A systematic review of UK‐based long‐term nonsurgical interventions for people with severe obesity (BMI ≥35 kg m−2)

Abstract Introduction The aim of this project was to systematically review UK evidence on the effectiveness of long‐term (≥12 months) weight management services (WMSs) for weight loss and weight maintenance for adults (≥16 years) with severe obesity (body mass index ≥35 kg m−2), who would generally be eligible for Tier 3 services. Methods Four data sources were searched from 1999 to October 2018. Results Our searches identified 20 studies, mostly noncomparative studies: 10 primary care interventions, nine in secondary care specialist weight management clinics and one commercial setting intervention. A programme including a phase of low energy formula diet (810–833 kcal day−1) showed the largest mean (SD) weight change at 12 months of –12.4 (11.4) kg for complete cases, with 25.3% dropout. Limitations or differences in evaluation and reporting (particularly for denominators), unclear dropout rates, and differences between participant groups in terms of comorbidities and psychological characteristics, made comparisons between WMSs and inferences challenging. Conclusions There is a persistent and clear need for guidance on long‐term weight data collection and reporting methods to allow comparisons across studies and services for participants with severe obesity. Data could also include quality of life, clinical outcomes, adverse events, costs and economic outcomes. A randomised trial comparison of National Health Service Tier 3 services with commercial WMSs would be of value.


Introduction
In the UK, obesity is managed on a tiered path by National Health Service (NHS) and community services.Tier 1 includes universal prevention services, Tier 2 includes lifestyle interventions in primary care, Tier 3 includes specialist multidisciplinary weight management services (WMSs) and Tier 4 includes bariatric surgery (1-3)   .Although people with severe obesity are likely to attend Tier 2 WMSs, having severe obesity (with or without comorbidities), may be a referral criterion for Tier 3 WMSs, prior to Tier 4 services (4,5) .Although adults with severe obesity may require more support with weight management, current National Institute for Health and Care Excellence (NICE) and Scottish Intercollegiate Guidelines Network (SIGN) guidance on WMSs provides little additional information for this group, apart from very-low-energy formula diets (VLEDs) (providing ≤800 kcal day À1 ) for people who need to lose weight quickly (e.g. for joint replacement or fertility treatment) (3)(4)(5)(6)(7)(8)(9) .VLEDs are rarely used in the NHS, although there is increasing interest in the use of low energy formula diets (LEDs) (800-1200 kcal day À1 ).Prior attendance at a Tier 2 service may be a criterion for entering a Tier 3 service.
Effective services could reduce the numbers of patients moving on to higher tiers of weight management or contribute to the long-term effectiveness after bariatric surgery.Our aim was to systematically review the UK evidence base for long-term (≥12 months) behavioural interventions for weight loss and weight maintenance for adults with severe obesity [body mass index (BMI) ≥35 kg m À2 ] and evaluate their effectiveness.

Materials and methods
The present study comprises an analysis of WMSs that are Tier 3 services or similar to Tier 3 services (e.g.participants with a spread of obesity-related comorbidities and/ or BMI ≥35 kg m À2 ) and is an updated version and a subgroup of results from the National Institute for Health Research funded REview of Behaviour And Lifestyle interventions for severe obesity: AN evidenCE synthesis (REBA-LANCE) (10) project.A protocol was registered a priori (PROSPERO No CRD42016040190).This systematic review is reported following the PRISMA standard (11) .

Inclusion criteria
Full-text reports of UK WMSs of any study design published since 1999, in NHS clinical settings (e.g.primary care, secondary care) or commercial organisations, with a mean or median duration of ≥12 months of follow-up, which included adults (mean or median age ≥16 years) with a mean or median BMI ≥35 kg m À2 , were included.Studies focusing on participants with only one type of morbidity, as indicated by study inclusion and exclusion criteria, were excluded to reflect generalisable interventions for people with obesity and a range of comorbidities, rather than condition-specific interventions, which would also have a behaviour change focus tailored for specific diseases, such as type 2 diabetes and weight management and blood sugar monitoring.Weight loss or prevention of weight regain after weight loss interventions (including VLEDs and LEDs), other dietary treatment, physical activity, behavioural counselling or a combination of these interventions were included.Interventions that included a pharmacological component (e.g.orlistat) were included only if this was offered as part of a WMS (i.e.studies were excluded for which the purpose was to evaluate orlistat).
The primary outcome was weight change or BMI change.Changes in secondary outcomes (e.g.cardiovascular risk factors) can be found in the full REBALANCE report (10) .

