Randomised controlled trial evaluating the effectiveness and cost-effectiveness of 'Families for Health', a family-based childhood obesity treatment intervention delivered in a community setting for ages 6 to 11 years

Care pathway for weight management of children and young people

Government guidelines describe a three-tier pyramid model for weight management, with three levels of service, from universal services, offered to everybody, through to specialist services, provided to those with particular needs. 8 Level 1 represents the core preventative services that all children, young people and their families should have access to, while level 2 represents targeted weight management services. The Families for Health intervention fits into level 2. This level of intervention is often aimed at children and young people with a BMI between the 91st and 98th centile, and services typically take the form of multicomponent family-based interventions, often taking place in community settings. They may also be called ‘early intervention services’. Level 3 is specialist support, offered to children with a BMI ≥ 99.6th centile, with a medical cause of obesity, significant comorbidity or complex needs. The Department of Health has recommended that overweight or obese children should be referred to appropriate weight management services. 9

Evidence for effectiveness of level 2 child obesity interventions

A Cochrane systematic review of interventions to treat obesity identified 64 randomised controlled trials (RCTs). 10 Thirty-seven studies were lifestyle interventions for children aged < 12 years (four dietary, nine physical activity and 24 behavioural). The authors concluded that it is difficult to recommend any particular intervention, but indicated that family-based lifestyle interventions combining dietary, physical activity and behavioural components can produce a significant and clinically meaningful reduction in overweight. Parental involvement was identified as useful with children aged < 12 years.

Recent UK-based trials of interventions targeting children who are obese have covered a variety of approaches. A RCT of the 9-week family-based community programme MEND (Mind, Exercise, Nutrition . . . Do it!; Mytime Active, Bromley, UK) with 116 children (aged 8–12 years) showed a between-group difference in the BMI z-score at the 6-month follow-up of –0.24 [95% confidence interval (CI) –0.34 to –0.13; p < 0.0001; n = 82] in favour of MEND over the waiting list control. 11 A further RCT compared paediatric dietitians using a one-to-one behavioural approach (5 hours) with standard dietetic care (1.5 hours) in 134 children (aged 5–11 years). 12 No significant differences in BMI z-scores were found at 6 or 12 months. A feasibility RCT that compared the community-based Watch It programme (delivered by health trainers) with a waiting list control, in 70 children and adolescents, found no significant change in BMI z-score and treatment. 13 A RCT of Epstein’s family-based behavioural treatment in a UK hospital, with 72 participants, found that there was no significant difference between the intervention and waiting list groups in BMI z-score, although both treatment and control groups showed significant reductions. 14

A National Institute for Health and Care Excellence (NICE) guideline on managing overweight and obesity among children and young people suggested that targeting both parents and children, or whole families, is effective in reducing BMI z-scores. 15 By the end of a programme, evidence on interventions involving families showed no negative effect on well-being and, in some cases, showed positive effects. Similarly, a review of the limited research on interventions focusing on parenting to treat childhood obesity shows a small positive effect on weight-related outcomes. 16

Family-based interventions rooted in behaviour theory have been shown to achieve better treatment effectiveness than those rooted in family systems theory. 17 A recent review of published studies18 investigating family-based childhood obesity interventions in the UK included 10 studies. The majority of programmes reviewed lasted 12 weeks, with only three studies providing follow-up data at 12 months or longer. Change in adiposity was a short-term benefit of participation, but there was insufficient robust evidence to suggest that this benefit was long-lasting and many studies were methodologically weak with limited internal validity. The authors concluded that there was insufficient evidence on how the inclusion of parents and the wider family may impact on the effectiveness of UK community-based weight management programmes for children and young people. 18

Evidence for cost-effectiveness of child obesity interventions

Oude Luttikhuis et al. 10 point to a paucity of cost-effectiveness studies in this area. A recent US study of families in which both children and parents were overweight or obese found that the cost per unit weight lost by parents and children was lower for family-based group treatment than for an intervention that treated the parent and child separately. 19 Evidence from seven short-term health economic analyses suggests that, in the short term, lifestyle weight management programmes will result in an increased cost to the NHS in terms of BMI z-score gains when compared with routine care. However, overall small (and in some cases non-significant) improvements in BMI z-scores can be achieved. 15 Interventions that lead to even small reductions in BMI can be cost-effective in the long term at conventional cost-effectiveness thresholds, provided that the short-term effects on BMI can be sustained into adulthood. The NICE guidance suggested that evidence from these studies is directly applicable, but there were some potentially serious limitations to the studies. 15

From randomised controlled trial to practice

Two recent studies have investigated the roll-out of an intervention tested in a RCT into the community. 20,21 Both were based on the multicomponent family-based weight management intervention, MEND, which addresses diet and physical activity through education, skills training and motivational enhancement. When delivered at scale, the rolled-out intervention resulted in the improvement of BMI and psychosocial outcomes but worked less well for some groups of children (certain ethnic groups and those from more deprived areas). This was not because of variations in uptake of the programme, which was good in deprived groups, but because of differing experiences in completing, with more disadvantaged groups being less likely to complete the programme. The intervention, therefore, had the potential to widen inequalities. The authors suggested that further research should investigate how completion rates could be improved for particular groups. 21 They felt that the intervention should be implemented in such a way as to achieve sustained impact for all groups by modifying content, training and implementation. 20

Parenting programmes

It has been argued that parenting programmes have potential to improve the mental health and well-being of children, as well as improve family relationships and benefit the community as whole. 22 A public health approach to parenting is needed to ensure that more parents benefit and that a societal-level impact is achieved. 22 The effectiveness of parenting interventions in the treatment and prevention of childhood obesity has shown a small positive effect on weight-related outcomes,16 but this approach has not been investigated in a RCT in the UK.

Development of the Families for Health intervention

‘Families for Health’ is a manualised group-based family intervention for the treatment of children aged 6–11 years who are overweight or obese, which could be an option in the care pathway as a targeted ‘early-intervention’ service. The programme was developed at Warwick University and supported by a Department of Health career development award to the principal investigator. The programme differs from other interventions for childhood obesity being used or tested in the UK by offering a greater emphasis on parenting skills, relationship skills and emotional and social development, combined with information about lifestyle. Parents and children attend separate groups, meeting mid-way for a healthy snack and activity. The groups are led by two trained facilitators.

The development and evaluation of the Families for Health programme has followed the Medical Research Council (MRC)’s framework for complex interventions. 23 A pre–post pilot study in Coventry of 27 children showed that a mean reduction in children’s BMI z-scores from baseline was sustained at 9 months (–0.21, 95% CI –0.35 to –0.07; p = 0.007) and 2 years (–0.23, 95% CI –0.42 to 0.03; p = 0.027). 24,25 There were also other health-related improvements. Interview data showed that parents found the parenting approach helpful, providing the tools for them to become ‘agents of change’ in the family. The NHS costs to deliver the programme were £517 per family or £402 per child. 25 The process evaluation of the pilot study showed the need for minor modifications to the programme including more physical activity, removal of some unnecessary material, a reduction in the number of sessions and addition of follow-up sessions. These changes to the intervention might increase effectiveness.

The materials for version 1 of this programme were developed by Candida Hunt and the University of Warwick team, including investigators in this trial (WR and SS-B). 26 Following development and evaluation of the original programme Families for Health version 1,24,25 changes were made including shortening the delivery from 12 to 10 sessions, adding two follow-up sessions and enhancing the information given to families on healthy eating and pedometers. The programme combines information on parenting skills, social and emotional development, as well as lifestyle change. The parenting aspects are based on the Nurturing Programme from Family Links (Family Links, Oxford, UK),27 and the circle time elements in the children’s programme have parallels with the Family Links Nurturing Programme for schools. 28

Delivery is group based involving 8–12 families, with children and parents attending parallel groups. The intervention is delivered in a community setting (e.g. a leisure centre or school), to enhance access and ensure adequate space and facilities for physical activity. The parents’ group covers both support with parenting skills and family lifestyle, which are integrated in the weekly sessions. The approaches include facilitated discussion, role-play, goal-setting, skill practice, a solution-focused approach and homework. The children’s programme includes a focus on healthy eating using the ‘Eatwell plate’ (Food Standards Agency, London, UK) as the basis; circle time to discuss emotional aspects of their lives and enhancing self-esteem; and physical activity aimed at increasing activity levels by participation in games, the use of pedometers and introduction to new physical activities. The parents and children meet mid-way in each session for a healthy snack and an active game. This gives facilitators an opportunity to act as role models, for example showing parents how they might reward or praise their children, and to introduce ways in which children and parents can interact at home. It also provides an opportunity for children to prepare healthy snacks and to try new foods.

The main principles underpinning the Families for Health intervention are that the parents are identified as the agents of change responsible for implementing lifestyle change in the family. 29 The parenting aspects aim to support and increase parental capacity to implement and maintain the lifestyle changes that they would like to try each week. The focus is on healthy eating (not dieting) and activity for the whole family (that is, not just for the child who is overweight), with an emphasis on children growing into their weight rather than weight loss. The programme aims to promote a sustainable, healthy approach to family-wide lifestyle change. The programme puts greater emphasis on parenting skills, relationship skills, and emotional and social development than other similar interventions, and combines this with information about lifestyle. These form a core theme throughout and are integrated alongside topics in each weekly session.

