The clinical effectiveness and cost-effectiveness of long-term weight management schemes for adults: a systematic review

Article history

The research reported in this issue of the journal was commissioned by the HTA programme as project number 08/49/01. The contractual start date was in July 2009. The draft report began editorial review in February 2010 and was accepted for publication in June 2010. As the funder, by devising a commissioning brief, the HTA programme specified the research question and study design. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the referees for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

none

Copyright statement

© 2011 Queen’s Printer and Controller of HMSO. This journal is a member of and subscribes to the principles of the Committee on Publication Ethics (COPE) (http://www.publicationethics.org/). This journal may be freely reproduced for the purposes of private research and study and may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NETSCC, Health Technology Assessment, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

2011 Queen’s Printer and Controller of HMSO

Description of the underlying health problem

It has been estimated that globally, in 2005, approximately 1.6 billion adults (aged 15 years and over) were overweight and at least 400 million adults were obese. 7 A recent systematic review estimated that across European countries the prevalence of obesity (BMI > 30 kg/m²) ranged from 4% to 28.3% in men, and 6.2% to 36.5% in women. 8 A 2009 NHS Information Centre report,9 drawing on data from the Health survey for England 2007 (HSE)10 states that in England in 2007, 65% of men and 56% of women were overweight or obese. In Wales in 2007, 57% of adults were classified as overweight or obese, including 21% obese. 11 In 2008, 68.5% of men and 61.8% of women aged 16 years or over in Scotland were overweight or obese. 12 Data from Northern Ireland in 2005/6 show that 59% of adults were either overweight (35%) or obese (24%), and rates were similar across genders. 13

Prevalence of obesity varies by age as well as by gender. In the 2009 NHS Information Centre report9 prevalence of overweight and obesity varied by age, with high rates in older age groups (Table 2). In the Northern Ireland health and social wellbeing survey 2005/06, obesity (BMI > 30 kg/m²) was most prominent among those aged between 35 and 64 years. 13

Overweight or obesity in women is more common in lower income households (defined using equivalised household income which takes account of the number of people in the household) than in women in the highest income households. 9 In men this pattern is not seen. Other demographic characteristics potentially associated with overweight and obesity include urbanisation, marital status and ethnicity. Available data for rates of obesity from the HSE 2004 report showed that men from Bangladeshi and Chinese minority ethnic groups had the lowest prevalence (both 6%) while Black Caribbean and Irish men had the highest prevalence (both 25%). 9 There are also regional differences in England in the prevalence of overweight and obesity, with generally higher levels in the North. 14 Estimates from survey data also suggest these geographical differences are widening in women but not in males (based on Health Survey data from 1993 and 2004). 14

The prevalence of obesity (BMI > 30 kg/m²) among adults is increasing. Estimates from England in 2007 reported obesity prevalence was 24% for both men and women. This shows a clear increase from the figures shown in 1993, which were 13% for men and 16% for women. 9 In Wales there has also been an upward trend in obesity over time among people aged over 16 years. 10 Data from Scotland also show that there has been a steady upward trend (data is for overweight and obesity) in adults since 1995 (55.6% of men and 47.2% of women aged 16–64 years in 1995 compared with 66.3% and 59.6%, respectively in 2008). 12

The prevalence of obesity is predicted to rise in the future. The World Health Organization (WHO) has projected that by 2015 more than 700 million adults worldwide will be obese. A projection of the UK prevalence of obesity (BMI > 30 kg/m²) extrapolating data from 1993 to 2004 onto 2012, showed that the prevalence would be 32.1% in men and 31.0% in women, suggesting nearly 13 million people will be obese by 2012. 15 The projected prevalence for adults was higher for those in manual social classes (48%) than non-manual social classes (35%). Another estimate of obesity prevalence, modelled in the UK’s Foresight project ‘Tackling obesities: future choices’, showed that if current trends persist, 36% of men and 28% of women aged 21–60 years will be obese in 2015. 16

Current service provision

Description of technology under assessment

Weight management programmes aim to improve the eating behaviour and the quality of a person’s diet, to reduce energy intake, and to also increase people’s physical activity levels or decrease inactivity. 35 A multicomponent intervention is defined as one which combines more than one mutually reinforcing strategy to achieve a common outcome. It is thought that addressing multiple influences on overweight and obesity will be an effective way to reduce and maintain healthy weight. The components may be distinguished from each other in terms of setting, location, provider, format, media and content. As previously stated, good practice guidelines promote weight management programmes that are multicomponent interventions that usually involve a combination of diet, exercise and behavioural therapy with ongoing support. In this sense the components differ primarily in terms of content (i.e. focus on diet or physical activity), but may also differ, to varying extents, in terms of setting (e.g. clinics, leisure centres), provider (e.g. dietitians, behavioural therapists) and other attributes. They work towards the common goal of encouraging weight loss and maintenance.

Dietary approaches incorporated in weight management programmes used in the UK may involve strategies such as calorie restriction, VLCDs, low fat, low carbohydrate, high fibre, meal replacement, food combining, or low glycaemic index foods. As described above (see Current service provision), dietary goals, in terms of the number of kilocalories per day permitted, may be set according to the type of diet followed. Nutritionists, dietitians or trained nurses may deliver dietary interventions. Broader lifestyle approaches may also accompany dietary goals, including education around food labelling, cookery skills and identifying where healthier foods can be purchased.

Physical activity elements of weight management programmes include exercise training and endurance exercises (e.g. running, swimming) and resistance training (e.g. use of weights). Physiotherapists, specially trained staff or fitness coaches may deliver physical activity sessions to individuals or groups with differing degrees of supervision in settings such as leisure centres or community centres. Physical activity may also be self-supervised, with participants given exercise goals to be reached through activities of their own choosing, for example daily living activities, walking or cycling.

Behavioural therapy may include a number of specific techniques including: self-monitoring (e.g. systematically observing and recording one’s own behaviour); problem analysis (e.g. dealing with situations that might interfere with reaching dietary or exercise goals); alteration of cognitive patterns such as cognitive reframing and coping imagery; cognitive strategies for replacing negative thinking with positive statements (e.g. modifying self-defeating thoughts about dieting, constructive self-statements); relapse prevention (e.g. identifying when lapses might occur in diet or physical activity and how to deal with high-risk situations); goal-setting; menu planning; stimulus control (e.g. avoiding stimuli that might encourage eating, such as reducing the visibility of food in the home, slowing the pace of eating, avoiding fast food restaurants); and may be delivered by psychologists, trained interventionists and counsellors. Ongoing support may take the form of personal contact, meetings, telephone calls and internet technology, and be individual, group or family based. Therapy may be theoretically based, for example drawing on the stages of change (SOC) approach of the Transtheoretical Model, or Social Cognitive Theory. 54–56 However, it is recognised that definitions of behavioural techniques are not always standardised, or explicitly linked to a theoretical model of behaviour change. 57

Multicomponent interventions may also include surgery and prescribed or over-the-counter (OTC) weight loss drugs, although this is more often as a second-line treatment if weight loss goals have not been attained through diet, physical activity and behavioural therapy strategies. There has been recent interest in the use of OTC drugs for obesity and their place within multicomponent approaches to weight management. In 2009 the European Union approved the use of 60 mg orlistat per day (Alli™, GlaxoSmithKline) for purchase following consultation with a pharmacist for people with a BMI over 28 kg/m². It is possible that people participating in multicomponent interventions may use OTC orlistat as an additional weight loss strategy, though it has been suggested that some may use it in place of diet and exercise. Research is needed to establish how it might be best used in practice. 58

Weight management programmes may be delivered in the primary care setting, by GPs supported by specialists as mentioned above, and tutors trained specifically to deliver such programmes, or through professionals in specialist hospital clinics. For example, the Counterweight Programme54 is a primary care-based programme in the UK that aims to help obese people to achieve weight loss (e.g. 5%–10% over 3–6 months) and weight maintenance through changes in diet and exercise. In addition to NHS-based programmes, commercial and self-help programmes are available such as Weight Watchers^®, Slimming World, Rosemary Conley™, LighterLife™ and others. LIs in overweight working populations are also available. 59

Weight management programmes vary in terms of their duration. Generally there will be an initial period where the aim is to achieve a desired weight loss goal (e.g. 6–12 months). Thereafter there may be a weight maintenance phase where the aim is to sustain weight lost in the first stage. There is some evidence to suggest that weight is lost rapidly at first, and the point of greatest loss occurs 6 months after beginning treatment; weight is then slowly regained and often reaches near the original level. 60 For example, approximately a third of lost weight may be regained in the first year after treatment and often continues with the average loss of about 1.8 kg remaining at 4 years after treatment. 60 Therefore, it is important that any weight management scheme facilitates weight maintenance after the target weight has been reached. Resulting changes to lifestyle must become part of everyday life for long-term health benefits to be realised. 35 It is important to acknowledge, however, that weight gain in the general population may occur throughout life and therefore people who have participated in a weight loss intervention may still be at lower weight than if they have not participated.

Well-managed, multicomponent weight management schemes are not thought to lead to any specific adverse events for participants;35 however, there are some suggestion that negative outcomes for some participants may occur from weight management programmes. This appears to be largely based on ‘dieting’ in general, where evidence suggests that for some people dieting can lead to eating disorders, negative psychological and emotional effects and increased social stigma. 61

Overall aims of this assessment

The aim of this review was to assess the long-term clinical effectiveness and cost-effectiveness of multicomponent weight management schemes for adults in terms of weight loss and maintenance of weight loss.

Search strategy

The search strategy was developed, tested and refined by an experienced information scientist. Separate searches were conducted to identify studies of clinical effectiveness, cost-effectiveness and epidemiology. Sources of information and search terms are provided in Appendix 2 and a flow chart of identification of studies can be seen in Figure 1.

FIGURE 1.

Flow chart of identification of studies for inclusion in the review. (a) Details of the interventions in six RCTs (seven publications) nearly met our threshold of containing enough detail (see Methods). (b) Five studies (six publications) were assessed as ongoing and are described as such in Assessment of effectiveness of multicomponent interventions versus non-active intervention comparators.

Searches for clinical effectiveness and cost-effectiveness literature were undertaken from database inception to December 2009. Electronic databases searched included: MEDLINE; EMBASE; MEDLINE In-Process & Other Non-Indexed Citations; The Cochrane Library including the Cochrane Database of Systematic Reviews, Cochrane Central Register of Controlled Trials, Database of Abstracts of Reviews of Effects, NHS Economic Evaluation Database and HTA databases; Web of Science; PsycINFO; BIOSIS; and databases listing ongoing clinical trials.

Searches were restricted to the English language. Bibliographies of related papers were screened for relevant studies, the expert advisory group were also contacted for advice and peer review and to identify additional published and unpublished references.

Inclusion process

Titles and abstracts identified by the search strategy for the clinical effectiveness section of the review were assessed for possible eligibility by two reviewers independently. The full texts of relevant papers were then obtained and inclusion criteria were applied by one reviewer and checked by a second reviewer using a previously piloted inclusion flow chart (see Appendix 3). Any disagreements over eligibility were resolved by discussion or by recourse to a third reviewer.

