The clinical effectiveness and cost-effectiveness of bariatric (weight loss) surgery for obesity: a systematic review and economic evaluation

Adults

A recent study10 evaluated trends in the incidence of overweight and obesity in the USA from 1950 to 2000 using data from the Framingham study participants. The results indicated that the overall incidence rates of overweight increased twofold, and that of obesity more than threefold over five decades. Per decade, there was an increase in the incidence of overweight of 25% in women and 20% in men. The corresponding per decade increases in women and men were 34% and 29% for the incidence of obesity, and 31% and 97% for the incidence of class II obesity.

A second study has compared trends in the incidence of overweight and obesity in a rural population from Sweden, and one from the USA between 1989 and 1999. 11 The 10-year incidence of overweight was similar in the two countries (337/1000 in Sweden, 336/1000 in the USA). However, the 10-year incidence of obesity was greater in the USA (173/1000) where 21.3% were obese in 1989 rising to 32.3% in 1999, than in Sweden (120/1000) where the prevalence of obesity was lower at both time points (9.6% in 1989 and 18.4% in 1999).

Adolescents and children

Calculating the incidence of overweight or obesity in adolescents and children is particularly problematic. For adolescents who are approaching adulthood the difficulty lies in ensuring that the data sets employed in the study have used a definition of obesity that is comparable for the age groups of interest. Gordon-Larsen and colleagues6 have used data from a longitudinal, nationally representative, school-based study of US adolescents and estimated that obesity incidence over the five-year study period was 12.7%. In contrast, fewer than 2% of the total sample of young adults who were obese as adolescents became non-obese.

The generalisability of the findings reported above to adults and children in the UK is unknown.

Adults

The World Health Organization (WHO)’s projections indicated that globally in 2005 approximately 1.6 billion adults (age 15 +) were overweight and at least 400 million adults were obese. 12 In England in 2006 the prevalence of overweight in people aged 16 and over was 38% (approximately 15.4 million people), with 24% obese (approximately 9.8 million people). 13 In Wales in 2007, 57% of adults were classified as overweight or obese, including 21% obese. 14

The prevalence of obesity (BMI > 30) among adults in England and Wales is increasing. In 2006 reported obesity prevalence in England was 23.7% for men and 24.2% for women. The increase was clear when the 2006 figures are compared with those for 1998 which were 17.3% for men and 21.2% for women. 15 The 2006 prevalence of morbid obesity (BMI > 40)16 was 2.1% (just under 863,000 people) with women being more likely to be morbidly obese than men (2.7% of women versus 1.5% of men). In comparison, the 1998 figures for morbid obesity were 0.6% for men and 1.9% for women. For a standard primary-care trust (PCT) population of 250,000, there would be 5250 cases of morbid obesity (based on the overall 2006 population value for England of 2.1% morbid obesity).

Prevalence of obesity increases with age, until age 55–64 years in men and until age 65–74 years in women, when it begins to decline16 (Table 2). The number of men and women with obesity in England is shown in Table 3. 17,18 Obesity in women is more common in households where the current or former occupation of the household reference person is classified as routine and manual than in those households classified as intermediate, or managerial and professional (Table 4). 16 For morbid obesity in women the prevalence was 1.6% in managerial and professional households, but 4.0% in routine and manual households. Differences for men by category based on occupation was less marked. 16 Data from the Welsh Health Survey also indicate a rise in obesity among people aged 16 years and over from 18% in 2003–4 to 19% in 2005–6. 19

The prevalence of obesity is predicted to rise in the future. WHO has projected that by 2015 more than 700 million adults will be obese. In the UK, the Foresight programme provides visions of the future using science-based methods. The Foresight project ‘Tackling Obesities: Future Choices’ produced a report made up of a number of documents which forms a long-term vision of how a sustainable response to obesity can be delivered in the UK over the next 40 years. The modelling section of the Foresight Report predicts that in England, if current trends persist, 36% of men and 28% of women aged 21 to 60 will be obese in 2015. 20 Predicting trends in morbid obesity is more problematic. The Foresight modelling projection to 2050 suggests figures of 1% for males and 4% for females. 20 In contrast a different Foresight project output has estimated that the proportion of morbidly obese English males and females will reach nearly 3% and 6%, respectively, in 2030. 21

Children

As noted, overweight and obesity in children and young people are not defined according to a particular BMI. This means that varying prevalences of overweight and obesity in children and young people will be obtained depending on the particular definitions used. 22 This not only makes it difficult to obtain an accurate prevalence estimate, but also creates problems when trying to compare prevalences reported by studies employing different definitions. The NICE obesity guideline states that BMI measurement in children and young people should be related to the UK 1990 BMI charts to give age- and gender-specific information. 9

Despite the difficulties in assessing the prevalence of overweight and obesity in children and young people there is general agreement that these are increasing. The National Statistics for England16 report that between 1995 and 2005 the proportion of boys aged 2 to 15 years who were obese increased from 10.9% to 18.3%. For girls a similar pattern was seen with the proportion of obese 2–15-year-olds increasing from 12.0% in 1995 to 18.3% in 2005 (Table 5). In the 11–15 years age group the prevalence of obesity in 2005 was very similar in boys and girls at 20.4% and 20.8% respectively. In general the proportion of children who are obese increases with age until age 13–15 (Table 6). 16 These prevalence figures were derived using the UK National BMI percentile classification to describe childhood overweight and obesity. This uses a BMI threshold for each age above which a child is considered overweight or obese. The classification estimates were produced by calculating the percentage of boys and girls who were over the 85th (overweight) or 95th (obese) BMI percentiles based on the 1990 UK reference population.

Increased mortality

Obesity significantly increases the risk of mortality at any given age (including after adjustment for other risk factors such as smoking) and those who have been overweight for the longest are at the highest risk. The National Audit Office (NAO) has reported that evidence from studies suggests that young adults with a BMI of 30 have a mortality risk that is about 50% higher than that of a young adult with a healthy BMI (18.5–25). 24 For a young adult with a BMI of 35 the mortality risk is more than doubled. One study reported that 40-year-old obese (BMI greater than or equal to 30) women and men lost 7.1 and 5.8 years of life, respectively, compared with 40-year-old non-smoking women and men of normal weight. 25 In 2004, a House of Commons Select Committee report estimated that 6.8% of all deaths in England were attributable to obesity. 26

Type 2 diabetes

Type 2 diabetes is the most common form of diabetes, accounting for over 90% of all diabetes in the UK. 13 It is characterised by insulin resistance and is a serious life-shortening condition. The ‘first-line’ treatment is diet, weight control and physical activity but drug therapy, e.g. with metformin, sulphonylurea drugs, thiazolidinediones (commonly called glitazones), or insulin therapy may become necessary. The risk of developing diabetes rises with increasing BMI even below the threshold of clinical obesity. The Health Survey for England (HSE) data have been used to examine the association between a number of risk factors and Type 2 diabetes in adults aged 35 years and over. Men and women who were obese had approximately double the odds of having Type 2 diabetes compared with those who had a BMI of less than 25 (after adjusting for other factors). A raised waist circumference was also linked to increased odds of having Type 2 diabetes, in men the odds were doubled, whereas women with a raised waist circumference were four times more likely to have Type 2 diabetes than those without a raised waist circumference measurement. 13

Cardiovascular disease and CVD risk factors

The term CVD encompasses ischaemic heart disease (IHD) [also known as coronary heart disease (CHD)], stroke and peripheral vascular disease. Obesity is an independent risk factor for CVD. Data from the HSE 200317 demonstrate that a relationship between IHD and BMI is present in men and women. The prevalence of IHD or stroke was lowest among people with a normal BMI, but increased for people in the overweight category and was highest among obese women and men. However, in women, the higher prevalence in the obese was no longer significant following age standardisation (Table 10). A raised waist circumference is also linked to an increased prevalence of CVD in men and women.

Hypertension is a key risk factor for CVD and the positive association between blood pressure and BMI is well documented. Data from the HSE 200317 shows that overweight men and women (BMI between 25 to less than 30) and obese men and women (BMI 30 or more) both have a higher blood pressure than those with a normal BMI (Table 11). The link between high blood pressure and obesity was also observed during the HSE 2005 which focused on people aged 65 years and over. 27 In this age group hypertension was twice as common in obese men and women, and more prevalent in overweight women, compared with those with a weight in the normal range.

Abnormalities in serum lipid levels [raised total cholesterol, triglycerides and low-density lipoprotein (LDL), with reduced high-density lipoprotein (HDL)] are a further risk factor for CVD. The National Health and Nutrition Examination Survey (NHANES) III28 found that the prevalence of raised cholesterol (≥ 240 mg/dl) in obese men and women was 22% and 27%, respectively, compared with 13% of adults with BMI < 25. HDL-cholesterol decreased with increasing BMI. The prevalence of low HDL-cholesterol (< 35 mg/dl men, < 45 mg/dl women) in obese adults was 31% of men and 41% of women compared with 9% and 17%, respectively, in adults with desirable weight. A more recent publication employing data from the survey of health, ageing and retirement in Europe (SHARE) also found that the odds ratios for high cholesterol were significantly increased for overweight and obese adults. 29

Cancer

A systematic review and meta-analysis of prospective observational studies reported that in men, a 5 kg/m² increase in BMI was strongly associated with oesophageal adenocarcinoma [relative risk (RR) 1·52, p < 0·0001] as well as thyroid (1·33, p = 0·02), colon (1·24, p < 0·0001) and renal (1·24, p < 0·0001) cancers. For women, the strong associations were between a 5 kg/m² increase in BMI and endometrial (1·59, p < 0·0001), gall bladder (1·59, p = 0.04), oesophageal adenocarcinoma (1·51, p < 0·0001) and renal (1·34, p < 0·0001) cancers. There were also weaker positive associations between increased BMI and some other cancers in both men and women. 30 As health risks increase with increasing obesity, increases in BMI greater than 5 kg/m² may be associated with greater cancer risks.

