Bariatric surgery, lifestyle interventions and orlistat for severe obesity: the REBALANCE mixed-methods systematic review and economic evaluation

Systematic reviews

We undertook four systematic reviews, which were of:

1. RCTs of weight-loss or weight-maintenance programmes (including orlistat or comparisons between usual care/controls or WMPs and bariatric surgery) for adults with obesity (with a BMI of ≥ 35 kg/m²) with follow-up durations of ≥ 1 year, in any setting (see Chapter 3)

2. UK interventions of weight loss or weight maintenance of any study design with people with a BMI of ≥ 35 kg/m² with follow-up durations of ≥ 1 year, in any setting (see Chapter 4)

3. qualitative and mixed-methods research exploring adults’ experiences of living with obesity and receiving weight-loss or maintenance interventions for obesity (BMI of ≥ 35 kg/m²) (including research exploring the views of professionals involved in their care) (see Chapter 5)

4. economic evaluations of weight-loss and weight maintenance interventions for adults with a BMI of ≥ 35 kg/m² (see Chapter 6).

Primary outcome

The primary outcome was weight change.

Secondary outcomes

The secondary outcomes were:

• change in BMI

• waist circumference

• cardiovascular risk factors [total cholesterol, high-density lipoprotein (HDL) cholesterol, low-density lipoprotein (LDL) cholesterol, triglycerides, fasting glucose, glycated haemoglobin (HbA_1c) and systolic and diastolic blood pressure]

• disease-specific outcome measures (e.g. development of type 2 diabetes mellitus and reduction in sleep apnoea)

• changes in medication and adverse events

• psychological well-being

• adverse events

• quality of life

• process outcomes (e.g. staff involvement, setting, type of intervention, timing, frequency, individual and/or group setting, couple or family setting, proportion recruited and dropping out and participants’ evaluations)

• costs and economic evaluations.

Organisational issues

• What interventions are effective in increasing uptake of weight-loss services? (See Chapters 4 and 5.)

• What factors seem to affect people’s choices of programme? (See Chapters 3–5.)

• Does the effectiveness of programmes vary by socioeconomic status, ethnic group (including south-east Asian), BMI or other patient group? Is uptake related to these factors? (See Chapters 3, 4 and 6.)

• Do programmes incorporate users in design, evaluation or delivery, and how do they do this? (See Chapters 3–5.)

• What do people think about commercial weight-loss organisations? What is the influence of paying for their services? (See Chapter 5.)

• What is the effectiveness of weight maintenance interventions? (See Chapter 3.)

• Is an intensive/inpatient programme of value at the start of the WMP, particularly for people with very high BMI? (See Chapter 3.)

Diets

• Which factors (type of diet, interventions for adherence, types of classes and characteristics of the participants) help explain weight loss and maintenance? (See Chapters 3–5.)

• Does having a choice of reducing diet matter? Do people pick diets that are more effective for them? (See Chapters 4 and 5.)

Physical activity

• Which factors (type of physical activity programme, interventions for adherence, types of classes and characteristics of the participants) help explain weight loss and maintenance? (See Chapters 3 and 4.)

• Should physical activity classes and programmes designed specifically for people with a BMI of ≥ 35 kg/m² be provided? (See Chapter 5.)

Intervention characteristics

• Are certain psychological theories and BCTs more useful to people with severe obesity for losing weight and maintaining that weight loss? (See Chapters 3 and 5.)

• Are group-based interventions more effective for weight loss than those delivered to individuals? (See Chapters 3, 5 and 6.)

• What is the effect of financial/other incentives on participation, completion and weight loss? (See Chapters 3 and 4.)

• Are programmes involving partners/families/friends more effective? (See Chapters 3 and 5.)

• Do people say who they prefer to be the best person(s) to help with/deliver the weight-loss programme? Does this agree with what the interventions show? (See Chapter 5.)

• How often should people be seen/contacted, and for how long? (See Chapters 3 and 5.)

• What role does information technology have, particularly for monitoring and long-term follow-up? (See Chapters 3, 5 and 6.)

Drug therapy

• Is current drug treatment(s), including long-term results post treatment, useful/cost-effective for people with severe obesity? (See Chapters 3–6.)

Surgery

• Should more bariatric surgery be offered at the expense of lifestyle WMPs, if no extra funding is available? (See Chapters 3, 6 and 7.)

Risk of bias

It was unclear whether randomisation and allocation concealment were adequate for the majority of the 19 studies, or whether health-care providers and outcome assessors were blinded to treatment allocation. Only one trial was judged to be at a low risk of bias owing to blinding of participants receiving orlistat or placebo for weight maintenance following a VLCD weight-loss period. Only three trials were judged to be at low risk of bias owing to blinding of outcome assessors. Just over half of the trials were considered to be at either a high or an unclear risk of bias for incomplete outcome data or selective reporting. Nine studies were judged to be at high risk of bias, mainly owing to financial associations with the manufacturers of the VLCD products used in the trials (see Report Supplementary Material 4, Figure E4).

Equity

Four out of the nineteen trials targeted a specific population: two150,200 targeted people with diabetes mellitus or other obesity-related comorbidities and two161,172 targeted people receiving outpatient care. Just over half of the trials were considered representative of people with severe obesity. Those that were considered unrepresentative mainly excluded potential participants with obesity-related comorbid conditions. Most trials (79%) reported one or more PROGRESS characteristic. None of the studies clearly reported strategies to address diversity or disadvantage. Three studies107,161,186 reported whether or not there were sociodemographic differences between trial completers and those who withdrew/were excluded. Only one trial161 reported conducting a fidelity check and only two82,200 reported collecting process measure data. Sustainability, partnerships and political or organisational contexts of the interventions were generally poorly reported. Only eight trials76,93,155,166,171–173,184 reported harms or unintended effects, and only one trial161 was considered to have no potential for author conflict (see Report Supplementary Material 4, Figure E5).

Risk of bias

Seven RCTs93,155,172,186,193,198,200 examined adding a VLCD to a dietary intervention. The method of randomisation was unclear for all trials, and adequacy of allocation concealment was unclear for all but one study. 186 Four93,155,172,186 out of the seven RCTs were considered to be at high risk of bias owing to financial associations with the manufacturers of the products used (see Report Supplementary Material 4, Figure E6).

Equity

The trials mainly recruited women, who represented 55.8%172 to 100%193 of participants. The trials did not report on ethnicity or socioeconomic status. Torgerson et al. 186 reported the educational attainment of participants, with the majority having attained intermediate or higher education. Just over half of the trials93,155,172,186 were conducted in a hospital outpatient setting, with the remainder conducted in university settings. 198,200 The weighted mean age of all participants in the trials was 49.9 years. The youngest reported mean age of the trial groups was 36.8 years (SD 8.9 years) and the oldest was 64.1 years (SD 10.5 years). Participants had a weighted mean weight of 105.9 kg and a weighted mean BMI of 37.9 kg/m². The lowest and highest mean weights for the groups were 92.1 kg and 116.6 kg, respectively, and the lowest and highest mean BMIs were 35.1 kg/m² and 40.5 kg/m², respectively (see Report Supplementary Material 4, Figure E7).

Descriptions of the trials

Bliddal et al. 93 evaluated a VLCD for improving the symptoms of primary knee osteoarthritis. Participants were randomised to receive either dietary instruction with fewer contacts or a VLCD for 12 weeks over 1 year with frequent consultations with a dietitian. Participants randomised to the VLCD received a formula diet [Speasy (Dansk Droge, Ishoj, Denmark)] dissolved in water and taken as six daily meals, providing 810 kcal/day for 8 weeks. Participants then received group dietary guidance with the aim of achieving a 1200 kcal/day intake for 24 weeks. Following this, a period of weight maintenance was reinforced by the same VLCD for 4 weeks, followed by group instruction for 16 weeks. Participants in the control group received nutritional advice from a dietitian at baseline and at weeks 8, 32, 36 and 52. Participants received a total of 66 hours of instruction in the intervention group and 10 hours of instruction in the control group. The primary outcome was the mean group difference in total Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC^®) score, a validated, disease-specific questionnaire addressing the severity of joint pain, stiffness and limitation of physical function using a visual analogue scale. The mean BMI was 35.6 kg/m² (SD 5.0 kg/m²). The most frequent adverse events reported by the intervention group were constipation (five participants, 11%), increased flatulence (four participants, 9%), dizziness (two participants, 5%) and heightened sensitivity to cold (two participants, 5%). Both groups showed improvement in reported WOMAC scores but there was no reported statistically significant difference in total WOMAC scores between groups (reported p = 0.11), although the VLCD intervention resulted in less WOMAC pain, with a reported group mean difference of 7.2 mm (reported 95% Cl 1.0 to 13.4 mm; p = 0.022). By trial completion, two of the participants in the control group were diagnosed with type 2 diabetes mellitus and one had died from a heart condition. The authors93 concluded that weight loss is effective at reducing symptoms of pain in knee osteoarthritis but there was no indication of clinical improvement in joint stiffness.

Moreno et al. 155 compared a very low-calorie ketogenic diet, usually followed for 30–45 days followed by a low-calorie diet, with a low-calorie diet alone. The mean BMI of participants was 35.4 kg/m² (SD 4.9 kg/m²). Both groups received behavioural support and counselling. The control low-calorie diet was calculated to have a caloric value of 10% below the total metabolic expenditure of each individual (range 1400–1800 kcal/day). The very low-calorie ketogenic diet followed a commercial weight-loss programme (PronoKal^® Method). The VLCD programme had three stages: active, re-education and maintenance. The active stage provided 600–800 kcal/day. In phase 1, participants consumed protein preparations five times a day as well as vegetables with a low glycaemic index. In phase 2, one of the protein servings was replaced with natural protein (e.g. meat and fish). In phase 3, a second serving of protein that was low in fat replaced the second biological protein preparation. The active stage was maintained until the participant achieved most (ideally 80%) of their weight-loss target. The VLCD induced more frequent asthenia, fatigue, headache, muscle weakness, constipation, hyperuricamia and nausea than the low-calorie diet. The authors reported that these effects were mild and transient.

Ryttig et al. 172 compared a low-fat diet of 1600 kcal/day for 26 months with a VLCD (Nutrilett^®, Nycomed Pharma, Oslo, Norway) of 420 kcal/day for an initial 2 months. The VLCD participants were then randomised to the 1600-kcal/day diet with or without 1 mJ (240 kcal) of VLCD formula taken as three sachets daily for 26 months of weight maintenance. Participants received help with behaviour modification, including strategies to prevent relapse, dietitian-led training in low-energy cooking, a group or individual physiotherapist-led exercise programme and other activities, such as swimming and shopping exercises. The initial BMIs for the groups ranged from 37.6 kg/m² (SD 5.7 kg/m²) to 37.7 kg/m² (SD 3.9 kg/m²).

Torgerson et al. 186 randomised participants to receive a VLCD for 12 initial weeks plus regular dietary and behavioural support over 2 years, or 2 years of the same supportive programme only. The initial mean BMI was 40.2 kg/m² (SD 3.3 kg/m²) for the VLCD group, and 40.5 kg/m² (SD 4.3 kg/m²) for the non-VLCD group. Participants in the non-VLCD group were advised to consume an individualised diet aiming for 1200–1400 kcal/day for women or 1400–1800 kcal/day for men, with 25–30% of energy from fat. Participants in the VLCD group were provided with Modifast (Novartis Nutrition, Bern, Switzerland) and were recommended to consume 456–608 kcal/day (a higher level was recommended to men with high energy expenditure). Following 12 weeks of the VLCD, the hypocaloric diet was gradually introduced. Participants received 6-monthly individual nutritional counselling sessions with a dietitian and a physical examination with a physician, plus minor visits with a dietitian/nurse for support and further nutrition education weekly for weeks 1–2, fortnightly for weeks 2–8 and monthly thereafter. VLCD participants received three additional visits to facilitate refeeding following the 12-week VLCD. Participants were taught behavioural control techniques and attempts were made to identify risk circumstances for overeating and strategies to avoid these situations. Group swimming and physical training sessions were offered and all participants were encouraged to take part in physical activity.

Wadden et al. 193 compared a 1200-kcal/day diet (≤ 30% of energy as fat) and behavioural therapy with and without a VLCD in women. The mean BMI of groups ranged from 38.8 kg/m² (SD 5.4 kg/m²) to 40 kg/m² (SD 5.7 kg/m²). Following this, participants adjusted caloric intake during a 26-week maintenance programme, depending on their desired weight change (the minimum intake was 1200 kcal/day). Group sessions provided instruction on recording eating behaviour, stimulus control, modifying self-defeating thoughts and emotions and eliciting social support. An incremental walking-based exercise programme began at week 8, leading to 3–5 times/week at 60–70% of the maximum heart rate by 12 months. Biweekly ‘upkeep’ sessions were provided during the 26-week weight maintenance period to prevent and reverse weight gains and dietary lapses. Women in the VLCD group received a 1200-kcal/day diet for the first week and then a liquid VLCD formula (Optifast 70, Sandoz Nutrition, Minneapolis, MN, USA) during weeks 2–17. Conventional foods were gradually reintroduced during weeks 8–23, so that by week 23 participants were consuming a 1000-kcal/day diet of conventional food. Refeeding was supervised by the dietitian who co-led groups from weeks 18–27 and provided further information on food preparation and nutrition (also provided for the comparison group). VLCD participants received the same exercise and cognitive–behavioural treatment programme as the comparison group throughout the study, but materials were presented in a different order in the first 26 weeks.

