Improving the assessment and management of obesity in UK children and adolescents: the PROMISE research programme including a RCT

Objectives

The aim of this study was to scope the impact of the NCMP on families and on health service use in overweight and obese children, with four specific objectives:

1. to estimate the effects of NCMP feedback on parental recognition of childhood overweight, parental perceptions of the health risks associated with their child’s weight status, and weight-related lifestyle behaviours and health service use

2. to compare the effects of different forms of weight feedback (letter vs. proactive feedback)

3. to identify barriers to behaviour change and health service use, particularly among deprived and ethnic minority groups

4. to produce preliminary data on NHS costs associated with NCMP feedback.

National Childhood Measurement Programme measurement and feedback

During the study, PCTs carried out their usual NCMP measurement and feedback procedures: PCTs sent letters to parents outlining the aims of the NCMP and measurement process, and provided an opportunity for parents to withdraw their child from the measurement. Eligible children had their height and weight measured at school by trained staff. Within 6 weeks of the measurement, written feedback was mailed to parents with information about their child’s BMI category, defined using centiles of the UK 1990 growth curves28 and cut-off points at the 2nd, 91st and 98th BMI centiles to define underweight, healthy weight, overweight and obese (described to parents as ‘very overweight’), respectively. Parents of underweight, overweight and obese children were provided with information about the health risks associated with their child’s weight status (see Appendix 1). Feedback also included information about healthy lifestyles from the DHSC’s Change4Life campaign,29 and information about local health and leisure services. A telephone number for the NCMP team was provided for parents who required further information and support. Redbridge, BANES and Sandwell PCTs supplemented the written feedback with proactive feedback in the form of telephone calls to parents of children identified as obese, during which parents were informed by a member of the measurement team (usually a school nurse) about their child’s result, and were able to get advice and support as required. Parents in Redbridge PCT were also offered a face-to-face appointment with a school nurse in a local clinic, during which families were provided with advice to support lifestyle change.

Survey questionnaires

Parents of all children participating in the NCMP in the five selected PCTs in 2010–11 (n = 18,000) were invited to participate in the study. Questionnaires were administered at baseline (approximately 6 weeks before parents received weight feedback), and at 1, 6 and 12 months after weight feedback (Figure 1). Baseline questionnaires were distributed through schools on the day of the NCMP measurements, between February and July 2011, and responses were returned to the study team in prepaid envelopes. The questionnaire assessed parental knowledge and perceptions of their child’s weight and health, and lifestyle behaviours (see Report Supplementary Material 1). The questionnaire was made available online also, and reminder postcards and e-mails were sent to parents via schools. To encourage participation, all respondents were entered into a free prize draw to win a Nintendo Wii (Nintendo Co. Ltd, Kyoto, Japan).

FIGURE 1.

Timeline for data collection in NCMP evaluation study, showing sample sizes and response rates. Figures in brackets show response rates as percentage of target population. Modified from Falconer et al. 15 © Falconer et al. ; licensee BioMed Central Ltd, 2014. This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.

Modified from © Falconer et al. ;15 licensee BioMed Central Ltd, 2014. This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited

Follow-up questionnaires at 1, 3 and 12 months were mailed directly to parents who had completed a baseline questionnaire. In addition to the questions in the baseline questionnaire, follow-up questionnaires included questions to assess parents’ feelings and behaviours in response to NCMP feedback (see Report Supplementary Material 2). Up to two reminder contacts in the form of e-mails, text messages and telephone calls were made to non-responders. Questionnaire responses were linked to children’s NCMP data, including anthropometric measures, ethnicity and deprivation data [Index of Multiple Deprivation (IMD) score, a measure of local area deprivation based on a respondent’s postcode30] using a Microsoft Access^® (Microsoft Corporation, Redmond, WA, USA) macro to match children based on their name and date of birth. Children who could not be matched automatically were identified using manual searches. The matching was carried out within the PCT and an anonymised linked data set was provided to the study team. All survey data were checked and entered electronically into a Stata^®, release 12 (StataCorp LP, College Station, TX, USA) database.

Main outcome measures

The following outcomes were assessed at baseline and at 1, 6 and 12 months’ follow-up:

1. Parents’ general knowledge of childhood obesity as a health problem, assessed using the question ‘Do you think that being overweight increases a child’s future risk of any of the following: diabetes mellitus (DM), cancers, heart disease, high blood pressure, and arthritis?’. Parents who correctly identified four or more conditions were considered to have good knowledge.

2. The child’s diet, based on parent-reported frequency of consumption of fruits, vegetables, sugary drinks, sweet and savoury snacks (categories ranged from less than once a week to at least three times a day). 31 Each food category was assigned a score from 1 to 7, with a higher score indicating more frequent consumption of fruits and vegetables, and lower consumption of sugary drinks and snacks. A healthy eating score was derived as a mean of these subscores, with a score of ≥ 5 indicating a healthy diet.

3. The child’s daily physical activity, assessed with the question ‘On average, how many minutes of physical activity (described as any activity that increases heart rate and makes the child get out of breath) does your child do?’. Children who met the national physical activity recommendation of at least 1 hour per day were categorised as engaging in adequate physical activity.

4. The child’s daily screen time (the number of hours spent watching television or playing video games). Responses were categorised by whether or not children met screen time recommendations of < 2 hours per day. 32

The following outcomes were assessed in the parents of overweight and obese children only:

1. Parental recognition of their child’s overweight status (child described as ‘overweight’ or ‘very overweight’) in response to the question ‘How would you describe your child’s weight at the moment?’.

2. Parental perception of the health risks associated with their child’s overweight status (parent answered ‘yes’ to the question ‘Do you think your child’s current weight is a health risk?’).

3. Weight-related teasing, assessed using the Teasing/Marginalisation subscale from Sizing Them Up, a validated parent-proxy measure of obesity-specific health-related quality-of-life (HRQoL) scale. 33 A score of ≥ 50 represented frequent weight-related teasing.

4. The child’s self-esteem, assessed using the Emotional Functioning subscale of the Sizing Them Up scale. A score of ≥ 50 indicated frequent episodes of low self-esteem.

At follow-up, all parents were asked whether or not they had sought further information regarding their child’s weight, from sources including the school nurse, GP, pharmacist, or friends and family. Parents were also asked about the emotions that they experienced in response to the feedback (e.g. surprised, guilty, proud, pleased, upset, angry, ashamed, judged or indifferent).

Qualitative interviews

Qualitative interviews were conducted with a subsample of parents of overweight or obese children, who were purposively selected on the basis of their responses to the questionnaire surveys at follow-up; the sample was selected so that equal proportions did and did not perceive their child’s weight to pose a risk to their health. The sample was also selected to ensure that a range of socioeconomic and ethnic backgrounds were represented (around half of the sample were from deprived and ethnic minority households), and equal proportions of parents with children in Reception year and Year 6 were selected. Semistructured interviews were conducted by a researcher with considerable experience in conducting qualitative interviews with parents of school-aged children. Interviews were conducted either at the participant’s home or by telephone, at the parent’s request. The majority (81%) of interviews took place at the participant’s home. The interview schedule was developed to explore the barriers to behaviour change following receipt of the NCMP feedback, and consisted of open-ended questions with prompts as required (see Appendix 2). Each interview lasted ≈ 30 minutes on average. Interviews were audio-recorded using a digital voice-recorder and transcribed verbatim. A thematic analysis was carried out using the software package NVivo Version 9 (QSR International, Warrington, UK) to identify emergent themes: transcripts were read and reread, and initial codes were drawn from the data. With input from three experienced researchers, these initial codes were collated into themes and a coding frame was developed for data analysis.

Economic evaluation

The costs of delivering NCMP feedback were estimated using data on staff time and resource requirements associated with the feedback provided by PCTs. Staff costs were estimated using the NHS Agenda for Change 2012 pay rates. 34 Total costs for each PCT were divided by the number of children measured for the NCMP in the PCT, to derive a cost per child, and costs associated with different forms of feedback (written and proactive) were compared. The mean cost per child of providing routine written feedback across all PCTs was calculated to provide an England-wide estimate.

Statistical analysis

The characteristics of survey respondents were described using summary statistics (means and proportions), and compared with anonymised data on weight status, ethnicity, age and IMD score for all children participating in the NCMP in the five PCTs.

Parental perceptions were explored in analyses comparing parent-perceived and objectively derived assessments of child weight status in the study sample at baseline. A multinomial model of parental-perceived weight status was used, with three categories of perceived weight status: (1) underweight, (2) healthy weight and (3) overweight (including obese). A child’s BMI (calculated from measured weight and height), expressed as a z-score from the UK 1990 reference population,35 was the only independent variable included in the model. Using this model, cut-off scores were obtained that represented the points at which a child’s BMI z-score was equally likely to be classified by parents into two adjoining weight status categories, identifying BMI z-scores (expressed as centiles) at which parents were equally as likely to describe their child as underweight as healthy weight, or healthy weight as overweight. Analyses stratified by school year and sex produced little variability in the estimates across models, indicating that an unadjusted model was appropriate. To explore whether or not ethnic group, IMD quintile, child’s sex or school year might predict parental misclassification of child weight status (two outcomes: underestimation or overestimation), logistic regression was used. Standard population-level weight status cut-off points (underweight < 2nd, overweight > 85th, very overweight > 95th centile), rather than clinical cut-off points, were used for comparison with parent-reported categories, as parents are more likely to have been exposed to these population cut-off points before receiving NCMP feedback. Standard errors were adjusted for school-level clustering.

The main analyses (objective 1) were restricted to respondents with longitudinal data (data at baseline and at least one follow-up at 1 or 6 months). If parents had completed a questionnaire at the 1- and 6-month follow-ups, precedence was given to data at 1 month to reflect the immediate impact of feedback. A sensitivity analysis to assess the effect of including responses at 6 months (n = 452; 24.5%) showed no difference compared with restricting to responses at 1 month only; therefore, results for the combined data are presented. The proportions of parents reporting each outcome at baseline and follow-up were calculated. The difference between pre- and post-feedback proportions was assessed using McNemar’s test, and results are presented as the change in proportion between baseline and follow-up (a positive value indicates an increase from baseline to follow-up, and a negative value indicates a decrease).

Analyses of the differences were stratified by the following sociodemographic characteristics to assess potential effect modification: PCT, child’s sex, school year, ethnicity (categorised as white or non-white owing to small numbers from individual ethnic minority groups), deprivation (quintiles of IMD score) and child’s weight category (three categories: healthy weight and underweight combined, overweight or obese). Among parents of obese children, analyses were also stratified by type of feedback: letter only or letter plus proactive feedback (objective 2). Differences in the outcomes by sociodemographic characteristics and type of feedback were assessed using chi-squared tests. The potential effect of seasonality on outcomes, particularly physical activity and sedentary behaviour, was also explored by stratifying and comparing outcomes by season (March–May, June–August, September–November, December–February). To account for potential clustering by PCT, random-effects logistic regression models were fitted for each outcome, including PCT as a variable in the model. Interaction terms were included to assess potential modification of the main effect by feedback type, gender, ethnicity and deprivation. All analyses were conducted using the statistical software Stata, version 12.

Ethics

Ethics approval was obtained from the London School of Hygiene & Tropical Medicine Ethics Committee.

Comparison with target population

Compared with the whole population of children participating in the NCMP in the five PCTs, the study sample had lower proportions of overweight and obese children (15.4% vs. 22.1%; p-value from chi-squared tests for difference < 0.01), Year 6 children (44.5% vs. 50.9%; p < 0.01), parents from the most deprived areas (43.7% from the two most deprived quintiles vs. 49.1%; p < 0.01) and children from ethnic minority groups (34.0% vs. 45.5%; p < 0.01) (Table 1). Parents from these groups were also more likely to drop out of the study, resulting in greater differences at follow-up: at 6 months, 13.4% of children were overweight or obese, 40.2% were from the most deprived areas and 30.0% were from ethnic minority groups.

Of the 285 parents of overweight or obese children who had responded to the questionnaire survey, 108 were invited (by letter) to participate in the qualitative interviews. Interviews were carried out with 52 parents (43%). Theoretical saturation of the data was achieved with 40 interviews, with no new themes emerging in the final 12 interviews. The demographic characteristics of interview participants are presented in Table 2.

Parental knowledge, perceptions and lifestyle behaviours at baseline

At baseline, 72.3% of parents were able to identify the common health conditions associated with childhood overweight; this proportion was higher among parents with healthy weight and underweight children [74.8%, 95% confidence interval (CI) 72.5% to 77.1%] than among those with overweight (61.9%, 95% CI 54.2% to 69.7%) or obese (57.8%, 95% CI 47.4% to 68.2%) children. Parental knowledge also varied with ethnicity; just 59% of parents of children from non-white ethnic groups correctly identified common childhood obesity comorbidities, compared with 79% of parents of white children.

About half of the children were reported to meet dietary (48.7%) and screen time (53.4%) recommendations, and about one-third (36.3%) achieved recommended levels of physical activity. Compared with their healthy weight and underweight peers, levels of physical activity were lower and screen time was higher among overweight and obese children; the proportion achieving a healthy diet was lower among obese (but not overweight) children.

Among the parents of overweight and obese children (n = 285), the proportion who considered their child to be overweight was low: 14% of parents of overweight children and 35% of parents of obese children described their child as overweight. Similarly, 11.0% of parents of overweight children and 38.6% of parents of obese children considered their child’s weight status to pose a health risk. A small proportion of parents of overweight and obese children (8.9%; n = 10) reported frequent weight-related teasing, and 1.8% (n = 2) reported low self-esteem.