A study assessing the cost of obesity to the UK estimated that the cancers with some relationship with overweight and obesity (breast cancer, colon/rectum cancer and corpus uteri cancer) were responsible for 6.2% of all mortality. 31

Osteoarthritis

Osteoarthritis, or degenerative disease of the knee and other weight-bearing joints, and lower back pain are common in obesity. Some effects are thought to be the result of excess weight; BMI has been associated with the incidence and progression of knee osteoarthritis, but not associated with the incidence and progression of hip osteoarthritis. However, it has also been postulated that some obesity-related osteoarthritis may be the result of a metabolic effect. 32,33 Whatever the cause, an estimated 27% of hip replacements and 69% of knee replacements in middle-aged women in the UK are attributable to obesity. 34

Respiratory disorders

Respiratory disorders such as obstructive sleep apnoea are associated with obesity. It has been estimated that approximately 17% of adults aged 30–69 years in the USA have sleep-disordered breathing graded as ‘mild or worse’ and it is further estimated that in 41% of these adults their sleep-disordered breathing can be attributed to having a BMI ≥ 25. The study has similarly estimated that 5.7% of adults have sleep-disordered breathing graded ‘moderate or worse’ which in 58% of cases may be attributed to excess weight. 35

Reproductive disorders

Reproductive disorders are common in obesity, occurring in both women and men. Cohort studies have identified obese women as having a higher risk of complications during pregnancy, such as pregnancy-induced hypertension, antepartum venous thromboembolism, induction of labour, caesarean delivery and wound infection, pre-eclampsia, premature labour, gestational diabetes and macrosomia (large-for-date fetuses). 36–40 Some studies have found evidence for an increased risk of adverse pregnancy outcomes such as perinatal death,41 but others have not found evidence for an increased risk. 38

Allocation

Nine of 23 RCTs described adequate allocation sequence generation,108–110,115–117,123,125,127,128,130,132 and just five had adequate concealment of allocation. 110,115,116,118,130,132 The method of allocation sequence generation and concealment was not reported by the remaining studies, therefore they were judged to be of uncertain risk of bias.

Blinding

Only one RCT reported that outcome assessors were blinded to the intervention assignment. 125 Outcome assessors were not blinded to the intervention assignments in three RCTs,115–117,126 therefore they were judged to be at high risk of bias. This information was not reported by the remaining RCTs.

Six RCTs assessed outcomes self-reported by participants. In four of these studies participants were not blinded to the intervention received,113–116,126,131 and in two studies blinding of participants was not reported or was unclear. 118,124

Incomplete outcome data

Incomplete outcome data for weight loss were adequately addressed by 14 RCTs (Table 14.3). 107–110,112,117,119,120,123,125–129,131,132 The remaining nine RCTs were judged to be at uncertain risk of bias. The withdrawals in the study by O’Brien and colleagues115,116 were uneven between groups, but as reasons were not provided for all withdrawals it was not clear whether withdrawals were related to outcome. In the studies by Lujan and colleagues130 and Bessler and colleagues,118 the number of participants included at each follow-up was not reported. The study by Puzziferri and colleagues was reported in two publications,113,114 the most recent of which contained a smaller number of participants, but no explanation for this was given. Information in the study by MacLean and colleagues121,122 was contradictory between tables and text. Presence or absence of missing outcome data was not reported by four studies. 73,106,124,129

Three RCTs assessed QoL. 113–116,124 Incomplete outcome data for QoL were not reported by Lee and colleagues. 124 O’Brien and colleagues115,116 analysed QoL data only for those who completed the study, which suffered from uneven withdrawals between the groups as previously stated. QoL in the study by Puzziferri and colleagues113,114 was reported only for 44 of 155 patients originally randomised, with no explanation given.

Comorbidity was assessed by seven RCTs. Incomplete outcome data for comorbidity were adequately addressed by three studies,107,117,127,128,132 but the remaining four studies were judged to be of uncertain risk of bias. 113–116,118,129

Selective reporting

The study by Bessler and colleagues118 was judged not to be free of selective outcome reporting, as per cent excess weight loss was reported at 6, 12, 24 and 36 months follow-up (with a statistically significant difference at 36 months), while BMI was reported at 12 and 24 months only, with no measure of variance or statistical analysis. The remaining studies were judged to be of uncertain risk of bias. For example, the studies by Agren and Naslund73 and VanWoert and colleagues106 are reported as abstracts only, therefore limited data were provided. Lujan and colleagues130 reported BMI in a figure only, with no exact data reported and no measure of variance. Mingrone and colleagues119 did not report adverse effects.

Other potential sources of bias

Four RCTs were judged to be at high risk of bias because they used block randomisation in an unblinded trial which can mean it is possible to predict future assignments. 113,114,117,123,131 In addition, recruitment to the RCT by Sugerman and colleagues123 was stopped early (after nine months) following an a priori stopping rule which stated that when a significant difference (p < 0.05) in weight loss was noted for either treatment, patient recruitment would cease until patients had achieved the same follow-up after surgery. The study would have reopened if statistical significance p < 0.01 was not present when all patients had reached an equivalent time frame after surgery. Studies that are stopped early are more likely to show extreme treatment effects than those that continue to the end. 135 Sundbom and Gustavsson112 was also judged to be at high risk of bias, as there appeared to be higher comorbidity on one of the treatment arms. The remaining RCTs were judged to be of uncertain risk of bias, because there was either insufficient information to assess whether an important risk of bias exists, or insufficient rationale or evidence that an identified problem will introduce bias. For example, Olbers and colleagues108,109 excluded 17 patients after randomisation either because they expressed a preference about the surgery they received, or were found to have a BMI > 50; the effect of the exclusion of these patients was unclear. Also, Angrisani and colleagues107 reported that they were in the early phase of the learning curve for one intervention, whereas the senior author had more experience with the comparator.

Allocation of participants in the cohort studies

In the SOS study, participants could volunteer for conventional or surgical treatment. For each surgical case a control was matched by computer taking into account 18 variables. The study by Stoeckli and colleagues102–104 included participants opting for surgical treatment, choosing either adjustable silicone gastric banding or gastric bypass. The control group were patients attending diet consultation, although the reasons for the patients not undergoing surgery are not given. Buddeberg-Fischer and colleagues105 included patients applying for bariatric surgery. However, the comparability of the surgery and no-surgery groups is unclear because although one paper states that all but three participants met the criteria for surgery (one had BMI 31.6, two were aged > 60 years),105 about nine participants in an earlier publication said that the reason for not undergoing surgical treatment was that their BMI was under the limit for morbid obesity. 133

Randomised controlled trials

The two RCTs that compared laparoscopic adjustable gastric banding with non-surgical interventions in obese people (BMI 40 or less) with identifiable comorbidities, reported statistically significant benefit on measures of weight change for those receiving laparoscopic adjustable gastric banding115–117 (Table 17). In a comparison of laparoscopic adjustable gastric banding with non-surgical interventions in people with a BMI ranging from 30 to 35 and identifiable comorbidities, O’Brien and colleagues115,116 reported a statistically significant (p < 0.001) difference in the weight of participants at 12, 18 and 24 months. While people in the laparoscopic adjustable gastric banding group consistently lost weight during the two-year follow-up, those in the non-surgical group increased in weight, despite an initial loss of weight at six months. The differences in weight change were reflected in their respective BMIs, with statistically significant (p < 0.001) differences at beyond the six-month follow-up. Participants in the laparoscopic adjustable gastric banding group experienced a decrease in their BMI from 33.7 at baseline to 26.4 at two years compared with a decrease from a BMI of 33.5 at baseline to 31.5 at two years for those in the non-surgical group. By two years people receiving laparoscopic adjustable gastric banding had lost 87.2% of excess weight, statistically significantly (p < 0.001) more than the 21.8% lost by people in the non-surgical group. Some 98% of those people with a laparoscopic adjustable gastric banding had achieved a satisfactory weight loss (greater than 25% of excess weight loss) at two years, compared with 35% of people in the non-surgical group.

Dixon and colleagues,117 who assessed the effectiveness of laparoscopic adjustable gastric banding and conventional therapy on obese people (BMI 30 to 40) diagnosed with Type 2 diabetes at two years follow-up, found a statistically significantly (p< 0.001) greater mean percentage weight loss following laparoscopic adjustable gastric banding (20.0%) compared with conventional therapy (1.4%). This equated to a statistically significant (p < 0.001) difference in mean weight loss with those receiving laparoscopic adjustable gastric banding losing an additional 19.6 kg. The change in weight resulted in a reduction in the mean BMI for people in the laparoscopic adjustable gastric banding group from 36.9 to 29.5, while those in the conventional therapy group declined from a BMI 37.1 to 36.6. Dixon and colleagues reported that the loss of weight represented a loss of 62.5% of excess weight (using BMI 25 as ideal weight) for people with the laparoscopic adjustable gastric banding and 4.3% for people receiving conventional therapy. Similar benefits were noted on measures of waist circumference and waist :  hip ratio for those in the laparoscopic adjustable gastric banding group compared with the conventional therapy group (Appendix 5).

