Treatment of anorexia nervosa: a multimethod investigation translating experimental neuroscience into clinical practice

Broad aims

Responding to the need for high-quality research into the management of AN, the overarching aims of the ARIADNE programme were to:

1. produce, validate and disseminate improved evidence-based interventions for AN

2. collaborate with patients and carers throughout the project

3. improve clinical outcomes in AN, by early detection and intervention, by reducing chronicity and relapse, and by improving carer outcomes

4. improve acceptability and cost-effectiveness of AN treatments

5. deliver standardised, trainable and disseminable AN treatments

6. assess service utilisation and NHS costs of AN and implications of changes in clinical practice for patient care and resources.

Objectives

Our objectives were: WP1a, to develop a training programme for school staff to enable them to detect and manage EDs; WP1b, to develop and test a school-based prevention programme for risk factors for EDs; WP2a, to develop an improved treatment for adults with AN that targets disease-maintaining factors, is matched to symptoms, personality and neuropsychological profile and which can be used as a first-line treatment in outpatient settings, and to evaluate the efficacy and cost-effectiveness of this treatment; WP2b, to test and validate components of this treatment, designed as intensive modules for inpatients with AN (i.e. those with severe, chronic or treatment-resistant AN); WP3, to evaluate the efficacy and cost-effectiveness of a carer skills training intervention, and to assess its impact on carer outcomes (e.g. distress, caregiving efficacy) and patient outcomes; WP4, to improve understanding of the nature of a debilitating core symptom of AN (i.e. hyperactivity); WP5, to develop and test a relapse prevention programme for inpatients with AN; WP6, to obtain information on the needs of mothers of children with an ED and the risks of the maternal ED for their offspring and to use this information to inform the development of an intervention for mothers with an ED to minimise the impact of their ED on their children; WP7a, to study existing care pathways for AN, with a focus on the impact of having access to specialist ED services; and WP7b, to study service utilisation and cost of illness in EDs.

Example 1: early intervention in schools

Wok package 1a involved the development and pilot testing of a teacher training programme for EDs (see Chapter 2). This research was devised and led by a former ED service user. It involved an extensive period of public consultation, which was used to assess the needs of both school staff and school students in this area. Thorough consultation was achieved through online surveys reaching over 800 school staff and over 500 students. Intervention materials for the training programme were then developed using an iterative process in which two panels of school staff (six members per panel) reviewed draft materials and their feedback was incorporated. PPI ensured that the training materials being developed were responding to a genuine need and were aligned with the needs of the school staff that would be using them.

Example 2: prevention

The intervention development in WP1b for the prevention programme (see Chapter 3) was informed by focus groups with 22 adolescent girls, who provided their experiences of body dissatisfaction, disordered eating and their recommendations for a preventative intervention. The intervention materials were then developed in conjunction with a panel of key stakeholders, which included a young person with a history of an ED, two young people without a history of an ED, and three secondary school teachers. Feedback provided by this panel was incorporated into the materials in an iterative process.

Example 3: carers interventions

Work package 3 evaluated an intervention for carers of people with AN, which was used as an adjunct to inpatient treatment (see Chapter 6). It included extensive PPI, with several members of the research team having personal experiences of an ED. The self-help materials (Experienced Carers Helping Others; ECHO) were developed in collaboration with patients and carers. 49 In addition, the majority of telephone coaching sessions offered during the intervention were provided by trained individuals with personal experience of an ED (either having recovered from the disorder themselves or as a carer). The findings from this trial are being disseminated to the public through a website dedicated to carers of those with an ED [URL: www.thenewmaudsleyapproach.co.uk (accessed 3 July 2017)], a newsletter, the database of ED volunteers and the annual carer’s conference which we run with Beat – the main national organisation for people with EDs and their families.

From the above, it is clear that patients and carers were involved in designing the programme, implementing the WPs and disseminating findings. Such collaboration between researchers and service user representatives has been highlighted as exemplifying good practice in service user involvement by the Mental Health Research Network (MHRN). 50

Work package 1: prevention and early intervention

Chapters 2 and 3 focus on prevention, detection and early intervention of EDs. In Chapter 2, we present the development and evaluation of a learning package for school staff on how to recognise symptoms of EDs, how to communicate about EDs sensitively and how to assess risk. Chapter 3 outlines the development of a teacher-delivered prevention programme for EDs, and the results of a cluster randomised controlled trial (RCT) evaluating its efficacy.

Work package 2: treatment

Chapters 4 and 5 focus on treatment of AN. In Chapter 4 we present the evaluation of the Maudsley Model of Anorexia Nervosa Treatment for Adults (MANTRA) using a large RCT with individuals in an outpatient setting. Chapter 5 explores the use of components of this treatment in an inpatient setting, designed as intensive modules for severe, chronic or treatment-resistant AN. Here, we evaluate this approach through case report, qualitative evaluation and a pilot trial comparing it to treatment as usual (TAU).

Work package 3: carers interventions

Chapter 6 presents an intervention for those caring for individuals with AN. We present findings from a large RCT assessing the impact of a guided self-help intervention for carers of individuals with AN (ECHO), in addition to standard inpatient care. We report on outcomes for both carers and for patients up to 12 months post discharge from inpatient care.

Work package 4: physical activity in anorexia nervosa

Chapter 7 focuses on the assessment of activity levels and endocrine changes in individuals with AN. We present data from an observational study. Individuals with AN (inpatients and outpatients) are compared with individuals with anxiety and healthy controls (HCs) using a range of methods, including body composition, endocrine measurements, self-report and actimetry.

Work package 5: relapse prevention

Chapter 8 focuses on relapse prevention. Here, we present a feasibility RCT on a novel e-mail-guided manual-based intervention to be used in the post-hospitalisation aftercare of patients with AN.

Work package 6: mothers with an eating disorder

In Chapter 9, we present research on mothers with an ED, a special population that may need tailored services. We report findings regarding fertility difficulties in women with an ED, and associations between maternal EDs and their children’s diet and growth trajectories.

Work package 7: care pathways and economic evaluations

Chapter 10 presents results regarding service use, focusing on how access to specialist services affects rates of referrals, admissions for inpatient treatment, continuity of care and service user experiences. Chapter 11 uses data from ARIADNE WPs (WPs 2, 3, 5 and 7a), plus the British Cohort Study (1970) (BCS-70), to identify the costs of services and treatments used by people with AN and to estimate its annual costs for England.

General discussion

Chapter 12 draws together the findings from the seven WPs and highlights clinical implications and recommendations for future research based on this programme.

Methods

First- and second-hand student experiences of suffering with an ED at school were explored using an online questionnaire (see Appendix 1, Student questionnaire), completed by 511 students aged 11–19 years [mean = 15.4 years, standard deviation (SD) = 2.3 years, 72% female]. Note that n varies by question as not all participants responded to all questions.

The questionnaire included free-text responses, which were coded using content analysis. 61,62 A categorisation system was developed by analysing responses and classifying them into categories ,with care being taken to ensure that the coding system was comprehensive, while avoiding overlapping of categories. A second researcher independently applied the categories, blind to the original researcher’s decisions. An inter-rater reliability of 94% was achieved.

Results

Discussion

This was the first UK study of student experiences of EDs. Respondents provided valuable insight into the experiences of UK school students with EDs and, further, provided actionable recommendations about improvements that could be made.

A strength of the study was that both male and female students were surveyed, which has not previously been the norm. 64 Future work could explore moderators (e.g. gender, school type) of student experiences.

The opportunity for students to recognise and respond to early ED symptoms in their friends, and for schools to provide a safe and supportive recovery environment, was clearly highlighted by respondents. However, it is clear that more work needs to be done if this potential is to be realised in UK schools. Psychoeducation and training for students and teachers was suggested by respondents as a key method for addressing this.

Methods

A total of 1250 UK school staff were invited to respond to an anonymous online survey exploring their experiences of EDs within the school setting (see Appendix 1, Staff questionnaire). A total of 826 (66%) of the convenience sample chose to participate.

Free-text responses were analysed using content analysis. 61,62 A categorisation system was developed by analysing responses and classifying them into categories, with care being taken to ensure that the coding system was comprehensive while avoiding overlapping of categories. A second researcher independently applied the categories, blind to the original researcher’s decisions. An inter-rater reliability of 89% was achieved (κ = 0.471; p < 0.001).

Results

Multiple-choice questions generated quantitative data, which have been summed and recorded in Table 3. Content analysis was used to interrogate the large amount of qualitative data produced in the form of free-text responses to open questions. This is summarised in Table 4.

Below, both forms of data are considered together under four salient themes which emerged during data analysis:

1. supporting students with EDs

2. ED training and policies

3. teaching about EDs

4. reintegrating students who were absent as the result of an ED.

Discussion

This was the first UK study of school experiences of EDs. Respondents provided valuable insight into the experiences of UK school staff with EDs and, further, provided actionable recommendations about improvements that could be made.

Methods

Participants for study 3 were recruited from those in study 2. Of 826 respondents in study 2, there were 109 expressions of interest in participating in further studies. Sixty-three of these went on to take part in the focus groups conducted in study 3.

A total of eight focus groups were held with 63 members of school staff from 29 schools in the UK. Topic guides were developed to explore school staff experience and opinion with relation to:

• school culture

• knowledge of school staff about EDs

• understanding of school staff about EDs

• communication with students

• support strategies

• working with parents

• working with external agencies.

The interviews were recorded and transcribed verbatim. Following initial familiarisation with the transcripts, a categorisation system was developed using content analysis. 61,62 Transcripts were analysed and classified into categories, with care being taken to ensure the categorisation system was comprehensive while avoiding overlapping of categories. The transcripts were independently catergorised by two researchers and inter-rater reliability was 87%, with a few minor discrepancies caused by one researcher applying more categories to some items than the other. There were no instances of the two researchers failing to agree on the primary category for a transcript section.

Results

Five salient themes emerged from the focus group discussions:

1. There was little general knowledge about EDs among staff.

2. Mental health issues, including EDs, were not openly talked about among staff.

3. School staff do not feel confident or comfortable teaching students about EDs.

4. When they exist, positive relationships with parents contribute to ED recovery, but sometimes relationships with parents are very negative.

5. More support is needed for school staff involved in the care of students undergoing ED recovery.

These themes are outlined briefly below. An extended discussion, including supporting quotes, is provided by Knightsmith et al. 65

Discussion

Focus group participants provided detailed, actionable insights into the current of confidence, understanding, knowledge and attititudes of UK school staff from a range of geographically and socioeconomically diverse schools.

Although the group reported themselves to be more than usually interested in and informed about EDs and other mental health issues, they drew widely on the knowledge, experience and attitudes of colleagues during the course of the focus groups.

Content outline

Based on school staff feedback about what is feasible in terms of in-service training days, the intervention was designed to be delivered in four 90-minute sessions, which could be delivered within the space of 1 day.

Each 90-minute session had clearly designed objectives and focus. These were:

• session 1 – EDs introduction

• session 2 – when and how to talk to students causing concern

• session 3 – working with parents, staff and students

• session 4 – providing a supportive environment during recovery.

Outcome measures

An outcome measure of school staff ED attitude, confidence and knowledge was developed for use during the current study as there was no existing tool. The new tool drew on existing measures of GP attitudes,67 and the feedback of school staff and students shared in studies 1–3.

A self-report style tool was developed which captured attitudes, confidence and knowledge about EDs. A copy of the self-report measure is included in Appendix 1, Eating disorders attitudes and knowledge questionnaire.

Methods

Forty-five members of UK school staff completed a 1-day face-to-face training programme designed to improve their knowledge about, attitudes towards and confidence in managing EDs. Participants completed self-report measures of knowledge, attitudes and confidence at the beginning of the day, before training commenced (T₁, baseline), at the end of the day, once training was completed (T₂, post intervention) and again 3 months later (T₃, follow-up).

The significance of intragroup changes was determined using generalised estimating equation models.

Results

The full results of the statistical analyses are shown in Tables 5 and 6. There was a statistically significant improvement (all p-values < 0.001) in participants’ self-reported knowledge, attitude and confidence scores post intervention (T₂) compared with baseline (T₁), with a large ES on all three comparisons. 68 These differences were maintained at the 3-month follow-up (T₃) and there was no significant difference between the knowledge, attitude or confidence scores measured post intervention (T₂) and at the 3-month follow-up (T₃).

Participants all completed a post-course evaluation form designed to assess the acceptability of the intervention. All delegates (n = 45) considered the course either good (16%) or very good (84%) in terms of course content, course materials and for providing practical strategies they could use at school.

Discussion

A 1-day training programme for UK school staff aimed at improving attitudes towards, confidence in supporting, and knowledge about EDs was tested for feasibility and acceptability and was found to have a positive, significant impact with medium and large ESs, which were maintained after 3 months. However, the study was limited by the use of outcome measures, which were entirely subjective/self-reported and in the lack of a longer-term measure of maintenance of positive outcomes.

Strengths

All four studies conducted as part of this WP were the first of their type within the UK setting. The contribution of male viewpoints is a significant strength of the current work as this has not previously been the norm. 64 A further strength of the studies was the large number of participants for studies of this type (511 student survey responses in study 1 and 826 school staff responses in study 2), and the depth of responses provided by these participants. A key strength of study 4 was the immediate and lasting impact of the intervention on participant attitudes, confidence and knowledge about EDs despite its brevity. A programme that took longer to deliver or the effects of which were not lasting would not be feasible or relevant for use in UK schools. Data collection at baseline, post intervention and 12 weeks post intervention is recommended, but not widely practised. 69,70

Limitations of the studies

It was not feasible to include both experimental and control groups within each school participating in study 4 owing to the possibility of trial arm contamination resulting from sharing of information between school staff, especially post intervention. In future studies, a stepped-wedge design could be employed to overcome this, with every participant completing both the control and intervention conditions.

The positive outcomes reported from study 4 in terms of an improvement in staff attitudes, confidence and knowledge were based entirely on self-report measures. Future studies could include some additional objective measures.

Study 4 looked only at school staff outcomes; the impact of the intervention was not extrapolated to explore the impact on student outcomes. Similar studies have similarly failed to provide such evidence,69,70 but appropriate measures should be considered for inclusion in future studies.

Future directions

The implementation of a fully powered stepped-wedge design in order to fully test the training programme developed in study 4 is a key future direction. Another possibility for exploration is the development of online training materials that could be accessed remotely in order to increase the reach and cost-effectiveness of the programme.

Conclusions

The aim of the current WP was to develop and feasibility test an ED training programme for school staff aimed at improving attitudes, confidence and knowledge.

This aim was achieved through a series of studies that drew extensively on the experiences and understanding of UK school staff and students in relation to EDs.

The work of all four studies was unique within the UK setting and the outcomes were promising and provide clear pathways for future research.

Objectives

To design and evaluate a universal prevention programme for EDs in secondary schools.

Trial design

Clustered RCT comparing intervention lessons with curriculum-as-usual control.

Participants

Students in years 8 and 9.

Intervention

Six-session Me, You & Us programme delivered by teachers.

Outcomes

Questionnaires at baseline, post intervention and at the 3-month follow-up, assessing body esteem (primary outcome), eating pathology, thin-ideal internalisation, appearance conversations, peer support, depressive symptoms and self-esteem.

Randomisation

Unrestricted randomisation of intact classes using random number generator.

Blinding

Teachers, students and researchers were not blinded to group assignment.

Numbers randomised

Sixteen classes were allocated to the intervention (nine classes, 261 students) or control group (seven classes, 187 students). No participants dropped out.