Informed consent

Informed consent was obtained in three steps, giving parents and children time to consider whether or not they wished to participate. Each potential participant was given, or sent by post, information sheets about the trial (child and parent versions). After a minimum of 3 days, parents were contacted by telephone to ask whether or not they were interested in taking part in the trial and to answer any questions. A researcher then visited the parent(s) and child(ren) at their home, and obtained the parent’s written consent and the child’s written assent. All research staff were trained in informed consent, including methods for assessing competence for consent, agreement to participate and obtaining assent from children.

Parents

A parent who had been a participant in the pilot of the Families for Health intervention was a member of the TSC. She attended seven of the eight TSC meetings and contributed to discussions on the intervention and study design.

Children

The aim for user involvement, as stated in the protocol, was met using the following approach. Claire Callens and Carly Tibbins, from the user involvement team of the West Midlands Medicines for Children’s Research Network carried out patient and public involvement consultation exercises with children from the West Midlands prior to recruitment from January to February 2012, comprising two stages: stage 1, two focus groups with hospital-based young person’s advisory groups (aged 8–18 years), which aimed to develop and pilot questions for the second stage, and also included obtaining views on the acceptability of wearing accelerometers to measure physical activity; and stage 2, consultation on the proposed intervention and the research measurements (including accelerometers) with children in the target age range for the trial (key stage 2, aged 7–11 years) from two primary schools from the areas in which the trial was to be delivered.

A specific area of inquiry in the second stage of consultation was about the types of activities children most enjoyed, which was used to inform the ‘activity tasters’ within the Families for Health intervention. Older children (years 5 and 6) completed a questionnaire and younger children (years 3 and 4) voted, by show of hands, on the options given for physical activity. Once they understood the concept of a children’s gym, all children felt that they would like to attend one to see what it was like.

Across the two stages of consultation, 35 young people aged 7–18 years were consulted on accelerometers in five focus groups. Prior to participation in the focus group discussion, 18 children were given the opportunity to wear an accelerometer (GT1M ActiGraph; ActiGraph LLC, Pensacola, FL, USA). First impressions of the acceptability of the accelerometer were often negative relating to comfort and unwanted attention from friends, indicating that this may be a difficult tool to use. Others felt that an increased attention from others would be positive. Ways to increase use by participants were suggested as presenting the accelerometer in a very positive way, include a trial wearing period and providing incentives. 34 A full report of the user involvement consultation with children is available. 35

The intervention: Families for Health

Families for Health version 2 is a family-based programme aimed at the treatment of children (aged 6–11 years) who are overweight or obese. Delivery is group based, and the aim was to recruit groups of 8–12 families, with children and parents attending parallel groups. In families randomised to this arm, both parents were invited, together with all overweight and non-overweight siblings in the target age range. Where necessary, the start of a programme was delayed so that viable attendance numbers could be obtained. The programme was manualised, with detailed handbooks available to facilitators, parents and children. Groups were run on a Saturday morning or afternoon for 2.5 hours each week for 10 weeks. Follow-up sessions were planned for 1 and 3 months post intervention.

The main principles underpinning the Families for Health intervention were that the parents were identified as the agents of change responsible for implementing lifestyle change in the family. 29 A solution-focused approach was employed, focusing on solutions rather than the problem. The programme emphasises parenting skills, relationship skills, and emotional and social development, and combines this with information about lifestyle, all of which are key to implementing and maintaining behaviour change. Table 2 shows an outline of the main content of the parallel parents’ and children’s groups for the 10 weeks. The weekly topics in the Families for Health programme were broadly the same for both parents’ and children’s groups, in order to promote greater understanding and discussion at home. The parents and children met mid-way each week for a healthy snack and an active game, with an aim of introducing ways in which children and parents could interact at home.

The parents programme of the Families for Health programme shared some of parenting skills topics from the Nurturing Programme from Family Links,27 and included both behavioural (e.g. positive discipline, family rules) and relationship (e.g. giving praise, raising self-esteem, emotional health) approaches to parent training. The support with parenting skills were integrated with family lifestyle topics around healthy eating and physical activity in the weekly sessions. The approaches include facilitated discussion, role-play, goal-setting, skill practice, a solution-focused approach and homework.

The children’s programme included a focus on healthy eating using the Eatwell plate as the basis; circle time to discuss emotional aspects of their lives and enhancing self-esteem; and physical activity aimed to increase activity levels by participation in games, the use of pedometers and introduction to new physical activities.

Families in the intervention arm were also eligible for usual-care interventions and any usual care they received was documented.

Training and selection of facilitators: Families for Health

Four facilitators, two for the children’s group and two for the parents’ group, were required to run each Families for Health programme. Facilitators were identified from the local NHS or other services, and selected on the basis of personal attributes including empathy for families with overweight children and previous relevant experience, for example with running groups with parents and children. In site A, a formal application process was set up, whereas for sites B and C nominations were received through the leads for obesity. Professional backgrounds of facilitators included community nursing, teaching, youth work, leisure services and nutritionists.

Nineteen facilitators attended a 4-day training course in February 2012 provided by two trainers from Family Links. Trainers from Family Links were used to deliver the training because of the close association of the Families for Health programme with the Nurturing Programme from Family Links. 27,28 The training covered the content, philosophy and logistics of running the programme. For some sessions facilitators were divided up into subgroups of whether they were facilitating the children’s or parents’ group, and practised facilitating parts of the programme. Participants completed an evaluation form for each day of the training, including the level of confidence in delivering the Families for Health programme and rating the usefulness of the topics covered. Training increased facilitators’ confidence in delivering the programme and most found the training a positive and useful experience. Further details of the evaluation of the training of the facilitators are in Appendix 1.

Usual-care control group

Families assigned to the control arm were offered any usual care that was available in their area. During the duration of the study, usual care for each locality varied.

Site A had the One Body One Life (Coventry City Council, Coventry, UK) 10-week programme, which is a group-based family intervention that has been subject to published evaluation. 36 This was available throughout the study. The eligibility criteria for the One Body One Life programme was children aged 7–16 years, but was offered to the whole family if one or more member of the family was an ‘unhealthy weight’. The programme took a solution-focused approach, with the 1.5-hour sessions comprising a 45-minute physical activity workshop and a 45-minute healthy eating workshop.

Site B had Change4Life (Department of Health, London, UK) advisors who offered one-to-one support for weight management for children aged 4–13 years who were overweight or obese. Recruitment to their service was via self-referral, referral from school health or other health professionals, and via the National Child Measurement Programme (NCMP). Visits were mainly undertaken at the child’s home. This service was available for the majority of the study, though in the final few months of recruitment funding for Change4Life advisors was withdrawn and a single telephone call was instead being offered.

In site C, usual care was one of the following: (1) a weight management programme for children and young people aged 7–15 years, comprising a two-step programme, MEND and Choose It (Wolverhampton City Council, Wolverhampton, UK), focusing on taster sessions for physical activity and healthy eating. Funding for this programme was withdrawn halfway through the study and so two alternatives were offered as ‘usual care’ in site C, depending on the age of the child; (2) Weight Watchers^® (Weight Watchers UK Ltd, Maidenhead, UK) for young people aged ≥ 10 years, who had to be accompanied by a parent; and (3) a referral to the school nurse for children aged 6–9 years, where children would be weighed and measured and offered advice. This type of usual care was not standard at the start of the study and, therefore, did not always occur as hoped. Table 3 gives further details of the usual-care programmes.

Inclusion criteria

Families were considered for inclusion if:

• they had at least one overweight (≥ 91st centile for BMI) or obese (≥ 98th centile for BMI) child aged 6–11 years, based on the UK 1990 BMI37

• at least one parent or guardian and the overweight child was willing to take part.

Exclusion criteria

Families were excluded if:

• the parent or child had an insufficient command of English and would find it difficult to participate in the group

• the child had a metabolic or other recognised medical cause of obesity

• the child had severe learning difficulties and/or severe behavioural problems, and would find it difficult to participate in a group-based programme.

Anthropometric

Physical activity and diet

Physical and mental health

Family relationships

Economic outcomes

Services received by child

A modified version of the CSRI was used to record hospital and community health and social services received by each child, as well as broader service utilisation (adaptation of CSRI from Beecham and Knapp54), at baseline, 3 and 12 months. The unit costs of each resource item were valued using both primary research, based on established accounting methods, and data collated from secondary national tariff sets. 61–64 Further details of the methods used are provided in Chapter 5.

Socioeconomic classification

Families’ socioeconomic classification was measured using details of parental employment. Employment status and type of employment of the mother and/or father was recorded, including their actual occupation and whether they were self-employed or an employee. This information was used to code families into eight socioeconomic classes using the The National Statistics Socio-Economic Classification. 55 As the number of families was small, this was subsequently collapsed into three classes, plus a fourth category of ‘never worked and long-term unemployment’ (Table 6).

Original sample size justification

Power calculations assumed a residual SD in the BMI z-score of 0.22, a SD of the random family effects of 0.14, an intracluster correlation of 0.1 in the intervention groups, a two-sided significance of 5% and that 60% of participating families have one overweight or obese child and 40% have two with a within-family intracluster correlation of 0.27. Allowing for clustering effects by family and for group effects in the intervention arm, a sample size of six groups of 10 families (60 families) in the intervention arm and 60 families in the control arm gives a power of 94% to detect an intervention effect of 0.2 in BMI z-scores based on pilot study data and relevant literature. 24,65 If 30% of families were to drop out, the study retains a power of 88%. As a check, the power was calculated when no families have more than one overweight or obese child. In this case, the power is 92% or 83% if 30% of families drop out. Power values were estimated using a simulation study including 10,000 simulated trials.

