Changing eating behaviours to treat childhood obesity in the community using mandolean: The community mandolean randomised trial (COMMANDO), A pilot study

Current evidence base for weight loss programmes

Current weight management programmes for children are directed at multicomponent interventions addressing healthy nutrition and improved levels of physical activity. Unfortunately, the evidence for the clinical effectiveness of such programmes remains scarce. The Cochrane meta-analysis of lifestyle interventions for obesity in children under 18 years old, including behavioural interventions, identified a mean difference in body mass index standard deviation score (BMI z-value) above ‘standard or minimal intervention’ of –0.04 [95% confidence interval (CI) –0.12 to 0.04], at termination of the intervention (12 months). 4 Such changes are unlikely to be clinically or metabolically advantageous to improving the health of an obese child. These statistically significant, but clinically unimportant, changes mirror data not included in the Cochrane review for community-based interventions, such as the Watch it and Lifestyle Education for Activity Program (LEAP) studies. 7,8 The LEAP study, the only randomised controlled trial (RCT) published prior to the commencement of the Community Mandolean RCT (ComMando) on primary care management of childhood obesity, found that the intervention did not achieve sustained weight reduction. 8

There is limited evidence that family-based approaches to weight management in childhood are more likely to be effective than those aimed at the child alone. 4 However, there is strong epidemiological evidence that children living in a family with an obese parent are more likely to develop obesity themselves and, thus, an intervention aimed at improving the weight of both obese child and parent seems a logical strategy to more fully engage a family in fundamental changes to lifestyle behaviour. 9,10

Speed of eating and its relation to obesity

There is a growing body of evidence that eating behaviours other than simple nutritional decision processes may be a determinant of obesity. Studies in adults and children suggest that increased speed of eating is associated with an increased risk of obesity. 11,12 Theoretically, eating food swiftly leads to a state of relative ‘satiety unresponsiveness’ by which an individual fails to respond to normal satiety signals such as gastric distension and gut peptide release. 13

Mandometer/Mandolean: a new device for weight management

Mandolean^® [previously Mandometer^® (Mikrodidakt AB, Lund, Sweden)] is a novel weight management method. It teaches patients how to eat and recognise hunger and satiety with the help of a small computer that receives information from a small scale beneath the patient’s plate of food. The Mandolean system allows the patient to see a rate of eating displayed on the screen that describes the rate at which normal individuals eat that amount of food and feel satiety as they eat. At the same time, the patient’s own eating speed and perception of satiety is shown on the screen. The patient then gradually learns to model his or her pattern of eating to a more normal and slower pattern of eating.

Evidence base for Mandolean

In a pilot study of obese adolescents referred to a hospital, we were able to demonstrate that speed of eating was faster than for normal-weight adult controls and that it was possible to slow down speed of food consumption using a Mandolean. 14 On the basis of these pilot data, we undertook a hospital-based, randomised trial of Mandolean in adolescents (aged 9–17 years), eliciting a mean BMI z-value reduction of –0.4 (95% CI –0.30 to –0.51) and a mean difference of 0.27 (95% CI 0.14 to 0.41) compared with standard care. Average self-determined portion size in the Mandolean group reduced at 12 months post randomisation by 45 g (95% CI 7 to 84 g) with maintenance of satiety compared with baseline. Importantly, the overall BMI z-value reduction was maintained 6 months after termination of therapy. 15 This device proved as effective as an adjunct to standard lifestyle modification in treating obesity in adolescents as pharmacotherapy, with maintenance of benefit post therapy which had not been described for any other weight loss intervention in children.

Pilot study objectives

To develop the infrastructure to deliver a full trial including:

1. establishing recruitment methodologies and ensuring strategies were successful in terms of rate of recruitment

2. engaging three hub practices to deliver treatment and a number of spoke general practices to feed into the hubs

3. training nurses at the three hubs to deliver treatment

4. developing a protocol for treatment delivery

5. refining trial paperwork and study design

6. testing the acceptability of the Mandolean intervention for families

7. examining patient and parental views and experience of treatment

8. testing adherence to Mandolean therapy

9. testing adherence to regular consultation attendance.

Success of the pilot study was to be determined by meeting the following criteria:

1. recruitment of at least 36 families who would be eligible for the main study

2. 90% of patients randomised to Mandolean would be successfully eating off the device at least five times a week

3. at least 60% of those using Mandolean would demonstrate a decrease in speed of meal consumption (longer meals) since baseline within 3 months of starting therapy

4. at least 80% would attend the 3-month nurse appointment for both study groups.

