Randomised controlled trial and economic analysis of an internet-based weight management programme: POWeR+ (Positive Online Weight Reduction)

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 09/127/19. The contractual start date was in January 2012. The draft report began editorial review in January 2016 and was accepted for publication in August 2016. 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 reviewers 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

Paul Little is Editor-in-Chief of the Programme Grants for Applied Research journal and a member of the National Institute for Health Research (NIHR) Journals Library Board. James Raftery is a member of the Health Technology Assessment (HTA) and Efficacy and Mechanism Evaluation editorial boards and a member of the NIHR Journals Library Editorial Group and was previously Director of the Wessex Institute and Head of NETSCC. Barrie Margetts states that since retiring from the University of Southampton (in September 2014) he has undertaken paid consultancies for the World Bank, UNICEF and the World Health Organization. Up until December 2015 he was an unpaid trustee of the international charity ‘Riders for Health’ and up until May 2015 he was unpaid president of the World Public Health Nutrition Association. Lucy Yardley is a member of the NIHR Public Health Research Funding Board and the HTA Efficient Study Designs Board.

Copyright statement

© Queen’s Printer and Controller of HMSO 2017. This work was produced by Little et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) 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: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

Scientific background

Obesity is a major threat to public health,1,2 the prevalence has risen sharply3 and the vast majority of obese patients are managed in primary care. 4 For some time systematic reviews,4 including those from the National Institute for Health and Care Excellence (NICE),5 have advocated a dietary and physical activity intervention supported by intensive behavioural techniques. The NICE economic model6 estimated that if trial results from other settings could be extrapolated to the skills mix, resource availability and setting of primary care, then intervention is likely to be cost-effective. The trials identified in the NICE review5 mostly had expert lifestyle and behavioural input, and intensively followed up patients (on average 13 times per year during the first 12 months). 5

An average practice with three general practitioners (GPs) and 5500 patients will have > 1300 obese patients. There is a demand from obese patients for primary care to provide longitudinal programmes of personalised support, but dropout from intensive face-to-face programmes is typically high and most practice staff have neither the training nor the time to implement intensive obesity management programmes based on one-to-one counselling, or even group counselling, to cope with such numbers. 7 The problem will become worse as the obesity epidemic progresses. Therefore, it is not feasible to apply current NICE recommendations in primary care because the high-level dietetic expertise required is not available and it is impossible to free up enough staff time for counselling and follow-up, particularly given the progressively resource-constrained environment in primary care. A research priority identified by NICE in 20065 was to develop and assess interventions in primary care, and also to assess the scope for less resource-intensive interventions in a primary care setting.

A review of studies in primary care8 using strict criteria for inclusion (e.g. excluding those with > 30% attrition at 12 months, which is common in studies of obesity and so may not be an appropriate exclusion) also concluded that there was very little evidence of studies using appropriately intensive behavioural counselling in primary care and suggested the need for using trained interventionists.

Rationale for this research

An alternative, or addition, to a cadre of highly trained interventionists is to harness the capacity of the internet to help support behaviour change.

A systematic review of internet-based behavioural health-related interventions9 concluded that web-based interventions were effective for supporting behaviour change, although with considerable heterogeneity, and that interventions employing multiple theory-based techniques achieved the best results.

Heterogeneity in the effectiveness of internet interventions may be, in part, due to the level of support, with a trend towards better outcomes in interventions with additional personal support,10,11 so it remains very unclear how much facilitation is needed for effective weight reduction, whether face-to-face support is necessary and/or whether briefer remote support could suffice. It is also unclear whether such interventions work better than using novel but simple written materials that do not emphasise cutting out or cutting down favourite foods, but instead promote healthy food swaps and increased fruit and vegetable consumption, which we have shown aid modest weight reduction. 12

Development of the intervention

The POWeR+ web-based intervention

POWeR+ is a theory- and evidence-based intervention that is designed to teach patients self-regulation and cognitive–behavioural techniques and that aims to help them to form sustainable eating and physical activity habits for long-term weight management. POWeR+ was developed, as with POWeR, using the person-based approach to maximise acceptability, feasibility and engagement. 22–25 We finally created a series of 24 web-based sessions lasting up to 6 months with novel content, links to external content and e-mail reminders to encourage patients to continue to use the website weekly to track their weight, set and review eating and physical activity goals, and receive personalised advice. After entering their weight and whether or not they had achieved the goals they had set themselves the previous week, patients receive tailored feedback giving encouragement if maintaining weight loss (e.g. reminders of health benefits accrued) and meeting goals. Weight gain and failing to meet goals triggers automated personalised advice, such as appropriate goal-setting and planning, boosting motivation, overcoming difficulties and recovering from lapses (see Appendix 1).

Having developed the internet-based intervention, we chose to trial it in comparison with an active control group – a brief approach using brief written materials to support food swaps and increased fruit and vegetable consumption. We have previously shown this to control weight well, achieving in the order of a 2% reduction in weight compared with providing a simple advice booklet. 12

Objectives

Compared with a brief intervention promoting brief written materials to support food swaps and increased fruit and vegetable consumption, we report the incremental effectiveness of (1) an internet-based behavioural intervention (POWeR+) with face-to-face support as needed (POWeR+F) and (2) an internet behavioural intervention with remote support (POWeR+R).

Study design

This was an individually randomised parallel-group study.