When parent-perceived weight cut-off points were estimated, the point at which a parent was equally likely to recognise underweight as healthy weight was when their child had a BMI at the 0.8th centile (95% CI 0.4 to 1.1 centile). A parent was more likely to classify their child as overweight rather than healthy weight when their child had a BMI of ≥ 99.7th centile (95% CI 99.3 to 99.9 centile). Based on population-based cut-off points for weight status, 915 (31%) parents underestimated and 25 (< 1%) overestimated their child’s weight status. Parents were more likely to underestimate their child’s weight status if they were black [reference group: white, odds ratio (OR) 1.5, 95% CI 1.1 to 2.1], south Asian (OR 1.6, 95% CI 1.3 to 2.0), male (OR 1.3, 95% CI 1.1 to 1.6) or in the older age group (OR 1.3, 95% CI 1.1 to 1.5). Parents from less deprived areas were less likely to underestimate their child’s weight status (OR for each quintile less deprived, OR 0.8, 95% CI 0.75 to 0.9). Overestimation of weight status was more likely for Year 6 than Reception children (OR 3.1, 95% CI 1.2 to 8.5). Ethnic group, deprivation and a child’s sex did not predict parental overestimation of weight status.

Objective 1: the impact of National Child Measurement Programme feedback

Parental perceptions

The proportion of parents of overweight and obese children who recognised their child’s overweight status increased post feedback (+15.8%, 95% CI 9.8% to 21.7%), with a greater increase among the parents of obese children (+23.5%) than among parents of overweight children (+11.1%) (Figure 2). However, 41% of parents of obese children and 75% of parents of overweight children continued to perceive their child to have a healthy weight after receiving NCMP feedback.

FIGURE 2.

The proportion of parents of overweight and obese (very overweight) children who considered their child to be overweight, before and after receiving NCMP feedback.

The increase in the proportion of parents who considered their child’s overweight status to pose a health risk was modest for both overweight (+7.0%, 95% CI 1.4% to 12.6%) and obese children (+5.0%, 95% CI –6.9% to 16.9%) (Figure 3).

FIGURE 3.

The proportion of parents of overweight and obese (very overweight) children who considered their child’s weight to pose a health risk, before and after receiving NCMP feedback.

Parents’ reactions to National Child Measurement Programme feedback

Parents reported experiencing a wide range of emotions in response to receiving NCMP feedback, which varied considerably by the child’s weight category. The most common reaction among parents of healthy weight children was ‘pleased’ (reported by 65.0% of parents) and 21.0% felt proud of their child’s result. Among the parents of overweight and obese children, the most commonly reported emotions were surprise (reported by 35.0%), upset (24.1%), guilt (15.4%) and shock or anger (14.8%).

Despite some negative emotional reactions to the feedback among the parents of overweight or obese children, the majority of parents across all child weight categories (87.2%) reported that they found the NCMP feedback to be either ‘somewhat helpful’ or ‘very helpful’. Parents of overweight children were most likely to find the feedback ‘not at all helpful’ (Figure 4). The majority of parents (70.0%) reported that they would encourage future participation in the NCMP for their child or child’s siblings. Only 1.8% said that they would withdraw their child from the programme in the future.

FIGURE 4.

The proportions of parents who found the NCMP feedback to be helpful, by child’s weight category.

Objective 2: the effects of different forms of weight feedback

Proactive feedback was implemented in Redbridge, BANES and Sandwell PCTs. In Redbridge PCT, approximately 60% of parents of obese children were successfully contacted by telephone and offered a face-to-face appointment; 69% of these parents accepted the appointment. Corresponding numbers for BANES and Sandwell PCTs could not be ascertained. There was some evidence to suggest that proactive forms of feedback may be more effective than written feedback alone in changing perceptions among the parents of obese children (Table 5). Parents who lived in PCTs offering proactive feedback demonstrated greater improvements in parental recognition of their child’s overweight status (32% vs. 10%) and greater recognition of the associated health risks (13% vs. –8%) than patients in all PCTs in the study taken as a whole. There was no effect of type of feedback on lifestyle behaviours or health service use.

Parents expressed a strong preference for written feedback over other forms of feedback, with 84.5% of all parents preferring feedback by letter, whereas 3.0% preferred feedback by telephone. Among the parents of obese children, the type of feedback received during the study was not associated with preferences for future forms of weight feedback.

Objective 3: barriers to behaviour change and health service use

Qualitative interviews with parents of overweight and obese children identified several barriers to health-related behaviour change: (1) parental perceptions of their child’s health, weight and appearance, (2) cultural differences, (3) access to and availability of local services, (4) contradictory advice from health professionals, (5) relevance of lifestyle advice and (6) limits to parental control.

Objective 4: NHS costs associated with National Child Measurement Programme feedback

The cost of providing routine written feedback letters to parents was estimated at £1.24 per child, compared with £9.50 per child for telephone feedback and £41 per child for face-to-face appointments. The higher costs were attributable to the additional staff time required to deliver proactive feedback (Table 6). Based on these estimated costs and the national prevalence of obesity among children taking part in the NCMP, the mean cost per child of providing written feedback letters to all parents and a telephone call to the parents of all obese children was £2.03 (on average an additional £0.79 per child compared with written feedback only).

Strengths and limitations

This study was conducted in a large, population-based and sociodemographically diverse sample, which enabled examination of the effects of weight feedback in different weight and sociodemographic groups. To account for the lack of a comparison group, a short time interval was selected between baseline and follow-up questionnaires (≈ 3 months) to minimise potential changes in outcomes that may have occurred independently of the intervention.

The main limitation of this study was the low response rate and high attrition, which raise the possibility of a biased sample. It is plausible that parents who are more engaged with issues relating to their child’s health would be more likely to participate in the study and respond to feedback than parents who are not engaged with these issues, which would lead to an overestimation of the effect of weight feedback. Comparison of the study sample with all children taking part in the NCMP in the five PCTs revealed a slight under-representation of children from ethnic minorities, children from Year 6, children from deprived areas and overweight/obese children in our sample, which may limit the generalisability of the study findings to the UK NCMP population. However, the lifestyle behaviours of the study sample were similar to those previously observed in national surveys of primary school-aged children. 46–48 To keep the questionnaires concise, brief measures of behaviour and potential harms were used, not all of which are validated. Lifestyle behaviours are difficult to measure accurately,49 and self-reported measures of behaviours may be subject to misreporting. The reported lack of a change in lifestyle behaviours in this study argues against this being an issue;50 however, this does not discount the possibility of systematic social desirability bias, for example parents over-reporting fruit and vegetable consumption at both time points. The dietary measures used have been previously assessed using test–retest methods and found to be reasonably reliable. 31 In addition, parental perceptions of overweight and health risk were evaluated using questions that have been used in previous evaluations of weight feedback. 43 Questionnaires were not translated into other languages, as it was not standard practice to translate the feedback letters in any of the PCTs involved in the study. Consequently, the study does not reflect the impact of NCMP feedback on non-English-speaking households.

We did not collect data on parental weight as this would have only been by self-report, which is highly prone to bias. We did not collect data on disordered eating among children because of contraints of questionnaire length; however, this would have added to our understanding of the impact of NCMP feedback.

Objectives

The objectives of this study were to:

• improve primary care management of childhood obesity through the development and pilot study of a brief evidence-based assessment and management tool

• improve primary care decision-making about onwards referral through the development of evidence-based algorithms for identifying the risk of obesity comorbidities.

Phase 1: development of the Computer-Assisted Treatment of CHildren tool

Online application development

The web-based CATCH tool was written using Ruby on Rails, an open-source web application framework for the Ruby programming language. It was designed to be accessed through standard internet browsers and incorporates secure tunnel technologies to allow secure data transfer and storage. The CATCH tool was designed to follow the standard flow of a clinical consultation and guide a health-care professional to record the child’s demographic information (e.g. age, sex, ethnicity) and take height and weight measurements. The tool computes the child’s BMI and age- and sex-specific BMI centile, displays a sex-specific BMI chart and the child’s weight status, identifies children for referral and provides lifestyle advice. Figures 5–7 show screenshots from a consultation using the CATCH tool to illustrate the main stages and user interface. The functionality and usability of the CATCH tool were pretested with anticipated end-users (primary care clinicians) during the web development stage using a paper-based version of the application.

FIGURE 5.

Screenshot of the BMI and height-for-age features of the CATCH tool, for a 12-year-old boy with a BMI of 24.9 kg/m². Reproduced from Park et al. 51 Published by the BMJ Publishing Group Limited. 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/.

Reproduced from Park et al. 51 Published by the BMJ Publishing Group Limited. 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/

FIGURE 6.

Screenshot of the risk assessment results for a 12-year-old boy with a BMI of 24.9  kg/m², who is at high risk of cardiovascular comorbidity and medium risk of emotional and behavioural difficulties. Sources: unpublished data from the PROMISE programme; Dr Min Hae Park from the Department of Non-Communicable Disease Epidemiology, London School of Hygiene & Tropical Medicine, London, UK, personal communication; and reproduced from Park et al. 51 Park et al. 51 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/. The figure includes minor additions and formatting changes to the original figure.

Reproduced from Park et al. 51 Published by the BMJ Publishing Group Limited. 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/

FIGURE 7.

Screenshot of an excerpt of the lifestyle assessment results for a 12-year-old boy who reports that he eats breakfast every day, eats meals away from home four times per week and drinks 22 sugary drinks each week. Sources: unpublished data from the PROMISE programme; Dr Min Hae Park from the Department of Non-Communicable Disease Epidemiology, London School of Hygiene & Tropical Medicine, London, UK, personal communication; and reproduced from Park et al. 51 Park et al. 51 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/. The figure includes minor additions and formatting changes to the original figure.

Reproduced from Park et al. 51 Published by the BMJ Publishing Group Limited. 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/

Phase 2: evaluation of the Computer-Assisted Treatment of CHildren tool

Performance of risk estimation models

Table 8 provides the risk scores (between 0 and 100) that correspond to each risk category and the performance measures for each risk estimation model, by risk category.

When tested in the databases from which they were derived, both risk prediction models had high sensitivity for identifying children (see Table 8) who are at risk of cardiovascular disease and are emotional and having behavioural difficulties. When used to identify high-risk children, the sensitivity of the models was lower: 23.3% of children who had emotional and behavioural difficulties were detected by the model and classed as being at high risk, and the cardiovascular risk factor model classified 52.7% of children with at least one cardiovascular risk factor as being at high risk. The positive predictive value of the cardiovascular risk factor model was moderate: just over half of children whom the model identified as being at high risk actually had one or more cardiovascular risk factors. The emotional and behavioural difficulties model had high negative predictive values for both risk categories, meaning that the vast majority (95.3%) of children whom the model identified as being at medium or high risk did have emotional and behavioural difficulties, and > 90% of children who were classified as not being at high risk did not have emotional and behavioural difficulties.

Family questionnaire

Family interviews

Table 10 shows the frequency of the main themes in the analytical framework. In line with the questionnaire responses, family interviews revealed that experiences with the CATCH tool were generally positive.

Parents described the consultation as informative, non-judgemental and non-intrusive:

I think the questions were very good because there are a lot of things that we were doing and we didn’t know that it was good or bad. It was good. The whole consultation was very beneficial.

Mother of boys aged 10 (obese) and 12 (overweight) years

All interviewees found the tool’s outputs useful. Even though most parents had been concerned about their child’s weight previously, several referred to the consultation as a ‘wake-up call’ that alerted them to the severity of the problem:

. . . I would know that [my daughter was] putting on weight but I would not say, I didn’t expect that it was as bad as it was and especially when they weighed her and they found she was 61 [kg] when she’s 9. It was a wake-up call. It was good.

Mother of girl aged 9 years, obese

However, two mothers described that the results of the consultation had generated some anxiety in their child:

. . . when she saw what was on the computer, that she was on the red line, and they explained to her what it means for her health . . . she really [got] scared.

Mother of girl aged 9 years, obese

[My son] was a little bit worried [when hearing the results of the consultation] like, ‘Oh what’s going to happen next?’. [But in the end] everything was fine.

Mother of boy aged 9 years, obese

All of the respondents found the lifestyle advice informative and instructive. In particular, specific advice on diet was highlighted as being useful:

[I found the lifestyle advice] very, very helpful because I was buying some stuff, like, I wanted them to eat breakfast, they eat breakfast every morning, but I was buying them wrong breakfast, the cereals . . . You know the sugary stuff so now . . . I’ve cut down on those things and the drinks that I was buying, I thought they are orange juice and stuff but [it’s] not 100% juice . . .

Mother of boys aged 10 (obese) and 12 (overweight) years

The use of visual aids, such as the BMI chart and the printed lifestyle advice, was described as reinforcing the advice given:

It just confirmed everything that we already knew, but you know, when it’s on paper, it’s sort of a bit more . . . I wouldn’t say serious, but it’s a bit more in your face.

Mother of boy aged 7 years, overweight

Five interviewees said that they had made lifestyle changes following the tool-based consultation. In at least two cases, changes had been introduced for the whole family. A theme that emerged was the need for practical support following the consultation. Five parents suggested that follow-up appointments for monitoring and guidance would be beneficial:

I would probably say that now, as a stepping-forward thing, although the study thing may be over, maybe to follow up and do like a plan on how to maintain or reduce the weight or some kind of thing like that . . . To be honest, I don’t mind doing it for myself but then there’s no monitoring or anything like that through the GP, which might be beneficial for them as well so they can keep an eye on whether we’re doing it correctly and there’s no other side effects or anything like that.

Mother of boy aged 7 years, overweight

Practitioner questionnaire

Practitioner interviews

The analytical framework used for practitioner interviews is shown in Table 11.