Titles and abstracts identified by the search strategy for the systematic review of cost-effectiveness were assessed for potential eligibility by two reviewers independently. Economic evaluations were considered for inclusion if they reported both health service costs and effectiveness of multicomponent adult weight management programmes, or presented a systematic review of such evaluations. Two reviewers formally assessed full papers independently, with respect to their potential relevance to the research question. Any differences in judgement were resolved through discussion.

Data extraction and quality assessment strategy

Data were extracted by one reviewer using a standardised and pre-piloted data extraction form and checked by a second reviewer. Any disagreements were resolved by discussion, if necessary involving a third reviewer.

Within the clinical effectiveness review the quality of included studies was assessed using criteria based on those recommended by the Centre for Reviews and Dissemination (CRD)63 (see Appendix 4). Each RCT was assessed by one reviewer and checked by a second reviewer. Any disagreements were resolved by discussion, if necessary involving a third reviewer. The reviewers assessed the adequacy of reporting: randomisation; allocation concealment; population baseline characteristics; blinding of assessors, care providers and participants; imbalances in attrition; outcomes; intention-to-treat (ITT) analyses; and analyses of missing data. For ease of presentation and interpretation of the quality assessment of RCTs the responses were then translated into judgements of ‘high’, ‘low’ or ‘uncertain’ risk of bias based on the Cochrane risk of bias criteria. 64 For example, where a positive response to a question indicated an adequate procedure to minimise bias (e.g. ‘was the allocation adequately concealed?’) this was translated into a ‘low’ risk of bias. Similarly, where a positive response to a question indicated an inadequate procedure to minimise bias (e.g. ‘is there any evidence to suggest that the authors measured more outcomes than they reported?’) this was translated into a ‘high’ risk of bias.

For the systematic review of cost-effectiveness, the methodological quality of the cost-effectiveness studies was assessed using a critical appraisal checklist based on that by Philips and colleagues,65 Drummond and Jefferson66 and the NICE reference case requirements. 67 These checklists were combined to include the key elements that were relevant to internal and external validity, and the generalisability of the studies to the UK NHS.

Data synthesis

Data were synthesised through a narrative review with tabulation of results of all included studies. Full data extraction forms for the clinical effectiveness review are presented in Appendix 5. Within the clinical effectiveness section studies using similar interventions were grouped together to aid interpretation. Studies were categorised according to which of the intervention components (diet, exercise, and/or behaviour modification) were of primary interest and how they differed from the comparator intervention. This led to four groups: studies in which the comparator was a non-intervention group; studies in which the intervention focused on the diet aspect versus another active comparison; studies in which the intervention focused on the exercise aspect versus another active comparison; and studies where other variables were the focus of the study.

Where trials reported several interventions, decisions were made by consensus among the review group as to which was the ‘key’ intervention of relevance to the scope of this review (as noted above, these are multicomponent interventions involving a combination of diet, exercise and behavioural therapy) and which comparators were relevant. For some trials there was little discernible difference between the interventions tested and as such the focus of our report is limited to those which appeared to be the most relevant to our research protocol, with complete details of all interventions reported (see Appendix 5).

Our pre-defined inclusion criteria excluded studies with < 18 months’ follow-up from randomisation, in order to focus the review on long-term weight loss and maintenance of weight loss given that weight regain in the long term is an intractable problem. Therefore, where included studies report interim data before 18 months’ follow-up these have not been included. However, we do provide comment in the text where these data are available in the primary studies.

It was considered inappropriate to combine the results of the studies in a meta-analysis due to differences in the interventions, comparators and populations in the included studies.

Quantity and quality of research available

Searching identified 3358 references after deduplication. After initial screening of titles and abstracts, 159 references were retrieved for further inspection. The total number of published papers included at each stage of the systematic review is shown in the flow chart in Figure 1. References to the studies retrieved for further inspection but subsequently excluded can be seen in Appendix 6. The most common reason for exclusion was the inadequate length of follow-up (54 studies); lack of, or only minimal, detail reported on one or more of the components of the intervention (see Chapter 2) led to 20 studies being excluded; another 21 were not deemed to be multicomponent interventions, and 19 were not RCTs. In addition, one study did not have an appropriate comparator; two studies did not report weight loss outcomes and the participants in seven studies did not meet the inclusion criteria. The level of agreement between reviewers assessing study eligibility was generally good although this was not formally measured.

Six RCTs (seven publications) appeared to meet all of the inclusion criteria except that the details of the multicomponent interventions were judged to be below our threshold (see Chapter 2). Authors of these six studies were contacted for further detail. A response from one author was received; however, the information provided did not clarify any further whether the interventions did indeed meet our criteria. These studies were undertaken between 2 and 15 years ago, which may explain the poor response from the authors. As such there were six studies that we were unable to identify whether they met the inclusion criteria of the review. These studies can be seen in Appendix 7.

Twelve RCTs met all of the inclusion criteria,68–79 of which five had two intervention arms, one had three arms, four had four arms and two had five arms (Table 7). In many of these RCTs data were reported in multiple publications; for ease of presentation the primary reference is used here, with full details of all secondary publications given in Appendix 5. Not all of the intervention arms were considered relevant to the scope of the current review; this is outlined in the subsequent sections where this is the case. The key methodological attributes of these trials are summarised in Table 8, with full details of the interventions, outcomes and quality assessments provided in Appendix 5. All included trials were individual-level interventions; no community-level interventions met the inclusion criteria.

All RCTs that met the inclusion criteria were conducted in the USA. Most (83%) of these were published between 1993 and 2008, with two older RCTs, published in 198468 and 1988. 69 The total number of participants randomised ranged from 5969 to 1191,70 while the number of participants per intervention group ranged from 1869 to 596. 70 Only four of the 12 RCTs had sample sizes > 100 participants per intervention group. 70,72–74 The earliest of the RCTs68 did not report the number of participants randomised. Power analyses to calculate the necessary sample sizes were reported in only five trials. 70–74

In all but one73 of the 12 RCTs the target population was stated as being overweight. The RCT by Simkin-Silverman and colleagues73 enrolled middle-aged women and aimed to prevent the rise in weight and low-density lipoprotein (LDL) cholesterol observed during the menopause. The baseline weight, however, indicated that participants were overweight (mean BMIs > 25 kg/m²). The two trials reported by Stevens and colleagues70,74 were conducted as part of a larger study of risk factors for hypertension [Trials of Hypertension Prevention (TOHP)] and included overweight participants.

Nine of the 12 RCTs reported the starting BMI of their participants. 70–78 According to the international classification of overweight or obesity,6 participants would be considered overweight (pre-obese) in two RCTs,73,74 class I obese in five RCTs,70,71,75–77 and class II obese in one RCT. 78 The remaining RCT72 contained a mixture of pre-obese (18%–22%), class I obese (32%–36%), class II obese (21%–24%) and class III obese (22%–24%) participants (see Appendix 5). Seven of the 12 RCTs specified that their participants, aside from being overweight, were in good health. 68,70,73–77 One RCT, by Weinstock and colleagues,78 excluded women with bulimia but not those with binge eating disorder. The remaining four RCTs69,71,72,79 did not include health status as an inclusion criterion.

Three studies did not mention age among their inclusion criteria (however participants were all adults). 68,69,78 The upper age limit specified for inclusion of participants was 45 years in two RCTs,75,79 50 years in two RCTs,73,77 54 years in two RCTs,70,74 55 years in two RCTs,71,76 and 69 years in the remaining RCT. 72 None of the trials specifically included elderly populations.

Descriptions of the baseline characteristics of the populations included in the RCTs varied considerably in their detail (see Table 7). In eight (67%) of the RCTs the mean age of participants within each intervention group was in the range 35.7–47.0 years. 69–71,73–76,78 Two RCTs did not report the age by intervention group but stated that the overall mean age was 42.2 years77 or that most participants (≥ 80%) were aged 40–59 years. 72 The remaining two RCTs did not report the age of their participants. 68,79 In two of the RCTs the participants were all female,73,78 in five RCTs they were mostly (67%–91%) female,68,69,71,72,76 in two RCTs they were mostly (63%–73%) male,70,74 and in two RCTs the gender balance was nearly equal (42%–53% male). 77,79 One RCT75 did not report the gender balance of its participants.

Weight change from baseline was reported as a primary outcome in 11 RCTs68–77,79 and as a secondary outcome in one RCT. 78 Two RCTs also reported changes in BMI, either as a primary outcome75 or as a secondary outcome. 71 The proportion of participants maintaining weight loss was also a primary outcome in one RCT69 (see Table 7).

In one study the setting of the intervention was reported to be in primary care. 72 Six studies70,71,73,74,76,78 reported that their study was clinic based; in some cases these were clinical practice-based research units within universities. Five studies68,69,75,77,79 did not report the setting of their intervention.

The duration of follow-up (post randomisation) ranged from 18 to 54 months. Eight68,69,71,72,74,76,78,79 of the 12 RCTs provided weight outcome follow-up data for a single time point after randomisation, three RCTs70,75,77 provided follow-up data for two time points, and one RCT70 provided follow-up data for four time points. Six68–70,73,75,77 of the RCTs included follow-up beyond 2 years (see Table 7). The duration of interventions is discussed more fully in the following three sections.

The methodological detail reported varied considerably among the RCTs. As an indication of methodological rigour, the risk of bias was assessed for each of the 10 methodological criteria (see Table 8). Quality assessment of each individual trial can be seen in Appendix 5.

Trials that do not properly randomise their participants or conceal the allocation of interventions could be at risk of selection bias. Trials were assessed according to whether they used an adequate method to generate random allocations and to conceal their intervention allocations. Four trials provided information on their randomisation sequence70,72,73,79 and three described their allocation concealment,70,73,74 and all of these were judged to have a low risk of selection bias (see Table 8). For the remaining trials there is an uncertain risk of bias as there was either inadequate or no information reported from which this could be assessed.

Blinding of participants, care providers and outcome assessors can help to reduce the risk that participants and trial personnel become aware of intervention assignment, so potentially influencing their performance in the trial (detection bias). None of the trials clearly reported blinding of their participants or care providers, so it is unknown whether blinding occurred. It seems unlikely that care providers and participants would have been blinded as it would have been difficult to conceal the identity of the weight loss interventions without substantial spatial and/or temporal separation of the care providers and participant groups that were allocated to different interventions. Blinding of outcome assessors, however, is more feasible but was only reported in two trials. 70,73 In both of these the method of blinding was judged as adequate, with a low risk of bias (see Table 8).