Literature searching
Literature searches were undertaken in four databases (MEDLINE, EMBASE, PsycINFO and Clinical Trials.gov) for interventions from 1999 to October 2018 (10,12,13) .ClinicalTrials.gov was searched for ongoing studies and reference lists of included studies were scanned to identify additional potentially relevant studies.Nineteen relevant NHS and commercial organisations, including Dietitians in Obesity Management, and the REBALANCE advisory group were contacted to help identify further published and unpublished reports.See REBALANCE report (10) for full search strategies.
The first, second and last author of the main included publications were contacted to identify additional materials (e.g.protocols, trial materials) that would assist data extraction.

Data extraction and quality assessment
Three reviewers (MA-M, CR and FS) independently screened titles, abstracts and full text reports, with a 10% check for agreement.The Template for Intervention Description and Replication (TIDieR) checklist was used for data extraction (14) .Each reviewer extracted details of study design, methods, participants, interventions and outcomes, and TIDieR (14) .A second reviewer (AA) checked numerical data extraction.Data for weight change are presented for complete cases, imputed estimations, last observation carried forward or baseline observation carried forward, as presented by authors.
Three reviewers (MA-M, CR and FS) conducted a double-blinded quality assessment of the included studies.The Cochrane risk of bias tool was used to assess randomised controlled trials (RCTs) (15) and a 17-question quality assessment tool (ReBIP) was used to assess nonrandomised comparative and case series studies (16) .An adapted version of the Campbell and Cochrane Equity Methods Group checklist (17) was used to assess the effect of interventions on disadvantaged groups and/or their impact on reducing socio-economic inequalities.

Characteristics of the participants
In total, 22 406 participants started interventions and 8982 were included in the analyses at final follow-up, although numbers were sometimes unclearly reported.Two studies included only women (30,31) .Sample size varied from 84 (31) to 6715 (22) participants.Women represented 76.1% of the total population.The average participant age (weighted mean) was 48.4 years.The youngest reported mean age was 39.9 years (33) and the oldest was 55.8 years (23) .The average BMI (weighted mean) of all participants was 39.9 kg m À2 , the lowest (31) reported mean BMI was 35 kg m À2 and the highest (37) was 50 kg m À2 .Of note, 8.2% of women included in the study by Cartwright (20) had a BMI ≥60 kg m À2 .
Three studies (21)(22)(23) did not report exclusion criteria.One trial (18) and one study (32) excluded participants using pharmacological treatment for obesity (e.g.orlistat), whereas three others offered orlistat as an optional drug treatment within the intervention. (25,28,29)One of the primary care trials excluded participants with a BMI ≥45 kg m À2 (29) and one trial excluded participants with a perceived incapability of walking 100 m (26) .One trial (26) and one study (31) reported excluding participants with psychiatric conditions (including eating disorders).

Assessment of risk of bias
The overall methodological quality was poor across studies (Figs 2 and 3).In four RCTs (18,26,29,31) , many of the domains were assessed as being at a high risk of bias (Fig. 2).Only just over half of these studies (52.6%) provided information on participant dropouts.

Assessment of equity and sustainability
Half (50%) of the studies were conducted in settings that might target or exclude specific populations.Most (65%) did not report socio-demographic differences between completers and withdrawals/dropouts, although 75%, reported details for some PROGRESS categories (Place of residence, Race/ethnicity, Occupation, Gender, Religion, Education, Socio-economic status, or Social capital).Few (25%) considered sustainability, although 60% discussed their interventions in organisational contexts.Five studies (22,25,30,32,36) reported organisational partnerships (e.g.NHS, commercial organisations, local authorities and community groups) (Fig. 4).Few studies assessed the fidelity of intervention delivery or participant adherence to interventions, and few reported intervention-related adverse events.Potential for conflict of interest was unclear in 15% of studies.
One trial (Cambridge Weight Plan UK) (18) and one study (LighterLife Company) (25) received partial or full financial funding from the intervention manufacturer.In two further studies (24,36) , no conflict of interest was declared, but Cambridge Weight Plan UK donated products.

Assessment of effectiveness
As a result of study heterogeneity, a narrative overview is presented according to the setting where the intervention was delivered.