Wing et al. 200 investigated a VLCD in people with type 2 diabetes mellitus. Participants were randomised to a low-fat diet (calorie goal of 1000–1500 kcal/day) and behaviour therapy for 20 weeks, or to the same behavioural therapy for 20 weeks with a VLCD of 400 kcal/day of lean meat, fish or fowl, and the option of using Optifast 70 for occasional meals for 8 weeks. The mean BMI of the participant groups ranged from 37.3 kg/m² (SD 4.7 kg/m²) to 38.1 kg/m² (SD 5.7 kg/m²). After 8 weeks on the VLCD, other foods were gradually reintroduced to reach 1000–1500 kcal/day by week 17. Behaviour modification strategies included self-monitoring behaviour, stimulus control, removing food cues from the environment, separating eating from other activities and modifying cognitions for relapse prevention and self-reinforcement. Participants were taught to increase their walking and were given weekly exercise goals. All participants deposited US$150 at the start of the programme, which was earned back weekly for meeting homework goals.

Wing et al. 198 evaluated an initial VLCD in people with type 2 diabetes mellitus compared with a low-fat 1000- to 1200-kcal/day diet, both with behaviour change support. Participants in the VLCD group were prescribed a diet of 400–500 kcal/day for weeks 1–12 and 24–36. The mean BMI of the participant groups ranged from 37.4 kg/m² (SD 6.1 kg/m²) to 38.3 kg/m² (SD 6.52 kg/m²). During VLCD periods, participants were instructed to consume no more than 500 kcal/day, either as liquid formula (Optifast 70) or lean meat, fish and fowl, with most participants choosing combinations of the food and liquid formula. After 12 weeks on the VLCD, other foods were gradually reintroduced, and the prescribed calories gradually increased over 4 weeks until participants were consuming the 1000–1200 kcal/day. Participants were restarted on the VLCD at week 24 unless their ideal body weight had been reached. Behavioural therapy focused on self-monitoring, goal-setting (including exercise goals), stimulus control, preplanning, relapse prevention and modifying cognitions during weekly group meetings conducted by a multidisciplinary team of therapists. All participants deposited US$150 at the start of the programme, which was refunded for reaching behavioural goals and attending assessment sessions. Diabetes mellitus medications were stopped at baseline and a survival analysis was used to compare the number of weeks that participants maintained blood sugar of < 13.3mmol/l without restarting medication. Fasting glucose remained below the cut-off level for a significantly longer time in the VLCD than in comparison group (reported p = 0.05) and the number of participants remaining on the diet without medication was significantly greater in the VLCD group (55%) than in the comparison group (31%) at 2 years (reported p = 0.01).

A summary of the delivery format of the interventions for these trials is provided in Report Supplementary Material 5, Table E2.

Meta-analyses

The results of the meta-analysis of the mean difference in weight change for the VLCD and dietary intervention versus dietary intervention are presented in Table 2. There was a difference in weight change between the VLCD intervention and control treatments at 12 months in favour of the VLCD for both adjusted and unadjusted data; however, at subsequent time points there is no evidence that VLCD treatments are effective and CIs rule out worthwhile weight loss.

The weight change in the VLCD arms alone at 12 months was –9.69 kg (95% CI –12.55 to –6.82 kg).

Overall, the direction of the effects of the intervention on risk factors tended to favour VLCD groups. However, there were only reductions in diastolic blood pressure (–5.00 mmHg, 95% CI –8.66 to –1.34 mmHg; p = 0.007) at 12 months, and reductions in HbA_1c (–2.60%, 95% CI –4.44% to –0.76%; p = 0.006) and glucose (–4.50 mmol/l, 95% CI –6.88 to –2.12 mmol/l; p = 0.000) at 18 months (the HbA_1c and glucose results mainly reflecting participants with type 2 diabetes mellitus). No changes were found at 24 months’ follow-up. Table 3 provides further details.

Behaviour change technique coding summary (see Report Supplementary Material 1, Section 8)

Suboptimal reporting of the behavioural support made coding BCTs difficult. Generally, the same BCTs were provided to both arms in trials (particularly goal-setting), but descriptions were provided for few BCTs. VLCD and dietary intervention arms had a mean of 8.7 BCTs (range 4–13 BCTs) and dietary intervention only arms had a mean of 7.5 BCTs (range 1–12 BCTs).

Descriptions of the trials

Stenius-Aarniala et al. 180 randomised people with asthma to either an initial VLCD treatment or control. The initial mean BMIs for groups ranged from 35.8 kg/m² (range 31.3–39.4 kg/m²) to 36.7 kg/m² (range 32.8–41.8 kg/m²). The VLCD (Nutrilett, Nycomed Pharma, Oslo, Norway) was taken for 8 weeks (420 kcal/day). The VLCD programme also included 12 group sessions on behaviour change over 14 weeks. The control group had sessions at the same intervals as the active group on education about asthma and allergy. After 1 year, the mean weight change was –13.40 kg (95% CI –17.21 to –9.59 kg; p < 0.001). The per cent of predicted forced expiratory volume in 1 second increased more in the active group than in the control group after the treatment period and at 1 year (mean difference 7.6%, 95% CI 1.5% to 13.8%; reported p = 0.02). Similarly, the per cent of predicted forced vital capacity was different between groups at 1 year (mean difference 7.6%, 95% CI 3.5% to 11.8%; reported p = 0.001). There was no evidence of differences in peak expiratory flow between the groups. After 1 year, the mean differences in the changes in the St George’s Respiratory Questionnaire between the two groups were –12 for symptom scores (range –1 to –22; reported p = 0.04) and –10 for total scores (–18 to –1; reported p = 0.02). During the 1-year follow-up period, the median number of exacerbations was 1 (range 0–4) in the treatment group and 4 (range 0–7) in the control group (reported p = 0.001). Thirteen control participants and 10 treatment group participants needed a course of oral steroids (reported p = 0.07) (Table 4).

Soenen et al. 177 compared four energy-restricted diets in a factorial design: high versus low protein and low versus normal carbohydrate. Initial energy intake was restricted to 33% of energy requirements during the weight-loss phase and, therefore, this could be classed as a VLCD. For further details, see Comparisons of diets.

Descriptions of the trials

Melin et al. 150 examined a VLCD with varying intensities of behavioural therapy. Participants had a mean group BMI of 35.2 kg/m² (SD 4.6 kg/m²) to 35.6 kg/m² (SD 4.5 kg/m²). The intensive treatment group received group meetings every fortnight for the first year and six meetings during the second year. Participants in the control group had group meetings every third month and less contact with supervisors, fewer repetitions with self-monitoring and less opportunity for nutrition counselling. Both groups underwent two VLCD periods, lasting 25 days each, in between following a diet with a 600-kcal/day deficit. Although favouring more meetings, there were no significant differences in weight change between the groups (see Table 4).

Purcell et al. 166 recruited volunteers without comorbidities, with a mean BMI of 35.3 kg/m² (SD 3.8 kg/m²), and investigated the effect of rapid weight loss with a VLCD and gradual weight loss without a VLCD. In the rapid weight-loss programme, participants consumed OPTIFAST^® (Nestlé Nutrition, Vevey, Switzerland) meal replacements of between 450–800 kcal/day, producing approximately 15% weight loss over 12 weeks. In the gradual weight-loss programme, participants followed a 400–500-kcal/day deficit low-fat diet, including one to two Optifast meal replacements daily, with the aim of 15% weight loss over 36 weeks. After 36 months, there was no evidence that rapid weight loss led to more weight being regained. Attrition in the weight-loss phase was significantly less with the VLCD (reported p = 0.002), and did not differ between groups in the weight maintenance phase. Two participants in the VLCD group developed cancer (see Table 4).

Rössner and Flaten171 compared three VLCDs of 420 (Nutrilett), 530 and 880 kcal/day for 6 weeks and a 2-week booster at 26 weeks. Between weeks 6 and 26, participants followed a diet of 1600 kcal/day and received sessions with a research nurse or dietitian. The initial group mean BMIs ranged from 38.4 kg/m² (SD 4.3 kg/m²) to 39 kg/m² (SD 5.2 kg/m²). At 1 year, weight loss was similar in the three groups (see Table 4).

Torgerson et al. 184 compared three approaches to VLCD use over 16 weeks in a 1-year programme in groups with mean BMIs of 37.7 kg/m² (SD 4.3 kg/m²) to 38.5 kg/m² (SD 4.5 kg/m²). All participants consumed three sachets of Modifast, providing 456 kcal/day. The VLCD-strict group was prescribed a strict outpatient VLCD for 16 weeks, followed by a 36-week hypocaloric diet. The metabolic ward (VLCD-mw) participants received the same treatment, but were hospitalised in a metabolic ward for the first week. The VLCD-plus participants were allowed two small meals weekly, but otherwise received the same recommendations as the VLCD-strict group. At 12 months, there were no significant differences in weight loss between the groups (see Table 4).

Descriptions of the trials

Details of the mean weight change in the trials are presented in Table 5. In the Ryttig et al. 172 trial, the VLCD followed by meal replacement group lost significantly more weight than the VLCD plus diet group at 12 months (p = 0.037). However, another trial from this group173 did not find as marked an effect from meal replacements (see Randomised controlled trials of meal replacements for weight loss and weight maintenance). In the trial by Christensen et al. ,99 a dietary programme was more successful at preventing weight regain than exercise or usual care. In the trial by Richelsen et al. ,76 new cases of type 2 diabetes mellitus were reported to have been significantly reduced with orlistat (8 cases out of 153 participants) versus placebo (17 cases out of 156 participants) (reported p = 0.041). Varying protein content107 or regular intermittent VLCD use compared with VLCD use on demand139 did not appear to affect weight regain. In one trial, Pekkarinen et al. 161 found little added benefit from a specific weight maintenance programme versus no follow-up.

Risk-of-bias assessment

Most trials (10/11, 90.9%) were considered to have adequate randomisation sequence generation, although it was unclear whether or not allocation concealment was adequate in seven trials (63.6%). Owing to the nature of the interventions, blinding of participants and personnel was not possible. Only one trial101 reported blinding outcome assessment. Four trials were rated as being at risk of bias owing to author links with surgical manufacturers (see Report Supplementary Material 4, Figure E8). 110,111,154,174

Equity assessment

Trials failed to report information for the majority of items on the Cochrane Public Health Equity assessment checklist. 61 Fewer than half of the trials were considered to have included a representative sample and none reported sociodemographic details of participants. Trials did not report information concerning diversity or disadvantage in their intervention design or delivery, nor did they discuss the sustainability of interventions. Trials did not report process measures or details of any fidelity checks (see Report Supplementary Material 4, Figure E9).

Descriptions of the trials

The trial conducted by Courcoulas et al. 100 compared RYGB, LAGB and an intensive lifestyle WMP (ILI). Participants in the ILI [mean BMI 35.7 kg/m² (SD 3.3 kg/m²)] followed an adapted 12-month programme based on the Diabetes Prevention Program (DPP)205 and the Look AHEAD trial. 41 ILI participants were prescribed a 1200- to 1800-kcal/day diet. Moderate intensity exercise was for prescribed 5 days per week, beginning at 20 minutes per day and progressing to ≥ 60 minutes per day. Surgical participants underwent clinical follow-up assessments at 2 weeks and 3, 6, 9 and 12 months postoperatively in the RYGB group and at 2 weeks and 2, 4, 6, 8, 10 and 12 months or more frequently, as necessary, for band adjustment in the LAGB group. The RYGB group had a mean initial BMI of 35.5 kg/m² (SD 2.6 kg/m²). RYGB was performed with a standard retrocolic–retrogastric technique using a linear stapled and hand-sewn gastrojejunal anastomosis. Participants were counselled on a diet programme consistent with postbariatric surgery recommendations and were encouraged to exercise for a minimum of three or four times each week. LAGB participants had mean BMI of 35.5 kg/m² (SD 3.4 kg/m²).

After 3 years, RYGB was the most successful treatment for achieving partial or complete remission of type 2 diabetes mellitus. Partial/complete remission was achieved by 40%, 29% and no participants in the RYGB, LAGB and ILI groups, respectively, and complete remission was achieved by 15%, 5% and no participants, respectively. The number of participants in remission declined in the RYGB group, from 60% at year 1 to 45% at year 2 and 40% at year 3, whereas remission remained stable at 29% for LAGB participants and none for ILI participants over the 3 years. The RYGB group had the greatest reductions in HbA_1c and fasting glucose from baseline to 3 years (reported p < 0.0013 for RYGB vs. ILI, and reported p < 0.05 for RYGB vs. both LAGB and ILI, for HbA_1c and fasting plasma glucose, respectively). No adverse events were reported for the ILI group during the 3 years. One serious adverse event (anastomotic ulcer) was reported for the RYGB group. One case of nausea and emesis requiring intravenous hydration and one case of renal lithiasis were reported for the RYGB group.