The control group received a brief but active intervention; the rationale was to minimise the pressure to cut down favourite foods, but instead to swap less healthy foods for healthier choices (the swap sheet) or to increase healthy foods (fruit and vegetables sheet). It was chosen for its simplicity to provide brief advice, which we have previously shown to control weight well. 12

We chose an individually randomised design because, with structured advice in each group, we have been able to minimise contamination in previous trials and an individually randomised design has the additional advantage of minimising cluster effects.

This was a very pragmatic trial to provide the most realistic estimates of effectiveness, so participants were not constrained from pursuing other activities in any study group (e.g. attending commercial slimming clubs).

Trial registration and ethics

The trial was registered on 16 March 2012 [International Standard Randomised Controlled Trial Number (ISRCTN) 21244703] and ethics approval was given by the National Research Ethics Service Committee South Central Southampton B First Multicentre Research Ethics Committee (approved 19 December 2011; reference number 11/SC/0455).

Participant inclusion and exclusion criteria

Patients with a BMI of ≥ 30 kg/m² (or ≥ 28 kg/m² with additional risk factors of hypertension, diabetes mellitus or hypercholesterolaemia) documented in GP case records4 were eligible.

Patients were excluded if they had current major mental problems, such as psychosis, or were very ill (e.g. severe left ventricular failure), that is, they had difficulty completing outcomes, were unable to change diet, were pregnant or breastfeeding, or had a perceived inability to walk 100 m (physical activity difficult).

Recruitment: invitation of patients/recruitment

General practices in the south of England (recruited around the centres of Southampton and Oxford) identified participants from their electronic records. Up to 100 patients from each practice were randomly chosen and invited, by letter, to a screening appointment to confirm eligibility. Patients could also be referred opportunistically when seen in clinics run by practice nurses or doctors (Figure 1).

FIGURE 1.

The CONsolidated Standards Of Reporting Trials (CONSORT) flow diagram: flow of participants in the trial. This figure has been reproduced from Little et al. 26 under the terms of the CC BY 4.0 licence (https://creativecommons.org/licenses/by/4.0/).

Informed consent

During the screening appointment, the study was explained and informed written consent was taken. Participants were given details of how to log in and register with the LifeGuide website (www.lifeguideonline.org/), POWeR+.

Randomisation/group allocation

On registering with the study website, participants were presented with baseline questionnaires, on completion of which they were automatically randomised using computer-based random numbers by the website to one of three feasible interventions in primary care:

1. Control group: evidence-based advice and simple materials to support behaviour change, and follow-up. Those randomised to the control group were taken to pages of the POWeR+ website containing two brief, structured, printable pages of advice about a healthy diet (healthy food swaps and NHS 5-a-day leaflet27). These materials had previously been developed by the Institute of Food Research, and had been trialled by us in primary care, resulting in modest weight loss (around 2%) compared with a generic advice booklet. 12 To enhance retention in the control group, participants were informed that this intervention had been shown to support weight loss. For follow-up, nurses arranged brief 5- to 10-minute appointments with sufficient time to measure weight at 6 months and 12 months, but not to provide extensive counselling.

2. POWeR+F: web intervention with face-to-face appointments for nurse support. The rationale for this intervention was to provide automated behavioural counselling with just three scheduled (and four optional) face-to-face nurse support sessions, thus requiring substantially less health professional skill and time than the evidence-based lifestyle interventions documented in the NICE review,5 and, hence, much easier to implement in the NHS. In addition to 6-monthly weighing, as in the control group, participants had three scheduled face-to-face appointments in the first 3 months and then up to four more during the next 3 months, if needed. Weight gain on two consecutive logins triggered an automated e-mail to the nurse advising that the patient required further support. Patients could also request additional support.

3. POWeR+R: web intervention with remote support. The rationale here was to test whether or not even briefer professional support for the web intervention could be effective. Patients could access the same web-based intervention as in the face-to-face group. In addition to 6-monthly weighing, as in the control group, participants had three scheduled telephone or e-mail contacts and up to two optional telephone/e-mail contacts during the first 6 months (triggered by weight gain or patient request as in the face-to-face group). This level of support was confirmed as acceptable and helpful during the development and piloting stages.

Data collection and management

Measures

Weight reduction averaged over a 12-month period, measured lightly clothed, without shoes, where possible at the same time each day, and using automated Tanita digital scales (Tanita Corporation of America Inc., Arlington Heights, IL, USA), was the primary outcome.

A secondary weight outcome was the proportion of participants maintaining a 5% weight loss or loss of around 4–5 kg . This is very likely to be important clinically (a reduction of 3–5 kg has a significant impact on the incidence of diabetes mellitus28,29) and it allows direct comparison with a previous UK trial from primary care settings,30 which published its results after our study commenced.

Sample size

We wished to compare each of the intervention groups primarily with the control group, but also potentially with each other, so we allowed for an alpha of 0.017 (i.e. 0.05/3). For the primary outcome (weight) we estimated that a standardised effect size of 0.33 [equivalent to a 2–3 kg difference, assuming a standard deviation (SD) of change of 6.5–7.5 kg36,39] and 80% power required 174 patients per group with complete data, or 654 patients in total allowing for a 20% loss to follow-up. Following liaison with both the funder (NIHR’s Health Technology Assessment programme) and the Trial Steering Committee, the power calculation was revised to allow for modest clustering at practice level if significant clustering was found; we assumed, on average, a minimum of 18 patients per practice to achieve 15 patients at follow-up, of whom roughly five or six were in each of the two intervention groups at follow-up, assuming five participants per group in each practice, for an intracluster correlation coefficient of 0.05 {i.e. a design effect of 1.2 = 1 + [(5 – 1) × 0.05]}. This resulted in a minimum of 654 × 1.2 = 785 patients. This was a conservative calculation because we made no allowance for repeated measures.

