals of normal weight (2).

Options for treatment of severe obesity include behavioural modification (enhanced physical activity, dietary modification, caloric restriction and psychosocial interventions required to induce behavioural change), medications such as orlistat and bariatric surgery. Because observational studies (10,11) have shown that it is the most effective way of achieving sustained weight reduction in such patients, bariatric surgery has become the preferred therapy in many centres for severely obese patients who are refractory to medical therapy (12). According to current guidelines, patients who fail non-surgical intervention and who have severe obesity (or moderate obesity with a major obesity-related comorbidity) are potentially eligible for surgery (12,13).

Demand for bariatric surgery worldwide is increasing at an exponential rate and nearly 350 000 procedures were performed globally in 2008. Currently, the most frequently performed procedures globally are adjustable gastric banding (adjustable gastric banding [AGB]; 42%), Roux-en-Y gastric bypass (RYGB; 39%) and sleeve gastrectomy (SG; 5%) (14). Although multiple different bariatric procedures are available, few studies directly compare the various procedures and hence there is uncertainty about their relative merits. Similarly, although there have been prior systematic reviews on this topic, none have used a network approach to allow comparisons between all available surgical procedures.

Because developing additional surgical capacity will require allocation of scarce healthcare resources (and choices about which procedures are offered), a comprehensive assessment of the efficacy and complications associated with bariatric surgery would be useful to decision-makers. We performed a comprehensive Health Technology Assessment (HTA) to examine these issues (15). This report details the results of one section of this HTA – a systematic review and network meta-analysis assessing the clinical efficacy and safety of bariatric surgical interventions and standard care in adult patients with severe obesity.


This systematic review was conducted and reported according to established guidelines (16); an a priori protocol was followed throughout. The current manuscript was part of a larger report for the Canadian Agency for Drugs and Technologies in Health reviewing the costs and consequences of bariatric surgery (15).

Data sources and searches

Search strategies created by a MLIS qualified librarian were used in MEDLINE (1950 to 3 February 2009), EMBASE (1980 to 3 February 2009), and CENTRAL (3 February 2009) without language and publication restrictions (see Appendix: Search strategies). The searches were supplemented by hand-searching the reference lists of relevant systematic reviews, trial registries and HTA websites; we also contacted manufacturers of gastric bands. A basic PubCrawler alert was run until March 2010.

Study selection

Randomized controlled trials (RCT) examining the efficacy and safety of any bariatric surgery were potentially eligible for inclusion. Inclusion criteria are summarized below.


Eligible trials studied severely obese adults (≥16 years) with an accepted indication for bariatric surgery: BMI ≥ 40 kg m−2 (or BMI ≥ 35 kg m−2 with at least one obesity-related comorbidity). In trials that did not report inclusion criteria for BMI, we used the following formulations of study-level BMI to approximate National Institutes of Health criteria in order to determine trial eligibility: mean BMI ≥ 43 kg m−2, or mean BMI ≥ 38 kg m−2 with comorbidity; mean excess weight ≥ 72%, or mean excess weight ≥ 52% with comorbidity, and mean weight ≥  113 kg, or mean weight ≥ 99 kg with comorbidity. These approximations were based on an ideal BMI of 25 kg m−2 and the mean height of an average 40-year-old Canadian woman (1.618 m) (17).


Eligible trials studied at least one bariatric surgical intervention such as: gastric banding, gastroplasty, gastric bypass, jejunoileal bypass, biliopancreatic diversion and sleeve gastrectomy. Non-surgical bariatric procedures such as intragastric balloons were excluded.


Eligible trials compared a bariatric surgical intervention as above with standard care (i.e. diet and exercise) or with another bariatric surgery.


Eligible trials reported at least one of: weight change (primary outcome), all-cause mortality, control of comorbidities, hospitalization, reoperations, gastrointestinal disturbances and serious surgical sequelae.

Each record was screened by an academic clinician and one other reviewer. Any trial considered relevant by one reviewer was retrieved for further review. The full text of each potentially relevant article was independently assessed by two reviewers for inclusion using the above selection criteria and a pre-printed form. Disagreements were resolved with a third party through consensus.

Data extraction and quality assessment

One reviewer extracted data from selected trials and a second reviewer checked for accuracy. A statistical reviewer checked the numerical results for accuracy of statistical information. We used a standard data extraction method to record the following properties of each trial into a database: characteristics (country, sample size, duration of follow-up, obesity inclusion criteria, number of participating surgeons); participants (age, gender, smoking status, BMI, comorbidities); bariatric surgical intervention (type, laparoscopic or open); control intervention and outcomes. We captured trial-level data on the following outcomes: change in BMI, all-cause mortality, control of comorbidities (asthma, cardiovascular disease, diabetes mellitus, hypertension, urinary incontinence, sleep apnea, gastroesophageal reflux disease, dyslipidemia, arthritis, mental health), hospitalization (length of post-surgical stay, readmission, number of visits), reoperations (any including revisions of the initial procedure, those indicating failure of a specific bariatric surgery – conversion to a different type of bariatric procedure, and reversals), gastrointestinal disturbances (dumping syndrome, dysphagia, gastritis, reflex, gastric ulcer, and vomiting), and serious surgical sequelae (anastomatic leak, bowel obstruction, hernia, myocardial infarction, respiratory failure, serious bleeding, slippage/dilatation of the band, staple line breakdown, luminal stenosis, venous thromboembolic disease and wound infection).