Overall, practitioners described satisfaction with the tool. A theme that emerged was that the tool could enhance the impact of practitioners’ advice by adding authority to the message:

. . . I thought that was really good. We talked about it all the time. And because patients can take it away with them, it is quite good for them. I literally go over what the [lifestyle advice] told them: what you should eat [etc.]. And they got something in writing from the computer. They see ‘I’m not just saying that’ . . . that [there is truth] to it.

Nurse, practice 3

All practitioners found the tool straightforward to use, and none reported having problems interpreting the tool’s outputs. One respondent (doctor, practice 1) felt that children reacted to the consultation with indifference, but others felt that patients reacted positively. The BMI chart and lifestyle advice were highlighted as generating the strongest responses from children and parents:

I think for some people they probably underestimated the time that their kids were spending watching TV or playing games, or computer games or whatever. So that was good. The thing about sleep was good. How much sleep that they need on average, was good . . . The other thing was about fruit juices. Sugar. I think that was a revelation to a lot of people. I think, you know, there was nothing negative about it [the tool].

Doctor, practice 2

Although none of the practitioners described serious problems with the tool, two respondents described that it can be uncomfortable to broach the subject of a child’s weight-related health risk with parents:

[The risk assessment component] was tricky as far as the parent’s concerned . . . The parents found it uncomfortable. Maybe it opened something up for them to say, you know, . . . ‘Is this going on [with my child]? Am I not being told?’

Nurse, practice 2

All four respondents were in agreement that the tool (or a version of it) would be something that they would continue to use in the future and would like to see integrated into their clinical software system. In particular, the child-specific BMI chart was seen to be desirable, as current tools on the system allow calculation of only adult weight status. Three of the four respondents suggested that the tool-based consultations would benefit from the addition of more tangible support, such as a structured programme of activities and follow-up consultations to monitor weight management.

Strengths and limitations

The development of the CATCH tool was informed by the evidence base: risk prediction models were based on analyses of large, well-established, population-based cohorts, whereas the lifestyle behaviours for which healthy recommendations were provided were identified through comprehensive literature reviews, and the generalisability of findings was confirmed in analyses of a UK data set. The tool was also extensively piloted among future user groups in order to obtain feedback relevant to analysis, design and development. Limitations of the CATCH tool include features of the data sets in which the risk estimation models were created: the cardiovascular risk factor model was developed within the ALSPAC data set, which comprises data from a relatively healthy population with a low prevalence of overweight and ethnic minority groups. Similarly, the emotional and behavioural difficulties model was developed in the RELACHS data set, which comprised mainly deprived children of ethnic minorities. For both data sets, data were available for adolescents only; therefore, the validity of the prediction models in younger children is unknown. AUC values indicated that the cardiovascular risk factor model was not good at correctly classifying those with and without elevated risk. However, when compared with similar measures for algorithms designed to detect children at risk of glucose intolerance84 or having atherosclerotic lesions,85 specificity was comparable and positive predictive values were higher, although sensitivity was lower. Further validation of the models in larger, more diverse data sets is required. Statistical analyses of lifestyle behaviours in the ALSPAC data set showed that, even though the six chosen behaviours explained a similar amount of variation as a model with all 17 lifestyle behaviours, lifestyle behaviours in general explained very little (7–10%) of the observed variations in BMI among overweight and obese children. This could indicate that most of the variation in the degree of overweight is actually due to unknown or unmeasured factors.

The generalisability of findings from the pilot study is limited by its small sample, which is not representative of the wider target population. In particular, owing to the ethnically diverse population in north-west London, the majority of participants in our study were from Asian or black ethnic groups (in the boroughs of Harrow and Brent, around two-thirds of the population are from black and minority ethnic communities, predominantly South Asian groups86). However, children from these ethnic groups are at increased risk of obesity87 and associated health problems39 and may therefore be priority groups for obesity interventions. Parents and practitioners who participated in the study are likely to be those who are engaged with weight-related health issues; experience of the tool-based consultation may be different if delivered by practitioners, or administered to families, with less interest in these issues. We did not collect data on parental weight and, thus, could not comment on implications of parental weight for management of child obesity. We did not collect data on disordered eating behaviour in this small pilot; this would be useful in a future larger evaluation.

Despite these limitations, the evaluation has identified key themes relating to the feasibility, acceptability and usefulness of an online tool for the assessment and management of overweight children in primary care. The responses from families and practitioners are promising and provide a basis for further intervention development and evaluation in a larger study.

A community-based motivational personalised lifestyle intervention to reduce body mass index in obese adolescents: results from the Healthy Eating and Lifestyle Programme randomised controlled trial

The trial was conceptualised by Deborah Christie and Russell M Viner. The intervention was developed by Deborah Christie with input from Russell M Viner. Russell M Viner and Deborah Christie developed the trial design, with input from Sanjay Kinra, Ian C Wong, Irwin Nazareth and Tim J Cole. Tim J Cole and Ulla Sovio developed the statistical plan, with input from Russell M Viner and JG. Anthony S Kessel, Anne Mathiot and Lee Hudson contributed to the trial design and writing of the protocol. Steve Morris and MP contributed to the trial design and undertook the economic analysis. RH supervised the providers and analysed the fidelity data and process evaluation. All authors read and approved the final manuscript.

Abstract

Background

Approximately 7% of CYP aged 5–15 years in the UK have obesity at a level likely to be associated with comorbidities. 1 Over 1% of adolescents have extreme obesity, with a BMI of > 3 standard deviations (SDs) above the mean, equivalent to a BMI of approximately 35 kg/m² in adults. 90 The treatments for childhood obesity recommended by NICE are multicomponent lifestyle modification programmes. 11 The majority of published programmes are, however, aimed at primary school children, with existing adolescent studies largely carried out in non-UK academic tertiary care centres with highly specialised staff involving white, middle-class, motivated families. 42,91 The applicability and generalisability of these studies for British adolescents, therefore, is limited, and reviews42,91 conclude that there is a need to conduct adequately powered high-quality trials in representative populations with integral process evaluation and appropriate lifestyle tools.

Major cognitive and psychosocial changes during adolescence mean that children’s family-centred weight management programmes are inappropriate for adolescents. During adolescence, young people spend more time outside the home, start making their own choices about food and exercise, and develop diet and activity patterns that will stay with them into adult life. Issues related to body image and self-esteem become important, and behaviour change depends as much on individual motivation as parental authority. Yet few weight management programmes recognise these issues, as programmes are generally designed for younger children or older adults.

The Healthy Eating and Lifestyle Programme was developed as a specific adolescent-focused intervention, designed for obese 12- to 18-year-olds seeking help to manage their weight. HELP was developed from best practice,11 and is a multicomponent intervention focused on enhancing motivation to change and developing self-efficacy and self-esteem (rather than focusing on weight only). Unpublished pilot data from application of HELP by psychologists in a clinical setting for 20 subjects aged 13–17 years showed a mean BMI reduction of 1.7 kg/m² over 6 months, equivalent to a 0.4 SD effect size. We report a RCT of HELP in a community sample of obese adolescents. The primary aim was to assess whether or not HELP delivered by graduate mental health workers in the community was more effective in reducing BMI in obese adolescents than enhanced standard care. Secondary aims were to assess the cost-effectiveness of HELP and its impact on cardiometabolic risk and psychological function.

Methods

Results

There were 519 contacts from young people or their families interested in participating in the trial, of whom 352 completed a telephone screening interview and the remainder were unwilling to participate further, were not eligible or could not be contacted futher after the first contact (Figure 8). Of those young people who completed the telephone screen, 210 were considered potentially eligible and invited to attend a baseline assessment. A total of 174 young people (88%) were randomised into the trial, with 87 in each arm. Primary outcome data on BMI at 26 weeks were collected on 145 (83%) participants (Figure 9): 119 at the hospital and 26 at home. At study entry, mean age was 15 years and mean BMI was 32.0 kg/m²; 63% of the participants were female, and participants were from a range of ethnic groups (Table 14). The two arms were well balanced with respect to baseline BMI and other baseline characteristics.

FIGURE 8.

The CONSORT recruitment flow diagram.

FIGURE 9.

The CONSORT flow chart of follow-up of the 174 participants randomised in the HELP study.

On average, participants in the intervention arm attended 10 of the 12 HELP sessions; 27 young people (31%) attended all 12 sessions and 69 (79%) met the criteria for compliance. Of observed sessions, 76% were rated as good, 22% as adequate and only 2% as poor.

Body mass index at 26 weeks was available for 74 (85%) participants in the intervention arm and 71 (82%) participants in the control arm (Table 15). The mean BMI at 26 weeks was 32.9 kg/m² in the intervention arm and 32.4 kg/m² in the control arm, representing an increase in BMI of 0.18 kg/m² and 0.25 kg/m², respectively, compared with baseline (Table 16). The effect of the intervention on BMI at 26 weeks adjusted for baseline age, sex and BMI was –0.11 kg/m² (95% CI –0.62 to 0.40 kg/m²; p = 0.7), indicating that the effect of the intervention was non-significant (Table 17). The effect among compliers was –0.14 kg/m² (95% CI –0.78 to 0.51 kg/m²; p = 0.7) and the distributions of BMI change at 26 weeks in the intervention and control arms were very similar (Figure 10).

FIGURE 10.

Distribution of changes in BMI from baseline to 26 weeks in control and treatment groups among 145 participants with measurements at both time points. Reproduced from Christie et al. ,88 licensee BioMed Central Ltd, 2011. This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.

Reproduced from Christie et al. 88 © Christie et al. ; licensee BioMed Central Ltd, 2011. This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited

Discussion

It was found that HELP, a multicomponent adolescent-specific weight management intervention based on published best practice and designed to enhance motivation to change and develop self-efficacy and self-esteem, did not reduce BMI or improve psychological or cardiometabolic function when compared with a single educational session in a community sample of obese adolescents. There was no evidence of harm from the intervention.

A high-quality randomised trial of a complex intervention, previously shown to be promising in clinical practice, was conducted. The HELP intervention was based on published best practice and was designed specifically for adolescents struggling to manage their weight; it included parents as well as young people and included an element of personalisation. The primary outcome was a reproducible quantitative measure recorded by blinded observers. ‘Raw’ BMI was used rather than BMI SD score (zBMI), as the skewness in the BMI distribution means that the same change in kg/m² will produce a smaller change in zBMI the more obese the subject is. 106 A community sample of obese young people was recruited and the intervention was tested against a standardised educational session. Randomisation was conducted independently of the research team. Attrition was low compared with other paediatric obesity studies,107 and the recruited sample was sufficient to identify the desired effect size with high power. Fidelity of delivery of the intervention was good and ≈ 80% of participants received at least the intended ‘dose’ of the intervention, which is good for complex interventions.

The study was subject to a number of limitations. Recruitment was difficult and the eligibility criteria were widened to include adolescents with morbid obesity. However, it is unlikely that this influenced the results, as only nine participants (4.6%) had a baseline BMI of between 40 and 45 kg/m². Attrition was significant by 52 weeks; although the smaller sample size reduced the power to identify an effect of the intervention, it is likely that the direction of bias from attrition would have been to preferentially retain those in whom the intervention had been effective and inflate any identified effect. Although full data on the primary outcome were available in all who attended for follow-up, it was not possible to collect sufficient planned follow-up data on physical activity and psychological morbidity. However, given the absence of any significant effects on other secondary outcomes, it is unlikely that the intervention would have had an effect on physical activity and psychological morbidity.

Conclusions

The HELP intervention was no more efficacious than standard care of a single educational session for reducing BMI in a community sample of obese adolescents. This adds to the existing evidence base of lack of effectiveness of weight management trials for children and adolescents, despite HELP being based on best practice and designed for adolescent development needs. Further work is needed to understand how weight management programmes can be delivered effectively to young people from diverse and often deprived backgrounds in which childhood obesity is common. At minimum, this study reinforces the need for higher-level, structured interventions to tackle the growing public health burden of obesity in the UK and internationally.

Abstract

Introduction

Childhood obesity has increasingly become one of the most serious global public health challenges. Globally, an estimated 42 million children aged < 5 years were overweight or obese in 2010. 111

The prevalence of obesity in England remains high. The NCMP found in 2014 that 9.5% of children in the Reception year at school were obese and a further 13% were overweight; 19% of children aged 10–11 years were obese and 14% were overweight. 112 Figures from the Health Survey for England (HSE) for 2012 show that ≈ 28% of children aged 2 to 15 years were overweight or obese. 113,114

Childhood obesity has both immediate and long-term effects on health and well-being. Obese children are more likely to have risk factors for cardiovascular disease, such as high cholesterol or high blood pressure,115 pre-DM,116 bone and joint problems,117 sleep apnoea118 and social and psychological problems (such as stigmatisation and poor self-esteem). 119,120

More than half of children who are obese will grow up to be obese as adults121 and are therefore more at risk of adult health problems such as heart disease, type 2 DM (T2DM), stroke, some cancers and musculoskeletal and psychosocial problems. 122–125

As well as having a substantial impact on health, obesity is also associated with significant health expenditures. In England, the costs of overweight and obesity to the health system have been calculated to be £4.2B per annum. 126 Obese individuals have been found to have medical costs ≈ 30% greater than their peers of normal weight. 127 In terms of non-health-care costs, productivity losses due to obesity have been estimated to be £15.8B. 126 Childhood obesity in particular imposes a substantial cost to the health system, with estimated costs of ≈ US$14B per annum in the USA. 128

There are few economic evaluations of childhood lifestyle obesity interventions and, among those that have been conducted, differences in the type of lifestyle intervention, age of participants, cost components included, country of setting, outcomes of interest and methodologies mean that it is difficult to draw conclusions about whether or not such interventions are cost-effective. Table 19 shows findings from a literature review of economic evaluations of studies in childhood obesity.