Statistical analysis

Patient and public involvement

Weight Concern was a partner in this project, and the study team requested named individuals to be part of the team. Weight Concern preferred, instead, to provide a panel-based input, arguing cogently that a more balanced input would be provided. The input was indeed very helpful, particularly during the development of the intervention. Nevertheless, in retrospect, patient and public involvement input would still have been strengthened by engaging named individuals to contribute, perhaps sourced through additional channels.

Subgroup analyses

We explored estimates of effect in subgroups according to baseline waist measurement (high vs. low waist circumference) and the presence of metabolic syndrome. We assessed outcome (weight, reported behaviour change, measured behaviour change) according to website usage and to the type of diet chosen by patients.

Other analyses

We assessed the relationship between weight changes compared with the changes in metabolic variables.

Changes to the protocol

Based on the advice of the Trial Steering Committee we increased the sample size to allow for clustering, as necessary. A repeated measures analysis of variance for the principal continuous outcome (i.e. weight) was originally proposed but could be undertaken only in those who completed both the 6- and 12-month follow-up, thus unnecessarily losing data; therefore, the analysis was changed to mixed multivariable regression modelling. We originally planned for one-third of the sample to be randomised to wear activity monitors to allow for more precise estimation of physical activity, but the logistics of organising this subsample at a time of intensive attempts to achieve high follow-up for the primary outcome became too difficult. Similarly, piloting suggested that intensive follow-up for anything but the primary outcome would result in participants dropping out, so all the effort of follow-up concentrated on achieving as high a rate of follow-up for the primary outcome as possible. Although the original protocol justified the sample size based on a clinically important weight reduction, the proportion maintaining a clinically important weight loss was not originally specified. We rectified this omission by specifying a secondary weight outcome as the proportion maintaining a 5% weight loss, or around 4- to 5-kg weight loss. This is important clinically (a reduction of 3–5 kg significantly reduces the incidence of diabetes mellitus28,29) and it allows a direct comparison with the previous UK trial from primary care settings,30 which published its results after our study commenced.

Compliance with intervention

Compliance with the use of the POWeR+ intervention using the LifeGuide software version 1.0.7.30 (University of Southampton, Southampton, UK) was documented automatically by the website, which recorded the number of sessions completed and the time taken.

Practice recruitment

Fifty-six general practices were recruited in the south of England around the main study centres (Southampton and Oxford).

Participant recruitment

A total of 818 eligible individuals were recruited and randomised between January 2013 and March 2014.

Baseline comparability

Table 1 shows the baseline characteristics of the groups and shows that groups are mostly well balanced.

Primary outcome

Secondary outcomes

A range of secondary outcomes is shown in Tables 11–22. Owing to the priority to obtain follow-up for weights, and the clear feedback from piloting that pressurising participants to have blood taken was offputting, achieving high follow-up for blood samples was necessarily a secondary priority for the trial team and so a minority of participants had follow-up blood measurements. As a result, even though estimates use multiple imputation, the results for blood samples must be interpreted cautiously. What results are available suggest a generally positive direction of outcomes in the POWeR+ groups (i.e. raised high-density lipoprotein cholesterol, lower aspartate transaminase and alanine transaminase levels).

Fat mass and BP were recorded in the majority of participants at follow-up (in both cases > 450 individuals) and, although there were no consistent changes in BP, fat mass reduced slightly in both POWeR+ groups (see Tables 18 and 19). Participants in the control group reported healthy changes in many areas of food consumption – with modest changes in sweets, fatty foods, and fruit and vegetables, and in the ratio of high- to low-fat foods. In comparison with the control group, in the POWeR+ groups there were no significant changes but there were trends in the reduction in sweets and cereals, and increasing low-fat dairy foods. Participants felt significantly more enabled to manage their weight problem in the POWeR+ groups (see Table 22).

Subgroups

Methods

The costing comprised two elements: one measuring the intervention cost alone, based on the cost of internet plus that of nurse support; and the other measuring total cost, that is, the cost of the intervention plus differences in the cost of NHS services. The perspective was that of the NHS and Personal Social Services. The time horizon was 12 months. The implications of different cost scenarios were explored.

The main cost of the intervention was the nurse support time comprising face-to-face consultations, telephone and e-mail contact. The cost of nurse support was based on trial data that documented the number of contacts by type (face to face and by telephone and e-mail, all with the practice nurse). Face-to-face contacts linked to the trial interventions were planned to be brief and last half as long as normal visits; hence, these were assumed to cost 50% of standard practice nurse consultations. The time taken for telephone contacts relative to face-to-face visits was based on the split for GPs between face-to-face and other contacts provided in Personal Social Services Research Unit (PSSRU) unit costs. 41 Telephone contacts were shorter than most telephone consultations as no new diagnosis and management plan was being instituted; instead, patients were given brief support and signposted back to the POWeR+ resources, as appropriate. The time taken for e-mail contacts was put at 50% of that of telephone contacts and, because e-mail contacts are normally quick, this may be an overestimate. These assumptions were tested in sensitivity analyses. Nurse costs were costed using national costs per consultation for a practice nurse. 41 The unit costs used in costing the intervention are shown in Table 24.