Results

There were significant between-group differences in body esteem favouring the intervention group at post intervention (d = 0.12) and 3-month follow-up (d = 0.19). There were also significant between-group differences in thin-ideal internalisation (d = 0.17, maintained at follow-up, d = 0.16) and self-esteem (d = 0.20, not maintained at follow-up). There were no between-group differences for the other outcomes. Fidelity to intervention material and acceptability of the programme varied across the three schools.

Harms

There was no evidence of reduction in body esteem.

Conclusions

The Me, You & Us programme improved body esteem, thin-ideal internalisation and self-esteem, but no other outcomes. Further work to increase efficacy across the range of outcomes and improve fidelity would be valuable.

Trial registration

Current Controlled Trials International Standard Randomised Controlled Trial Number (ISRCTN) 42594993.

Lessons 1 and 2: media literacy

The aim of the first two lessons was to help participants to critique media presentations of ideal beauty through exploring how conceptions of beauty have varied over time and place, what messages are hidden in media images and how to take action.

Lessons 3 and 4: fat talking

Lessons 3 and 4 focused on perceived peer pressure towards thinness, introducing the concept of fat talking, and examining why we might fat talk as well as possible consequences. The lessons went on to challenge negative appearance-related commentary through exploring the giving and receiving of compliments.

Lessons 5 and 6: personal strengths and well-being

The remaining lessons focused on tackling negative affect and low self-esteem, through learning about personal strengths and how to use them. Simple exercises to promote well-being were explored in the final lesson, including writing a gratitude letter and carrying out small acts of kindness.

Main hypothesis

1. Students receiving the intervention will show significant improvements in body esteem, internalisation, peer support, appearance conversations, depressive symptoms, self-esteem and eating pathology compared with students in the control group at post intervention and at a 3-month follow-up.

Subsidiary hypotheses

1. Students will find the material in the intervention acceptable, in that they will report enjoying the lessons and perceive them as useful.

2. It will be feasible to train secondary school teachers to deliver an ED prevention programme from a manual and student workbook with high fidelity.

The aims of the trial were to determine the acceptability, feasibility and efficacy of this universal prevention programme for EDs.

Trial design

The study used a cluster RCT design with intact classes of students allocated in a 1 : 1 ratio to intervention or control arms (trial registration: ISRCTN42594993, including protocol).

Participants

Settings and location of data collection

All data were collected within the school setting. School staff administered questionnaires within regular school hours, based on protocols provided by the researcher. Data collection took place between September 2011 and May 2012.

Intervention

The programme involved six 50-minute lessons which teachers delivered to existing classes. The intervention content was described in a Facilitators’ Manual and Student Workbook, both of which are available from the first author (HS). As discussed above, the material targeted risk factors for EDs and body dissatisfaction, namely thin-ideal internalisation, peer factors, depression and low self-esteem. 6,91 Participating teachers received a standardised 2-hour session, involving education about EDs and introduction to the Me, You & Us manual.

The control group received their usual curriculum. The content of these lessons was not determined by the research team.

Outcomes

Age, ethnicity and parental education were provided by participant self-report 1 week before the intervention began (‘pre-intervention’). Participants also completed an ED screening tool, the Eating Disorder Diagnostic Scale,92 which identifies Diagnostic and Statistical Manual of Mental Disorders-Fourth Edition (DSM-IV) diagnostic criteria for AN, bulimic nervosa and binge ED.

All primary and secondary outcome measures were administered at pre-intervention, 1 week following the intervention period (‘post intervention’) and at approximately 3 months following the intervention period (‘3-month follow-up’).

Sample size

To account for clustering, the sample size calculation was increased by an inflation factor [1 + (average cluster size – 1) intracluster correlation (ICC)]. The inflation factor for this trial was based on a small ICC (ICC = 0.05). 98 With an average class size of 28 students the estimated inflation factor was 2.35.

G*Power 3 (University of Düsseldorf) was used for sample size calculations. 99 Assuming a 1 : 1 ratio, a small ES (d = 0.20)100 and power set to 0.80, the basic sample size requirement was 394 participants per group, which increased to 926 per group when the estimated inflation factor was taken into account.

Randomisation and blinding

Intact classes were randomly allocated to intervention and control arms. As classes were enrolled into the trial, an unrestricted random allocation was generated by an online random number generator. 101 One researcher (HS) carried out the enrolment of classes, the generation of the random allocation sequence and the allocation of classes to conditions. Participants’ allocation in trial arm was based on their class membership. Informed consent for participants was obtained from all participants’ parents/carers following randomisation. In addition, participants provided written assent post randomisation, when completing the pre-intervention questionnaire measures.

The trial design precluded blinding of school staff or students participants as intervention materials, such as the student workbooks, would have been identifiable to staff and students as being distinct from their usual curriculum. Researchers were also unblinded.

Statistical analyses

Full details of statistical analyses, including managing of missing data, are reported by Sharpe et al. 71 All analyses were based on originally assigned groups. The main hypothesis was tested using linear and logistic mixed-effects models. Two continuous outcomes that were not normally distributed (depression, peer support) were dichotomised for all analyses.

In addition to significance testing, ESs (d) for continuous outcomes were calculated using the differences in adjusted means at each time point. Reliable and clinically significant changes were also computed102 using reliability data and clinical cut-off points from previous work. 103

Participant flow and characteristics

Three schools agreed to participate in the trial. One additional school refused participation because it did not want to trial previously untested resources. Each of the schools was state maintained and had 100% female intake. The three schools varied in their average level of deprivation (free school meal eligibility ranged from 2% to 24%), and in the ethnic background of their students (the percentage of students from black and ethnic minority backgrounds ranged from 28% to 77%). Random allocation of 16 intact classes from year 8 or 9 in these schools resulted in nine classes allocated to the intervention arm and seven classes allocated to the control arm. Of the 479 students from these classes, 31 were excluded because of lack of parental consent. This resulted in 261 students in the intervention arm and 187 students in the control arm. Between 92% and 98% of students provided data at each data collection point. All missing data were due to school absence. No participants withdrew from the trial.

There were no significant differences between the two trial arms at the beginning of the trial. Participants had a mean age of 13.06 (SD = 0.59) years in the intervention group and 12.99 (SD = 0.54) years in the control group. Equal proportions of participants came from ethnic minority backgrounds (intervention = 47%, control = 53%), and had parents with university-level education (intervention = 76%, control = 78%). There was also no difference on reports of body esteem (intervention: mean = 2.30, SD = 0.75; control: mean = 2.27, SD = 0.70), or across any of the other clinical outcomes for the trial. Eight participants scored above the cut-off point in the ED screening tool and so were excluded from analyses on the grounds that the aim of the programme was prevention of future difficulties.

Acceptability and feasibility

Students in the intervention arm were asked to rate how useful and enjoyable they found the lessons in the intervention. The results are shown in Table 7. Looking across both ratings of lessons being enjoyable and useful, the acceptability are notably higher in two schools (A and C) compared with the third school (B). Whereas few students (3–16%) in schools A and C rated the programme negatively, this figure was nearer 50% for school B.

These findings mirror the results of the feasibility assessment, in which the teachers’ fidelity to the intervention manual was assessed. A greater amount of intervention content was delivered in schools A and C than in school B (78% activities rated as ‘completed’ in schools A and C compared with 50% in school B). School B did not deliver material outside the programme, but instead took longer to complete each activity, meaning that fewer activities were completed within the set lesson time.

Body esteem

Results from the mixed-effects models showed an overall marginal effect of the intervention on improvements in body esteem [β = 0.09, standard error (SE) = 0.05; p = 0.08]. Post hoc analyses at each time point showed a marginal difference between the groups post intervention (intervention: mean = 2.31, SE = 0.32; control: mean = 22.2, SE = 0.39; p = 0.07) and a significant difference between the groups at the 3-month follow-up (intervention: mean = 2.37, SE = 0.32; control: mean = 2.22, SE = 0.39; p = 0.006). In each case the results favoured the intervention over the control condition. The ESs for these group differences were small (d = 0.12–0.19).

Reliable and clinically significant change from baseline to post intervention was calculated separately for those above and below the clinical cut-off point at baseline. As this was a community sample, few participants were above the clinical cut-off point at baseline (n = 60). Of those above the cut-off point initially, 50% (n = 17) of participants in the intervention group showed clinically significant improvement, compared with 38% (n = 10) in the control group. However, this difference was not statistically significant [χ²(1) = 0.79; p = 0.37]. When considering reliable change, 32% (n = 11) of participants in the intervention group showed reliable improvement, compared with 8% (n = 2) in the control group. This difference was statistically significant [χ²(1) = 5.97; p = 0.02].

Considering those participants that began the trial in the normal range for body esteem, it was notable that there were no participants that experienced clinically significant worsening of symptoms. Some participants did, however, show reliable improvements in body esteem. In the intervention group, 12% participants (n = 20) experienced reliable improvement compared with 4% in the control group (n = 4). This difference was statistically significant [χ²(1) = 6.58; p = 0.01].

Secondary outcomes

A linear mixed-effects model for continuous secondary outcomes showed that there was a main effect of group for thin-ideal internalisation (β = –1.53, SE = 0.74; p = 0.04) and self-esteem (β = 0.19, SE = 0.09; p = 0.04). There was no significant main effect for appearance conversations (β = –0.04, SE = 0.32; p = 0.90). For thin-ideal internalisation, looking separately at each time point showed significant differences between the groups both post intervention (intervention: mean = 21.94, SE = 0.47; control: mean = 23.47, SE = 0.57; p = 0.04) and at the 3-month follow-up period [χ²(1) = 3.84; p = 0.05]. At each time point the results favoured the intervention. For self-esteem, the comparisons between the groups at each of the time points revealed a significant difference post intervention (intervention: mean = 3.52, SE = 0.06; control: mean = 3.33, SE = 0.07; p = 0.04), but no difference by the 3-month follow-up (intervention: mean = 3.47, SE = 0.06; control: mean = 3.34, SE = 0.07; p = 0.15).

Logistic mixed-effects models were used to explore intervention effects for binary outcomes. There were no main effects of group for binge eating [odds ratio (OR) = 4.44, 95% CI 0.39 to 51.22; p = 0.23], compensatory behaviours [OR = 1.69, 95% CI 0.74 to 3.89; p = 0.22], peer support [OR = 1.40, 95% CI 0.64 to 3.06; p = 0.40] or depressive symptoms [OR = 1.49, 95% CI 0.46 to 4.78; p = 0.50]. Rates of each of these outcomes at each time point are reported in detail elsewhere. 71

Strengths and limitations

A number of factors limit the conclusions of this trial. First, there was a risk of control group contamination, due to random allocation of classes within the same school. Second, we did not use an active control condition. Potentials for sham interventions that have been used in previous similar trials include activities such as expressive writing,107 educational brochures108 and healthy eating programmes. 109 Third, the sample size recruited was below that required for desired statistical power. Post hoc estimates of achieved power using the ES for the primary outcome at the 3-month follow-up (d = 0.19) show that the trial only had 49% power to detect group differences. Fourth, the assessment of intervention fidelity was coarse (two lesson observations per school). Ideally, the trial would have involved recording all sessions and having these rated by independent researchers. Finally, the follow-up period in this trial was limited to 3 months. Valuable information about whether or not effects are maintained in the medium to long term would be gained from longer follow-up.

The generalisability of the results was bolstered by the fact that the schools involved represented a wide range of different participants, including many from ethnic minority backgrounds as well as a range of economic backgrounds. It is also of significance that the trial was conducted in state-funded schools, and was not delivered by specialist school staff. However, it should be pointed out that the reliance on girls’ schools does limit the generalisability of these findings, as typical state-maintained schools in the UK are coeducational. Replication of these findings in coeducational schools is a priority going forwards with this work.

Conclusions

Overall, this study suggests that it is possible to design a teacher-delivered ED-prevention programme that is efficacious, manageable for teachers and liked by students. The between-school differences in acceptability and fidelity suggest that further work is needed to increase the suitability of the materials across a range of school settings and also to examine best practice around training teachers for taking on the delivery of this programme. Efficacy was not found across all outcomes and so there was also scope for improvements to the materials for which intended impacts were not observed. In addition, further trials will be essential to replicate these findings and to ensure that materials provided for schools are a safe and effective use of school resources in tackling EDs.

Objectives

To assess the efficacy of a new psychological treatment for AN (MANTRA) versus specialist supportive clinical management (SSCM) in a RCT.

Trial design

Multicentre two-arm superiority trial.

Setting

Four specialist ED services within the south of England.

Participants

Adult outpatients, aged between 18 and 65 years, meeting DSM-IV criteria for AN or an eating disorder not otherwise specified – anorexia nervosa subtype (EDNOS-AN), with a body mass index (BMI) of < 18.5 kg/m².

Intervention

Twenty once-weekly treatment sessions of MANTRA or SSCM (extended to 30 sessions in those with a BMI of < 15 kg/m²) and optional extras (carers’ sessions, dietetic sessions).

Outcomes

Randomisation

Completed independently from trial team using a restricted stratified randomisation algorithm with (1) BMI of > or ≤ 15 kg/m², (2) AN subtype and (3) previous ED admission, as stratifiers.

Blinding

Research assessors were blind, but patients and therapists were not blinded to treatment allocations.

Numbers randomised

Seventy-two patients allocated to MANTRA and 70 allocated to SSCM.

Results

There was no difference in outcome between groups. In both treatments, patients showed significant improvements in BMI, ED symptoms and other clinical outcomes. Patients rated MANTRA as more acceptable and credible than SSCM at 12 months.

Harms

One SSCM patient died during treatment.

Conclusions

Both MANTRA and SSCM can be used as outpatient treatments in adults with AN.

Trial registration

Current Controlled Trials ISRCTN67720902.

Aims

• The central aim of this work is to test the efficacy and cost-effectiveness of MANTRA in adult outpatients with AN versus SSCM using a randomised controlled design. (Only efficacy data are reported here; cost-effectiveness data are not yet available and will be reported elsewhere when complete.)

• The subsidiary aim is to assess the impact of the experimental treatment on the requirement for any intensive (i.e. day care or inpatient) treatment and to identify mediators and moderators of treatment outcome. (These results are not currently available and will be published elsewhere when complete.)

Hypotheses

• MANTRA will be superior to SSCM in producing higher weight increase and greater reduction in ED symptoms in adults with AN at 6 and 12 months.

• MANTRA will be more cost-effective than SSCM, being less costly at 6 and 12 months. Specifically, MANTRA patients will need fewer and briefer hospitalisations during the study period compared with SSCM. (Only the hospitalisation aspect of this hypothesis is addressed here; the full cost-effectiveness data are not yet available and will be published elsewhere when complete.)

Trial design

This is a superiority trial conducted in four NHS specialist ED units. The trial evaluated the efficacy and cost-effectiveness of MANTRA versus SSCM in consecutively referred adult outpatients with AN. Patients were allocated to one of two treatment arms: MANTRA or SSCM. Further details on the randomisation procedure are given below.

Outcomes were measured pre-randomisation, at 6 months post randomisation (i.e. designed to broadly coincide with the end of weekly treatment) and at 12 months post randomisation (follow-up). Outcomes were assessed by researchers who were not involved in the treatment and efforts were made to keep researchers blind to patients’ treatment allocation. As much as possible, patients who had terminated treatment early were followed up to enable intention-to-treat (ITT) analysis.

Ethics approval

We obtained ethics approval for the MOSAIC trial from Central London Research Ethics Committee (REC) 4, National Research Ethics Service, Royal Free Hospital, London (NHS REC Reference: 10/H0714/9). All participants provided informed written consent prior to entering the study. The study was conducted in compliance with the Helsinki Declaration.