Main trial objectives

The main trial objectives were as follows:

1. to determine if Mandolean therapy could be delivered in the community by trained nurses in order to obtain a BMI z-score improvement at least 0.25 greater than standard care in obese children

2. to examine the clinical effectiveness of Mandolean therapy in obese adults/parents

3. to explore the effect of parental usage of Mandolean on child and parental weight loss

4. to examine the cost-effectiveness of Mandolean therapy compared with standard care

5. to assess patient and practitioner experience and acceptability of Mandolean and standard care

6. to develop a ‘toolkit’ for future users of this technology in other health-care settings.

General practitioner-initiated recruitment

General practitioners who identified a child as a potential participant were asked to inform the child and their parents of the trial. If the patients were interested in participating, the GP then completed a screening form to determine if the child met the required eligibility criteria. If the child was eligible to take part, the GP asked the parents to sign a ‘permission for researcher contact’ form to confirm that they were happy for their contact details to be released to the research team. The GP also sent the researchers the completed screening information. If the child was not eligible, the GP explained this to the child and his or her parents and an anonymised screening information sheet was sent to the research team. As soon as an eligible patient had been referred, the researcher mailed a copy of the adult information sheet and age-appropriate child information sheet to the potential participant’s home address. Within 1 week of mailing out the information sheet, the researcher contacted the patients to answer any questions they had about the study and to arrange a recruitment appointment, which took place in the patient’s home. If the patient did not attend the appointment or was no longer interested in participating after considering the patient information sheet, a ‘referral outcome’ form indicating this was sent to the patient’s GP. Patients who decided not to take part were asked for their reason for declining, if they were willing to provide this. The researcher made clear that they were under no obligation to provide a reason if they preferred not to. They were also asked if they would be willing to take part in a short telephone interview with a qualitative researcher, to explain their reason for declining in more detail. If they agreed, their contact details were retained. If not, they received no further contact from the research team.

Patients who attended a recruitment appointment had the opportunity to discuss the study further with the local researcher, who went through the patient information sheet with them and obtained consent from the parent for the baseline height and weight measurements of the child to be taken to recheck eligibility. If the child was still eligible, the researcher obtained consent from the parent for them and the child to enter into the trial. Consenting patients were then asked to complete a baseline questionnaire. The local researcher also provided the participants with two ‘blinded’ Mandoleans, which recorded portion size and eating rate, but did not display this information on the screen. The researcher demonstrated how to use the device and asked both parent and child to use them for one meal a day for a week (this was subsequently reduced to 3 days as the families found this task onerous). An appointment to see the practice nurse for treatment to begin was arranged for the end of the baseline data collection week. Participants were asked to return the Mandolean at this appointment. At the initial nurse appointment, the nurse used an online randomisation system to determine treatment allocation while with the participants. The nurse was then able to inform the participants immediately of their allocation and arrange for the appropriate treatment to begin. The outcome of the recruitment and initial nurse appointments were sent to the GP along with details of where the patient was receiving treatment if they had been recruited into the study.

Health-care professional-initiated recruitment

Health-care professionals (HCPs) based at the practice that saw children potentially eligible for the trial could also refer. They were asked to introduce the trial to the child and parent and ask the parent to complete the ‘permission for researcher contact’ form. They also completed the screening and referral form and faxed both the ‘permission to contact’ form and the referral form to the local researcher. The local researcher then let both the GP and the HCP know of the outcome of the referral.

Recruitment through advertisements

Advertisements through various media, including posters in GP practice waiting rooms and a schools magazine (Primary Times), were used to aid recruitment. These advertisements invited patients to approach their GP about the trial or to get in touch directly with the research team. Patients responding to GP waiting room posters who contacted their GP were referred to the study as per routes 1 and 2 (GP- or HCP-initiated recruitment). Potential participants who contacted the research team directly, were briefly screened by telephone to ensure their registered GP practice was within the BNSSG area and that the child was within the correct age range for the study. The parent was asked if they were aware of the child meeting any of the exclusion criteria. The researcher explained that they would still need to be screened for their child’s BMI when they first met with the child, and explained that there was a threshold BMI level which the child must be over before they could enter the trial, as we were unable to offer the intervention to all children.

If, following screening, the patient was likely to be eligible, the researcher arranged a time to meet with the child and their parent to talk further about the study. The potential participants were asked for their contact details and were sent the adult patient information sheet, and an age-appropriate children’s information sheet.

The research meeting would then continue as per recruitment routes 1 and 2. If the patient was eligible and consented to take part, a letter was sent to the participant’s GP informing them that their patient was taking part in the study. The GP letter asked the GP to contact the study team if they knew of any reason the patient should not take part, with specific emphasis on the study’s exclusion criteria.