Mingrone and colleagues119 randomised patients to either biliopancreatic diversion or a diet of 20 kcal/kg fat-free mass, 55% carbohydrates and 15% proteins that was modified every six months according to analysis of fat-free mass. Weight, BMI, fat-free mass and fat mass were significantly reduced in both men and women 12 months following biliopancreatic diversion compared with baseline (p < 0.0001). Weight loss in women and men following surgery was 35 kg and 52 kg, respectively, and 7 kg and 9 kg, respectively, following the diet. The study did not present a statistical comparison of surgery versus diet, and because the results reflect a before and after comparison only they should be treated with caution.

Cohort studies

In 3505 participants who completed two years follow-up, the SOS study reported a significantly greater weight loss among gastric surgery patients (23.4%) than for those receiving conventional treatment (0.1% gain) [difference 22.2, 95% confidence interval (CI) 21.6 to 22.8, p < 0.001]97 (Table 17). Among 1276 patients followed for 10 years, patients in the surgical group had a 16% (SD 12.1) weight loss compared with a 1.5% (SD 9.9) gain in weight for patients receiving conventional treatment. This equates to a mean reduction in weight and BMI of – 19.7 kg (SD 15.8) and – 6.7 (SD 5.4), respectively, for the surgical group versus a gain in weight and BMI of 1.3 kg (SD 13.8) and 0.7 (SD 4.9), respectively for the conventional treatment group, a statistically significant difference between groups (p < 0.0001). 100 Weight loss after 10 years was greater following gastric bypass (25%, SD 11) than following vertical banded gastroplasty (16%, SD 11) or gastric banding (adjustable or non-adjustable) (14%, SD 14), although it should be noted that this was not tested statistically and may be subject to selection bias (systematic differences between the groups) as the groups were not randomised. After 15 years, the weight loss was 27% (SD 12), 18% (SD 11) and 13% (SD 14), for the three surgical groups respectively; although it should be noted that the numbers followed for this duration were much smaller (Appendix 5). 99

One small cohort study, reported by Buddeberg-Fischer and colleagues,105 compared participants undergoing laparoscopic gastric banding or laparoscopic gastric bypass (surgery group) with a no-surgery control group. While mean BMI was not shown to be statistically significantly different between the two groups at end point (mean 3.2 years follow-up), the mean change in BMI was (p < 0.001), indicating greater BMI reduction in the surgery group (Table 4.6). The percentage of excess weight loss was also seen to be statistically significantly better in the surgery group compared with the no-surgery group (p < 0.001). It should be noted, however, that there is some potential for bias as these measures were self-reported and the rates of dropout were different between the two groups over time. The study also assessed BMI and percentage excess weight loss between the two types of surgical procedures used within the surgery arm. The study showed that the mean change in BMI was greater in those undergoing laparoscopic gastric bypass than those undergoing laparoscopic gastric banding [– 27.7 (SD 12.6) versus –17.2 (SD 12.5) for the two groups respectively, p = 0.002], and that the percentage excess weight loss was greater in the laparoscopic gastric bypass subgroup than the laparoscopic gastric banding group [52.8% (SD 17.0) versus 36.0% (SD 24.5) respectively, p = 0.005] (Appendix 5). Care should be used when interpreting these results, however, because the sample sizes were small, there was some degree of crossover between surgical options, this was not a planned comparison, and the surgical groups were not randomised and therefore were subject to selection bias.

The small cohort study with just 20 participants, reported by Stoeckli and colleagues,102–104 also found a statistically significant lower mean BMI two years following surgery [gastric bypass 32.9 (SD 6.7); laparoscopic adjustable silicone gastric banding 33.2 (SD 4.7); compared with a control group who did not undergo surgery 41.0 (SD 3.4)].

Randomised controlled trials

O’Brien and colleagues compared changes in the short-form health survey (SF-36) domain scores from baseline to two years follow-up for people undergoing laparoscopic adjustable gastric banding and non-surgical therapy. 115,116 Although no point estimates were reported, O’Brien and colleagues noted improvements in scores on all eight domains for the laparoscopic adjustable gastric banding group and on three domains (physical function, vitality and mental health) for the non-surgical therapy group. Statistically significantly greater improvements were reported for five of the eight domains for laparoscopic adjustable gastric banding compared with the non-surgical group (Table 18).

Quality of life was not reported by either Dixon and colleagues117 or Mingrone and colleagues. 119

Cohort studies

The SOS study assessed Health Related Quality of Life (HRQoL) using several measures, including general health perceptions from the General Health Rating Index, social interaction from the Sickness Impact Profile, overall mood from the Mood Adjective Check List (MACL), the obesity-related problems scale and the Hospital Anxiety and Depression scale. At baseline the patients in the surgery group had generally worse HRQoL than those in the conventional treatment group. 86,100 These differences may reflect the significant differences in BMI and prevalence of hypertension that developed between matching of controls and start of treatment, or may indicate bias in the selection of patients for surgery.

The two-year results of 974 participants have been tabulated and discussed in the previous report15 and can be seen in Appendix 5. In brief, at two years follow-up gastric surgery patients had significant improvements in all HRQoL measures compared with patients receiving conventional treatment. These changes were significantly related to the magnitude of the weight loss and may have been expected given that the patients in the surgical group had significantly higher BMI at the time of treatment compared with the controls. 86

A more recent report of 1276 participants found that improvements in HRQoL, which peaked one year after surgery, were followed by a gradual decline between one and six years, and then observations were relatively stable between six and ten years follow-up. 100 All HRQoL measures were improved at 10 years compared with baseline for the surgery group, but for the conventional group some had improved while others had worsened. After 10 years follow-up, the mean level of current health perception, social interaction, obesity-related problems, overall mood and depression did not differ significantly between the surgery and conventional treatment groups, although the surgery group had more anxiety (p < 0.01). However, statistically significantly greater 10-year change was observed in the surgery group for current health perceptions, social interaction, obesity-related problems and depression. There was no statistically significant difference in 10-year change for overall mood and anxiety (Table 18).

Buddeberg-Fischer and colleagues105 applied a range of validated questionnaires related to quality of life to participants via telephone interview. The study reported that the Psychosocial Stress and Symptom Questionnaire (PSSQ) was used which incorporated the Hospital Anxiety and Depression Scale (HADS), the Bing Scale Questionnaire (BSQ) and the Psychosocial Assessment Questionnaire (PAssQ), although no results were reported for the overall PSSQ questionnaire. After a mean follow-up of 3.2 years there were no statistically significant differences between groups on mean scores from any of the three questionnaires (Table 18).

Buddeberg-Fischer and colleagues105 also presented self-reported assessment of overall physical and mental health. This was also undertaken via telephone interview and while the results give an indication of the individuals’ perception of their health this was not a validated measure and is likely to be measuring the state of the individual at that particular point in time only. Higher proportions of participants in the surgical group (79.3%) rated their physical health as good compared with the no-surgery group (64.5%), but this was not statistically significantly different (p = 0.10). A similar pattern emerged for ratings of mental health (77.6% versus 67.7% for the two groups respectively, p = 0.22). No analysis was presented of the proportions rating their health as poor (Appendix 5).

Quality of life was not reported by Stoeckli and colleagues. 102–104

Randomised controlled trials

Dixon and colleagues assessed the effects of laparoscopic adjustable gastric banding compared with conventional therapy on measures of glycaemic control and use of diabetes medication among 60 obese people with Type 2 diabetes. 117 Remission of Type 2 diabetes at two years follow-up was statistically significantly (p < 0.001) higher following laparoscopic adjustable gastric banding (73%) than conventional therapy (13%) (RR 5.5; 95% CI 2.2, 14.00) (Table 19). Similarly, people undergoing laparoscopic adjustable gastric banding were statistically significantly less likely to suffer from metabolic syndrome (70% versus 13%, p < 0.001). Measures of glycaemic control improved more following laparoscopic adjustable gastric banding than conventional therapy with statistically significantly greater decreases in mean levels of HbA1c (glycosylated haemoglobin; difference – 1.43, 95% CI – 2.1 to 0.80, p < 0.001), plasma glucose (difference – 31.8, 95% CI – 53.1 to – 12.3, p = 0.002), plasma insulin (difference – 13.4, 95% CI – 19.6 to – 7.3, p < 0.001), and HDL-cholesterol (difference 10.0, 95% CI 5.8 to 14.2, p < 0.001). There were greater improvements in other indices of glycaemic control (e.g. blood pressure, total cholesterol) for the laparoscopic adjustable gastric banding group compared with the conventional therapy group, although these were not statistically significant (Appendix 5). The benefits of laparoscopic adjustable gastric banding compared with conventional therapy were evident in the reduction in the use of diabetes medication. At two years follow-up a greater proportion of those receiving laparoscopic adjustable gastric banding no longer required diabetes medication compared with conventional therapy (change from baseline 83% versus 15% respectively, not tested for statistical significance). There were similar improvements from baseline to two years follow-up for those in the laparoscopic adjustable gastric banding group compared to the conventional therapy group in their use of metformin (86.3% versus 30.8%), other hypoglycaemics (27.6% versus 3.2%), insulin (3.4% versus 11.5%), anti-hypertensives (48% versus 0%) and lipid-lowering agents (27.6% versus 3.9%) although these differences between the groups were also not tested for statistical significance.