We assessed the risk of study bias using the following evidence-based criteria: method of allocation concealment (18), randomization technique, double-blinding and description of withdrawals/dropouts (19). Finally, we extracted data on funding sources (20), as well as reports of a priori sample size calculations, interim or preliminary analyses, intention-to-treat designs, and reports of surgical sequelae. Two reviewers independently assessed each study. An initial set of five eligible trials was used to calibrate risk of bias assessment between reviewers. Disagreements were resolved with a third party through consensus.

Data synthesis and analysis

We analyzed data using Stata 10.1 (College Station, TX, USA), R 2.8.1 (, and WinBUGS 1.4.3 (MRC Biostatistics Unit, Cambridge, UK). Results were pooled using the risk ratio or the risk difference (RD) to summarize dichotomous results, and the MD to summarize continuous results. The most frequently reported intervention RYGB was framed as the control group for every meta-analysis wherever possible. For adverse events (mortality, gastrointestinal disturbances and surgical sequelae), we report results at early and late (≤30 days and >30 days postoperatively, respectively) time points. Participants without documented comorbidity at baseline were not included in the analysis of comorbidity resolution and/or improvement.

The primary outcome was change in BMI. Per cent excess weight loss was converted to change in BMI by defining excess as a BMI >25 kg m−2. Single-value imputations were used to approximate missing standard deviations (21). Change in BMI is reported for year 1, year 2 and after year 2 where we selected the result at the latest time point available. Due to the differences expected between trials, we decided a priori to combine results using a random effects model (22). Statistical heterogeneity was quantified using the I2 statistic (23,24), which approximates the per cent of total variation (within- and between-study) due to between-study variation. Publication bias (25) was not assessed and meta-regressions (26) were not explored due to small numbers of trials within any particular comparison.

We also explored the relationship in change in BMI between available surgical intervention (or standard care) using network analysis (27) (specifically, Markov chain Monte Carlo methods within a Bayesian framework). We used non-informative prior distributions: uniform for the between-study variance and Gaussian for the other parameters. Varying the range of the uniform prior did importantly change the results. Network analysis (otherwise known as mixed-treatment comparisons) extends meta-analysis from simply pooling directly compared treatments (direct evidence) to pooling data from studies not directly compared but linked via one or more common comparators (indirect evidence). This technique facilitates the comparison of any two surgical interventions (and standard care) not directly compared in any one study and therefore for the purposes of this report, the ranking of all available bariatric surgical interventions (and standard care). In network analysis, the control group was selected according to lowest ranking of efficacy. We report Bayesian credible intervals for mixed (direct and indirect) and indirect evidence rather than frequentist confidence intervals as per the framework used. Incoherence, where direct and indirect evidence do not agree (beyond chance), was quantified using Bucher's method (28,29).


Quantity of research available

The literature searches resulted in 1838 articles; 223 were retrieved for further scrutiny, of which 31 (30–60) met inclusion criteria (Fig. 1). Disagreements regarding inclusion of studies occurred in 7% of the articles (kappa =  0.84); 11 of 15 disagreements resulted in inclusion.

Figure 1.

PRISMA flowchart of clinical trials included in the systematic review.

Trial characteristics

Characteristics of the included articles are shown in Table 1. The following surgical interventions/standard care were reviewed: gastroplasty (vertical banded gastroplasty (31,32,35,38,40,41,46,48,49,51,54,55,58,60)[VBG], horizontal banded gastroplasty (32)[HBG], horizontal (unbanded) gastroplasty (33,44,45,52,56)[HG]), gastric bypass (RYGB (34,36–38,40,42,44–47,49,52,53,55–58), jejunoileal bypass (30,36,37)[JB] biliopancreatic diversion(50,57)[BPD], mini-gastric bypass (31,47)), adjustable gastric banding (34,35,39,41,43,48,51,53,54,59,60), sleeve gastrectomy (42,43), gastrogastrostomy (38), and standard care (30,33,39,50) (one used a very low calorie diet (33), two used diet ± exercise and the last simply referred to ‘medical management’, which did not appear to include pharmacotherapy). One banding trial (59) included an intervention group whom underwent omentectomy. In 16 trials, surgery was performed using an open approach, in eight trials the approach was laparoscopic, three trials used a combination, and in four trials the approach was unspecified.

Table 1.  Description of included trials Thumbnail image of

The median year of publication was 1997 and median duration of follow-up was 24 months (range from 6 to 60 months). Sample size ranged from 16 to 310 (median 59). The mean age range of trial participants was 30 to 48 years and the proportion of females ranged from 44% to 97%. Mean baseline BMI ranged from 42 to 58 kg m−2. Few trials reported comorbidities among participants.

Items indicating potential risk for bias are shown in Table 2. Although double-blinding would have been difficult if not impossible in many of the trials, the revised upper limit of 3 on the Jadad score was rarely achieved. Except for adequate reporting of adverse events, the trials were poorly conducted.

Table 2.  Quality assessment of included trials
Author, yearAllocation concealmentJadad scoreSample size calculationIntention-to-treatInterim/preliminary analysisAdverse events reportFunding
  1. Not reported.