In an accompanying paper,138 the efficacy of a community-delivered motivational multicomponent intervention compared with enhanced standard care for obese adolescents was evaluated. The aim of this paper was to assess the costs and cost-effectiveness of the intervention from the perspective of the UK NHS.

Methods

Results

Discussion

The economic analysis of the HELP trial showed that the intervention had similar outcomes but higher costs when compared with enhanced standard care. Resource use during follow-up was similar between the trial arms; therefore, the cost of the HELP intervention was the major driver of the cost difference.

Given that there were no differences in outcomes or resource use during follow-up between groups, and given the higher intervention costs, these findings indicate that the HELP intervention is not cost-effective.

As noted, there are few evaluations of childhood obesity interventions that include an economic evaluation with which to compare these results; however, differences between studies in terms of age of participants, country, components of the interventions and outcome measures make comparisons with other studies difficult.

Summary

The HELP trial showed that changes in BMI in obese adolescents who received the intervention are similar to those seen in adolescents who received enhanced standard care. The complementary economic analysis has further shown that there were no differences in HRQoL and health-care resource use during 12 months’ follow-up between the two groups. The HELP intervention was more costly than the control because of higher intervention costs, and so is unlikely to be cost-effective.

Key findings across study C

• Across all young people in the study, approximately one-third reduced their BMI, one-third increased their BMI and one-third retained the same BMI at 6 and 12 months after starting the study.

• Recruitment to the trial was difficult, with a large degree of work required to recruit obese young people into the trial.

• No difference was found in the BMI of young people between the HELP group and the control group at 6 or 12 months. This means that HELP and the single nurse-delivered session had very similar effects on the BMI of young people and their cardiovascular risk over the 6 and 12 months of the study.

• Regarding secondary outcomes, there was no difference in young people’s waist circumferences, blood pressure and other cardiovascular risk markers (e.g. cholesterol, glucose and insulin), measures of QoL, self-esteem or problem eating behaviours at 6 or 12 months between those young people who received HELP and those who received the nurse session.

• Regardless of trial group, young people and their families reported that joining the trial had helped them make positive changes to their diet, activity levels and motivation.

• Mean intervention costs per subject were £918 for HELP and £68 for enhanced standard care. Adjusted costs were significantly higher in the intervention group [mean incremental costs for HELP vs. enhanced standard care £1003 (95% CI £837 to £1168)]. There were no differences in adjusted QALYs between HELP and the nurse session (standard care).

Background

Three drugs were approved for obesity treatment in the UK between 1998 and 2006 (although the last two were subsequently withdrawn from the market): orlistat (Alli^®; GlaxoSmithKline plc, Brentford, UK), sibutramine (Meridia^®; Abbott Laboratories, Abbott Park, IL, USA) and rimonabant (Acomplia^®; Sanofi, Paris, France). In 2014, the revised NICE guidelines recommended that orlistat was appropriate for obesity management in children and adolescents aged ≥ 12 years in certain circumstances. 12 The aims of this study were to (1) gain a better understanding of young people’s and families’/carers’ experiences of using AODs, (2) use this improved understanding to develop interventions that optimise AOD usage in young people and (3) analyse GP prescribing patterns of AODs and compare their prescribing to current recommendations.

Methods

To assess the aims, a series of studies were conducted: (1) a systematic review and meta-analysis evaluating the efficacy and safety of AODs from published RCTs in children and adolescents; (2) cohort studies to investigate prescribing patterns [for orlistat, sibutramine, metformin (Glucophage^®; Merck Serono, Darmstadt, Germany)] to young people in primary care settings; (3) a survey study to determine current practice in prescribing these drugs to young people for obesity treatment, with examining GP-completed questionnaires across UK general practices (in England, Scotland, Wales and Northern Ireland); and (4) semistructured interviews with young people (i.e. aged 13–18 years) and their parents from three specialist obesity clinics, to understand children’s and families’ perspectives of treatments.

Key results

Evidence from published RCTs showed that orlistat and sibutramine (in combination with behavioural therapy) significantly reduced BMI compared with placebo in young people.

Prescribing of AODs (orlistat, sibutramine) to children and adolescents increased 15-fold between 1999 and 2006, but approximately 45% of orlistat and 25% of sibutramine prescriptions were discontinued after 1 month of treatment, and before patients could see any weight benefits.

There was also a steady increase in metformin prescribing to obese and overweight young people, particularly in girls aged 16–18 years.

In 121 returned questionnaires (80% response rate), 61% of GPs reported issuing the drugs to their patients without secondary or tertiary care team advice. Despite comprehensive guidance provided in the NICE guideline, GPs expressed a need to develop a new guide for prescribing AODs to young people.

Young people and parents described side effects as a significant experience and few adhered to prescribed regimens, independently changing lifestyle and dosage to tolerate medications.

Conclusions

General practitioners require more support from a specialist in managing the treatment of obesity in children in the community. Pharmacological treatment is unlikely to be effective without further assistance, such as lifestyle modification or psychological support. There should be further research to develop a collaborative, prospective, obesity management surveillance network for children and adolescents across other specialist centres in the UK. This could examine effectiveness, safety and prescribing patterns of medications along with other interventions (e.g. psychological intervention, bariatric surgery) for weight management in clinical practice.

Literature search

Search strategies drew on published optimal search strategies for drug trials. 156,157 The following databases were search between January 1996 and July 2008 for clinical trials investigating AOD and body weight reduction: MEDLINE, EMBASE, and the Cochrane Central Register of Controlled Trials (CENTRAL). We also searched the trial registers: the metaRegister of Controlled Trials (www.controlled-trials.com), WHO clinical trial registered (www.who.int/ctrp/en/) and the Clinical trials government (www.clinicaltrials.gov/). Hand-searching was also carried out to examine the reference lists of identified studies.

Eligibility criteria

The inclusion criteria for systematic review were published double-blind randomised placebo-controlled clinical trials investigating the effects and safety of AOD (orlistat, sibutramine, rimonabant) for body mass index (BMI) reduction in children and adolescents aged < 20 years, with duration ≥ 6 months. We excluded quasirandomised, open-label crossover trials and studies published only in abstract form. There was no restriction on language. Note that, given that a range of definitions of childhood obesity exist, we included any trials which used an established definition of overweight or obesity (BMI ≥ 85th, 95th or 98th centile; BMI > International Obesity Taskforce definitions). 158 Given our age range, we examined BMI reduction rather than weight loss as the primary outcome, as growing children with a stable weight may exhibit reduced BMI.

Data extraction and quality assessment

Two reviewers (YH, TT) performed the electronic searches and screened the articles independently. Articles that clearly did not meet eligibility criteria were rejected on initial review. Articles marked for potential inclusion were then obtained electronically or in paper copy, and assessed again for inclusion. Disagreement was resolved by consensus.

Those included studies deemed to meet inclusion criteria by both reviewers were appraised. A standardised form was used to record all details of the papers reviewed. 159 The standard form included study design, blinding status, trial duration, mean age of participants, gender, number of participants in treatment and placebo group, interventions and the assessment of intention-to-treat (ITT) analysis. The Quality of Reporting of Meta-analyses (QUOROM) guideline was used for reporting our review. 106

Statistical analysis

We expressed the primary outcome as change in raw BMI rather than in BMI standard deviation score (SDS), as use of BMI SDS masks significant loss of body mass in the very obese during adolescence. 106 We calculated weighted mean differences for continuous outcomes (e.g. BMI) and risk difference for dichotomous outcomes at the end of study follow-up. The meta-analysis used a random-effects model with RevMan 5.0.16 (The Cochrane Collaboration, The Nordic Cochrane Centre, Copenhagen, Denmark). The primary outcome analysis (BMI) was based on ITT data from the completion of the randomised trial, prior to any crossover or open-label extension. However, data on secondary outcomes and adverse events were taken from the same trial end point as the BMI data, using the highest-quality data reported in each trial (whether the ITT population or for completers). Where SDs were not reported, these were obtained from standard errors, confidence intervals (CI), t-values or p-values that relate to the differences between means in two groups. The DerSimonian and Laird Q-test was performed to assess the degree of heterogeneity between studies, and the I²-statistic was used to describe the percentage of total variation across studies due to heterogeneity. Owing to the small number of studies, we were unable to assess publication bias by inspection of Funnel plots. Secondary outcomes were included in the meta-analysis if each outcome was reported in more than two studies for each drug. Findings were reported in accordance with the QUOROM consensus statement. 106

Subjects and co-interventions

Subjects across all trials included in the meta-analysis had similar demographic profiles: the majority were aged 12–18 years, mean BMI was between 30 and 40 kg/m² and subjects were predominantly white or Hispanic. In each trial, subjects with secondary causes of obesity were excluded, as were those with DM. All trials included a standardised low-fat, low-energy diet and encouragement to exercise, with a variable element of behavioural modification in some trials.

Methodological quality

Studies were all of similar quality. All studies included an ITT analysis, reported eligibility criteria, and co-interventions were similar in intervention and control arms. The main limitation to quality was moderately high attrition rates, averaging 19% for sibutramine studies and 25% for orlistat studies. Most studies did not describe the randomisation process nor comment on allocation concealment or blinding of outcome assessors. Because there was little variation in quality, we did not perform sensitivity analyses according to study quality. Secondary end points were reported inconsistently, and frequently in a subgroup of patients or not in an extractable fashion.

Orlistat

Two studies fulfilled criteria to be included in the meta-analysis, with a total sample size for BMI outcomes of 573 adolescents (see Table 24). One ran for 6 months and one for 12 months. Each used the recommended dose of 120 mg three times daily, and participants also received behavioural, dietary and exercise counselling. Participants in both studies also received multivitamin supplements.

Figure 12 shows the pooled estimate of mean BMI change with orlistat was a reduction of 0.83 kg/m² (95% CI 0.47 to 1.19 kg/m²) compared with placebo. There was no evidence of heterogeneity. We were unable to undertake an analysis of proportions achieving 5% and 10% reductions in BMI or weight as this was reported in only one study.

FIGURE 12.

Mean reduction in body mass index (kg/m²) with sibutramine and orlistat. IV, instrumental variable. Source: Viner et al. 155

Secondary outcomes for orlistat compared with placebo are shown in Table 25. There were no significant differences in fasting lipids, glucose or insulin between the orlistat and placebo groups. As waist circumference, body fat and blood pressure were each reported in only a single study, these outcomes were not included in the meta-analysis. The effect of orlistat and placebo on changes in vitamin A, D and E levels was not included in the meta-analysis as the dose of multivitamin used in each trial was not specified; however, neither study reported any significant difference in levels of each vitamin between groups during the trial. We were unable to undertake sensitivity analyses for orlistat due to the low study numbers.

TABLE 25 - Secondary outcomes and adverse reactions with orlistat

HDL, high-density lipoprotein; LDL, low-density lipoprotein.

Source: Viner et al. 155

Secondary outcomes (fasting)	Number of studies (sample size: orlistat, placebo)	Weighted mean difference (95% CI)
Triglycerides (mmol/l)	2 (567: 368, 199)	0.00 (–0.17 to 0.18)
Cholesterol Total (mmol/l)	2 (520: 339, 181)	0.03 (–0.17 to 0.23)
HDL	2 (520: 339, 181)	0.00 (–0.02 to 0.03)
LDL	2 (518: 338, 180)	–0.05 (–0.11 to 0.01)
Glucose (mmol/l)	2 (452: 298, 154)	0.02 (–0.25 to 0.28)
Insulin (mU/l)	2 (437: 287, 150)	–0.41 (–4.83 to 4.01)
Adverse reactions	Number of studies (sample size: orlistat, placebo)	Risk difference (95% CI)
Fatty/oily stool	2 (467: 368, 199)	0.53 (0.27 to 0.79)
Oily spotting	2 (467: 368, 199)	0.49 (0.00 to 0.99)
Oily evacuation	2 (467: 368, 199)	0.51 (–0.08 to 1.10)
Faecal urgency	2 (467: 368, 199)	0.10 (0.04 to 0.16)
Flatus with discharge	2 (467: 368, 199)	0.17 (0.12 to 0.21)
Flatulence	2 (467: 368, 199)	0.05 (0.01 to 0.09)
Faecal incontinence	2 (467: 368, 199)	0.08 (0.05 to 0.11)

Adverse reactions for orlistat compared with placebo are shown in Table 25. Those taking orlistat were significantly more likely to experience a range of gastrointestinal side effects. It was not possible to assess the risk of any gastrointestinal event or study discontinuation due to gastrointestinal side effects.

Comparison with the literature

Our findings are similar to another meta-analysis of AODs in children and adolescents, by McGovern et al. ,153 who reviewed drug trials as part of a wider systematic review of childhood obesity treatment. They reported a mean BMI reduction for sibutramine of 2.4 kg/m² from three randomised controlled trials, similar to our finding of 2.2 kg/m², and a mean BMI reduction for orlistat of 0.7 kg/m², again similar to our finding of 0.83 kg/m². However, this review included unblinded studies, failed to identify eligible trials, included sibutramine and orlistat as subcategories within a larger random-effects meta-analysis and did not undertake sensitivity analyses or examine secondary outcomes in detail. A second (Cochrane) systematic review by Oude Luttikhuis et al. 42 reported similar findings for orlistat (mean reduction: 0.76 kg/m²). This review included only two small studies in the meta-analysis for sibutramine, reporting a mean reduction of 1.66 kg/m², which is considerably lower than our estimate. Neither of these studies undertook a meta-analysis of adverse events with orlistat or sibutramine. Our findings also align with an update of the Cochrane review in November 2016,176 which again reported similar findings for orlistat but noted the poor overall quality of the literature.