Record screening

Where there was agreement from the GP, practice staff were asked to review patient records to identify any potential participants from all families known to have children within the age range of the study. All families with children in the appropriate age range were contacted by post. They were sent an invitation letter, brief information sheet, and a ‘permission for researcher contact’ form. On return of a ‘permission for researcher contact’ form, the researcher contacted the potential participants and recruitment continued as per route 3 (recruitment through advertisements).

Recruitment through schools

School nurses in local primary schools were contacted to help identify potential participants. School nurses could send information about the trial to parents of potential participants who were encouraged to contact the study team for enrolment into the trial, as detailed previously in Recruitment through advertisements.

Inclusion

The target population was obese children aged 5–11 years with a BMI ≥ 95th percentile [definition of obesity in National Child Measurement Programme (NCMP)].

Exclusion

Parents were required to read and understand information written in English for consent purposes, data collection and for treatment. Therefore, children whose parents were unable to read English were excluded from the study.

General practitioners used an adapted screening tool to identify the need for the child to consult a secondary care paediatrician. This tool was previously developed for the Research for Patient Benefit (RfPB) trial18 (see Appendix 1 ).

Red flag indicators (i.e. need for secondary care evaluation) were:

1. possible genetic cause for obesity – learning difficulties, visual or hearing difficulties, dysmorphic features

2. possible associated endocrine disorder – weight and height disproportionate (height > 50th percentile), features of delayed or precocious puberty, features of Cushing’s syndrome

3. possible comorbidity – recurrent severe headaches (Idiopathic Intracranial hypertension), features of sleep apnoea syndrome, polyuria or polydypsia (type 2 diabetes)

4. features of an overt eating disorder – history of suggestive of bulimia

5. iatrogenic causes of obesity – cranial surgery, anticonvulsant therapy.

Standard care

Participants allocated to the control arm received a package of care specifically tailored for the treatment of childhood obesity in a primary care setting. This was referred to as standard care, though this reflected an enhanced level of care to that usually received in primary care settings.

In standard care, emphasis was placed on implementing changes to increase levels of enjoyable physical activity to national recommended levels (60 minutes’ exercise a day for children) alongside a balanced diet, based on the ‘Eatwell Plate’. 19 Families were encouraged to set their own dietary goals and targets, with practical advice and guidance from the practice nurse. In encouraging activity, the approach was one of facilitation rather than prescription. Motivational interviewing techniques were used to engage patients and families in the decision-making process for lifestyle changes which were consistent with self-determination principles and, therefore, more likely to lead to responsibility for long-term change. 20

Standard care appointments were delivered at 3-monthly intervals over a 12-month period, resulting in five appointments in total. In addition, three supportive telephone calls were provided to help patients to engage in behaviours discussed in the face-to-face sessions.

Experimental intervention

Participants allocated to receive the Mandolean intervention received the same standard care package as the control group with the addition of Mandolean therapy. Mandolean therapy required the patient to attend appointments with the nurse specialist fortnightly for the first 2 months of treatment. This meant that there were four additional Mandolean appointments, resulting in a total of nine appointments and three telephone calls for patients allocated to the intervention arm.

What is Mandolean therapy?

Mandolean is a portable weighing scale connected to a small computer which can generate a graphical representation of food removal from the plate with weight of food (grams) on the y-axis and time (minutes) on the x-axis. The patient puts a measured portion of food determined by a therapist on the scale and the computer records and displays, in real-time graphics, the removal of food from the plate as the patient eats; time zero on the graph effectively displaying total portion size. Removing food from the plate generates a gradually developing, red line on screen which is visible to the patient and can be compared and matched to a preset eating line on screen displaying the speed at which the therapist wants the patient to eat. Deviation from the training line by eating too quickly or slowly elicits a spoken request from Mandolean to slow down or eat faster. At regular intervals, a rating scale appears on the monitor of the computer and the patient rates their level of fullness (satiety): from 0 (no satiety) to 100 (maximum satiety). Patient-rated satiety appears as a dot on screen yielding a ‘development of satiety’ curve allowing comparison of the development of fullness to a ‘normal’ fullness curve again preset on screen. During ‘Mandolean training’ the patient gradually adopts a more normal pattern of eating and satiety by following these training lines and curves ( Figure 1 ).

FIGURE 1.

Representation of graphics seen on a Mandolean screen during training. (a) The blue training line is presented to patient to follow while eating. (b) The green line develops as food is consumed. In this case, food is being consumed too quickly, so the computer would encourage the patient to eat a little more slowly to approximate the green line to the training line. (c) At regular intervals during the meal, the patient is asked by the computer screen to rate satiety. (d) Dark green satiety ratings are plotted by the computer on to the screen. The patient learns to associate feeling of ‘fullness’ with the dark green S-shaped satiety training line. When eating quickly, as in this case, satiety is rated low. Note: no numerical values are displayed on the axes during training.