Sources of measurement bias in clinical trials include losses of participants to follow-up, unequal dropout rates between interventions, selective reporting of outcomes (missing outcomes) and failure to explain why participants are missing (e.g. whether they are missing at random). Six RCTs were judged to have low risk of bias from dropout (no dropout imbalance);69,73,78 five, however, provided inadequate or no information to assess risk of bias from dropout,68,74–77 and one had unequal dropout between interventions, indicating a high risk of bias. 79 All 12 trials provided sufficient information to permit an evaluation of reporting bias,68–79 of which three had a high risk of bias (some outcomes not reported)71,72,75 and nine had a low risk of bias (all outcomes were adequately reported). 68–70,73,74,76–78,79 Four trials were judged to have a high risk of bias because they did not report an ITT analysis,70,71,76,78 and seven trials were judged to have a high risk of bias because their analysis and interpretation did not account for missing data68,69,71,73,76,78,79 (see Table 8).

Given the lack of methodological information reported it was not possible to rank the RCTs according to their methodological rigour or risk of bias. The extent to which methodological criteria were reported did not appear to show any patterns in relation to the publication date, suggesting there has not been an obvious improvement or deterioration in rigour through time. Where appropriate, the implications of risk of bias are considered in more detail for specific trials in the following three sections of this report.

The 12 RCTs varied considerably in the structure and content of their interventions and comparators and were consequently grouped into four categories (see Chapter 2). The four categories are: trials in which the comparator was a non-active intervention group, for example described as usual care (five trials);70,72–75 trials where the focus of the study was predominantly on diet (two trials);69,71 trials where the focus of the study was predominantly on exercise (four trials);76–79 trials where the focus was on other factors (one trial). 68 Studies are reported in the following sections grouped in these four categories. In these subsequent sections of the report we describe the key aspects of the interventions and results of the studies. As described above there were differences between studies in terms of quality, follow-up, participant characteristics and sample sizes that are referred to where relevant in the subsequent sections. Studies were statistically powered to detect a difference between groups unless stated.

Assessment of effectiveness of multicomponent interventions versus non-active intervention comparators

Five trials70,72–75 compared multicomponent interventions against a non-active intervention comparator group, and can be seen in Table 9. The nature of the comparator differed slightly between these studies. In two of the RCTs, reported by Stevens and colleagues,70,74 the comparator was described only as a usual-care control group, with no further details provided. In one RCT, reported by Jeffery and Wing,75 the comparator was a control group whose participants received no instruction or guidance and were free to act independently to achieve weight loss. The remaining two RCTs, reported by Logue and colleagues72 and Simkin-Silverman and colleagues,73 provided participants in their comparator group with limited general guidance on improving diet72 or reducing cardiovascular risk factors73 (see Table 9). The trial by Jeffery and Wing75 differed in that it had five arms (four active interventions and one non-active comparator), with only 40 or 41 participants randomised per arm. The four active intervention arms reported by Jeffery and Wing75 were standard behavioural therapy/treatment (SBT) alone, or SBT in combination with food provision (FP) and/or incentives (see Table 7). For the purpose of this review we consider SBT as the multicomponent intervention of most relevance. Details of the other interventions can be found in Appendix 5.

One trial did not explicitly report an aim. 72 The trial reported by Simkin-Silverman and colleagues73 was restricted to premenopausal women and aimed to prevent the natural gain in weight and adverse changes in lipid profiles that occur during the menopause. The remaining three trials stated that they aimed to achieve specific weight loss goals, which were to lose 4.5 kg during the first 6 months,70,74 or to achieve participants’ self-selected weight loss goals of 14, 18, or 23 kg together with an exercise energy expenditure goal of 1000 kcal/week by the final week. 75 Three of the RCTs also included weight maintenance in their aim (see below). 70,73,74

Only Logue and colleagues72 explicitly stated that their intervention was based on a theoretical model (the Transtheoretical Model). Their intervention was from the Reasonable Eating and Activity to Change Health (REACH) trial (not included in this systematic review) and the diet component was based on the US Department of Agriculture (USDA) Food Guide Pyramid Dietary Guidelines for Americans. The intervention for perimenopausal women reported by Simkin-Silverman and colleagues73 was described as an LI (the Women’s Healthy Lifestyle Project). Jeffery and Wing75 reported interventions that they described as standard behaviour treatment. The two remaining interventions by Stevens and colleagues70,74 were referred to as weight loss interventions (see Table 9) which were from the TOHP study.

The RCTs ranged in duration from 18 to 54 months. In the three trials that specifically referred to weight maintenance, the weight maintenance phase was assisted with ongoing support (details are given below). The RCTs reported by Stevens and colleagues70,74 consisted of a 6-month weight loss phase followed by a weight maintenance phase of 12 months74 or 36 months (up to 48 months for some data). 70 The trial reported by Simkin-Silverman and colleagues73 consisted of a 20-week initial phase of weight stabilisation followed by a maintenance phase up to 54 months.

Four of the RCTs specified quantitative calorie goals in the diet component of their interventions. 70,73–75 These were daily limits of: < 1500 kcal for men or < 1200 kcal for women;70 1000 or 1500 kcal according to body weight;75 ≥ 1200 kcal (no upper limit specified);74 or 1300–1500 kcal, gradually increased until weight stabilised. 73 Three RCTs specified dietary composition,70,72,73 but only one, by Simkin-Silverman and colleagues,73 mentioned quantities (dietary fat reduced to 25%, saturated fat to 7% and cholesterol to 100 mg/day). The remaining two trials specified an increase in fruit and vegetables, and reduction in fat,72 or decreased consumption of excess fat, sugar and alcohol70 (see Table 9).

The same four RCTs that specified quantitative calorie goals also provided quantitative exercise goals. 70,73–75 In two of these, exercise intensity was expressed as approximately 40%–55% of the heart rate reserve, achieved principally by increasing the frequency and duration of walking. 70,74 In one trial, exercise energy expenditure was specified as 1000 or 1500 kcal week, depending on participants’ initial activity levels, with maintenance of 1500 kcal/week encouraged, but the mode of exercise was not specified. 73 The remaining trial specified an exercise energy expenditure goal of 50 kcal/day for 5 days/week, increasing to 1000 kcal/week, achieved by walking or cycling. 75 Two of the RCTs72,73 did not specify the mode of exercise (such as walking or cycling) (see Table 9).

Four RCTs reported that the mode of behavioural therapy involved group delivery of the intervention. 70,73–75 Group sizes were 7–20,74 approximately 20,73,75 or 11–34 participants. 70 The frequency of group meetings was initially weekly for the first 10 weeks,73 14 weeks,70,74 or 20 weeks. 75 Subsequent meetings were then fortnightly for 10 weeks,73 fortnightly for 12 weeks, then monthly for 3–4 months,70 or monthly. 75 The behavioural therapy intervention reported by Logue and colleagues72 was not described in detail but appeared to involve an individual mode of delivery, as telephone counselling was provided every month (see Table 9).

The personnel who delivered the behavioural therapy interventions were described only briefly. Behaviour therapy in the trial by Logue and colleagues72 was delivered by a registered dietitian and a weight loss advisor, but it is unclear how many people delivered these two roles. None of the other RCTs specified the number of people who delivered the intervention. In four RCTs the behavioural therapy interventions were delivered partly by dietitians,70,72,74 or nutritionists. 73 These RCTs also included delivery by a weight loss advisor,72 a health educator and psychologist,70 behavioural psychologists,73 or a psychologist or exercise physiologist. 72 The remaining trial used trained interventionists with advanced degrees in nutrition or behavioural science to deliver the intervention75 (see Table 9). None of the studies reported where the behaviour therapy was undertaken.

At least 10 different components of behaviour therapy were identifiable among descriptions provided in the RCTs and are summarised in Table 10. As mentioned above, Logue and colleagues72 did not specify which components they included, but instead reported the behavioural model they used. The model employed is widely documented and accessible, and five individual target behaviours for change were specified for the intervention. Among the remaining four RCTs, the behavioural components represented most commonly were goal-setting or action plans (four studies),70,73–75 relapse prevention (four studies), 70,73–75 problem solving (three studies),73–75 self-monitoring (three studies),70,73,74 and social support (three studies). 70,74,75 Two RCTs included cognitive restructuring,73,75 motivation enhancing, and stimulus control, whereas only one trial included behaviour modelling and none explicitly included self-reinforcement or skill development (see Table 10).

Ongoing support for participants in the intervention group was reported in four RCTs and varied in its type and frequency. 70,72–74 Logue and colleagues72 provided written materials, which were mailed to participants on request. Stevens and colleagues70,74 provided a variety of options, including refreshing or redelivering the intervention content and encouraging a number of participant contacts, the type and number of which varied according to the needs of individual participants. Simkin-Silverman and colleagues73 provided five meetings for participants during months 7–14, after which they were offered refresher programmes to assist maintenance of behaviour change (see Table 9).

The population characteristics of these RCTs limit the generalisability of their findings. The RCT by Simkin-Silverman and colleagues73 is relevant only to perimenopausal women and based on a subgroup of the total population who were mildly overweight (total population mean BMI 20–34 kg/m²; aged 44–50 years). The remaining RCTs were mostly limited to people with an upper BMI of approximately 34–37 kg/m², with an upper age limit of 45–54 years. This precludes generalising the findings to very obese and/or older people. In contrast, the trial by Logue and colleagues72 represented overweight participants whose BMI was ≥ 40 kg/m² and included people aged up to 69 years. The trials, which were all conducted in the USA, recruited predominantly white volunteers who responded to media advertisements and, who apart from being overweight, were free from serious disease. Such people are unlikely to be representative of the UK overweight population. Where reported, participants were relatively well educated and employed. Three of the trials used incentives to enhance participation, which could also influence their generalisability. 72,73,79

The starting weight and BMI of the trial participants were reported for four of the five trials (excluding Logue and colleagues72) (see Table 7). The participants would be classed as pre-obese in two RCTs73,74 and class I obese in two RCTs. 70,75

The appraisal of methodological quality (see Quantity and quality of research available) does not enable any of these five RCTs to be considered ‘best’ or ‘worst’ in terms of risk of bias. This mainly reflects the lack of, or unclear, reporting of the methodological variables. However, for three70,72,73 of these studies the overall risk of bias is generally low, with many key attributes required to reduce potential bias present, while for the remaining two studies74,75 the risk of bias remains uncertain.

Assessment of effectiveness for multicomponent interventions with a focus on diet

Two69,71 included studies compared a multicomponent intervention which had a focus on the dietary component compared with another active comparator. The two studies are briefly described below and the main components are summarised in Table 12.

Burke and colleagues71 hypothesised that choice of either a standard calorie and fat-restricted diet (standard diet, STD-D) or a calorie- and fat- restricted vegetarian diet (lacto-ovo-vegetarian diet, LOV-D) would result in greater weight loss than having one of these diets randomly assigned. The authors’ secondary hypothesis was that LOV-D would result in greater weight loss than STD-D. There were four arms to the trial but for the purpose of this review we consider just two of these arms to be the most relevant, those where participants were randomised to either the intervention with STD-D or LOV-D and therefore had no choice regarding the type of diet they received. Wadden and colleagues69 compared three active interventions. For our review we consider the SBT the intervention of relevance to the scope. The other two groups were SBT and VLCD, and VLCD alone.