National Health Service primary care
Across primary care services, 10 eligible studies (18,21- 23,25,26,28,29,31,32) were identified.Most of these studies were undertaken in England, with the exception of two (21,28) undertaken in different sites across the UK, as well as two (22,25) in Scotland.In all cases, primary care practices were involved as the main setting of the studies, except one (28) that not only mainly recruited participants from primary care settings, but also included participants from commercial services (i.e. one commercial weightmanagement service and recruitment through eight freelance Counterweight-Plus trained practitioners).Women made up the majority of participants in primary care studies (over 60%) and one study recruited only women. (31)he interventions were mainly delivered in primary care practices to individuals.One study also applied the intervention in pharmacies and community settings (22) .The main care providers were nurses (21)(22)(23)25,26,28) , dietitians (21,22,25,28,31,32) , general practitioners or psychologists (29,31) . One tral described the intervention provider as a 'LED counsellor' (18) . One study of primay care interventions incorporated other professionals, such as an exercise scientist (31) .In most cases, the interventions were delivered individually, although three studies implemented group sessions (29,31,32) .
One trial (18) and two (25,28) studies evaluated the efficacy of LEDs in primary care, the latest in addition to the Counterweight programme (25,28) .In these three cases, the Cambridge Weight Plan/Counterweight PRO800 UK LED was offered (LED with 810-833 kcal day À1 ) and, in the study by Lean et al. (25) , an option of an 810 kcal day À1 homemade LED was also available.Few of the interventions defined the nutritional characteristics of the dietary advice/or nutritional programme in depth (18,21,22,25,28,31) .Similarly, only one intervention provided in depth detail on the physical activity plan offered to participants (28) .General characteristics of the included studies delivered in the primary care studies are provided in Appendix 1. Overall weight, percentage of weight and BMI change are presented in Table 1.
In primary care, studies that provided LEDs were those with the higher weight loss.For example, after 12 months of follow-up, Lean et al. (25) reported a mean (SD) weight loss of 12.4 (11.4) kg for completers with 25.3% drop out from baseline.A similar result was reported by Astbury et al. (18) where those participants randomised to LED were reported to have a mean (SD) weight loss of 10.7 (9.6) kg for completers [10.2 (9.7) kg by multiple imputation] and a dropout rate of 24.6%.In another study incorporating a LED, McCombie et al. (28) reported a mean (SD) weight loss of 14.2 (11.6) kg at 12 months for complete cases [À10.5 (9.5)The  ). † Complete cases.‡Analysis adjusted for missing data.
From those interventions that did not include VLEDs or LEDs, higher reported weight losses were associated with higher dropout rates, which reflected selective reporting of results.Most primary care studies that did not include VLEDs or LEDs achieved weight losses at 12 months of 2-4 kg mostly for complete cases and dropout rates of 20-30%, with the exception of Counterweight studies where dropout rates were 55-72% at 12 months.
All WMSs included multidisciplinary teams (mainly a physician with a special interest in obesity, dietitians, and psychologists) and offered a similar service (behavioural therapy, including reduced calorie diets, LEDs, VLEDs and, in some cases, orlistat).Some interventions were delivered as individual sessions (20,27,33,37) , two were delivered as group sessions (19,30) , and three were delivered as both individual and group sessions (24,35,36) .Some of the interventions were delivered in general practitioner practices in the community (20) or in local gyms (24) .Only four studies provided weight data after 12 months of follow-up (19,20,24,37) .Dropout rates, where clearly provided, ranged from 45% (24) to 78.3% (35) over the first 12 months.Some interventions included an initial period with LED, and a follow-up period with psychological and dietetic support. (18,25,28)The number of contacts followed a similar pattern: intensive initial care (approximately the first 3 months) and then fortnightly or monthly meetings, comprising five to 15 contacts in the first 12 months.
Overall weight, percentage of weight and BMI change are presented in Table 2.
Rolland et al. (33) implemented a RCT.Patients initially underwent a dietary treatment with a low-fat, 600 kcal day À1 deficit diet for 3 months.If patients responded well to this method, it was continued for the next 9 months.If patients failed to lose weight, they were randomised either to LighterLife VLED (550 kcal day À1 ) plus a weekly group support activity or a low carbohydrate/high protein (800-1500 kcal day À1 ) diet for the next 9 months with six contacts over 9 months.After 12 months, participants who responded well to the initial low fat, 600 kcal day À1 deficit diet (and were not randomised), had the highest weight change of all participants within this trial [À17.5 (6.4) kg] and across the other studies set in secondary care clinics, although the dropout rate was unclear for this group.12-month weight loss in the VLED group was 16.1 (19.0) kg compared to 3.0 (6.7) kg for the low carbohydrate high protein diet.Dropout rates were also unclear for these groups.
Across other studies that included a LED or VLED, weight loss varied from 5.1 kg (30) to 13.4 kg (19) after 12 months; however, the dropout rates were either unclear or over 69%.