The CROSSROADS (Calorie Reduction Or Surgery: Seeking to Reduce Obesity And Diabetes Study) trial, conducted by Cummings et al.,101 randomised people to receive RYGB or an ILI. The ILI group [mean BMI 37.1 kg/m² (SD 3.5 kg/m²)] had a 12-month in-person and telephone-based intervention. A dietitian delivered nutrition sessions based on the DPP205 with modifications for people with diabetes mellitus. Reduced calorie intake and weight loss were encouraged but participants were not given weight-loss goals. Participants attended physiologist-led supervised exercise sessions at least three times a week and were asked to exercise on ≥ 2 additional days per week at home for the first 6 months. Participants were then asked to attend the supervised exercise sessions for 1 day per week and to exercise on ≥ 4 days per week at home for the remaining 6 months. Participants in the RYGB group [initial mean BMI 38.3 kg/m² (SD 3.7 kg/m²)] underwent laparoscopic RYGB and a 4-week preoperative and 10-month postoperative behavioural programme, which focused on diet and nutrition counselling, behaviour modification and exercise recommendations. Participants received weekly telephone-based appointments and attended two or three bariatric support group meetings during the preoperative phase. Telephone appointments with the health educator continued during the postoperative phase.

Diabetes mellitus remission was achieved by 60% of RYGB participants and 5.9% of ILI participants (reported p = 0.002). There were 64 adverse events reported in the ILI group and 31 adverse events reported in the RYGB group. There were no reported deaths or serious surgical adverse events for either group. More hypoglycaemic events were reported in the ILI group than in the RYGB group (43 vs. 16). One serious adverse event involving a hospital emergency room visit for acute alcohol intoxication was reported for a RYGB participant. Minor musculoskeletal complaints were reported by seven ILI and two RYGB participants. The authors reported that there was no difference between groups for other minor adverse events.

The trial by Halperin et al. 126 compared RYGB with a 12-month intensive diabetes mellitus medical and WMP: the Weight Achievement and Intensive Treatment (Why WAIT) programme. Why WAIT included a diet of 1500–1800 kcal/day and was delivered by a multidisciplinary team in 2-hour weekly group sessions during the first 12 weeks, focusing on diabetes mellitus medication adjustments, cognitive–behavioural therapy (CBT), group education and supervised exercise. Participants received individual monthly counselling thereafter. The initial mean BMI of Why WAIT participants was 36.5 kg/m² (SD 3.4 kg/m²), and it was 36.0 kg/m² (SD 3.5 kg/m²) for RYGB participants. RYGB participants showed greater improvements in quality of life, measured by the Impact of Weight on Quality of Life (IWQOL)-Lite, than Why WAIT participants at 12, 18 and 24 months, but there were no between-group differences for other quality-of-life measures.

The trial by Mingrone et al. 154 randomised people to RYGB, BPD or a diet and lifestyle modification. Participants in the lifestyle intervention group had an initial mean BMI of 45.6 kg/m² (SD 6.2 kg/m²). Participants were advised to reduce their overall energy and fat intake (< 30% total fat and < 10% saturated fat) and increase exercise (approximately 30 minutes/day of brisk walking with additional moderate intensity aerobic activity twice a week). Participants in the surgical arms had an initial mean BMI of 44.5 kg/m².

At 5 years, 37% of participants in the RYGB group and 63% in the BPD group achieved and maintained diabetes mellitus remission, compared with none of the medically treated participants. Five major complications of diabetes mellitus (including one fatal myocardial infarction) were reported in four participants (27%) in the medical group, compared with only one complication in the RYGB group and no complications in the BPD group. Two participants (13%) in the medical treatment group reported persistent diarrhoea associated with metformin. One intestinal obstruction requiring reoperation occurred 6 months after RYGB. The incidence of metabolic adverse events was higher in the surgical groups than in the medical treatment group. No late complications or deaths were reported in the surgery groups.

The STAMPEDE (Surgical Treatment and Medications Potentially Eradicate Diabetes Efficiently) trial, conducted by Schauer et al.,174 randomised people to receive intensive medical therapy or medical therapy plus RYGB or SG. Participants in all groups received lifestyle counselling and weight-management support. Participants were encouraged to participate in the Weight Watchers^® (New York, NY, USA) programme. Surgical participants were evaluated by a psychologist. The primary outcome of the trial was HbA_1c of ≤ 6.0% with or without the use of diabetes mellitus medications. The initial mean BMI was 37 kg/m² (SD 3.3 kg/m²) for GBP participants, 36.2 kg/m² (SD 3.9 kg/m²) for SG participants and 36.8 kg/m² (SD 3.0 kg/m²) for intensive medical therapy participants. At 5 years, the primary end point was reported by 2 out of 38 participants (5%) who received intensive medical therapy, 14 out of 49 GBP participants (29%) (reported unadjusted p = 0.01, reported adjusted p = 0.03 and p = 0.08 in the intention-to-treat analysis) and 11 out of 47 participants (23%) who underwent SG (reported unadjusted p = 0.03, adjusted p = 0.07 and p = 0.17 in the intention-to-treat analysis). Quality of life, as measured by the RAND 36-Item Health Survey, was significantly better in the RYGB group than in the intensive medical treatment group for general health scores and bodily pain scores (reported p < 0.05 for both). Four surgical participants required subsequent surgical interventions, one intensive medical therapy participant died as a result of myocardial infarction and one participant in the SG group had a stroke.

Meta-analyses

The five trials100,101,126,154,174 were included in a meta-analysis of weight change and used to inform the economic evaluation of RYGB in Chapter 7. The results of the meta-analysis of the mean difference in weight change between the intervention and control treatments are presented in Table 6. At all time points, up to 60 months following surgery, there is evidence of a difference in weight loss in favour of RYGB for both adjusted and unadjusted data.

A further RYGB trial was identified for inclusion in our review. 169 This trial compared RYGB and behavioural and psychological counselling with a control group who received only written and oral information about a healthy lifestyle, but not a diet and lifestyle WMP. It was not included in the weight change or risk factor meta-analyses. Reis et al. 169 examined the effects of weight loss on erectile function in men. Men in the RYGB group had a mean BMI of 55.7 kg/m² (SD 7.8 kg/m²). Those in the control group had an initial mean BMI of 54.0 kg/m² (SD 6.1 kg/m²). None of the men was taking phosphodiesterase type-5 inhibitors at baseline. At 24 months, men in the RYGB group had significantly greater reduction in weight and BMI than men in the control group (reported p = 0.0006 for both). Surgery increased erectile function quality, measured by the International Index of Erectile Function (IIEF)-5 questionnaire (reported p = 0.0224), and increased total testosterone (TT) (reported p = 0.0043) and free testosterone (FT) (reported p = 0.0149) levels.

Over 5 years, RYGB consistently improved HbA_1c, glucose and HDL cholesterol. Triglycerides and systolic blood pressure were also improved at 3 years by RYGB, but were not improved at 5 years. LDL cholesterol was lower with lifestyle interventions, and other results were mixed (Table 7 provides more information).

Descriptions of the trials

Details of the earlier of these trials, by Schauer et al. ,174 were reported in Roux-en-Y gastric bypass surgery and details of the interventions for all surgery trials are presented in Report Supplementary Material 5, Table E5.

The trial conducted by MacLaughlin et al. 145 compared laparoscopic SG with a lifestyle intervention in people with obesity and stages 3/4 chronic kidney disease to determine the effect on kidney function. The medical care group had an initial mean BMI of 37.4 kg/m² (range 35.8–40.0 kg/m²). The intervention was delivered by an experienced renal dietitian and physiotherapist, supported by a nephrologist and renal pharmacist. Participants were prescribed a 1400- to 1800-kcal/day low-fat, renal diet; regular exercise; CBT; and 120 mg of orlistat three times daily. Dietary protein intake was 0.8–1.0 g/kg of ideal body weight. Participants randomised to the SG group [initial mean BMI of 40.3 kg/m² (range 37.3–43.0 kg/m²)], received dietary and renal information. Energy intake was restricted to approximately 1000 kcal/day following surgery. Protein intakes were similar to those of the medical care group. The SG group showed greater improvement in quality of life than the medical care group, measured by Hospital Anxiety and Depression Scale scores and increased Short Form questionnaire-36 items (SF-36) physical domain scores (reported p < 0.05 for both). There were no significant differences in unadjusted body surface estimated glomerular filtration rate (eGFR). There was no change in diabetes mellitus prevalence throughout the study. One participant on insulin therapy in the SG group decreased their insulin dose by 59%, and four participants requiring insulin in the lifestyle group decreased their median insulin dose by 7% over 12 months.

Meta-analyses

The results of the meta-analysis of SG versus lifestyle intervention, presented in Table 8, show that, for data adjusted and unadjusted for participant withdrawal, SG produces greater weight loss than lifestyle intervention at all time points.

Favourable changes from SG were seen for all risk factors, with the exception of total cholesterol, LDL cholesterol and diastolic blood pressure (Table 9).

Descriptions of the trials

Details of the Courcoulas et al. 100 trial are provided in Roux-en-Y gastric bypass surgery and details of the interventions of all surgery trials are presented in Report Supplementary Material 5, Table E5. Participants in the LAGB group followed the same diet and lifestyle programme as participants in the RYGB group.

Ding et al. 109 randomised people with type 2 diabetes mellitus to receive LAGB or the same Why WAIT intensive medical and WMP delivered by Halperin et al. ,126 described previously in Roux-en-Y gastric bypass surgery. The initial mean BMI was 36.7 kg/m² (SD 4.2 kg/m²) for the Why WAIT participants and 36.4 kg/m² (SD 3.0 kg/m²) for the LAGB participants. At 12 months, both groups showed similar improvements in health and quality of life.

In the 2008 trial by Dixon et al. ,110 participants with type 2 diabetes mellitus were randomised to conventional therapy and had meetings with a physician, a dietitian, a nurse and a diabetes mellitus educator every 6 weeks for 2 years. Participants were encouraged to reduce their fat intake to < 30%, and to consume low glycaemic index foods. VLCDs and drug treatment were used. Participants were encouraged to engage in 200 minutes per week of structured physical activity. Surgical participants followed conventional therapy and underwent the LAGB procedure. The mean BMI was 37 kg/m² (SD 2.7 kg/m²) for participants in the LAGB group, and 37.2 kg/m² (SD 2.5 kg/m²) for conventional treatment participants. After 2 years, more LAGB participants (73%) than conventional therapy participants (13%) had achieved type 2 diabetes mellitus remission (reported p < 0.001).

The 2012 trial by Dixon et al. 111 compared LAGB with conventional weight-loss therapy for obstructive sleep apnoea (OSA) in participants with an apnoea–hypopnoea index of ≥ 20 events per hour. Participants in the conventional therapy group [mean BMI 43.8 kg/m² (SD 4.9 kg/m²)] were encouraged to engage in 200 minutes per week of structured physical activity and have a daily energy deficit of 500 kcal/day. Both groups saw a bariatric physician, a sleep physician and a dietitian, and were reviewed every 4–6 weeks for 2 years. LAGB participants reportedly had greater reduction in apnoea–hypopnoea index than conventional therapy participants (a decrease of 25.5 events/hour vs. a decrease of 14.0 events/hour), but the difference between groups was not significant. The LAGB group showed greater improvement in SF-36 physical activity scores than the conventional therapy participants (reported p = 0.04). The authors reported that 20 participants had type 2 diabetes mellitus at baseline (10 in each group) and two participants in the conventionally treated group developed type 2 diabetes mellitus during the study; however, there were no within- or between-group differences in the use of diabetes mellitus medication.

Meta-analyses

The results of the meta-analysis of gastric band and lifestyle intervention versus lifestyle intervention only, presented in Table 10, show that gastric band produces greater weight loss than lifestyle intervention at all time points.

Risk factor data show no consistent benefit at each time point over 3 years with bariatric surgery, although HbA_1c is lower at 3 years (Table 11 provides more details).

Descriptions of the trials

Details of the Mingrone et al. 154 2012 trial are presented in Roux-en-Y gastric bypass surgery and details of the interventions in the Mingrone et al. 153 2002 trial are presented in Report Supplementary Material 5, Table E5. In both trials, participants in the surgical groups lost more weight than those in the diet groups.

Behaviour change technique coding summary for all trials of bariatric surgery (see Report Supplementary Material 1, Section 8)

Surgical interventions were compared with lifestyle interventions, which usually included a dietary component and – in all but one trial – physical activity ‘advice’ or supervised physical activity. Control groups received, on average, 4.9 BCTs (range 1–10 BCTs) and intervention (surgery) groups received slightly fewer BCTs (average 2.6 BCTs, range 1–6 BCTs) targeting these behaviours. The most commonly used BCTs were setting a behavioural goal and a respected person explaining the importance of diet/physical activity. Comparator group support varied substantially between trials, with the least supported comparator group receiving only advice from a credible person, versus the most extensively supported comparator group, in which participants also received instructions, goal-setting, graded tasks, self-monitoring and problem-solving BCTs, as well as adding/removing objects from the environment, help with reducing negative emotions and social support. These results must be interpreted with caution, however, as intervention and control group support was generally described very poorly (only two trials were rated as adequately described and none was rated as well described) (see Report Supplementary Material 1, Section 8).

Descriptions of the trials

In the study by Davidson et al. ,105 all participants received placebo plus a controlled-energy diet in a 4-week lead-in phase. After 4 weeks, the diet was continued and participants were randomised to receive placebo three times daily or 120 mg of orlistat three times a day, for 12 months, after which time participants began a weight maintenance diet. The placebo group [mean BMI 36.5 kg/m² (SD 2.4 kg/m²)] continued to receive placebo. Orlistat-treated participants [mean BMI 36.2 kg/m² (SD 2.6 kg/m²)] were rerandomised to receive placebo, 60 mg of orlistat or 120 mg of orlistat three times a day for an additional 12 months. Participants who received 120 mg of orlistat during year 2 regained significantly less of their first-year weight loss (35.2% regain) than those who received 60 mg of orlistat (51.3% regain) or placebo (63.4% regain) during the second year (reported p < 0.001).