In the Burke and colleagues71 study the interventions only differed in the type of diet participants were asked to adhere to. For the duration of the 12-month interventions participants in both groups who weighed < 90.5 kg at baseline aimed to reduce their maximum daily calorie intake to 1200 kcal (women) or 1500 kcal (men). Those who weighed over 90.5 kg at baseline had goals of 1500 kcal (women) and 1800 kcal (men). Participants were advised to maintain a minimum daily intake of at least 1000 kcal. All participants also aimed to reduce their fat intake to 25% of total kilocalorie intake, but were advised not to consume < 10% of total kilocalorie intake as fat. The only difference between the two groups was that the LOV-D group eliminated meat, poultry and fish first from breakfast, then lunch, and then dinner so that by the sixth week they had eliminated these foods from their diet. In the Wadden and colleagues study69 participants in the SBT group consumed a 1000–1200 kcal/day ‘balanced’ diet of their choosing for the 6-month intervention period. The mean [standard deviation, (SD)] baseline weight of participants in this study was 112.2 kg (21.5kg) in the SBT group, 108.0 kg (21.5kg) in the SBT and VCLD group, and 106.4 kg (18.4kg) in the VLCD alone group. So for participants over 90.5 kg the weekly calorie goal was lower than in the Burke and colleagues study. 71 Participants in the SBT and VLCD, and VLCD only groups consumed a 1000–1200 kcal/day ‘balanced’ diet of their choosing for the first month of intervention, then they began a 2-month VLCD protein-sparing modified fast of only 400–500 kcal/day. This diet consisted of lean meat, fish and fowl plus potassium, sodium chloride and calcium supplements. In the fourth month fruits and vegetables, breads and cereals, and finally fats were reintroduced to the diet, in that order, to return to a more conventional diet. The VLCD intervention ended at this point (4 months), whereas the SBT and VLCD group continued at 1000–1200 kcal/day diet for a further 2 months (months 5 and 6).

All participants engaged in physical activity for the duration of the Burke and colleagues71 study. Participants were encouraged to walk at least 50 minutes per week initially, gradually increasing this to at least 150 minutes per week, by week 6. In the Wadden and colleagues study69 only two of the three groups had an exercise component. The SBT and SBT and VLCD groups each received the same instruction on increasing lifestyle activity by walking and using the stairs. However, unlike the study by Burke and colleagues,71 it does not appear that there was a specific activity goal.

For the first 6 months of the Burke and colleagues71 study participants attended weekly 60-minute sessions in groups of about 10–20 participants. Sessions were held in the evenings biweekly for months 7–9 and then monthly for the final 3 months of the intervention period. Dietitians, exercise physiologists and nurse behavioural scientists were involved in delivering the sessions, with the LOV-D group also receiving advice from a vegetarian nutritionist for four sessions on how to adopt the eating plan. The behavioural aspect of the intervention was based on several models of motivation and behavioural change such as Social Cognitive Theory. The range of features and strategies included in the behavioural component of the interventions are summarised in Table 13. It is not reported in the study where the behaviour therapy was undertaken. At the end of the 12-month intervention participants had no further contact with the interventionists until the final 18-month assessment. In the Wadden and colleagues study69 the intervention periods of 4 months for VLCD alone, and 6 months for the SBT and SBT and VLCD groups were all shorter than for the Burke and colleagues71 study. All participants were treated weekly for 90 minutes during the intervention period in groups of four to seven participants, a smaller group number than in the Burke and colleagues71 study. A doctoral-level clinical psychologist led the groups. Only two groups, SBT and SBT and VLCD, were taught traditional behavioural methods of weight control following a published manual, although it was not clear whether this was based on a particular behavioural theory. The range of features and strategies included in the behavioural component of the intervention can be seen in Table 13. The setting for the behaviour therapy was not reported. The SBT and VLCD group received additional instruction in the final 3 months of the intervention on maintenance of the weight loss achieved with the VLCD, including relapse prevention training and strategies for handling weight regain. The VLCD alone group did not receive any formal instruction about modifying their eating and exercise habits, and did not receive any behaviour therapy. Following the intervention period participants in the SBT group and SBT and VLCD group attended 11 follow-up meetings in the year after treatment. The timing of these meetings was not reported. Participants in the VLCD only group attended six follow-up meetings at 1, 2, 3, 6, 9 and 12 months post treatment. There was then no further contact with participants in any of the groups until the 3-year follow-up.

The Burke and colleagues71 study enrolled a much higher proportion of women than men, and the majority of participants (about 70%) were white with average socioeconomic status. The majority of the participants recruited by Wadden and colleagues69 were women, but the ethnicity of participants was not reported. The generalisability of the results to men and people of different ethnicities is therefore unknown. The mean ages of the participants in both trials were similar, at approximately 43 years. Wadden and colleagues69 charged a fee to cover medical tests and all other expenses, and in addition participants had to pay a deposit which was refundable after the first year of follow-up. This may have affected the type of people willing to be recruited into the trial.

The starting weight of the trial participants was reported by both trials69,71 and Burke and colleagues71 also reported starting BMI. In the Burke and colleagues study71 starting weight was slightly over 90 kg in both groups (Table 7) and participants would be categorised as having class I obesity according to their starting BMIs. Participants in the Wadden and colleagues study69 were heavier than those in the Burke and colleagues study with baseline weights over 100 kg in each group. Although BMI was not reported, if height is assumed to be 170 cm, then these participants would be categorised as having class II obesity.

Details of the methods of the Burke and colleagues71 trial either were not reported or were unclear (see Quantity and quality of research available). In the Wadden and colleagues69 study the majority of items contributing to the assessment of study quality were not reported. This may be a reflection of the age of the study, which was undertaken over 20 years ago. Therefore, when interpreting the results of these studies the likelihood that they may be associated with a high risk of bias should be borne in mind (see Table 8).

Assessment of effectiveness for multicomponent interventions with a focus on exercise

Four76–79 of the included studies compared a multicomponent intervention to an active comparator in which the focus of the study in general was on exercise. The four studies are briefly described below and the main components are summarised in Tables 15 and 16.

Of the four studies, those by Tate and colleagues77 and Jeffery and colleagues76 were the most similar in terms of intervention characteristics and design, having been conducted by some of the same investigators (these are therefore discussed first in this subsection, with the characteristics of the remaining two studies presented afterwards). Both studies evaluated an 18-month-long ‘SBT intervention’ comprising similar behavioural change techniques, alongside exercise and diet components. The most intensive phase in both studies appeared to be the first 6 months with weekly meetings to encourage weight loss, followed by biweekly or monthly meetings to promote weight maintenance. Final outcomes were measured at the end of the 18-month period in the study by Jeffery and colleagues study,76 but in the Tate and colleagues77 study they were measured 1 year following the end of the intervention (i.e. at 30 months).

The Tate and colleagues study77 compared SBT with a ‘high physical activity (HPA)’ arm, identical to the SBT group but with higher weekly energy expenditure goals and a greater level of supervision and social support (see Appendix 5). The aim of the study was to determine whether higher levels of energy expenditure (2500 kcal/week) would result in greater weight loss compared with SBT (1000 kcal/week). In contrast, Jeffery and colleagues76 compared SBT with four other arms including SBT plus supervised exercise (SBTE), SBT and supervised exercise plus a personal trainer (SBTT), SBT and supervised exercise plus incentives to exercise (SBTI), and SBT and supervised exercise plus both personal trainer and incentives to exercise (SBTTI). The hypothesis was that a greater number of exercise sessions would be attended by those receiving support from a personal trainer or financial incentives than SBT alone, and that a combination of trainer and incentives would result in greater exercise adherence and therefore greater energy expenditure in exercise and greater weight loss.

For both studies we considered the SBT arm as being the multicomponent intervention of most relevance to the scope of this review. In the Jeffery and colleagues76 study we have compared SBT with SBTE only, hence the characteristics and results of the other three arms are not reported here (for further details of these see Appendix 5).

The diet component, which was the same in both arms in each study, was similar between the two studies, with an emphasis on calorie restriction to between 1000 and 1500 kcal per day, with ≤ 20% of calories from fat. Both studies encouraged a gradual increase in exercise (e.g. walking, cycling) to build up energy expenditure to at least 1000 kcal per week (or, as mentioned, to 2500 kcal in the HPA arm of the Tate and colleagues study77). Exercise activities were conducted either individually or as part of group sessions.

Likewise, the behavioural component, which was the same in both arms in each study, was similar between the two studies, with therapy delivered in small groups (10–20 participants) on a weekly basis for the first 6 months, then biweekly or monthly thereafter until the end of the 18-month intervention. The sessions were led by interventionists trained in exercise, nutrition or the behavioural sciences and incorporated motivation enhancement, problem solving, stimulus control (e.g. reducing the visibility of food in the home), and relapse prevention techniques (e.g. recognising precursors and consequences of dietary lapses) (see Table 16).

The third study in this subsection was a trial of a cognitive behavioural weight control intervention by Skender and colleagues79 The trial compared a multicomponent ‘combination’ intervention comprising diet, exercise and behaviour therapy with an intervention focused on exercise, and with another intervention focused on diet (a waiting list control group was also included, but no outcome data were reported for this group so it is not mentioned any further in this subsection). All three interventions included the same behavioural component. The aim was to assess whether the addition of exercise to diet would lead to greater weight loss. For the purpose of this review the ‘combination’ intervention was considered to be most relevant to the scope.

The interventions took place over 1 year with weekly 1-hour meetings over a 12-week period, then three biweekly sessions followed by the maintenance phase comprising eight monthly meetings. Final outcomes were measured at 2 years from baseline (i.e. a year after the end of the intervention). The dietary component featured in the combination intervention and the diet-focused intervention encouraged a caloric intake to attain weight loss of up to 1 kg per week, with 30% of calories to come from fat, 50% from carbohydrate and 20% from protein. Participants were instructed to plan their meals based on the Help Your Heart Eating Plan (HYHEP) low-cholesterol diet. In the exercise focus group participants were asked to maintain their current eating habits and nutrition was not discussed.

The exercise component featured in the combination intervention and the exercise-focused intervention. Information was provided on the benefits of exercise and instruction was given on correct methods of walking. Participants were instructed to self-regulate the intensity of brisk walking based on heart rate, breathing difficulty, and perceived effort. They were instructed to exercise at a level that felt ‘vigorous’ but never ‘strenuous’, and were encouraged to exercise for three to five sessions per week for 45 minutes or more per session. Those in the diet-focused intervention were asked to maintain their relatively sedentary lifestyles and not to begin any new exercise programme.

The behavioural therapy component (common to all three interventions) involved the use of self-monitoring contracts to reward behaviour change (contingency contracting), stress management, stimulus control, goal-setting and maintenance techniques (see Table 16). The behavioural component followed the principles outlined in the LEARN^® programme for weight control. This is a 12-week educational lifestyle weight management programme (‘Lifestyle, Exercise, Attitudes, Relationships and Nutrition’). It aims to enable people to lose between 1 and 2 lb per week through a decrease in dietary caloric intake and an increase in exercise to burn stored calories. The interventions were delivered by registered dietitians who were trained and experienced in behaviour modification.