Mandolean training

During week 1, the child and parent were trained in how to use the Mandolean. This was a stepped process and is outlined below.

Step 1: the practice nurse used the ‘blind’ (no visual or oral feedback provided to participant) Mandolean data collected in the previous week to assess baseline food intake, rate of eating and satiety. Participants’ eating behaviours were then compared with previously established normal eating patterns. 14 Individualised training lines for portion size, eating speed (reflecting weight removal from plate on y-axis) and satiety were then programmed on to the participant’s Mandolean.

Step 2: the participants were encouraged to eat at least one cooked meal (usually evening meal) a day from their Mandolean, matching as closely as possible their red eating line to the current preset ‘optimal’ eating line.

Step 3: participants received up to four new training lines over 3 months, effectively reducing total portion size and slowing down food removal from the plate in order to ‘normalise’ portion size, eating rate and satiety response. The treatment goals aimed for the patient to feel ‘full’ after eating 300–350 g of food over 12–15 minutes. Normal-weight healthy volunteers stop eating and rate their level of satiety at about 50–60 on the Mandolean rating scale (random units: 0 having no satiety and 100 being completely satiated) after eating this amount of food in this period of time. 21

Qualitative study

A qualitative study was designed alongside the pilot trial. It entailed holding interviews with families randomised to the Mandolean intervention arm, and with individuals who had been approached about the trial but who had declined to take part.

Thirteen families randomised to the intervention arm were interviewed. These families were to be interviewed at three time points throughout the 12 months they were receiving treatment. The first interview was conducted within the first 2 weeks of the family receiving the Mandolean. The second interview occurred 4 weeks later (when the Mandolean training line should have been established) and the last interview took place at around 10 weeks (when the study parent and child should have been eating to an established training line for a month). The purpose of the interviews was to explore the views and experiences of families who were using the Mandolean and receiving standard care. Five patients who declined to take part in the study were interviewed to identify possible barriers to participation. These patients were provided with a ‘decliner’ form by their GP or this form was mailed to them by a member of the study team. The form asked the participants for details of their reason for declining to take part and also asked if they would be willing to be contacted by a researcher for a brief telephone interview to explore these reasons in more detail. Patients were asked to return the form directly to the research team using a prepaid envelope. Patients did not have to release their contact details and could choose to complete the form anonymously if they did not want to be contacted further.

Pilot trial outcomes

The outcomes for the pilot study consisted of a series of process measures: recruitment of practices, families and training nurses; adherence to protocol in terms of attending appointments; and altering Mandolean training lines.

A further objective of the pilot trial was to review the feasibility and acceptability of the secondary outcome measures for use in the main trial (see Main trial outcomes).

Main trial outcomes

The primary outcome measure was the child’s BMI – accurate height and weight – converted to BMI z-values at 12 months. In weight research in children it is important to use BMI z-values. A z-value is defined as the number of standard deviations (SDs) above or below the mean, determined using age- and gender-matched population reference data, and is calculated as (x – mean)/SD, where x is the case value and the mean and SD are derived from control data. 25 The expected mean z-value for a normal population is 0. It is vital to evaluate where an individual lies on the age–gender distribution of BMI curves rather than using absolute values for BMI such as used in adult studies, as children grow both linearly and in terms of weight at differing rates throughout normal childhood.

All height and weight measures for the trial were obtained at the recruitment location and follow-up visits, usually in the child’s home, by a Criminal Records Bureau-checked trained researcher. Weight was measured without shoes in light clothing to the nearest 0.1 kg, using a portable Tanita floor scales (WB 100 S MA, Tanita Europe BV, the Netherlands). The scales were calibrated on a quarterly basis. Height was measured without shoes to the nearest 0.1 cm, using a Seca Leicester stadiometer (Seca, UK).

The secondary outcome measures were:

1. Height and weight of parents of index child converted to BMI for adiposity analysis.

2. Maintained BMI or BMI z-value improvement at 12 months post therapy (24 months after study entry).

3. Quality-of-life measures in child [Pediatric Quality of Life Inventory (PedsQL),26 Child Health Utility 9-Dimensions (CHU9D)27 and European Quality of Life 5-Dimensions – youth (EQ-5D-Y)28] and parents [European Quality of Life 5-Dimensions (EQ-5D)]29 for self-completion at 0, 3, 6, 9, 12 and 24 months.