In the trial by Sjöström et al. ,176 participants who completed a 4-week placebo lead-in period on a 600-kcal/day deficit diet, followed by randomisation to double-blind treatment with 120 mg of orlistat or placebo three times a day in conjunction with the hypocaloric diet for a 1-year period, were then rerandomised to orlistat or placebo with a weight maintenance diet for a second period of 12 months. Mean BMI ranged from 36.0 to 36.1 kg/m². At the end of the weight maintenance phase, participants who continued with orlistat regained half as much weight as those participants who were switched to placebo (reported p < 0.001). Participants who switched from placebo to orlistat lost an additional 0.9 kg during year 2, compared with a mean regain of 2.5 kg in participants who continued on placebo (reported p < 0.001). As shown in Table 12, weight loss was significantly greater in the orlistat group than in the placebo group at 24 months.

Descriptions of the trials

Participants in the Kumanyika et al. 137 trial received a 10-week group counselling Healthy Eating and Lifestyle Programme (HELP) adapted for African American participants. Participants were then randomised to receive further group counselling, staff-assisted self-help or clinic visits only. The mean BMI of participants was 37.0 kg/m². At 18 months, weight loss was greater in the group help condition but there was no evidence of an important difference between the two conditions, as shown in Table 12.

Latner et al. 140 compared standard care and self-help continuing care for weight maintenance. The mean BMI of participants was 35.8 kg/m². At 24 months, weight maintenance was more successful in the standard care group than in the self-help group, but there was no evidence of a difference between groups (see Table 12).

Participants in the trial by Lowe et al. 143 were randomised to one of four treatment groups involving the presence or absence of meal replacements and a low-fat energy-reducing diet: MR+/ED+ (meal replacements and an energy-reducing diet), MR–/ED+ (energy-reducing diet only), MR+/ED– (meal replacements only) and MR–/ED– (healthy eating advice control group). The mean BMI of participants was 39.5 kg/m² (SD 6.6 kg/m²). Table 12 shows the combined data for participants in the groups that did/did not follow the energy-reducing diet (ED+ and ED–) and in those that did/did not receive meal replacements (MR+ and MR–) at 24 and 36 months. Weight regain did not differ between the interventions.

In the trial by Perri et al. ,162 women from rural communities who had completed an initial 6-month weight-loss programme were randomised to two extended-care counselling interventions delivered by telephone or face to face, or to an education control group for a further 6 months. The participant group mean BMI ranged from 36.2 kg/m² (SD 4.3 kg/m²) to 37.1 kg/m² (SD 4.5 kg/m²). As shown in Table 12, participants receiving the telephone and face-to-face interventions regained less weight than those in the education control group at 12 months.

In an earlier trial conducted by Perri et al. ,164 two 12-month extended behavioural therapy treatments were compared with a standard behavioural therapy with no further contact. The two extended behavioural therapy conditions used different approaches to weight maintenance. The first focused on relapse prevention training and the second focused on problem-solving. The control group received standard behavioural therapy. The mean BMIs of the groups ranged from 35.0 to 36.4 kg/m². Table 12 shows weight change at 17 months. When missing data were ignored, relapse prevention or problem-solving resulted in a weight change of –4.37 kg (95% CI –8.69 to –0.05 kg), but when accounting for missing data this was reduced to –2.45 kg (95% CI –6.21 to –1.31 kg).

West et al. 197 randomised African American women with urinary incontinence to a WMP or education control. Women received the WMP for 6 months and were than randomised to standard skill-based or motivation-based weight maintenance programmes for 12 months. The mean BMI was 36.0 kg/m² (SD 6.0 kg/m²). Table 12 shows weight change at 12 and 18 months. The skill-based programme produced more favourable results at 12 months, whereas the motivation-based programme was more successful at 18 months. There is no evidence of a difference between groups at either time point.

Risk-of-bias assessment

Random sequence generation was considered to be adequate for the majority (80%) of the trials, although details of allocation concealment were unclear for most (80%). Only one trial89 clearly reported blinding of outcome assessors. Most trials were rated as being at low risk of other types of bias (see Report Supplementary Material 4, Figure E10).

Equity assessment

Factors relevant for assessing equity were poorly reported in trials. All trials reported one or more PROGRESS category at baseline, usually sex, and most trials (80%) reported unintended effects or harms associated with the intervention. The remaining equity checklist items either were not considered or were so poorly reported that they were judged as unclear (see Report Supplementary Material 4, Figure E11).

Descriptions of the trials

Bazzano et al. 89 examined the effects of a low-carbohydrate diet versus a low-fat diet for 12 months; both groups received regular dietary counselling throughout. No serious adverse events were reported. The incidence of minor adverse events was similar between groups, with the exception of headaches at 3 months, which were reported by more participants in the low-fat diet group (reported p = 0.030).

Foster et al. 120 compared low-carbohydrate and low-fat dietary interventions with the same group behavioural intervention weekly for 20 weeks, every other week for 20 weeks and then every other month over 24 months. All participants were given the same instructions for increasing their levels of physical activity.

Davis et al. 106 compared low-carbohydrate and low-fat diets for people with type 2 diabetes mellitus over 12 months. Participants initially underwent a 3- to 4-week pre-randomisation phase, during which participants were asked to self-monitor their diet and blood glucose levels. During the pre-randomisation phase, 31.8% discontinued the study. The low-carbohydrate diet was based on the Atkins diet and the low-fat diet was based on the DPP intervention. 205 Both groups received booklets detailing the carbohydrate or fat content of common goods and were given general recommendations to increase physical activity. All participants received six counselling sessions with a dietitian.

Iqbal et al. 129 randomised people with type 2 diabetes mellitus to a low-carbohydrate or low-fat diet. All participants received dietitian sessions, which were weekly for the first month and monthly thereafter for 24 months. All participants were given educational handouts specific to their diet and were encouraged to engage in ≥ 30 minutes of moderate physical activity at least five times per week. Two people in the low-carbohydrate group and three in the low-fat group died during the course of the trial. One participant in each group died of myocardial infarction. The authors stated that the cause of death for the other three participants was not available. One participant was diagnosed with breast cancer (group unknown). No severe hypoglycaemic episodes were reported.

Participants in the trial by Stern et al. 181 received a low-carbohydrate or low-fat diet with weekly counselling sessions for 4 weeks followed by 11 monthly sessions. Participants were given diet handouts. Two people in the low-carbohydrate group and four people in the low-fat group developed diabetes mellitus at 1 year (reported p > 0.2). One person on the low-carbohydrate diet was reported to have been hospitalised with non-cardiac chest pain. Two people in the low-carbohydrate group died: one died of complications of hyperosmolar coma 5 months into the study and another had severe ischaemic cardiomyopathy and died suddenly 10 months after study enrolment.

In an additional study by Soenen et al. ,177 a low-carbohydrate diet (5% of energy from carbohydrate) was compared with a normal-carbohydrate diet (35% energy from carbohydrate) as part of a factorial design to also investigate high- versus low-protein diets. The Soenen et al. 177 trial is also discussed in Higher-protein diets versus lower-protein diets. The authors concluded that weight loss and weight maintenance depend on protein but not on the low-carbohydrate component of the diet.

Meta-analyses

The results of the meta-analyses of the low-carbohydrate versus low-fat diets are presented in Table 13. The meta-analysis of weight loss favoured the low-carbohydrate intervention for both adjusted and unadjusted data, with evidence of a benefit from low-carbohydrate diets at 12 months only. Subgroup analyses were conducted for interventions with and without a specified energy goal for the low-fat diets (see Table 13). In the Bazzano et al. 89 trial, participants were not given a specified energy goal for the low-fat diet, whereas the other four trials had an energy goal. Both subgroup analyses favoured a low-carbohydrate diet, with the trial by Bazzano et al. ,89 with no energy restriction for the low-fat diet, having a much larger effect.

Participants randomised to low-carbohydrate diets had lower triglyceride levels and higher HDL cholesterol concentrations after 12 months. Only increased HDL cholesterol and decreased diastolic blood pressure were present at 24 months. None of the other risk factor changes, including HbA_1c and glucose levels, was different between the groups (Table 14).

Behaviour change technique coding summary (see Report Supplementary Material 1, Section 8)

Participants in low-carbohydrate diet groups received, on average, 7.2 BCTs (range 3–12 BCTs) and those in comparator groups received 6.6 BCTs (range 3–10 BCTs). Three BCTs were used routinely, namely setting a behavioural goal, providing instruction on how to perform the behaviour and a respected person explaining the importance of diet/physical activity. Comparator group support varied between trials, with the least supported comparator receiving instructions, goal-setting and advice from a credible person versus the most extensively supported comparator group additionally receiving feedback on and a review of behaviour goals, self-monitoring and social support BCTs, as well as additions to the environment and information about health consequences. These results must be interpreted with caution as group support was generally poorly described in both arms; only one trial was rated as well described and another was rated as adequately described.

Risk-of-bias assessment

Most trials (71.4%) were considered to have adequate randomisation sequence generation, although in 57.1% of trials it was unclear whether or not allocation concealment was adequate. Owing to the nature of the interventions, blinding of participants was not possible. Only two trials134,160 reported blinded outcome assessment, and one160 reported blinded personnel. Two trials107,119 were rated as being at risk of bias owing to author links with dietetic product manufacturers (see Report Supplementary Material 4, Figure E12).

Equity assessment

The trials did report some information for the majority of items on the Cochrane Public Health Equity assessment checklist. 61 Fewer than half reported the sociodemographic details of their participants. Similarly, most of the trials did not report information concerning diversity or disadvantage in their intervention design or delivery, nor did they discuss the sustainability of the interventions. The trials also did not report process measures or details of any fidelity checks (see Report Supplementary Material 4, Figure E13).

Descriptions of the trials

Dalle Grave et al. 102 randomised participants (mean BMI 45.6 kg/m²) to energy-restricted diets (both 1200–1500 kcal/day, with 20% of energy from fat): a higher-protein diet (34% of energy from protein) or a lower-protein diet (17% of energy from protein). The first stage was 3 weeks of inpatient treatment and the second stage was 48 weeks of outpatient treatment. All participants received a CBT manual. Participants in both groups received 55 sessions that included weekly CBT group sessions, aerobic exercise sessions and callisthenic sessions. Psychosocial outcomes (i.e. body image dissatisfaction, binge eating, depression and quality of life) were reported as not statistically significantly different between the groups. No adverse events were reported.

Delbridge et al. 107 investigated a 3-month intensive weight-loss period with a VLCD (Optifast, described in Very low-calorie diets) for all participants (mean BMI of 39 kg/m²). For the 12-month weight maintenance phase, participants were randomised to diets with either 30% or 15% of energy being from protein. Participants of both groups were recommended a low glycaemic index diet with fat intake providing < 30% of energy. Twelve monthly individual face-to-face sessions were provided.

Flechtner-Mors et al. 119 recruited participants meeting at least three of five criteria for metabolic syndrome. Participants (mean BMI of 36 kg/m²) were randomised to low-fat diets with either 30% or 15% of energy from protein. In both diets, 30% of energy was from fat and there was a 500-kcal/day deficit. In the first 3 months, participants in the higher-protein group consumed two protein-enriched meal replacements, one conventional meal and two snacks (either a protein bar or a low-fat curd with fruit). After these 3 months, participants consumed one protein-enriched meal replacement, two meals and two snacks. The conventional diet group consumed three meals and two snacks with no meal replacements for the first 3 months. After these 3 months, participants consumed one standard meal replacement, two meals and two snacks each day. Participants received 15 sessions in groups and individually. After 12 months, 64.5% of the participants in the higher-protein group and 34.8% in the lower-protein group no longer met three essential criteria for metabolic syndrome. The difference between the groups was statistically significant (reported p < 0.05). It was reported that adverse events could not be related to the diets.

Pedersen et al. 160 recruited participants with type 2 diabetes mellitus, microalbuminuria and an eGFR of > 40 ml/minute/1.73 m² (mean BMI of 36 kg/m²). After a run-in phase, volunteers were randomly assigned to either a higher-protein (30% of energy) or a lower-protein (20% of energy) low-fat (30% of energy) weight-loss diet for 12 months. The energy content was approximately 1430 kcal/day. Participants attended a visit every 2 weeks during the 16-week weight-loss phase and then every month. Participants were provided with diet information booklets and a sample daily meal plan. Weight loss improved renal function, but there was no reported statistically significant difference between the diets.

Krebs et al. 134 recruited people with type 2 diabetes mellitus (mean BMI of 36.6 kg/m²), who were randomised to either a 30% protein and 30% fat energy diet or a 15% protein and 30% fat energy diet. Individuals participated in group sessions led by dietitians. Energy was reduced by approximately 500 kcal/day using an individualised dietary prescription based on estimated energy requirements. Portion charts and sample diet plans were provided. There was no difference in the number of adverse renal events between the groups. Two deaths (one in the higher-protein group and one in the lower-protein group) and health problems (two in the higher-protein group and three in the lower-protein group) were reported by 6 months; one more death (in the higher-protein group) and two more health problems (in the lower-protein group) were reported by 24 months.