Commercial setting
Only one study was conducted outside the NHS setting.Rolland et al. (34) retrospectively assessed the effect of LighterLife Total VLED with group-based behaviour therapy for self-referred participants who completed 1 year of treatment.The initial weight loss phase could vary from weeks to several months, continued by weekly group meetings.The mean (SD) weight change from baseline was À12.9 (11.3) kg at 36 months, presumed for completers; dropout rates were unclear.Over 50% of participants returned to the weight loss phase for a second attempt during the 36-month period (Table 3).

Discussion
We attempted to comprehensively review studies relevant to Tier 3 WMSs for adults with higher BMIs.One previous systematic review of Tier 3 weight loss services for adults by Brown et al. (38) included 14 studies with wider BMIs and shorter follow-up.Our focus was somewhat different, looking at longer-term outcome data from services relevant to adults with a BMI ≥35 kg m À2 .The distinction between Tier 2 and Tier 3 services appears to be blurred.Two specialist weight management services (27,35) explained that participants needed to undertake a programme similar to Tier 2 services before entering their Tier 3 programme.Primary care services offered programmes to participants whose mean was BMI ≥35 kg m À2 with a range of comorbidities; these programmes were difficult to distinguish from those for participants in secondary care specialist weight management services in the studies reported here.
Only 35% of our included studies reported data beyond 12 months; the absence of long-term data in the remaining studies is problematic with repect to evaluating the long-term effectiveness of these interventions.Limitations or differences in evaluation and reporting, as well as differences between participant groups in terms of comorbidities and psychological characteristics, made comparisons and inferences between studies and interventions challenging, and precluded meta-analysis.There is a need to improve data collection data in these interventions.Long-term data collection has been a challenge, in  Across studies, LEDs were associated with the greatest weight losses; for example, a mean weight change of -12.4 kg at 12 months in the study by Lean et al. (25) , with a reported dropout rate of 25.3% (25) , as well as similar results in the study by Astbury et al (18) .Dropout rates tended to be lower with LEDs, which could suggest that better weight loss with these diets provided participants with more motivation to continue in the weight management programme.Unclear denominators in studies with the LighterLife VLED do not allow comparisons with other VLED (19,33) studies.Only one trial (26) described expressly considering participants' choices or motivations for improving engagement with starting or continuing services.By contrast, one study (24) reported excluding participants 'by their lack of motivation'.Motivation (or lack of it) is sometimes assessed before participants are included in services, and so it would be helpful for authors to be explicit about this assessment and the referral pathway.Changing dietary advice according to how the weight of participants responds to different dietary interventions also appears to be beneficial for weight loss (33) .
Socio-demographic characteristics were often not reported and few studies appeared to include hard to reach or disadvantaged groups (e.g.ethnic groups, people with disabilities, younger or older people) or participants with a BMI >40 kg m À2 .
All studies included both men and women, except for two women-only studies (30,31) .Overall, more women (76.1%) were recruited than men in the remaining studies.Evidence was insufficient to assess whether specific services for men or women would be more effective.One study, which was not included in this review, reported the results obtained in a community intervention delivered in football clubs to men with mean BMI of 35 kg m À2 (39) .Exceptionally, this trial showed little evidence of weight regain by 12 months; weight loss 5.6 (8.1) kg, 11.0% dropout at 12 months.The results of this study indicate that WMSs that are tailored for men could be particularly effective.Few interventions reported considering 'emergency plans' or contact after the intervention, if needed.
Dietary and physical activity interventions were poorly described, making programme reproduction difficult.One study (19) and one trial (29) did report participants' weight loss history (including number of past weight loss attempts, methods used, average weight lost).Some studies excluded participants with eating disorders (31,33,34) .In one trial, participants were able to choose their diets (26) .Important features of the diets (e.g.availability; affordability; preferences; behavioural, social and economic costs for participants) were not described.These factors could impact on intervention effectiveness and adherence.Similarly, the extent to which diets were tailored may influence not only their success, but also their ease of delivery.
One study (24) and one trial (26) provided information on physical activity advice provided to participants; however, in most cases, details of physical activity advice were either poorly reported or not reported at all.One trial excluded participants with inability to walk more than 100 m (26) .Others included participants with arthritis (20,37) or joint pain (36,37) , factors to consider when recommending physical activity.
Scaling up interventions to reach more participants is important, particularly from an NHS perspective.Little et al. (26) showed that remote delivery produced much the same 12 month weight change compared to faceto-face delivery with a dropout rate of under 20% (mean À3.2 kg and À3.8 kg, respectively, for completers).This is comparable to the 12-month weight loss in the Counterweight evaluations (21,22) , which had dropout rates of 54.8% to 72%, although these are smaller weight losses than those reported in UK RCTs of commercial WMSs in primary care, with dropout rates from 11% to 29.5% (38,40) .Similarly, given that primary care referral to a commercial provider for participants of mean BMI 34.6 kg m À2 (in a RCT excluded from our review) demonstrated a weight loss of 4.9% from 12 weeks of programme at 12 months (100% of participants) and 7.1% from 52 weeks of programme (data for all participants), the role of commercial providers for people with higher BMIs could be explored further (41) .A comparison of Tier 3 services with commercial WMSs would be of value, considering the possible methodological challenges that this might comprise (particularly data collection and drop-out rates).Long-term UK data are urgently needed for participants with severe obesity (e.g.LighterLife, Cambridge Weight Plan, Weight Watchers, Slimming World, Counterweight Ltd) with weight outcomes taking account of dropouts.Randomised evaluations of comparisons of different approaches, including existing Tier 3 specialist WMSs, or allowance for the choice of reducing diet, would be valuable.
None of the included studies reported adapting the intervention to the needs of participants.Interventions appear to have been designed according to the resource availability or capability of the weight management system.For example, none of the studies reported attending participants out of the practice's regular attendance hours (e.g.evening or weekends), to facilitate participation.
There is a clear need for guidance on weight data collection and reporting to allow comparisons across studies and services.It was difficult to make comparisons between services, particularly when data were not provided for all participants (e.g. by last observation carried forward or baseline observation carried forward, which correct for differences in dropout rates).Services should be funded to collect data for longer than 1 year, preferably for 5 years.Public Health England has guidance for the evaluation of weight loss services (42) and a core outcome set has been developed in the UK using consensus methods, including advice on weight change data collection and statistical analysis (43,44) .Data should include quality of life, clinical outcomes, adverse events, costs and economic outcomes in a standard format.More detailed guidance on the content of reported WMSs would be very valuable, aiding with replication and evaluation.
In summary, our searches identified 20 studies, which were mostly noncomparative.A programme including a phase of low energy formula diet low energy diet showed the largest mean weight change at 12 months of À12.4 (11.4) kg with 25.3% dropout.Differences in evaluation and reporting (particularly for denominators), unclear dropout rates, and differences between participant groups in terms of comorbidities and psychological characteristics, make comparisons between different programmes very challenging.There is a persistent and clear need for guidance on long-term weight data collection and reporting methods to allow comparisons across studies and services for participants with severe obesity.