Participants

Locations of data collection

Patients were recruited from several centres: South London and Maudsley NHS Foundation Trust; North East London Foundation Trust Eating Disorders Service; Barnet, Enfield & Haringey Mental Health NHS Trust; Oxford Health NHS Foundation Trust.

Interventions

Therapists

Twenty-eight ED therapists delivered the trial therapies across the four centres. Therapists were trained in MANTRA and SSCM (2 days were provided as an introduction to each treatment). Additionally, further training days were offered periodically to ensure uniform delivery of treatments across centres and to protect against ‘therapeutic drift’. Therapists had to deliver both treatments to minimise allegiance effects. Supervision was provided to therapists on a weekly basis by designated trained supervisors in each team. To avoid cross-therapy contamination, this was delivered separately for MANTRA and SSCM. Patients were allocated to therapists based on availability. To ensure that treatments were delivered competently and as planned, audio-recordings of sessions were carried out and three tapes per patient were chosen randomly to assess treatment adherence in both groups.

For each of their patients, therapists recorded details, including number, duration and content of sessions, and details of any additional treatment.

Outcomes

All outcome measures were collected at baseline, 6 months after randomisation (i.e. approximately end of weekly sessions) and 12 months after randomisation (i.e. follow-up). Treatment credibility and acceptability ratings were taken only at 6 and 12 months. Potential outcome mediator variables were also taken at 3 months (mid-treatment). These results are not yet available and will be published elsewhere when complete.

Sample size

Full details of our sample size calculation, including the data and assumptions on which this is based, can be found in our protocol paper. 128 In brief, we assumed that we would need a sample size of 69 patients per group to have 90% power to detect a difference in mean weight gain of 2.5 kg, using an independent-samples t-test with a significance level of α = 0.05 and correcting for 20% attrition.

Randomisation and implementation

The randomisation sequence was generated and implemented independently from the study team by the King’s College London Clinical Trials Unit (CTU). After participant recruitment and baseline assessment, the research assessor entered patient identification and stratification details into the online CTU system. Participants were then allocated to one of the two treatments based on a restricted stratified randomisation algorithm. The strata were (1) BMI below or above 15 kg/m², (2) restricting or binge/purge subtype of AN and (3) previous hospitalisation in an ED unit. These factors are known to affect treatment outcome and rates of potential future inpatient treatment. The stratification was implemented by minimised randomisation with a random component. The first n cases (n would not be disclosed) were allocated entirely at random to further enhance allocation concealment.

Blinding

Although patients and therapists were not blinded, the research assessors were blind to treatment allocation. In order to assess whether or not researcher blinding was successful, they had to make a guess at the end of 12 months’ assessment as to which treatment they believed the person had received.

Statistical methods

The objective of the statistical analyses was the comparison of those offered MANTRA with those offered SSCM on a number of outcomes. The primary clinical outcome was BMI (kg/m²) at 12 months after randomisation. Secondary outcomes were continuous measures at 6 or 12 months’ follow-up (see above). All statistical analyses were based on the ITT principle; thus, patients were analysed in the treatment arm to which they were randomised irrespective of whether or not they received the allocated treatment. All analyses were carried out in Stata^® 12 (StataCorp LP, College Station, TX, USA).

All outcomes were analysed using linear mixed models. Full details of the analysis plan can be found elsewhere. 110,128 Outcome variables contained considerable numbers of missing values. Full details of the treatment of missing variables can be found elsewhere. 110

Participant flow and recruitment

Flow of participants through the trial is shown in Figure 1. Of the 319 patients approached, 142 agreed to participate in the trial. Following baseline assessments, random allocation of the 142 participants across all four study sites resulted in 72 patients in the MANTRA arm and 70 in the SSCM arm. Three patients did not receive their allocated therapy (see Figure 1).

FIGURE 1.

Consolidated Standards of Reporting Trials flow diagram (WP2a). Copyright © 2015 by the American Psychological Association (APA). Reproduced with permission. The official citation that should be used in referencing this material is Schmidt U, Magill N, Renwick B, Keyes A, Kenyon M, Dejong H, et al. (2015). The Maudsley Outpatient Study of Treatments for Anorexia Nervosa and Related Conditions (MOSAIC): Comparison of the Maudsley Model of Anorexia Nervosa Treatment for Adults (MANTRA) with specialist supportive clinical management (SSCM) in outpatients with broadly defined anorexia nervosa: a randomized controlled trial. Journal of Consulting and Clinical Psychology, 83(4), 796–807. 110 The use of APA information does not imply endorsement by APA.

Treatment completion was defined at the start of the study as attending ≥ 15 therapy sessions (i.e. receiving more than three-quarters of the weekly treatment sessions). This definition is in line with other studies’ definitions of treatment adherence. 127,131 Overall, 66.9% of participants were treatment completers, with 75% of MANTRA participants and 59% of SSCM participants completing treatment [χ²(1) = 1.19; p = 0.28].

Non-completion of treatment predicted loss to follow-up at 12 months. At the 12-month follow-up, primary outcome data were missing for 45.2% of non-completing participants, compared with 10.5% of completers. Details of the reasons for missing data can be found in Figure 1.

All 142 participants were included in the primary outcome analysis.

The trial was conducted between June 2010 and November 2013. Between June 2010 and November 2012, participants were recruited into the trial via their initial clinical assessment with the EDs service, which was followed by a pre-intervention assessment. Delivery of the intervention occurred between June 2010 and November 2013. End of weekly treatment session follow-up assessments (month 6) were completed during December 2010 to May 2013 and end of treatment follow-up assessments (month 12) occurred throughout June 2011 to November 2013.

Service utilisation

Use of additional dietitian sessions was not different between groups [MANTRA 33/72 (46%); SSCM 31/70 (44%)]. Sixty-seven of 72 (93%) MANTRA patients and 64 of 70 (91.4%) SSCM patients provided information on additional service utilisation (outside the study protocol) during the study period. Eighteen of 131 patients for whom this information was available (13.7%) had additional treatment during the study period. This was categorised into ED inpatient or day care treatment, general psychiatric inpatient treatment or other treatment. Eight of these patients had been allocated to MANTRA. Two of these had ED inpatient treatment (217 and 66 days, respectively), one of these also had 30 days of ED day care treatment and 1 day of ‘other treatment’. Three further patients only had ED day care (175 days, 144 days and 1 day, respectively). One of these also had a 10-day inpatient alcohol detoxification programme. Two further patients had brief general psychiatric admissions (in one case 7 days and in the other 10 days), because of suicidal or self-harm behaviour; one of these also had 1 day of ‘other treatment’. One further patient had 1 day of ‘other’ treatment only.

Ten patients who needed additional treatment had been allocated to SSCM (15.6%). Five of these had ED inpatient care (of 45, 62, 110, 140 and 198 days), with one of these having a second 8-day ED admission. Three patients had ED day care (for 2, 54, 183 days); one of these also had a 50-day general psychiatric admission following depression and suicidal behaviour. One further patient had a 1-day general psychiatry admission followed by 11 days of home treatment, and one patient had 1 day of ‘other treatment’. Proportions of people needing additional treatment were not different between MANTRA and SSCM. There was also no significant difference between the two treatment arms in the number of days per admission or admission days per patient.

Baseline data

The baseline characteristics of patients in both groups were comparable (Table 8) Overall, the mean baseline BMI was 16.6 (SD 1.2) kg/m² and the duration of illness was 8.3 (SD 7.3) years. Sixty-three (44.4%) patients presented with restrictive AN.

Primary outcome

Tables 9 and 10 show that the two treatments did not differ in their effect on BMI at 6 months (p > 0.05) or at 12 months (p > 0.05) (primary outcome).

Tables 11 and 12 demonstrate, however, that there was a significant overall effect of receiving treatment, with BMI in the total sample increasing from baseline to month 6 by 0.74 [95% confidence interval (CI) 0.4 to 1.08] and from baseline to month 12 by 1.19 (95% CI 0.59 to 1.79).

Secondary outcomes

Tables 9 and 10 show that the two treatments did not differ in their effect on the EDE global score at either the 6- or 12-month time point (all p > 0.05).

Tables 11 and 12, however, demonstrate that there was a significant overall effect of receiving treatment, with mean EDE global score decreasing from baseline to month 6 by 0.63 (95% CI –0.85 to –0.42) and from baseline to month 12 by 0.84 (95% CI –1.08 to 10.59).

Tables 9 and 10 show that there was no difference between the two treatments at 6 or 12 months’ follow-up for all other secondary outcomes (all p > 0.05).

Again, Tables 11 and 12 demonstrate that there was a significant overall effect of receiving treatment for all EDE subscales, DASS-21, CIA, WCST and Brixton Spatial Anticipation Test scores after month 6 (all p < 0.01), whereas OCI-R, REY and RMF did not change significantly (all p > 0.05).

At month 12 there was a significant overall effect of treatment for all EDE subscales, DASS-21, CIA, Brixton Spatial Anticipation Test and REY scores (all p < 0.01), whereas OCI-R, WCST and RMF did not change significantly (all p > 0.05).

Recovery rates

At 12 months’ follow-up, participants were split into three groups: recovered, partially recovered and not recovered. These were defined as (1) recovered, BMI of > 18.5 kg/m² and EDE global score of < 2.77; (b) partially recovered, BMI of ≤ 17.5 kg/m² and EDE global score of < 2.77, BMI of > 17.5 kg/m² and ≤ 18.5 kg/m² or BMI of > 18.5 kg/m² and EDE global score of > 2.77; and (c) not recovered, BMI of ≤ 17.5 kg/m² and EDE global score of > 2.77.

The proportions of participants falling into these categories at 12 months were as follows:

• MANTRA: recovered, 13 of 72 (18.1%); partially recovered, 36 out of 72 (50%); and not recovered, 9 out of 72 (12.5%);

• SSCM: recovered 8 out of 70 (11.4%); partially recovered, 32 out of 70 (45.7%); and not recovered, 9 out of 70 (12.9%).

There was no association between treatment allocation and recovery [χ²(3) = 2.80; p = 0.42].

Baseline body mass index as a potential moderator of primary and secondary outcomes

We also examined whether or not baseline BMI was a potential moderator of treatment effect on BMI at 6 and 12 months, dividing participants into those with a baseline BMI of < 17.5 kg/m² (77% of MANTRA and 70% of SSCM participants) and those with a BMI of > 17.5 kg/m² (18.1% of MANTRA and 25.7% of SSCM participants). (This information was missing for 4.2% and 4.3% of participants in MANTRA and SSCM groups, respectively.)

There was no evidence of a statistically significant interaction between treatment group and whether baseline BMI was above or below 17.5 kg/m² when BMI was the outcome. However, on visual inspection of the estimated effects, it does appear that the effect of treatment (favouring MANTRA) was larger among those with lower baseline BMI (Table 13).

Baseline BMI was also investigated as a potential moderator of secondary outcomes. No evidence of a statistically significant interaction between treatment group and whether baseline BMI was more or less than 17.5 kg/m² was found for these other outcomes.

Harms

One SSCM patient died. No other harms were noted.

Treatment acceptability and credibility

There were no significant differences in acceptability and credibility ratings between MANTRA and SSCM at 6 months {acceptability: MANTRA = 8.5 [2.0], SSCM = 8.0 [2.2] [t(100) = 1.33; p = 0.18]; credibility: MANTRA = 6.4 [3.1], SSCM = 5.8 [2.7] [t(100) = 1.1; p = 0.29]}. However, at 12 months, MANTRA was given significantly higher ratings on both acceptability and credibility than SSCM {acceptability: MANTRA = 8.6 [1.8], SSCM = 7.8 [2.3] [t(91) = 2.01; p < 0.05]; credibility: MANTRA = 6.8 [3.1], SSCM = 5.5 [2.7] [t(91) = 2.24; p < 0.05]}.

Main findings

First, overall, participants showed significant improvements in terms of BMI, ED symptoms, depression, anxiety and stress and psychosocial impairment after 6 months of weekly treatment sessions. These improvements were either maintained or increased at 12 months post randomisation. Second, there were no differential effects between the two treatments in terms of primary and secondary outcomes and further service utilisation. Therefore, our main hypotheses were not confirmed. However, BMI, EDE and proportions of recovered participants all somewhat favour MANTRA at 12 months, and exploratory moderator analysis suggested that the treatment effect in relation to BMI (favouring MANTRA) increased (non-significantly) at 12 months, especially in those patients who were more severely ill at baseline.

There was a mixed picture of results in terms of neurocognitive and social cognitive performance. There was no effect of treatment on social cognition; however, participants were not particularly impaired on this task at baseline,155 so this lack of improvement is unsurprising. Central coherence and cognitive flexibility as measured by the Brixton Spatial Anticipation Test improved significantly by month 12; however, cognitive flexibility as measured by the WCST did not, but again participants were not particularly impaired at baseline. 15

Comparison of main findings with those of other trials

The degree of BMI change observed in the MOSAIC trial is comparable to that found in our earlier single-centre pilot RCT (total n = 72), which used identical study inclusion criteria and which also compared MANTRA and SSCM,136 suggesting that these treatments can be disseminated to other UK centres with relatively brief initial training (2 days) of therapists and booster training (1 day).

Comparison against other trials needs to be approached cautiously because of differences in patient populations. Perhaps the findings from the present study can be compared most easily against the large German psychotherapy trial by Zipfel et al. ,131 which compared focal psychodynamic therapy, CBT and optimised TAU in outpatients with AN. Patients in this trial had comparable baseline BMIs (between 16.6 and 16.8 kg/m²), but received a somewhat higher ‘dose’ of psychotherapy (40 sessions over 10 months). Patients were followed up at 13 months post randomisation and at 22 months post randomisation. Thirteen-month BMI outcomes in the Zipfel et al. 131 study were between 17.6 kg/m² (focal psychodynamic therapy) and 17.4 kg/m² (CBT and optimised TAU), whereas our 12-month BMI outcomes were 18.0 kg/m² in MANTRA and 17.7 kg/m² in SSCM. Thus, our BMI improvements stand up well in this international comparison.

Two other previous trials have used SSCM. 127,156 Comparison of our data against these trials is difficult. One of these trials was carried out in patients with a significantly milder form of AN (baseline mean BMI of 17.3 kg/m²),127 and the other specifically focused on chronic patients (mean illness duration of 15.5 and 17.7 years in the CBT and SSCM group, respectively). 156 In the McIntosh study, SSCM BMI outcomes were superior to CBT and IPT BMI outcomes at end of treatment, but not at long-term follow-up. 157 In the Touyz et al. 156 study, at 12-month follow-up, BMI gains were modest (0.5 kg/m²) and identical for both groups.

Other findings

Harms

One SSCM patient died during the study. This patient had very severe and chronic AN. It has been documented in many studies that chronically ill, low-weight AN patients such as the one in our study do have a much increased mortality risk. 72 The causes of death typically are related directly to consequences or complications of the illness (via malnutrition or abnormal weight control methods), and this is thought to be the case here. Previous trials of AN outpatient treatment in adults have reported deaths. For example, in a trial by Dare et al. ,33 one patient (out of 84) died over the course of the trial, and in the study by McIntosh et al. 127 there was one death among 56 participants.