Soenen et al. 177 recruited from an outpatient WMP and randomised participants to four groups: (1) a high-protein, low-carbohydrate diet (60% of energy from protein for weight loss and 30% of energy from protein for weight maintenance), (2) a high-protein, normal-carbohydrate diet (60% of energy from protein for weight loss and 30% of energy from protein for weight maintenance), (3) a normal-protein, low-carbohydrate diet (30% of energy from protein for weight loss and 15% of energy from protein for weight maintenance) and (4) a normal-protein, normal-carbohydrate diet (30% of energy from protein for weight loss and 15% of energy from protein for weight maintenance). For the weight-loss period, participants were prescribed 33% of their energy requirement. Participants had 19 sessions over 12 months.

Meta-analyses

Higher-protein diets produced favourable changes at 12 months102,107,119,134,160,177 but not in one trial at 24 months134 (Table 15). Consistent beneficial changes in risk factors at 12 and 24 months were not seen (Table 16).

Behaviour change technique coding summary (see Report Supplementary Material 1, Section 8)

Only one trial included supervised physical activity,102 and one provided physical activity advice. 107 The higher-protein diet groups received, on average, 7.2 BCTs (range 3–12 BCTs) and the comparator groups received, on average, 6.8 BCTs (range 4–12 BCTs). Two BCTs were used routinely and provided to both arms, namely setting a behavioural goal, and providing instruction on how to perform the behaviour. Dietary groups in the same trial largely received the same BCTs; however, in Soenen et al. ,177 the high-protein, normal carbohydrate group received three more BCTs than the comparator group (normal protein, normal carbohydrate): self-monitoring of outcomes, habit formation and information from a credible source. These results must be interpreted with caution, as group support was generally poorly described in both arms: only one trial was rated as well described134 and one was rated as adequately described. 102

Descriptions of the trials

Cheskin et al. 98 randomised people with type 2 diabetes mellitus to receive a standard diet or a portion-controlled diet of Medifast^® Plus for Diabetics (Medifast, Inc., Owings Mills, MD, USA) meal replacements. After 34 weeks, the standard diet group continued their diet at maintenance energy levels, whereas the portion-controlled participants were rerandomised to 26 weeks of meal replacements followed by the maintenance standard diet, or vice versa. The mean BMI was 36 kg/m².

The trials by Reichard et al. 168 and Shikany et al. 175 randomised people to follow a programme with meal replacements [Modified Stop Light and Medifast 5 & 1 Plan (Medifast, Inc., Owings Mills, MD, USA), respectively] or a food-based diet. The group mean BMI ranged from 40.4 to 45.6 kg/m².

The Flechtner-Mors et al. 119 trial is discussed in Higher-protein diets versus lower-protein diets.

Participants in all weight maintenance trials82,143,172,173 followed a VLCD for an initial weight-loss phase and were then randomised to weight maintenance conditions, in which VLCD products were used as meal replacements. These trials are discussed in Very low-calorie diets.

Meta-analyses

The meta-analyses for all studies and subgroups of studies reporting weight loss and weight maintenance are presented in Table 17. The analyses favoured meal replacements at 12 months but not thereafter.

Behaviour change technique coding summary (see Report Supplementary Material 1, Section 8)

Usually, each trial gave the same number and type of BCTs to both arms, with the exception of Shikany et al. ,175 which reported participants in the meal replacement group receiving 15 more BCTs than those in the diet group. Weight-loss arms were reported to receive fewer BCTs than weight maintenance arms. The meal replacement arms received an average of 9.5 BCTs, and the dietary comparators received an average of 5.25, compared with an average of 13.5 in the meal replacement arms and 13 in the diet arms of the weight maintenance trials. In all trials, instruction was given on how to perform a behaviour for all arms, as well as information from a credible source in all but the dietary arm of Shikany et al. 175 Other commonly used BCTs were goal-setting and adding objects to the environment. For weight loss, the number of BCTs given to the meal replacement arms varied greatly in number and type, ranging from 4 to 18 BCTs, and similarly in the dietary arms, ranging from 3 to 11 BCTs between trials. For weight maintenance, the use of BCTs was similarly varied, and ranged from 6 to 32 BCTs in the meal replacement arms and from 4 to 32 in the dietary arms. Only one trial aimed at weight loss and two aimed at weight maintenance reported providing physical activity advice to participants. The BCT use in the trials was well described in Shikany et al. ,175 adequately described in Lowe et al. 143 and poorly described in all other trials.

Descriptions of the trials

Participants with type 2 diabetes mellitus were randomised to receive a high-monounsaturated-fat diet (40% fat) or a high-carbohydrate diet (25% fat), with a deficit of 200–300 kcal/day, for 1 year in the trial by Brehm et al. 94 The mean participant BMI was 35.9 kg/m² (SD 3.3 kg/m²). At 12 months, weight loss was not different between the groups, as shown in Table 18.

Ebbeling et al. 115 randomised young adults aged 18–35 years with group mean BMIs of 36.6–37.2 kg/m² to a low glycaemic load (35% of energy from fat) or a 20% fat energy diet, with a deficit of 250–500 kcal/day. Both groups attended group workshops with one private counselling session and five motivational telephone calls with a dietitian over a 6-month intensive phase. As shown in Table 18, there was no difference in weight between the glycaemic load diet and the low-fat diet.

Participants with type 2 diabetes mellitus were randomised to receive a vegetarian or conventional diet with similar caloric restriction (500-kcal/day deficit) for 24 weeks. 130 All participants received instruction in nutrition and cooking and in aerobic exercise. Group mean BMIs ranged from 35.0 kg/m² (SD 4.6 kg/m²) to 35.1 kg/m² (SD 6.1 kg/m²). The vegetarian diet produced better weight loss at 12 and 18 months when data are both adjusted and unadjusted, but differences between groups were estimated with some uncertainty, as shown in Table 18.

Khoo et al. 133 randomised men with type 2 diabetes mellitus to a high-protein, low-fat, 600-kcal/day deficit diet for 12 months or to receive the same diet but with an initial 900-kcal/day formula diet (KicStart™; Pharmacy Health Solutions, Sydney, NSW, Australia) for 8 weeks. Mean BMIs of groups ranged from 35.1 kg/m² (SD 4.3 kg/m²) to 35.6 kg/m² (SD 4.8 kg/m²). At 12 months, there was no difference between groups, as shown in Table 18. Erectile function, sexual desire and urinary symptoms improved similarly in both groups.

Nackers et al. 156 randomised women to receive 1000- or 1500-kcal/day balanced diets, based on US guidelines. Both groups also received a standard behavioural lifestyle intervention over 12 months. The mean BMI of participants was 37.8 kg/m² (SD 3.9 kg/m²). The 1000-kcal/day diet produced better weight loss at 12 months, but the difference between groups showed some uncertainty, as shown in Table 18.

Women with type 2 diabetes mellitus and women with obesity and a family history of type 2 diabetes mellitus (FH participants) were separately randomised to receive a calorie restriction diet of 1000–1500 kcal/day (CAL) or the same calorie restriction and < 30% fat energy (CAL + FAT) in the Pascale et al. 159 trial. The mean BMI in the groups of participants with type 2 diabetes mellitus ranged from 36.3 kg/m² (SD 4.2 kg/m²) to 36.4 kg/m² (SD 4.7 kg/m²). The mean group BMI of FH participants ranged from 35.0 kg/m² (SD 4.4 kg/m²) to 36.1 kg/m² (SD 5.6 kg/m²) in the CAL + FAT group. Weight loss was greater in the CAL + FAT group at 12 months, but the difference between groups showed some uncertainty (see Table 18).

Rock et al. 170 randomised participants with type 2 diabetes mellitus to a higher-carbohydrate and lower-fat diet plan (LF diet), to a lower-carbohydrate and higher-fat diet plan (LC diet) or to usual care. Both plans were provided by Jenny Craig, Inc. (Carlsbad, CA, USA) as pre-packaged foods with weight-loss counselling. Mean group BMIs ranged from 36.2 kg/m² (SD 4.3 kg/m²) to 36.3 kg/m² (SD 4.4 kg/m²). Data at 12 months are shown in Table 18. Weight loss was better for the lower-carbohydrate and lower-fat groups than for the controls, but did not differ between diet plans.

Risk-of-bias assessment

The quality of randomisation sequence and allocation concealment were unclear in the majority of trials (83.3% and 91.7%, respectively). Most trials (83.3%) reported attempting to blind participants but were less clear on blinding personnel/health professionals delivering the intervention and outcome assessors. All trials were rated as being at high risk of bias owing to author links with companies that manufacture orlistat.

Eight trials96,105,117,127,135,152,176,183 included single-blind, placebo lead-in phases prior to randomisation of study participants, ranging from 2 to 4 weeks. The percentage of participants who did not continue to the randomised phase ranged from 4%183 to 28%. 96 Common reasons for not continuing included withdrawal of consent, loss to follow-up and adverse events. In two trials,105,117 only participants who achieved high pill-taking compliance (≥ 70%) were reported to be eligible for the randomised element of the trial. Participants who experienced difficulty with taking oral medication were therefore ineligible to be included in the randomised phases, or could have been more likely to withdraw prior to randomisation, thus limiting the generalisability of these trial results. The summary graph of the risk-of-bias assessments for the orlistat RCTs is presented in Report Supplementary Material 4, Figure E14.

Equity assessment

The trials either did not report or were considered to have not considered the majority of the equity assessment items, with the exception of unintended effects and harms, which were reported by all but one trial. 183 None of the trials reported sociodemographic differences between trial completers and participants who withdrew or were excluded, although all but one trial185 reported some baseline PROGRESS data, usually sex. Only one trial185 was considered to have excluded a specific population by excluding participants with diabetes mellitus and ongoing active CVD from the trial. As with the risk-of-bias assessment, all trials were considered to have a potential risk of author conflict. The summary of the equity assessment is presented in Report Supplementary Material 4, Figure E15.

Descriptions of the trials

In a 12-month trial of 120 mg of orlistat three times daily or placebo, Bakris et al. 86 reported no deaths in either group; 14 out of 267 participants (5.2%) in the control group and 15 out of 265 (5.7%) in the orlistat group reported serious adverse events requiring hospitalisation.

In a 12-month UK trial of 120 mg of orlistat three times daily or placebo, Broom et al. 96 reported serious adverse events in 13 out of 265 orlistat participants and in 17 out of 266 placebo participants, and one death attributable to cancer was reported in the orlistat group. In a second UK study, Broom et al. 95 randomised people with obesity and hypercholesterolaemia to receive 120 mg of orlistat or placebo three times daily for 24 weeks. Participants who completed the 24-week, double-blind phase were entered into a 28-week open-label phase and received 120 mg of orlistat three times daily and continued with the hypocaloric diet.

Davidson et al. 105 evaluated placebo or 120 mg of orlistat three times daily for 12 months. Participants receiving drug treatment for type 2 diabetes mellitus were not eligible to participate in the trial. Participants who had ≥ 70% compliance then entered a weight-maintenance phase for a further year. Participants treated with orlistat were rerandomised to receive placebo, 60 mg of orlistat or 120 mg of orlistat three times daily for the weight maintenance phase. Second year weight regain was reported to be significantly less with 120 mg of orlistat than with 60 mg of orlistat or placebo (reported p < 0.001). Participants who were originally randomised to receive placebo continued to receive placebo for the final year of the trial. One orlistat-treated woman was diagnosed with breast cancer during the first year and 1 out of 197 placebo-treated women and 3 out of 548 of women treated with 120 mg of orlistat were diagnosed with breast cancer during year 2.

Finer et al. 117 randomised participants to 120 mg of orlistat or placebo three times daily for 12 months in a UK trial.

Hauptman et al. 127 evaluated placebo, 60 mg of orlistat or 120 mg of orlistat treatment three times per day for 24 months. On four occasions during the first 12 months, participants viewed videos of behaviour modification techniques, and at four points during the second year, they followed the Live for Life programme (Johnson & Johnson Health Management Inc., New Brunswick, NJ, USA). With the exception of GI events, the incidence and types of adverse events were similar in all groups.

In the trial by Krempf et al. ,135 participants were randomised to 120 mg of orlistat or placebo three times daily for 18 months. People with type 1 or 2 diabetes mellitus were excluded from trial entry. Five serious adverse events were reported in the orlistat group and four were reported in the placebo group.

A 1-year trial by Miles et al. 152 allocated people with metformin-treated type 2 diabetes mellitus to either placebo or 120 mg of orlistat three times per day for 12 months. At 12 months, fasting serum glucose was reported as significantly decreased with orlistat compared with placebo (reported p = 0.001). Significantly more orlistat-treated participants than placebo-treated participants were reported to have discontinued treatment owing to an adverse event (reported p < 0.05), and more placebo group participants withdrew overall (reported p < 0.05).

Participants in the Sjöström et al. 176 trial were randomised to placebo or 120 mg of orlistat three times a day for 12 months. In year 2, participants were reassigned orlistat or placebo for a further 12 months (see Weight maintenance interventions without a very low-calorie diet weight-loss phase). The authors reported that, during year 2, participants on orlistat regained half as much weight as those who were changed to placebo (reported p < 0.001). In a subgroup of the trial, at 12 months, no significant difference was reported between groups for total body, forearm or lumbar spine bone mineral density. Only the urinary hydroxyproline-to-creatinine ratio was increased with orlistat compared with placebo, suggesting increased bone breakdown. Vitamin D status after 12 months did not differ between groups. Serious adverse events were reported by 24 participants in the placebo group and by 25 participants in the orlistat group during the first 12 months. One GI neoplasm occurred in a participant treated with placebo for 2 years.