Nguyen et al., 2009 (53)Adequate1YesYesNoYesPublic
Karamanakos et al., 2008 (42)Adequate5YesYesYesNoNR
Angrisani et al., 2007 (34)Adequate2NoNoNoYesNR
Skroubis et al., 2006 (57)Unclear1NoNoNoYesNR
Langer et al., 2005 (43)Unclear1NoNoNoNoNR
Lee et al., 2005 (47)Adequate2YesNoNoYesNR
Olbers et al., 2005 (55)Unclear3YesNoNoYesPublic
Van Dielen et al., 2005 (60)Unclear3NoNoNoYesPublic
Lee et al., 2004 (46)Adequate1NoYesNoYesNR
Morino et al., 2003 (51)Adequate3YesNoNoYesNR
Mingrone et al., 2002 (50)Unclear1NoNoNoNoNR
Thörne et al., 2002 (59)Unclear1NoYesNoNoPublic
Nilsell et al., 2001 (54)Adequate1NoNoNoYesNR
Ashy and Merdad, 1998 (35)Unclear1NoNoNoYesNR
Heindorff et al., 1997 (39)Unclear1NoNoNoYesNR
Husemann and Kluy, 1997 (41)Inadequate1NoNoNoNoNR
Lundell et al., 1997 (48)Unclear1NoNoNoNoPublic
Howard et al., 1995 (40)Unclear1NoNoNoYesNR
MacLean et al., 1993 (49)Unclear2NoNoNoYesNR
Hall et al., 1990 (38)Unclear2YesNoYesYesMixed
Agren and Naslund, 1989 (31)Unclear1NoNoNoNoNR
Andersen et al., 1987 (32)Unclear3NoNoNoYesPrivate
Sugerman et al., 1987 (58)Adequate2NoNoNoYesNR
Naslund et al., 1986 (52)Adequate2NoNoNoYesNR
Andersen et al., 1984 (33)Adequate2YesNoNoYesPrivate
Pories et al., 1982 (56)Adequate4NoNoNoYesNR
Laws and Piantadosi, 1981 (44)Adequate2NoNoNoYesNR
Lechner and Callender, 1981 (45)Unclear1NoNoNoYesNR
Buckwalter, 1980 (36)Unclear2NoNoNoYesPublic
The Danish Obesity Project, 1979 (30)Unclear1NoNoNoYesNR
Griffen et al., 1977 (37)Inadequate1NoNoNoYesNR

Data analysis and synthesis

Decreases in BMI

Fifteen trials (31,34,36,42,45–47,50–53,55,57,58,60) (n = 1103) reported changes in BMI at 1 year (Matrix 1). Network (mixed-treatment) analysis was used to rank the interventions in order of efficacy for reducing BMI (from most to least efficacious); the probabilities of being the most efficacious intervention are listed in the diagonals of the matrices. In mixed comparisons, mini-gastric bypass, BPD, SG and RYGB all reduced BMI at 1 year to a significantly greater extent than standard care (Matrix 1).

Table 3. Matrix 1 Change in body mass index at 1 year, in kg m−2Thumbnail image of

As compared with standard care, differences in BMI levels from baseline were as follows: JB (MD: −11.4 kg m−2[95% CI: −23.0; 0.8]), mini-gastric bypass (−11.3 kg m−2[−18.6; −4.1]), BPD (−11.2 kg m−2[−15.7; −6.9]), SG (−10.1 kg m−2[−17.8; −2.6]), RYGB (−9.0 kg m−2[−15.1; −3.1]), HG (−5.0 kg m−2[−12.0; 1.8]), VBG (−6.4 kg m−2[−12.9; 0.01]) and AGB (−2.4 kg m−2[−9.1; 3.9]). Therefore, mini-gastric bypass, BPD, SG and RYGB reduced BMI to a significantly greater extent than standard care, and JB and VBG appeared more effective than standard care but the difference was of borderline statistical significance.

Of the bariatric procedures we considered, only BPD was directly compared with standard care (Matrix 1): BPD significantly reduced BMI to a greater extent than standard care. There was no direct evidence comparing standard care with RYGB at 1 year. However, in direct comparisons RYGB reduced BMI to a significantly greater extent compared with HG, VBG and AGB. BPD led to significantly greater reductions in BMI than RYGB. Because there was only one closed loop in the network (AGB – VBG – RYGB), we could calculate only one incoherence statistic, which was non-significant (−2.2 kg m−2[−6.5; 2.1]; P = 0.31).

Standard care groups were not reported in the 11 trials (36,46,47,51–53,55,57–60) (n = 870) that reported changes in BMI at 2 years (Matrix 2a). However, compared with AGB, the lowest ranking surgery, BPD, JB, mini-gastric bypass, RYGB and VBG reduced BMI to a significantly greater extent. After including data from two initially excluded RCTs (61,62), comparing AGB with standard care (direct evidence: MD −3.1 kg m−2 (−4.1; −2.1) at 1 year and −6.1 kg m−2 (−7.6; −4.5) at 2 years), all available bariatric surgeries (BPD, JB, mini-gastric bypass, AGB and omentectomy, RYGB, HG, VBG and AGB) significantly reduced BMI to a greater extent than standard care (Matrix 2b).

Table 4. Matrix 2a Change in body mass index at 2 years, in kg m−2Thumbnail image of
Table 5. Matrix 2b Sensitivity analysis*: change in body mass index at 2 years, in kg m−2Thumbnail image of

Little evidence could be derived from the seven trials (33,34,36,40,51,53,58) (n = 416) that reported changes in BMI after follow-up of 3 to 5 years with respect to standard care (Matrix 3).

Table 6. Matrix 3 Change in body mass index at 3 to 5 years, in kg m−2Thumbnail image of

All-cause mortality and control of comorbidity

Twenty-three trials (30,32,34–38,40,42–47,49,51–54,56–58,60) (n = 2042) and six trials (34,42,52,56,57,60) reported all-cause mortality and resolution and/or improvement in comorbidity, respectively. Bariatric surgery did not significantly reduce mortality risk or the likelihood that comorbidity would resolve (data not shown), although follow-up time was 5 years or less in all trials. To note, the trials measuring comorbidity included few patients with comorbidity thus these outcomes (n = 3 to 100 depending on the comorbidity) were very underpowered.