4. Resource-use questionnaire, including child’s use of primary and secondary care services, to be completed by parent at 3, 6, 9, 12 and 24 months, to be used to inform economic analysis. These were to be given out at scheduled nurse appointments or posted with a prepaid envelope for return to the research team should the participant withdraw from treatment and, therefore, no longer be attending nurse appointments.

5. Change in eating speed and self-determined portion size in ‘blinded’ test meals for both child and parent at 0, 12 months (end of intervention) and 24 months (12 months post intervention). These were to be measured using a ‘blind’ Mandolean, which acted solely as a measuring device and did not provide any feedback on eating rate or portion size choice. Blind Mandoleans were issued at the same time as the anthropometric measurements were taken and collected a week later by the researcher. Parents and children were instructed how to use the blind Mandolean by the researcher. They were asked to eat three meals using the device over the course of 1 week.

6. Precise measures of changes in ‘ideal portion size’ and ‘expected satiety levels’ across a range of commonly consumed foods for both child and parent were to be compared between treatment groups at 0, 12 and 24 months. These were measured using a computer-based program that allows the child or parent to manipulate a photograph of a meal to increase or decrease the volume of food on the plate to represent the portion size they would choose. The measure was presented on a laptop in the patient’s home. The meal photographs were advised by a paediatric dietitian as foods likely to be consumed by children in that age range.

7. Changes in physical activity levels, measured as number of steps per day for 1 week, were collected at baseline using the New Lifestyles NL-800 pedometers (New Lifestyles Inc., MO, USA) which store 7 days of step-count data for a sample of children taking part in the pilot study (this was expected to be measured at 12 and 24 months also in the main trial). Children were instructed to wear the pedometer during the day for 7 consecutive days (except when bathing or swimming) and the pedometers were sealed so that the participant received no feedback on number of steps taken per day. The pedometers were issued at the same time as the anthropometric measurements and it was intended that they would be collected a week later by the researcher.

During the pilot phase it became apparent that it was not possible to collect the pedometer from every child a week later. Because the pedometer was able to store only 7 days of data in a loop-back memory, for every day after the 7 days the child wore the pedometer, one day of data was deleted in order to record a new day.

The study team identified an alternative pedometer, the Nl-2000i (New-Lifestyles Inc., USA), which would allow the research team to lock the memory in order to define the 7-day period. This would enable the research team to collect the pedometer at any time after the 7-week defined period. These new pedometers were not utilised before the main trial terminated.

1. (h) For children aged 8 years and over, dietary restraint measures were to be collected at 0, 12 and 24 months. This is a self-complete measure adapted by Brunstrom et al. 30 from the Dutch Eating Behavior Questionnaire,31 which was presented on the laptop screen following completion of the ideal portion size task.

2. (i) A measure of children’s diets over the last 2–3 months was collected using a paper food frequency questionnaire (FFQ) at baseline (this was also planned for completion at 12 and 24 months). During the pilot phase of the study, two FFQs were trialled: the Scottish Collaborative Group (SCG) FFQ version C2: Diet questionnaire for children32 and Frémeaux et al. ’s33 FFQ for children. As participants entered the study, they were provided with either the SCG FFQ or Frémeaux’s FFQ. The two FFQs were assessed for completeness, and informal assessment of acceptability of the different scales was carried out by the researcher by eliciting feedback from the parents after completion. It was identified that the Frémeaux’s FFQ was more acceptable for parents to complete.

A questionnaire and outcome measure schedule can be found in Appendix 2 detailing time of delivery of each measure for both parents and child.

Process evaluation

A process evaluation was conducted throughout the course of the study to monitor delivery of both the standard care and Mandolean intervention as per protocol. Study nurses kept an appointment log for each participant, recording the date of the appointment, attendance (attended/cancelled/did not attend), type of contact (telephone/face to face/e-mail/letter), duration of contact, topics covered and resources provided (as chosen from a checklist based on the manualised standard care package) and anthropometric measures for the child.

The appointment log enabled the research team to monitor that the content of the sessions were as expected, but also that the spacing and number of appointments were as per protocol for each group. In addition to the appointment log, nurses were also asked to complete a checklist for each appointment to ensure they had completed relevant study procedures. These include checking continued consent to participate, randomisation group, collection of Mandolean data, adjustment of Mandolean training lines, checking for adverse events since the previous appointment, completion of appointment log, record of appointment in patient’s medical notes, measurements of children’s height and weight, usage of other health services and quality-of-life questionnaires at the 3-, 6- and 9-month appointments and completion of a final progress report for the referring GP at the end of treatment. All study nurses were provided with a research manual and training in the study’s standard operating procedures to ensure effective delivery.

Peer meetings were arranged for the study nurses and were facilitated by a senior research nurse from the Western Comprehensive Local Research Network (WCLRN) who had received Mandolean training and was familiar with the study protocol. These meetings provided a further opportunity for nurses to share good practice and to discuss clinical issues arising.