The Swinburn et al. 183 trial sought to determine whether or not orlistat treatment could reduce the estimated absolute 10-year risk of having a CVD event versus placebo in people with type 2 diabetes mellitus in a trial of placebo or 120 mg of orlistat three times a day for 12 months. The authors reported that the baseline median 10-year risk of a CVD event was moderately low, at about 10% [0.089 (range 0.008–0.416) for the orlistat group and 0.105 (range 0.017–0.353) for the placebo group]. At 12 months, it was reported that, although there was no difference in the 10-year CVD risk between groups, the orlistat group had significant improvements in individual CVD risk factors compared with the placebo group. The orlistat group was also reported to have a significantly better score on the SF-36 vitality domain than the placebo group at 12 months (reported p = 0.006). Serious adverse events were reported for 9.4% of orlistat participants and 7.1% of placebo participants. The authors did not describe the nature of these events.

Torgerson et al. 185 randomised people with normal or impaired glucose tolerance to 120 mg of orlistat or placebo three times daily. Authors reported that at 4 years the cumulative incidence of diabetes mellitus was 9.0% with placebo and 6.2% with orlistat, corresponding to a risk reduction of 37.3% (reported p = 0.0032).

Meta-analyses

Table 19 shows the meta-analysis of 120 mg of orlistat versus placebo. For both unadjusted and adjusted data, there is evidence of a benefit favouring orlistat for weight loss at all time points. The results of the analysis of 60 mg of orlistat versus placebo, which also significantly favours orlistat for weight loss, are shown in Table 20. Table 21 shows the results of the comparison of 120 mg of orlistat with 60 mg of orlistat and showed no evidence of a difference in effect.

For all trials, the numbers of reported GI events were greater for participants treated with orlistat than for those treated with placebo. Table 22 presents the reported incidence of GI events for orlistat and placebo groups for each of the trials.

Orlistat treatment significantly improved all risk factors apart from triglyceride levels and HDL cholesterol at 12 months, with HDL cholesterol showing a fall (Table 23). At 24 and 48 months, risk factors, apart from HDL cholesterol, were still favoured by orlistat but with greater uncertainty around most estimates.

Behaviour change technique coding summary (see Report Supplementary Material 1, Section 8)

The WMP component reported physical activity advice in 8 out of the 12 trials, with no trial providing supervised physical activity. The orlistat groups received, on average, 5.1 BCTs (range 3–8 BCTs) and the comparator groups received, on average, 4.2 BCTs (range 2–7 BCTs) targeting behaviours. Setting a behavioural goal was used in both arms of all trials except one,152 and the same BCTs were given to both arms in all trials with the exception of pharmacological support. These results must be interpreted with caution as behavioural support was described poorly in both arms of all trials.

Descriptions of the trials

Mexican American women in the Poston et al. 165 trial were randomised to receive a 1-year culturally tailored lifestyle WMP with 120 mg of orlistat three times a day or a wait list control group. At 12 months, weight loss was greater in the orlistat group. Data are provided in Table 24.

Descriptions of the weight-loss trials

Participants in the trial conducted by Damschroder et al. 103 were randomised to receive a standard group-based weight-loss programme or to a lifestyle intervention [Aspiring for Lifelong Health (ASPIRE)], which was delivered either as in-person group sessions (ASPIRE-Group) or to individuals by telephone (ASPIRE-Phone) over 12 months. The mean BMI was 36.5 kg/m² (range 25.2–63.0 kg/m²).

Nilsen et al. 157 compared a WMP delivered to individuals by a physician with the same WMP combined with a group-based interdisciplinary programme. Participants had a mean BMI of 36.8 kg/m² (SD 6 kg/m²) and all were identified as having a high risk of type 2 diabetes mellitus as measured by the Finnish Diabetes Risk Score. 206,207

Participants in the Wadden et al. 192 trial were randomised to receive a group WMP, based on the LEARN (lifestyle, exercise, attitudes, relationships, nutrition) manual208 and conducted by a nutritionist, or the same treatment conducted by a psychiatrist experienced in weight control delivered to individuals. Group mean BMIs ranged from 36.2 kg/m² (SD 6.2 kg/m²) to 36.7 kg/m² (SD 3.6 kg/m²).

The SHINE (Support, Health Information, Nutrition and Exercise) trial, conducted by Weinstock et al. ,196 randomised people with metabolic syndrome to receive a telephone adaptation of the DPP intervention,205 delivered as individual calls or to groups via conference calls. The initial mean BMI was 38.9 kg/m² (SD 7.6 kg/m²) in the individual call group and 39.7 kg/m² (SD 8.3 kg/m²) in the conference call group. Six participants were diagnosed with diabetes mellitus during the trial, but their allocation was not given.

Descriptions of the weight maintenance trials

Kumanyika et al. 137 recruited African American women to receive a 10-week culturally adapted group counselling Health Eating and Lifestyle Programme (HELP) of 1200- to 1500-kcal/day. The mean BMI of participants was 37.0 kg/m². Following the weight-loss programme, participants were randomised to receive interventions designed to facilitate weight maintenance or additional weight loss [i.e. to HELP as group counselling classes, to individual self-directed HELP (Self-HELP) facilitated by HELP staff members (nutrition, exercise or behaviour change specialists) or to usual care, which involved semi-annual clinic visits with a physician but no further HELP intervention].

Latner et al. 140 evaluated a community WMP, randomly allocating participants to receive group-based continuing care following weight loss or to a standard individual-care condition. All participants then received a maintenance manual providing additional behavioural strategies and skills; however, the continuing care participants continued to meet in groups of 10–15 participants, with two participants volunteering to act as co-facilitators within each group. The mean BMI of participants was 35.8 kg/m².

Meta-analyses

Table 25 shows the meta-analysis of the comparison of group and individual programmes. Three trials140,157,192 had many more contacts with those participants receiving the group intervention, which may have affected results. For the overall comparison of group versus individual weight loss and weight maintenance programmes, there was a benefit from group programmes at 12 months, in unadjusted and adjusted analyses, and results favoured group programmes at other time points. Subgroups showed no consistent favouring of group or individual programmes.

Behaviour change technique coding summary (see Report Supplementary Material 1, Section 8)

Interventions comparing the effect of group programmes with individual programmes involved an in-person group meeting component in all trials but one (which used telephone calls196). Trials generally provided the same BCTs to both arms, with the exception of the Nilsen et al. 157 trial, which reported the group arm receiving six BCTs and the individual arm receiving only one (information from a credible source). Weight-loss trials typically reported giving more BCTs to trial groups than those aiming for weight maintenance, with an average of 15.1 BCTs (range 1–29 BCTs) versus 10 BCTs (range 9–11 BCTs), respectively. Only one BCT was used routinely in both weight loss and weight maintenance trials: a credible source providing arguments for weight loss or maintenance. Goal-setting, problem-solving and self-monitoring techniques were frequently used. In general, group support was described well or adequately in both the weight-loss and weight maintenance trials, apart from in one trial. 157

Descriptions of the trials

In the Weigh To Go trial, Dennison et al. 108 evaluated a computer-assisted nutrition programme to assist weight loss and weight maintenance in employees of a large US automobile manufacturing firm. The mean weight of participants was 100 kg, consistent with a BMI of ≥ 35 kg/m². Participants were randomised to receive the 8-week Weigh To Go programme classes either with computerised food intake and activity analysis entered by the participants (hands-on) or as a condition without computer assistance. The materials and activities were otherwise the same in both conditions. A control group received no nutrition programme. At 12 months, the authors reported that differences in weight between groups were not significant.

In the Choose to Lose study,114 participants were randomised to an enhanced lifestyle intervention or to a standard intervention, both including in-person counselling on a 500- to 1000-kcal/day deficit diet and physical activity based on the DPP. 205 Participants had a mean BMI of 37.8 kg/m² (SD 6.6 kg/m²). During the first 12 months, enhanced intervention participants received monthly telephone calls from the counsellor, and received weekly mailings of tailored and non-tailored videos and printed materials that focused on weight loss. In the second year, enhanced intervention participants received tailored and non-tailored materials for weight maintenance biweekly for the first 6 months and monthly thereafter. Enhanced intervention participants also received exercise feedback reports.

The trial by Spring et al. 178 evaluated the effectiveness of adding an electronic mobile personal digital assistant (+ mobile group) and telephone coaching to a group-based WMP. Participants had an initial mean BMI of 36.3 kg/m² (SD 4.6 kg/m²).

The trial conducted by Wylie-Rosett et al. 203 evaluated a WMP delivered via a workbook only, a workbook plus additional computerised assistance, and the addition of computers and staff consultations, including up to 18 telephone or face-to-face consultations with a registered dietitian and/or a cognitive–behavioural therapist. The mean BMI of participants was 35.6 kg/m² (SD 6.5 kg/m²).

Meta-analyses

Table 26 shows that adding telephone or internet support to an in-person WMP favoured weight loss at 12 months, but not at later follow-up times in the one trial114 longer than 12 months.

Behaviour change technique coding summary (see Report Supplementary Material 1, Section 8)

The in-person only groups received, on average, 12.8 BCTs (range 3–18 BCTs) and the enhanced intervention groups received, on average, 16.3 BCTs (range 8–24 BCTs). Only one BCT was used routinely and provided to both arms, namely using a credible source. The WMPs in both arms of the trials were the same. Eaton et al. 114 gave additional information about antecedents, demonstration of the behaviour, grade tasks, body changes, framing/reframing and self-talk to the enhanced intervention group. Spring et al. 178 also gave more BCTs to the enhanced intervention group, namely goal-setting, feedback on and self-monitoring of the behaviour, and graded tasks. Wylie-Rosett et al. 203 used the greatest number of BCTs of the four trials, with 24 in the additional support arm (providing extra goal outcome, action-planning and verbal persuasion BCTs) and 21 in the comparator arm. BCTs were well described in two trials,114,203 and adequately described in the remaining two. 108,178

Descriptions of the weight-loss trials

Appel et al. 83 compared delivery of a WMP by remote support via telephone, a study website and e-mail with remote support and in-person support during group and individual sessions over 24 months. The initial mean BMI of participants was 36.6 kg/m² (SD 5.0 kg/m²). A third self-directed control group received brochures and a list of websites promoting weight loss at baseline only.

The E-LITE (Evaluation of Lifestyle Interventions to Treat Elevated Cardiometabolic Risk in Primary Care) study84 compared usual care, usual care plus in-person, coach-led support and individual self-directed remote support WMPs in people with prediabetes and/or metabolic syndrome (mean BMI ≥ 35 kg/m²).

Participants in the trial by Damschroder et al. 103 (see earlier description in Group programmes versus individual programmes) were randomised to the Veterans’ Affairs standard MOVE! weight-loss programme delivered by expert clinicians, or to two ASPIRE programmes delivered by non-clinician lifestyle coaches.

Little et al. 142 randomised participants with hypertension, hypercholesterolaemia or diabetes mellitus to receive a weight-loss programme: Positive Online Weight Reduction (POWeR+), delivered via the internet with either remote (POWeR+ R) or face-to face (POWeR+ F) nurse support. The mean group BMIs ranged from 36.3 kg/m² (SD 5.7 kg/m²) to 37.1 kg/m² (SD 6.0 kg/m²). In addition to the web-based intervention, POWeR+ F participants received three scheduled and four optional face-to-face sessions, plus 6-monthly weighing sessions. Face-to-face contact was replaced with three scheduled telephone or e-mail contacts and two optional telephone or e-mail contacts for the POWeR+ R participants. A control group received 6-monthly nurse follow-up sessions only. The POWeR trial is also discussed in Chapter 4.

Description of the weight maintenance trial

In the Perri et al. ,162 women with from rural communities who had completed a 6-month weight-loss programme were randomised to two extended-care counselling interventions, delivered by telephone or face to face, or to an education control group for a further 6 months. The participants’ mean BMI was 36.8 kg/m² (SD 4.9 kg/m²).

Meta-analyses

Table 27 shows the meta-analysis of in-person versus telephone or internet delivery of a WMP. The trial by Azar et al. 84 was excluded from this analysis because it was not possible to identify the numbers of participants with a BMI of > 35 kg/m², but it reported a greater reduction in BMI and body weight in the coach-led intervention than in the self-directed intervention (p < 0.05). For the meta-analysis of interventions targeted at both weight loss and weight maintenance, in-person intervention delivery was favoured at 12 and 24 months for adjusted and unadjusted data, with evidence of a small effect.

Behaviour change technique coding summary (see Report Supplementary Material 1, Section 8)

Interventions comparing the effects of in-person support programmes with those delivered remotely by telephone or via the internet generally provided the same BCTs to both comparisons, with one exception84 in which the in-person group received a demonstration of the behaviour and the remote group did not. Weight-loss trials gave an average of 13.6 BCTs (range 1–17 BCTs), and the one weight maintenance trial162 gave 20 BCTs. Several BCTs were used routinely in weight loss and weight maintenance trials, namely goal-setting, problem-solving, self-monitoring and social support. BCT support was well described in all trials.