Eleven trials (34,35,38,45–47,51–53,55,60) (n = 1218) compared length of hospital stay after surgery. Comparisons of AGB and RYGB, and AGB and VBG found significantly shorter lengths-of-stay in AGB participants (MD: −1.7 days [−2.0; −1.3]; two trials (34,53), I2 = 0% and −3.1 days [−5.0; −1.2]; three trials (35,51,60), I2 = 0%, respectively). One comparison (47) of mini-gastric bypass vs. RYGB found significantly shorter lengths-of-stay in mini-gastric bypass participants (−1.4 days [−2.4; −0.4]; one trial); other pooled results were non-significant. Four trials (37,44,46,53) (n = 389) reported the incidence of readmission to hospital following surgery. JB led to a significantly elevated risk of readmission compared with RYGB (risk ratio: 2.96 [1.05; 8.39]; one trial (37)); other pooled results were again non-significant.


Twenty trials (30,31,34,36,37,40,41,45–47,49,51–58,60) (n = 1769) compared the incidence of reoperations between intervention groups. Compared with RYGB, JB had more late reoperations (RD: 28% [6.5; 50]; one trial (37)), and HG and VBG had more total reoperations (29% [6.7; 50]; three trials (45,52,56), I2 = 84%, and 17% [5.0; 30]; two trials (49,58), I2 = 0%, respectively). Twelve trials (34,36,37,49,51–54,56–58,60) (n = 1018) compared the incidence of repeat surgeries for conversions and reversals between intervention groups. Compared with RYGB, JB and AGB had more late conversions/reversal surgeries (32% [9.9; 53]; one trial (37), and 8.3% [2.8; 14]; two trials (34,53), I2 = 0%, respectively), and HG and VBG had more total conversions/reversals surgeries (38% [27; 50]; two trials (52,56), I2 = 0%, and 29% [1.7; 57]; two trials (49,58), I2 = 85%, respectively). Seven trials (34,37,49,51,53,54,57) (n = 696) compared the incidence of surgical reversals between intervention groups. Compared to RYGB, VBG had significantly more reversals (9.3% [0.9; 18]; one trial (49)).

Serious surgical sequelae

Twenty-two trials (30,32–34,36–38,40,44–47,49,51–58,60) (n = 3391) reported incidences of various serious surgical sequelae. AGB groups had significantly lower risk of late stenoses than RYGB groups (RD: −15% [−22; 8.3]; one trial (53)). AGB had significantly lower risk of late hernia than RYGB and VBG (−4.5% [−8.4; −0.5]; two trials (34,53), I2 = 0%, and −16% [−27; −5.4]; one trial (60)). AGB had significantly higher risk of late slippage/dilatation than RYGB and VBG (6.1% [1.3; 11]; two trials (34,53), I2 = 0%, and 20% [12; 28]; two trials (51,60), I2 = 0%). AGB led to significantly lower risk of late staple line breakdown than VBG (−25% [−36; −14]; two trials (54,60), I2 = 20%). HG groups had significantly higher risk of luminal stenosis over the course of follow-up than RYGB groups (11% [1.0; 21]; one trial (56)). AGB led to a lower risk of early wound infection than RYGB (−6.3% [−11; −1.4]; one trial (53)). No trials reported the comparative incidence of peri-operative myocardial infarction.


Our network meta-analysis of studies comparing bariatric surgeries and standard care reaches three key conclusions. First, high-quality data from large, adequately powered, long-term RCTs are lacking – especially for the effects of bariatric surgery on mortality and cardiovascular disease. Second, bariatric surgery appears substantially more effective than standard care for inducing weight loss in severely obese adults. In fact, a gradient of efficacy for weight loss (as assessed at 1 year postoperatively) is apparent across the spectrum of bariatric surgical procedures. Operations that are predominantly diversionary in nature (JB, BPD) result in the greatest amounts of weight loss; hybrid procedures (RYGB) are of intermediate efficacy and purely restrictive procedures (VBG, HBG, AGB) result in the least amounts of weight loss. Third, we identified some other potentially important differences between the different procedures. Compared with RYGB, early forms of restrictive (HG, VBG) and malabsorptive procedures (JB) are associated with a greater failure rate and/or need for downstream revisions or reoperations. This is probably a major reason why these surgeries are infrequently performed at present. AGB was associated with a higher risk of slippage/dilation and procedure reversals/conversions, but a lower risk of luminal stenosis, ulceration and herniation compared with RYGB as shown in summary (Fig. 2). AGB had shorter lengths-of-stay compared with RYGB; and JB carries a higher risk of reoperation compared with RYGB.

Figure 2.

Pooled comparisons of clinical outcomes for adjustable gastric banding vs. Roux-en-Y gastric bypass. AGB, adjustable gastric banding; CI, confidence interval; I2, percentage of between-study statistical heterogeneity; MD, mean difference; n, number of participants; N, number of trials; RD, risk difference; RYGB, Roux-en-Y gastric bypass.

Of the different surgical procedures mentioned in our review, only RYGB, AGB and SG are frequently performed in contemporary practice. SG is the latest to be developed of the procedures that were examined in this meta-analysis (63). Current data suggest that RYGB reduces BMI compared with standard care; that RYGB may reduce BMI to a greater extent than AGB; and that the risk of post-operative complications may be lower with AGB than with RYGB (Fig. 2). SG is increasingly used and appears similar to RYGB in efficacy although further study is needed as only two small trials, reporting 1-year or less outcomes, were identified.

These results imply that the choice between the two most commonly performed surgeries (RYGB and AGB) is a trade-off between safety and efficacy, and the information synthesized here may be helpful to patients and providers when making decisions regarding the type of surgery to undergo or perform. We feel that the quantification of BMI change provided by our network analysis may also help patients and providers gain perspective on the relative efficacy of different procedures and incorporate this into their decision-making in conjunction with additional important data elements (e.g. local expertise, adverse effect profile). When selecting a procedure, it is also important for patients and providers to consider procedure-specific differences in post-operative nutritional requirements and care (64).