O’Brien and colleagues noted that the group receiving laparoscopic adjustable gastric banding experienced a statistically significant reduction in the proportion of people with metabolic syndrome at two years follow-up (baseline 37.5%, two years 2.7%, p < 0.001). 115,116 For those in the non-surgical group the proportion with metabolic syndrome decreased, but not significantly (baseline 37.5%, two years 24%, p = 0.22). There was a statistically significant difference in the number of participants with metabolic syndrome between the two interventions, at two years p = 0.006 (Table 19).

Comorbidities were not reported by Mingrone and colleagues. 119

Cohort studies

Earlier publications from the SOS study of two-year and eight-year data on the incidence of diabetes and hypertension in 483 participants89 and two-year data on lipid disturbances in 1449 participants88 have been summarised previously15 and can be seen in Appendix 5. More recently reported data including 3505 participants at two-year follow-up and 1268 participants at 10-year follow-up97 are discussed here.

The incidence of diabetes (2 years: 1% versus 8%, p < 0.001; 10 years: 7% versus 24%, p < 0.001), hypertriglyceridaemia (2 years: 8% versus 22%, p < 0.001; 10 years: 17% versus 27%, p = 0.03) and hyperuricaemia (2 years: 4% versus 16%, p < 0.001; 10 years: 17% versus 28%, p < 0.001) was significantly lower in the surgery group at both 2 and 10 years follow-up (Table 19). The incidence of low HDL-cholesterol was significantly lower in the surgery group at two years (2% versus 10%, p < 0.001), but not 10 years (3% versus 6%, p = 0.12) follow-up, and there was no statistically significant difference in the incidence of hypertension (2 years: 24% versus 29%, p = 0.06; 10 years: 41% versus 49%, p = 0.13) and hypercholesterolaemia (2 years: 27% versus 24%, p = 0.11; 10 years: 30% versus 27%, p = 0.57) at 2 or 10 years follow-up. Participants who underwent surgery were more likely to recover from diabetes (2 years: 72% recovered versus 21%, p < 0.001; 10 years: 36% versus 13%, p < 0.001), hypertension (2 years: 34% recovered versus 21%, p < 0.001; 10 years: 19% versus 11%, p = 0.02), hypertriglyceridaemia (2 years: 62% recovered versus 22%, p < 0.001; 10 years: 46% versus 24%, p < 0.001), low HDL-cholesterol (2 years: 76% recovered versus 39%, p < 0.001; 10 years: 73% versus 53%, p < 0.001) and hyperuricaemia (2 years: 71% recovered versus 31%, p < 0.001; 10 years: 48% versus 27%, p < 0.001) than those with conventional treatment, at 2 and 10 years follow-up, but there was no statistically significant difference in recovery from hypercholesterolaemia between groups (2 years: 22% recovered versus 17%, p = 0.07; 10 years: 21% versus 17%, p = 0.14) (Table 19).

Of patients who were on diabetes medication at baseline, significantly fewer surgery patients were on diabetes medication at six years follow-up compared with controls [68.8% versus 100%, RR 0.71 (95% CI 0.56 to 0.89), p < 0.05]. This was also the case for patients who were not on diabetes medication at baseline [2.1% versus 11.3%, RR 0.20 (95% CI 0.10 to 0.38), p < 0.05]. 136

Of patients who were on CVD medication at baseline, significantly fewer surgery patients were on CVD medication at six years follow-up compared with controls [64.7% versus 86.4%, RR 0.77 (95% CI 0.67 to 0.88), p < 0.05]. Of patients who were not on medication at baseline, the difference in the proportion on medication was statistically significant at two years follow-up [surgery 3.1%, control 10.1%, RR 0.28 (95% CI 0.14 to 0.56), p < 0.05], but not at six years follow-up [surgery 13.3%, control 16.7%, RR 0.80 (95% CI 0.56 to 1.16)]. 136

At two years follow-up, men with surgery had more cholelithiasis [4.0% versus 1.2%, p = 0.011, OR 4.2 (95% CI 1.5 to 12.0)], cholecystectomy [3.4% versus 0.7%, p = 0.008, OR 5.4 (95% CI 1.5 to 19.6], cholecystitis [2.5% versus 0.7%, p = 0.058, OR 4.5 (95% CI 1.2 to 17.1)] and total biliary disease [4.1% versus 1.5%, p = 0.024, OR 3.5 (95% CI 1.3 to 9.2)] than male controls. There was no difference in pancreatitis among men, and there were no statistically significant differences in these diseases among women. 96

Incidence of cancer during an average of 11 years follow-up in the SOS study has been reported in a recent abstract. 101 In the surgery group (n = 2010) there were 126 cases of first-time cancers and in the control group (n = 2037) there were 173 cases. The unadjusted hazard ratio for overall cancer incidence was 0.71 (p = 0.003) and the hazard ratio adjusted for risk factors was 0.74 (p = 0.011). The unadjusted hazard ratio for men (n = 1178) was 0.98 (95% CI 0.63–1.51, p = 0.91) and for women (n = 2867) was 0.63 (95% CI 0.48–0.82, p = 0.001) (Appendix 5).

Only data on medication use were reported by Buddeberg-Fischer and colleagues. 105 In the last three months of the study those in the surgery group were found to use statistically significantly fewer numbers of different obesity-related drugs than those in the no-surgery group (p < 0.001). There were no statistically significant differences in the number of different medications being taken for somatic comorbidity or psychiatric comorbidity (Table 19). Caution is required in interpreting these outcomes as these were medications used only in the last three months of a study with a mean of 3.2 years follow-up, no measures of variance around the mean values were reported, results were based on self-reports and no further definition of the drugs was given.

Comorbidities were not reported by Stoeckli and colleagues. 102–104

Randomised controlled trials

O’Brien and colleagues found a higher proportion of adverse events among those people in the non-surgical therapy group (58%, n = 31) than in the laparoscopic adjustable gastric banding group (18%, n = 39)115,116 (Table 20). For those receiving non-surgical therapy the most common adverse events were intolerance to orlistat (26%), acute cholecystitis (13%), the need for operative interventions (13%) and intolerance to a VLCD (3%). Adverse events reported by people in the laparoscopic adjustable gastric banding group included operative interventions (13%), laparoscopic revision (prolapse or posterior) (10%), 5-mm port site infection (2.6%) and acute cholecystitis (2.6%). Loss to follow-up was higher in the non-surgical group (16%) compared with the laparoscopic adjustable gastric banding group (2.6%) (but reasons were not given).

Dixon and colleagues reported several adverse events among people in the laparoscopic adjustable gastric banding group (n = 30), including a superficial wound infection (one patient), gastric pouch enlargement requiring revisional surgery (two patients), eating difficulties and persistent regurgitation requiring band removal (one patient), postoperative febrile episode (one patient), hypoglycaemic episode (one patient) and gastrointestinal tract intolerance to metformin (one patient)117 (Table 20). People in the conventional therapy group (n = 30) suffered minor adverse events associated with their medication which resolved following discontinuation of treatment, including gastrointestinal problems (two patients), persistent diarrhoea with metformin (one patient), and vasculitic rash (one patient). Other adverse events included multiple hypoglycaemic episodes (one patient), angina and transient cerebral ischaemic episode requiring admission to hospital (one patient) and intolerance to very-low-calorie meal replacement (two patients). Dixon and colleagues noted that the mean procedure time for placement of the laparoscopic adjustable gastric banding was 54 minutes and that 80% of patients were kept in hospital for only one day.

Complications and additional operative procedures were not reported by Mingrone and colleagues. 119

Cohort studies

Within 90 days of surgery in the SOS study there were five deaths (0.25%) in the surgery group (four from peritonitis with organ failure, one from sudden death) and two deaths in the control group (one from pancreatic cancer, one from alcohol-related causes). 99

Perioperative complications were experienced by 13% of 1164 patients in the SOS study, these included bleeding (0.9%), thromboembolic events (0.8%), wound complications (1.8%), abdominal infection (2.1%), pulmonary symptoms (6.2%) and miscellaneous (4.8%) (Table 20). Postoperative complications requiring reoperation were experienced by 2.2% of patients in the surgery group. The patients in the SOS study underwent vertical banded gastroplasty, gastric banding or gastric bypass, but the complications are only reported for the surgery group as a whole. Surgical reoperations or conversions (excluding operations caused by postoperative complications) were reported for 1338 patients followed for at least 10 years, and occurred in 31% of gastric banding patients, 21% of vertical banded gastroplasty patients and 17% of gastric bypass patients.

The Buddeberg-Fischer and colleagues105 cohort study reported reoperations only. Seven of 69 participants in the surgery group were reported to have a reoperation; five participants with laparoscopic gastric banding were converted to Roux-en-Y gastric bypass and two had their bands removed. Nine patients in the no-surgery group underwent gastric bypass. No further details are reported.