The cost of the intervention comprised the cost of the website and the support offered to users. The cost of providing a web intervention that could be used widely would be spread among many thousands of individuals. We therefore put the cost of the website at £1 per person, based on the cost of providing and maintaining the website, but this may overestimate its cost, were it to be made widely available.

The brief verbal intervention and supporting advice sheets in the control group were not costed separately, but were assumed to be included in primary care consultations. Resource use data were extracted from GP case notes 12 months after recruitment – covering medication, primary care visits, outpatient consultant, accident and emergency attendance and hospital admission. Intervention-specific resource use was recorded by study nurses.

The costing of medications was limited to those plausibly associated with obesity, that is, medications for diabetes mellitus, BP, clotting/antiplatelet, musculoskeletal problems (back, hip and knee pain), and lipid lowering and weight management. The names of these medications, dosage and days of use were recorded. Prices were the listed pack price, the cost to the NHS. For drugs started before the trial or for which no end date was recorded, we assumed the duration of use to be that of the trial (12 months). For all other obesity-related medications, if no duration of use data were available, we assumed the recommended duration from the starting date. The unit costs of medications were those in the British National Formulary in 2013–14 prices. 43

Consultations in primary care, walk-in centres, and accident and emergency departments and outpatient attendances were recorded and priced using PSSRU 2013/14 unit costs41 (see Table 24).

Data on all hospital admissions were also extracted. We used Healthcare Resource Group cost per episode linked to cause of admission. 43 Data on these unit costs and those for medications are available on request.

The outcomes were weight lost and QALYs. Data on weight were collected at 6 and 12 months, when patients were weighed by the practice nurse. Health-related quality of life was measured by the EQ-5D, completed by patients at baseline and at 6 and 12 months. The EQ-5D scores were expressed as utilities using the national tariff, with interpolation using area under the curve.

Missing data were imputed using multiple imputation, as in the clinical analyses in Chapter 3. Bootstrapping was applied to the imputed data to estimate costs, weights, QALYs and incremental cost-effectiveness ratios (ICERs). Weights were analysed by repeated measure analyses. Estimates of the differences in QALYs were adjusted by baseline EQ-5D scores. Pooled results of the bootstrapped analyses were used to generate the 95% CIs around the estimates. In line with the approach that matches the effectiveness analysis (where a reduction is indicated by a negative value), where costs are reduced we have also indicated this with a negative value (in particular, the negative values for ICERs indicate that there was an estimated reduction in costs for every kilogram of weight lost).

The base case for the cost per weight lost analysis was that for the imputed case based on repeated measures analyses, linked to the statistical analysis report in Chapter 3, chosen in view of the low completion of the EQ-5D (< 50%). We also present the cost per QALY analysis based on imputed cases, but given the extent of missing data caution should be used in interpreting such results.

Sensitivity analyses were conducted on complete cases, the cost per percentage achieving weight loss of > 5% from baseline and without hospital admission costs.

Results

The pattern of use of the nurse support in the two intervention groups is shown in Table 25. Approximately 70% of those in each intervention group had at least one contact [POWeR+F, 72% (193/269); POWeR+R, 67% (181/270)]. Contacts could be face-to-face, e-mail or telephone contacts. As expected, the mean face-to-face contacts were highest in the POWeR+F group at 1.50 per patient, but a few also occurred in the POWeR+R group (mean 0.10). The mean total number of contacts was slightly higher in the POWeR+F group, at 3.23, than in the POWeR+R group at 2.85.

Cost-effectiveness acceptability curves

The cost-effectiveness acceptability curves (Figures 2 and 3) for weight loss show the probability of each intervention being cost-effective compared with the control group at various of levels of willingness to pay per kilogram lost, including NICE’s suggested willingness-to-pay threshold of £100 per kilogram lost. Compared with the control group, this puts the probability of being cost-effective at 88% and a 98% for the POWeR+F and the POWeR+R groups, respectively.

FIGURE 2.

Cost-effectiveness acceptability curve of POWeR+F compared with control group based on average weight loss (kg) from baseline during 12 months.

FIGURE 3.

Cost-effectiveness acceptability curve of POWeR+R compared with control group based on average weight loss (kg) from baseline during 12 months.

The incremental cost per QALY analysis is complicated by the negative QALY increments for both interventions versus control, which is unreliable as a result of missing data. For POWeR+F, the control group was dominant, as it also cost less. For POWeR+R, which cost less than the control group, the ICER was positive as a result of the division of two negative values. Given the uncertainty of these results, with half of the EQ-5D values missing, cost-effectiveness acceptability curves are not reported.

Conclusions

Overall, both interventions were cost-effective in terms of weight loss, but less so in terms of incremental cost per QALY. This was the case for our base-case analyses of cost per kilogram lost and per QALY, and showed little variation in other analyses. The former result is more robust for several reasons, including the relationship between QALYs and weight loss and the level of missing EQ-5D data. The cost per kilogram lost is highly likely to be below NICE’s threshold of £100 per kilogram lost, but this conclusion is limited by our lack of data on the maintenance of weight loss beyond 12 months.

Design

We used semistructured qualitative interviews to gain a rich, in-depth understanding of HCPs’ experiences of supporting patients using POWeR+.