Patient and therapist experience of Maudsley Model of Anorexia Nervosa Treatment for Adults and specialist supportive clinical management

The present trial is so far the only RCT of treatments for AN to include qualitative feedback from patients and therapists, which we published separately prior to the outcome evaluation, as is recommended for such process evaluations. 158–160 In particular, process data from therapists, who of course were able to compare and contrast these treatments, identified marked differences between these two treatments, in terms of the therapy focus, therapeutic strategies and applicability to a wide range of patients. 159 Process data from patients largely agree with these findings, and in addition strongly point to the quality of the therapeutic relationship, as the basis for a successful therapy. 158 Both treatments were perceived as having strengths and weaknesses. A further process evaluation examined written qualitative feedback from all study participants at 12 months. 160 Eighty-two study participants provided such written feedback, significantly more from the MANTRA group than from the SSCM group. MANTRA patients also tended to write in more detail and give more positive feedback than individuals receiving SSCM. Taken together with quantitative acceptability and credibility ratings, these process evaluation findings suggest that patients prefer MANTRA and therapists like it too. It thus provides a good alternative to SSCM. Of note, the greater acceptability of MANTRA may mean that patients are more willing to have further treatment if required or have better outcomes in the longer term. Assessment of these long-term effects of MANTRA and SSCM is currently under way.

Strengths

The strengths of the present study include that it is the largest RCT of first-line psychological treatments for adults outpatients with AN to be carried within the UK and the second largest in the world. 131 The study had good participation rates at follow-up data points, good treatment completion, and high acceptability and good outcomes compared with other studies. As such, the MOSAIC trial improves our limited evidence base regarding outpatient treatments in this hard-to-treat population. This study is also novel in that it is the first AN trial to incorporate both qualitative and quantitative measures of the therapeutic process and therefore offers a greater insight into patient and therapist perceptions of the two treatments.

Limitations

The MOSAIC trial also had limitations. First, as the study compared two active treatments and did not include a no-treatment control group, improvements in clinical outcomes cannot be attributed definitively to treatment effects. Second, although the trial is the largest ever to be carried out within the UK, it is still not large enough to perform conclusive mediation and moderation analyses of the core illness maintenance factors, which MANTRA attempts to target. However, we will be able to amalgamate the data from the present trial with the data from our earlier smaller RCT,136 which used an identical design and overlapping assessments, to allow more definitive moderator and mediator analyses. The combined sample size from both these trials will be 214 patients.

Third, although the ‘lost to follow-up’ rate within our overall sample was acceptable, the rate for participants assigned to SSCM is 10% greater than for those receiving MANTRA. Therefore, information on recovery in individuals in this group is more limited.

Finally, we were restricted by funding to a relatively short-term follow-up of 12 months. Further (2-year) follow-up is necessary to measure the longer-term effects of these two treatments on primary and secondary outcomes.

Generalisability

The MOSAIC trial included four specialist ED services and the interventions were delivered by a large number and broad range of therapeutic staff, with different professional backgrounds and different degrees of experience, so the study can be seen to reflect usual clinical practice. The trial also had very few exclusion criteria, making the sample as generalisable as possible and leading to the inclusion of a sample with a very broad range of clinical severity and BMI. As in usual clinical practice, trial therapists were able to conduct both treatments. Through therapist interviews159 we have been able to confirm that both treatments were experienced very differently and delivered with fidelity. Further confirmation of this will be reported separately through analysis of the therapeutic session recordings.

Conclusions and future outlook

The significant improvement of patients in both groups suggests that both of these two very different treatments have value as first-line outpatient treatments for adults with AN. In the original trial from New Zealand in which SSCM was first used,127 SSCM was found to outperform two active treatments (CBT and IPT) against the investigators’ hypothesis, and therefore provides a very high standard of control treatment. Comparable weight gain and comparable reductions in ED psychopathology between MANTRA and SSCM are therefore encouraging.

The current report includes an ITT analysis only. In future, we will also conduct a more complete analysis and an exploratory analysis of potential treatment mediators. As part of this WP we are also in the process of completing a cost-effectiveness analysis based on the current trial. The results of these analyses will be published as soon as they are available. Finally, we will also conduct analyses of session recordings to assess treatment fidelity and therapeutic process. We have completed a 2-year follow-up of the MOSAIC patients, showing that improvements across clinical outcomes are maintained or increase further. 161

Objectives

To evaluate the benefits of Cognitive Remediation and Emotional Skills Training (CREST) for patients with AN receiving inpatient treatment.

Design

This project/chapter has three studies: (1) a case report, (2) a qualitative study of service users’ views about CREST and (3) a quasi-experimental comparison of CREST and TAU in two inpatient settings.

Participants

Adult inpatients with AN.

Interventions

The main novel intervention used in this project is CREST, which consists of simple cognitive exercises and simple emotion regulation, recognition and processing skills training, promoting micro skills in social communication and self-regulation as well as facilitating better self-awareness.

Main outcome measures

Outcomes included qualitative feedback and neuropsychological tests targeting flexibility of thinking, central coherence and theory of mind. Clinical outcomes (BMI, EDE-Q) were also reported.

Results

Qualitative assessments demonstrated CREST to be acceptable and perceived as beneficial by patients. Additionally, self-report questionnaires demonstrated improvements. However, quantitative data were less promising, showing no clear differences between the CREST and TAU groups in terms of cognitive or clinical outcomes.

Conclusions

Future work will focus on revisions to the CREST manual based on the outcomes of this work, as well as greater refinement of the measures used in subsequent quantitative evaluations.

Setting and referral

Anna is a 21-year-old Caucasian British woman informally admitted to the ED unit with AN and self-harming behaviour.

Assessment

On admission, Anna had a BMI of 13.1 kg/m² (the healthy range is between 19 and 25 kg/m²). She fulfilled diagnostic criteria for the restricting subtype of AN. She was significantly restricting her diet and exercising excessively prior to her admission, which led to significant weight loss. She also described a number of rigid rules and ritualistic behaviours surrounding her eating, and would become distressed if her routines were disrupted. Anna had a distorted body image and body dissatisfaction, and described herself as ‘disgusting’ and ‘fat’, despite being objectively emaciated. She seemed to lack a sense of personal identity, but felt that AN gave her an identity and a sense of specialness. She rated the importance of change as 2/10 and her perceived ability to change as 2/10.

Prior to her admission Anna was frequently engaging in self-harm, including punching, scratching, and burning herself (superficially) on her arms and hands. She described using these behaviours as a means of managing adverse emotional states. She did not have any current suicidal ideation and denied plans to take her own life.

Anna reported growing up in an emotionally invalidating environment, one in which her feelings and thoughts were often marginalised by her parents, and she felt that she was ‘second best’ to her younger brother.

Key triggers for Anna’s eating disorder appeared to be her failing her General Certificate of Secondary Education (GCSE) exams and being bullied by her peers, who told her she was not ‘good enough’ for her boyfriend. These events seem to activate and reinforce negative core beliefs that Anna had about herself. AN may have provided Anna with a sense of achievement and specialness at a time when she was feeling like a failure in other domains. AN and her subsequent self-harm also gave her a way of managing her emotions, which she found confusing and aversive.

Intervention

Anna was offered CREST over 10 individual sessions. This was deemed to be a particularly useful treatment option for Anna, given her difficulties in managing her emotions. Anna’s mother also attended a series of family meetings with Anna, in which they were supported to gain insight into each other’s perspectives. These sessions also helped in supporting Anna’s mother in her own struggle to support Anna.

Outcomes

Anna did not complete the outcome measures given pre and post CREST. However, over the course of her 14-week admission her weight did increase such that her BMI reached 14.7 kg/m². Her self-reported rating of the importance of change rose from 2/10 to 5/10, and her perceived ability to change also rose from 2/10 to 5/10. Anna reported that she found the intervention (CREST and family work) very useful in helping her to reduce her self-harming and in managing her emotions, as well as allowing for a change in the relationship with her mum to develop. Anna’s mother also said that she found the CREST work very helpful and felt more confident in her ability to support Anna in her recovery.

Aims

The aim of this study was to explore patients’ views and experience of individual CREST.

Methods

Discussion

The main aim of the qualitative study was to evaluate the acceptability of the CREST intervention for inpatients with AN. In the national inpatient ward where this study took place, the majority of the patients are nutritionally compromised (according to our annual audit data the mean BMI at admission is 13.0 kg/m²). In a severely ill AN patient, treatment engagement and meaningful psychological intervention delivery is challenging. Research shows that acutely ill AN patients have a major difficulty in correctly identifying and expressing emotions. 16,163,170,171 CREST includes information about basic emotion processing skills and aims to give patients simple tools and strategies to notice their own emotions and express their needs in a safe way. Alongside enhancing these skills, CREST also had a benefit in terms of increasing assertiveness. Many of the patients who received this intervention reported that CREST was helpful and that the strategies they learnt will be useful in the future. This highlights an improvement in patients’ confidence and perceived ability to express thoughts and emotions. From the qualitative outcome analysis we could also conclude that CREST has a positive impact on reducing interpersonal problems known to be present in this population. 19,172

Using qualitative outcomes from the study and learning from patients’ personal experiences were very useful steps in terms of gathering evidence for CREST.

Aims

The aims of this study were to use cognitive (neuropsychological) assessments to (1) investigate changes in performance on cognitive flexibility, central coherence and emotion processing tasks in inpatients with AN following CREST and (2) investigate the magnitude of the changes on a cognitive and emotion test battery used in the study with patients who have received CREST plus TAU (referred to from now on as the CREST group), compared with patients who have received TAU only (the TAU group).

Hypotheses

1. In both groups we hypothesised weight gain, since both inpatient programmes focus on nutritional stabilisation.

2. In terms of cognitive and emotion task-related performance, our hypothesis was that improvement after treatment would be greater in the CREST group than in the TAU group.

Method

Ethics approval was obtained from the Oxfordshire REC NHS Ethics Committee (reference 08/H0606/58). Study participants were recruited from specialist inpatient ED units receiving national referrals. Patients receiving CREST were recruited from the South London and Maudsley NHS Foundation Trust inpatient ward and patients receiving TAU were recruited from Manchester (Cheadle Royal Hospital).

Participants

Consecutively referred adult patients were offered participation in the trial if they had a DSM-IV166 diagnosis. Patients were excluded from the study if they had an additional diagnosis of a learning disability, psychosis or major physical illness. Each participant was given a detailed information sheet about the study, and signed consent forms were collected from each study participant.

Interventions

Participants in the London inpatient ward received CREST as detailed above, in addition to TAU. Participants in the TAU group received standard inpatient care with individual and group therapy. At both sites TAU involved typical inpatient care: nutrition management; dietetic input; occupational therapy; family work; individual cognitive behavioural therapy; and group work on self-esteem and body image. In addition, TAU in London included individual work on psychological assessment and formulation. In Manchester, TAU additionally included individual cognitive analytic therapy and schema therapy.

Assessments

Participants in both groups were assessed 10 days after admission to the inpatient wards (within the first 2 weeks, T₁). Participants were reassessed 10–11 weeks later [i.e. at the end of treatment (T₂)].

Statistical analysis

All statistical analyses were performed in Stata 11. A linear mixed model was used to compare the two treatments. BMI and the six neuropsychological tasks were the outcome measures. Severity of illness (measured by BMI), duration of illness (years since diagnosis was established), intelligence quotient (IQ; measured by National Adult Reading Test total score) and age were included as covariates.

Results

Discussion

This study aimed to assess the efficacy of the individual format of CREST for inpatients with AN. Both groups showed significant improvements in BMI, as expected. Our hypothesis, that in within-group analyses neuropsychological measures would not change over time in the TAU group and would improve in the CREST group, was only partially supported by the results. Although the TAU group showed no change over time on central coherence (more ‘bigger picture’ thinking, less attention to detail) or emotion processing outcomes (less bias towards angry faces or recognition of emotions from the RME task), set-shifting task performance did improve in this group (this supports the idea that cognitive flexibility improves with weight restoration). The CREST group also showed set-shifting improvements and additional improvements in global processing (this could be explained by the fact that CREST has introductory sessions on cognitive styles addressing global thinking style strategies). The results of the study showed that the TAU group became more detail focused over time, whereas the CREST group remained stable in their information processing style (central coherence). There were no changes in emotion processing experimental outcomes in the CREST group (e.g. less bias towards angry faces and better recognition of emotions). In the between-group analyses there were no significant differences between the CREST and TAU groups on any of the neuropsychological domains or weight gain.

This study does not support previous findings suggesting that TAU does not improve set-shifting,180 as the TAU group improved in set-shifting in this study. This suggests that more longitudinal studies are required to assess the effects of TAU on set-shifting across different inpatient groups.

It is possible that differences between the groups explain the improvements in cognitive flexibility following TAU and the non-significant differences between the groups on all measures. Although the groups in this study were comparable on sociodemographic and clinical characteristics, there was a trend in the CREST group towards a lower lowest lifetime BMI and a lower highest ever BMI than in the TAU group. Moreover, although the two groups were recruited from specialist inpatient services, the TAU group received intensive individual psychological input from a variety of treatment models as part of their TAU package (e.g. cognitive analytical therapy and schema therapy).

Background

Families express a need for information and help to support people with severe and/or enduring AN. We have developed a guided self-help, skills training intervention for carers (ECHO).

Objectives

To examine the impact of the addition of ECHO to standard inpatient care on carer and patient outcomes over time.

Method/design

Patients (aged > 12 years) with a primary diagnosis of AN, and their carers, were ascertained from 15 inpatient services in the UK. Patients were randomised either to receive ECHO [a book, digital versatile discs (DVDs) and 10 telephone coaching sessions per family] or TAU. Patient (n = 178) and carer (n = 268) outcomes were measured at discharge and at 6 and 12 months after discharge.

Results

Carers in the ECHO group spent less time caregiving and had less burden and expressed emotion at discharge and/or 6 months. At 6 months, patients experienced a decrease in ED symptoms and improved quality of life.

Discussion

Skills-based training improves both patient and carer outcomes.

Primary hypotheses

Secondary hypotheses

Design

This is a pragmatic, two-arm, multicentre, parallel-group RCT. A detailed account of the trial protocol, including the description of the ECHO carer intervention, has been published. 204 Consenting carers from dyads meeting the inclusion criteria were randomly assigned to either ECHO (plus TAU) or TAU alone. Guidance for the ECHO intervention was delivered by ‘experienced’ coaches [people with lived experience of EDs (n = 15) and postgraduate psychologists without clinical training (n = 5)], who were specifically trained and supervised. Recruitment involved 15 inpatient ED units within the NHS. Baseline and discharge measures were collected, along with follow-up data at 6 and 12 months post discharge. After ceasing intensive treatment, patients were considered ‘discharged’. To meet this criteria, they had to receive inpatient, day patient, or specialised residential treatment on ≥ 4 days per week, and maintain this status for a minimum of 4 weeks. Those who transferred to general supported housing were included too. Data regarding whether patients completed treatment or discharged themselves were not accessible for all sites and are not reported.

Ethics and governance

Main ethics approval was granted by the Royal Free Hospital Ethics Committee (08/H0720/41) with site-specific ethics and governance approval for all participating sites. The study was adopted by the MHRN.

Randomisation

Carers of a patient were randomly allocated to one of the two trial arms (ECHO or TAU). The CTU at King’s College London conducted the randomisation and minimisation was used, stratified by study site (15 NHS England hospital units) and disease severity categories (one or both of BMI < 15 kg/m², presence of compensatory vomiting). Those in the TAU arm could access the intervention after finishing the study.

Recruitment

Participating sites

This project involves 15 different centres. Fourteen of the sites are ED specialist inpatient wards in the UK (13 adult, one adolescent), and one is a general psychiatric ward with specialist ED staff (adolescent). One of the specialist sites makes local referrals to specialist ED inpatient wards in the UK. All were NHS units, except one, which is a private collaboration accepting NHS referrals.