Complications and additional operative procedures were not reported by Stoeckli and colleagues. 102–104

Summary

Three RCTs (one with a low risk of selection bias, one with a high risk of selection bias and one of uncertain risk of bias) and three cohort studies (of variable size and quality) were included. Regardless of the surgical intervention used or the type of patients included, all studies reporting a statistical comparison found statistically significant benefits on measures of weight change compared with no surgery at two to three years follow-up. A large cohort study found weight loss was still significantly greater at 10 years follow-up compared with conventional treatment. One RCT found statistically significantly greater improvements in five of eight domains of the SF-36 following laparoscopic adjustable gastric banding compared with no surgery, but one cohort study found no statistically significant difference in mean scores of the PSSQ between surgery and no-surgery groups. The SOS study found mixed results in HRQoL at 10 years follow-up, with significantly greater 10-year change following surgery observed in some measures, but not others. The RCT of people with Type 2 diabetes found significantly higher remission of the disease following laparoscopic adjustable gastric banding than conventional therapy, and two RCTs reporting metabolic syndrome found significantly fewer people with the syndrome two years after surgery. The SOS study found a statistically significant reduction in the incidence in three of six comorbidities assessed at 10-year follow-up after surgery compared with conventional therapy. Significantly fewer surgery patients than conventional therapy patients were on diabetes medication at two and six years follow-up, and on CVD medication at two, but not six years follow-up. Two RCTs reported adverse events from following surgery (e.g. operative interventions, revisional surgery, port-site infection) and from conventional therapy (e.g. intolerance to medication, acute cholecystitis, need for operative intervention, gastrointestinal problems). Within 90 days of surgery in the SOS study there were five deaths (0.25%) in the surgery group and two deaths in the control group. Perioperative complications occurred in 13% of patients.

Weight change

Three of the seven RCTs that compared vertical banded gastroplasty with gastric bypass (open or laparoscopic surgery) demonstrated statistically significantly greater percentage of excess weight loss with gastric bypass108,109,120,123 (Table 22). Howard and colleagues120 found that patients with open gastric bypass (n = 20) had 78% excess weight loss compared with 52% excess weight loss for patients undergoing open vertical banded gastroplasty (n = 22) (p < 0.05) at 12 months follow-up. At five years excess weight loss was 70% and 37% for gastric bypass and vertical banded gastroplasty, respectively (p < 0.05), although only six patients in each group were followed for this length of time. All gastric bypass patients had lost at least 50% of excess weight at 12 months and 60 months follow-up, whereas only 55% of vertical banded gastroplasty patients had achieved this at 12 months, and none by 60 months (p-value not stated, Appendix 6). Sugerman and colleagues123 found that excess weight loss for open gastric bypass was significantly greater than for open vertical banded gastroplasty at 12 months [68% (SD 17) versus 43% (SD 18), p < 0.001], 24 months [66% (SD 29) versus 39% (SD 24), p < 0.001] and 36 months [62% (SD 18) versus 37% (SD 19), p < 0.001]. As previously stated, this study has a high risk of bias because recruitment was stopped early when a significant difference (p < 0.05) in weight loss was noted in favour of gastric bypass. At this point 20 patients had been recruited to each arm of the study, and were followed up for three years. Before surgery, Sugerman and colleagues classified patients as ‘sweets eaters’ or ‘non-sweets eaters’. They noted that gastric bypass surgery led to a significantly greater excess weight loss for sweets eaters than did vertical banded gastroplasty (p < 0.0001). For non-sweets eaters gastric bypass caused greater decreases in excess weight compared with vertical banded gastroplasty, but differences were not significant [p = ns (not statistically significant)] (Appendix 6). The authors attribute this difference to the development of dumping syndrome symptoms in sweets eaters with gastric bypass. However, caution is required when interpreting these results because sample sizes were small and the comparisons were not randomised. A third study, Olbers and colleagues,108,109 which compared laparoscopic gastric bypass with laparoscopic vertical banded gastroplasty also reported that excess weight loss for gastric bypass was significantly greater than for vertical banded gastroplasty at 12 months [78.3% (SD 20) versus 62.9% (SD 28.4), p = 0.009], and at 24 months [84.4% (SD 22.1) versus 59.8% (SD 29.6), p < 0.001]. A greater proportion of patients (34 of 36) who received laparoscopic gastric bypass achieved an excess weight loss of at least 50% without remedial surgery, in comparison to those receiving laparoscopic vertical banded gastroplasty (21 of 35) (Appendix 6). Weight loss was also reflected in mean BMI values which had fallen to 29 in the laparoscopic Roux-en-Y gastric bypass group and 32 in the laparoscopic vertical banded gastroplasty group (p-value not reported), but there was little further change at two years when BMI values were 28 and 32, respectively (p-value not reported).

Two of the seven trials reported greater weight loss in the gastric bypass group, but did not indicate whether or not this difference was statistically significant (Table 22). The trial by Agren and Naslund,73 which was reported only as an abstract, found greater mean excess weight loss with loop gastric bypass (76.6%) than with vertical banded gastroplasty (59.8%) at 18 months follow-up. Lee and colleagues124 compared laparoscopic gastric bypass and laparoscopic vertical banded gastroplasty and reported that at two years follow-up the gastric bypass group had lower BMI (28.5 versus 31.9) and greater excess weight loss (71.4% versus 53.1%).

The remaining two trials that compared gastric bypass with vertical banded gastroplasty found that there was no significant difference in weight loss between the groups (Table 22). VanWoert and colleagues106 reported, in an abstract only, per cent of ideal body weight at 36 months to be 121% in the gastric bypass group, and 123% in the vertical banded gastroplasty group. Success rates, defined as a BMI < 35 or < 50% excess weight and reoperation not required, were compared for Roux-en-Y gastric bypass and vertical banded gastroplasty with surgical isolation of the gastric pouch by MacLean and colleagues. 121,122 When compared at three years and five to six years follow-up, there was no significant difference in success rates between gastric bypass and vertical banded gastroplasty [about three years: 58% versus 39% (p = 0.08); five to six years: 34% versus 16%, (p = 0.112)]. Failures were converted to isolated gastric bypass, which had a success rate of 63% at five to six years. Although comparisons of the three procedures were reported to show a significantly greater success rate for isolated gastric bypass compared with gastric bypass (p < 0.0009) and vertical banded gastroplasty (p = 0.0001), these were not valid because the periods of follow-up differed.

Quality of life and comorbidities

Only one RCT comparing gastric bypass with vertical banded gastroplasty reported on QoL measures, and this study performed both procedures laparoscopically. 124 Using the gastrointestinal quality of life index (GIQLI), patients in both groups had significant improvements in physical function, emotional function and social function domains, but patients with vertical banded gastroplasty had a significant decrease in the domain of symptoms (p-values not reported) (Table 23). Patients with gastric bypass scored significantly better on seven of 19 symptom items, four of five emotional items, one of seven physical items and two of five social items (p-values ranged from p = 0.04 to p < 0.001, see Appendix 6). Patients with vertical banded gastroplasty scored better on the symptom of abdominal flatulence (p = 0.02).

Data on comorbidities were not assessed.

Complications and additional operative procedures

Of the seven RCTs comparing vertical banded gastroplasty with gastric bypass surgery, five reported no deaths. 106,108,109,120–122,124 Three of these trials were comparisons of open procedures,106,120–122 the other two were comparisons of laparoscopic procedures. 108,109,124 One trial reported no deaths in the vertical banded gastroplasty group but two deaths (10%) in the gastric bypass group, occurring after three days and 13 months as the result of assumed arrhythmia. 123 Agren and Naslund73 reported one death, but did not state when this occurred or whether following gastric bypass, vertical banded gastroplasty or gastric banding (unbanded, intervention not included in systematic review) (Table 24).

Complications were not reported by all studies, but when they were reported they differed between the various surgical procedures (Table 24).

Of the two studies that compared laparoscopic procedures, Olbers and colleagues108,109 reported no conversions to open surgery, but Lee and colleagues124 reported that conversion to open surgery occurred in one (2.5%) patient with laparoscopic gastric bypass and none of the patients with laparoscopic vertical banded gastroplasty (p = ns). Operative time was significantly less with laparoscopic vertical banded gastroplasty in both studies [126 minutes (SD 38) versus 209 minutes (SD 50), p < 0.001124, and 105 minutes (SD 35) versus 138 minutes (SD 41), p < 0.001108,109]. In one study mean postoperative stay was also significantly shorter for the laparoscopic vertical-banded gastroplasty group [3.5 days (0.9) versus 5.7 days (2.2), p < 0.001124], whereas in the other study there appeared to be little difference in the length of hospital stay (LOS) [median three days (range 1–16 days) versus three days (range 2–15 days)108,109].

In one study124 early postoperative complications were significantly more common following laparoscopic gastric bypass (17.5%, seven patients) than laparoscopic vertical banded gastroplasty (2.5%, one patient). Three of the seven early postoperative complications experienced by laparoscopic gastric bypass patients were major complications. These required interventional management and hospitalisation for over 14 days. Two with anastomotic leakage required reoperation, the third had an abdominal abscess. The remaining four early postoperative complications in this group were classed as minor and included upper gastrointestinal bleeding, a sutured nasogastric tube, and minor leakage from a drainage tube. Only one early postoperative complication occurred in the laparoscopic vertical banded gastroplasty group and this was a wound infection that was classed as a minor complication. Analgesic use was also higher [2.4 units (3.0) versus 1.4 units (1.5), respectively (p < 0.05)] in this study. 124 The second study that compared laparoscopic procedures108,109 reported five early reoperations in the gastric bypass group (three for haemorrhage, one for stenosis and one for suspected leak) and one early reoperation in the vertical banded gastroplasty group (for suspected leak), but this difference was not significant (p = 0.080). A statistical comparison of the perioperative complications reported by Olbers and colleagues108,109 (in addition to the reoperations) is not provided, but these appear similar (gastric bypass group: two minor bleeding, one deep infection; vertical banded gastroplasty group: four minor bleeding, one deep infection) (Table 24). Neither group experienced a thrombotic complication, and there was no difference in the incidence of pulmonary complications between the groups (p = 0.888). One patient in the laparoscopic Roux-en-Y gastric bypass group experienced an intra-abdominal abscess after discharge.