Transparency declaration
The lead author affirms that this manuscript is an honest, accurate and transparent account of the study being reported.The reporting of this work is compliant with PRISMA guidelines.The lead author affirms that no important aspects of the study have been omitted and that any discrepancies from the study as planned (protocol PROSPERO No CRD42016040190).This project is part of the National Institute for Health Research funded REview of Behaviour And Lifestyle interventions for severe obesity: AN evidenCE synthesis (REBALANCE) (10) project.

Figure 2
Figure 2 Risk of bias of randomised controlled trials (RCTs).

Figure 3
Figure 3 Risk of bias of nonrandomised comparative studies and case series.

Table 1
kg imputed data] with a dropout rate of 44.2%.Weight change, percentage weight change and body mass index (BMI) change in National Health Service (NHS) primary care interventions PROGRESS= Place of residence; Race/ethnicity/culture/language; Occupation; Gender/sex; Religion; Education; Socioeconomic status; Social capital

Table 2
Overall weight change, percentage weight change and body mass index (BMI) change in specialist weight management clinics

Table 3
Overall weight change, percentage weight change and body mass index (BMI) change in commercial setting, presumed data for completers Inclusion criteria: Age ≥18 years and BMI of ≥30 kg m À2 or ≥ 28 kg m À2 with associated comorbidities Exclusion criteria: BMI of >45 kg m À2 , had lost > 5% of weight in the previous 6 months,