Participants

Patients admitted to the treatment facility with a primary diagnosis of AN or an EDNOS with anorexic symptoms were approached. The agreement of at least one carer (proposed by the patient) was required for participation in the study. {Carers are defined as someone who provides unpaid help and support to a child, partner, relative, friend, or neighbour, who could not manage without their help [URL: www.carers.org (accessed 3 July 2017)].}

Recruitment

Patients were approached after admission by participating staff on the wards and offered study details. Informed consent was from both patient and carer. Clinical studies officers (CSOs) from the MHRN supported recruitment of patients and administration of the project on eligible sites.

Interventions

Intervention delivery

All correspondence with carers on the randomisation outcome was by post. Families in the ECHO group received an intervention pack with the book, five DVDs and supporting documentation (a letter explaining their contact with the telephone coach, action/goal sheets, details on the transtheoretical model of change, frequently asked questions and details for technical support). Those receiving TAU were informed they could access ECHO after completing the study, and were given contact details for Beat, the leading ED charity in the UK.

Data collection

All participants (patients and carers) completed self-report assessments by post at inpatient admission, discharge, and at 6- and 12-month time points post discharge. Finally, for the year following discharge, patients completed a short monthly assessment on core eating symptoms by telephone, e-mail or post. Detailed information about data management is described in the published protocol. 204

Measures

Those completed by the patient are marked with a (P) and those completed by carers are marked with (C).

Reimbursement

Participants were reimbursed on completion of each set of questionnaires (patients received £25, carers £10).

Statistical analysis

The objective of the statistical analyses was the comparison of those offered ECHO with those allocated to TAU on a number of primary and secondary outcomes. The primary patient clinical outcome is ‘relapse’. A linear change was assumed between monthly measurements of BMI and this was used to extrapolate the day on which two points were estimated to be lost. The variable is right censored since a relapse might not have occurred by the end of the study period (1 year) or follow-up data for BMI could be missing. If five consecutive BMI measurements were missed, time to relapse was censored.

The primary carer clinical outcome is distress (measured by the DASS-21) at 12 months after discharge. To account for two primary outcomes, group differences on these outcomes were tested at a significance level of 2.5%. Secondary patient and carer outcomes are continuous measures at discharge, 6- or 12-month follow-up (see Secondary outcomes). All statistical analyses were based on the ITT principle, that is, participants were analysed in the treatment arm to which they were randomised irrespective of whether or not they received the allocated treatment. Additionally, the statisticians were kept blind to treatment allocation as long as possible.

Time to relapse was analysed using Cox regression. Explanatory variables in this model were the variable of interest (treatment arm) and randomisation stratifiers (site and illness severity categories, based on binary BMI and vomiting). The effect of treatment is estimated by the hazard ratio of relapse comparing ECHO with TAU.

The continuous secondary patient outcomes were analysed using linear mixed models. The dependent variable is the outcome at the relevant time point (e.g. BMI at 12 months post discharge) and (fixed) explanatory variables are given by treatment arm, baseline values of the variable under investigation (e.g. BMI at pre-randomisation) and randomisation stratifiers. The models can contain random intercepts for coaches (in ECHO) to allow for correlation in outcomes due to treatment being facilitated by the same coach. The models were used to estimate differences between treatment arms at each time point. Standardised treatment effect estimates were calculated by dividing group differences by the common pre-randomisation SD of that outcome.

Outcome variables contained considerable numbers of missing values (see Results for details). We empirically identified a number of baseline variables that were predictive of missing values in outcome and also found that the primary carer not adhering to ECHO (coded ‘1’ = completed at least five coaching sessions or read at least half the coaching manual, ‘0’ = did not complete the intervention) was predictive of loss-to-follow up (see Results for details). To allow for these processes driving missingness, in addition to allowing randomised group and values of the outcome under investigation at different time points being predictive of missingness multiple imputation (MI) using chained equations as implemented in the Stata command ice. 215 This allowed us to include predictors of missingness (including the post-randomisation variable adherence) in the imputation step without having to condition on these variables in the analyses models. 216

The continuous carer outcomes were analysed in a similar fashion except that analyses needed to include up to two carers per family (a nominated first and second carer). To deal with this, the analysis and imputation models described for patients were extended for carer outcomes. First, the analysis models contained additional random intercepts that varied at the level of the patient to allow for similar outcomes for carers of the same patient. Second, to ensure that correlations were also reflected, the imputed values imputations in ice were carried out at the level of the patient allowing for two outcome variables – one for the primary carer and another for secondary carer (set to missing when there was no second carer, with resulting imputed values discarded before analysis).

To evaluate the effect on primary carer outcome of receiving a sufficient level of intervention (here an adherence score of ‘1’) without bias, the complier-average causal ECHO effect (CACE) on carer distress was estimated using a two-stage least squares (instrumental variables) estimator. 217

Participant flow

This study represents a unique collaboration of major UK ED treatment centres and describes a large group of severely ill patients with AN (178 patients and 268 carers). The Consolidated Standards of Reporting Trials diagram for the study is shown in Figure 2.

FIGURE 2.

Consolidated Standards of Reporting Trials flow diagram (WP3).

Sample characteristics

Randomisation

Patient and carer outcomes were well balanced across groups. Roughly 25% at discharge, 30% at 6 months after discharge and slightly over one-third of patients at 12 months showed missing questionnaire outcome data. At 12 months, the proportion ranged between 0.33 and 0.37 for the different scales. For the carers, the proportion of missing data was 2–4 percentage points greater at all outcome time points. The proportion of missing outcome data at 12 months ranged between 0.32 and 0.43. Logistic regression was used to explore the relationships between a dependent variable that represented whether outcome data were present or missing at 12 months after discharge, and a number of baseline demographic and clinical variables, such as participant gender and lowest ever BMI. Any variable that showed a statistically significant association with the dependent variable was included in the imputation step of the MI procedure. Non-adherence with ECHO was strongly associated with missingness, with ORs from 1.5 to 2.0 for patients’ data and from 3.3 to 6.2 for carers’ data.

A comparison of inpatient treatment post discharge of the patient groups (carers with Experienced Carers Helping Others and carers with treatment as usual)

The ECHO group had shorter admission length (median = 148 days, range 28–991 days) than the TAU group (median = 163 days, range 33–570 days; Mann–Whitney U-test, z = –0.88, p = 0.38). The readmission rate was 26.7% (n = 23) in the ECHO group and 32% (n = 29) in the TAU group. Relapse in terms of readmission and/or fall in 2 BMI points occurred in 43% of the ECHO group and 52% of the TAU group. The median time to relapse for the ECHO group was 262 days and 240 days for TAU. Survival plots of the time to relapse showed two similar curves (Figure 3). The TAU showed greater relapse-free survival initially. However, after approximately 8 months, the curves crossed over. Thus, the study did not detect a statistically significant difference in survival curves and the main assumption of a Cox regression model, that the hazards are proportional, was broken (a Cox regression which relies on proportional hazards would give a hazard ratio of 0.82 in favour of ECHO and a p-value of 0.42 for the group difference).

FIGURE 3.

Kaplan–Meier curves for time to relapse.

Table 16 summarises the clinical outcomes and Table 17 provides estimated outcome differences between the two treatment arms for both patients and carers at all three time points.

A comparison of carers’ well-being and burden in the Experienced Carers Helping Others and treatment as usual condition

In all of the outcomes measured, the carers in the ECHO group had improved functioning. Carers in the ECHO group reported less distress than those in the TAU group, but this difference was not statistically significant at any time point.

Time spent care giving was reduced 6 months post discharge and was significantly lower in the ECHO group (p = 0.05; ES = –0.34). Expressed emotion was also significantly lower in the ECHO group at 6 months (p = 0.05; ES = –0.25). Burden was significantly lower in the ECHO group at discharge (p = 0.04; ES = –0.29).

At discharge, we estimated that carers’ perception of ED burden in the ECHO group was 3.98 points (95% CI 0.16 to 7.80 points) lower than in the TAU group. At 6 months after discharge, time spent caregiving in the ECHO group was 63% (95% CI 40% to 100%) of that spent in the TAU group. At the same time point, expressed emotion in the ECHO group was estimated as 2.24 points (95% CI 0.05 to 4.43 points) less on the FQ than in the TAU group.

A comparison of the clinical status post discharge of the patient groups (carers with Experienced Carers Helping Others and carers with treatment as usual)

Eating disorder psychopathology and quality of life were significantly better in the ECHO group at 6 months (ES = –0.38 and 0.38, respectively), but these differences were not significant at 12 months post discharge. At 6 months after discharge, we found ED psychopathology among the ECHO group to be 0.47 points (95% CI 0.03 to 0.92 points) less than in the TAU group. At the same time point, we estimated quality of life as measured on the WHOQOL-100 to be 0.91 points (95% CI 0.04 to 1.78 points) lower in the ECHO group compared with the TAU group. Estimated differences in distress and BMI pointed towards a beneficial effect of ECHO, but none of these effects could be shown to be statistically significant.

Important harms or unintended effects

Two patients (one from each trial arm) died during the course of the study.

Semistructured open feedback

Limitations

This sample is not necessarily representative of the target population as it includes only people who agree for their carers to be involved and who have carers who are willing to help. Moreover, the study includes patients at various stages of illness (19% early and 55% enduring), and it is possible that these subsets have different responses to the interventions. For example, family therapy is effective only as an early intervention. 194 Over half of the patients in this sample (n = 83; 55%) fulfil the criteria for being in a severe and/or enduring stage of illness. Such patients are considered to be resistant to all forms of treatment. 221,222 The sample was powered to detect only moderate overall differences in patient outcome. With 134 carers in each group and an alpha level of 0.025 (adjusting for multiple primary outcomes), we had 80% power to detect an ES of 0.38.

Although there are standard quality criteria for inpatient care, the treatment ethos does vary between services. For example, some services involve carers more than others do. This may have decreased the size of the effect. However, as the sample was stratified by site, it would not have led to any bias in the interpretation of the results. The comparison with TAU (rather than an active control), does mean that we cannot rule out that the apparent benefits of ECHO are non-specific effects of participants receiving a greater intensity of intervention.

Strengths

This study used the strengths of a pragmatic randomised controlled design to examine the effects of adding a low-intensity psychoeducational and skills training intervention for carers to inpatient care for AN. Furthermore, this is a multicentre study in which the majority of NHS England specialised ED centres participated. Considering that the number of patients with AN requiring inpatient care is small, the sample size of this study is unique.

Clinical implications

The possible added benefit for patients and for services attained from giving carers psychoeducational information has been rarely studied, but this study suggests that this is an area that merits more attention. It is possible that increasing the intensity of the intervention and including it within the service model of inpatient care might produce improvements for services, patients and carers. This would mean that carers might become more involved in treatment for adults.

Conclusion

A low-intensity psychoeducational intervention for the carers of patients at a severe and/or enduring stage of AN produced a decrease in burden and time spent caregiving and a small decrease in distress. Unhelpful carer behaviours such as high expressed emotion were also decreased. Carers reported higher levels of functioning in the people with EDs. Moreover, patients had greater improvement in eating psychopathology and quality of life and noted positive changes in carers’ behaviours. Patients felt more empathy for their carer. Moreover, carers’ dissatisfaction with services was lower in the ECHO group. It is possible that increasing the intensity of work with carers may have benefits for clinical services for patients in addition to benefits for carers themselves.

Objectives

This WP examined physical activity (PA) in people with AN and associations between drive to exercise, ED psychopathology, anxiety, endocrine measures and energy expenditure (EE).

Method

Four groups of women were recruited: AN outpatients (n = 37), AN inpatients (n = 18), those with anxiety symptoms (n = 34) and HCs (n = 30). Actigraphy and self-report were used to measure PA, together with the drive and reasons for exercise, ED and general psychopathology, body composition and BMI. Salivary cortisol was also measured and EE was estimated. Measures were made over 24 weeks.

Results

Objective PA levels did not differ significantly between groups cross-sectionally or longitudinally: nonetheless, AN groups reported higher total PA than HCs. AN outpatients reported more walking and moderate PA than HCs. AN inpatients reported more walking but less moderate and vigorous PA than other groups. In the two AN groups, drive to exercise was significantly higher. These groups also rated ‘improving tone’ as being an important motivator to exercise. Estimates of EE due to PA indicated a maximum of 400 kcal/day in patients with AN, out of a total EE of approximately 1600 kcal/day.

Discussion

Elevated PA has been reported in AN, but this may reflect patients’ peception of their activity rather than what is actually done. Drive to exercise in AN seems to be more closely linked to ED psychopathology than to anxiety. Estimates of EE due to PA provide guidance for therapists who want to introduce exercise into therapy.

Participants

Women were recruited as four groups: inpatients (n = 18) and outpatients (n = 37) with a diagnosis of AN, individuals experiencing moderate anxiety (n = 34) and HCs (n = 30). AN patients (binge-purging or restricting subtype) with a BMI < 17.5 kg/m² were recruited from in and around London (from specialist ED services). These services included outpatient treatment (n = 26), day care (n = 10) and inpatient treatment (n = 18). Patients attending day care treatment were grouped with the outpatient participants. Anxiety and control participants were recruited via an e-mail circular to students and staff at King’s College London. Anxiety participants were screened using the Generalised Anxiety Disorder questionnaire, and those with a score of ≥ 10 were recruited into the study. Participants had a mean age of 27 years in the anxiety group (range 18–54 years), 29 years in the combined AN groups (range 18–67 years) and 29 years in the HCs (range 20–52 years). All were English speaking. Recruitment commenced following approval by the local REC (09/H0807/4). All participants provided informed consent.

Measures

All participants provided data on the following measures.

Data analysis

Five participants from the HCs with a BMI < 18.5 kg/m² and six from the anxiety group with EDE-Q global scores > 2.8270 were not used in the analyses. Group means were compared using one-way analyses of variance (ANOVAs) and Tukey post hoc tests. Kruskal–Wallis tests and post hoc Mann–Whitney tests using Bonferroni corrections were used to compare means for variables that violated parametric assumptions [EDE-Q, DASS-21, International Physical Activity Questionnaire (IPAQ) and body fat]. Pearson’s rank correlations were used to infer relationships between variables before hierarchical multiple regressions were performed, with drive to exercise as the dependent variable.

Group differences in outcome measures

Associations between outcomes

Longitudinal studies related to physical activity

We also investigated PA longitudinally over 24 weeks in patients with AN. The four groups of participants were the same and the methodology was similar to that used in the cross-sectional study. PA was measured continuously for 8 weeks and at follow-up (weeks 12 and 24) by actimetry and self-report (IPAQ). Measures of general and ED-related psychopathology were made in parallel.

Eating disorder psychopathology and physiological measures remained consistent with diagnosis. Drive to exercise, depression, stress and anxiety were, as might have been predicted, consistently higher in the two AN groups than in the HCs and the group with anxiety symptoms. The highest values were seen in the inpatient AN group. In all four groups, there was a wide range in activity, especially on the self-report (IPAC) scores, but mean actimetry values were relatively constant over time. In the AN groups (compared with HCs and the group with anxiety symptoms), self-reported activity was higher on measures of total, walking and moderate PA. In the two AN groups, the drive to exercise was correlated with self-reported exercise. The objective measures (actimetry) showed no statistically significant difference in levels of PA between the four groups.