The fourth study, a 48-week trial by Weinstock and colleagues,78 randomised participants to either a diet only intervention or one of three exercise interventions, diet with aerobic training, diet with strength training and diet with aerobic and strength training. In the exercise interventions the diet and behaviour components were the same across the arms. For the purpose of this review we considered that the diet and aerobic training intervention was the multicomponent intervention of most relevance to the scope. In this intervention the exercise type was ‘step’ aerobics. Details of the other exercise interventions can be seen in Appendix 5. An increasing daily kilocalorie goal was used through the 48-week intervention, increasing to 1500 kcal by week 22 which remained then until week 48. In the first 17 weeks a 900–925 kcal/day goal was used and this was as a liquid meal replacement. For the rest of the duration of the study participants were advised on the proportions of their diet to be made up from protein, carbohydrate and fat (see Table 15). Participants maintained weekly diet diaries.

Participants met weekly in groups of 7–10 people until week 28 for their behavioural therapy and biweekly thereafter. The components of the behavioural aspect of the intervention were poorly described (see Table 15). The features and strategies covered appeared to be centred on skill development. The intervention was delivered by clinical psychologists, a dietitian and graduate students in exercise physiology. For the first 28 weeks participants exercised with members of their behavioural treatment groups by step aerobics. They were required to participate in three sessions per week for 28 weeks and then two sessions per week during weeks 29–48. The time spent undertaking the step aerobics increased from 12 minutes at week 1 to 40 minutes by week 14. In the diet alone arm no prescribed exercise was undertaken, although the participants were allowed to maintain any lifestyle activities as normal. After the 48-week intervention participants were followed up at week 96. In the interim, group sessions occurred every 3 months as a form of ongoing support.

The multicomponent interventions included in the four studies appear to be similar in terms of content and structure. All studies encouraged participants to choose low-calorie diets and to reduce the proportion of calories from fat. They were also supported to take regular exercise, primarily through walking and cycling, according to set weekly goals. Behavioural techniques such as stimulus control, self-monitoring and goal-setting were also common aspects of the interventions, typically led by trained interventionists in small groups of < 20 participants. None of the studies reported where the behaviour therapy was undertaken. The level of detail in which the interventions were reported precludes a full assessment of generalisability. However, all studies were conducted in the USA with mostly white, educated populations who were paid monetary incentives to participate. The mean ages of participants were between 40 and 42 years for three studies,76–78 but in one study were not reported. 79 With the exception of Tate and colleagues,77 the interventions were delivered in the early to mid-1990s. It is not clear whether the LEARN^® programme or the HYHEP diet are necessarily representative of current UK practice.

The value of the Weinstock and colleagues78 study to this systematic review is limited for a number of reasons. The trial predominantly reported changes in body composition and resting energy expenditure (REE), appetite and mood, and the long-term weight loss outcomes reported in the study were from a subgroup of the overall trial population. In addition, the number of participants randomised to each intervention was not explicitly reported (but has been estimated by our reviewers to be between 29 and 31 per intervention group) and the results presented were for three of the groups combined rather than for all randomised interventions.

In terms of starting weight, two of the studies included participants classified as BMI obesity I,76,77 and in a third study78 the participants were classified as BMI obesity II. The fourth study79 did not report baseline BMI; however, it did report weight in terms of kg that appeared to be reasonably similar to the other studies.

Poor reporting of methodological details of the studies hampers an informed judgement regarding potential risks of bias (see Quantity and quality of research available) and therefore these risks are generally regarded as uncertain, which needs to be considered when interpreting results presented below.

Assessment of effectiveness for multicomponent interventions where the focus was on other variables

One included study by Dubbert and Wilson68 did not appear to fit into the categories of active intervention versus no-active comparator, focusing on the dietary component, or focusing on the exercise component. This study is therefore considered separately in this section. The trial tested two variables – goal-setting (daily or weekly), and spouse involvement (or not) – in their multicomponent approach to weight loss. The intervention is briefly described here and in Table 19. BMI at baseline was not reported in this study (see Table 7); however, the participants were required to be at least 15 lb (approximately 7 kg) overweight and participants in both of these groups had mean weights around 208 lb (∼  94 kg). If height is assumed to be 170 cm, then these participants would be categorised as having class I obesity.

In the trial68 the interventions lasted for 19 weeks, with follow-up 30 months later. Participants were randomised to essentially the same multicomponent intervention undertaken either as an individual or with a spouse, and with the calorie and exercise goals set either daily or weekly. For the purpose of this review we consider the interventions with individuals and with daily goals as the multicomponent intervention of most relevance to the scope. The individual treatment with weekly goals is treated as the comparator intervention. The details of the couples’ interventions can be seen in Appendix 5. Multicomponent interventions often involve family members; however, in this study the spouse monitored adherence and progress with the weight loss programme and was involved in the problem solving and goal-setting.

In the daily goal-setting treatment group participants were recommended to set a goal of 1215 kcal per day (females) or 1525 kcal per day (males) from week 5. Participants were encouraged to divide the daily goals into subgoals for portions of the day. No detail was given for the proportions of the diet but participants were encouraged to record their weight daily as a form of monitoring. The exercise component of this intervention also set daily goals for energy expenditure from an aerobic exercise and walking programme. Participants started at an initial 145 kcal per day (equivalent to a 1.5-mile walk or 1.5 hours of active housework) above their initial baseline of activity and this increased by 25 kcal/day if the previous week’s goals had been met on at least 4 days. In addition, participants were instructed to walk for at least 30 minutes on 5 days each week and to monitor their heart rate to ensure that they were exercising within a 70%–80% range of their age-predicted heart rate (see Appendix 5). The participants recorded daily records of activities. The behavioural modification element of the interventions was undertaken in groups (size not reported) and included lectures and small group discussions in weeks 1–4; weekly 15- to 20-minute individual sessions in weeks 5–7; and then fortnightly sessions thereafter. The setting for the behaviour therapy was not reported. Table 20 describes the range of features and strategies covered in the behavioural treatment component of the study. These appeared to centre on goal-setting, cognitive restructuring and problem-solving skills. Clinical psychology graduate students delivered the sessions. The intervention with weekly goal-setting was very similar to the daily goal-setting intervention, with a few exceptions. The participants were encouraged to set weekly calorie goals of 8500 kcal for women and 10,675 kcal for men and the exercise component was to meet a weekly exercise goal equivalent to the sum of the same number of days for those in the daily goal-setting intervention (1000 kcal in the first week) (see Appendix 5). Participants were instructed to weigh themselves only once per week.

There are questions over the generalisability of the findings of this study to UK practice. Participants in this study had answered a media advertisement and paid a refundable fee to take part in the study, which may have affected the types of people taking part. The number of participants allocated to each treatment arm was not explicitly reported in the study report, and there was no report of results of any statistical significance testing of the data. The mean age of the participants in the study was not reported. Another factor that needs to be considered when interpreting results of this study is that the daily goal-setting for the diet component was likely to be more detailed than common practice. Also, the exercise component used a daily goal. The study was also undertaken some 25 years ago, before the advent of standards for reporting of RCTs, and this is reflected in the assessment of the trial quality that suggests an uncertain risk of bias (see Table 8).

Ongoing studies

The following studies, which may meet the review eligibility criteria, were identified in searches and are currently ongoing:

Ross and colleagues80 Prevention and Reduction of Obesity through Active Living (PROACTIVE). This is a Canadian RCT of 491 adults with BMI 25–39.9 kg/m². The goal of the study is to reduce abdominal adiposity, increase physical activity, and maintain and improve diet quality. Participants are randomised to the intervention group or a usual-care control group. Trial outcomes were unpublished as of December 2009.

Stolley and colleagues81 Obesity Reduction Black Intervention Trial (ORBIT). This is a US RCT of a culturally proficient weight loss intervention compared with a general health control group in 213 obese black women. Short-term (6-month) outcomes were published in January 2009.

van Weir and colleagues82 ALIFE@work trial aims to evaluate a lifestyle intervention among a Dutch overweight working population. Participants (BMI > 25 kg/m²) (n = 1386) were randomised to either the LI via the telephone, the internet or to a control group. Trial outcomes were unpublished as of December 2009.

Williams and colleagues83 Work, Weight, and Wellness: The 3W program. This is a worksite obesity prevention and intervention trial being conducted on 11,559 employees on the island of Oahu in Hawaii, USA. Work sites are randomised to either a minimal intervention or an intensive intervention. Trial outcomes were unpublished as of December 2009.

Østbye and colleagues84,85 Active Mothers Postpartum. This is a trial being undertaken in the USA of women with a BMI > 25 kg/m² post partum. A total of 450 women were randomly assigned to either an intervention group (healthy eating classes, physical activity classes, telephone counselling) or a control group. Short-term (12-month) outcomes were published in 2009.

Modifying Obesogenic Homes: Impact on Weight Maintenance, is an RCT which aims to deliver 18 months of SBT or 18 months of SBT plus direct modifications to the home environment (SBT + Home) to people who are BMI > 25. The study of approximately 200 participants is currently recorded as ongoing, but no longer recruiting participants and is funded by the National Heart, Lung and Blood Institute in the USA.

Internet Assisted Obesity Treatment (iReach) is an RCT comparing three weight loss interventions: internet alone, internet and periodic in-person support, and in-person support alone. A US study this is funded by the National Institute of Diabetes and Digestive and Kidney Diseases. The study aims to follow-up participants with a BMI > 25 kg/m2 at baseline for up to 18 months. This study is recorded as ongoing but no longer recruiting participants.

The Take HEED (Healthy Eating and Exercise Decisions) RCT is a US study which aims to examine the effect of a culturally-adapted weight loss programme in African American females aged 40–65 years. Take HEED is a combination of two active interventions, Therapeutic Lifestyle Changes diet and the ‘change’ exercise programme. The study is ongoing but no longer recruiting and is funded by Kaiser Permanente and the Garfield Memorial Foundation.

Maintenance-Tailored Obesity Treatment is an RCT comparing state-of-the-art, SBT for weight loss with a maintenance tailored treatment and will follow participants until 30 months. Participants are required to be BMI between 30.0 and 37.0 kg/m² to be included in this study which is ongoing but no longer recruiting participants. The study is a US study funded by the National Institute of Diabetes and Digestive and Kidney Diseases.

Participants in an 18-month study funded by the National Heart, Lung and Blood Institute will be randomly assigned to either a standard behavioral weight loss programme or a stepped-care weight loss programme. The study is currently recruiting participants with a BMI between 25.0 and 39.9 kg/m² to the ‘Step-up Study’ which is expected to complete in January 2012.

Another ongoing study is comparing a wellness-centred intervention with a weight-centred intervention (LEARN^® programme) in people with a BMI between 30 and 45 kg/m². The interventions are delivered over 6 months with follow-up at 18 months. The study is active but no longer recruiting participants and is funded by The Reading Hospital and Medical Center and The Edna G. Kynett Memorial Foundation in the USA.