POWeR+

POWeR+ consists of two parts: one for patients and one for the HCPs who supported both intervention groups. The development of the patient website is explained more fully elsewhere21,25,46 but, in brief, it provided patients with a choice of two eating plans (low calorie or low carbohydrate) and two physical activity plans (a walking plan, with free pedometer, or a plan for any other physical activity). POWeR+ aimed to help patients to develop self-regulation skills by promoting weekly weighing, weekly reviews of eating and physical activity goals, as well as by providing cognitive–behavioural techniques in weekly online sessions.

The HCP website provided brief information about POWeR+ and how to provide support to patients (this information was also given on paper in HCP study files). The website also allowed HCPs to view patients’ recorded weekly weight and goals, and to send patients support e-mails.

In the first version of POWeR22 we gave HCPs access to detailed information about the patient website, hoping that practitioners would therefore be able to give advice that was consistent with POWeR. However, HCPs reported that they lacked the time to look at this information and usage analysis revealed that very few had looked at these support pages. Taking a person-based approach,21 we redesigned the HCP support for POWeR+ based on practitioners’ feedback. In POWeR+, practitioners were asked not to give advice to patients; instead, all advice came from POWeR+ and the practitioners simply provided a supportive relationship to promote adherence.

We developed the Congratulate, Ask, Remind (CARe) approach to facilitate a non-directive supportive relationship, which would be easy to deliver and would fit with practitioners’ busy schedules. In addition to being developed using a person-based approach, the CARe approach is also based on self-determination theory. 47 The CARe approach aims to provide an autonomy-supportive relationship, which can raise patients’ autonomous motivation for behaviour change by promoting feelings of autonomy, competence (feeling effective) and relatedness (understood and cared for by others). 48 Autonomy-supportive relationships with HCPs predict better weight-loss outcomes,45,48 as well as predicting outcomes in a range of other health conditions. 49 The theorised mechanisms of the CARe approach are explained in Table 35.

Sampling and recruitment

The 54 HCPs who supported POWeR+ were nurses (n = 53) or health-care assistants (n = 1), and the majority were female (n = 53). All HCPs were sent a study invitation letter, information sheet and consent form. Nineteen female nurses expressed an interest in taking part and 13 were interviewed by telephone. No significant new themes emerged with later interviews, implying that saturation had been achieved. 53 Practitioners who chose not to participate noted that they did not have time.

Data collection

The telephone interviews were conducted between April and June 2014 by LP, a health psychology Master of Science student who was given training in qualitative interviewing and analysis. LP had no prior relationship with any of the HCPs interviewed. HCPs who agreed to participate posted a signed consent form to the researcher prior to the interview.

Before each interview, LP explained that she was a postgraduate student, with no previous involvement in the design or evaluation of POWeR+. The interviews explored HCPs’ experiences of providing support for POWeR+ participants, their experiences of using the POWeR+ website and of the study procedures. All interviews were audio-recorded and lasted between 23 and 46 minutes.

Analysis

All interviews were transcribed verbatim and imported into NVivo 10 (QSR International, Warrington, UK), to allow systematic comparisons to be made across the data set. An inductive thematic analysis54 was carried out, augmented with procedures from grounded theory. 55 First, the researchers familiarised themselves with the data. The interviews were then coded and a coding manual was created. This coding manual was continually updated to reflect the ongoing analysis. Constant comparison was used to ensure that codes were being used consistently and reflected the data. 55 Codes that identified similar aspects of the data were clustered together into themes. Inter-rater agreement on all the final codes and themes was agreed with ES (a post-doctoral researcher), LY (a health psychologist) and LP. Deviant case analysis was used to ensure that perspectives that diverged from dominant trends were not overlooked.

Results

Four themes were identified: HCPs’ perceptions and use of POWeR+, supporting patients in their use of POWeR+, the impact of POWeR+ and comparisons to existing weight management services. These are discussed in detail below.

Methods

Results

Fifteen (48%) women and 16 (52%) men aged between 45 and 88 years (mean 61 years) took part in individual telephone interviews. Full demographic details are shown in Table 37.

Four themes were identified: reactions to the POWeR+ website content and function, changes to thinking and lifestyle, motivators of and barriers to change, and using POWeR+ with and without support (see Table 36).

Summary of main findings

As far as we are aware, this is one of the few studies in primary care that has compared three feasible briefer behavioural interventions using primary care staff for support to manage obesity. It demonstrates that some participants managed in primary care can achieve clinically important weight loss over a 12-month period by combining simple written materials with occasional brief nurse follow-up, but that significantly more will maintain clinically important weight reduction with the POWeR+ behavioural internet programme and brief follow-up.

Strengths and limitations

The individuals who take part in trials are likely to be a relatively well-motivated group, but this is no different from other trials in obese populations, and this is also the intended target group for which intervention is most likely to be helpful. The study was large and pragmatic, mimicking the everyday conditions in primary care settings, which had the disadvantage that the control group were not closely controlled and so they undertook other activities to lose weight. This may have reduced the estimates of benefit from the internet intervention, but this is also a strength as estimates are more realistic – this is precisely what would happen in practice. The data also suggest that when controlling for such other activities there was a modest change to the estimates.