Readmission for late complications reported by Lee and colleagues124 was similar between the laparoscopic procedures (10% versus 5%, p = ns). Late complications associated with laparoscopic gastric bypass included marginal ulcer requiring blood transfusion (two patients), anastomotic stricture (one patient), and pyothorax (pus in the chest cavity) (one patient). Two patients (5%) with laparoscopic vertical banded gastroplasty experienced the late complication of reflux oesophagitis, one of which required laparoscopic revision surgery. Olbers and colleagues,108,109 however, reported that remedial surgical intervention was required for eight participants in the vertical banded gastroplasty group (four in the first year and four in the second year, all conversion to Roux-en-Y gastric bypass, due to: one migration of outlet restricting band, five vomiting and insufficient weight loss, two vomiting and excessive weight loss), but for none in the gastric bypass group.

In the comparisons of open procedures73,106,120–123 gastric bypass patients suffered from symptomatic ulcer disease (25% of patients),120 staple-line fistula (23%),121,122 stomal ulcers (13% of patients),121,122 intractable vomiting and stomal stenosis (25%),123 marginal ulcer of jejunal side of gastrojejunostomy (5%),123 cholelithiasis (gallstone formation) (13%),106 and peptic gastro-oesophagitis (33%). 106 Vertical banded gastroplasty patients suffered stenosis (20%),121,122 enlarged orifice (13%),121,122 staple-line fistula (4%),121,122 clinical failure (4%),121,122 abscess (2%),121,122 superficial stomal erosions (5%),123 cholelithiasis (24%)106 and peptic gastro-oesophagitis (18%). 106 Intraoperative cholecystectomy and postoperative cholecystectomy were reported for gastric bypass (20% and 29% respectively) and vertical banded gastroplasty (14% and 29% respectively) (Table 24). 120

In the trial by MacLean and colleagues,121,122 failures of vertical banded gastroplasty (due to stenosis and enlargement of the gastroplasty orifice), and failures of gastric bypass (due to perforation of the vertical staple line), were converted to normal (9% versus 2%) or isolated gastric bypass (44% versus 37%). In another trial by Sugerman and colleagues,123 a total of four (20%) vertical banded gastroplasty patients were converted to Roux-en-Y gastric bypass. One patient was converted at one month due to staple-line disruption, and a further three patients were converted due to a failure to lose weight, one at 18 months, and two at 38 months following surgery. After 2 years there were no significant deficiencies in most vitamins, electrolytes, renal or liver function tests. However, vitamin B₁₂ levels were lower in patients with Roux-en-Y gastric bypass than those with vertical banded gastroplasty (286 pg/ml versus 461 pg/ml, p <0.05) (Appendix 6). Agren and Naslund73 found just one patient (4%) with vertical banded gastroplasty required reoperation for staple-line disruption. Agren and Naslund73 reported that morbidity was low and not significantly different between the groups, but no data were presented.

Summary

Seven RCTs were included. Six trials were of uncertain risk of bias as many factors were not reported. One study by Sugerman and colleagues had a high risk of bias because of an a priori stopping rule. On measures of weight, participants who underwent gastric bypass had a better outcome, particularly at later time points, in five of the seven trials than those who underwent vertical banded gastroplasty (although a statistically significant difference was only reported in three of these trials). Only Lee and colleagues reported on QoL, which they found to be better for the gastric bypass group. Data on comorbidities were not assessed. Reporting of complications varied between studies. Evidence from two studies108,109,124 suggests that the laparoscopic vertical banded gastroplasty surgery is quicker, and is associated with fewer early postoperative complications than laparoscopic gastric bypass surgery. However, two studies of open surgery121–123 report that after approximately three years, conversions to an alternative bariatric procedure occurred more often in the vertical banded gastroplasty groups, but neither study tested this for statistical significance.

Weight change

At one and two years follow-up, Bessler and colleagues118 reported per cent excess weight loss was greater for participants undergoing banded than non-banded long limb gastric bypass, but the differences were not statistically significant (Table 25). After three years, the per cent excess weight loss was statistically significantly higher in the banded gastric bypass group compared with the non-banded gastric bypass group (p < 0.05). There is however an uncertain risk of bias for this outcome as the paper reports that this was based on the small number of patients reaching the 36-month follow-up period (numbers not presented). The proportion achieving a BMI of less than 35 was higher in the banded gastric bypass groups than the non-banded gastric bypass groups at one and two years follow-up but the difference was not statistically significant. No rates were reported for the three-year follow-up.

Quality of life and comorbidities

Data on QoL were not assessed. Comorbidities were reported by Bessler and colleagues. 118 The study reported baseline values for stated comorbidities and the proportion with resolution of these comorbidities; however, the study does not state which follow-up period is being reported, or give any numerators and denominators for these calculations.

The proportion of participants with resolution of their respective comorbidities was not shown to be statistically significantly different between the two interventions (Table 26).

Complications and additional operative procedures

There were no deaths reported in the study by Bessler and colleagues118 and no statistically significant differences between banded gastric bypass and non-banded gastric bypass in rates of any postoperative complications. Postoperative gastrointestinal symptoms were scored according to a subjective scale where 0 = none, 1 = mild, 2 = moderate, 3 = severe. In this study only the scores for emesis were statistically significantly better (lower score) in the non-banded gastric bypass group than in the banded gastric bypass participants (0.35 versus 0.13 respectively, p = 0.0002) (Table 27). No other gastrointestinal symptoms were statistically significantly different between the two interventions. The proportion of participants with food intolerance was higher in the banded gastric bypass group (79% versus 33%, p < 0.05). Care is required in the interpretation of these results, not only because of the subjective nature of these outcomes, but also because the study does not state which follow-up period is being reported or what the participant numbers were for these outcomes.

Summary

One study (with a low risk of selection bias but at high risk of bias through selective reporting) found similar weight loss after banded and non-banded gastric bypass in people with BMI greater than 50. Although a statistically significant difference in BMI was found at 36 months follow-up, this was based on a small number of participants. Rates of improvement of existing comorbidities were not different between the two groups and complications were also generally similar. The evidence suggests that in these high BMI participants there is little added benefit from the addition of banding to the intervention.

Weight change

At one and three years of follow-up Angrisani and colleagues107 found that the per cent excess weight loss was greater for those undergoing laparoscopic Roux-en-Y gastric bypass than laparoscopic adjustable gastric banding, but no statistical significance was reported (Table 28). After 5 years of follow-up the per cent excess weight loss and mean weight were statistically significantly better for participants in the laparoscopic Roux-en-Y gastric bypass group compared with those in the laparoscopic adjustable gastric banding group (p < 0.001 for both outcomes). At one and three years follow-up the mean BMI was also lower in the laparoscopic Roux-en-Y gastric bypass group (but no p-values were reported) than the laparoscopic adjustable gastric banding group. After five years the mean BMI was statistically significantly better in the laparoscopic Roux-en-Y gastric bypass group than the laparoscopic adjustable gastric banding group (p < 0.001). The proportion classed as weight loss failures, defined as those with a BMI > 35 at 5 years, were statistically significantly lower in the laparoscopic Roux-en-Y gastric bypass than the laparoscopic adjustable gastric banding group (p < 0.001) and the proportion with a BMI of less than 30 at 5 years were statistically significantly higher in the laparoscopic Roux-en-Y gastric bypass group (p < 0.001) than in the laparoscopic adjustable gastric banding group.

Quality of life and comorbidities

QoL was not assessed by Angrisani and colleagues. 107 Baseline rates of comorbidities were low in this study, with two participants in the laparoscopic Roux-en-Y gastric bypass group having hyperlipaemia, one hypertension and one Type 2 diabetes. In the laparoscopic adjustable gastric banding group three participants had hypertension and one sleep apnoea. The study reports that at five years there was resolution of the diabetes and hyperlipaemia (in the laparoscopic Roux-en-Y gastric bypass group) and the sleep apnoea (in the laparoscopic adjustable gastric banding group) (Appendix 8).