Descriptions of the trials

Bartels et al. 87,88 compared a 12-month fitness and weight-loss programme tailored for people with serious mental illnesses (In SHAPE) with a 12-month membership of a fitness club in two trials. The In SHAPE intervention was delivered in weekly sessions by a certified fitness trainer/health coach who received instruction on healthy eating and nutrition training in tailoring individual plans for people with serious mental illness. The intervention included healthy eating discussions with a registered dietitian, including group cooking classes. The control group participants received 12-month membership to the same local fitness club and educational materials on the benefits of exercise and healthy eating. In SHAPE participants received 56 intervention sessions, whereas control participants received four. From the two trials, the mean group BMI ranged from 36.2 kg/m² (SD 7.6 kg/m²) to 38.3 kg/m² (SD 8.5 kg/m²).

Meta-analyses

Table 28 shows the meta-analysis of fitness club trainer (In SHAPE programme) versus fitness club membership. The results show no differences at either time point.

Behaviour change technique coding summary (see Report Supplementary Material 1, Section 8)

Fitness trainer groups in both trials reported receiving six BCTs and the comparator groups reported receiving three or four BCTs. Both trials gave demonstration of the behaviour and a material reward to both arms of the trial, and the fitness trainer arms of both trials additionally received goal-setting, social support, and instruction on how to perform the behaviour and use a credible source. The comparator groups of both trials received information about health consequences, which was not reported for the fitness trainer groups. The BCTs were adequately described across both trials.

Descriptions of the trials

Hakala et al. 125 compared a 12-month group weight-loss programme, comprising an initial 2-week intensive weight reduction period in an inpatient rehabilitation setting and then a community programme, with an individual, community-based programme. Mean group BMIs ranged from 41.7 kg/m² (SD 3.1 kg/m²) to 43.6 kg/m² (SD 4.8 kg/m²). The inpatient intervention comprised intensive behavioural and educational group sessions along with a prescribed physical activity programme and occupational therapy, as well as individual nutritionist (1200 kcal/day) and physician counselling. The community intervention involved the same dietary intervention and individual physician counselling.

In a second trial, Hakala124 conducted a similar comparison between an initial 3-week intensive inpatient rehabilitation programme followed by a community programme, and a community-based weight-loss programme delivered over 2 years. The mean group BMI ranged from 37.7 kg/m² (SD 2.3 kg/m²) to 40.5 kg/m² (SD 3.9 kg/m²).

The trial by Torgerson et al. 184 (see Very low-calorie diets) compared delivery of a 456-kcal/day VLCD weight-loss programme starting with a week as a hospital inpatient (VLCD – metabolic ward) with the same VLCD treatment delivered without the hospital stay (VLCD – strict). A third group followed the VLCD treatment but participants were allowed two small meals weekly (VLCD – plus). All participants underwent VLCD for 16 weeks, followed by a 500-kcal/day deficit diet for 36 weeks.

Meta-analyses

Table 29 shows evidence that favoured inpatient sessions at 12 months for unadjusted data, but did not favour inpatient sessions thereafter, with evidence against provision at 24 months only. When data were adjusted for dropouts and withdrawals, meta-analysis did not favour inpatient sessions at any time point, again with evidence against provision at 24 months.

Behaviour change technique coding summary (see Report Supplementary Material 1, Section 8)

The inpatient groups received, on average, 8.7 BCTs (range 7–12 BCTs) and the comparator groups received, on average, 6 BCTs (range 3–10 BCTs). All trials used a credible source in both arms and gave instruction on how to perform the behaviour. The 1993 trial by Hakala et al. 125 gave additional social support, behavioural rehearsal and restructuring of the physical environment to the inpatient group, and the 1994 trial in the same group of participants124 gave additional behavioural rehearsal, restructuring of the social and physical environments and additional environmental object BCTs to the inpatient group. The trials used largely the same BCTs in the inpatient arms. Torgerson et al. 184 gave five additional BCTs targeting goal-setting, action-planning, problem-solving and feedback on and self-monitoring of behaviour to the inpatient arm. However, these results should be interpreted with caution because group support was poorly described across all three trials.

Mindfulness

Daubenmier et al. 75 evaluated adding mindfulness techniques to a WMP compared with a WMP-only intervention. The authors reported that mindfulness techniques were based on stress reduction,209 raising awareness about eating210–212 and walking. 213 Both intervention arms attended 16 sessions lasting 2–2.5 hours and one day session spread over 5.5 months. Content in the control intervention was reduced by 30 minutes in sessions 9–16. Mindfulness was delivered by meditation instructors and one dietitian. The control intervention was run by dietitians. The group mean BMI ranged from 35.4 kg/m² (SD 3.5 kg/m²) to 35.6 kg/m² (SD 3.5 kg/m²). At both 12 and 18 months, adjusted and unadjusted data favour the mindfulness intervention (Table 30).

Skill-based counselling

Yeh et al. 77 compared skill-based counselling with standard counselling. Skill-based counselling aimed to provide skills to support overcoming barriers to healthy eating. Both interventions were delivered by a dietitian [it was unclear if the same dietitian(s) were involved in both], with skill-based counselling participants having 19 contacts and standard counselling participants having 10 contacts during the first 12 months. Participants were contacted once in the second year. The mean BMIs of women in the trial arms ranged from 36.3 kg/m² (SD 5.4 kg/m²) to 37.9 kg/m² (SD 6.7 kg/m²). Data at 12 and 24 months are shown in Table 30. There was little difference between types of counselling.

Motivational interviewing

West et al. 78 investigated adding motivational interviewing to a group-based behavioural WMP in women with type 2 diabetes mellitus. All participants received the WMP, which was delivered in 42 sessions by a team including a behaviourist, a nutritionist, an exercise physiologist and a diabetes mellitus educator. The mean BMI of participants was 36.5 kg/m² (SD 5.5 kg/m²). Weight loss was emphasised for the first 6 months and weight maintenance was emphasised for the following 12 months. All participants were prescribed a low-fat diet of 1200–1500 kcal/day and were encouraged to achieve a goal of ≥ 150 minutes per week of physical activity. Motivational interviewing was delivered in five sessions by a counsellor who focused on eliciting change talk, commitment language and how to motivate change. Individual health education sessions acted as an attention placebo and matched motivational interviewing sessions for frequency and duration. Data for 12 and 18 months are shown in Table 30. Motivational interviewing was superior to the attention control at 12 and 18 months.

Cognitive–behavioural therapy

Two trials141,146 investigated CBT compared with standard behavioural therapy for women. Participants in the Linde et al. 141 trial had a group mean BMI of 39.5 kg/m² (SD 7.7 kg/m²) to 38.6 kg/m² (SD 6.8 kg/m²). Manzoni et al. 146 had groups with mean BMIs from 39.2 kg/m² (SD 5.3 kg/m²) to 41.8 kg/m² (SD 6.3 kg/m²). Manzoni et al. 146 also investigated using virtual reality enhanced CBT (CBT + VR), in which participants were able to develop and practise weight-management techniques in response to trigger situations within virtual environments. The authors reported that weight-loss results from CBT + VR did not differ from those from CBT. In one trial,141 women had clinical depression. Both trials included a prescribed calorie diet (1200–1500 kcal/day) and physical activity goals. A brief description of the interventions is provided in Report Supplementary Material 5, Table E13. Meta-analysis favoured CBT at 12 months for both unadjusted and adjusted data (Table 31).

Psychodynamic approach

Beutel et al. 79 compared a psychodynamic approach with a behavioural approach, both delivered in an inpatient rehabilitation setting to mostly female (85%) participants with psychiatric and somatic comorbidities over 7 weeks. In both arms, the therapists were psychological psychotherapists or psychiatrists. Both approaches used group therapy. The psychodynamic approach had more emphasis on individual therapy. Trial participants had a mean BMI of 44.3 kg/m² (range 35.1–73.5 kg/m²). Adjusted and unadjusted data at 12 and 26 months favoured the psychodynamic treatment (mostly at 12 months) (Table 32).

Counselling

The Steps to Health study80 evaluated two existing employee health programmes. Employees were randomised to an intensive behavioural intervention (weight management plus) or to a healthy lifestyle educational programme (weight management). In addition to the different emphasis on behaviour versus education, the programmes differed in intensity, frequency and resource requirements. Weight management included one face-to-face meeting with a counsellor in the first month and two telephone counselling and feedback sessions at months 6 and 12, with monthly mailings of educational material. Weight management plus participants were offered monthly counselling sessions (face to face for months 1, 4, 8 and 12, with the rest by telephone), met with an exercise physiologist for months 2 and 5 and received quarterly biometric feedback and educational materials. Weight management plus participants were also encouraged to use eHealth trackers for diet and weight. Weight management plus participants received an additional 10 sessions during the first 12 months. The mean BMI of the trial participants was 37.2 kg/m² (SD 6.4 kg/m²). Table 32 shows that the weight management plus programme was superior when data were adjusted and unadjusted at 14 months, but the differences were small and rule out any meaningful difference between the strategies.

Acceptance-based counselling

Lillis et al. 74 randomised people to an acceptance-based behavioural (counselling) intervention (ABBI) or standard behavioural treatment. Psychologists, exercise physiologists and nutritionists delivered both programmes in 32 group sessions over 12 months. Participants received the same diet of 1200–1800 kcal/day and exercise goal of 250 minutes per week. All participants were taught standard behavioural strategies. ABBI was based on Acceptance and Commitment Therapy214 and used acceptance and mindfulness techniques to promote detachment from problematic thoughts and commitment to values consistent with the target behaviour. Standard behavioural treatment used cognitive and emotional control strategies to address negative thoughts and emotions. The mean BMI of trial participants was 37.6 kg/m² (SD 5.3 kg/m²). Data for 12, 18 and 24 months are shown in Table 32. Standard behavioural treatment was favoured at 12 and 18 months, and ABBI was favoured at 24 months.

Social cognitive theory-based programme

The trial by Annesi81 recruited women who were physically inactive to two different WMPs based on social cognitive theory or self-efficacy theory. 215,216 The social cognitive theory programme had group sessions, emphasising self-regulatory skills for weight loss and weight maintenance. Assigned daily energy intake was based on an individual’s weight (e.g. 79–99 kg = 1500 kcal/day).

Participants in the self-efficacy group reviewed chapters in the LEARN manual,208 and were recommended to limit energy intake to 1200 kcal/day. Lessons were followed by 15-minute telephone support sessions with a wellness leader. Both interventions were delivered over 6 months and addressed the concept of body image and its role in weight management. All participants were recommended to do 150 minutes per week of moderate exercise. The mean BMI of trial participants was 35.4 kg/m² (SD 3.3 kg/m²). Data for 12 and 24 months are shown in Table 32. The social cognitive theory-based programme was favoured at both time points.

Descriptions of the trials

The SHARE (Supporting Health Activity and eating Right Everyday) trial conducted by Kumanyika et al. 138 with African American men and women (90% were women) chose to enrol participants alone (individual stratum) or with one or two family members or friends (family stratum). Participants were then randomised to high or low social support within each stratum. Participants’ mean group BMIs ranged from 38.4 kg/m² (SD 6.2 kg/m²) to 39.3 kg/m² (SD 6.2 kg/m²). The trial by Wing et al. 199 randomised people with type 2 diabetes mellitus, and their spouses, also with obesity with/without diabetes mellitus, to attend a 20-week behavioural weight-loss programme together with their spouses or alone. Group mean BMIs ranged from 35.7 kg/m² (SD 5.8 kg/m²) to 36.6 kg/m² (SD 5.8 kg/m²). Details of the interventions for both trials are presented in Report Supplementary Material 5, Table E14.

Meta-analyses

Table 33 shows that there was no evidence of a difference between groups at any time point for either adjusted or unadjusted data. In the Kumanyika et al. 138 trial, partner success with weight loss was associated with greater participant weight loss. Spouses of both sexes lost more weight in the together group than in the alone group in the Wing et al. 199 trial (reported p < 0.05).

Behaviour change technique coding summary (see Report Supplementary Material 1, Section 8)

Both the high-support and low-support groups received, on average, 13 BCTs in the trial by Kumanyika et al. ,138 which used four more BCTs in its groups than the trial by Wing et al. 199 The BCTs provided were largely similar in both treatment arms, with the exceptions of practical social support to the high-support group, and information about health consequences to the low-support group in the Kumanyika et al. 138 trial. Goal-setting, problem-solving, self-monitoring of the behaviour, social support, instruction on how to perform the behaviour, use of a credible source and a material reward were used in all trials. These results should be interpreted with caution as support was poorly described in one trial138 and adequately described in the other. 199

Descriptions of the trials

Three trials85,151,167 compared interventions emphasising weight loss with interventions that were weight neutral with an emphasis on healthy lifestyles rather than weight loss. In two of these trials,85,151 the WMPs were both based on the LEARN programme, and a standard cognitive–behavioural programme was delivered in the remaining trial. 167 The weight-neutral groups in all trials employed a range of behavioural and counselling techniques. All three trials recruited women and excluded those with obesity-related comorbidities (e.g. diabetes mellitus and hypertension). Mean BMIs in the groups in the trials ranged from 35.2 kg/m² (SD 6.1 kg/m²) to 38.6 kg/m² (SD 3.9 kg/m²). Details of the interventions are presented in Report Supplementary Material 5, Table E15.

Meta-analysis

Meta-analysis favoured weight-loss interventions, with evidence of a difference at all time points (unadjusted and adjusted data) (Table 34).