We are unable to directly compare our findings with adult studies of AOD, either in terms of absolute BMI loss or proportions who lost ≥ 5% or ≥ 10% of initial weight. We did not include weight loss in our review, as BMI reduction is the goal of treatment in childhood and adolescence,177 growing children may lose BMI while gaining weight, and BMI centiles continue to shift in later adolescence even after height growth has ceased. 65 However, approximate comparisons can be made for older adolescents who are at or near final height. Meta-analyses in adults show that weight loss from sibutramine and orlistat are limited to 3–4 kg over 12 months. 178 In contrast, the weight loss corresponding to a loss of 2.2 kg/m² related to sibutramine use in 14- and 15-year-old adolescents with a height on the 50th centile for sex is approximately 6 kg for both boys and girls, a relationship that holds true across the obese BMI range. The corresponding weight loss for orlistat is approximate 2.3 kg. These figures suggest the possibility that sibutramine therapy may be more effective in obese adolescents than in adults. The reasons for this are unclear. This may be an artefact of the paucity of data on adolescents and the lack of long-term data. Alternatively, this may suggest that sibutramine is more potent in suppressing appetite in adolescents, possibly because of developmental immaturity in hypothalamic appetite control systems.

We found sibutramine to have modest beneficial effects on triglycerides and HDL-cholesterol, similar to findings in adults. 178 However, we found no evidence of beneficial metabolic effects associated with orlistat use, in contrast to meta-analysis in adults, which suggests that orlistat has small beneficial effects on LDL and total cholesterol. 178 The reasons for this difference are unclear but may relate to the modest BMI loss seen with orlistat in adolescents and the small number of studies.

Study quality was relatively high; attrition rates (19% for sibutramine and 25% for orlistat) were moderately high, but lower than those reported in a meta-analysis of adult trials of these drugs (30% for orlistat, 40% for sibutramine). 178 However, the number of studies of each drug in adolescents is small, and trials have all been undertaken in secondary care settings, limiting the generalisability of findings. We identified no published studies with a duration ≥ 12 months, so we were unable to examine long-term maintenance of BMI loss.

Our findings apply to young people with simple or primary obesity. However, we note that our BMI effect size finding for sibutramine is similar to the 0.7-SD BMI reduction reported by Danielsson et al. 163 in a randomised controlled trial of sibutramine for 20 weeks in adolescents with secondary or monogenic obesity that was not eligible for our meta-analysis.

Safety

The safety profile of orlistat was similar to that noted in adults (i.e. a marked increase in unpleasant and anti-social gastrointestinal experiences but there was little evidence of significant health risk). Theoretical concerns about fatsoluble vitamin deficiencies were not supported although subjects in both trials were given multivitamins. In contrast, sibutramine was generally well tolerated by subjects but was associated with small rises in systolic and diastolic blood pressure and resting heart rate. The magnitude of these changes is highly similar to that seen in adult studies; a recent meta-analysis found that sibutramine increased adult systolic blood pressure by 1.7 mmHg, diastolic blood pressure by 2.4 mmHg and heart rate by 4.5 beats/minute. 178 Even small increments in blood pressure can have an adverse impact on cardiovascular risk in the long term, particularly in at-risk groups, such as obese adolescents, who often have high blood pressure compared with peers.

Authorities note that the long-term safety of anorectic agents has not been established in children and adolescents. 150 However, the clinical significance of small blood pressure increments over a short treatment period remains unclear, particularly when balanced against the beneficial cardiovascular effects of successful weight reduction.

Strengths and limitations

We undertook a rigorous systematic review and meta-analysis using independent reviewers adhering to the established Cochrane Collaboration methodology. In contrast to adult studies,178 studies in our review included a range of non-white ethnic groups.

However, our findings have several limitations which need to be considered. First, all published studies have demonstrated efficacy of BMI and body weight reduction in both orlistat and sibutramine. This suggests the possibility of publication bias; however, there were too few studies for either drug to warrant the generation of funnel plots to assess publication bias. Second, there was moderate but non-significant statistical heterogeneity between studies in BMI outcomes from sibutramine use. This was addressed by using a random-effects meta-analysis. It is likely that this heterogeneity is the result of differences in co-interventions, study duration and study populations. We included studies of differing length in the meta-analysis, as it was not possible to standardise duration owing to the differing timings of the ITT analysis. We did not have access to individual patient data to investigate the cause of this heterogeneity. Third, for sibutramine, we excluded one trial from the meta-analysis as study duration was only 3 months and an ITT analysis was not performed. 162 Repeating the meta-analysis for BMI including this study reduced the estimate of BMI reduction to –1.8 kg/m² (95% CI –2.65 to –0.95 kg/m²). Fourth, all included studies were conducted in specialist environments, and the generalisability of these findings to more general populations of obese adolescents is unclear. Finally, our analyses included only data that were extractable from studies, which may be a source of bias as studies may publish only secondary outcomes that differed significantly from placebo. This was the case for the largest included trial of sibutramine, by Berkowitz et al. ,170 which published only metabolic outcomes that differed significantly between sibutramine and placebo, and we were unable to include this study’s data on other not significantly different secondary outcomes in our analyses. However, we believe that this is unlikely to have been important, as we found no significant mean difference for any of these secondary outcomes in the studies that were included in meta-analyses for these outcomes.

The unlicensed use of metformin in children and adolescents in the UK

This work was reproduced and published as Hsia Y, Dawoud D, Sutcliffe AG, Viner RM, Kinra S, Wong IC. Unlicensed use of metformin in children and adolescents in the UK. Br J Clin Pharmacol 2012;73:135–9. https://doi.org/10.1111/j.1365-2125.2011.04063.x. 179

Licence to reproduce number (John Wiley & Sons): 4675940922641.

Abstract

Analyses

Analyses were conducted using Stata 11.0. Simple descriptive statistics were used for the majority of data, including chi-squared test and t-test (paired and independent) for parametric data, and Wilcoxon–Mann–Whitney and Wilcoxon-signed rank tests for unpaired and paired non-parametric data. Thematic analysis was used to analyse qualitative responses. 212

Governance

The study was reviewed and approved by the National Research Ethics Service (NRES) Committee London – Surrey Borders [Research Ethics Committee (REC) reference number 11/LO/1020].

Patient demographics

Figure 14 summarises patient sampling. A total of 151 patients were identified by THIN as having received orlistat or metformin during the study period. GP response rate was 78.8% (119/151). GPs identified 21 out of 119 patients as being ineligible for the study. An additional four subjects were excluded after inspection of questionnaires, owing to duplicate questionnaires received for the same drug with inconsistent responses (n = 1), missing patient demographics and drug details (n = 2), and an empty questionnaire (n = 1).

FIGURE 14.

Patient sampling. Adapted from White et al. 206 © Article author(s) (or their employer(s) unless otherwise stated in the text of the article) 2017. All rights reserved. No commercial use is permitted unless otherwise expressly granted. 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/.

Adapted from White et al. 206 © Article author(s) (or their employer(s) unless otherwise stated in the text of the article) 2017. All rights reserved. No commercial use is permitted unless otherwise expressly granted. 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/.

A total of 94 subjects were eligible for this study, of whom 86.2% (n = 81) were female. GPs identified ethnicity as British for 44.7% of patients (29 white, three Pakistani, one mixed, one British-African, one mixed African-white British and seven British), white for 30.9% (n = 29), Asian for 4.3% (n = 4) and other for 4.3% (n = 1 each from the Caribbean, Turkey, Iran and Afro-Caribbean/white mixed); 16.0% (15/94) were of unknown ethnicity. The majority of subjects came from England (78.7%, n = 74); the remainder came from Wales (11.7%, n = 11), Scotland (7.4%, n = 7) and Northern Ireland (2.1%, n = 2).

A total of 99 AOD initiations occurred in the 94 subjects (five subjects were prescribed both orlistat and metformin), consisting of 44 metformin (46.8% of sample) and 55 orlistat (58.5%) prescriptions. Drugs were initiated in 68 practices, with 46 practices prescribing one drug each, 15 practices prescribing two drugs each, six practices prescribing three drugs each and one practice prescribing five drugs.

Table 29 summarises baseline demographic and comorbidities by drug. Comorbidities appeared higher in those taking metformin. Sufficient data were provided to calculate BMI and zBMI during the study period for 90.9% (40/44) of those prescribed metformin and 89.1% (49/55) of those prescribed orlistat. All had a BMI above the 98th centile (> 2 SDs). Of the remaining 10, one had no recorded measurements, seven had no height data and two had no drug start date provided by GPs.

Drug initiation

Figure 15 summarises the frequency of drug prescription by age and drug initiator.

FIGURE 15.

Bar graph summarising age at AOD initiation, by drug and prescriber. Adapted from White et al. 206 © Article author(s) (or their employer(s) unless otherwise stated in the text of the article) 2017. All rights reserved. No commercial use is permitted unless otherwise expressly granted. 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/.

Adapted from White et al. 206 © Article author(s) (or their employer(s) unless otherwise stated in the text of the article) 2017. All rights reserved. No commercial use is permitted unless otherwise expressly granted. 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/.

Of 55 orlistat prescriptions, 89.1% (n = 49) were initiated in primary care, independent of specialist advice. Of this 89.1% of prescriptions, 66.7% were to those aged 12–15.9 years, 93.1% were to those aged 16–17.9 years and 94.1% were to those aged > 18 years. Orlistat was not prescribed to any patient aged < 12 years. In the case of orlistat treatment initiated in primary care on specialist recommendation, the recommending specialist was a paediatrician in three cases (50%) and an adult physician, a lipid clinic and a dietitian in one case each (16.7%).

Of 44 metformin prescriptions, 27.3% (n = 12) were initiated in primary care, independent of specialist advice. Of these, none was for those aged < 12 years, 23.5% were for those aged 12–15.9 years, 25% were for those aged 16–17.9 years and 51.5% were for those aged > 18 years. When metformin was initiated in primary care on recommendation of a specialist, 59.4%(19/32) of recommendations were from a paediatrician, 21.9% (7/32) were from a gynaecologist, 12.5% (4/32) were from an adult physician and 3.1% (1/32) were from an endocrinologist (some had multiple/joint consultations).

Indications for metformin initiation were obesity together with (1) PCOS (70.5%, 31/44), (2) insulin resistance (25.0%, 11/44), (3) impaired glucose tolerance/impaired fasting glucose (9.1%, 4/44) and (4) obesity without known comorbidity (6.8%, 3/44).

Drug monitoring

Of the 61 prescriptions initiated independently by a GP, 67.2% (n = 41) were subsequently reviewed by a GP (75% of metformin prescriptions and 65.3% of orlistat prescriptions), including three that were additionally reviewed by their practice nurse. One was reviewed by a gynaecologist, but not the GP, after drug initiation to confirm appropriateness of the drug prescription.

Of the 38 prescriptions initiated by a GP on specialist recommendation, 26.3% (n = 13) of patients had a GP consultation related to the drug for ‘supervision’ (n = 1), weighing (n = 2), psychological support (n = 1), prescribing (n = 2), answering queries (n = 1) and to discuss efficacy (n = 6). No patients discussed side effects with their GP. GPs were aware of monitoring by a clinician in 84.2% (32/38) of cases, of whom 81.8% (18/22) were followed up by paediatricians and 40% (2/5) were followed up by adult physicians.

General practitioners were made aware of adverse drug effects by two patients, both prescribed metformin: one had diarrhoea and the other had nausea.

Drug duration and termination

Duration of drug prescription is summarised in Figure 16.

FIGURE 16.

Kaplan–Meier survival curve demonstrating treatment duration of metformin and orlistat. Figure shows proportion still actively prescribed AOD for orlistat and metformin by time since initiation, beginning from time 0 (100% active prescriptions) out to 85 months (longest active prescription). Adapted from White et al. 206 © Article author(s) (or their employer(s) unless otherwise stated in the text of the article) 2017. All rights reserved. No commercial use is permitted unless otherwise expressly granted. 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/.

Adapted from White et al. 206 © Article author(s) (or their employer(s) unless otherwise stated in the text of the article) 2017. All rights reserved. No commercial use is permitted unless otherwise expressly granted. 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/.

Over half of metformin prescriptions (25/44), but only 5.5% of orlistat prescriptions (3/55), were active at the time of the survey; an active prescription was defined as a new prescription issued within the preceding 3 months. Twenty-seven patients had only a single prescription issued from primary care, being 45.5% (25/55) of all orlistat and 4.5% (2/44) of metformin treatments. The median duration for supply of metformin was 10.5 months (IQR 4.0–18.5 months), compared with 2.0 months (IQR 1.0–4.0 months) for orlistat (p ≤ 0.001). None of these single prescriptions was issued in the 3 months prior to the survey, and all were given a maximum of 1 month’s supply, making ongoing use highly unlikely. There was a disparity between the reported length of drug prescription and the amount of drug prescribed, suggesting non-continuous use at the prescribed dose.

The majority of all drug terminations were due to families not requesting repeat prescriptions (96.2% of all orlistat terminations and 89.5% of metformin terminations), rather than medically led terminations. GPs reported that in three cases orlistat cessation may have been due to a lack of drug supply in pharmacies. Of four prescriptions actively terminated by a doctor (n = 2 for metformin and n = 2 for orlistat), two were caused by lack of efficacy, one by lack of drug adherence and the other two for reasons unknown.

Adherence to the National Institute for Health and Care Excellence’s guidance

The compliance analysis by NICE12 was restricted to recommendations for children (Box 1). Using 1 January as their assumed birthday, 23 subjects were definitely aged < 18 years at drug initiation (different age criteria to analyses above).