Updated sample size calculation

Recruitment to the trial was slower than expected, leading to smaller groups for the Families for Health programme than planned; therefore, the addition of a seventh Families for Health programme was proposed. On 19 June 2014 this amendment to the Families for Health protocol was approved by Health Technology Assessment. The sample size calculation was updated to incorporate the additional programme and recruitment thus far (all other assumptions of the sample size calculation remained the same). A sample size of seven groups of eight families (56 families) in the intervention arm and 56 families in the control arm yields a power of 91% to detect an intervention effect of 0.2 in BMI z-scores if seven families (one with two children) were providing 12-month follow-up data (12.5% dropout) and 88% power if six families (one with two children) were providing 12-month follow-up data (25.0% dropout). If no families had more than one child, then the power would be 89% and 85% for the 12-month follow-up of seven and six families per programme, respectively.

General statistical considerations

The primary outcome analysis as well as all secondary outcome analyses were conducted at the conventional (two-sided) 5% level and, corresponding to this, all presented CIs are 95% CIs. As specified in the statistical analysis plan, no formal adjustment for multiple testing among the secondary end points was used as outcomes are likely to be highly correlated so that standard adjustment techniques, such as the Bonferroni method, would be conservative. All outcome measures and child/parent characteristics were summarised by the trial allocation group and for outcome measures by follow-up period. The distribution of outcome data was investigated and transformations applied, if necessary, before performing statistical tests and modelling. The main analysis of clinical outcome data was the comparison of change from baseline between treatment groups. Using the change from baseline has a ‘normalising’ effect so that standard techniques could generally be used.

Unless otherwise stated, all analyses were performed on an intention-to-treat basis, that is all participants were analysed in the arm they were allocated to and included regardless of whether or not the treatment and follow-up schedule was complied with. This reflects the pragmatic nature of the trial and ensures that conclusions drawn from the analysis will reflect the impact of the Families for Health programme in a real-world setting.

The statistician conducting the statistical analysis was unblinded for the final statistical analysis and the analysis presented at the final DMEC/TSC meeting. The multilevel models originally proposed to be adopted in the analyses (see Primary outcome analysis) consist of two hierarchical levels in the usual-care arm and three levels in the intervention arm. For these models to be fitted, the participants’ group affiliation needed to be revealed. Unblinding was agreed to by the TSC.

The entire statistical analysis was conducted using SAS v9.4 TSL1M2 (SAS Institute Inc., Cary, NC, USA) with the exception of the sample size calculation, which was conducted using R versions 2.10 and 3.0 (R Foundation for Statistical Computing, Vienna, Austria).

Primary outcome analysis

As indicated above, the primary outcome for the statistical analysis was the mean change in child BMI z-score after 12 months of follow-up compared between treatment arms. It was anticipated that these data would be correlated within families (if more than one child of a family participates in the trial) and within delivery groups in the Families for Health intervention arm. The analysis allowed for this clustering in order to obtain unbiased estimation of the treatment effect and its standard error (SE). 67

A three-level hierarchical mixed-effects model was proposed to be fitted in the statistical analysis plan. At the highest level of the model a random effect for delivery group was intended. Delivery group-level clustering would have been allowed for in the Families for Health arm only, as usual-care interventions varied by site, were not necessarily group based and precise details on usual-care treatment received were generally not available. As the statistical analysis showed, there was no evidence of delivery-group clustering and models comprising this random effect failed to converge (see Chapter 4, Main primary outcome analysis). The decision has therefore been made (and been approved by the TSC) to remove this effect from all hierarchical modelling and to use a two-level hierarchical mixed-effects model instead.

Correlation between measurements of children within family was allowed for in both arms.

The multilevel model was adjusted for the child-level characteristics, baseline BMI z-score and sex, and family-level variable ‘locality’ as fixed effects. Additionally, adjustment for the family-level characteristic socioeconomic status (SES) and child-level characteristic ethnicity was explored. Restricted maximum likelihood estimation was employed for estimating covariance parameters in the multilevel modelling. The Satterthwaite approximation68 was used for computing the denominator degrees of freedom for the test of a treatment effect difference between the allocation groups (and for tests of other fixed effects).

The primary analysis is a complete case analysis in the sense that if either the baseline or 12-month follow-up z-score was missing the subject had to be omitted from the analysis.

Sensitivity analyses for the primary outcome

Several sensitivity analyses were undertaken to assess the impact of areas of uncertainty surrounding the primary outcome analysis and its robustness. These involved re-estimating the treatment effect under the following scenarios: (1) conducting a per-protocol analysis in which families having participated in five or more sessions of the Families for Health programme are regarded as ‘programme completers’ (i.e. as having complied with the protocol sufficiently); and (2) (multiple) imputation of missing primary outcome data.

Three standard imputation techniques were employed to assess the sensitivity of the analysis to the missing data: first, simple regression imputation, in which missing values are imputed by predicted values from a linear regression model using the same predictors as the primary outcome analysis; second, multiple imputation methods Markov chain Monte Carlo;69 and, third, fully conditional specification regression. 70 For the two multiple imputation analyses, 200 burn-in iterations were used and estimates averaged over 100 imputed data sets. Baseline BMI z-score, age, sex and site were included as explanatory variables in the imputation models. Imputations were generated separately for each treatment group.

Subgroup analyses

To explore heterogeneity in the trial population, the following prespecified exploratory subgroup analyses were conducted with respect to the primary outcome:

• child’s sex (male or female)

• locality (site)

• SES (according to The National Statistics Socio-Economic Classification’s55 four-class standard classifications)

• parent’s BMI at baseline (normal, overweight or obese)

• age of child at baseline (6–8 years or 9–11 years).

The difference in treatment effects by subgroups was initially assessed by interaction tests. These were performed via significance tests of interaction terms in the hierarchical model utilised for the primary outcome analysis. Variables that have been categories for subgroup analyses (e.g. age, parent BMI) were also investigated as covariates on their original scale. Separate models were then fitted for each subgroup to obtain estimates of the treatment effects within subpopulations.

Repeated measures modelling

An exploratory analysis was performed for the investigation of the difference between arms in terms of change over time in the primary outcome measure rather than a comparison between arms at either 3- or 12-month follow-up. In this analysis, the time at which follow-up data were provided was fitted as a continuous variable, accounting for the fact that the actual times varied widely (especially for the 3-month follow-up). For this purpose a repeated measures mixed model was fitted. This model was based on the aforementioned hierarchical model for the primary outcome analysis comprising the same effect (plus time). Model complexity was increased as time became the new first-level (random) factor. The same model specification, as in the primary outcome multilevel model (restricted maximum likelihood estimation), was used where possible. An unstructured covariance matrix was assumed for the correlation between measuring time points.

Analysis of baseline demographic data

Baseline demographic outcomes were obtained on a family level or on a child/parent level. Summary statistics (mean and SD for continuous variables and absolute number and percentage for categorical variables) were calculated for all participants recruited to the trial and for the two treatment groups separately. The characteristics within the two treatment groups were compared using t-tests and chi-squared tests for continuous and categorical variables, respectively. The statistical comparison of baseline characteristics was considered exploratory rather than confirmatory and ignored clustering by delivery group or family. The main intention was to identify baseline variable differences that might be deemed relevant by the main investigators and would consequently require adjustment in the multilevel models fitted in the primary and secondary outcome analyses.

Secondary outcome analyses

All secondary outcomes, including subscales of questionnaires, were summarised by trial allocation group and follow-up period using mean, SDs and 95% CIs for continuous variables and absolute number, percentages and 95% CIs for categorical variables. Statistical tests of differences for child secondary outcomes were performed using the same hierarchical mixed-effects model as the primary outcome analysis. In addition, child BMI z-score was also compared as change from baseline to 3-month follow-up and change from baseline to 12-month follow-up using independent group t-tests. Parent outcomes were compared using t-tests and chi-squared tests, as appropriate, as generally only one parent provided data and within-family correlation could not be modelled. In case parents and children provided data for the same outcome, the simpler test was used to ensure comparability. Where the analysis of parent outcomes was adjusted, hypothesis testing was done within a linear regression model and the adjustment variables provided in the respective results section.

Primary outcome analysis

Exploratory unadjusted analysis

The primary end point for the statistical analysis is the mean change in BMI z-score from baseline to 12 months of follow-up. The primary analysis is a complete-case analysis in the sense that if either the baseline or 12-month follow-up z-score is missing the subject was omitted from the analysis. In the subsequent section, the primary end point was compared between treatment arms using a hierarchical mixed-effect model. Before that, simple unadjusted means are compared in this section. Table 12 shows the number of children and BMI z-score means at each time point by treatment group. The change over time of these means is illustrated in Figure 2.

TABLE 12 - Primary outcome: BMI z-score – sample sizes and unadjusted estimates

FFH, Families for Health; UC, usual care.

Treatment group	Sample size (n)			Unadjusted mean z-score (95% CI)
	Baseline	3 months	12 months	Baseline	3 months	12 months
FFH	63	48	45	2.69 (2.52 to 2.85)	2.62 (2.43 to 2.82)	2.72 (2.54 to 2.89)
UC	65	50	43	2.74 (2.57 to 2.91)	2.65 (2.45 to 2.85)	2.58 (2.37 to 2.79)

FIGURE 2.