This is the most up-to-date and comprehensive systematic review of the clinical implications of bariatric surgery for the treatment of severe obesity in adults. Furthermore, to our knowledge, no prior network meta-analysis has been performed. Nonetheless, the quantity and quality of identified evidence limited the strength of our conclusions. The trial quality was generally poor (see Jadad score (19) in Table 2) and most studies examined a relatively short time horizon (median on-treatment follow-up for head-to-head comparisons was 24 months). However, we also note that study quality was perhaps artificially lowered by the requirement of double-blinding, that this is rarely performed in trials evaluating surgical interventions and that others have suggested reducing the threshold to define a high-quality surgical study from 4 to 3 using the Jadad system (65). Certain procedures we evaluated in only a single study or a low number of studies. While there were two notable RCTs that did not meet our pre-defined inclusion criteria because they enrolled patients with lower BMI ranges than current guideline-concordant thresholds, we attempted to address this limitation by performing a sensitivity analysis that included these studies (61,62). Because this was a post hoc sensitivity analysis, results should be interpreted with caution. Nonetheless, we note that the combined average BMI levels of patients within these trials was above current thresholds for surgery; and that the trial investigators were not willing to enrol heavier patients because, in their opinion, the contemporary evidence suggested a clear benefit for surgery over non-surgical treatment as BMI levels increased. For these reasons, we feel that including these two trials in the overall analysis carries a minimal downside risk of introducing bias.

We also attempted to reduce the potential for bias by following recommendations for the conduct of systematic reviews, developing a review protocol a priori, using a defined and comprehensive literature search strategy, performing quality assessment and data extraction with duplicate reviewers, and using rigorous statistical methods. Given that the techniques used to perform bariatric surgery have evolved over time, there was understandably a lack of direct comparisons between different surgical techniques. Therefore, we used state of the art, rigorous methodology for mixed-treatment comparisons to estimate the relative efficacy of each procedure as compared with the others. The results of mixed-treatment comparisons were similar to those for which direct comparisons were available and statistically coherent for change in BMI from baseline to 2-year following surgery. Hence, we feel confident that our results are valid. However, we note that mixed-treatment comparisons are considered less reliable than direct comparisons (66) and should therefore be interpreted with greater caution.

We did not find evidence from randomized trials that surgery significantly reduces obesity-related comorbidity or mortality. Notably, only four studies compared the efficacy of surgery with standard care (which would be more likely than head-to-head trials to detect a difference in these endpoints) and morbidity/mortality end points were not commonly reported in these studies. In contrast, observational data from the Swedish Obesity Study, which used a matched cohort design in 4047 subjects, remains the longest and most rigorous non-randomized data available to-date. In Swedish Obesity Study, 68% of patients underwent vertical banded gastroplasty, 19% underwent gastric banding and 13% underwent gastric bypass. Compared with matched controls, surgery was associated with reductions in 15-year mortality rates (5.0% vs. 6.3%; HR: 0.71 [0.54; 0.92]) (67) and substantial improvements in cancer incidence and cardiovascular risk factors (10,68). Furthermore, a meta-analysis of primarily observational data (22 094 patients) reported resolution or improvement of type 2 diabetes, hypertension, dyslipidemia and sleep apnea in 70–86% of cases (11). One of the RCTs excluded from the primary analysis due to a low mean baseline BMI also reported resolution of type 2 diabetes with AGB in 73% vs. 13% of patients (OR: 5.5 [2.2; 14]) (61). Although data from observational studies may overestimate efficacy, large effect sizes are noted in these studies, and indicate that these effects warrant study in future trials.

We included evidence from open and laparoscopic comparisons of surgery although they have different adverse effect profiles (15,69). One trial (51) compared laparoscopic AGB with open VBG but did not report evidence from outcomes (e.g. wound infection, hernia, vomiting, luminal stenosis) that may be confounded by these different approaches.

One final limitation is that published RCTs tend to reflect the outcomes of high selected patient populations receiving care in specialized centres and therefore these results may represent the ‘best-case’ scenario and may not be reflective of less specialized centres and/or less selected patient populations.

In conclusion, we found that although data from large, adequately powered, long-term RCTs are lacking, bariatric surgery appears substantially more effective than standard care for the treatment of severe obesity in adults. More studies are required to directly compare the clinical benefits of different surgical procedures on clinically relevant outcomes over long follow-up periods – especially for newer procedures types. Studies that examine the relative benefit of bariatric surgery in different subpopulations (e.g. obesity-related comorbidities, older or younger patients); and the impact of more vs. less selective approaches for patient selection should also be a high priority.

Conflicts of Interest Statement

Dr Arya Sharma has received consulting and speaking honoraria from Allergan Canada Inc. and Johnson & Johnson Medical Products. Dr Daniel Birch has been an advisor and has received speaking honoraria and research funding from Johnson & Johnson Medical Products and Ethicon Endo-Surgery. Dr Padwal has received research funding from Coviden Inc.


The authors of this report are grateful to Ellen Crumley for librarian support and to Mohammad Karkhaneh, Natasha Krahn, Maria Ospina, and Sophanny Tiv for additional reviewer support.