The data arising from the longitudinal study are in agreement with the results from the cross-sectional investigation. For example, comparison of the objective and the self-report data suggests that, in the AN groups, self-reported overactivity is perceived rather than real and may arise from factors such as drive to exercise. It is also of note that levels of activity remain relatively constant across the duration of the study. Furthermore, the objective data suggest that levels of PA do not differ between the AN groups and the HCs. What is perhaps surprising is that levels of activity are not apparently decreased by being in an inpatient setting, for which limits on activity may have been set. This may be because people in an inpatient setting spend considerable time walking (e.g. pacing). It suggests that levels of PA are difficult to regulate in severely ill patients in whom there is a high drive to exercise. However, given that some of the patients have a very low BMI and high levels of ED pathology, diminished fat reserves and a loss of muscle tissue (i.e. the AN inpatient group), this relatively ‘normal’ level of activity may be of some clinical concern and should be considered. In this respect, it should also be seen in the context of setting a balance that recognises the beneficial effects of exercise on mood and anxiety, and its effects on bone health, EE and PA.

Energy expenditure and physical activity

A final element of the study was an investigation of EE asociated with PA in AN. The data were derived from the measures of self-reported activity and from physiological measures such as BMI. PA-related EE was estimated by converting IPAQ scores to ‘metabolic equivalent of task’ values and then to corrected metabolic equivalent of task values, to adjust for differences in resting metabolic rate (RMR).

Compared with the HC group, RMR was lower in the outpatient group (by 11%) and in the inpatient group (by 17%). When EE is corrected for RMR (corrected metabolic equivalent of task), no group differences in EE emerged. We estimate that EE (i.e. EE due to PA) is roughly 360 kcal/day in the outpatient group and 400 kcal/day in the inpatient group. Based on the use of the self-report (IPAQ) data, combining RMR with EE provides energy neutral points of 1620 kcal/day for the outpatient AN group and 1578 kcal/day for the inpatient AN group. As the two patient groups report that they undertake higher levels of exercise than that suggested by the objective actimetry scores, the actual total daily EE in these two clinical groups is likely to be lower than our estimate (e.g. by roughly 200 kcal/day).

Energy expenditure may be of concern when a patient is at low weight/not eating. However, given that they have a high drive to exercise and that PA may improve affect, reduce stress and possibly improve general health, it is arguable that awareness of the caloric consequences of exercising will enable the desire to exercise to be more easily negotiated in relation to food intake.

Summary

The data on psychopathology and on body composition were as expected (i.e. patient groups had significantly higher ED pathology, anxiety, depression and stress and lower BMI, fat and muscle than the HCs and the anxiety groups). This information was fairly consistent both cross-sectionally and longitudinally.

The PA data are less easy to interpret. Self-reported PA (IPAQ scores) showed that the two AN groups reported higher levels of activity across all levels of PA (especially walking) than the anxious and HC groups. These two patient groups reported slightly different exercise profiles: AN outpatients reported undertaking more moderate and vigorous activity than the two control groups and the AN inpatients reported the opposite (i.e. that they engaged in less moderate and vigorous activity than HCs). In contrast, actimetry showed that average and peak PA levels in the AN groups were quite similar to those reported by the HCs and anxious participants.

Both the groups with AN had significantly higher drives to exercise than the HCs and the anxious group, and the groups valued different reasons for exercising as more or less important. For example, the AN groups rated weight control, improving mood and improving tone as more important motivators for engaging in exercise. In contrast, the anxiety and HC groups rated health as the most important exercise motivator.

Physical activity

Physical activity measured by actimetry (both cross-sectionally and longitudinally) shows that the AN groups spent similar amounts of time exercising, and at similar intensities, as HCs, whereas, on the self-report measure, the outpatient AN group have higher PA scores at all intensity levels. The self-report data are therefore consistent with other findings indicating increased PA in AN,232,235,236 and also with our first hypothesis. The objective data do not support our original hypothesis, but do support the findings of Zipfel et al. 238 that individuals with AN and controls showed similar levels of PA when objective measures are used (e.g. the doubly labelled water method). Zipfel et al. also reported that not all individuals with AN have increased levels of PA. However, in the present study, although we saw a range in activity in the AN group, there was no relationship between illness severity and PA and, last, no evidence of activity-based subgroups emerged.

The data from the self-report (IPAQ) measure (both cross-sectional and longitudinal) suggest that there are group differences in the exercise type that is undertaken (e.g. inpatients reported spending significantly more time walking). This suggests that these individuals choose types of exercise that are appropriate for their weight and physical ability in order to satisfy their high drive for exercise. It may also be that it is the most feasible method in an inpatient setting.

Several potential explanations can be provided for the differences between subjective and objective measures of PA. For example, objective measures may not accurately detect different intensities or types of PA. This could occur when arm movements may not be consistent with the intensity of activity (e.g. uphill walking or cycling). There may also be some limitations to self-report measures in terms of recall accuracy. Additionally, it is possible that different groups estimate their levels of PA differently. HC participants may underestimate this whereas, conversely, individuals with AN may perceive themselves to be highly active. Furthermore, AN patients who are physically compromised may perceive PA at any level as a significant physical investment, and this may result in them overestimating their activity. This idea is consistent with the observed large SD and variance in our IPAQ data. Overall, the results suggest that individuals when they are ill (and possibly in an inpatient setting) are undertaking PA at roughly the same level as HCs and this in itself is of some clinical concern.

Drive to exercise and physical activity

Consistent with our second hypothesis, both the AN inpatient and outpatient groups had significantly higher drives to exercise than the HC and the anxiety groups. Second, drive to exercise was not different between the HC and the anxiety groups. It is therefore suggested that this drive to exercise is linked more to ED than to anxiety. The absence of a strong relationship between anxiety and the drive to exercise or objective/subjective measures of PA further supports this. Across the whole sample, there was a moderate correlation between drive to exercise and all measures of self-reported PA (i.e. with total activity, walking, moderate activity and vigorous activity). There were also moderate correlations between drive to exercise and global ED score, exercising to to improve mood, improve tone, for fitness and for weight control. Last, drive to exercise was inversely correlated with BMI and poorly with the DASS-21 anxiety score.

In the AN inpatient and outpatient groups, there was a significant relationship between drive to exercise and exercising to control weight, to improve mood, to improve tone, for fitness and health, and peak activity (based on actimetry). In the group with anxiety symptoms, drive to exercise was highly correlated with self-reported vigorous activity. Finally, drive to exercise was moderately correlated with exercising to improve mood in this group.

Hierarchical multiple regression analysis showed that the global EDE-Q score, exercising to control weight and to improve mood, predicted drive to exercise across the whole sample after controlling for muscle (%) and for BMI. In the anxiety group alone, stress (DASS-21 subscale) contributed most to the variance in drive to exercise. For the AN outpatient group, the biggest contributors to variance in drive to exercise were exercising to improve mood and global EDE-Q score.

Overall, results show that PA has several psychological drivers, as well as exercising to improve fitness and tone. Engaging in PA seems to be driven by the desire to cope with stress and to improve mood in the anxiety group. However, in people with AN, the drive may partly be to improve mood, but, in this group, the absence of a relationship between PA and anxiety scores suggests that drive to exercise is not primarily a response to anxiety. It appears that these behaviours are associated with body and weight concerns (i.e. central elements of ED psychopathology). Data thus support our hypotheses and previous findings242 that the drive to exercise in AN is substantially motivated by concerns at the core of ED psychopathology (e.g. the cognitive issues related to losing/maintaining low weight).

Endocrine findings

Anorexia nervosa inpatients have significantly higher cortisol levels than HCs, which is to be expected given the higher level of anxiety in this group. A small correlation was observed between cortisol level and DASS-21 anxiety score. Both AN groups had significantly lower leptin levels than HCs, and there was a strong and highly significant correlation between leptin and body fat percentage (both these findings are consistent with expectations).

Strengths and limitations

A strength of this investigation is that it allowed for PA to be directly compared in people with AN and with moderate anxiety. Second, participants in each group were of similar ages, which is important when comparing PA. Additionally, combining both objective and subjective measures allowed us to assess and compare different methodologies for studying PA. An additional advantage was that measures of ED psychopathology, general psychopathology and biology were collected along with the activity data. Finally, the study was conducted cross-sectionally and longitudinally (over 24 weeks).

The investigation had some limitations. First, the more complex analyses used in this study (e.g. group differences in predictors of reasons for exercise) are probably underpowered. In addition, some data were missing because participants occasionally forgot to wear the Actiwatches and also because self-report data were collected retrospectively over a period of 7 days, making such data slightly less reliable. In addition, there is the possibility of some recruitment bias in the AN groups (i.e. not all of the people with the problem might be willing to join an investigation assessing PA). Finally, participants were predominantly white females and, therefore, it might not be possible to generalise our data to men and/or to different ethnic groups.

Conclusions

Our activity measurements gathered from self-report data indicate that levels of PA are higher in invividuals with AN than in HCs or those with moderate anxiety, whereas the actimetry data show that these groups do not differ in their amount or level of PA. It is possible that self-report overestimates the level of PA in AN patients owing to differences in perception caused by the ED or because these individuals are in poor physical health. Findings also show that patients with AN have a higher drive to exercise than HC and anxiety groups. In AN patients, ED pathology, rather than anxiety, appears to be a more important contributor to their increased drive to exercise. Estimates of RMR and of EE arising from PA will inform decisions related to the clinical management of people with AN and will also inform discussions on the role and use of exercise programmes in treatment.

Background

Relapse rates after inpatient treatment of AN are high. The aims of this study were (1) to assess the feasibility of a relapse prevention programme for people with AN and (2) to acquire information to inform a large RCT of this intervention.

Methods

Participants were inpatients with AN aged ≥ 16 years. Participants were recruited from seven UK specialist ED units and randomly allocated to receive either manual-based e-mail-guided self-care for 12 months combined with TAU or TAU alone. The manual was based on the MANTRA. E-mail support was delivered weekly by therapists. Outcome assessments included BMI, EDE, depression, anxiety, quality of life and service utilisation at 6 and 12 months post randomisation. Treatment ESs (Cohen’s d) were calculated.

Results

Forty-one patients participated. At 6 months post randomisation (and post discharge) there was little difference between groups, with ESs between d = –0.08 and d = 0.3. At 12 months, patients receiving the experimental intervention had a higher BMI (d = 0.41) and lower scores on the DASS-21 (d = 0.64). Readmission rates were 5 out of 22 (22.7%) in the experimental group and 5 out of 16 (31.2%) in the TAU alone group. No harms were detected.

Conclusions

These findings suggest that this low-intensity relapse prevention may be beneficial in the aftercare of inpatients with AN and that a large-scale RCT of this intervention may be justified.

Trial registration

Current Controlled Trials ISRCTN18274621.

Design

The trial was a multicentre two-arm feasibility trial that compared iMANTRA added to TAU with TAU alone in consecutive referrals of patients discharged from inpatient treatment for AN. Ethics approval was obtained from the National REC West Midlands – Edgbaston, reference number: 08/H1208/33. The trial was registered with Current Controlled Trials ISRCTN18274621. Recruitment and follow-up procedures took place between March 2009 and December 2012.

Participants and recruitment

Patients were recruited from seven specialist adult or adolescent inpatient units in the UK. These were South London and Maudsley NHS Foundation Trust; Seacroft Eating Disorders Unit, Leeds Teaching Hospitals NHS Trust; St Ann’s Hospital, Barnet, Enfield and Haringey Mental Health Trust; South West London and St George’s Mental Health NHS Trust; Huntercombe Hospitals (Stafford, Edinburgh); The Priory Hospital, Roehampton; and Royal Victoria Hospital, Newcastle upon Tyne. Patients’ suitability for participation in the study was checked by a clinician from the EDs team towards the end of their inpatient treatment (i.e. within 1 month of discharge). If the patient was suitable for participation, their fully informed written consent was sought and after this she/he was introduced to the researcher who completed the research assessment.

Inclusion/exclusion criteria

Comparison groups

Measures

Randomisation, blinding and protection against bias

Randomisation was conducted independently from the trial team by the King’s College London CTU using a computerised system. Patients were randomised to the iMANTRA intervention plus TAU versus TAU alone, at a ratio of 1 : 1. Randomisation was carried out using minimisation to balance groups for prognostic factors (previous hospitalisation; illness subtype). Patients were told about the outcome of randomisation by an e-mail sent to them by a study administrator. Throughout the trial every effort was made to ensure that the researcher who conducted outcome assessments remained blind to patients’ treatment allocation.

Therapists

Three experienced ED therapists delivered the e-mail support. All had undergone a 1-day training programme preparing them to deliver the intervention. The focus of the training was to standardise delivery of the e-mail contact. Therapists were instructed to keep e-mails brief, be informal, warm and supportive in tone, point patients in the direction of appropriate materials from their manual that might be relevant in keeping them well or helping them to make further improvement and provide non-intrusive monitoring of symptoms via the traffic light relapse prevention tool.

Statistical analyses

Outcomes were analysed on an ITT basis, that is, participants were analysed in the group to which they were randomised irrespective of their compliance with the assigned treatment. No formal hypothesis testing was undertaken, as this is a feasibility study. ESs (Cohen’s d) were calculated for continuous outcomes at 6 and 12 months from means, SDs and sample sizes using the ES calculator of the Campbell Collaboration [URL: www.campbellcollaboration.org/effect-size-calculato.html (accessed 3 July 2017).

Patient flow

Figure 4 shows the participant flow through the study. Twenty-four participants were randomised to iMANTRA plus TAU and 17 to TAU alone.

FIGURE 4.

Consolidated Standards of Reporting Trials flow diagram (WP5).

Patient characteristics at baseline

Patients in the two treatment groups were similar in terms of baseline sociodemographic and clinical characteristics (Table 27), with the exception of the WHOQOL-100 environmental subscale, according to which the iMANTRA group had significantly poorer quality of life than the TAU group [t(32.6) = –2.34; p = 0.025]. All had severe AN (mean admission BMI of 14.1 kg/m²), with just over half (51%) having had at least one additional previous admission to hospital. Patients had made considerable gains in weight while in hospital (discharge BMI of 18.0 kg/m²) and prior to coming into the study.

Acceptability of internet-based Maudsley Model of Anorexia Nervosa Treatment for Adults

Twenty-one of 24 (87.5%) patients allocated to iMANTRA plus TAU took up this treatment. Five iMANTRA patients completed the 12-month acceptability questions. All five said that the e-mail contact was the most helpful element of the intervention. All five participants rated the frequency of e-mails as ‘the right amount’.

Treatment outcomes

Table 28 shows outcomes on interview and questionnaire variables at 6 and 12 months.

At 6 months there was little difference in outcomes between groups, with ESs between d = –0.08 and d = 0.3 (i.e. negligible to small). At 12 months, patients receiving the experimental intervention had a higher BMI (d = 0.41) and lower total scores on the DASS-21 (d = 0.64). However, CIs were wide and overlapped with zero.

Service utilisation

Ten patients required readmission to hospital during the trial, 22.7% (5/22) in the iMANTRA group and 31.2% (5/16) in the TAU alone group. This difference was not significant [χ²(1) = 0.347; p = 0.556]. Comparable numbers of patients in each group had GP or outpatient care during the study period [iMANTRA: 12/18 (66.6%); and TAU alone: 10/15 (66.6%)]. The proportion of patients with no additional treatment was also similar in both groups [iMANTRA: 1/18 (5.5%); and TAU alone: 1/15 (6.6%)]. No patient died during the study period.

Harms

No harms were detected.

Strengths

The study recruited patients from a range of specialist units in the UK including those in the NHS and private sector and the patients recruited seemed representative of a typical AN inpatient population.

Limitations

The sample size was small and TAU-only patients were harder to follow up than those allocated to iMANTRA. This may reflect TAU-only patients’ disappointment about not being offered the experimental treatment component. No attempt was made to standardise TAU. Furthermore, the use of a TAU alone comparison also means that observed effects may not be due to iMANTRA specifically, but rather to participants receiving additional support of any kind. Future studies should remedy this.