Withdrawals from treatment

There were eight withdrawals from treatment: four from the intervention arm and four from the control arm. All agreed to be recontacted for the 12-month follow-up.

Withdrawals from study

There were seven full withdrawals from the study: six from the intervention arm and one from the control arm.

Lost to follow-up

One additional patient was lost to follow-up, as contact details held by research team and registered practice were no longer current.

Appointment attendance

The standard weight management programme involved four follow-up treatment appointments at 3, 6, 9 and 12 months. These were attended by both control and intervention participants ( Table 7 ).

In addition to the standard weight management programme, intervention participants were required to attend a further four appointments to train them to use the Mandoleans at 2, 4, 6 and 8 weeks after their initial nurse appointment (week 1) ( Table 8 ).

Non-attendance and cancellation rates

Objective 1: recruitment of at least 36 families who would be eligible for the main study

A total of 182 children (families) contacted the study team, of who 85 were assessed for eligibility. A total of 74 were eligible and consented to the study and 61 were randomised over a period of 12 months ( Table 11 ).

This report will focus on the 26 children randomised to the standard care plus Mandolean arm.

Objective 2: at least 90% of patients randomised to Mandolean will successfully be eating off the device at least five times a week

This estimate involves two parameters:

• the number of meals recorded

• the number of days between randomisation and end of the study or withdrawal from treatment, whichever came first.

Children randomised to Mandolean therapy did not use the device as frequently as instructed ( Table 12 ). Use of the Mandolean device on at least 5 days each week equated to 71% usage. Only 5 of the 26 (19%) participants randomised to the Mandolean arm achieved this target.

As each participant has a number of observations (meals) in the data set, it is useful to present summary statistics both between and within individuals ( Table 13 ).

The overall and within individual statistics are calculated over 543 meals. The between individual statistics are calculated over 22 individuals and the average number of meals per person (24.7).

Let us examine the second block of results, actual meal duration, as an example. The average meal duration across all meals for all individuals was 10 minutes, with a range of 1.3–33.1 minutes. The lowest mean meal duration for each individual was 2.5 minutes and the highest 18.5 minutes. ‘Meal duration within’ varied between –3.7 minutes and 24.6 minutes. This tells us about deviation from each individual’s average, and we would expect some of these deviations to be negative. We have to take account of the global mean of 10 minutes and, when we do this, we find that one individual deviated from their average by 13.7 minutes (range –3.7 to 10 minutes), and one individual deviated from their average by 14.6 minutes (range 24.6–10 minutes). Finally, the reported SDs tell us that the variation in mean meal duration between individuals (3.0 minutes) is slightly less than the variation in meal duration within individuals over time (3.5 minutes). The intraclass correlation coefficient (between-participant variance divided by between- plus within-participant variance) indicates that around 40% of the variation in meal duration is because of differences between individuals.

It should also be noted that data for a number of meals suggested problems with use of the Mandolean device. For example, a meal starting weight that is less than meal end weight gives rise to a negative weight consumed and negative consumption rate. Similarly, extreme values of either weight consumed or meal duration (such as when the device was left on after the user had finished eating) or both gives rise to implausible meal consumption rates. Therefore, after examining distributions of these variables, it was decided to include only meals lasting between 1 and 40 minutes and with weight of meal consumed between 1 g and 700 g. This resulted in data from 17 out of 560 meals being discarded.

Objective 3: at least 60% of those using Mandolean will have demonstrated a decrease in speed of meal consumption (longer meals) from baseline measures within 3 months of starting therapy

The following graphs present participant-level data for each of the 24 participants randomised to Mandolean for weight consumed and consumption rate for every meal ( Figure 4 ). The treatment goals were to help the participant feel ‘full’ after eating 300–350 g of food over 12–15 minutes.

FIGURE 4.

Graphs displaying participant-level data for each of the 24 patients randomised to Mandolean for weight consumed and consumption rate for every meal. (a)–(x) Individual patients. Patient identification codes (IDs) are included in each figure part.

Participants 10246063, 10493122 and 10554133 did not complete any training or treatment meals (see Figure 4e , i and k ). Participant 10850157 completed six training meals, but no treatment meals. These four participants are, therefore, excluded from Tables 12 and 13 .

Note that the blue line (square symbols if printed in black and white) on the graphs represents eating rate and should be read against the left y-axis, and the red line (triangle symbols) represents weight of food consumed and should be read against the right y-axis. Up to seven meals (the first meals on each graph) were considered as training meals where duration and food weight was not prescribed, the purpose was to assess baseline food intake, rate of eating and satiety. This information alongside previously established normal eating patterns provided the patient with an individualised training line (see Chapter 2, Mandolean training). It also helped to familiarise the participant with using the device.