The following criteria were fully met. First, all subjects were aged > 12 years (recommendation 1.8.4). Second, no participants were prescribed orlistat for more than 12 months (1.9.11).

The following criteria were partially met. Four (17.4%) patients were prescribed orlistat following specialist advice (recommendation 1.8.5; see Box 3). Recommending specialists were paediatricians (n = 3) and an adult physician, with one known to be part of a specialist multidisciplinary team (MDT) (1.8.7). All prescriptions recommended by specialists were continued in primary care (1.8.8).

Comorbidities were reported in 56.5% of the sample (13/23), despite NICE requiring comorbidities to be present (recommendation 1.8.6; see Box 3). Comorbidities were emotional distress (7/23), hypothyroidism (3/23), T2DM (1/23), medulloblastoma (1/23), PCOS (1/23) and worsening of another chronic disease secondary to obesity (1/23). No patients had sleep apneoa. Low levels of comorbidity screening in primary care were reported, suggesting that higher numbers of comorbidities may have existed (6/23 were screened for psychosocial distress, five were screened for hypertension, two each for T2DM and dyslipidaemia, and one each and none for sleep apnoea).

No patient was prescribed a multivitamin (1.9.2). Screening for micronutrient intake was not assessed, nor was risk of vitamin deficiencies.

Improving prescribing in primary care

General practitioners reported their confidence in prescribing AODs using a 10-point Likert scale (10 = highest). Confidence was higher for prescribing to adults (median 8, IQR 8–9) than to children (median 3, IQR 1–5; p < 0.0001). There was no difference in confidence between prescribing orlistat or metformin in children (p = 0.8).

The most popular resources used by GPs to support prescribing were NICE guidance12 (orlistat, n = 20; metformin, n = 7), British National Formulary193,194 (orlistat, n = 17; metformin, n = 8), local prescribing guidelines (orlistat, n = 8; metformin, n = 11) and specialist guidance (orlistat, n = 1; metformin, n = 7).

General practitioners perceived that 27.3% (n = 12) of patients prescribed metformin and 12.7% (n = 7) prescribed orlistat benefited from the drug; half (50% of those prescribed metformin and 52.7% of those prescribed orlistat) reported not knowing if there had been any benefits for the patient.

Thirty-five GPs provided free-text reflections of their experiences of prescribing orlistat (n = 20) and metformin (n = 14). Three main themes arose. First, metformin was mostly prescribed to PCOS rather than as a weight loss drug. One GP stated that (s)he ‘wouldn’t normally prescribe this just for weight loss’.

Second, there was controversy about whether or not AODs should be prescribed in primary care in this age group, with one GP saying (s)he would ‘usually not prescribe for children’ and another saying (s)he avoided orlistat ‘where possible’. Either metformin was issued ‘on advice of specialist only’ or the patient received specialist follow-up after initiation.

Third, GPs noted concern about the efficacy of these drugs. ‘Inadequate counselling’, lack of drug availability and patient compliance (‘clearly patient was not able to comply’) were hypothesised reasons for ineffectiveness.

Sixty-two GPs wanted improved support, primarily split into two main themes. First, they requested improved age-related guidance for prescribing AODs that is ‘realistic’, with ‘clear [and] concise’ advice including ‘flow diagrams’ and ‘stepwise advice’. This would include instructions on assessment prior to initiation, indications, contraindications, monitoring, safety advice, duration, targets and indications for stopping treatment. Second, they wanted improved guidance for managing patients with obesity, namely advice about lifestyle management and details of available interventions. GPs requested details of ‘non-drug treatments’ including ‘community support for adolescents’ and ‘special clinics for monitoring and support of patients.’ One GP highlighted that ‘non-drug treatments need to be key alongside drug treatment.’

Recruitment and sample

Young people aged 12–18 years were eligible if they had BMI ≥ 98th centile215 and had been prescribed orlistat or metformin for weight control within the last 3 years. Exclusion criteria were (1) use of metformin for management of T2DM, pre-DM or PCOS in non-obese young people, (2) inability to participate in a face-to-face interview or (3) insufficient mastery of English language to participate. Young people were recruited from three paediatric obesity clinics in England (London, Bristol and Liverpool) through the Medicines for Children Research Network. We anticipated requiring approximately 10–20 families to achieve theme saturation. 212 Based on our previous difficulties recruiting participants into obesity studies, all 109 subjects fulfilling eligibility criteria were invited by their hospital doctor to enrol in the study. Those consenting were invited to interview, together with one parent, and each given a £10 gift voucher as participation compensation.

Face-to-face interviews were conducted in the participant’s own home by one researcher (LJ) experienced in interviewing young people. Parent/young person dyads were independently interviewed unless either of the dyad opted out or they wished to be interviewed together. Written assent for under-16-year-olds and consent for those over 16 were obtained from both the young person and parent/carer. Interviews were audio-recorded and field notes written immediately after the interview.

Analysis

Audio recordings from each interview were transcribed verbatim and anonymised by LJ. Transcripts and field notes were read and coded independently by LJ, and a subsample coded by BW using a general thematic coding methodology. 216 Memos were written to summarise and synthesise emerging themes. This initial coding framework was used to code the subsequent transcripts and new codes were added as they emerged using a constant comparative technique to compare new and previously collected data to understand emerging themes. To ensure reliability, BW read all transcripts and reviewed the coding. BW and LJ developed models through an iterative process, in which the initial model was reviewed using constant comparison techniques (in which successive items of data are appraised and compared to ensure that the code is reflective of all) and the models revised accordingly. The qualitative analysis was facilitated by the use of NVivo software [QSR International (UK) Limited, Southport, UK].

Model 1: factors influencing why young people commenced on an anti-obesity drug

Six themes fed into the decision by young people to take the AOD: passive acceptance of the AOD, enthusiasm and relief at the prospect of a drug treatment, medication as a last resort, fear as a motivating factor, AOD as a way out of obesity and their own perceived uniqueness (Figure 17).

FIGURE 17.

Model 1: factors influencing the commencement of an AOD. Source: White et al. 209

Model 2: management of side effects

The overall themes in model 2 can be seen in Figure 18.

FIGURE 18.

Model 2: management of side effects. Source: White et al. 209

Side effects from AOD were a key issue for many young people, although a minority did not experience any. Side effects of both metformin and orlistat were usually gastrointestinal, particularly abdominal cramps and diarrhoea. For some, this involved spending ‘many, many hours on the loo’ (young person 15, boy aged 15 years, using metformin), and ‘really weird, intense pain’ (young person 1, girl aged 18 years, using metformin). Some were taken aback by the severity of the side effects; one young person reported ‘I have never seen diarrhoea like it’ (young person 6, girl aged 14 years, using orlistat), while another said ‘we were told that there could be mild stomach upsets or whatever but I didn’t expect it to be as uncomfortable as it was’ (young person 15, boy aged 15 years, using metformin). A few young people attributed unusual symptoms to metformin, including hand tremor, headaches and change in moods.

Parents expressed concern about their children having to experience such unpleasant side effects. One mother felt that it was ‘totally wrong’ for her daughter to have such side effects at her age (parent 6, young girl using orlistat):

It was awful and he would do it [faecal incontinence] in his trousers and he would phone me up and say Mum, I need to come home.

Parent 5, young boy

Nearly all young people were prepared to endure some side effects as their doctor had forewarned them. However, there was personal variation as to how much each individual tolerated these side effects; some continued taking the medication despite episodes of faecal incontinence while others stopped with much milder side effects. Some young people were more resilient than others in coping with side effects.

Some initiated self-devised lifestyle strategies or changes in drug regimen to minimise side effects; we identified a number of key likely mediators that influenced how these side effects were managed.

Model 3: drug continuation – efficacy versus side effects

The overall model for the decision to continue or stop taking the prescribed AOD is shown in Figure 19.

FIGURE 19.

Model 3: patient-led decision process regarding drug continuation. Source: White et al. 209

Participants continued with the medication for between 1 month and 8 years. Nine out of 12 were prescribed metformin, and one of six participants prescribed orlistat continued beyond 6 months. Seven young people discontinued an AOD. The decision to either continue with, or stop, the AOD was frequently based on a decisional balance between the efficacy of the AOD and ongoing side effects. The decision to terminate treatment was frequently described as a balance between the perceived benefits of the AOD and its side effects. Various mediators influenced this decision, including perceived benefits and expectations, lack of support and understanding of drug action:

I don’t want to take something that I don’t think was working and making me ill.

Young person 6, girl aged 14 years

I just thought what is the point of taking it if it is not working and I am not eating the rubbish foods, there is no point as it weren’t really doing anything.

Young person 3, girl aged 16 years

Participants described efficacy in terms of body weight, body shape and metabolic parameters. Individual goals varied from going down a clothes size to weight stabilisation. Most young people had expectations of weight control that were aligned with published outcomes, although some hoped for outcomes that were faster, more extreme or more guaranteed. Weight stabilisation was acceptable for some participants, and they continued to take it, fearing that their weight would increase faster if they stopped taking the medicine:

He [clinician] said that they may help me lose weight and my mind crossed out the word ‘may’ and replaced it with the word ‘will’.

Young person 9, aged 17 years

It is keeping her weight not going up.

Parent 10, young girl

There was significant variation in the understanding of drug action, in terms of mechanism and efficacy. Some perceived that the AOD required a restricted diet and increased exercise to be effective, while others believed that lifestyle changes were not necessary. These views seemed to be unrelated to the drug prescribed. Some described the futility of taking an AOD as they were unable to undertake healthy behaviours, and subsequently stopped the drug:

. . . they said this [the medicine] is not a miracle worker it doesn’t help you lose weight. You help yourself to lose weight and it just gives you a little pat on the back every so often to help you carry on what you need to do . . .

Young person 10, girl aged 14 years

I don’t know whether the tablet actually makes you lose weight or just because it makes you stop eating, it makes you lose weight.

Young person 2, girl aged 17 years

Drug termination was an active decision by young people and their families, and not by their doctor. The decision was taken by the young person alone, or with the advice and support of their parent. Very few young people or parents reported adequate drug monitoring and support from the obesity specialist; this influenced their decision to independently stop the drug:

She [Mum] advised me not to take them because it wasn’t very nice for me experiencing this.

Young person 6, girl aged 14 years

I just stopped taking them, went cold turkey.

What put you off ringing up the clinic to discuss it?

I just didn’t think it was important.

Young person 15, boy aged 15 years

Support was a theme that spanned across all three models. One mother described the period taking the AOD as a ‘lonely’ time. Few reported adequate support from their obesity specialist, primary care physician or pharmacist. Two participants described disheartenment after being discharged from a specialist service due to inadequate progress, and reported subsequent weight gain. General practitioners mostly only issued repeat prescriptions. Many young people said that they would be happy to be monitored by their GP if they could not get an appointment at the specialist clinic, yet others felt that GPs had insufficient experience to support them. Emotional support mainly came from friends and family. Parental supervision, usually from a mother, ensured that younger adolescents took the medication. As young people matured, parents were more likely to step back and let the young person take responsibility for their medication:

I think there were times when she tried skipping it, but she had a dragon as a mother. So as long as I am aware it happens, and what I do now is put it all out in individual pill boxes for the day.

Parent 13, young girl

Limitations

Participants were recruited from three hospital clinics in England, with the majority from one hospital in which two of the authors are clinicians. This has the potential of limiting generalisability and introducing bias. However, AODs in young people are largely initiated in specialist centres, and those centres included in this study were among the largest of a very small number of specialist paediatric obesity clinics in the UK. To minimise bias, all data were collected by an independent researcher not part of any clinical team and responses were anonymised before analysis. Delays in the study may have led to problems with recall for those patients who stopped the medication some time previously. As with all qualitative studies, the researcher’s presence during interviews could have affected the subjects’ responses; every effort was made to reassure the participants that the researcher was both non-judgemental and not part of the clinical team, and their responses would be fully anonymised prior to analysis by the team. We aimed to include patients who were prescribed an AOD but never took it. However, no such young people responded to recruitment invitations. We are therefore unable to comment on those participants who were prescribed an AOD but who never initiated medication. Only a small minority of eligible subjects enrolled in the study, despite thorough attempts to contact them using clinical research nurses. It is highly possible that those with negative experiences felt more motivated to participate in the study. Largely negative responses indicate that participants are likely to have felt reassured about their anonymity, and did not fear reprisal from their clinicians.

Background

Bariatric surgery is currently the only evidence-based intervention resulting in highly meaningful weight loss for obese children and young people. 224 In the UK, national guidelines for adolescent bariatric surgery are contained within the most recent NICE obesity guideline12 summarised in Box 3. It recommends that bariatric surgery is considered only in exceptional circumstances, and only if young people have achieved or nearly achieved physiological maturity.

Published adolescent data have largely focused on weight outcomes of surgery. 1 The largest and most complete data sets have come from the USA155 and Sweden,226 with only very small case reports published from the UK. 222,227,228 A recent systematic review of adolescent bariatric surgery showed mean loss of 13.5 kg/m² at 1 year post surgery with greatest change seen in those undergoing the Roux-en-Y gastric bypass (RYGB), followed by sleeve gastrectomy (SG) and then adjustable gastric band (AGB). 224 Although data were drawn largely from observational studies, the magnitude of effects in these were similar to those reported in the only randomised controlled trial undertaken in adolescents. Such changes are very much greater than those seen with drug treatments, such as metformin (mean loss 1.4 kg/m²) and229 orlistat (–0.8 kg/m²),10 or with lifestyle interventions (–1.25 kg/m²). 224

Psychosocial outcomes after surgery have been less rigorously studied, with four large cohorts reporting outcomes for up to 2 years after surgery. 226,230–232 However, there are no published outcomes of eating disorders, and only one study exploring predictors of success after surgery.