Complications and additional operative procedures

No deaths were noted in either group of the Angrisani and colleagues study119 (Table 29). Operative time for laparoscopic adjustable gastric banding was significantly shorter than for laparoscopic Roux-en-Y gastric bypass [mean 60 (SD 20) minutes versus mean 220 (SD 100) minutes, p < 0.001]. Four (15.2%) participants in the laparoscopic adjustable gastric banding group underwent a reoperation (two pouch dilatation, two band removal because of inadequate weight loss: one of these was converted to biliopancreatic diversion, three waiting list for laparoscopic Roux-en-Y gastric bypass) and three (12.5%) from the laparoscopic Roux-en-Y gastric bypass group underwent a reoperation (for potentially lethal complications not further specified). Early complications requiring surgery in the laparoscopic Roux-en-Y gastric bypass group occurred in two (8.4%) [one posterior pouch leak intraoperatively causing conversion to open surgery, one (4.2%) sepsis caused by jejunal perforation (sutured and intestine resected)]. No early complications requiring surgery were noted in the laparoscopic adjustable gastric banding group. One participant undergoing laparoscopic Roux-en-Y gastric bypass had a late complication (small bowel obstruction as the result of internal hernia) and two participants undergoing laparoscopic adjustable gastric banding had gastric pouch dilatation (which was treated by band removal). No significance testing was undertaken for reoperation rates, or early or late complication rates between groups. Mean hospital stay was statistically significantly longer in the laparoscopic Roux-en-Y gastric bypass group than the laparoscopic adjustable gastric banding group [4 days (SD 2) versus 2 days (SD 1) for the two groups respectively, p < 0.05], although one participant in the laparoscopic Roux-en-Y gastric bypass group required an intensive care stay of 40 days (Appendix 8).

Summary

On a variety of measures of weight this small study showed that laparoscopic Roux-en-Y gastric bypass was superior to laparoscopic adjustable gastric banding. Reoperation rates were similar between the two interventions, and comorbidities were similar and few. Early complications requiring reoperation were seen in more participants undergoing laparoscopic Roux-en-Y than laparoscopic adjustable gastric banding, but the numbers were small and not tested for statistical significance. The risk of bias for this study is uncertain, although the risk of bias from incomplete outcome data for weight loss and comorbidities is likely to be low.

Weight loss

There were no statistically significant differences in BMI or weight loss between the two procedures at 12 months follow-up. However, laparoscopic sleeve gastrectomy led to a greater per cent excess weight loss at 12 months [69.7% (SD 14.6) versus 60.5% (SD 10.7), p = 0.05] in this small RCT (Table 30).

Quality of life and comorbidities

QoL was not assessed by this study. Two patients had diabetes at baseline (both in the laparoscopic Roux-en-Y gastric bypass group); this was resolved at 12 months in both patients. The outcome of other comorbidities noted at baseline (see Appendix 9) was not reported.

Complications and additional operative procedures

Karamanakos and colleagues reported that there were no conversions to open surgery and no intraoperative and postoperative complications. No further details were reported.

Summary

In this small RCT with uncertain risk of bias, BMI and weight loss at 12 months follow-up were similar between laparoscopic Roux-en-Y gastric bypass and sleeve gastrectomy. per cent excess weight loss was greater with sleeve gastrectomy at 12 months (p = 0.05).

Weight change

At one year follow-up, Nilsell and colleagues126 found that weight loss was greater for the vertical banded gastroplasty group, but these patients then began to regain weight. The patients with adjustable gastric banding experienced lower initial weight loss, but this continued over five years resulting in a weight reduction of 43 kg at five years compared with 35 kg for vertical banded gastroplasty (statistical significance not given) (Table 31).

When comparing laparoscopic procedures, Morino and colleagues110 also found significantly lower BMI with vertical banded gastroplasty at one year follow-up (30.1 versus 35.5, p < 0.05), but at two and three years follow-up the difference was not statistically significant (two years: 29.7 versus 34.8; three years: 30.7 versus 35.7). A similar result was seen in terms of percentage excess weight loss which was significantly greater with vertical banded gastroplasty at one year follow-up (62.3% versus 39.2%, p < 0.05), but was not statistically significantly different at later follow-up (two years: 63.5% versus 41.4%, three years: 58.9% versus 39.0%, p-values not reported). At three years follow-up, 25% of patients with laparoscopic adjustable silicone gastric banding and 63% of patients with laparoscopic vertical banded gastroplasty had an excellent or good result (residual excess weight < 50%, p = 0.056).

Van Dielen and colleagues127,128 reported a comparison of open vertical banded gastroplasty and laparoscopic adjustable gastric banding at one and two years follow-up, with data at a mean of 84 months follow-up reported in a recent abstract128 (Table 31). At one year, mean BMI was lower in the open vertical banded gastroplasty group than in the laparoscopic adjustable gastric banding group [31.1 (SD 6.2) versus 35.0 (SD 6.3), no p-value reported] and statistically significantly lower at the two-year follow-up [31.0 (SD 6.0) versus 34.6 (SD 6.5), p ≤ 0.002]. Percentage excess weight loss was statistically significantly greater in the open vertical banded gastroplasty group in comparison to the laparoscopic adjustable gastric banding group at both one and two years [one year: 71.1% (SD 24.0) versus 53.3% (SD 21.2), p ≤ 0.001; two years: 70.1% (SD 25.5) versus 54.9% (SD 23.3), p ≤ 0.001]. The difference was maintained at a mean of 84 months follow-up (per cent excess BMI loss 68.8% versus 56.9%, respectively). 128 However, it should be noted that at this time point 59% of vertical banded gastroplasty participants had been converted to gastric bypass, and 11% of laparoscopic adjustable gastric banding participants had been converted to another procedure. From the limited information presented in the abstract reporting these results128 it is not clear whether this has been taken into consideration.

Quality of life and comorbidities

One study assessed patient satisfaction126 and one study reported on comorbidities,127 whereas one study did not report QoL or comorbidity outcomes. 110 At five years follow-up, patients were asked if they were satisfied with, or regretted having undergone, the operation. 126 Only 56% of vertical banded gastroplasty patients were satisfied with the result of the operation, while 81% of the patients with adjustable gastric banding were satisfied (statistical significance not given). The comparison of open vertical banded gastroplasty and laparoscopic adjustable gastric banding127 revealed that although the overall number of patients with comorbidity in both groups significantly decreased following surgery at one and two years follow-up, no differences in comorbidities were observed between groups (open vertical banded gastroplasty: 82% at baseline, 30.4% at one year, 47.9%, at two years, p ≤ 0.001 versus baseline; laparoscopic adjustable gastric banding 78% at baseline, one year 37.5% at one year, 40% at two years, p ≤ 0.001 versus baseline) (Table 32). Significant improvements in both groups were seen in joint problems (p ≤ 0.001 compared with preoperative), pulmonary problems (p ≤ 0.05 compared to preoperative) and diabetes mellitus (p ≤ 0.05 compared with preoperative). No improvement in either group was observed between the percentage of patients with preoperative and postoperative hypertension, cardiovascular problems, hypercholesterolaemia, reflux disease, sleep apnoea or neurological problems (Appendix 10). After a mean of 84 months follow-up comorbidities had all significantly decreased except for gastro-oesophageal reflux disease (GERD), which increased in both groups, although no numerical data are provided in the abstract reporting these results. 128

Complications and additional operative procedures

Van Dielen and colleagues127 report that two deaths (4%) occurred, both in the open vertical banded gastroplasty group, one from sepsis in the first postoperative week, and one as the result of a pre-existing pneumonia that had not been reported to the surgeon preoperatively. No postoperative deaths occurred in the other two studies,110,126 and although one patient from each group died during follow-up in the study by Nilsell and colleagues,126 these deaths are reported to be unrelated to the surgery (Table 33).

Operative time was shorter with laparoscopic adjustable silicone gastric banding than laparoscopic vertical banded gastroplasty [65.4 minutes (35–120) versus 94.2 minutes (40–270), p < 0.05], as was hospital stay [3.7 days (2 to 6), versus 6.6 days (3 to 58), p < 0.05] in the Morino and colleagues study. 110 Hospital stay was also shorter for laparoscopic adjustable gastric banding than open vertical banded gastroplasty [mean 3.5 days (SD 1.5), range 2–9 days, versus mean 6.8 days (SD 10.4) days, range 2–56 days; p < 0.001] in the study by van Dielen and colleagues. 127

One reoperation on the third postoperative day because of an anastomotic leak in the vertical banded gastroplasty group is reported by Nilsell and colleagues. 126 Morino and colleagues110 found that early morbidity was similar between the two laparoscopic procedures (laparoscopic adjustable gastric banding 6.1% versus laparoscopic vertical gastric banding 9.8%, p = 0.754) and there were no conversions to open surgery. One patient experienced early postoperative band slippage (laparoscopically repositioned). There was also one port infection and one haematoma at the port site in the laparoscopic adjustable gastric banding group. In the vertical banded gastroplasty group there was one fistula at the staple line (treated with open gastric bypass), two instances of prolonged postoperative pyrexia, and two respiratory failures without evidence of pulmonary embolism. Van Dielen and colleagues127 reported that two conversions to open surgery were necessary in the laparoscopic adjustable banding group, and in an additional patient a conversion to gastric bypass was necessary. In the open vertical banded gastroplasty group nine patients experienced immediate postoperative complications: leakage in three patients required reoperation, two splenectomies were performed, and in two patients an obstruction that necessitated gastroscopy occurred. There were no infections in the laparoscopic adjustable banding group, but there were seven infections in five patients (including the two patients who died) in the open vertical banded gastroplasty group.

