Inpatient versus outpatient care, partial hospitalisation and wait-list for people with eating disorders

  • Protocol
  • Intervention

Authors


Abstract

This is the protocol for a review and there is no abstract. The objectives are as follows:

  1. To assess the effects of treatment setting (inpatient, partial hospitalisation or outpatient) on reduction in symptoms and increase in remission rates in people with:

    1. anorexia nervosa and atypical anorexia nervosa; and

    2. bulimia nervosa and other eating disorders.

  2. To assess the effects of continuing hospitalisation until the person is restored to a normal weight in people with anorexia nervosa.

  1. anorexia nervosa and atypical anorexia nervosa; and

  2. bulimia nervosa and other eating disorders.

Background

Description of the condition

Eating disorders are a group of diagnoses defined by severe disturbances in eating behaviour. They include anorexia nervosa, bulimia nervosa, binge eating disorder and other specified or unspecified eating disorder (APA 2013; WHO 1992). Anorexia nervosa is an eating disorder with features of weight loss or failure to gain weight. This leads to a weight that is less than 85% of that expected for height and age, an intense fear of gaining weight, a distorted body image and loss of at least three consecutive menstrual cycles. Bulimia nervosa, by definition, is a disorder of normal or above normal weight characterised by binge eating and extreme weight control methods to compensate for the binge eating. Recurrent binge eating is defined as eating unusually large amounts of food over which there is a sense of loss of control. The compensatory extreme weight-control behaviours may take the form of self induced vomiting with or without laxative or diuretic use (purging), or fasting with or without intense exercise (the non-purging form of bulimia nervosa). In addition to disordered eating behaviours, people with bulimia nervosa also have specific eating disorder psychopathology whereby their self view or self evaluation is unduly influenced by concerns about their weight, shape, body image or a combination of these. Other eating disorders may share the body image, shape and weight concerns of those people with anorexia nervosa or bulimia nervosa, or share the disordered eating or weight control behaviours, or share a combination of these. People with binge eating disorder have regular binge eating behaviour but do not have regular extreme weight control behaviours. Other specified eating disorders include atypical anorexia nervosa (where the person's weight is in or above the normal range); bulimia nervosa or binge eating disorder of low frequency or limited duration or low frequency and limited duration; purging disorder and night eating syndrome. Finally, unspecified feeding or eating disorder is a heterogeneous category where patients do not meet criteria for an eating disorder but have a clinical eating or feeding disorder syndrome. It most usually comprises people with disordered eating behaviours with or without body image disturbance characteristic of the better-defined disorders. This review will not address feeding disorders such as 'avoidant restrictive food intake disorder', which are categorised with eating disorders in the new Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) diagnostic scheme (APA 2013; WHO 1992).

Eating disorders are common (Hoek 2003; Hudson 2007), and increasing (Lucas 1991; Hay 2008). One national US survey found that lifetime prevalence estimates of DSM-IV anorexia nervosa, bulimia nervosa and binge eating disorder were 0.9%, 1.5% and 3.5%, respectively among women, and 0.3%, 0.5% and 2.0%, respectively among men (Hudson 2007). In Australia between 1995 and 2005, there was a two-fold increase in the point prevalence of eating disorder behaviours in the general community (Hay 2008).

Description of the intervention

Treatments for eating disorders have been developed for outpatient, inpatient and partial hospitalisation (the latter is also known as day hospital care). Clinical practice guidelines vary on recommendations, but there is some consensus that inpatient care is most often needed for people with anorexia nervosa (NICE 2004; RANZCP 2004). Bulimia nervosa and other eating disorders seldom require an admission unless there are medical complications, for example, hypokalaemia (reduced levels of potassium in the blood) or high suicide intent (NICE 2004; RANZCP 2004; APA 2006). Inpatient programmes are usually multidisciplinary (where treatment providers include psychologists or psychiatrists (or both), dieticians, nurses and other allied healthcare specialists) and involve a programme of nutritional counselling and supervised meals, combined with individual and group psychotherapy and medical care (La Puma 2009). Partial hospitalisation is similar to inpatient programmes with regards to multidisciplinary care, intensity of therapy, the capacity for regular supervision of meals and the direct provision of meals (Thornton 2009). However, there is no overnight stay with partial hospitalisation. In contrast, outpatient care does not provide regular meal supervision. In outpatient care, therapy is also usually less frequent (e.g. occurring one or two times a week) and care is less likely to include therapists of multiple disciplines. Indeed, in an outpatient programme, care may be delivered by a single therapist of one discipline (e.g. a psychologist). We present an overview of the different components of care across treatment settings in Table 1.

Table 1. Comparative components of treatments across different settings
  1. 1 Psychological care refers to individual, group and family psychotherapy. This may be delivered by psychologists, psychiatrists or people from other disciplines such as social work who have specific training in psychological therapies.