Furthermore, outcome studies have universally focused on those people who received surgery,224 with no studies of patient pathways into adolescent bariatric surgery programmes. It is unclear (1) how many young people who are referred for surgery actually receive it and (2) the reasons for not receiving it in those referred. Such data are important for planning and evaluating adolescent bariatric surgery programmes, particularly within state-funded systems.

Aim

The aim of this substudy was to describe the characteristics and outcomes of patients referred to an adolescent bariatric surgery programme within the NHS in England.

Methods

All patients referred to an adolescent weight management service for consideration of bariatric surgery between its inception in 2007 and 27 January 2014 were included. Subjects were referred by their GP orlocal paediatrician, or from within our medical weight management service. Patients were identified by means of an ongoing database. Body weight was measured to the nearest 0.1 kg using a Tanita BC-418MA scale (Tanita, Europe, Amsterdam, the Netherlands) or SECA 645 scale (Seca GmbH, Hamburg Germany) where body weight exceeded 200 kg. Height was measured to the nearest millimetre using a wall-mounted stadiometer. Body mass index (BMI) was calculated as weight divided by height squared (kg/m²). Patients selected the ethnicity with which they most associated, based on a predetermined list used by the hospital trust. This list included only one category for British, and did not differentiate between white British, black, Asian or other ethnic groupings. Socioeconomic status was derived from the patient’s home postcode using the 2007 Index of Multiple Deprivation30 (IMD) score and rank (England only). The 32,482 IMD ranks were split into 10 groups to allow us to rank patients by deprivation deciles (decile 10 being the most deprived). This study fulfilled National Research Ethics Service (NRES) criteria for service evaluation, and was registered in the UCLH Research & Development Department.

Results

Patient flow and characteristics

The patient flow is summarised in Figure 20. Fifty-two patients were referred for consideration of bariatric surgery. Two young people had bariatric surgery in the private sector while waiting for their initial assessment, and so were not eligible for the pathway and are not included here.

FIGURE 20.

Flow chart of patients through the bariatric pathway. Source: White et al. 223

Fifty young people were assessed for eligibility for bariatric surgery between 26 July 2007 and 27 January 2014. Demographics are summarised in Table 31. Thirty (60%) were aged 16 years or over at their first assessment. As of 27 January 2014, 42 had completed the preoperative pathway and eight were still undergoing assessment and/or preparation for surgery.

TABLE 31 - Demographics, BMI and cardiometabolic characteristics for the complete cohort, those completing the pathway and undergoing surgery and those completing the pathway but not undergoing surgery

Participants in the bottom two deciles of the 2007 Index of Multiple Deprivation.

Source: White et al. 223

Characterisitc	Whole cohort (N = 50)	Completed
		Surgery (N = 20)	No surgery (N = 22)
Female, n (%)	29 (58)	11 (55)	11 (50)
Mean (SD) age (years) at baseline	16.0 (1.3)	16.4 (1.1)	15.6 (1.3)
Age (years) range	12.8–18.5	13.3–17.8	12.8–17.2
Mean (SD) baseline BMI (kg/m2)	51.2 (7.5)	51.8 (8.7)	50.8 (7.5)
BMI range (kg/m2)	37.5–69.6	38.2–69.6	37.5–64.8
High–normal blood pressure, n/N (%)	10/48 (20.8)	3/20 (15)	4/20 (20)
High blood pressure, n/N (%)	16/48 (33.3)	8/20 (40)	6/20 (30)
High–normal lipids, n/N (%)	25/43 (58.1)	10/20 (50)	9/15 (60)
High lipids, n/N (%)	17/43 (39.5)	9/20 (45)	6/15 (40)
T2DM, n/N (%)	6/50 (12)	2/20 (10)	3/21 (14.3)
Ethnicity, n (%)
British	29 (58)	9 (45)	14 (70)
African/Caribbean	8 (16)	4 (20)	3 (15)
Asian	5 (10)	4 (20)	1 (5)
Other	8 (16)	3 (15)	2 (10)
Deprivation decile,a n (%)	18 (36)	6 (30)	8 (40)

Of the 42 young people, six (14.3%) were deemed not eligible for surgery and were discharged from the pathway, two after the initial assessment and four during the preparation phase. The principal reasons for exclusion were (1) resident outside the UK, which the clinical team deemed to be incompatible with safe postoperative monitoring (n = 1); (2) inability to consent owing to learning difficulties (n = 2); (3) severe needle phobia (n = 1); (4) inconsistent desire to undergo bariatric surgery (n = 1); and (5) complex behavioural difficulties (n = 1). No patients were excluded because of psychiatric disorders. A further 15 participants (35.7%) did not proceed along the bariatric surgery pathway, due to either active withdrawal after receiving further information about surgery (n = 8) or lack of engagement (n = 7). Two (5%) patients were lost in the transfer between adolescent and adult bariatric surgery services within the same hospital. We have been unable to ascertain if they underwent surgery at a different centre. Twenty (47.6%) patients underwent bariatric surgery during the study period, and their demographics can be seen in Table 31.

Bariatric surgery

All patients fulfilled criteria specified by NICE. All patients had BMIs greater than 40 kg/m² at the time of surgery planning. Deprivation data were available for all 39 participants from England. Graphically, it appears that patients from deprived populations were less likely to get to surgery (Figure 21); however, there was no difference in access to surgery between those patients in the most deprived quintile and the remainder of the cohort (p = 0.4).

FIGURE 21.

Distribution of socioeconomic status in those patients completing the bariatric pathway. Source: White et al. 223

The mean time between first appointment and surgery was 1.8 years (SD 0.7 years) (range 0.5–3.7 years). Those patients with higher BMIs had a longer time to surgery (0.04 years per kg/m², 95% CI 0.00 to 0.07 years per kg/m²; p = 0.08) after adjustment for age, although deprivation and duration of the programme did not influence time.

Patients underwent RYGB and SG in equal numbers (10 each), with no patients having LGB. The mean BMI in those patients undergoing SG was higher than in those undergoing RYGB [57.6 (SD 10.2) vs. 49.6 (SD 6.9) kg/m², respectively; p = 0.06]. There was no difference in age between those undergoing each procedure.

Body mass index trajectories

Figure 22 shows the BMI trajectory of each patient undergoing surgery in the adolescent pathway and the change in BMI relative to day of surgery. A substantial variability in preoperative BMI trajectories was seen across the subjects, with up to 10 kg/m² weight gain seen in the time before surgery for some, and minimal visible change for others. Varying changes in BMI were seen during the 2- to 4-week preoperative hypocaloric diet phase, with weight change varying from minimal change to over 5 kg/m².

FIGURE 22.

Individual BMI trajectories in relation to day of surgery. Source: White. 245

Similarly, a substantial variability in BMI outcomes was seen after surgery, with only limited data after 2 years. All patients rapidly lost weight in the first 3 months after surgery at a similar rate. The majority of patients lost weight rapidly in the first 6 months, with subsequent deceleration in the rate of weight loss in the next 6 months and weight stability in the second year. The majority lost between 10 and 20 kg/m² in this time period, with subsequent maintenance of weight loss. One patient lost over 30 kg/m² in the first year after surgery although this magnitude is probably exaggerated by his short stature secondary to achondroplasia. Two patients regained all weight lost after bariatric surgery in the second year after surgery.

The two patients with hypothalamic obesity secondary to craniopharyngioma had differing outcomes. One regained all weight lost in the second year after surgery and the other regained substantial weight, although not back to presurgical levels.

Nine patients had reached 1-year follow-up at our centre at time of analysis. The mean change in BMI at 1 year was –15.2 kg/m² (SD 9.3 kg/m²). There was no evidence of difference in means between the two surgical procedures [mean RYGB –12.7 kg/m² (SD 2.7 kg/m²), SG –17.2 kg/m² (SD 12.6 kg/m²); p = 0.5]. Greater variation in BMI change at 1 year was seen in patients undergoing SG than in those undergoing RYGB, with post-SG change ranging from −33.4 to 0.2 kg/m², compared with −16.3 to –10.4 kg/m² post RYBG.

Data were sufficient to develop a multilevel model of BMI change associated with surgery, including 249 data points on the 20 patients who had surgery (Table 32 and Figure 23). The best model included a linear term describing the period before surgery and linear and quadratic terms for the period after surgery. Figure 24 shows that the mean BMI trajectory postoperatively was very rapid loss in the first 3–6 months with a nadir at 12 months and regain in BMI thereafter. Owing to limited data we did not extend the model beyond 18 months or include predictors in the model.

TABLE 32 - Multilevel models derived for BMI trajectories relative to surgery (model 6 best fits the data)

AIC, Akaike Information Criterion; BIC, Bayesian Information Criterion.

Source: White. 245

	Model
	1	2	3	4	5	6
Fixed terms	No change	Time (years)	Time (years) model 2 + surgery	Model 3 + surgery × time (years)	Model 4 + surgery × time2 (years2)	Model 5 + surgery × time3 (years3)
Random terms	No change	Time (years)	Model 2 + surgery	Model 3 + surgery × time (years)	Model 4 + surgery × time2 (years2)	Model 5
Fixed effects
Composite model
Intercept	50.53 (46.69, 54.38)	55.26 (51.14, 59.14)	54.43 (50.49, 58.37)	52.21 (48.46, 55.95)	52.05 (48.35, 55.74)	52.06 (48.35, 55.78)
Time (linear)		3.09 (–4.39, –1.79)	–0.24 (–1.59, 1.11)	2.48 (1.40, 3.55)	2.60 (1.67, 3.53)	2.60 (1.66, 3.54)
Surgery			–7.37 (–9.02, –5.73)	–6.98 (–8.09, –5.88)	–4.94 (–5.80, –4.07)	–4.58 (–5.43, –3.74)
Surgery × time				–12.22 (–17.09, –7.37)	–26.87 (–34.47, –19.27)	–29.68 (–36.58, –22.78)
Surgery × time2					10.71 (6.72, 14.70)	14.07 (10.44, 17.71)
Surgery × time3						–1.22 (–1.72, –0.72)
Variance intercept
SD (intercept)	8.59 (6.22, 11.86)	8.97 (6.42, 12.55)	8.64 (6.20, 12.02)	8.35 (6.05, 11.52)	8.32 (6.06, 11.42)	8.38 (6.11, 11.49)
Years in bariatric
SD (years in bariatric)		2.33 (1.49, 3.65)	2.20 (1.40, 3.45)	1.61 (0.91, 2.83)	1.64 (1.05, 2.58)	1.66 (1.07, 2.58)
Correlation with baseline		–0.14 (–0.59, 0.39)	–0.26 (–0.66, 0.25)	–0.17 (–0.63, 0.39)	–0.06 (–0.53, 0.44)	–0.12 (–0.57, 0.38)
Correlation with surgery × time				–0.29 (–0.73, 0.33)	–0.24 (–0.68, 0.34)	–0.14 (–0.62, 0.42)
Correlation with surgery × time2					0.49 (–0.32, 0.89)	0.29 (–0.48, 0.81)
Surgery × time
SD				9.15 (6.04, 11.87)	15.22 (10.56, 21.95)	13.43 (9.18, 19.67)
Correlation with baseline				–0.31 (0.70, 0.22)	–0.50 (–0.79, –0.01)	–0.51 (–0.80, –0.04)
Correlation with surgery × time2					–0.95 (–0.99, –0.77)	–0.94 (–0.99, –0.74)
Surgery × time2
SD					7.60 (4.66, 12.38)	6.04 (3.65, 10.00)
Correlation with baseline					0.44 (–0.22, 0.82)	0.55 (–0.11, 0.87)
Residual SD	6.20 (5.70, 6.74)	4.93 (4.48, 5.42)	4.27 (3.88, 4.69)	2.70 (2.44, 2.98)	1.93 (1.74, 2.14)	1.85 (1.66, 2.05)
Number of observations	291	249	249	249	249	249
Log-likelihood	–676.53	–798.99	–765.62	–679.26	–621.85	–611.26
AIC	1959.06	1609.98	1545.24	1380.71	1275.7	1256.52
BIC	1970.08	1631.08	1569.96	1419.40	1331.98	1316.32

FIGURE 23.

Superimposed BMI trajectories of patients undergoing bariatric surgery, by type of surgery. Source: White et al. 223

FIGURE 24.

Modelled BMI trajectory for the average adolescent patient, with surgery 1.8 years after first assessment.

Limitations

We present data from a small number of young people having surgery at a single site and follow-up data were very limited. Our study is subject to the common limitations of service evaluation studies (e.g. data were not collected at standardised time points and collection of comorbidity data at follow-up was also not standardised, limiting data availability). We did not routinely collect data on quality of life or psychosocial functioning. Significant loss to follow-up postoperatively is a potential source of bias, suggesting that our findings may underestimate or overestimate BMI loss and complications. It is possible that those with poorer outcomes were less likely to return for follow-up. However, our clinical experience was that those with poorer BMI outcomes were often keen attenders at follow-up (to work with us on increasing weight loss). The majority of those who decided against surgery did not remain within our service, and we are unable to compare their outcomes with the outcomes of those who had surgery.

Aim

The aim of this substudy was to investigate the process of clinicians’ decision-making around bariatric surgery for adolescents, and understand how professionals manage in the context of limited research and guidelines for bariatric surgery for young people, in order to reach a decision as to which adolescents should be offered surgery and which should not.

Methods

Results

Following analysis of the data, extracts have been taken from various interviews, which are characteristic of the dominant themes emerging within the data analysis (Figure 25).