Quantity and quality of published research

A total of 419 potentially relevant references were identified in the cost-effectiveness searches. Of these, the full texts of 15 papers were retrieved but no studies met all of the a priori inclusion criteria. A summary of the selection process and the reasons for exclusion are presented in Figure 2 and a list of excluded studies in Appendix 6. Two studies86,87 met the core criteria of evaluating diet, exercise and behaviour interventions in overweight or obese populations using a lifetime economic model. However, one of these included the use of prescription antiobesity drugs in a small number of participants87 and the length of follow-up in the other study was < 18 months (12 months). In the absence of any other evidence a pragmatic decision was taken to describe these two studies in the review of cost-effectiveness, as these studies were considered relevant and useful to the research question. However, caution is required in their interpretation. Characteristics of the studies are shown in Table 22.

FIGURE 2.

Flow chart of identification of studies for inclusion in the review of cost-effectiveness. a, The abstracts provided insufficient details of methods and results to allow inclusion in the systematic review; b, a pragmatic decision was taken to describe these studies despite them not fully meeting the a priori inclusion criteria.

Description and results of the published economic evaluations

The Counterweight Programme

The Counterweight Programme is a weight management programme in primary care in the UK. 87 An economic evaluation was undertaken using an existing model developed for the NICE clinical guidelines on obesity. 35 The evaluation examines the impact of weight loss interventions on health outcomes and costs. The model incorporates the relationship between weight gain and the increased risk of developing a range of associated conditions. The analysis compared the costs and outcomes of the Counterweight Programme with no active intervention. The cost per patient of the Counterweight Programme was derived from an analysis of the budgetary impact of implementing the programme. 54 The analysis was conducted from the perspective of the UK NHS.

Modelling approach

A patient-level simulation model was used which had previously been developed to provide input to the UK national guidance on obesity. 35 The model works by randomly selecting an individual whose characteristics are based on those of the UK adult population (for example, BMI, age and gender) and following their health-care costs and outcomes until death. The model assesses people over 6-month cycles. During each cycle, individuals may experience a change in BMI (increase or decrease) or no change. Based on their characteristics, individuals can develop diabetes, CHD, or colon cancer depending on the prevalence of each disease at the BMI they are currently experiencing. There is a utility score associated with each health state. Patients are at increased risk of death if they experience one of these conditions. To determine whether the individual is experiencing each particular health state at any one time, the incidence of the chronic health states was calculated, based upon the prevalence of the condition for that individual’s characteristics. The model was run for a cohort of 10,000 people.

Assumptions

In the base-case analysis, patients are assumed to regain all the weight lost in 2 years following removal of the programme while those obese adults not on the programme are assumed to gain a background rate of 1 kg per year (Figure 3). 87 Those patients who were not attending at 12 months were assumed to gain weight at the same rate as those not on the programme.

FIGURE 3.

Assumptions of weight change for the economic evaluation of the Counterweight Programme.

Effectiveness of intervention

The estimates of the effectiveness of the Counterweight Programme were based on a prospective evaluation of 1906 obese adults in 65 UK general practices. 54 These patients had a mean age of 49.4 years, a mean BMI of 37.1 kg/m² and were mostly female (77%). All patients received the intervention, as the study was not a controlled trial, and the results are compared with weight gain in observational studies of obese adults. The intervention consisted of diet, exercise and behaviour modification components. First-line interventions were a prescribed eating plan, using a goal-setting approach, or a group intervention. These were all aimed at achieving an energy deficit of more than 500 kcal/day. Patients were asked to commit to nine appointments in 12 months, including six individual appointments (10–30 minutes) or six group sessions (1 hour each) over a 3-month period and then follow-up appointments at 6, 9, 12 and 24 months. Mean weight change at 12 months (n = 642) was –3.0 kg (95% CI –3.5 to –2.4 kg) and at 24 months (n = 357) was –2.3 kg (95% CI –3.2 to –1.4 kg) compared with original weight. Mean change in BMI was –1.1 kg/m² (95% CI –1.3 to –0.9 kg/m²) at 12 months. Untreated patients are assumed to gain 1 kg weight per year. There were a large number of dropouts from the programme. Of 1419 attendees, 54% of patients (761) provided data at or beyond 12 months. These dropouts are likely to have incurred additional costs to the programme.

Caution is required in the interpretation of the evidence of effectiveness from the Counterweight Programme study. The effect of the intervention is not based on evidence from an RCT or a study with a concurrent control group and as such there is a higher risk of bias than there would be with a study using a more rigorous design.

Estimation of QALYs

The Counterweight Programme economic evaluation uses QoL utility values stratified for BMI levels, age and gender based on a discussion paper by Macran88 which estimated HRQoL based on European Quality of Life-5 Dimensions (EQ-5D) scores in the HSE, 1996 (Table 24). The utility values were for the general population that included individuals with CHD, diabetes and cancer. These values were adjusted to derive utility values for the general population without these chronic health states for different BMI scores. For a person with a relevant comorbidity, multipliers were applied to these utility scores to obtain their individual utility estimate. The multipliers for diabetes and CHD (0.8661 and 0.867, respectively) were from a manufacturer submission to NICE for the cost-effectiveness of sibutramine (Meridia^®, Abbott Laboratories), published in NICE guidance for obesity. 89 Individuals with colon cancer had their utility score reduced by 5%. The utility values for the different health states do not appear to have been derived from a systematic search.

TABLE 24 - Quality of life utility scores by BMI and gender

BMI (kg/m2)	Male	Female
< 21	0.86	0.85
21–25	0.87	0.87
26–30	0.86	0.82
31–39	0.82	0.78
> 39	0.88	0.75

Estimation of costs

The cost of the Counterweight Programme was based on participants from the evaluation of the weight management programme and scaled up to the national population. The costs used in the economic evaluation were the actual costs observed and recorded, and they were limited to the services actually provided. Costs included remuneration for all clinicians’ time required for the intervention – training, GPs’ time for clinical and motivational assessment, practice nurse time for assessment, motivation, delivery of advice and review. Costs for the Counterweight Programme project team, training resources and patient information materials were also included in the analysis. Costs for the intervention were based on the optimal attendance rate of at least six visits in 12 months. In addition, the costs of weight management medications prescribed according to protocol were incorporated. Around 20% of patients followed up at 12 months or around 9% of the total intervention group were prescribed weight management drugs. The scaled-up national costs, assuming that the programme was available in all practices in the UK and would be taken up by 20% of eligible obese patients, included the following: (i) initial staff recruitment and training (£2M); (ii) weight management advisor team and support staff (£8.5M/year); and (iii) resources and practice staff time for meetings, GP discussion, intervention, equipment, medication, exercise referral and secondary care referral (£16M/year). After the initial year, each cohort was assumed to require 50% of year 1 costs. The total 5-year UK costs were about £200M for 3.3 million patients, which equates to about £60 per patient.

This is a very low intervention cost. Looking at the costs in more detail, they appear to have been calculated incorrectly. The cost for the intervention is £5707 per one GP practice for 92 patients for the first year. The authors assume the follow-up for the second year is one-half of the rate in the first year (e.g. three individual contacts instead of six). Management costs for the scheme are about £15.25 per person. Thus the cost per patient would be about £108.30. This is still a low estimate for the programme, as there does not appear to be any time allocated for administration by the practice nurse or the GP and the time in practice to run the programme may be considerably more than allocated. In the study, patients attended GP appointments for other clinical matters and as part of this the GP suggested practice nurse appointments to discuss weight management. In this way there is no cost allocated for this GP appointment. There is also no time allocated to specialist advisors such as dietitians, cognitive behavioural therapy counsellors or health trainers, except as part of the initial nurse training. There are also no costs included for the fitness programme, such as leisure centre costs. We suggest that the NHS or other government departments pay for part or all of exercise referral schemes. We suggest that the resource costs are more likely to be similar to those used by a recent NICE report for bariatric surgery. 33 This suggested that for a diet-only programme there would be four contacts with a GP per year and two additional contacts per year with a community dietitian, practice nurse and district nurse at a total cost of £282 per year.

As mentioned above, the practice nurse is expected to run almost the entire Counterweight Programme. The practice nurse only receives one full day of training before the start of the course but is expected to gain extensive competencies, in order to run diet, exercise and behavioural components. This training was provided by registered dietitians, who also provided subsequent peer support for about 6 months. It is likely to be the case that GP practices require more than one practice nurse to run the scheme and that they need more support from specialist health professionals and hence the cost incurred will be substantially higher.

The costs of the chronic conditions were based upon previous studies and converted to 2005 prices. Yearly costs included medical conditions such as CHD (£1637), diabetes (£653) and colon cancer (£7320).

Follow-up data were available for only 642 of 1419 participants at 12 months, with the remainder not reaching these data collection points. The costs and effect size have been estimated for these 642 participants. It is unclear what the costs and effect size are for these dropouts, and the weight loss may be lower for these individuals while they will still have incurred the intervention cost.

Cost-effectiveness results

The model considers the impact of the 12-month observed outcomes of the Counterweight Programme. For the non-active intervention, lifetime QALYs are 28.32. For the baseline intervention, lifetime QALYs are 28.38, i.e. a QALY gain of 0.056 for the intervention. The lifetime cost to the NHS of the non-active intervention is £1884 and for the baseline intervention is £1857. The incremental cost-effectiveness ratio (ICER) is –£473 per QALY gained, i.e. it is cost saving. No sensitivity analyses were conducted for structural, methodological or parameter uncertainty. Scenario analyses were conducted for background weight change for the untreated population. For background weight change of 0.5 kg/year and 0.3 kg/year the ICER was £2017 and £2651, respectively.

Summary of key issues

• The study has evaluated a diet, exercise and behavioural programme for obese people in the UK, using the actual costs observed in the Counterweight Programme.

• There appear to be some miscalculations in the intervention costs used and the costs recorded are lower than reported in other studies.

• The intervention effects are based upon a single observational study, rather than a controlled trial or systematic review.

• There are high dropout rates in this study that will have an added cost.

• The methodology of the model and the derivation of the QoL utility values seem reasonable.

Roux and colleagues86 conducted an economic evaluation of outpatient weight loss strategies in overweight and obese women in North America. Each weight loss intervention consisted of a 6-month intervention followed by a 6-month maintenance programme. The strategies were: routine care; diet only; diet and exercise; diet and pharmacotherapy; and diet, exercise and behavioural modification. In all strategies women who successfully completed the 6-month weight loss intervention were modelled to enter the 6-month maintenance phase. Women who were unable to lose weight or maintain weight loss were assumed to remain at their age-adjusted original BMI. The base case simulated the natural history of a hypothetical cohort of healthy, non-pregnant 35-year-old overweight and obese women with original BMI > 25 kg/m² and free from known CHD. The model was used to project lifetime costs and gains in life-years and QALYs, and the analysis was from a societal perspective. Costs were in US$ and the price year was 2001. The study used benefits from published RCTs and costs were derived from published data, Medicare reimbursement rates and primary data for indirect costs.