The data were not formally analysed to assess objective 3. However, it is apparent from the graphs that there was no obvious systematic reduction in eating speed during the study for these participants.

Recruitment and sampling of participants in qualitative study

Qualitative study data collection

Analysis of the interview data

All interview data (decliner and participating families) were transcribed and checked for accuracy. When checking for accuracy, all names and locations were deleted and replaced with appropriate markers (e.g. nurse, child, area) to ensure anonymity. The qualitative researcher then read each transcript to gain familiarity with the data. All data were then coded and organised using computer qualitative software, NVivo 8 (QSR International Pty Ltd, Victoria, Australia). Data were initially coded to reflect sections of the data relating to key topics for exploration, such as views of the Mandolean and eating to a training line. Once all the data were coded, summaries were produced and the data further analysed using thematic content analysis. This method of data analysis allowed patterns and themes within the data to be identified across the three time points. Analysing data longitudinally (i.e. across the three time points) allowed for a rich account of families’ experiences of a weight management intervention (i.e. using a Mandolean and its integration into family mealtimes) and specific issues relating to certain weeks (e.g. adapting to changes to training lines).

To ensure confidentiality and anonymity, all decliner and participating families were given identification codes (IDs) on entry into the trial, and these codes were used on all transcriptions to ensure no one family, nurse, GP practice, and location, could be identified by name. In this chapter, quotes are used to illustrate key findings and tagged according to participants’ ID code and age of the study child.

Interviews with decliners

Eight parents who declined to take part in the pilot or main trial agreed to take part in a telephone interview to discuss their views further. However, the researcher was unable to contact three of these parents and, thus, in the end, only five parents were interviewed. All the parents interviewed were mothers of children, four girls and one boy, aged 9–10 years, who fulfilled the entry criteria (BMI ≥ 95th percentile). Interviews lasted approximately 25 minutes and took place during the months of August and September 2012. All parents who took part in the interviews had declined between baseline and randomisation (the first nurse appointment) and had used the Mandolean for 1 week.

None of the parents interviewed had tried any other form of weight management treatment for their child. During the interviews, they talked about their views of the trial information and materials, their reasons for initially wanting to take part in the trial and then their reasons for declining.

The intervention families interviewed

Although the intention was to purposefully sample families, as recruitment to the trial was very slow, in the end, details of the first 13 families eligible to take part in the qualitative study were sent to the qualitative researcher. The qualitative researcher then contacted these 13 families by telephone to provide further details about the qualitative study and to ask if they were still willing to be interviewed. All were still happy to be interviewed.

All 13 families were interviewed at time point 1 ( Tables 15 and 16). At time point 2, two parents withdrew from the main trial and the qualitative study, one parent withdrew from the qualitative study only, and one parent was unable to be contacted. Thus, nine parents were interviewed at time point 2. At time point 3, three families withdrew from the main trial and the qualitative study. Thus, six families were interviewed at time point 3.

All 13 parents interviewed at time point 1 (12 mothers and one father) were white European/British. Age at randomisation ranged from 26.7 to 60.0 years (mean 41.4 years, SD 8.24 years). Four parents were divorced or separated, eight parents were married or living with partner and one was a single parent. Twelve parents had previously tried other treatments and diet plans in order to try and lose weight. The remaining parent had previously received treatment at a weight management clinic, as she had been very underweight prior to, during and after the birth of her child. One parent reported experiencing depressive symptoms.

Of the 13 parents interviewed, 12 parents had experienced, or were currently experiencing, problems with weight management. Parents had tried various slimming diets available; two parents had had gastric bands fitted and one parent was currently taking a prescribed diet pill and following a strict diet.

All 13 children (six girls and seven boys) were white European/British, and at randomisation were aged 6.2–11.6 years (mean 8.8 years, SD 1.71 years). Twelve children attended primary school and one child was in first year of senior school.

Eleven parents and their children were interviewed together and one parent and child (aged 11 years) were interviewed separately, with the parent available in the next room. While various attempts were made to direct questions to children, parents often interrupted and on occasion answered on their child’s behalf.

At each time point, interviews were conducted within the planned period, i.e. weeks 2–3, 4–6 and 7–10. All interviews at time points 1 and 3 were conducted face to face in parents’ own homes, with the exception of one parent at time point 3, who was interviewed by telephone. At time point 2, five interviews were conducted by telephone. All of these interviews had been arranged and then cancelled for various reasons. Thus, in order to keep to the timetable, interviews were rescheduled to be conducted by telephone with the parent’s agreement. The other time point 2 interviews were held on a face-to-face basis. When conducting the interview by telephone, in cases where both the parent and child were being interviewed, the participant’s telephone was placed on speaker mode so that both individuals could participant in the interview.