Participants with obesity in primary care settings are notoriously difficult to follow up, and we were not able to record weights in nearly 20% of individuals at 12 months. The estimates from the complete-case analysis, and the analysis using recorded weights only, are therefore both likely to provide overestimates of effectiveness; hence our primary analysis was the more conservative analysis using the imputed data. However, in practice, multiple imputation modified the estimates only slightly, which suggests that attrition bias was not a major issue. The qualitative study of health professionals used a rigorous approach to thematic analysis, involving constant comparison, deviant case analysis and multiple analysts. Another strength was that HCPs’ interview data could be triangulated with patient interviews, as well as engagement and weight-loss outcomes from the POWeR+ trial, providing a fuller understanding of the data. A weakness of this study was that only 19 practitioners volunteered to take part in this study, out of the 54 who had taken part in the original trial; the views of those who took part may not be the same as those who did not and could either over-represent or under-represent positive experiences of POWeR+. The larger qualitative study of participants may also not have been representative of the whole cohort, but in both qualitative studies there was a suitable range of participant characteristics and saturation of themes was reached, so it seems unlikely that major issues were missed. A significant limitation is that, although the intervention finished at 6 months and we were able to assess maintenance at 12 months, there was no longer-term follow-up beyond 1 year. This is relevant not only for clinical effectiveness, but also because the NICE modelling suggests that cost-effectiveness is sensitive to the duration of clinical effectiveness. The major concentration on weight outcomes meant that there was much less good coverage of the other outcomes (e.g. cholesterol, glucose, etc.).

A strength of the study was that Weight Concern was a partner in this project and chose to provide very helpful panel-based input, particularly during the development of the intervention. Although this was well justified, the project would have been strengthened by seeking additional patient and public involvement engagement from other named individuals, perhaps sourced through additional channels.

Effectiveness: main findings in the context of existing literature

We confirmed our previous findings12 that a brief intervention, using the novel approach of promoting the use of healthy food-swap sheets and with brief follow-up for weighing, can be useful. In this study, this strategy helped around 20% of participants to achieve a clinically important loss in weight. Many patients in primary care do not get such simple, positive interventions, so we have probably underestimated the effectiveness of POWeR+ in routine practice. As discussed above, the effectiveness in the control group may also be, in part, due to motivated individuals who want to lose weight, and will do other activities to help themselves (which about 50% did). In contrast, fewer individuals offered the POWeR+ intervention undertook other activities.

The systematic reviews of interventions both in other settings and primary care suggest that both intensive dietetic behavioural counselling and intensive follow-up are necessary to achieve effective weight reduction. 5,8,11 In contrast, we found that for a behavioural internet intervention, such as POWeR+, very restricted face-to-face and remote follow-up is effective (median of four contacts, comprising 6-monthly weighing, brief telephone calls and e-mails). The weight loss achieved with POWeR+ compares favourably with other internet-based interventions, which on average10 (albeit with high heterogeneity) have led to only short-term weight loss, of < 1 kg compared with no treatment controls or < 2 kg if combined with face-to-face support (in a review based mainly on studies of motivated volunteer samples). Better weight management results using an internet-based programme (around 5 kg) have been achieved only with much more intensive human support (weekly 20-minute contact for 12 weeks, and then monthly up to 24-month follow-up). Encouragingly, the weight loss achieved by POWeR+ was comparable to the best-performing interventions10 evaluated in a primary care setting, including that achieved by face-to-face commercial programmes,30 but much less costly. The difference in mean weight loss of the intervention groups when compared with the control group at the 12-month follow-up was smaller than the difference compared with the control group at 6 months, so, although clinically important reductions in weight were maintained from 6 to 12 months, particularly in the POWeR+R group, maintenance beyond 1 year is unclear.

It is difficult to be sure that physical activity did not mediate intervention effectiveness because reported leisure time physical activity is a relatively crude measure of physical activity. Nevertheless, the biases are such that over-reporting of physical activity is more likely in the intervention groups, so it seems likely that physical activity changed little. There was also little increase in the use of drugs, such as orlistat, for managing weight. Thus, most of the effect of the intervention observed is probably a result of better dietary management, which suggests that more effective engagement of participants for both physical activity and appropriate drug treatment could result in significantly greater weight loss.

Health economic analyses

The total cost per kilogram for both interventions compared with control was low relative to NICE’s suggested benchmark of £100. It was also well below the cost per kilogram loss in a recent systematic review of commercial weight-loss programmes:53 among the five most effective interventions, average cost per kilogram of weight lost ranged from US$155 (95% CI US$110 to US$218) for Weight Watchers^® (Weight Watchers International, Inc., New York, NY, USA) to US$546 (95% CI US$390 to US$736) for orlistat. Economic analysis of the Counterweight Programme^® (Counterweight Ltd; www.counterweight.org/Programmes),58 in which the intervention cost £59 per patient and achieved 3 kg of weight loss, and which modelled only impacts on diabetes mellitus, cardiovascular disease and colon cancer, demonstrated that this level of weight reduction is likely to be very cost-effective, even allowing for weight gain following the end of the programme. The likelihood of both POWeR+F and POWeR+R groups being cost-effective at the NICE threshold was high, well over 80% (88% and 98%, respectively). A similar figure applied to the probability of POWeR+R being cost-effective at this threshold relative to POWeR+F.

This pattern of result is robust for several reasons: it held regardless of whether costing was restricted to the intervention cost or extended to total cost, the results are based on data that are largely complete and the directions of change were consistent at 6 and 12 months.