Search strategies

Database and PlatformSearch strategiesDate
MEDLINE (OVID) 1 exp abdominal fat/12653 February 2009
 2 (abdominal adj3 (fat or adipos$) ).tw. 41351950–January week 3 2009
 3 adipos$.tw. 37038 
 4 Anastomosis, Roux-en-Y/2111 
 5 exp Body Fat Distribution/1533 
 6 ( (Fat or fatty or adipos$) adj3 Distribution).tw. 4776 
 7 body fat 31 
 8 Body mass index/42052 
 9 (body mass ind$ or body ban mass).tw. 49750 
 10 31842 
 11 42334 
 12 Obesity/80177 
 13 obesit$.tw. 69747 
 14 Overweight/2752 
 15 (overweight or over weight).tw. 17198 
 16 Obesity, morbid/6146 
 17 quetelet$ ind$.tw. 680 
 18 ( (skinfold or skin fold or skin) and (thickness or measurement) ).tw. 16703 
 19 Skinfold Thickness/4797 
 20 Waist-Hip Ratio/1428 
 21 (waist hip ratio$ or hip waist ratio$).tw. 1652 
 22 or/1-21 214646 
 23 ballobes balloon$.tw. 2 
 24 banded gastroplast$.tw. 541 
 25 Bariatric surgery/1061 
 26 (Bariatric adj4 (operation or surg$ or procedure$) ).tw. 2489 
 27 Biliopancreatic bypass$.tw. 48 
 28 Biliopancreatic diversion/517 
 29 Biliopancreatic diversion$.tw. 365 
 30 Duodenal Switch$.tw. 210 
 31 Garren-Edwards Gastric 15 
 32 5 
 33 Gastrectomy/21331 
 34 Gastrectom$.tw. 12719 
 35 Gastric balloon/195 
 36 (gastric adj5 balloon$).tw. 312 
 37 gastric band$.tw. 1239 
 38 288 
 39 (Swedish adj3 band*).tw. 83 
 40 Lapband$.tw. 18 
 41 Lap-band$.tw. 177 
 42 Intragastric band$.tw. 13 
 43 Laparoscopic adjustable 6 
 44 Gastric belt$.tw. 0 
 45 Gastric bubble$.tw. 40 
 46 Gastric bypass/2636 
 47 gastric bypass$.tw. 2654 
 48 long limb bypass$.tw. 3 
 49 gastric exclusion$.tw. 28 
 50 gastric partition$.tw. 105 
 51 Gastric Resection$.tw. 3043 
 52 gastric surg$.tw. 1518 
 53 21 
 54 Gastroenterostomy/2563 
 55 Gastroenterostom*.tw. 629 
 56 Gastro-ileal bypass$.tw. 0 
 57 (Gastro-ileal surg$ or Gastro-ileal operation$).tw. 0 
 58 (Gastroileal bypass$ or Gastroileal surg$ or Gastroileal operation$).tw. 1 
 59 Gastrojejunal 145 
 60 Gastrojejunostom$.tw. 1233 
 61 Gastrojejunum 1 
 62 Gastrojejunal Fixation$.tw. 0 
 63 Gastroplasty/2192 
 64 Gastroplast$.tw. 1299 
 65 Gastroresection$.tw. 24 
 66 112 
 67 ileojejunal bypass$.tw. 14 
 68 Ileum Bypass$.tw. 6 
 69 Intestin$ bypass$.tw. 459 
 70 Intragastric balloon$.tw. 177 
 71 Jejunogastric 2 
 72 Jejunoileac Bypass$.tw. 0 
 73 Jejunoileal bypass/501 
 74 Jejunoileal bypass$.tw. 738 
 75 jejuno-ileal bypass$.tw. 187 
 76 (Jejunoileal Shunt$ or Jejuno-ileal Shunt$).tw. 70 
 77 Malabsorptive bypass$.tw. 1 
 78 (Malabsorptive surg$ or Malabsorptive operation$).tw. 22 
 79 (obes$ surg$ or obes$ operation$).tw. 421 
 80 Pancreatobiliary Bypass$.tw. 0 
 81 Restrictive bypass$.tw. 2 
 82 ( (Roux-en-y or roux y) and (gastric bypass$ or loop$ or anastomos$ or gastric bypass$ or diversion$) ).tw. 2396 
 83 287 
 84 Stomach band$.tw. 1 
 85 Stomach bypass$.tw. 13 
 86 Stomach Extirpat$.tw. 2 
 87 Stomach Resection$.tw. 567 
 88 Stomach Transection$.tw. 4 
 89 Stomach 5 
 90 Vertical band$.tw. 583 
 91 or/23-90 39419 
 92 exp Obesity, Morbid/su [Surgery] 3862 
 93 and/22,91 8204 
 94 or/92-93 8469 
 95 randomized controlled 261036 
 96 controlled clinical 77826 
 97 meta 19624 
 98 exp Randomized Controlled Trials/57210 
 99 exp Random Allocation/62453 
100 exp Double-Blind Method/98197 
101 exp Single-Blind Method/12335 
102 exp Meta-Analysis/19624 
103 (metaanalysis or meta-analysis).tw. 18958 
104 ( (singl$ or doubl$ or tripl$ or trebl$) adj2 (blind$ or mask$) ).ti,ab. 95556 
105 random$.ti,ab. 423350 
106 (systematic adj2 (review$ or overview$) ).ti,ab. 16943 
107 Medline.ab. 26306 
108 clinical 445405 
109 exp Clinical Trial/553061 
110 (clin$ adj2 trial$).ti,ab. 127254 
111 exp Placebos/27289 
112 placebo$.ti,ab. 111962 
113 exp Research Design/243654 
114 or/95-113 1024606 
115 exp animals/13763865 
116 exp animal experimentation/3866 