Change over time for primary outcome BMI z-score (unadjusted means and CIs). FFH, Families for Health; UC, usual care. Adapted from Robertson W, Fleming J, Kamal A, Hamborg T, Khan KA, Griffiths F, et al. Randomised controlled trial and economic evaluation of the ‘Families for Health’ programme to reduce obesity in children [published online ahead of print 21 December 2016]. Arch Dis Child 2016. doi: 10.1136/archdischild-2016-311514. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited (see http://creativecommons.org/licenses/by/4.0/). 71

The literature suggests that a change of –0.25 in BMI z-score would be considered clinically significant. 65 Seven children (15.6%) in the Families for Health group and 10 children in the usual-care group (23.3%) achieved a reduction of ≥ 0.25. This difference is non-significant (p = 0.360). No reduction of > 0.5 was observed in the Families for Health group, whereas three children in the usual-care group had a reduction of > 0.5. The largest reduction observed was –0.746 and the largest increase was 0.895.

Table 13 presents (unadjusted) mean change from baseline estimates by treatment groups in child BMI and waist z-scores, and percentage body fat. These are called ‘longitudinal changes’ henceforth. A significant reduction from baseline to 12-month follow-up in the BMI z-score is achieved in the usual-care group with the p-value for the difference in reduction between groups being p = 0.052. The magnitude of the difference and p-value is consistent with those of the secondary (anthropometric) outcomes. Waist z-score and percentage body fat are also presented in Table 13.

TABLE 13 - Primary end point: BMI z-score, waist z-score and percentage fat in children

FFH, Families for Health; UC, usual care.

Note

Bold text denotes values for the reduced data set without child who had large weight increase.

Measure	Time period
	Baseline to 3 months			Baseline to 12 months
	Treatment group		p-value	Treatment group		p-value
	FFH	UC		FFH	UC
BMI z-score, mean change (SD) [95% CI]	–0.019 (0.253) [–0.093 to 0.054]	–0.042 (0.163) [–0.089 to 0.004]	0.593	0.013 (0.277) [–0.070 to 0.096] –0.001 (0.264) [–0.081 to 0.079]	–0.102 (0.268) [–0.184 to –0.019]	0.052 0.081
Waist z-score, mean change (SD) [95% CI]	–0.111 (0.276) [–0.191 to –0.031]	–0.048 (0.216) [–0.110 to 0.013]	0.211	0.002 (0.372) [–0.110 to 0.113] –0.011 (0.366) [–0.122 to 0.101]	–0.145 (0.295) [–0.235 to –0.054]	0.045 0.065
Percentage body fat, mean change (SD) [95% CI]	0.204 (2.492) [–0.520 to 0.928]	–0.084 (1.941) [–0.636 to 0.468]	0.524	0.642 (2.822) [–0.206 to 1.490] 0.480 (2.633) [–0.321 to 1.280]	–0.900 (4.381) [–2.248 to 0.448]	0.055 0.080

The Families for Health group comprised one child who gained a large amount of weight between baseline and 3-month follow-up for reasons unrelated to the intervention. Removing this child from the analysis (see Table 13, bold values) leads to larger p-values for all outcomes. However, the Families for Health RCT was specified to use an intention-to-treat analysis; therefore, this child is retained in all subsequent analyses.

Table 14 shows between-group comparisons of further anthropometric measures for children. All of them suggest a favourable development in the usual-care group compared with the Families for Health programme, albeit with small absolute differences.

TABLE 14 - Child anthropometric measures

FFH, Families for Health; UC, usual care.

Based on exact age also used in z-score calculation.

For estimates without the child in the FFH group who gained a large amount of weight between baseline and 3 months, the FFH BMI means would be 27.74, 25.97 and 27.10 for baseline, 3 months and 12 months, respectively.

Anthropometric measure	Time point
	Baseline		3 months		12 months
	FFH (n = 63)	UC (n = 65)	FFH (n = 48)	UC (n = 50)	FFH (n = 45)	UC (n = 43)
Mean child age, years (SD)a	9.46 (1.57)	9.43 (1.61)	10.01 (1.53)	9.56 (1.65)	10.50 (1.49)	10.27 (1.63)
Mean child BMI, kg/m2 (SD)b	25.79 (4.44)	25.93 (4.32)	26.18 (4.94)	25.60 (4.24)	27.30 (5.02)	25.82 (3.67)
Mean child percentage body fat (SD)	36.81 (6.18)	37.46 (6.72)	36.69 (6.63)	36.95 (6.15)	37.58 (6.86)	35.71 (6.74)
Mean child waist circumference, cm (SD)	86.17 (11.61)	86.30 (11.88)	86.44 (13.10)	85.62 (12.19)	90.36 (11.86)	86.47 (10.86)

Main primary outcome analysis (adjusted)

The primary outcome measure was the change in children’s BMI z-score from baseline to the 12-month follow-up. It was anticipated that the data would be correlated within delivery groups in the Families for Health intervention arm and within families should more than one child from a family participate in the trial. It was therefore stipulated that a three-level hierarchical mixed-effects model would be fitted for the primary outcome analysis. Furthermore, it was decided that this analysis would adjust for child-level characteristics baseline BMI z-score and sex, and the family-level variable locality (site).

Fitting a model with a random effect for the Families for Health delivery group yielded an estimate of 0 for this effect (Table 15); the model was unable to uniquely estimate any variation from one Families for Health programme to another, above and beyond the residual variance. This suggests that the responses of children from the same delivery group were no more similar than children from different groups. Several checks have been conducted to test this. The decision has therefore been made to remove this effect and proceed with a two-level hierarchical model.

For the two-level model for the primary outcome the intraclass correlation estimate is:

with a 95% CI of 0.3170 to 0.6256. CI calculation is based on Bonett. 72 This suggests that 47.13% of the observed variation in the dependent variable is attributable to family-level characteristics and 52.87% is attributable to child-level traits.

Table 16 presents covariate effect estimates in the hierarchical modelling. The estimates for the prespecified primary analysis model are shown in the upper part of the table (model 1). Expert input by TSC members recommended exploring additional adjustment for child ethnicity and family SES (model 2). In both models BMI z-score was the only statistically significant predictor. We therefore continued with model 1. F-test statistic p-values for predictor variables are not shown; however, the p-value of the only significant effect is identical to the p-value in the table (because it is a continuous variable).

The primary outcome analysis estimate of the between-group difference in terms of BMI z-score is 0.114 (95% CI –0.001 to 0.229), that is a non-significant difference in favour of usual-care treatment (Table 17). This estimate and its p-value are almost identical to the estimate obtained in the unadjusted analysis. Point estimates of the within-treatment group differences suggest a slightly larger reduction than in the unadjusted analysis. The different treatment effect estimates obtained for the primary outcome can be compared in Table 18. Unadjusted values are taken from Table 13, model 1 is the prespecified primary outcome analysis and model 2 is the analysis including two additional predictors. Treatment effect estimates and p-values are consistent across analyses.

Imputation analysis for primary outcome

To further investigate the consistency of the treatment effect estimates, an imputation analysis was conducted to explore the effect of the large number of missing data at 3- and 12-month follow-up. In the Families for Health group, 28.57% of 12-month primary outcome measures were missing, in the usual-care group 34.20% were missing. This difference in missing data is not significant (p = 0.520) by the chi-squared test.

Estimates presented in Table 19 for the multiple imputation models are averages over the 100 models. Treatment effect point estimates are, again, similar to the primary outcome analysis with only Markov chain Monte Carlo imputation reducing the effect size and as a result yielding a non-significant p-value.

Baseline covariates included in the analysis were fully observed and only outcome values were missing. If the assumption is made that data are ‘missing at random’, then the primary outcome analysis, which adjusts for baseline BMI z-score (and treatment group), provides an unbiased estimate of treatment effect. In this case, multiple imputation could be regarded as unnecessary or an additional sensitivity analysis. In this scenario the simple regression imputation method also provides unbiased estimates, albeit with an underestimated variability. This effect can be seen in Table 19, in which point estimates for simple regression imputation are very similar to the primary outcome analysis, but the p-value is smaller.

Per-protocol analysis

Families that participated in five or more sessions of the Families for Health programme are regarded as ‘programme completers’ (i.e. as having complied with the protocol sufficiently). Improvements in children from these families, in terms of BMI z-score change from baseline, are compared with improvements in children from families who attended fewer than five sessions. A comparison is made based on estimates obtained from the primary analysis multilevel model (model 1). An exploratory analysis comparing families who attended nine or more sessions (essentially full attendance) with those who attended fewer sessions can be found in Appendix 2.

Figure 3 presents the number of sessions attended by families and the number of follow-up appointments completed. The mean number of Families for Health sessions attended was 5.42. For those families who provided 12 months’ data, the mean number of sessions attended was 6.96. The figure suggests no clear relationship between sessions attended and the likelihood of being lost to follow-up.

FIGURE 3.

Attendance rate (bar height) and time points at which outcome data were provided (colour/pattern) for families attending Families for Health.

Table 20 shows the difference in BMI and waist z-score between completers and non-completers. Non-completers achieved a reduction similar to the reduction observed in the usual-care group, while the mean z-score increased in the completer population. There are several potential explanations for this including (a) those who achieved a weight reduction felt that they did not need to attend further sessions; and (b) families were able to access usual care (which appears to yield a better outcome) as well and non-completers might have been more likely to have accessed usual care. Non-completers also had a (non-significantly) higher mean baseline BMI z-score (Table 21).

Repeated measures analysis

An exploratory repeated-measures analysis was conducted for the investigation of the difference between arms in terms of change over time in the primary outcome measure rather than a comparison between arms at fixed 3- or 12-month follow-up time points. For this purpose, a repeated-measures mixed model was fitted. The model is based on the hierarchical model used above. Unlike previous models, time is fitted as a continuous variable here, allowing for the fact that actual measuring time points varied considerably rather than taking place at exactly 3 and 12 months from baseline. This is illustrated in Figures 4 and 5.