What is less expected was the finding that the POWeR+R group was more cost-effective than the POWeR+F group. This was because it achieved similar changes in weight loss (depending on the measure), but at lower cost. Its lower intervention cost was as expected, but the finding of a greater reduction in total cost was not expected. The difference in total costs was not, however, statistically significant, but the cost of the POWeR+R intervention was shown in sensitivity analysis to be highly likely to be less than that in POWeR+F. The difference is mainly because of the lower cost of e-mails compared with face-to-face contacts.

The cost per QALY results are more difficult to interpret, mainly because both intervention groups recorded small (not statistically significant) QALY losses relative to the control group. This is despite both losing more weight than the control group. The QALY losses combined with inevitably higher intervention costs in both intervention arms means that the control would dominate in an intervention cost analysis. When total costs were used, one intervention cost more, and the other cost less than the intervention.

These incremental cost per QALY results are highly uncertain for several reasons: the decreases in QALYs from baseline linked to greater weight loss may or may not be plausible; the high proportion of missing data precluded more detailed exploration of the POWeR+ trial results; and the incremental cost per QALY values were unstable, varying by type of analysis.

In summarising the economic data, both interventions deliver a low cost per kilogram lost, as judged by NICE’s suggested threshold of £100 per kilogram lost. This is the case regardless of whether intervention or total costs are used. The probability of each intervention being cost-effective compared with the control was > 80%, with a similar figure applying to POWeR+R being more cost-effective than POWeR+F. This result reflects the combination of statistically significant differences in weight loss but not in total cost. Incremental cost per QALY results also indicated both interventions were cost-effective by NICE standards, but the results were highly uncertain. Comparison of the two interventions generally showed an advantage for the POWeR+R group as a result of its lower cost, however defined.

Even if weight loss had been identical in the POWeR+ remote group at all stages when compared with the active control group, there was no evidence that health service costs increased. Furthermore, individuals also apparently gained utility in feeling more enabled to manage their weight, and they undertook fewer other weight-loss activities, suggesting that societal costs were probably lower in the POWeR+ remote group.

Qualitative health-care professional study

Very few studies have previously looked at HCPs’ perceptions of providing support for online weight-loss interventions. Two focus group studies have explored physicians’ perceptions of providing support for patients using an online intervention. 59,60 However, in both these studies the physicians provided infrequent support to patients (every 3–6 months), whereas a coach provided patients with weekly intensive support. Coaches were not interviewed in these studies, meaning that only the perceptions of physicians who provided relatively sparse support were explored. Still, these studies usefully show that some physicians perceive that they lack the skills to support patients losing weight59 and that others use dubious motivational strategies, such as expressing frustration to patients who are not making progress. 60 These findings suggest that a more structured approach to providing support, such as the CARe approach, may be a useful tool for practitioners wanting to support patients using online weight-loss interventions. The wider literature examining HCPs’ experiences of providing support for an online intervention in chronic health conditions tends to be of slightly lower quality, as a result of very small samples of practitioners (n = 4 or 5). This literature suggests that HCPs see a lack of time as a key challenge to providing support,61,62 which is consistent with the findings of our study and represents a challenge to implementing digital interventions into health-care settings.

Health-care professionals found it feasible and acceptable to support patients using POWeR+. As POWeR+ has been shown to be effective and can be provided at low cost, it may serve as a useful tool for the treatment of obesity in primary care and public health. Digital interventions are rising in popularity and additional human support is likely to maximise the effectiveness of these interventions. The CARe approach shows promise as a model to guide human support for digital weight-loss interventions, and may be a useful model to guide human support in digital interventions more broadly.

Qualitative study with patients

Overall, the POWeR+ website was viewed positively and as a trustworthy source of information relating to weight loss, particularly when compared with other sources of information. Confidence in the accuracy of the information was further enhanced by delivering the programme through participants’ usual GP practices. Participants appreciated the flexibility offered by an online intervention and were complimentary about the formatting and design of both the eating and exercise plans, although suggestions for improvement included the provision of structured meal plans and example recipes. Participants provided mixed views regarding the setting of short-term, realistic goals on a weekly basis, with some finding it beneficial to make and review their goals regularly, but others perceiving this to be an unnecessary inconvenience. Nurse support was also perceived positively by all participants, even those who reported feeling that the level of support received was low. Nurse records kept during the trial show that participants all received a regular amount of support in line with the trial guideline; however, there was an indication that participants did not always perceive remote support, particularly by e-mail, as a form of support. The findings suggest that participants would prefer support to include the chance to speak with a nurse, or at the very least to know that someone was available for them to talk to as required.

Interpretation of results

Some participants in primary care maintain clinically important weight loss over a 12-month period by promoting simple written materials without expert dietetic support, but significantly more will achieve clinically important weight reduction with a behavioural internet programme and brief remote follow-up, and resource use is not likely to increase. HCPs find POWeR+ both feasible and acceptable to use in daily practice.

Implications for health care

Practices could feasibly consider implementing a weight management approach with internet-based support for behaviour change (such as POWeR+R) among their obese patients, because it likely to be more effective and more efficient than brief interventions using simple dietary advice.

Future research implications

1. Many individuals did not continue using the POWeR+ website and, given the evidence that those who did lost rather more weight, a research priority is to identify clinically effective and cost-effective ways of continuing to engage individuals.