FIGURE 25.

Main themes and subthemes emerging from the interviews. Source: Doyle et al. 247

In order to protect the anonymity of participants, names are not stated. Clinicians will be referred to by a number indicating the order in which interviews were completed (C1–C9).

Discussion

The paper by Doyle et al. 247 demonstrates that making decisions about bariatric surgery for young people is a challenge for health-care professionals in bariatric surgery teams across the UK, characterised by multiple uncertainties. To our knowledge there has been only one other study investigating the views of medical professionals on this topic. Bailey et al. 249 conducted a qualitative study to understand Canadian physicians’ attitudes to the treatment of paediatric obesity. Physicians were more accepting of medical treatment and lifestyle management of obesity in young people than of surgical treatment and showed scepticism for this in the absence of long-term data. In our UK study, the challenges were described as coming from the perception of adolescent bariatric surgery as a controversial subject for the public, as a complex intervention unlike other medical treatments, with limited outcome data to guide decision-making. This is further complicated by a view of adolescence as being a turbulent developmental period, with some professionals querying whether adolescents are capable of future-orientated thinking that is thought to be necessary for good postoperative outcomes.

Social scientific research into decision-making amid uncertainty suggests that a range of different ‘tools’ are drawn on to manage uncertainty. 250 These include more rational-calculative approaches, based on empirically ‘proven’ knowledge or non-rational strategies (e.g. decisions based on beliefs, hopes or emotions). Trust, as placed in either the patient, other colleagues or self-judgement, enables uncertainty to be ‘bracketed away’ and complexity reduced, thus facilitating decisions and action. 251 In this study, professionals are unable to draw on the preferred method of making decisions based on empirically derived knowledge. In the absence of this, professionals manage uncertainty by looking for other knowledge (i.e. behavioural information and signs of an adolescent’s ability to make preoperative changes), signs that appear to make intuitive sense to the professional, building his/her confidence that it is safe to proceed to surgery. Trust in the workings of the multidisciplinary process is also one way in which the dilemmas around surgery are bracketed away, although this too is a source of uncertainty, with a number of professionals questioning the value of a consensus decision, doubt around the roles of different professionals and who ultimately has legal responsibility for the decision. Some professionals therefore have looked for other ways of managing uncertainty, arguing that adolescence should be bypassed completely. Furthermore, viewing surgery as inevitable and a treatment that can ‘normalise’ allowed teams to dispel many of the uncertainties connected to the provision of a treatment still relatively controversial to the public.

Current NICE guidelines with regard to bariatric surgery emphasise the need for a multidisciplinary approach to decision-making. However, up until now there has been very little written about the nuances and complexity of these interactions with adolescents and health-care professionals. At present, professional decisions appear to be driven by intuition and ‘judgement calls’ rather than empirical data. This is not to suggest that empirically derived knowledge is superior or that patients are deliberately being misled, however, it is argued that this process needs to be made transparent. Revealing the ways in which clinicians work through, and reach, decisions amid uncertainty facilitates greater awareness of difficulties and a more reflective practice as a result. Adolescents too ‘bracket away’ uncertainty regarding bariatric surgery. 252 A ‘shared decision-making model’ is one means of reaching an agreement regarding appropriateness of surgery,253 yet this approach potentially becomes more problematic when dealing with adolescents in terms of heightened power asymmetries. It is also potentially complicated by the active process of bracketing away uncertainty by both professionals and adolescents.

This study has made use of IPA as an insightful analytical framework for exploring participants’ experiences of the phenomenon in question, while recognising that this involves interpretation by the onlooker. However, many have argued that language is not merely a means to express reality but instead that it prescribes what we can feel and think (i.e. that language both reflects and shapes reality). 254 One medium by which professionals moderate the dilemma around surgery for young people is to accept the inevitability of surgery. Surgery is seen as something that can have profound effects to ‘normalise’ and there is an implicit acceptance that offering something (surgery) is possibly better than doing nothing. It is argued here that these assumptions need to be interrogated and questions need to be asked as to how this way of thinking came to be taken for granted. Brown255 has argued that that examining these interactions has implications for policy and practice in ‘underlining for practitioners the significant interpretive scope to which their actions/utterances are liable’. These questions are essential as the discourses around the inevitability of surgery for a normal life are ones that adolescents themselves take up and, therefore, within this context, the option not to choose surgery if it is offered is potentially rendered invisible. It is argued that the uncertainty that professionals experience, around what is known and not known, needs to be shared with potential patients. However, professionals must also question the social context that informs their thinking, such that, if surgery is offered, accepting and not accepting surgery are both regarded as legitimate decisions.

Moves towards a shared decision-making model could usefully include the following developments:

• professional agreement about the limits of knowledge as regards adolescent bariatric surgery and support for staff to communicate this information, in a context where patients often believe that ‘doctors know best’

• parent- and adolescent-friendly information leaflets outlining what is known and not known about adolescent bariatric surgery

• adolescent-specific support groups and web-based resources for young people battling overweight, such that adolescents are in a position to become more equal partners in medical consultations around managing overweight and considering the pros and cons of surgery.

There are a number of limitations to this study that need consideration. The research has taken place in the UK, where adolescent bariatric surgery is an emerging field. Therefore, the views expressed herein may not be representative of those of professionals where surgery for young people is more readily available, for example in the USA and Australia. Investigating whether these countries have been through a similar process in the early stages of the availability of bariatric surgery, and how this has been managed, would be of interest.

Background

In 2010, just over one-quarter of adults in England were classified as obese, with a BMI of ≥ 30 kg/m². A NIHR Health Technology Assessment report256 concluded that bariatric surgery is a clinically effective and cost-effective intervention for moderately to severely obese individuals, in comparison with non-surgical treatments. The NICE guidelines11 indicate that bariatric surgery is not generally recommended for children or young people and should be offered only in ‘exceptional circumstances’. There is no national database to record bariatric surgery performed specifically in adolescents, although anecdotal evidence suggests that this, too, is increasing.

Adolescent bariatric surgery is an area that is fraught with many ethical dilemmas. 257 Although it might seem logical to postpone surgery as long as possible, there is evidence that weight loss surgery may be less effective the longer it is delayed and the more the health risks of obesity increase. Despite this, many clinicians do not agree that bariatric surgery performed in adolescence is an appropriate intervention. One survey of > 300 family physicians and paediatricians based in the USA revealed that 48% would not ever consider referring an adolescent for bariatric surgery, and the most frequently endorsed age at which the doctors would consider referring was 18 years. 258 Furthermore, this research team’s study of bariatric surgery teams based in the UK revealed a great deal of uncertainty around bariatric surgery for young people and how to decide which adolescents should be put forward for surgery, in the absence of a firm evidence base or specific guidelines in this area. 247 Professional uncertainty about weight loss surgery for adolescents is likely to be compounded by the fact that adolescents who are delayed from having surgery often change their minds about proceeding.

Ogden et al. 259 conducted interviews with 15 adults who had bariatric surgery. The authors reported on the ‘paradox of control’ in that, although many turned to bariatric surgery because they no longer had faith that they could lose weight without medical assistance, the imposed restriction of surgery seemed to enhance the sense of self-efficacy and sense of control. A recent qualitative study260 explored attitudes and beliefs regarding bariatric surgery for young people among obesity specialists, morbidly obese adolescents and parents. These authors found that views on the aetiology of obesity (i.e. whether it should be considered primarily a medical condition or more a psychosocial problem) seemed to affect decision-making around bariatric surgery. Parents and adolescents who saw obesity as something that they could influence themselves were more in favour of non-surgical treatment, and vice versa. Apart from these two studies, however, there has been very little in-depth research on how adolescents make decisions about the appropriateness of bariatric surgery.

Methods

Analysis

Investigators PB, SC and JD analysed the nine written transcripts using IPA, following procedures outlined by Smith et al. 248 In line with IPA, the thematic analysis was especially focused on the various ways in which participants sought to make sense of and construct meaning within the situations being described, developing an ‘insider’s perspective’248 of how these professionals experienced the phenomena being described. Initial parallel analyses were completed by JD and SC and a combined preliminary outline of themes was produced. Investigator PB completed secondary analysis of transcripts and added to the conceptual framework to produce a final thematic analysis of the data. Critical discussions of coding and the development of themes took place between the researchers, as a means of enhancing reliability of coding and the construct validity of the themes emerging in the analysis.

Results

Following data analysis, extracts have been taken from various interviews that are characteristic of the emerging themes and subthemes (Figure 26), as understood in terms of both repetition and salience within and across participants’ accounts. 261 To protect their anonymity, participants are referred to by a number indicating the order in which interviews were completed [participant (P) 1–P9].

FIGURE 26.

Main themes and subthemes emerging from adolescents’ interviews.

Discussion

There have been very few qualitative studies of adolescents considering bariatric surgery. This research demonstrates that young people have a range of motivations for choosing surgery, including wanting to be normal, confident and healthy, as well as not wanting to die young. Surgery is seen as the last option when people have exhausted all other options; therefore, choosing the right method and ‘working at it’ is viewed as vital. Despite the apparent advantages of surgery, the young people interviewed had several reservations and faced multiple uncertainties. The dilemmas were partly personal in nature, involving the relative importance of weight loss versus having excess skin, the mortality risks and also dilemmas around personal maturity. These and other dilemmas were also social in nature, as some young people worried about the impact of changes in eating behaviour on their relationship with family, the views of others about bariatric surgery and the perception of it as a form of cheating. These findings cast doubt on the assertion that adolescents are perhaps not sufficiently future orientated enough to decide about such an intervention. 247 Adolescents, however, do appear to actively bracket away these concerns, by focusing on the positives (hope), referring to and (re)interpreting probabilistic expressions of risk, trusting professionals and developing knowledge and familiarity through the influence of family and social media.

Puhl and Heuer262 have clearly documented negative attitudes towards obesity and the pervasive nature of weight bias in Western society. The apparent perception of obesity as a ‘lifestyle choice’ is prevalent, irrespective of growing evidence of strong genetic influences on adiposity despite the obesogenic environment. 263 It has also been hypothesised that weight stigma is stressful and that this ultimately begets weight gain through a complex relationship between physiological responses (e.g. increased cortisol production) and behavioural responses, such as increased eating behaviour. 264 Rees et al. 265 conducted a review of 28 qualitative studies that reported on the perspectives of UK children aged 4–11 years on obesity or body size, shape or weight. The authors demonstrated that, instead of a focus on health, children emphasised the social impact of body size, describing experiences and awareness of abuse and isolation for children with a greater weight. A second review266 among those aged 12–18 years pointed to a similar trend. Regardless of their own size, young people were judgemental of individuals who were overweight, and those who had experience of obesity described an environment that contained multiple barriers to weight loss. The authors concluded that the studies revealed a ‘picture of a stigmatising and abusive social world’. 266 Interestingly, this is more implicit than explicit in the accounts presented in this report. In this study, young people alluded to such attitudes in their accounts of their desire for surgery to help ‘normalise’, that is, in the words of one participant, surgery was seen as a way of ‘feeling welcome . . . like not an outcast’. It is, perhaps, in this context that the need or desire to ‘bracket’ away uncertainty about surgery should be viewed.

In terms of the strategies used, the young people do not appear to follow procedurally rational approaches to manage uncertainty. They do reference probabilistic information, but, although this is decidedly useful in planning and predicting across large numbers of cases, the impossibility of knowing whether any one person is the one or the 99 leaves residual uncertainty that still has to be overcome. 267 This remaining uncertainty, along with anxiety about the unknowable future, still needs to be bridged over or ‘bracketed away’. 268 The young people, therefore, make use of other methods, such as trusting in professionals, media and family, in order to cope with both aspects of the dilemma. Alongside uncertainty pertaining to the future, the complexity pertaining to various different outcomes, and the need to decide between these, may be further neutralised by viewing surgery as the last and only resort. In the context of these young people having tried many other options that have failed to bring about lasting and sustained weight loss, perceiving a particular format of surgical intervention as the only option provides a very useful way forward in reaching a decision while minimising anxiety. 255

There are a number of implications of this study. In order to consent for treatment in their own right, the treating clinician has to ascertain that the adolescent is competent; is able to understand and retain the information pertinent to the decision, the purpose and possible consequences of treatment and the consequences of not having treatment; and is able to use this information to consider whether or not they should consent to the intervention offered. This study demonstrates that the process of ‘informed consent’ around weight loss surgery is a complex one. In the eyes of the young people interviewed, it makes sense to have surgery, lose weight and be ‘normal’ because all other non-surgical options have failed. Individual change rather than social change appears obvious and the effects of weight bias and ubiquitous discrimination are not interrogated. This is despite growing evidence of strong genetic influences on adiposity, despite the obesogenic environment. 263 There are multiple consequences of not having weight loss surgery, not just in terms of health, but also in terms of having to face the difficulties of living in a social world that is less than tolerant of obesity. The stakes are high and the option not to have surgery is invisible. It is argued that clinicians working with young people deciding on weight loss surgery need to make these processes visible in supporting them to make their decisions.

Eligibility criteria

Studies were eligible if they fulfilled the following criteria:

• Subjects were children, adolescents or young adults aged up to 21 years undergoing bariatric surgery of any type.

• The study reported either:

  • psychological outcomes of bariatric surgery, including QoL

  • psychological or social predictors of outcomes of surgery. Outcomes were defined in the broadest sense, and included changes in weight and medical comorbidities as well as psychological outcomes.

• The study used either:

  • validated questionnaires before and after surgery, or questionnaires that compared postoperative with preoperative state or qualitative studies

  • a validated diagnostic system to detect prevalence of mental health disorders.

Search strategy