FIGURE 4.

Child BMI z-scores and time points (Families for Health).

FIGURE 5.

Child BMI z-scores and time points (usual care).

Analysis of the covariance matrix for repeated observations revealed that the baseline and 12 months observation were more highly correlated than was the 3 months observations and either of the other time points. An unstructured covariance matrix leads to the best model fit and was therefore used for the mixed model.

The least squares estimates of group BMI z-scores and the treatment effect presented in Table 22 are averages over time. This includes the baseline time point, which dilutes a difference because of treatment. Had discrete time points (3 months and 12 months) been used, the differences between arms at different time points could have been estimated. However, the time by treatment interaction term in the repeated measures model is significant (see Table 23, bottom row). Consequently, it can be claimed that Families for Health and usual care have a statistically significantly different effect on children’s BMI z-score, with Families for Health yielding higher z-scores. Unlike in previous models presented in Primary outcome analysis, locality and sex are significant predictors of the outcome (Table 23). Note, however, that the outcome variable in this repeated measures model is z-score change over time [i.e. (slightly) different from ‘change from baseline to 12-month follow-up’ used in the primary outcome analysis].

Secondary outcomes analysis

Parenting Styles and Dimensions Questionnaire

Parenting style was measured using the PSDQ. 50 The PSDQ is a 32-item, parent-report questionnaire. The 32 items form three patterns of parenting: authoritative, authoritarian and permissive. Higher scores indicate more frequent use of the described behaviour. Authoritarian and permissive parenting styles are considered to be less effective parenting styles than authoritative. Parents are classified according to their highest score in the three parenting domains.

Table 38 shows the mean score for each parenting style at each time point. The authoritative style always scored higher than the other two styles. For 53 parents in the Families for Health group, authoritative was the highest scoring parenting pattern at baseline, with the remaining three parents scoring highest on permissive (5.4%). In the usual-care group, 53 parents endorsed behaviours indicative of a primarily authoritative parenting style and six parents (10.2%) exhibited permissive styles at baseline. At the 3-month follow-up, 43 parents scored highest on authoritative style, whereas two parents (4.4%) scored highest on permissive behaviour in both treatment groups. At the 12-month follow-up, one parent (2.3%) in the Families for Health group and two parents in the usual-care group (5.3%) scored highest on permissive style. Authoritarian never emerged as the primary parenting style in this trial. Very few changes of the primary parenting style occurred over the course of the study, which is desirable as most parents are already in the preferred group (authoritative). In total, five parents scored higher for a parenting style different from their primary baseline parenting style over the course of the study: three in the Families for Health group and two in the usual-care group. Four parents changed from primarily permissive to authoritative, and one parent in the Families for Health group changed from authoritative to permissive.

TABLE 38 - Sample sizes and time point averages in parenting style (PSDQ)

FFH, Families for Health; UC, usual care.

Parenting style	Time point average (95% CI)
	Baseline		3 months		12 months
	FFH (n = 56)	UC (n = 59)	FFH (n = 45)	UC (n = 45)	FFH (n = 44)	UC (n = 38)
PSDQ authoritative	4.08 (3.92 to 4.23)	4.02 (3.84 to 4.21)	4.13 (3.97 to 4.29)	4.04 (3.82 to 4.26)	4.19 (4.04 to 4.34)	3.97 (3.74 to 4.21)
PSDQ authoritarian	1.70 (1.58 to 1.83)	1.58 (1.47 to 1.69)	1.60 (1.46 to 1.74)	1.56 (1.45 to 1.66)	1.59 (1.45 to 1.72)	1.58 (1.46 to 1.69)
PSDQ permissive	2.34 (2.14 to 2.55)	2.28 (2.07 to 2.50)	2.21 (1.98 to 2.43)	2.19 (1.96 to 2.43)	2.11 (1.90 to 2.32)	2.09 (1.86 to 2.33)

There were no significant differences between or within groups for authoritative, authoritarian and permissive parenting style scores over the course of the study (Table 39). Figure 6 presents the correlation of reduction in child BMI z-scores and authoritative parenting style scores at baseline and 12 months. Despite being generally regarded as the most effective parenting style, the correlation was weak (r = –0.142 and r = –0.128 for baseline and 12 months, respectively).

TABLE 39 - Longitudinal changes in parenting style (PSDQ)

FFH, Families for Health; UC, usual care.

Parenting style	Treatment group		p-value
	FFH	UC
PSDQ authoritative
Change from baseline to 3 months (95% CI)	0.033 (–0.115 to 0.182)	0.085 (–0.074 to 0.244)	0.633
Change from baseline to 12 months (95% CI)	0.106 (–0.042 to 0.254)	0.073 (–0.082 to 0.228)	0.756
PSDQ authoritarian
Change from baseline to 3 months (95% CI)	–0.052 (–0.184 to 0.081)	0.011 (–0.076 to 0.097)	0.431
Change from baseline to 12 months (95% CI)	–0.072 (–0.182 to 0.039)	–0.028 (–0.084 to 0.140)	0.204
PSDQ permissive
Change from baseline to 3 months (95% CI)	–0.030 (–0.258 to 0.198)	–0.010 (–0.166 to 0.146)	0.884
Change from baseline to 12 months (95% CI)	–0.205 (–0.391 to –0.016)	–0.147 (–0.360 to 0.065)	0.684

FIGURE 6.

Scatterplot showing the correlation of BMI z-score change from baseline to 12 months with an authoritative parenting style at (a) baseline; and (b) 12 months. FFH, Families for Health; UC, usual care.

Subgroup analysis

The difference in treatment effects between subgroups was initially assessed by interaction tests. These were performed via significance tests of interaction terms in the hierarchical model used for the primary outcome analysis. Each interaction has been fitted separately in a model with the individual fixed effect of this subgroup covariate also included (if it was not already included). Finally, variables that have been categories for subgroup analyses (age, parent BMI) were also investigated as covariates on their original scale. Table 50 shows that none of the interaction effects was significant. Therefore, there is no evidence of a differential treatment effect between prespecified subgroups for any of the subgroup-generating covariates.

The treatment effects in terms of BMI z-scores estimated separately within each subgroup are provided in Appendix 3.

Measurement of resource use and costs

A comprehensive strategy was adopted to estimate the incremental costs associated with the Families for Health programme. This encompassed two broad strands of research: (1) estimation of costs associated with the experimental and usual-care programmes targeted at overweight and obese children; and (2) estimation of broader health and Personal Social Service resource inputs and broader societal resource inputs.

Costing of Families for Health and usual-care programmes

A particular focus of the economic evaluation was the assessment of the cost of delivering the Families for Health programme in community settings, including the costs of programme development, training of facilitators, staff-related expenses, and revenue and capital overheads. This primarily involved asking each of the Families for Health facilitators to complete a detailed proforma outlining the cost of delivering each Families for Health programme, including costs associated with preparation time, programme delivery time, travel, parking, refreshments and miscellaneous expenditures; these costs were subsequently converted into programme-specific estimates of cost per session per attending child using separately collected attendance data. The cost of delivering the alternative usual-care programmes [e.g. One Body One Life,36 Change4Life and Wolverhampton Inspiring and Supporting Health (WISH; Royal Wolverhampton NHS Trust, Child Healthy Lifestyle Services, Wolverhampton, UK),74 see Chapter 2, Usual-care control group] was also estimated using the same primary research methods.

Collection of broader resource-use data

Data were also collected about all health and Personal Social Service and broader societal resource inputs over the 12-month time horizon of the study (i.e. the period between randomisation and 12 months post randomisation). The main parent was asked to complete detailed resource-use questionnaires via researcher-administered interviews at baseline and at 3 and 12 months. The questionnaires administered at 3 months covered the period between baseline and 3 months post randomisation, whereas the questionnaires administered at 12 months covered the period between 3 and 12 months post randomisation. The data collected from the main parent at each time point included their child’s use of hospital care services, community-based health care, community-based social care, and medicines and drugs. Information was also collected regarding educational support, family expenditures and parental lost productivity attributable to the child’s health status, over the relevant time horizons. The use of hospital-based services included hospital inpatient admissions, hospital day-care services, paediatric outpatient department appointments, other outpatient department appointments, and visits to accident and emergency units. The use of community-based health care included contacts with GPs, dentists, opticians, dietitians, psychiatrists, psychologists and general practice nurses. It also included contacts with child development centres, child guidance units, family therapists and individual therapy or counselling sessions. The use of community-based social care included contacts with social workers and after-school activities. A list of all medications prescribed was also included in the resource-use questionnaires and items were identified for inclusion in the economic analyses. Medications were categorised by chemical entity, mode of administration, dosage and duration of use. Educational support included contacts with school nurses, educational psychologists, educational welfare officers and special education needs co-ordinators. Family expenditures accounted for changes in expenditure attributable to the child’s weight-related problems. They also covered direct non-medical costs borne by families, for example travel costs and child-care costs, associated with the study child’s use of health and social care services. Parents were asked to report the number of days they had been absent from work during the relevant time horizons as a result of the child’s weight-related problems. The resource-use questionnaires had been piloted to assess their acceptability and parents’ comprehension levels of the resource-use questions. The categories of cost included in analyses from the NHS and Personal Social Services perspective were hospital care, community-based health care, community-based social care and medication use. Analyses from a societal perspective additionally encompassed educational support, additional family expenditures, lost productivity and travel costs.