2. Given the utility of POWeR+ with brief support in primary care, the question then arises of what public health benefit might accrue from using POWeR+ in community settings and what magnitude and nature of support is necessary, such as a central facilitator or pharmacy support, in order to achieve effective weight control.

3. Very few participants were engaged in using drug management of their weight problems despite this being intended as part of the package, and few individuals increased physical activity; an implementation study to develop and trial a complex intervention to address both these issues is warranted.

4. Future research to test the effectiveness, acceptability and feasibility of the CARe approach as a method for providing human support for digital interventions in a range of health conditions is warranted.

Contributions of authors

Paul Little (GP and Professor of Primary Care Research) had the original idea for the protocol, led the protocol development and the funding application, supervised the running of the lead study centre and co-ordination of centres, contributed to the analysis and led the drafting of the report.

Beth Stuart (Study Statistician) developed the analysis protocol, performed the quantitative analysis and contributed to the drafting of the report.

FD Richard Hobbs (GP and Professor of Primary Care) developed the protocol for funding, contributed to the management of the study, supervised the Oxford study centre and contributed to the drafting of the paper.

Jo Kelly (Senior Trial Manager) developed the protocol, provided day-to-day overall management of the study, co-ordinated recruitment in the lead study centre and undertook co-ordination of another centre, and commented on drafts of the report.

Emily R Smith (Senior Trial Manager) developed the protocol, provided day-to-day overall management of the study, co-ordinated recruitment in the lead study centre and undertook co-ordination of another centre, and commented on drafts of the report.

Katherine J Bradbury (Health Psychologist) contributed to protocol development, was responsible for the day-to-day development and piloting of the intervention, and commented on drafts of the report.

Stephanie Hughes (Health Psychologist) contributed to protocol development, was responsible for the day-to-day development and piloting of the intervention, and commented on drafts of the report.

Peter WF Smith (Professor of Statistics) developed the analysis protocol, performed the quantitative analysis and contributed to the drafting of the report.

Michael V Moore (GP and Professor in Primary Care) developed the protocol for funding, contributed to the management of the study, and contributed to the analysis and to the drafting of the report.

Mike EJ Lean [Professor of Human Nutrition (Medicine)] developed the protocol for funding, contributed to the management of the study and contributed to the drafting of the report.

Barrie M Margetts (Professor of Public Health Nutrition) developed the protocol for funding, contributed to the management of the study, led the analysis of the food frequency data and contributed to the drafting of the report.

Christopher D Byrne (Professor of Endocrinology and Metabolism) developed the protocol for funding, contributed to the management of the study and contributed to the drafting of the report.

Simon Griffin (Professor of General Practice) developed the protocol for funding, contributed to the management of the study and contributed to the drafting of the report.

Mina Davoudianfar (Trial Manager) co-ordinated the Oxford study centre.

Julie Hooper (Trial Administrator) provided co-ordination and data management for the Southampton study centre.

Guiqing Yao (Health Economist) developed the protocol for analysis of the notes review data and contributed to the drafting of the report.

Shihua Zhu (Health Economist) developed the protocol for analysis of the notes review data and contributed to the drafting of the report.

James Raftery (Professor of Health Economics) developed the protocol for funding, contributed to the management of the study, supervised the analysis of resource use data, developed the protocol for analysis of the notes review data and contributed to the drafting of the report.

Lucy Yardley (Professor of Health Psychology) developed the protocol and funding application, led the development of the intervention, contributed to daily supervision of website issues, contributed to broader study management and contributed to drafting the report.

Publication

Little P, Stuart B, Hobbs FDR, Kelly J, Smith E, Bradbury K, et al. An internet-based intervention with brief nurse support to manage obesity in primary care (POWeR+): a pragmatic, parallel-group, randomised controlled trial. Lancet Diabet Endocrinol 2016;4:821–8.

Data sharing statement

There are no further data available.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health.

Overview of the four stages of POWeR+

Stage 1: users are tunnelled (i.e. they cannot proceed to a session until they have completed the previous session) through three core sessions introducing them to POWeR+ and guiding them through the process of setting and reviewing weekly goals (Figure 4). There are three components to the goal review:

1. recording weight

2. choosing an eating plan from:

   • a low-calorie diet

   • a low-carbohydrate diet

3. choosing a physical activity plan:

   • increasing walking using a provided pedometer and recording the weekly step count

   • choosing any other physical activity and setting weekly goals.

FIGURE 4.

Overview of weekly goal review. This figure has been reproduced from Little et al. 26 under the terms of the CC BY 4.0 licence (https://creativecommons.org/licenses/by/4.0/).

Participants may opt to change their plans after their weekly goal review at any point during the intervention.

Stage 2: users have free-choice access (i.e. they can access the sessions in any order) of sessions 4–10. Goals can now be reviewed weekly and users can view their progress on a graph. If a user misses 1 week of goal review, they may still review their goals the following week. After all sessions are completed or a period of 7 weeks has passed since the stage was started, stage 3 sessions become available.

Stage 3: users have free-choice access to sessions 11–17 and continue to complete weekly goal reviews. After all sessions are completed or a period of 7 weeks has passed since the stage was started, stage 4 sessions become available.

Stage 4: users have free-choice access to sessions 18–25 and continue to complete weekly goal reviews.

Throughout the intervention, weekly e-mails are sent to remind users to complete their goal review. Set periods of inactivity at various points during the intervention also trigger e-mails prompting users to return to either complete sessions or complete their goal review.