Behaviour change technique coding summary (see Report Supplementary Material 1, Section 8)

The BCTs provided in the trials varied between the arms, both in type and in number, with the weight-loss arms typically receiving more BCTs than those aiming for weight neutrality [i.e. an average of 14 BCTs (range 5–28 BCTs) vs. seven BCTs (range 5–9 BCTs), respectively]. In one trial,151 the weight-loss group received 28 BCTs and the comparator group received only five. Two BCTs were used routinely in weight-loss and weight-neutral groups, namely instruction on how to perform a behaviour and information from a credible source. In general, group support was described well or adequately across all trial arms.

Descriptions of the trials

The Think Health! study136 was set in health-care practices that reached African American and Hispanic people in the USA. Participants had an initial mean BMI of 37.2 kg/m² (SD 6.4 kg/m²) and were randomised to receive counselling from a primary care provider and lifestyle coach or to receive counselling from a primary care provider only.

In the Pounds Off With Empowerment (POWER) trial,148 people with diabetes mellitus living in medically underserved rural communities received an ILI delivered primarily via individual counselling or a condensed ‘reimbursable’ intervention delivered in group and individual sessions. The initial mean BMI of participants was 36.7 kg/m².

The rural LITE (Lifestyle Intervention Treatment Effectiveness) trial163 was set in Cooperative Extension Offices in rural communities of the USA and randomised participants to low, moderate or high doses of behavioural treatment or control. Participants had a mean BMI of 36.3 kg/m² (SD 4.0 kg/m²).

Practice based Opportunities for WEight Reduction trial at the University of Pennsylvania (POWER-UP)195 randomised people with at least two out of five components of the metabolic syndrome to receive brief or enhanced brief lifestyle counselling. In both conditions, participants received counselling from primary care providers and lifestyle coaches; in the enhanced arm, participants were also able to choose sibutramine, orlistat or meal replacements to increase weight loss (data without sibutramine used here).

In the remaining LITE trial,132 participants were randomised to receive intensive counselling from a nutritionist over 10 visits or to short-term counselling over two visits. The initial mean BMI of participants was 35 kg/m² (SD 5 kg/m²). Brief details of the interventions are presented in Report Supplementary Material 5, Table E16.

Meta-analyses

When data are unadjusted, meta-analysis favours intensive counselling for weight loss for all studies, and the subgroup of studies based on the Look AHEAD/DPP-based interventions, with evidence of an effect at all available time points. When data are adjusted, meta-analysis also favours intensive counselling at 12 and 24 months for all studies and the Look AHEAD/DPP-based interventions (Table 35).

Behaviour change technique coding summary (see Report Supplementary Material 1, Section 8)

The intensive trial arms received, on average, 13.4 BCTs (range 6–19 BCTs), and the less intensive arms received 11.6 BCTs on average (range 4–19 BCTs). The BCTs routinely used in both arms of the trials were goal-setting, social support, instruction on how to perform the behaviour and information from a credible source. In the trial by Keränen et al. ,132 the more intensive intervention arm received six more BCTs than the less intensive arm. However, in general, the trials provided the same BCTs to both arms but had more sessions in the more intensive arm. All trials except Keränen et al. 132 provided physical activity advice to both trial arms. BCTs were described well in three of five trials;136,163,195 the remaining two were described poorly. 132,148

Descriptions of the trials

In the Villareal et al. 189 trial, participants’ group mean BMIs ranged from 36.9 kg/m² (SD 5.4 kg/m²) to 37.3 kg/m² (SD 4.7 kg/m²). Bone loss at the total hip was relatively less in the diet plus exercise group (–1.1%) than in the diet group (–2.6%), and bone mineral density increased in the exercise group (reported p < 0.001). The authors also reported that cognitive function, as measured by the Modified Mini-Mental State Examination, word fluency test and IWQOL, improved more in the active interventions than in the control interventions (reported p = 0.0001 to 0.04) and improved more in the diet plus exercise group than in the diet group, but not more than the exercise group.

Wadden et al. 194 randomised participants to a diet-only group, a diet plus aerobic exercise group, a diet plus strength training group and a diet plus aerobic and strength training group. The mean BMI of participants was 36.5 kg/m² (SD 5.7 kg/m²).

Wing et al. 201 recruited people without type 2 diabetes mellitus [mean BMI 35.9 kg/m² (SD 4.3 kg/m²)] who had one or both parents with diabetes mellitus. The authors reported that, at 24 months, 21 participants (17%) developed diabetes mellitus over the course of the study. In the control group, 7% (n = 2) developed type 2 diabetes mellitus, compared with 30.3% of diet (n = 10), 14% of exercise (n = 4) and 15.6% of diet plus exercise conditions (n = 5) (reported p = 0.079).

Meta-analyses

The results presented in Table 36 show that our meta-analyses favour diet and exercise at 12 and 24 months when data are both adjusted and unadjusted.

Behaviour change technique coding summary (see Report Supplementary Material 1, Section 8)

All diet plus exercise arms of the trials provided supervised physical activity, and the comparator groups were typically instructed not to change their activity levels. The diet plus exercise groups received 16 BCTs on average (range 11–25 BCTs) and the comparator groups received 13.3 BCTs on average (range 7–25 BCTs). Goal-setting, self-monitoring of behaviour and information about antecedents were used in both arms of all trials, as was the use of a credible source. Wing et al. 201 provided the same BCTs to both groups; however, Villareal et al. 189 and Wadden et al. 194 provided fewer BCTs to their diet-only groups. BCTs were adequately described in two trials189,201 and poorly described in the third. 194

The effect of Look AHEAD on body weight

Long-term weight differences between the ILI group and diabetes mellitus support and education are presented in Table 38. Absolute weight loss in both groups is likely to be less than in participants without type 2 diabetes mellitus, particularly because some medicines for type 2 diabetes mellitus (e.g. sulphonylureas and insulin) promote weight gain.

Mean weight loss from baseline was 6.0% in the intervention group and 3.5% in the control group. Participants with BMIs of 35 to < 40 kg/m² or ≥ 40 kg/m² had similar percentage weight losses to participants with BMIs of 30 to < 35 kg/m². 43 The authors discussed whether or not the educational sessions in the control group may have lessened differences between groups. Statins were used more often in the control group. 41 Of the 10% of Look AHEAD participants reporting recent binge eating at recruitment, two-thirds did not report binge eating after 1 year, and 3% reported new binge eating at 1 year. Participants who continued to binge eat or started to binge eat had significantly less weight loss at 1 year. 218

Gorin et al. 218 reported that, at 1 year, partners of participants in the intensive lifestyle arm lost 2.2 kg (SD 4.5 kg), compared with partners in the control arm, who lost 0.2 kg (SD 3.3 kg) (reported p = 0.012).

The effect of Look AHEAD on cardiovascular disease

The trial was stopped early at a median follow-up of 9.6 years on the basis of a futility analysis for the primary composite cardiovascular outcome.

Occurrence of death from cardiovascular causes, non-fatal myocardial infarction, non-fatal stroke or hospitalisation for angina was reported in 403 participants in the intervention group and 418 in the control group, with no between-group difference (HR 0.95, 95% CI 0.83 to 1.09; p = 0.51). 41 The intervention did not affect the incidence of atrial fibrillation (HR 0.99, 95% CI 0.77 to 1.28). 219

Differences in cardiovascular risk factors diminished over time, with HbA_1c and systolic blood pressure showing the most beneficial sustained difference. LDL was lower in the control group. 41

The effect of Look AHEAD on type 2 diabetes mellitus

After a median of 8 years, the incidence of very-high-risk chronic kidney disease was lower in the intervention group than in the control group, with incidence rates of 0.91 cases per 100 person-years in the control group and 0.63 cases per 100 person-years in the intervention group (difference of 0.27 cases per 100 person-years, HR 0.69, 95% CI 0.55 to 0.87; reported p = 0.0016). 44 Over 4 years, 3.5% (95% CI 2.7% to 4.3%) of intervention participants had a continuous remission of type 2 diabetes mellitus, compared with 0.5% (95% CI 0.2% to 0.8%) of control group participants (reported p = 0.02). 52

Participants were followed up for neuropathy over 9–11 years. There was no reported significant between-group difference for the Michigan Neuropathy Screening Instrument physical examination score (≥ 2.5 being indicative of diabetic neuropathy). 220

The effect of Look AHEAD on obstructive sleep apnoea

A total of 264 participants were recruited at four Look AHEAD sites to study the intervention effects on OSA. Remission of OSA at 48 months was reported to be five times greater in the intervention group (20.7%) than in the control group (3.6%). 51

The effect of Look AHEAD on hepatic steatosis

A substudy examined weight-loss effects on hepatic steatosis and incident NAFLD in one centre. Proton magnetic resonance spectroscopy was performed in 151 participants at baseline, and 96 (46 in the intervention group and 50 in the control group) were evaluated at 12 months, after dropouts and exclusions. The median per cent decrease in steatosis was –50.8% in the intervention group and –22.8% in the control group (reported p = 0.04). At 12 months, 1 out of 31 intervention participants (3%) and 6 out of 23 control participants (26%) without NAFLD at baseline developed NAFLD [reported odds ratio (OR) 0.07, 95% CI 0.007 to 0.71]. 45

The effect of Look AHEAD on knee pain and physical function

A subsample of 2889 participants who reported no knee pain at baseline were followed up for 48 months. Pain was assessed with the WOMAC questionnaire. At 48 months, knee pain was newly reported in 25.9% of the intervention participants and 27.2% of the control participants (reported RR 0.95, 95% CI 0.84 to 1.07). 46 At all annual assessments to year 8, intervention participants reported less knee pain than controls, but more pulled or strained muscles. 221

Physical function, assessed by the Medical Outcomes Survey Short-Form (SF-36), led to a significant improvement in the intervention group compared with the controls at 1 year, which was maintained for 8 years, with greater benefit for older participants. 221

The effect of Look AHEAD on bone loss and fractures

After 4 years, there was a small increase in bone loss at the hip from the intervention in men but not in women. 222 After a median follow-up of 11.3 years, a total of 731 participants had a confirmed incident fracture: 358 in the control group and 373 in the intervention group. Compared with the control group, the intervention group had a reported statistically significant 39% increased risk of a frailty fracture (HR 1.39, 95% CI 1.02 to 1.89), although this difference was reported as not statistically different for all fractures (HR 1.03, 95% CI 0.89 to 1.19) or hip fractures (HR 1.69, 95% CI 0.93 to 3.07). 223

The effect of Look AHEAD on urinary incontinence and erectile dysfunction

In a substudy, 1910 men were examined for urinary incontinence at 1 year. The intervention group had an increased odds of incontinence resolving (OR 1.93, 95% CI 1.04 to 3.59) and reduced new-onset urinary incontinence (OR 0.66, 95% CI 0.42 to 1.02). 49 In a substudy at five sites, there was evidence after 1 year of less worsening in erectile dysfunction. 50

In another substudy, 2739 women were examined for urinary incontinence at 1 year. The intervention was associated with a statistically significant reduction in having urinary incontinence at 1 year (OR 0.80, 95% CI 0.65 to 0.98), but not of resolution of urinary incontinence (OR 1.20, 95% CI 0.97 to 10.21). 48

The effect of Look AHEAD on depression and cognitive function

At 10–13 years after randomisation, the intervention was not significantly associated with overall or domain-specific cognitive function. 224 Using the Beck Depression Inventory over 8 years, the intervention reduced the incidence of mild or more serious depression (HR 0.85, 95% CI 0.75 to 0.97). 41

Meta-analyses

Subgroup analyses were performed in accordance with the physical activity and diet prescription. The three major categories refer to (1) trials in which participants were provided with a physical activity programme to attend, (2) trials that provided physical advice only and (3) trials that did not report providing either. Within these three groups, trials were classified in accordance with their dietary prescription into the following subgroups: (1) trials in which a specific number of calories was prescribed, (2) trials in which a caloric deficit was prescribed or (3) trials in which the caloric deficit or calorie intake was not described.

Results for the meta-analyses are presented for weight data adjusted for dropouts at 12 months and 24 months (Figures 1 and 2). Only Look AHEAD provided data after 30 months. In both meta-analyses, the marked heterogeneity of weight loss is very apparent (I² = 96% in both cases). Funnel plots did not suggest small-study bias.

FIGURE 1.

Meta-analysis of weight change (kg) in trials of WMP vs. usual care or attention control at 12 months. WMD, weighted mean difference.

FIGURE 2.

Meta-analysis of weight change (kg) in trials of WMP vs. usual care or attention control at 24 months. WMD, weighted mean difference.

At 12 months, those trials with a physical activity programme providing a specific calorie intake (n = 2 trials)41,201 had a mean difference weight change of –7.03 kg (95% CI –9.21 to –4.84 kg), mainly influenced by Look AHEAD. Trials with a prescribed calorie deficit and physical activity programme had a mean difference weight change of –7.44 kg (95% CI –11.13 to –3.75 kg).

Trials with physical activity advice only generally had lower weight loss. Two small trials that did not describe providing exercise advice, prescribing a VLCD180 or a 1200- to 1300-kcal/day diet with a CBT programme179 (n = 2 trials, mean difference –11.02 kg, 95% CI –14.93 to –7.11 kg) had the greatest effect on weight.

To provide sufficient data for meta-analysis, data from 18 and 24 months of follow-up were combined in Figure 2 (whichever was the later available time point for each trial). At 24 months, Look AHEAD41 continued to show good weight loss. In addition, the trial by Esposito et al. ,116 with a calorie intake of 1700–1900 kcal/day for men, with supervised, individually tailored physical activity sessions, demonstrated very good weight loss. The small trial by Stahre and Hällström179 also continues to demonstrate effective weight loss.