Valuation of resource use

Resource inputs were valued using a combination of primary research, based on established accounting methods, and data collated from secondary national tariff sets (NHS Reference Costs 2013/14, Unit Costs of Health and Social Care;61 £, 2013–14 prices). Inpatient admissions over this time horizon were delineated by type and duration, and valued using per-diem costs extracted from the NHS Reference Costs 2013/14. 61 Use of other hospital-based care was valued by applying unit costs extracted from national tariffs. 62 Costs for the community-based services were calculated by applying unit costs from national tariffs to resource volumes. NHS net prices per mg for the medications were obtained from the British National Formulary for Children. 63 Costs for individual children were estimated based on their reported doses and frequencies if these were available, or otherwise on an assumed daily dose based on British National Formulary for Children recommendations. 63 The costs to parents of taking time off work to care for the child(ren) were estimated by applying sex-specific median earnings data64 to occupational classifications derived from self-reported work status information. Other family-borne costs were valued using data reported by the parents as part of the follow-up resource-use questionnaires. Unit costs were inflated, where necessary, to 2013–14 prices (£) using the NHS Hospital and Community Health Services Pay and Prices Index75 or Consumer Price Index. 75 No discounting of costs or benefits was applied as the time horizon was < 12 months. Table 51 shows a list of resource valuations and their sources.

Calculation of utilities and quality-adjusted life-years

The economic evaluation made use of QALYs to measure preference-based health outcomes. The health-related quality of life of the study children was assessed using the EQ-5D-Y51,53 obtained from both parents and children at baseline and 3 and 12 months after randomisation as secondary outcomes of the trial. The standard UK (York A1) tariff values52 were applied to these responses at each time point to obtain health utility scores. QALYs were calculated using linear interpolation between baseline and follow-up utility scores and form the main health outcome measure of the economic evaluation. The baseline analysis was conducted using the self-reported EQ-5D-Y data completed by the child.

Missing data

Multiple imputation was used to impute missing data and avoid biases associated with complete-case analysis. Missing data were a particular issue for costs and health utility scores collected at the 3- and 12-month follow-up time points. Multiple imputation using chained equations78 and predicted mean matching was carried out on the two main outcome measures, the BMI z-score and EQ-5D-Y self-reported by the child, as well as cost estimates, at both the 3- and 12-month follow-ups. Predicted mean matching is a semiparametric imputation approach and generally performs better than linear regression despite the similarities in method. 79 Age group, sex, SES and location were included as explanatory variables in the imputation models. In addition, the baseline BMI z-score was included as an explanatory variable in the models predicting BMI z-scores at the follow-up points; the baseline EQ-5D-Y utility score was included as an explanatory variable in the models predicting EQ-5D-Y utility scores at the follow-up points; and baseline costs were included as an explanatory variable in the models predicting costs at the follow-up points. Five imputed data sets were generated, as this is deemed sufficient to obtain valid responses. 80,81 Siblings within families (n = 24) were treated as clusters in the analyses and reflected in the multiple imputation. Twenty-three per cent (30/128) of the 3-month BMI z-scores and 31% (40/128) of the 12-month BMI z-scores were imputed. With respect to the EQ-5D-Y, 23% (30/128) of the 3-month EQ-5D-Y utility scores and 34% (44/128) of the 12-month EQ-5D-Y utility scores were imputed with the identical method of chained equations and predicted mean matching. Fifteen per cent (20/128) of the 3-month costs and 31% of the 12-month costs were also imputed.

Analyses of resource use, costs and outcome data

Resource-use items were summarised by trial allocation group and follow-up period, and differences between groups were analysed using a t-test for continuous variables and a chi-squared test for categorical variables. Mean (SE) costs by cost category and mean (SE) total costs were estimated by trial allocation group for all time periods. Total costs were estimated from both a NHS and Personal Social Services perspective, and from a broader societal perspective. Cost comparisons were carried out using Student’s t-test. Differences in mean total costs and their CIs were estimated. Non-parametric bootstrap estimates73 based on 1000 replications were also calculated for these differences in mean costs, and their CIs calculated.

In addition, bivariate regression was carried out for both costs and outcomes. These analyses explored the determinants of costs and outcomes using seemingly unrelated regression, and included the prespecified prognostic factors of trial intervention (referent: usual care), child age (referent: 6–8 years), child sex (referent: males), site (referent: site A), family SES (referent: higher managerial) and number of siblings within the trial population (referent: none).

Cost-effectiveness analyses

The main cost-effectiveness analyses were conducted for complete cases (i.e. those with complete cost and outcome data) on an intention-to-treat basis. The cost-effectiveness results were expressed primarily in terms of an incremental cost-effectiveness ratio (ICER). This was calculated as the difference in mean costs divided by the difference in mean outcomes (QALYs or change in BMI z-score between baseline and 12 months) between the trial comparators. The primary analyses took the perspective of the NHS and Personal Social Services. The non-parametric bootstrapping approach was used to determine the level of sampling uncertainty surrounding the mean ICER by generating 10,000 estimates of incremental costs and benefits. These were represented graphically on four-quadrant cost-effectiveness planes. Cost-effectiveness acceptability curves (CEACs) showing the probability that the Families for Health programme was cost-effective relative to usual care, across a range of cost-effectiveness thresholds, were also generated based on the proportion of bootstrap replicates with positive incremental net benefits. Unless otherwise stated, all statements about cost-effectiveness are based on a £20,000 per QALY gained threshold. The probability that the Families for Health programme is less costly or more effective than usual care was based on the proportion of bootstrap replicates that had negative incremental costs or positive incremental health benefits.

Secondary analyses were also conducted where the outcomes remained unchanged from the main cost-effectiveness analyses, but for the costs a wider societal perspective was taken that included broader economic costs.

Sensitivity and subgroup analyses

Several sensitivity analyses were undertaken to assess the impact of areas of uncertainty surrounding components of the economic evaluation. These involved re-estimating the main cost-effectiveness outcomes under the following scenarios: (1) conducting a per-protocol analysis where families having participated in five or more sessions of the Families for Health programme are regarded as ‘programme completers’ (i.e. as having complied with the protocol sufficiently); (2) multiple imputation of all missing cost and outcomes data; (3) parent-reported EQ-5D-Y values for the study child(ren) substituted for child self-reported values in the formulation for QALYs; and (4) incorporation of EQ-5D values reflecting the main parent’s self-reported health within calculations of overall QALYs gained.

Subgroup analyses were conducted for the main cost-effectiveness results to explore heterogeneity in the trial population. These were conducted by (1) age group (6–8 and 9–11 years); (2) sex (males or females); and (3) site (site A, B or C).

Long-term cost-effectiveness model

This study focused on the short- and medium-term costs and consequences of the Families for Health programme in the treatment of overweight and obese children in a community setting. The study protocol82 had allowed for extrapolation of costs and consequences over a longer time horizon if the results had demonstrated a difference in medium-term outcomes. This longer-term modelling would have been based on the natural history of the disease and additional evidence from the literature in the event that the trial yielded significant benefits for the Families for Health programme.

Study population

A total of 128 children were randomised into the Families for Health trial, 63 to the Families for Health programme and 65 to usual care. The 128 children were recruited from 115 families (56 allocated to the Families for Health programme arm and 59 to the usual-care arm). A complete profile of resource use was collected for 94 children at 3 months post randomisation (representing 73.4% of the trial population). A complete profile of resource use was collected for 88 children at 12 months post randomisation (representing 68.8% of the trial population). A complete QALY profile between randomisation and 12 months post randomisation was available for 79 children (representing 61.7% of the trial population).

Resource use

Resource-use measures for the period 3 months pre-randomisation are summarised in Table 52 by trial allocation group. Over-the-counter medication purchases and additional meetings with tutors were significantly higher for the usual-care group compared with the Families for Health group. There were no significant differences between the trial arms across the remaining resource-use categories.

Resource-use measures for the first 3 months post randomisation are summarised in Table 53 by trial allocation group. The mean (SE) number of GP contacts at surgery was 0.70 (0.11) in the Families for Health group, but was significantly lower in the usual-care group [0.28 (SE 0.07); p = 0.002]. The Families for Health group also had more contacts with social workers at home, on average, with a mean (SE) number of contacts of 0.09 (0.04) compared with none for the usual-care group (p = 0.044). There were no significant differences between the trial arms across the remaining resource-use categories. For the period 3–12 months post randomisation, there were no significant differences between the trial arms across all resource-use categories (see Appendix 4).

Attendance rates for the Families for Health programme and a range of usual-care programmes targeted at overweight and obese children during the first 3 months post randomisation are summarised in Table 54. The mean number of contacts with the Families for Health programmes was higher than for comparator usual-care programmes. This difference may be explained, at least in part, by the sources of information for the two comparator arms. The attendance data for the Families for Health programme were based on the recorded attendance information collected as part of the trial. In contrast, attendance at comparator usual-care programmes was based on parent-reported attendance recorded as part of the resource-use questionnaires. Nevertheless, the researchers administering the resource-use questionnaires were trained to prompt parents about their children’s use of a range of services available in their locality. The attendance rates for the comparator usual-care programmes over the period 3–12 months post randomisation are presented in Appendix 5.

Costing of Families for Health and comparator usual-care programmes

The costs of developing the Families for Health programme and training staff to deliver the programme were measured and valued. These costs are presented in Table 55. The costs directly attributable to the delivery of the programme encompassed the value placed on staff preparation time, programme delivery time, travel, parking, refreshments and miscellaneous expenditures. The programme delivery costs over the duration of the trial, broken down by cost category, are presented in Table 55.