117 exp models animal/295951 
118 exp vertebrate/13344053 
119 or/115-118 13771114 
120 exp humans/10444338 
121 119 not 120 3326776 
122 114 not 121 940371 
123 94 and 122 1032 1032 
EMBASE (OVID) 1 Randomized Controlled Trial/1650713 February 2009
 2 exp Randomization/264671980–2009 week 5
 3 Double Blind Procedure/71178 
 4 Single Blind Procedure/7923 
 5 or/1-4 207410 
 6 Clinical Trial/528455 
 7 (clin$ adj25 trial$).mp. 604884 
 8 ( (singl$ or doubl$ or trebl$ or tripl$) adj25 (blind$ or mask$) ).mp. 118726 
 9 exp Placebo/122540 
 10 (placebo$ or random$).mp. 524524 
 11 or/6-10 877925 
 12 5 or 11 877925 
 13 exp animals/18266 
 14 exp animal experimentation/1277321 
 15 exp models animal/484741 
 16 nonhuman/3175046 
 17 exp vertebrate/8354259 
 18 or/13-17 9358623 
 19 exp humans/6438975 
 20 18 not 19 2919648 
 21 12 not 20 810795 
 22 exp abdominal fat/1193 
 23 (abdominal adj3 (fat or adipos$) ).tw. 3380 
 24 Abdominal obesity/835 
 25 adipos$.tw. 30249 
 26 Body Fat Distribution/1108 
 27 ( (Fat or fatty or adipos$) adj3 Distribution).tw. 4118 
 28 body fat 28 
 29 Body mass/63149 
 30 (body mass ind$ or BMI or body ban mass).tw. 58425 
 31 (obese or obesit$).tw. 78705 
 32 Obesity/85968 
 33 (overweight or over weight).tw. 15181 
 34 Morbid obesity/4336 
 35 quetelet$ ind$.tw. 585 
 36 ( (skinfold or skin fold or skin) and (thickness or measurement) ).tw. 15247 
 37 Skinfold Thickness/3672 
 38 Waist-Hip Ratio/1437 
 39 (waist hip ratio$ or hip waist ratio$).tw. 1492 
 40 or/22-39 193757 
 41 ballobes 1 
 42 banded gastroplast$.tw. 576 
 43 Bariatric surgery/3155 
 44 (Bariatric adj4 (operation or surg$ or procedure$) ).tw. 2628 
 45 biliopancreatic bypass/653 
 46 biliopancreatic diversion$.tw. 391 
 47 Duodenal Switch$.tw. 220 
 48 (Garren-Edwards Gastric Bubble or gegb).tw. 14 
 49 exp Gastrectomy/10888 
 50 Gastrectom$.tw. 7395 
 51 (gastric adj5 balloon$).tw. 278 
 52 gastric banding/1669 
 53 gastric band$.tw. 1239 
 54 292 
 55 (Swedish adj3 band*).tw. 91 
 56 Lapband$.tw. 17 
 57 Lap-band$.tw. 170 
 58 Intragastric band$.tw. 13 
 59 Laparoscopic adjustable 3 
 60 Gastric belt$.tw. 0 
 61 Gastric bubble$.tw. 31 
 62 Gastric bypass$.tw. 2591 
 63 Long limb bypass$.tw. 3 
 64 Gastric exclusion$.tw. 17 
 65 Gastric partition$.tw. 99 
 66 Gastric Resection$.tw. 1247 
 67 Gastric surg$.tw. 958 
 68 23 
 69 Gastroenterostomy/385 
 70 Gastroenterostom*.tw. 297 
 71 Gastroileal bypass$.tw. 1 
 72 (Gastroileal surg$ or Gastroileal operation$).tw. 0 
 73 (Gastroileal bypass$ or Gastroileal surg$ or Gastroileal operation$).tw. 1 
 74 Gastrojejunal 101 
 75 Gastrojejunal 0 
 76 Gastrojejunostomy/1370 
 77 Gastrojejunostom$.tw. 960 
 78 Gastrojejunum 1 
 79 Gastroplasty/1616 
 80 gastroplast$.tw. 1221 
 81 Gastroresection$.tw. 20 
 82 Hemigastrectom$.tw. 57 
 83 ileojejunal bypass$.tw. 7 
 84 Ileum Bypass/176 
 85 ileum bypass$.tw. 3 
 86 intestine bypass/465 
 87 intestin$ bypass$.tw. 263 
 88 Intragastric balloon$.tw. 175 
 89 Jejunogastric 2 
 90 Jejunoileac Bypass$.tw. 0 
 91 jejunoileal bypass/304 
 92 jejunoileal bypass$.tw. 480 
 93 jejuno-ileal bypass$.tw. 110 
 94 (Jejunoileal Shunt$ or jejuno-ileal shunt$).tw. 17 
 95 Malabsorptive bypass$.tw. 1 
 96 (Malabsorptive surg$ or Malabsorptive operation$).tw. 26 
 97 (Obes$ surg$ or obes$ operation$).tw. 383 
 98 Pancreatobiliary Bypass$.tw. 0 
 99 Restrictive bypass$.tw. 1 
100 Roux y Anastomosis/2679 
101 ( (Roux-en-y or roux y) and (gastric bypass$ or loop$ or anastomos$ or gastric bypass$ or diversion$) ).tw. 2161 
102 285 
103 Stomach band$.tw. 0 
104 Stomach bypass/2960 
105 stomach bypass$.tw. 4 
106 Stomach Extirpat$.tw. 1 
107 Stomach Resection$.tw. 79 
108 Stomach Transection$.tw. 3 
109 Stomach 2 
110 Vertical band$.tw. 606 
111 or/41-110 24587 
112 and/40,111 7281 
113 exp Obesity, Morbid/su [Surgery] 2679 
114 or/112-113 7429 