Assumptions

It was assumed that the probability of long-term weight maintenance was 20% with lifestyle modification programmes and 10% without lifestyle modifications based on published literature. Women who successfully maintained their weight loss for at least 5 years were assumed to remain at their postintervention BMI for the remainder of their lifetime. Figure 4 shows the assumptions used in the model for weight change for the intervention versus routine care (no intervention), although the study did not clearly describe how weight changed over time in the routine care group. Age- and BMI-specific obesity-related disease complications were assumed to be lifelong. It was assumed that a woman who achieves only short-term success accrues the clinical benefit associated with a lower BMI until she regains weight. Women who do not lose weight or who lose weight but do not maintain the weight loss are subject to age-adjusted cardiovascular disease risks based on their original BMI.

FIGURE 4.

Assumptions of weight change for the economic evaluation by Roux and colleagues86 Weight changederived by assuming average height in cohort of women of 1.63m and those who do not maintain weight loss return tooriginal weight.

Comparison of cost-effectiveness analyses

The intervention effect varied between the evaluations. The mean weight change was –3.0 kg at 12 months (–1.1 BMI kg/m² at 12 months) and –2.3 kg at 24 months for the Counterweight Programme. 87 Roux and colleagues86 reported weight change of –3.11 kg/m² change after 6 months intervention and thereafter the probability of maintaining a 10% weight loss was 67% at 1 year and 20% at 5 years. The weight loss from the Counterweight Programme87 seems consistent with that seen in the trials in our clinical effectiveness systematic review (see Chapter 3), while the weight loss from Roux and colleagues86 appears optimistic.

There was a range of assumptions used to extrapolate weight change beyond the follow-up period of the intervention. These are shown graphically in Figure 5. Roux and colleagues86 assumed that after 5 years, 20% of women on the diet, exercise and behaviour programme would maintain weight loss and that they would remain at their postintervention weight or maintain their weight loss post intervention for the remainder of their lifetime. In the base case, the Counterweight Programme87 assumed that those on the programme return to the same weight as the control population within 2 years of finishing the programme. In a ‘best-case’ scenario, they assumed that individuals remain at a lower weight than the control population lifelong. From the evidence presented in our systematic review of clinical effectiveness, we suggest that the control group are likely to gain weight but at less than 1 kg/year, perhaps in the region of 0.3–0.5 kg/year. Individuals finishing the weight loss programme would be likely to regain weight within 2–4 years after the end of the intervention, and thereafter gain weight at a similar rate to the control group. Figure 5 shows the weight loss of the intervention group compared with the control groups in the two economic analyses and one of the studies from the systematic review of clinical effectiveness that seemed the most robust (Stevens and colleagues70). It is likely that although people may regain lost weight eventually, because the population gains weight throughout life, people may still be at a lower weight than those who never received a weight loss intervention. Compared with the Stevens and colleagues study,70 the assumptions used to extrapolate beyond the follow-up period of the intervention seem reasonable for Roux and colleagues,86 but for the Counterweight Programme study87 seem pessimistic, as they do not assume any long-term difference between the intervention and control groups. On the other hand, in the best-case scenario of the Counterweight Programme87 the long-term weight difference between the intervention and control group seems optimistic.

FIGURE 5.

Assumptions of weight loss difference between the intervention group and control group for the economic evaluations. Weight change derived for Roux and colleagues86 by assuming average height in cohort of women of 1.63 metres and those who do not maintain weight loss return to original weight.

The gain in QALY compared with routine care varies between 0.06 and 0.243 for the two studies. These differences are mainly due to the differences in the assumptions for long-term weight loss maintenance and the utility values chosen for weight change. Neither study provided a clear explanation of the derivation of HRQoL values or was based upon a review of QoL literature. The studies adopt very different estimates for the effect on QoL of a reduction in weight. Roux and colleagues86 use a HRQoL value of 0.93 for individuals with obesity and 0.87 for those who achieved a 10% weight loss, i.e. a change in BMI of roughly 3 kg/m². The Counterweight Programme study87 used values shown in Table 24, for example females have a utility of 0.82 and 0.78 for BMI between 26 and 30 and between 31 and 39 kg/m², respectively. The change in utility values associated with a unit change in BMI varied between about 0.006 for the Counterweight Programme87 and 0.02 for Roux and colleagues. 86

The cost of the intervention varies between around £60 per patient in the Counterweight Programme study87 and US$3040 (£1820) for the diet, exercise and behaviour programme in the Roux and colleagues study. 86 Roux and colleagues86 itemised the costs included in their programme and costs are included that would not be included in UK economic evaluations, for example non-medical and time per patient costs. There is also a large difference between the costs due to the personnel used to deliver the programmes. For example, Roux and colleagues86 state that the medical cost of the programme includes consultation costs for physician, nutritionist, exercise therapist and psychologist while the Counterweight Programme used six 20-minute consultations with a practice nurse to cover diet, exercise and behavioural components (total cost £43). Furthermore, the Roux and colleagues86 programme included other investigations such as laboratory tests, X-ray, electrocardiogram and exercise stress test which were not included in the Counterweight Programme. 87

Summary of cost-effectiveness studies

• Two cost-effectiveness studies were described in our review of economic evaluations.

• Each study used a lifetime chronic disease model to evaluate the effect of changes in individuals’ weight. The models included the costs and benefits from avoiding chronic illnesses such as CHD and diabetes.

• Omissions in reporting details of the modelling methodology and data inputs reduced transparency and make it difficult to draw conclusions about the results; however, the results and methodology of the studies seemed reasonable.

• The Counterweight Programme study87 is of most relevance to the UK as the health-care system in the study was the NHS and the study was conducted from the perspective of the UK NHS.

• There were limitations to each study. For the Roux and colleagues North American study,86 the costs were much higher than for the UK NHS perspective as non-medical costs have been included. For the UK study,87 the intervention effect is not based upon an RCT, and there is a likely underestimate of the actual costs for the intervention. Neither study has fully explored the uncertainty around the results.

• Both studies found the interventions to be cost-effective, with estimates varying between –£473 and £7200 ($12,640) per QALY gained.

• One study compared86 a diet, exercise and behavioural modification strategy with strategies for routine care, diet only, diet and exercise, and diet and pharmacotherapy, and found diet, exercise and behavioural modification to be the most effective and efficient strategy.

• Sensitivity analyses conducted in one of the studies86 showed that the results were most influenced by the probability of weight loss maintenance and utility values for obesity.

Statement of principal findings

General discussion

Even though studies in the review of clinical effectiveness were grouped to try to keep the most similar studies together, the studies were still rarely comparable. Differences in the types and durations of interventions, and any subsequent weight maintenance strategies, the length of follow-up, issues around generalisability to the UK and the risk of bias of the studies mean that it is difficult to draw robust conclusions as to the effectiveness of multicomponent weight management programmes. It is also difficult to establish what the core components of such programmes may need to be to maximise and sustain weight loss. With such complex interventions it is difficult to establish with any precision what the ‘active ingredient(s)’ causing any demonstrated effect is. It may be that there are necessary elements to successful weight loss, but with so few data, and so few similarities between interventions, it is difficult to draw any conclusions on this.

As noted earlier we informally assessed all of the studies included in this evaluation in terms of their length of follow-up, sample size, risk of bias, degree of reproducibility of components of the interventions, generalisability and overall effect size (see Appendix 8). Speculatively, the weight loss interventions evaluated by the Stevens and colleagues70,74 studies may offer a useful model for long-term weight management. In particular the Stevens and colleagues70 TOHP-II study, which has a large sample size, reported a statistically significant effect on weight loss at least 24 months after the active intervention, and the intervention appears to be reproducible. In terms of overall methodological rigour, the study was judged to have a generally low risk of bias. It is unclear whether the study was statistically powered to detect an effect of weight change (because the study was powered for reduction in BP); however, the large sample size provides no reason to suspect that the weight change outcome would not have been adequately powered. Its generalisability to the UK is uncertain. Although this study, based on the Trials of Hypertension, had some shortcomings it appears on balance to be best placed among those included in this systematic review to provide a model for further exploration and testing with overweight populations in the UK.

This study had two main phases: an intensive phase (14 weeks) and a maintenance phase (16–18 months), with some maintenance continuing until follow-up at 36 months. The participants’ goals were to lose at least 4.5 kg during the first 6 months of the intervention and to maintain that weight loss for the remainder of the trial (participants did regain weight over time but their weight at 36 months remained just lower than at baseline). Key features of the intervention were a calorie-controlled diet with a focus on decreasing consumption of excess fat, sugar and alcohol, a gradual increase in walking, monitoring through self-report (diaries) and at group sessions, and a relatively simple behavioural therapy intervention. This comprised goal-setting or action plans, relapse prevention, self-monitoring and social support. These components appear consistent with recent NICE obesity guideline advice. The extended support and relatively straightforward goals for diet and exercise may be contributing factors to the weight change results shown in the intervention group compared with the usual-care control group but we are unable to test this. However, the resource requirements to deliver an intervention such as this are unclear at the present time as few details were reported in the study publication.

In the studies included in our review the interventions varied in terms of their length, their components, the personnel involved and the ongoing maintenance/support mechanisms involved. For example, if we look at the number of contacts with participants, personnel involved in the interventions saw or had contact with participants weekly for at least 14 weeks in most of the included studies. It is unclear whether this is realistic in terms of the likely resource availability or as an expectation of the participants. In the intervention that we feel may have the most potential for testing in a UK setting70 the intervention was delivered by dietitians and health educators with some input from psychologists. It included 14 weekly sessions, then six biweekly sessions and then three to four monthly sessions during the 18-month intervention and then at least three contacts during the 18-month ongoing support phase. Weight change was generally positive and dropout rates were around 10% in both the intervention and a usual-care control group.

Throughout our report we have commented on issues of generalisability of the populations where appropriate. For example, where the populations are of a certain age or gender, have answered media advertisements for entry into the studies and/or paid deposits to do so and the fact that none of the studies were carried out in a UK setting and so may have different health systems and populations. A potential example of a more generalisable intervention is the Counterweight Programme. 87 This is currently being rolled out in primary care in many areas of the UK. However, the evaluation of the Counterweight Programme87 intervention was based on a non-randomised study and thus did not meet the inclusion criteria for our systematic review of clinical effectiveness. We are therefore unable to ascertain whether the intervention is clinically effective. The economic evaluation of the Counterweight Programme87 was summarised in our systematic review of cost-effectiveness studies and appeared to be cost saving. However, caution in interpretation is required as the study did not meet the inclusion criteria for this review and there were uncertainties around estimates of the costs of the programme and around assumptions about the extrapolation of effectiveness estimates over time.

Comparison to existing systematic reviews

We have identified 10 existing systematic reviews that focused on weight management interventions for adults (see Table 6). All of these systematic reviews had included studies of interventions comprising diet, exercise and/or behavioural components, but none specifically included only studies with all three components as recommended in the NICE obesity guideline. There appeared to be a consensus from these systematic reviews that the most effective weight loss interventions were those that were multicomponent. Only one of these existing systematic reviews looked specifically at long-term evidence of the interventions, but its aims and inclusion criteria were different to those of our systematic review, precluding comparisons.