The costs of delivering the alternative usual-care programmes were also estimated using comparable primary research methods. Estimates of cost per attending child per programme session are presented for each programme in Table 56. The costs per attending child per programme session were variable among the programmes and ranged from £4.58 for the Weight Watchers programme to £95.94 for Families for Health. This is not entirely surprising because the programmes were not homogeneous in terms of their constituent components, mode or location of delivery, or staff requirements.

Economic costs

Economic costs for the 3-month period pre-randomisation are summarised in Appendix 6 by trial allocation group and cost category. There were no significant differences between the trial allocation groups in any cost category. Economic costs during the first 3 months post randomisation are summarised in Table 57 by trial allocation group and cost category. The mean total costs of GP visits, social worker contacts at home, overweight- and obesity-related programmes, and overall community-based social and other care were significantly higher for the Families for Health group than for the usual-care group. Mean total NHS and Personal Social Services costs, and mean total societal costs were significantly higher for the Families for Health group than for the usual-care group [£688 vs. £385 (p = 0.007) and £792 vs. £459; (p = 0.006), respectively]. For the period between 3 and 12 months post randomisation, there were no significant cost differences in any cost category between the two trial allocation groups (see Appendix 7).

Economic costs from randomisation to 12 months post randomisation are presented by cost category in Table 58 for the Families for Health and usual-care groups. The costs are presented for complete cases (i.e. for those children for whom cost data were available at all follow-up periods). Mean total social and other non-health community care costs, total NHS and Personal Social Service costs, and total societal costs were significantly higher for the Families for Health group than for the usual-care group. There were no significant differences in mean total hospital care costs, community health-care costs, medication costs, educational support costs, family expenditures, parental lost earnings and travel costs between the trial allocation groups. Mean (SE) total NHS and Personal Social Service costs were estimated at £998 (£72) for the Families for Health group, compared with £548 (£73) for children in the usual-care group. The mean NHS and Personal Social Service cost difference was £450 (bootstrap 95% CI £249 to £650; p < 0.001). Similarly, mean (SE) total societal costs were estimated at £1311 (£100) for the Families for Health group, compared with £882 (£112) for children in the usual-care group. The mean societal cost difference was £429 (bootstrap 95% CI £110 to £713; p = 0.005).

Analyses of incremental costs and incremental health outcomes

A bivariate regression, in the form of a seemingly unrelated regression, was carried out with the view to estimating the incremental costs and incremental health outcomes associated with the Families for Health programme. The results of this bivariate regression are presented in Table 59. The adjusted incremental costs associated with the Families for Health programme were £483.63 when NHS and Personal Social Services costs were considered and £410.64 when societal costs were considered. These incremental costs were statistically significant at the 0.1% and 1% levels, respectively. Children from all comparator socioeconomic backgrounds incurred higher adjusted costs, regardless of study perspective, than the reference category (highly managerial). Children with siblings in the trial population incurred lower adjusted NHS and Personal Social Services costs and lower adjusted societal costs than children without siblings in the trial population, although these cost decrements were not statistically significant. Similarly, children aged 9–11 years incurred lower adjusted NHS and Personal Social Services costs and lower adjusted societal costs than children aged 6–8 years, although these cost decrements were not statistically significant. Regarding health outcome measures, the Families for Health programme was associated with a mean increment in QALYs of 0.04 (95% CI –0.04 to 0.11), but also a difference in longitudinal change in BMI z-score of 0.11 (95% CI 0.00 to 0.22). However, it should be noted that these results were based on 88 children with complete data for all covariates.

Cost-effectiveness

The incremental cost-effectiveness of the Families for Health programme is shown in Table 60 for the children with complete costs and health outcomes data over the trial follow-up period, by study perspective and health outcome. When a study perspective of the NHS and Personal Social Services was adopted (i.e. that adopted for the baseline analysis) and health outcomes were measured in terms of QALYs, the average total cost was £1019 in the Families for Health group compared with £507 in the usual-care group, generating a mean incremental cost of £512. The mean incremental cost-effectiveness of the Families for Health programme was estimated at £552,175 per QALY gained. The bootstrapped mean ICERs fell in the north-east and north-west quadrants of the cost-effectiveness plane (Figure 7). The CEAC shown in Figure 7 indicates that regardless of the value of the cost-effectiveness threshold, the probability that the Families for Health programme is cost-effective does not exceed 40%. If decision-makers are willing to pay £20,000 for an additional QALY, the probability that the Families for Health programme is cost-effective is approximately 28% (see Table 60). Broadening the study perspective to that of society as a whole had little effect on these cost-effectiveness results. In particular, the mean ICER remained relatively static at £559,115 per QALY gained and the probability that the Families for Health programme is cost-effective at a £20,000 cost-effectiveness threshold remained unchanged at 28% (Table 60, Figure 8).

FIGURE 7.

(a) Cost-effectiveness plane; and (b) CEAC for QALY outcome: complete cases from a NHS and Personal Social Services perspective. Adapted from Robertson W, Fleming J, Kamal A, Hamborg T, Khan KA, Griffiths F, et al. Randomised controlled trial and economic evaluation of the ‘Families for Health’ programme to reduce obesity in children [published online ahead of print 21 December 2016]. Arch Dis Child 2016. doi: 10.1136/archdischild-2016-311514. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited (see http://creativecommons.org/licenses/by/4.0/). 71

FIGURE 8.

(a) Cost-effectiveness plane; and (b) CEAC for QALY outcome: complete cases from a societal perspective.

When a study perspective of the NHS and Personal Social Services was adopted and health outcomes were measured in terms of longitudinal change in BMI z-score, the average total cost was £998 in the Families for Health group (for those with complete data), compared with £548 in the usual-care group, generating a mean incremental cost of £450. The mean incremental cost-effectiveness of the Families for Health programme was estimated at –£3935 per unit change in BMI z-score. The bootstrapped mean ICERs largely fell in the north-west quadrant of the cost-effectiveness plane (Figure 9), indicating that, on average, the Families for Health programme is dominated by usual care in health economic terms. The CEAC shown in Figure 9 indicates that regardless of the value of the cost-effectiveness threshold, the probability that the Families for Health programme is cost-effective does not exceed 2% (on the basis of the BMI z-score). If decision-makers are willing to pay £20,000 per unit change in BMI z-score, the probability that the Families for Health programme is cost-effective is < 1% (see Table 60). Broadening the study perspective to that of society as a whole had little effect on these cost-effectiveness results. In particular, the mean ICER remained relatively static at –£3748 per unit change in BMI z-score and the probability that the Families for Health programme is cost-effective at a £20,000 cost-effectiveness threshold remained unchanged at < 1% (Figure 10).

FIGURE 9.

(a) Cost-effectiveness plane; and (b) CEAC for BMI z-score outcome: complete cases from a NHS and Personal Social Services perspective.

FIGURE 10.

(a) Cost-effectiveness plane; and (b) CEAC for BMI z-score outcome: complete cases from a societal perspective.

Sensitivity analyses

Several sensitivity analyses were undertaken to assess the impact of uncertainty on the cost-effectiveness results. Table 61 presents our recalculations of cost-effectiveness estimates when a per-protocol analysis was performed. These analyses defined ‘programme completers’ as families who participated in five or more sessions of the Families for Health programme and non-completers as families who participated in less than five sessions of the Families for Health programme. These analyses were also restricted to the baseline NHS and Personal Social Services perspective. Of particular note is that for programme completers, the mean incremental cost per QALY gained attributable to the Families for Health programme declined to £27,790 and the probability that the programme is cost-effective at a £20,000 cost-effectiveness threshold increased to 43%. For comparison, the mean incremental cost per QALY gained among non-completers was –£6441 (indicating that usual care is dominant in health economic terms) and the probability that the programme is cost-effective at a £20,000 cost-effectiveness threshold was 17% (Figures 11–14).

FIGURE 11.

(a) Cost-effectiveness plane; and (b) CEAC for QALY outcome and per-protocol analyses: completers.

FIGURE 12.

(a) Cost-effectiveness plane; and (b) CEAC for QALY outcome and per-protocol analyses: non-completers.

FIGURE 13.

(a) Cost-effectiveness plane; and (b) CEAC for BMI z-score outcome and per-protocol analyses: completers.

FIGURE 14.

(a) Cost-effectiveness plane; and (b) CEAC for BMI z-score outcome and per-protocol analyses: non-completers.

Table 62 presents our recalculations of cost-effectiveness estimates following multiple imputation of all missing cost and outcomes data (i.e. estimates are calculated for all 128 study children). Of particular note is that following multiple imputation, the mean incremental cost per QALY gained (assuming a NHS and Personal Social Services perspective) attributable to the Families for Health programme declined to £9119 and the probability that the programme is cost-effective at a £20,000 cost-effectiveness threshold increased to 67%. Similar results were observed when a societal perspective for costs was adopted. In contrast, when the BMI z-score was considered as the health outcome measure, the Families for Health programme remained dominated by usual care in health economic terms and the probability that the programme is cost-effective did not exceed 6%, regardless of the value of the cost-effectiveness threshold (Figures 15–18).

FIGURE 15.

(a) Cost-effectiveness plane; and (b) CEAC for QALY outcome: imputed analyses from a NHS and Personal Social Services perspective. The cost-effectiveness plane is based on 1000 bootstrap replicates of the data set.

FIGURE 16.

(a) Cost-effectiveness plane; and (b) CEAC for QALY outcome: imputed analyses from a societal perspective. The cost-effectiveness plane is based on 1000 bootstrap replicates of the data set.