115 and/21,114 1029 
CENTRAL (OVID) 1 exp abdominal fat/61February 3, 2009
 2 (abdominal adj3 (fat or adipos$) ).tw. 3014th Quarter 2008
 3 adipos$.tw. 774 
 4 Anastomosis, Roux-en-Y/75 
 5 exp Body Fat Distribution/72 
 6 ( (Fat or fatty or adipos$) adj3 Distribution).tw. 250 
 7 body fat 0 
 8 Body mass index/2940 
 9 (body mass ind$ or body ban mass).tw. 4361 
10 2652 
11 3477 
12 Obesity/3582 
13 obesit$.tw. 2619 
14 Overweight/219 
15 (overweight or over weight).tw. 1566 
16 Obesity, morbid/234 
17 quetelet$ ind$.tw. 27 
18 ( (skinfold or skin fold or skin) and (thickness or measurement) ).tw. 1128 
19 Skinfold Thickness/228 
20 Waist-Hip Ratio/89 
21 (waist hip ratio$ or hip waist ratio$).tw. 282 
22 or/1-21 12566 
23 ballobes balloon$.tw. 1 
24 banded gastroplast$.tw. 33 
25 Bariatric surgery/16 
26 (Bariatric adj4 (operation or surg$ or procedure$) ).tw. 55 
27 Biliopancreatic bypass$.tw. 0 
28 Biliopancreatic diversion/8 
29 Biliopancreatic diversion$.tw. 9 
30 Duodenal Switch$.tw. 3 
31 Garren-Edwards Gastric 3 
32 1 
33 Gastrectomy/474 
34 Gastrectom$.tw. 496 
35 Gastric balloon/18 
36 (gastric adj5 balloon$).tw. 37 
37 gastric band$.tw. 63 
38 10 
39 (Swedish adj3 band*).tw. 6 
40 Lapband$.tw. 3 
41 Lap-band$.tw. 14 
42 Intragastric band$.tw. 0 
43 Laparoscopic adjustable 0 
44 Gastric belt$.tw. 0 
45 Gastric bubble$.tw. 7 
46 Gastric bypass/94 
47 gastric bypass$.tw. 131 
48 Long limb bypass$.tw. 0 
49 gastric exclusion$.tw. 3 
50 gastric partition$.tw. 8 
51 Gastric Resection$.tw. 69 
52 gastric surg$.tw. 94 
53 6 
54 Gastroenterostomy/31 
55 Gastroenterostom*.tw. 14 
56 Gastro-ileal bypass$.tw. 0 
57 (Gastro-ileal surg$ or Gastro-ileal operation$).tw. 0 
58 (Gastroileal bypass$ or Gastroileal surg$ or Gastroileal operation$).tw. 0 
59 Gastrojejunal 1 
60 Gastrojejunostom$.tw. 42 
61 Gastrojejunum 0 
62 Gastrojejunal Fixation$.tw. 0 
63 Gastroplasty/90 
64 Gastroplast$.tw. 81 
65 Gastroresection$.tw. 1 
66 2 
67 ileojejunal bypass$.tw. 0 
68 Ileum Bypass$.tw. 1 
69 Intestin$ bypass$.tw. 8 
70 Intragastric balloon$.tw. 20 
71 Jejunogastric 0 
72 Jejunoileac Bypass$.tw. 0 
73 Jejunoileal bypass/20 
74 Jejunoileal bypass$.tw. 38 
75 jejuno-ileal bypass$.tw. 4 
76 (Jejunoileal Shunt$ or Jejuno-ileal Shunt$).tw. 2 
77 Malabsorptive bypass$.tw. 0 
78 (Malabsorptive surg$ or Malabsorptive operation$).tw. 1 
79 (obes$ surg$ or obes$ operation$).tw. 19 
80 Pancreatobiliary Bypass$.tw. 0 
81 Restrictive bypass$.tw. 0 
82 ( (Roux-en-y or roux y) and (gastric bypass$ or loop$ or anastomos$ or gastric bypass$ or diversion$) ).tw. 83 
83 14 
84 Stomach band$.tw. 1 
85 Stomach bypass$.tw. 1 
86 Stomach Extirpat$.tw. 0 
87 Stomach Resection$.tw. 8 
88 Stomach Transection$.tw. 0 
89 Stomach 0 
90 Vertical band$.tw. 35 
91 or/23-90 1195 
92 22 and 91 382 
93 swedish obese subject$.tw. 9 
94 92 or 93 386 
95 exp Obesity, Morbid/su [Surgery] 132 
96 or/94-95 388 
Center watchbariatric surgery 327 November 2008
gastroplasty 0 
gastrectomy 42 
gastric bypass 6 
gastric partition* 0 
gastric band* 0 
biliopancreatic diversion 0 
Clinical trials (US National Institutes of Health)bariatric surgery 10327 November 2008
gastroplasty 4 
gastrectomy 41 
gastric bypass 82 
gastric partition* 0 
gastric band* 20 
biliopancreatic diversion 6 
Current controlled trialsbariatric surgery 8327 November 2008
gastroplasty 12 
gastrectomy 120 
gastric bypass 87 
gastric partition* 0 
gastric band* 29 
biliopancreatic diversion 2 
UK Clinical Research Network: Portfolio Databasebariatric surgery 127 November 2008
gastroplasty 0 
gastrectomy 0 
gastric bypass 0 
gastric partition* 0 
gastric band* 0 
biliopancreatic diversion 0 
Agency for Healthcare Research and Quality (AHRQ)browsed their online collection22 January 2009
National Institute for Health and Clinical Excellence (NICE)browsed their online collection22 January 2009