Nilsell and colleagues126 reported that late reoperations were necessary in a third of vertical banded gastroplasty patients because of staple-line disruption with rapid weight regain or strictures of the stoma with vomiting or intolerance of solid food. The total incidence of staple-line disruption was 18.5% (five patients), but three patients were not reoperated on for various reasons. 126 Three (11.5%) adjustable gastric banding patients were reoperated; two as the result of dilatation of the gastric pouch, and one patient requested that their band be removed for reasons that were unclear. Morino and colleagues,110 found that late complications were more common following laparoscopic adjustable silicone gastric banding (32.7% versus 14%, p < 0.05). However, no patients with laparoscopic vertical banded gastroplasty required late reoperation, whereas 24.5% of patients required late reoperation following laparoscopic adjustable silicone gastric banding (p < 0.001), most commonly because of bands slipping. The percentage of late complications occurring in the open vertical banded gastroplasty group and the laparoscopic gastric bypass group appear similar in van Dielen and colleagues127 (no statistical comparison reported). Revisional surgery (conversion to gastric bypass) was necessary in 36% of those who had undergone open vertical banded gastroplasty, in most cases (15/18) this was the result of vertical staple-line disruption. A further eight (16%) patients required a surgical repair of an incisional hernia. Six patients (12%) required gastroscopy at least once for outlet stenosis or obstruction. In the laparoscopic adjustable gastric banding group, 40% of participants required reoperation. The majority of reoperations (16/20) were major reoperations for pouch dilatation or slippage, band leakage, or band erosion. At a mean of 84 months follow-up, the authors reported that long-term complications ‘were mainly’ staple-line disruption (51%) and incisional hernia (27%) for patients with open vertical banded gastroplasty, and pouch dilatation (24%) and anterior slippage (15%) for patients with laparoscopic adjustable gastric banding. 128 Major reoperation was required in 59% of vertical banded gastroplasty patients, who required conversions to gastric bypass, and in 46% of laparoscopic adjustable gastric banding patients, who required refixation or band replacements (35%) or conversion to another procedure (11%). 128

GERD was slightly more common in patients with vertical banded gastroplasty (14.8%) than adjustable gastric banding (11.5%) in the study by Nilsell and colleagues. 126

Summary

Three studies were included: one had a low risk of selection bias and two were of uncertain risk of bias for several items, although missing outcome data for weight loss were adequately addressed. Weight loss results were inconclusive. One study found that weight loss was initially greater with vertical banded gastroplasty, but weight regain meant that by three years patients with adjustable gastric banding had a lower mean weight, and this was still the case at five years (statistical significance not reported). The second study found lower BMI and greater per cent excess weight loss following laparoscopic vertical banded gastroplasty, but this was statistically significant only at year one and not at years two or three. The third study found statistically significant lower BMI and greater per cent excess weight loss at one and two years following open vertical banded gastroplasty, and greater per cent excess BMI loss seven years after open vertical banded gastroplasty (statistical significance not reported). However, the impact of participants being converted to another procedure in this study is unclear. More patients who had undergone adjustable gastric banding reported being satisfied with the results at five years, but this apparent superiority was not tested statistically. Resolution of comorbidities was similar in the two groups in the only study that reported on this outcome. 127 It is difficult to draw conclusions regarding complications and additional operative procedures because one study compared two open procedures, one two laparoscopic procedures, and one compared an open procedure with a laparoscopic procedure. In the comparison of open procedures more reoperations were necessary following open vertical banded gastroplasty than open adjustable gastric banding, but a statistical comparison was not reported. Laparoscopic adjustable gastric banding was associated with a statistically shorter operative time and hospital stay than laparoscopic vertical banded gastroplasty, but there were statistically more late complications and reoperations. The third comparison found laparoscopic adjustable gastric banding was associated with a statistically significant shorter hospital stay while open vertical banded gastroplasty led to more infections. Late complications requiring further surgery were similar but a statistical comparison was not reported.

Weight loss

In this study Himpens and colleagues129 report that the proportion of excess weight loss was statistically significantly greater in participants in the laparoscopic isolated sleeve gastrectomy group than the laparoscopic adjustable gastric banding group at one year (57.7% versus 41.4%, p = 0.0004) and three years (66% versus 48% p = 0.0025) (Table 34). Weight loss (three years: 29.5 kg versus 17 kg, p < 0.0001) and reduction in BMI (three years: 27.5 versus 18, p < 0.0001) were statistically significantly improved in laparoscopic isolated sleeve gastrectomy participants in comparison with the laparoscopic adjustable gastric banding participants. All of these data were presented by the trial authors as median and range so care should be taken when interpreting the results.

Quality of life and comorbidities

QoL was not assessed in this study. At baseline in the Himpens and colleagues129 study GERD requiring drug therapy with proton pump inhibitors was a problem for 15% (6/40) of the laparoscopic adjustable gastric banding participants and 20% (8/40) of the participants in the laparoscopic isolated sleeve gastrectomy group. After one year GERD had disappeared in 83% and 75% of these participants in the two groups respectively, and this remained the same at three years. Statistical significance was not reported. In those without GERD at baseline, no statistically significant differences in rates of appearance of GERD between the intervention groups were observed at one year [laparoscopic adjustable gastric banding 3/34 (8.8%), versus laparoscopic isolated sleeve gastrectomy 7/32 (21.8%), p = ns] or three years [laparoscopic adjustable gastric banding 7/34 (20.5%) versus laparoscopic isolated sleeve gastrectomy 1/32 (3.1%), p = ns] (Appendix 11).

Complications and additional operative procedures

No early postoperative complications were seen in the laparoscopic adjustable gastric banding group of the Himpens and colleagues129 study, two participants in the laparoscopic isolated sleeve gastrectomy group had an early postoperative complication, both required revisional surgery, in one this was a total gastrectomy for gastric ischaemia (Table 35). Late complications requiring surgery were observed in the laparoscopic adjustable gastric banding participants, with three pouch dilatations (treated with band removal in two and conversion to Roux-en-Y gastric bypass in one); one gastric erosion (treated with Roux-en-Y gastric bypass) and three disconnections of the port (treated with reconnection). There were no late complications requiring surgery in the laparoscopic isolated sleeve gastrectomy group. Complications not requiring surgery that were observed at one and three years can be seen in Table 35. There appeared to be higher frequencies of complications in the laparoscopic adjustable gastric banding group than in the laparoscopic isolated sleeve gastrectomy group but this is based on observation of the data only, no statistical analysis was undertaken.

In addition, two participants in each group had ‘insufficient weight loss’ noted as a complication in the Himpens and colleagues129 study. The two participants in the laparoscopic adjustable gastric banding group were converted to Roux-en-Y gastric bypass and the two participants in the laparoscopic isolated sleeve gastrectomy group were converted to laparoscopic duodenal switch.

Summary

On measures of weight, participants undergoing laparoscopic isolated sleeve gastrectomy showed more improvement than participants undergoing laparoscopic adjustable gastric banding in one study with an uncertain risk of bias. Rates of complications were observed to be lower in the laparoscopic isolated sleeve gastrectomy group, apart from rates of early postoperative complications.

Weight change

Puzziferri and colleagues113,114 demonstrated a slightly higher percentage of excess body weight loss following laparoscopic gastric bypass (one year: 68%, SD 15) compared with open gastric bypass (one year: 62%, SD 14), but the difference was not statistically significant at one year (p = 0.07) or at the three-year or four-year follow-ups114 (Table 36). Similarly, a non-significant difference in reduction of BMI was found at one year by Westling and Gustavsson131 [laparoscopic 14 (SD 3), open 13 (SD 3)], Sundbom and Gustavsson112 (BMI reduction of 15 in both groups, BMI: laparoscopic 29, open 30) and at three years follow-up by Lujan and colleagues130 (laparoscopic 31, open 35.5). Incomplete weight loss outcome data were adequately addressed by Westling and Gustavsson131 and by Sundbom and Gustavsson. 112 However, the risk of bias from incomplete weight loss data was uncertain in the studies by Lujan and colleagues130 and Puzziferri and colleagues. 113,114

Quality of life and comorbidities

The one- to six-month QoL outcomes for Puzziferri and colleagues113 have been tabulated and discussed previously,15 and can be seen in Appendix 12. In brief, early differences in some components of the SF-36 score (at one month) and Moorehead–Ardelt quality of life questionnaire (MAQoL) (at three months) were no longer significant at three months or six months respectively. At the three-year follow-up there continued to be no significant difference in MAQoL scores, and there was also no significant difference in the proportion of participants with Bariatric Analysis and Reporting Outcome System (BAROS) scores of good, very good or excellent (Table 37).

Westling and Gustavsson131 reported that 92% of all patients described themselves as ‘very satisfied’ with the result of the operation after one year, while the remaining patients described themselves as ‘satisfied’. The authors report no significant difference between the groups, but data were not provided.

Only one study reported on comorbidities. 114 At the three-year follow-up the improvement of obesity-related comorbidities was significantly different between the two groups for only two of the reported comorbidities: osteoarthritis symptoms improved more in the laparoscopic gastric bypass group than the open gastric bypass group (p < 0.05), whereas the open gastric bypass group experienced a greater improvement of dyslipidaemia (p < 0.01). For other obesity-related comorbidities improvements were not statistically significantly different between the groups (Table 38).

Complications and additional operative procedures

One postoperative death in the laparoscopic gastric bypass group was reported by Westling and Gustavsson;131 this was the result of malignant hyperthermia which developed during surgery leading to the death of the patient one week later. Lujan and colleagues130 reported three postoperative deaths. Two occurred in the laparoscopic group, one on postoperative day 32, and the other (unrelated to surgery) six months after surgery. One death occurred in the open surgery group within the first 30 postoperative days. No postoperative deaths occurred in the open versus laparoscopic113,114 or open versus hand-assisted laparoscopic112 gastric bypass studies (Table 39).