FIGURE 17.

(a) Cost-effectiveness plane; and (b) CEAC for BMI z-score outcome: imputed analyses from a NHS Personal Social Services perspective. The cost-effectiveness plane is based on 1000 bootstrap replicates of the data set.

FIGURE 18.

(a) Cost-effectiveness plane; and (b) CEAC for BMI z-score outcome: imputed analyses from a societal perspective. The cost-effectiveness plane is based on 1000 bootstrap replicates of the data set.

Finally, Table 63 presents our recalculations of cost-effectiveness estimates when alternative sources and inputs for EQ-5D utility values were incorporated into the analyses. Substituting parent-reported EQ-5D-Y values for the study child(ren) for child self-reported values in the formulation for QALYs removed the incremental QALY benefit associated with the Families for Health programme. The probability that the programme is cost-effective at a £20,000 cost-effectiveness threshold declined to 23% when a NHS and Personal Social Services perspective was adopted, and to 25% when a societal perspective was adopted. Furthermore, the reduction in incremental QALYs associated with the Families for Health programme increased when EQ-5D values reflecting the main parent’s self-reported health were also incorporated within calculations of overall QALYs gained. The probability that the programme is cost-effective at a £20,000 cost-effectiveness threshold declined to 2% (see Table 63 and Figures 19–22).

FIGURE 19.

(a) Cost-effectiveness plane; and (b) CEAC for QALY outcome: proxy-reported EQ-5D-Y values, NHS and Personal Social Services perspective.

FIGURE 20.

(a) Cost-effectiveness plane; and (b) CEAC for QALY outcome: proxy-reported EQ-5D-Y values, societal perspective.

FIGURE 21.

(a) Cost-effectiveness plane; and (b) CEAC for QALY outcome: incorporation of EQ-5D values for parent’s health, NHS and Personal Social Services perspective.

FIGURE 22.

(a) Cost-effectiveness plane; and (b) CEAC for QALY outcome: incorporation of EQ-5D values for parent’s health, societal perspective.

Subgroup analyses

Several subgroup analyses were conducted to explore the heterogeneity in our cost-effectiveness results. These are summarised in Appendix 8. See Table 93, which presents the cost-effectiveness results by subgroup when a NHS and Personal Social Services perspective is adopted for costs, and see Table 94 which presents the cost-effectiveness results by subgroup when a societal perspective is adopted. The subgroups considered in our analyses were (1) age group (6–8 or 9–11 years); (2) sex of child (male or female); and (3) site (site A, B or C). The cost-effectiveness results for two particular subgroups of children are worthy of comment. First, when the QALY measure was considered as the primary health outcome measure, the probability of cost-effectiveness of the Families for Health programme was notably higher for girls than for boys: 67% versus 15% at a £20,000 cost-effectiveness threshold from a NHS and Personal Social Services Perspective and 63% versus 16% at a £20,000 cost-effectiveness threshold from a societal perspective. Second, when the QALY measure was considered as the primary health outcome measure, the probability of cost-effectiveness of the Families for Health programme was notably higher in site A than in site B or C: 61% versus 11% versus 36% at a £20,000 cost-effectiveness threshold from a NHS and Personal Social Services perspective, and 64% versus 7% versus 35% at a £20,000 cost-effectiveness threshold from a societal perspective. These patterns were not replicated for the BMI z-score outcome measure.

Design

Table 64 shows the topics covered by the process evaluation, together with a definition of the component in the context of this trial and the associated research question.

Three members of the research team (WR, JF and AK) collected both quantitative and qualitative data from parents, children and facilitators. Table 65 shows the data used to assess each component. Interviews with families took place at baseline, and at the 3- and 12-month follow-ups. In addition, weekly feedback forms during the seven Families for Health programmes, attendance registers, focus groups with facilitators and hours worked by facilitators were also collected. Follow-up family visits were conducted by a researcher who was blinded to treatment allocation.

Data collection methods

This section describes the data collection methods that were used solely for the process evaluation, grouped by the five components. Other data used were collected as described in Chapter 2.

Data management and analysis

Recruitment

There were 194 families who expressed interest in the trial and were sent further information. Of those, 115 (59.3%) went on to be recruited and randomised (Figure 23 and Table 68). Active recruitment yielded 85 potential participants, with 43 recruited; passive recruitment yielded 99 potential participants, with 72 recruited (Table 69). A higher proportion of enquires from families recruited passively resulted in recruitment to the study (72.7% vs. 50.6%; p = 0.002). The most productive active method of recruitment was a referral from a health professional, with 69 ‘expressions of interest’ resulting in 30 families being recruited to the study. Information seen at school or the GP surgery accounted for 55 enquiries and over one-third (37.4%) of final recruitment. Media-generated enquiries were few, but all nine resulted in recruitment. Further details of the recruitment to the trial via passive and active methods have been published, alongside examples of the materials. 93

FIGURE 23.

Flow chart of participant recruitment to the FFH study. FFH, Families for Health; UC, usual care.

Reach

The intended audience for this trial was families with an overweight or obese child aged 6–11 years. All participating children were overweight or obese, mostly obese (Table 70). There was no significant relationship between site and the proportion of obese and overweight children (p = 0.420), but there was a significant difference in baseline BMI z-scores, with site C having the highest mean BMI z-score [site A: mean 2.57 (SD 0.62); site B: mean 2.64 (SD 0.66); site C: mean 2.99 (SD 0.71); p = 0.010]. Asian, black and mixed ethnicity children were a higher proportion of the children recruited than would have been expected from census data for the three sites. However, the prevalence of obesity is higher for Asian, black and mixed ethnicity children in England in the NCMP. 3

The families recruited to the trial approximately reflected the local population for SES. By contrast, it was clear that a higher proportion of single-parent families were recruited than would have been expected from 2011 census data88 (site A, 46% recruited vs. 9.6% in the census; site B, 25.8% vs. 6.1%; site C, 44.1% vs. 9.4%). In addition, the number of families in which the parents were unemployed was higher than expected from the census data (site A, 16% vs. 7.4%; site B, 6.5% vs. 3.4%; site C, 32.4% vs. 10%).

Dose delivered

Seven Families for Health programmes were delivered: three of them ran for the full 10 weeks and four for 9 weeks. Session cancellations were because of bad weather (two programmes), shortage of facilitators (one programme) and poor attendance (one programme).

Fidelity

As described in Methods, a number of fidelity checks were carried out. The programmes varied in the extent to which they were delivered as prescribed (Table 72). Three Families for Health programmes (site A, FFH1, FFH2; site B, FFH1) had only very minor changes, including adapting songs and games for children so that they were age-appropriate and giving parents as well as children pedometers.

Partial adjustments to programme delivery included incorporating activity taster sessions across the entire programme instead of during weeks 7–9 for one Families for Health programme (site A, FFH3). In one programme the parent and child sessions were combined because of poor attendance and one programme added visual aids to support content delivery (site C, FFH1).

Two Families for Health programmes varied the way in which programme content was delivered. One programme (site C, FFH2) adapted content delivery to include a theory hour followed by a more structured physical activity hour because there was a gym available on the site. This adjustment to programme content appeared to work well:

Gym sessions have enhanced the weight loss and motivated the group lots.

Facilitator weekly evaluation form, site C, FFH2, week 6

Another programme (site B, FFH2) adjusted delivery because of poor attendance, by combining parent and child sessions and facilitating whole-group discussions instead of small-group activities. Activity taster sessions were also brought forward in this programme in the hopes of improving attendance, but this failed.

All facilitators who delivered the Families for Health programme attended the 4-day training programme, with the exception of one facilitator who joined the programme as a facilitator after the training workshop had been delivered. The training they received was perceived positively (see Appendix 1). Almost all facilitators delivered more than one Families for Health programme and facilitators’ confidence increased as they became more experienced in the delivery of Families for Health programmes (site A, FFH2, FFH3; site B, FFH1, FFH2; site C, FFH2). They found they could modify games easily and avoid issues they encountered previously:

It seemed to run a lot smoother this time than it did last time, and I think because that we had so many negatives the last time we’ve kind of turned it into a positive this time.

Site C, FFH2, week 10

The Families for Health programme is designed to be delivered in a group setting with a minimum of eight families not waiting more than 3 months to join a programme, and four facilitators (two for parents and two for children) delivering each session. This was not always achieved (Table 73). Three Families for Health programmes (site B, FFH1 and FFH2; site C, FFH2) could not be delivered as intended because of small numbers of families starting the programme and subsequent poor attendance. Two Families for Health programmes (site B, FFH1 and FFH2) lacked continuity of facilitators and one session (site B, FFH1) had to be cancelled because of a lack of facilitators.

Delivering the programme as a single facilitator for the parents group for one programme that had a small number of families restricted interaction with families as the facilitator was too busy delivering the programme activity to attend to other issues, such as families that were experiencing problems, late arrivals or preparing for the next activity (site B, FFH1).

Of the 56 families that were randomised to the Families for Health intervention, 21 (37.5%) waited more than 3 months before starting a Families for Health programme (see Table 73). Other issues that were encountered included external service providers for the physical activity tasters undermining the message of the programme by rewarding children with chocolate (site B, FFH1), room booking issues at one venue, resulting in there being no suitable rooms to deliver one of the children’s sessions (site A, FFH3), shorter activity sessions because of the late arrival of the coach for one of the activity tasters (site A, FFH3) and cancelling one session because of snow (site A, FFH2; and site C, FFH1).