Objectives

Three studies examining the effects of maternal EDs on (1) fertility, (2) their offspring’s dietary patterns and (3) growth trajectories.

Design

Data from the Avon Longitudinal Study of Parents and Children (ALSPAC).

Participants

A total of 11,088 women from the ALSPAC birth cohort and their children.

Main outcome measures

(1) Maternal self-report on time taken to conceive and whether they had ever received help to conceive, (2) maternal report of child diet at 38, 54, 81 and 103 months, and (3) child weight and height at 1, 2, 5 and 10 years.

Results

(1) Women with AN were more likely than women without an ED to have sought fertility treatment and to have had an unintentional pregnancy; women with AN and BN were more likely than women without an ED to have received help to conceive and to have taken longer than 6 months to conceive. (2) Children of mothers with AN and BN were more likely to be ‘health conscious’ in terms of diet and less likely to have a ‘traditional’ type of diet. (3) There was a complex pattern of differences in trajectories of height, Ponderal Index (PI) and BMI in children of women with an ED compared with children of HCs.

Conclusions

Women with an ED experience difficulties in the pre-conception period and show postnatal differences in their diet patterns. Growth trajectories in their children were found. Continuity of care from pre-conception to the postnatal period is paramount for women with an ED.

Psychiatric illness and parenting

Children of parents with a mental illness have an increased risk of developing psychological disturbances and psychiatric disorders themselves. 299,300 Within the study of EDs the topic of pregnancy and motherhood has been investigated only more recently. Nevertheless, there is evidence that an ED can affect parenting ability and have adverse outcomes on children’s development. 301 Furthermore, the offspring of mothers with an ED are at an increased risk of developing disordered eating themselves, and it has been proposed that a cycle of risk may be occurring, perpetuating EDs across generations. 302 Gaining a greater understanding of the associations between EDs, pregnancy and child development may extend our current understanding of the risk of intergenerational transmission of an ED.

Eating disorders and fertility

The presence of an ED can cause significant disruption to a woman’s menstrual cycle, and impaired fertility has been shown to exist across the spectrum of EDs. Women with a history of AN have shown to have greater disturbance in their menses than HCs. 303 Furthermore, these differences have been shown to continue in up to 30% of recovered women. 304 Women with a history of BN or EDNOS have also shown to frequently have greater disturbance in their menses,305,306 even though their body mass is usually within the normal range.

There is not enough research to understand the effect of EDs on fertility. Past research, based on two small studies of women looking for infertility treatment, found similar prevalences: 16–20% of women seeking infertility treatment met clinical criteria for an ED. 256,307 However, there have also been contradictory reports with regards to long-term follow-up studies. For example, Bulik et al. 308 found that the rates and frequencies of pregnancies were similar in women with AN and HCs; furthermore, Crow et al. 309 found the same results in women with BN. 310 On the other hand, when looking at births in women with AN, the number has been reduced to one-third of the expected rate.

A better and more precise measure of fertility may be time taken to conceive. 311,312 A majority of women will become pregnant after trying to conceive for between 3 and 6 months, with 90% of them succeeding after 12 months. 313 Therefore, and based on the data, infertility has been defined as inability to become pregnant after trying to conceive for a period of 12 months or longer,314 with delays in conception being a good measure of a range of underlying fertility problems.

Dietary patterns in children of mothers with eating disorders

There is evidence that mothers with an ED may find catering for the nutritional needs of their children especially challenging. However, only a small number of studies have examined the impact of a maternal ED on their child’s diet. The literature indicates a high risk for feeding difficulties in children of women with an ED. 315 These difficulties may lead to different nutritional intake in their children and have long-term implications for health and development.

Only one previous study has investigated the dietary intake in children whose mothers have an ED;316 although this study found the diet of the children to be generally unaffected, there was some suggestion that children of women with an ED consumed less junk food than children in the control group. However, this particular study was limited by the small sample size and the fact that the sample comprised children at different ages at a single time point. No previous studies have longitudinally investigated dietary patterns and nutritional intake in children of women with an ED.

Growth in the offspring of women with an eating disorder

Taking into consideration the possibility of an association between a maternal ED and their offspring’s feeding and eating habits, it is important to also consider the possibility of an association between a maternal ED and their offspring’s growth.

Maternal EDs, particularly AN, at the time of pregnancy are associated with higher risk of fetal growth restriction and low birthweight. 317,318 Conversely, it has recently been reported that women with binge ED during pregnancy are more likely to deliver babies that are large for gestational age. 319 Although preliminary studies have suggested that these children may reach an adequate postnatal growth in favourable environments,320 others indicate that altered growth patterns may continue throughout infancy and childhood. 321

The large majority of studies in the past have used small samples and included large age ranges, making it difficult to generalise findings. To the best of our knowledge, only one study has investigated growth in the offspring of mothers with an ED in a longitudinal manner, reaching the conclusion that children of women with an ED gained less weight at 1 year compared with HCs;322 however, their BMI was similar to those of controls when the children reached the age of 10 years. 323

Aims

The overall aim of the studies undertaken was to further investigate the effects of EDs on women’s fertility and the dietary patterns and growth trajectories of their offspring. Three separate studies are outlined to achieve the overall aims. The specific aims of each study are outlined below.

Design and participants

Data for this study were collected from the ALSPAC. 324 ALSPAC is a longitudinal birth cohort study, based in Avon, England, which includes all women who were willing to participate and who were due to have a child between 1 April 1991 and 31 December 1992.

A total of 14,663 women were initially enrolled in the study when they were 9 weeks pregnant. Using postal questionnaires, data were obtained from 14,472 women. In the current studies we only included singleton live births (n = 12,254) and excluded all women who were missing information on psychiatric history (n = 2019).

Outcomes and measures

Statistical analyses

Study 1: time taken to conceive among women with eating disorders

Study 2: longitudinal dietary patterns in children of women with lifetime eating disorders

Results from full models are reported in Easter et al. 329

Study 3: growth trajectories in children of women with lifetime eating disorders

Girls

The daughters of women with AN (between birth and 5 years of age) and of women with both AN and BN (between birth and 10 years of age) had lower predicted heights than girls in the unexposed group (Table 32). After adjusting for covariates, daughters of women with AN were expected to be, on average, 0.47 cm shorter than the daughters of women in the unexposed group.

TABLE 32 - Mean predicted anthropometry for girls across categories of maternal ED, with adjustment for standard confounders (gestational age, maternal age, maternal education, parity)

*p < 0.05; **p < 0.01.

Values are predicted from the multilevel models, and represent the predicted anthropometry for offspring of mean gestational age (39.4 weeks) and with a mother with the following characteristics: mean age 28.2 years, less than Ordinary level education, parity of zero. p-values are from z-tests comparing each group to the offspring of women with no eating or other psychiatric disorder.

p ≤ 0.1.

This table is reproduced from Easter et al. 330 This is an Open Access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 3.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See http://creativecommons.org/licenses/by-nc/3.0/.

Anthropometry	Mean predicted anthropometry (SD) of offspring of unexposed womena	Mean difference (SD) from unexposed for offspring of women with
		AN	BN	AN + BN	Other psychiatric disorders
Height (cm)	n = 4416	n = 65	n = 82	n = 30	n = 432
Birth	49.7 (5.3)	–0.47 (1.9)b	–0.08 (2.1)	+ 0.23 (2.0)	–0.16 (2.2)b
1 year	74.3 (5.3)	–0.27 (2.4)	+0.05 (2.4)	–0.73 (2.4)b	–0.18 (2.6)b
2 years	85.6 (5.3)	–0.24 (3.0)*	+0.02 (3.0)	–0.93 (3.0)b	–0.18 (3.2)
5 years	109.6 (6.0)	–0.31 (4.3)	–0.18 (4.3)	–0.94 (4.2)	–0.17 (4.8)
10 years	138.9 (8.6)	–0.57 (7.0)	–0.65 (6.9)	–0.32 (6.7)	–0.12 (7.5)
PI (kg/m3)	n = 4363	n = 64	n = 81	n = 30	n = 424
Birth	26.2 (5.3)	–0.07 (1.9)	+0.48 (2.2)*	–0.16 (2.2)	–0.09 (2.3)
1 year	23.2 (5.3)	+0.39 (2.4)b	+0.06 (2.5)	+0.39 (2.7)	–0.22 (2.7)b
BMI (kg/m2)	n = 4117	n = 61	n = 78	n = 29	n = 398
2 years	16.6 (3.9)	–0.35 (1.9)b	+0.30 (1.9)	+0.25 (1.8)	–0.07 (2.1)
5 years	16.0 (3.9)	–0.01 (1.5)	+0.26 (1.5)b	+0.32 (1.5)	+0.16 (2.1)*
10 years	18.1 (5.1)	+0.03 (3.5)	–0.29 (3.4)	–0.52 (3.3)	+0.56 (3.7)**

On the other hand, early in childhood, the daughters of women with BN were expected to be of similar height to the daugters of mothers in the unexposed group. At the age of 5 years, the daughters of women with BN were slightly shorter, and from 5 to 10 years of age the differences in predicted height were larger, with the daughters of women with BN predicted to be 0.64 cm shorter than the daughters of women in the unexposed group (see Table 32).

Children of women with other psychiatric disorders had a tendency to be shorter during childhood; however, when compared with the daughters of women with AN, or those with AN and BN, these differences were comparably smaller for this group (see Table 32).

Boys

Figures 5 and 6 show group differences in the trajectories of PI and BMI in boys. There was no statistical evidence of growth differences for all periods; in a large number of cases we found large SDs, which could be attributable to the fact that maternal ED groups were small.

FIGURE 5.

Average fractional polynomial curve of PI trajectories for girls and boys by maternal ED, birth to 2 years, adjusted for confounders. Note: values are predicted from the multilevel models, and represent the predicted anthropometry for offspring of mean gestational age (39.4 weeks) and with a mother with the following characteristics: mean age 28.2 years, less than Ordinary level education, parity of zero. (a) Boys’ PI and (b) girls’ PI. This figure is reproduced from Easter et al. 330 This is an Open Access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 3.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See http://creativecommons.org/licenses/by-nc/3.0/.

FIGURE 6.

Average fractional polynomial curve of BMI trajectories for girls and boys by maternal ED, aged 2–10 years, adjusted for confounders. Note: values are predicted from the multilevel models, and represent the predicted anthropometry for offspring of mean gestational age (39.4 weeks) and with a mother with the following characteristics: mean age 28.2 years, less than Ordinary level education, parity of zero. (a) Boys’ BMI and (b) girls’ BMI. This figure is reproduced from Easter et al. 330 This is an Open Access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 3.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See http://creativecommons.org/licenses/by-nc/3.0/.

After adjusting for covariates, the sons of women with AN had lower PI than the sons of unexposed women. The children of women across all exposed groups had a higher BMI between the ages of 2 and 10 years than the children of unexposed women. At 5 years of age, the predicted BMI of children of women in the maternal AN group was 0.34 kg/m² higher than the BMI of children of unexposed mothers. The BMI of sons of women with AN, and women with AN and BN, at 10 years of age was similar to the BMI of the sons of unexposed women (see Table 31). A tendency towards a higher PI at 2 years was found in the sons of mothers with other psychiatric disorders; however, their BMI trajectories a few years later were similar to those of the sons of mothers in the unexposed group.

Girls

The daughters of women with BN had a higher PI at birth than the daughters of unexposed women; however, this difference was reduced by the time they turned 12 months of age.

The daughters of women with BN, and of women with AN and BN, continued to have higher BMI than the daughters of unexposed mothers until the age of 5 (see Figure 6); however, their BMI was comparable to the BMI of daughters of unexposed women by the time they turned 10 years old. On the other hand, the daughters of women with AN had a lower BMI than the daughters of unexposed women during early childhood; however, their BMI was comparable to that of the daughters of unexposed women years later (late childhood) (see Figure 6). The BMI changes through childhood of the daughters of women who reported other psychiatric disorders were comparable to those observed in children of women in the AN group (see Figure 6); conversely, at the age of 10 years, their expected BMI was on average 0.56 points higher than that of the daughters of unexposed mothers.

Eating disorders and fertility

Study 1 highlighted that women with an ED experience more difficulties conceiving. The results showed that women with a lifetime diagnosis of AN (either AN or AN + BN) were more likely than unexposed women to visit a doctor because they have problems conceiving. Although only a small percentage of women in this sample actually received treatment during the time trying to conceive, and we found that women with AN and BN were twice as likely to undergo treatment for problems with conception. Based on these results, it could be hypothesised that problems with fertility in this sample may be secondary to low BMI, which is present in both groups.

Our results also showed that the odds of taking longer than 6 months to conceive were higher for women who reported both AN and BN than for unexposed women. Importantly, women with both AN and BN had lower BMI before conception and higher prevalence of lifetime purging than women in the other two groups. 318 The severity in this group may lead to a higher risk for both fertility problems and delayed conception.

Approximately 40% of women in the sample who reported having AN reported that their pregnancy was unintentional. Research has shown that ovulation and pregnancy can still occur in women with an ED who have irregular cycles. 331 Therefore, the large percentage of unintentional pregnancies found in this sample may be due to erroneous beliefs about their ability to become pregnant.

Diet and growth in children of women with eating disorders

Studies 2 and 3 explored potential differences in diet and growth between the ages of 3 and 9 years, and growth trajectories between birth and 10 years, in children of women with an ED.

The children of mothers with an ED were found to be less likely to adhere to a ‘traditional’ dietary pattern than children of women without an ED. This pattern is synonymous with a British ‘meat and two veg diet’, which would typically be eaten at mealtimes. Family mealtimes are particularly difficult for mothers who have experienced an ED,332 and such women have been found to be less likely to regularly cook or eat with their children. 333 Reduced adherence to the ‘traditional’ dietary pattern in this study may reflect less frequent mealtimes in families in which the mother has an ED.

In line with suggestions from previous research that children of women with an ED may consume less junk food,316 the children of mothers reporting either AN or BN were found to consume a more ‘health-conscious’ diet. We have previously found that women with AN in this sample were more likely to follow this dietary pattern in pregnancy themselves;334 these findings might imply a stronger maternal desire for women with an ED to provide a healthy diet for their children.

It is currently not known what effects increased exposure to a health conscious/vegetarian diet during childhood may have on the long-term development and health of the children of women with an ED; previous evidence from this sample suggests associations with positive developmental outcomes. 335,336

Conclusions

In conclusion, the investigations described in this chapter highlight novel findings regarding the effects of ED on women’s fertility and their children’s development. Women with an ED were found to experience difficulties in the pre-conception period and postnatal differences in the dietary patterns and growth trajectories of children born to women with an ED were highlighted.

Our study is the first to investigate longitudinal growth and diet patterns in the children of women with lifetime EDs, and, therefore, replication and additional examination of the pathways are necessary. Nonetheless, health professionals need to be conscious of the associations between EDs in women and their children’s growth so that they can better guide both the mother’s health and the development of the fetus. Continuity of care for this group of women is paramount from before conception into the postnatal period. Furthermore, and at a public health level, it is of extreme importance to further analyse underlying mechanisms of growth differences in children of women with an ED. Following on from this study, we are now developing an intervention for pregnant women with an ED, aimed at preventing adverse child outcomes.

Background

There is now a significant body of research demonstrating the efficacy of specific treatments for adolescent AN. Little is known about the way in which different service contexts impact treatment outcome. The present study investigates the role of specialist outpatient services and how the availability of such services influences the rates of referrals, hospital admissions, continuity of care and service user experiences.