All interviews lasted between 35 and 55 minutes and took place between August 2012 and July 2013.

Reasons for withdrawal from the main trial and the qualitative study included personal reasons undisclosed, too busy and problems with the Mandolean; one family felt uncomfortable with the interview because of questions regarding their child’s weight management, and two children (ages 9 and 10 years) were getting ‘fed up’ of using the Mandolean and did not want to continue. The parents of these two children (who withdrew at time point 3) were very positive about the trial and had made some changes to food shopping, portion sizes, speed of eating (i.e. they said their children were feeling full and eating less at mealtimes), and had lost a little bit of weight. They also said that using the Mandolean had helped to make these changes and hoped they would maintain them.

Results of qualitative interviews with families randomised to intervention arm

This section starts with families’ reasons for taking part in the trial. It then details key findings across the three time points (time points 1, 2 and 3), relating to:

• views of the Mandolean

• portion sizes and weighing food

• eating to the training line

• voice commands

• measuring fullness

• nurse appointments

• changes to dietary and eating habits

• barriers to taking part, and improving the experience of taking part in the trial.

The intervention

The number of clinic appointments associated with using the Mandolean, including details of who was seen and for how long, were recorded as part of the trial.

Outcomes

The CHU9D and the EQ-5D-Y were completed by the children at each of their 3-monthly clinic visits. At the same time, the parents were asked to complete the EQ-5D-5L.

Resource use

Data on NHS and personal resource use associated with the child’s weight were collected at the 3-monthly appointments face to face using a questionnaire designed specifically for this trial, based on one previously used in a similar patient group. 22 The questionnaire addressed the use of primary care, secondary care and personal expenditure as follows:

• Primary care resources included all face-to-face, telephone and home visit consultations with a GP or a practice nurse. We also collected data on any use of NHS Direct, walk-in centres, out-of-hours services and prescribed medication.

• Secondary care resources included accident and emergency (A&E) attendances, outpatient appointments and inpatient stays.

• Cost to the families included travel, expenditure on other weight-related services (e.g. counselling, complementary and alternative therapies), private health care, expenditure on exercise/physical activity and associated travel, extra expenditure on food as advised in the clinics, over-the-counter remedies, loss of earnings, carer support.

The intervention

Information on the number of appointments attended by the children in the intervention group to train them in the use of the Mandolean and to monitor their progress was documented by the nurses involved; these data were complete.

Outcomes

We report here the response rates for the CHU9D, the ED-5D-Y and the parent-completed ED-5D-5L. Table 17 gives the number of families responding to each item of each instrument at each follow-up.

At baseline all 61 children completed the CHU9D and provided data for all nine items. At the 3-month follow-up, 19 of the 20 children who attended their appointment provided some data, although data were complete in only 18 cases. This represents 44% of those still involved in the trial. Data were complete for all children who attended at both 6 and 9 months.

Fifty-nine (97%) children who were randomised completed all five items of the EQ-5D-Y at baseline, although two did not respond to the visual analogue scale (VAS). At 3, 6 and 9 months the response rates for those attending were 18 out of 20 (90%), four out of five (80%), and one out of one (100%) respectively. Fewer parents completed the EQ-5D-5L. At baseline, 56 (92%) answered the five items and 53 (88%) gave a VAS score; 13 out of 20 (65%) responded at 3 months; three out of five (60%) at 6 months; and one out of one (100%) at 9 months.

Resource use

At 3 months, 19 out of the remaining 41 participants involved in the trial completed resource-use questionnaires. At 6 and 9 months, data were provided by all families still involved, i.e. five and one respectively. The following analysis of missing data is based on the 19 questionnaires returned at the 3-month follow-up.

Full details of the completeness of each requested resource-use item are given in Table 18 . From an NHS perspective, details for a total of nine key questions covering hospital-based services, primary care services and medication use were requested. Sixteen respondents returned complete data sets that could be analysed. Of the 171 (19 × 9) possible answers, four (2%) were missing. Although the medication section was poorly completed, sufficient detail was supplied in all cases to enable a cost to be applied with some basic assumptions.

From the perspective of the child and parent, 13 questions covering travel and other financial costs to parents and carers were asked. Only 10 participants provided complete data sets for analysis; data in one or more areas were missing in nine. Overall, 15 of the possible 247 (19 × 13) responses (6%) were missing. In an additional five participants, GP travel data that might have been required at later follow-ups were missing.