Other issues and methodological concerns

The scope of this review was to focus on studies that would be of most use to policy and practice. We therefore sought studies with long-term follow-up and in which the interventions were clearly described (to enable replicability). Despite this, we identified evaluations of a range of different multicomponent interventions, making it difficult to compare results and to make meaningful conclusions. There are a number of other factors that should also be considered in the interpretation of the studies.

Weight regain was common among participants in the included interventions, even in studies with extended ongoing support. Unfortunately, the data available do not allow us the opportunity to offer any inference as to what might be causing this and it is unclear whether this is clinically meaningful in some way.

Incentives to recruit people to participate were used in three of the included studies. 72,73,79 Although not directly tested it does not appear that there was any relationship between using incentives and the likelihood of weight change.

Our review aimed to comment on any particular barriers to or facilitators of weight loss that may help to establish what the key components of these types of interventions should be, but the evidence we reviewed offers no insights into this. From the available data it is not possible to determine whether the weight loss interventions had any negative effects on participants.

Weight loss success may depend on whether trial participants have previously attempted weight loss or how long they have experienced overweight or obesity. Data on these factors could potentially be used to target interventions to certain populations. We attempted to capture these data; however, treatment history was only reported in three72,75,76 of the included studies and the duration that participants had been overweight or obese was not reported in any study. As such there are not enough data for us to reflect on any patterns in the results seen.

Even in those studies with a non-active comparator (e.g. usual care) it was difficult to establish the true estimate of the effects of an intervention as very few studies adequately described their non-active comparator. It seems likely that participants assigned to control arms would still have received some contact with health-care professionals, which may have the potential to underestimate the treatment effect.

The methodological quality of the included studies varied. None of the studies reported all the details necessary to assess all 10 of the quality assessment criteria that we used. In most studies fewer than half of the criteria could be adequately assessed. Similarly, between the included studies there were no individual quality assessment criteria that were reported by all of the studies. A recent systematic review91 of 63 RCTs of any intervention for weight loss found that while reporting seemed to have improved since the publication of the revised consolidated standards of reporting trials (CONSORT) statement in 2001, reporting of some key aspects was still poor. In the review, 60% of the overall CONSORT criteria were satisfied by the RCTs, but the reporting of criteria relating to the methods varied, with just 19% of studies satisfying the reporting criteria relating to treatment allocation. The authors of that systematic review suggest that there is considerable room for improvement in the adherence to the CONSORT reporting criteria.

Many studies included in our evaluation were undertaken more than 10 years ago. It is likely that this would have had a bearing on the reporting of methods used in the studies and the interpretation of results. For example, the general population today is more exposed (e.g. by mass media) to the issue of overweight and obesity and, as a result, likely to be more aware of healthy eating and physical activity messages. It is therefore possible that the effect of interventions conducted a decade or more ago might not be so large in the present day. Over time the roles of health professionals may also have changed, so the types of people delivering weight management interventions today may be different from those who would have delivered such interventions in the past. Practice may also vary geographically, including within countries. The studies in our review provided little detail on the process of training providers.

In the two studies86,87 that were described in our cost-effectiveness review the costs of the multicomponent weight loss programme varied between £60 and US$3040 (£1820) per participant. The costs reported in the other reviews for weight programmes appear also to be within this range. A NICE guidance report35 for the management of obesity included a review of diet, exercise and behavioural treatment and the resource costs for these components. In the NICE report, physical activity interventions cost between £532 and £737; diet interventions cost between £103 and £621; and a diet and behavioural intervention cost £672 per patient. One study included in the NICE review of a pharmacological trial for sibutramine included diet and exercise advice as the control arm with a cost of £243.48 per person. 92 Avenell and colleagues30 reviewed treatment for obesity and constructed an economic model for a diet and exercise intervention for individuals with impaired glucose tolerance. The total estimated cost for the LI was £324 per person in the first year and £178 per person for subsequent years. The resource costs reported for other weight loss interventions seem to concur with our conclusion that the costs in the Counterweight Programme study87 are too low, and those from the Roux and colleagues North American study86 do not reflect UK costs.

The gain in QALY compared with routine care varies considerably between the two studies86,87 due to the differences in the assumptions for long-term weight loss maintenance and the utility values chosen for weight change. The change in utility values associated with a unit change in BMI varied between about 0.006 for the Counterweight Programme87 and 0.02 for Roux and colleagues. 86 Neither study based these utility estimates upon a review of the QoL literature. A recent Health Technology Assessment report33 on the clinical effectiveness and cost-effectiveness of bariatric surgery for obesity conducted a targeted search to identify published utility estimates for BMI values relevant to obese adults. The authors of this study stated that the values from the study by Hakim and colleagues93 represented the most methodologically sound estimates derived. Hakim and colleagues93 found that a one-unit decrease in BMI in obese individuals without diabetes was associated with a gain of 0.017, which was independent of age or gender. The utilities values from this study are consistent with those chosen by Roux and colleagues,86 but much higher than those used by the Counterweight Programme study, suggesting that the Counterweight Programme study may be underestimating some of the health gain associated with weight loss. However, although the QALY gains are small in these studies, the interventions are still cost-effective by generally accepted thresholds, because the interventions have such low costs.

One of the studies discussed in our review (Roux and colleagues86) included a utility decrement associated with the effort involved with participating in a weight management scheme. This resulted in the situation where participants of a diet-only strategy had lower lifetime QALY than those having routine care. As the primary data analysis to derive these data was not reported, it is unclear how these data were derived and whether the assumptions are valid. Removing this assumption would result in a more favourable cost-effectiveness estimate.

The cost-effectiveness results in the two studies described in our cost-effectiveness review are consistent with other cost-effectiveness studies for diet and exercise interventions (but without a behavioural component). Gallani and colleagues94,95 described a Markov model for a diet and exercise intervention in Switzerland in overweight and obese people. In their analysis the diet and exercise intervention was cost-effective for all subgroups with an ICER < £5000 per QALY. Bemelmans and colleagues96 described a cost–utility study in the Netherlands in overweight and obese people. The ICER for the diet and exercise intervention was €7400 (£6700) per QALY saved. Furthermore, a NICE review35 for diet, exercise and behavioural treatments concluded that ‘the cost per QALY in the best-performing non-pharmacological studies ranges from £174 to £9971’. This suggests that even though there were limitations to each of the studies in our cost-effectiveness review, the results appear to be consistent with other studies.

Strengths and limitations

This review has the following strengths:

• It is independent of any vested interest.

• It has been undertaken following the principles for conducting a systematic review. The methods were set out in a research protocol (see Appendix 1), which defined the research question, inclusion criteria, quality criteria, data extraction process and methods to be employed at different stages of the review.

• An advisory group has informed the review from its initiation. The research protocol was informed by comments received from the advisory group and the advisory group has reviewed and commented on the final report.

• The review brings together the evidence for the clinical effectiveness and cost-effectiveness of long-term weight management schemes for adults who are overweight or obese. This evidence has been critically appraised and presented in a consistent and transparent manner.

In contrast, this review also has certain limitations:

• Synthesis of the included studies was through narrative review. Although 12 studies were included in the review of clinical effectiveness, differences in the interventions meant that meta-analysis was inappropriate.

• No cost effectiveness or cost-effectiveness studies were found which met our full inclusion criteria and we therefore took a pragmatic approach to discuss two studies that met many of the attributes our review was looking for. Caution is therefore required in the interpretation of our results in terms of cost-effectiveness.

• Searches were limited to the English language. Therefore, we may have omitted non-English language, but otherwise includable, studies from our review.

• Many studies were excluded from our review because they either did not include a behavioural therapy element or they reported that there was a behavioural element but did not provide any details of how this was implemented. Although we attempted to contact authors of a number of studies to try to ascertain further details, this was unproductive. It may be that there are other studies with a long-term view of a weight management intervention that we were therefore unable to include. However, it is also apparent that the term ‘behavioural’ appears to be used in many situations to mean that a study aimed to change behaviours, but that this was as a result of the diet and/or exercise intervention components rather than of a specific behavioural therapy. It is unclear whether this is the case in any of these situations.

Need for further research

There have been a number of RCTs of multicomponent weight management interventions, and despite a number of differences between the many studies, there does appear to be some degree of success in general. It is uncertain whether any one particular intervention is best; however, the results of our review suggest that the TOHP intervention70,74 may be useful for testing further in a UK population. If any future RCT takes place, then replicating the intervention used in the TOHP study should be considered.

As most interventions succeeded in short-term weight loss, more research into the most appropriate long-term support to improve long-term maintenance of weight loss is required. Researchers should consider the Medical Research Council framework for developing evaluations of complex interventions97 and the National Obesity Observatory standard evaluation framework for weight management interventions98 when they design their studies.

There is a need for information on barriers to and facilitators of weight change in weight management interventions to be reported and, if possible, evaluated in clinical trials. As well as providing useful evidence on individual factors associated with greater or lesser weight change, such information could assist understanding of which of its components a multicomponent intervention should focus on.

There is a need for better reporting of behaviour therapy interventions, in terms of clear details of the techniques used, theoretical model, the format, setting and provider. 62 Authors of studies should more critically consider the reproducibility of their interventions. A taxonomy of a variety of behaviour change techniques has been devised and tested, to facilitate a common classification by intervention providers and researchers. 57 It would be advantageous for future evaluations to employ such a system in the reporting of interventions to facilitate a greater understanding of the specific components of interventions associated with effective health-related behaviour change.

An economic model for weight management interventions is needed for the UK NHS that includes a complete explanation of model structure, assumptions and data inputs. This model should be based upon evidence of the intervention effect, from an RCT or systematic review, with UK-relevant data inputs, particularly with regard to costs.

Contributions of authors

E Loveman (Senior Research Fellow) developed the research protocol, drafted the background section, assisted in the development of the search strategy, assessed studies for inclusion, extracted data from and quality assessed included studies, synthesised evidence, drafted and edited the final report, and project managed the study; G Frampton (Research Fellow) drafted the background section, assessed studies for inclusion, extracted data from and quality assessed included studies, synthesised evidence, drafted the final report; J Shepherd (Principal Research Fellow) drafted the background section, assessed studies for inclusion, extracted data from and quality assessed included studies, synthesised evidence, drafted the final report; J Picot (Research Fellow) drafted the background section, assessed studies for inclusion, extracted data from and quality assessed included studies, synthesised evidence, drafted the final report; K Cooper (Senior Research Fellow), assessed studies for inclusion, extracted data from and quality assessed included studies, synthesised evidence, drafted the report; J Bryant (Principal Research Fellow) developed the research protocol, assessed studies for inclusion, extracted data from and quality assessed included studies, synthesised evidence, drafted the report; K Welch developed the search strategy, undertook literature searches and edited the final report; and A Clegg (Professor/Director of SHTAC) developed the research protocol, drafted and edited the final report and acted as guarantor.

Disclaimers

The views expressed in this publication are those of the authors and not necessarily those of the HTA programme or the Department of Health.