Methods

Mental health services in London were asked to identify adolescents who presented for treatment for an ED over a 2-year period. Retrospective data about service use, including the need for hospital admissions, were collected from clinical case notes. A small sample of adolescents and parents were interviewed about their experiences of services.

Results

Direct access to specialist outpatient services was associated with higher referral rates, lower admission rates and greater consistency of care. Service users identified a number of advantages of specialist service provision.

Conclusions

Data from the present study suggest that facilitating direct access to specialist services for adolescents with AN may result in better outcomes, lower costs and higher satisfaction among service users.

Study 1: quantitative study

The design of the study was a naturalistic, retrospective cohort study, using clinical case notes as the primary data source. Our main hypotheses were (1) accounting for population size, more adolescents with AN and restricting EDNOS from the specialist care pathway than from the general care pathway would present to services beyond primary care and (2) controlling for severity of presentation, a lower proportion of adolescents with AN and EDNOS from the specialist care pathway than from the general care pathway would be admitted for inpatient treatment during the first year of treatment.

Study 2: qualitative study

For the qualitative part of the study, a subsample of adolescents was selected from those who allowed us to access their case notes for the quantitative study. The main aim of sampling was to select people with experience of the different care pathways. Fifty-six adolescents were asked whether or not they would be willing to take part in an interview about their experiences of treatment and services, and were also asked for permission to contact their parents about taking part in a similar interview. Interviews were based on an interview guide, rather than a fixed interview schedule. The focus was on the participant’s journey through services and experience of each stage of treatment. Questions were non-directive and open ended.

Data were analysed by two researchers using thematic analysis. The procedure was based on six phases of analysis, described by Braun and Clarke. 359 In phase 1, interviews were transcribed verbatim, and each researcher read through the transcripts multiple times to become familiar with the content, while making notes about anything they found interesting at a descriptive or interpretative level. Phase 2 involved labelling data with ‘codes’, which highlighted points of interest, and identifying areas of agreement and disagreement between participants. The researchers discussed their coding and, if they had made different interpretations of the data, revisited the transcripts together. Phase 3 involved tentative grouping of codes into themes, which were reviewed and reorganised when necessary in phase 4, and finalised and defined in phase 5. The data were organised to reflect patterns, similarities and differences in participants’ experiences. Themes were then grouped into superordinate themes. Phase 6 was the production of a written report of the findings, including sections of the transcripts (quotes) that led to the development of themes.

Study 1: quantitative study

Presentation rates to services beyond primary care

Primary care trust groups

Three PCT groups were identified, based on the care pathway they typically commissioned to provide outpatient treatment following referral from primary care: (1) a specialist Child and Adolescent Eating Disorders Service (CAEDS), (2) a specialist CAMHS and (3) a non-specialist CAMHS. In the CAEDS group, some PCTs allowed referrals to bypass CAMHS completely, some required CAMHS to authorise referrals, and others required CAMHS to assess adolescents before referring them for treatment in an ED service. In the specialist and non-specialist CAMHS groups, assessment and treatment were provided by the same service. Specialist CAMHS differed from CAEDS in that they each served a single PCT, whereas CAEDS each served a number of PCTs. All PCTs had access to general paediatric medical units. Depending on the area, longer-term inpatient treatment was provided by public or private ED units or general adolescent psychiatric units.

Cases by primary care trust group

Services identified 641 potentially suitable cases. Using initials, dates of birth and PCTs, duplicates were identified and removed, leaving 489 cases. After checking eligibility for the study, 111 cases were excluded. The final number of cases meeting the inclusion criteria was 378. Services from four PCTs did not provide any data and were excluded from the analyses. Therefore, a total of 27 PCTs and 367 cases were included in the subsequent analyses. The grouping of these is shown in Table 34.

TABLE 34 - Identified cases by PCT group

Cases when services did not provide a more detailed diagnosis than ‘eating disorder’.

PCT group	AN and EDNOS-AN		BN and EDNOS-BN		EDa		Total
	Female	Male	Female	Male	Female	Male
CAEDS (12 PCTs)	112	8	62	1	25	1	209
Specialist CAMHS (5 PCTs)	50	3	15	0	5	0	73
Non-specialist CAMHS (10 PCTs)	38	3	10	3	28	3	85
Total	200	14	87	4	58	4	367

The total number of males in the overall sample was small (n = 22), making it inappropriate to include them in the analyses. Data from the services could be used to provide a clear diagnosis for 287 females. Of these, 200 had AN or EDNOS-AN and 87 had BN or an eating disorder not otherwise specified – bulimia nervosa subtype (EDNOS-BN). In some of the sample (58 females), however, the services gave a diagnosis only of ‘eating disorder’ and there was insufficient data for a researcher to place these individuals into one of the specific diagnostic groups. Presentation rates were therefore calculated in two ways: first, using cases with a known diagnosis, to give an ‘observed presentation rate’, and, second, based on the assumption that the same proportion of missing and known cases had either AN or EDNOS-AN (69.7%) and BN or EDNOS-BN (30.3%) to give an ‘estimated presentation rate’. As the focus of this study is restricting EDs, results are presented for this subgroup only.

Figure 7 shows the presentation rates of AN and EDNOS-AN by PCT group. Unadjusted pairwise comparisons show that the presentation rates for the non-specialist CAMHS group were significantly lower than those of the specialist ED service group (observed rate: p < 0.01; estimated rate: p = 0.04). The presentation rates of the specialist CAMHS group did not differ significantly from those of the CAEDS group (observed rate: p = 0.41; estimated rate: p = 0.98).

FIGURE 7.

Observed and estimated presentation rates of AN and EDNOS-AN by PCT group.

Study 2: qualitative study

The number of cases seen in services beyond primary care

This is the most robust finding as it is based on the full sample of 367 cases. The low level of case identification in areas with no specialist services is consistent with previous research. 26 The striking finding from this study is that the rate of referral to specialist services is close to the level that epidemiological studies suggest is the actual incidence of new cases seen in primary care. 51,361 It is possible that GPs with links to specialist services have better knowledge of EDs and their treatment and therefore identify and refer more cases than GPs with no specialist links. Alternatively, the visibility and easy access to specialist services may encourage referrals rather than a ‘wait and see’ approach that is sometimes seen in primary care. The second explanation is supported by Currin’s362 primary care study, which showed that GP’s working in areas where inpatient treatment was the only specialist service provision have higher referral thresholds than those in areas with specialist outpatient services. Our own clinical experience in a specialist context of receiving a growing number of very early referrals also supports this explanation. A third explanation is that availability of services has an effect on patient and/or parent preferences and changes the threshold for seeking help.

Admissions for inpatient treatment

Although there have been previous clinical reports of reductions in admission rates following the development of specialist outpatient services,363 this is the first study that provides clear empirical support for this. Our study showed that admission for inpatient treatment was not predicted by service-level agreements between PCTs and service providers, but was predicted by participants commencing treatment in a non-specialist CAMHS. The high level of admission in the non-specialist care pathway raises a number of concerns. Inpatient treatment of AN is costly,354,364 has high relapse rates,228,365 and there is evidence that hospital treatment is associated with poor outcomes in adolescent AN even when severity of symptoms are controlled for. 353

Continuity of care

In the present study, the differences in the continuity of care between the specialist care and non-specialist care pathway was striking. This is important as continuity potentially reduces duplication of work, and avoids the need for managing transitions between services, which can be problematic for patients and their families. 36

Service user experiences

Adolescent and parents highly valued the level of clinician expertise that they found in specialist services. Those who were dissatisfied with CAMHS were frustrated by barriers to specialist services. Movement between different services could disturb the delicate process of building and maintaining helpful therapeutic relationships, and a lack of continuity of care could be experienced as being ‘sent from pillar to post’ – this was one of the most frequent areas of dissatisfaction for those who were referred to CAMHS. Group interventions were particularly important for parents, and these are likely to be more easily facilitated by specialist services because of the resources available and the greater number of families that can be seen at any one time. Service user reports suggest that work may need to be done with GPs to increase their understanding of EDs beyond the physical symptoms – again, specialist services may be best placed to undertake such work.

Study limitations

Although the differences between the specialist and non-specialist care pathways found in the present study are striking, a degree of caution is needed as we do not know the extent to which this will generalise beyond London, as the study area differs from the rest of the UK in some important ways. The most relevant is that London contains considerably more specialist ED services than rest of the country. We do not know what impact this may have on the attitudes of referrers or the confidence of clinicians working in generic CAMHS. The finding concerning the differences in the continuity of care is probably most susceptible to the specific conditions in London and may not generalise well to areas with no specialist services.

The study did not include every service that could conceivably accept a referral of an adolescent with an ED, but rather focused on services that handle the large majority of such referrals. Although the exclusion of services such as specialist CAMHS targeted at other disorders may have resulted in comorbid cases being excluded, the study was able to obtain data from most of the ED services, CAMHS that specialised in EDs and non-specialist CAMHS in London, with 84% of eligible services providing data. The impact of the different methods of participant identification used by different services is unclear.

Although the sample used to test hypothesis 1 was relatively large, and included cases from the majority of the PCTs in London, the subsample of consenters used to test hypothesis 2 was comparatively small (42.3% of eligible cases), and the analyses of admission rates were underpowered. The logistic regression analyses did not take into account any PCT clusters, so the effect of belonging to an individual PCT on admission was not explored. The differential recruitment rate for specialist and non-specialist services could have biased the sample and influenced the results. However, comparisons between consenters and non-consenters showed no significant differences in the number of services accessed or admissions for inpatient treatment during the study period. Moreover, the findings of high admission rates from general services is consistent with previous reports30,351 as is the lower rate in specialist services. 363 The short follow-up period for this study is a further limitation, as findings relate only to participants’ service use for a year after presentation.

Although the results of the qualitative study cannot be taken as representative of a wider population, they suggest a number of reasons why service users may find specialist services preferable to non-specialist services. These are best described as hypotheses at present, and warrant further study.

Implications for service development and future research

The results of this study are potentially of considerable significance for the way treatment of child and adolescent EDs is provided in the NHS. The data suggest that establishing care pathways in which referrals from primary care go directly to specialist outpatient CAEDS could significantly increase levels of case identification, while at the same time reducing overall costs by reducing admissions for inpatient treatment and the associated readmissions, which are likely to increase cost differentials over time. This study did not examine the nature of the treatments provided within different services or the impact that specific treatments may have had on outcomes. However, it is likely that part of the explanation of the differences found is a greater use of effective, evidence-based treatments by specialist services. One could argue that a further factor in favour of developing specialist services is that this would facilitate dissemination of such treatments.

The fact that CAEDS and specialist CAMHS appeared to be relatively comparable in this study suggests that general CAMHS may be able to achieve similar outcomes to specialist services if they contain some organisation around EDs and clinicians have opportunities to develop necessary expertise. The disadvantage of this arrangement, in comparison with specialist ED services, is that developing a truly multidisciplinary team is often difficult when only a small group of clinicians is involved and such teams are also more vulnerable to the impact of staff turnover (in fact, during the service mapping exercise we heard of several instances when a mini-specialist team was unable to continue when a key member of staff retired or changed jobs).

Conclusion

The study presented here can be considered one of the first stages in the investigation of outcomes following treatment in services with different levels of specialisation in EDs. Findings suggest that starting treatment in a specialist outpatient service may have some advantages. Questions remain, however. The difficulty in recruiting participants from CAMHS in particular raises the question of how to evaluate treatment outside of specialist contexts. Future research will need to address this issue.

Background

We aimed to identify services and treatments used by people with AN and associated costs, to estimate the unit costs of ED treatments, to explore cost variations by patient characteristics, to explore the economic consequences of AN and to estimate the annual costs of AN for England.

Methods

We analysed service use data collected alongside three clinical trials and estimated costs by attaching unit costs to service use. Individual variations in costs were explored using regression analyses. Unit costs of ED treatment and service costs for a cohort of young people with AN over 1 year were estimated using data from the Care Pathways Study (see Chapter 10, Study 1: quantitative study). Economic consequences of AN in adulthood were analysed using BCS-70 data,366 comparing women with lifetime AN and women without eating problems. Combining publicly available data and results from the ARIADNE studies, we estimated the annual costs of AN in terms of health service costs, private health-care costs, benefit receipt and years of potential life lost (YPLL).

Results

Service costs were driven by hospital admissions and vary by age, illness severity and treatment history. Those treated in non-specialist outpatient services incurred higher costs than those treated in specialist services only, with no differences in outcome. Women with lifetime AN were more likely to be long-term sick or disabled in adulthood, to receive benefits and to have completed a degree, with no differences in weekly income or employment status compared with cohorts with no eating problems. The annual costs of AN for England were estimated to be between £45M and £230M in 2011.

Methods

Results

Service costs

Table 38 shows the costs associated with service use, summarising the service use data presented in Table 37 into service categories. Reflecting the pathways of recruitment and the location of the interventions to be evaluated in the RCTs, average costs per person were highest for inpatient treatment in iMANTRA and CASIS, whereas outpatient treatment costs were the largest contributor to total costs in MOSAIC.

TABLE 38 - Costs per person by service category (three trials)

A&E, accident and emergency; AMHS, adult mental health services; n/a, not applicable.

Note

Hospital includes inpatient, outpatient, A&E. Mental health includes psychiatrist, psychologist, community psychiatric nurse, psychotherapist, family therapist, multiple-family day treatment, individual therapist, CBT, CAMHS, AMHS, crisis team, residential rehabilitation. Primary care includes: GP, practice nurse, community nurses, dentist, optician, dietitian/nutritionist. Community services include counsellor, alternative therapy, solicitor/lawyer, physiotherapy, occupational therapy, osteopathy, police. Self-help and advice includes: self-help group, support group, Citizens Advice Bureau, helplines, websites. Social care includes: social worker, outreach worker, family support worker, family centre, carer.

Service category	Trial
	CASIS			MOSAIC			iMANTRA
	Mean (£)	SD	Range	Mean (£)	SD	Range	Mean (£)	SD	Range
Hospital	21,045	30,370	0–215,172	4547	14,403	0–87,794	81,304	12,029	21,672–87,651
Mental health	1687	2261	0–11,232	709	1564	0–11,458	1062	2894	0–13,224
Primary care	1650	1409	0–7360	1046	1304	85–12,272	229	493	0–2076
Community services	286	489	0–1752	115	247	0–1460	3	16	0–105
Self-help and advice	32	99	0–673	59	394	0–4500	0	n/a	n/a
Social work	95	397	0–3780	70	383	0–3549	49	293	0–1872
Total costs	24,795	31,121	318–224,025	6546	15,316	138–96,287	82,647	11,296	25,200–95,124

The proportion of total cost absorbed by each cost category for each study can be seen in Figures 8–10. Total costs over the 6 months prior to baseline were highest for the iMANTRA group, in which all participants had used inpatient care over the previous 6 months; hospital costs accounted for 99% of total costs. In the CASIS group, members of which were also recruited from an inpatient population, hospital costs still accounted for 85% of total costs, and only around 6% of this was due to outpatient visits, used by over 60% the participants. In the MOSAIC group – recruited through outpatient services – only 16% reported a hospital admission in the 6 months prior to the intervention. For all three studies, community, self-help and social work services contributed a very small proportion to total costs even though some of these services were used by over 10% of the study samples.

FIGURE 8.

Distribution of costs by service category: MOSAIC.

FIGURE 9.

Distribution of costs by service category: CASIS.

FIGURE 10.

Distribution of costs by service category: iMANTRA.