 InpatientPartial or day hospitalOutpatient 
Anorexia nervosaMultidisciplinary medical, nursing, dietetic (re-feeding) and psychological care1Multidisciplinary medical, nursing, dietetic (re-feeding) and psychological careMultidisciplinary medical, nursing, dietetic (re-feeding) and psychological care or psychological care incorporating re-feeding and other elements delivered variably 
Bulimia nervosaMultidisciplinary medical, nursing, dietetic and psychological careMultidisciplinary medical, nursing, dietetic and psychological careMay be multidisciplinary but is most usually individual psychotherapy with limited medical care of complications from purging 
Binge eating disorderMultidisciplinary medical, nursing, dietetic and psychological careMultidisciplinary medical, nursing, dietetic and psychological careMay be multidisciplinary but is most usually individual psychotherapy with dietetic care for those with obesity 
Other eating disordersMultidisciplinary medical, nursing, dietetic and psychological careMultidisciplinary medical, nursing, dietetic and psychological careMultidisciplinary medical, nursing, dietetic and psychological care or psychological care and other elements delivered variably 

There is no one psychotherapeutic approach that is applied consistently in outpatient care worldwide, although cognitive behavioural therapy has the best evidence base for bulimia nervosa (NICE 2004; Hay 2009). In adolescents and children, a family-based approach is regarded as the first line in treatment (NICE 2004). It is superior to 'treatment as usual' but there is insufficient evidence for any single form of family therapy (Fisher 2010). Derivations of the Maudsley model (as described in Le Grange 2005 and in Rhodes 2009) have gained ascendancy worldwide.

Medical care comprises physician and nursing management of physical aspects of eating disorders. This includes re-feeding but also management of osteopenia and other effects from starvation. A comprehensive account of medical management is provided in the text by Birmingham and Treasure (Birmingham 2010). Medications for physical or psychological co-morbidities are seldom used as a stand-alone treatment but will often be used in either inpatient or outpatient settings. Evaluations of the relative efficacy of medications in eating disorders are also found in Flament 2012 and Hay 2012. These reviews reported that the evidence of efficacy of drug treatments is mostly weak or moderate with generally low recovery rates. However, there was support for the use of antidepressants, particularly high-dose fluoxetine in bulimia nervosa, and anticonvulsants (topiramate) for binge-eating disorder. Attrition rates were usually higher than for psychotherapies and combined treatments in bulimia nervosa and binge eating disorder had better outcomes than either approach alone. Low-dose antipsychotic medication was considered to be possibly clinically useful as adjunct treatment of anxiety in anorexia nervosa but more trials were needed.

Further details of psychological approaches in treatment may be found in complementary Cochrane reviews of family therapy (Fisher 2010), and individual outpatient psychotherapies for anorexia nervosa (Hay 2003), and other eating disorders (Hay 2009). There are also three Cochrane reviews of antidepressant medication use, one in bulimia nervosa (Bacaltchuk 2003), one in anorexia nervosa (Claudino 2006), and one of combination medication and psychotherapies in bulimia nervosa (Hay 2001).

However, while there is a growing evidence base for treatments (albeit still sparse in anorexia nervosa (Lock 2009)), many people with eating disorders do not access treatment. In particular, up to half of people with anorexia nervosa may never present for treatment (Keski-Rahkonen 2007), and attrition in anorexia nervosa treatment trials can be unacceptably high with reasons for dropout difficult to identify (Halmi 2005). Stigma and perceived fear of hospitalisation may well contribute to underutilisation of services. Where trials have compared inpatient versus outpatient care, preference for the outpatient care (Freeman 1992) and higher attrition in the inpatient care have been found (Gowers 2007). In addition, there is concern from professional bodies as, for example, expressed in position statements by the Australian and New Zealand Academy for Eating Disorders that inpatient care may be being underutilised, undervalued, or both, leading to a reduction in services (www.anzaed.org.au/uploads/7/3/9/2/7392147/positionstatementinpatient.pdf). While many jurisdictions are not now providing inpatient services, lack of evidence hampers the ability to argue for these.

How the intervention might work

There is consensus that therapy in eating disorders includes dietetic, medical, nursing and psychological care. In anorexia nervosa, multidisciplinary care is expected (NICE 2004). This is in order to be able to provide the level of expertise in medical or physical care, individual or group psychotherapy, and re-feeding that is needed. Medical or physical care may be administered by a paediatrician, general practitioner or adult physician; individual and group therapy from a psychologist, psychiatrist or other therapist; and re-feeding from a dietician (Steinhausen 2002; Zipfel 2000).

Inpatient and partial hospitalisation have the benefit of being able to provide care for extended periods (6 to 24 hours a day). Notably, meals can be directly supervised and staff can respond quickly to psychiatric or physical emergencies (such as re-feeding syndrome) (Treasure 2005; La Puma 2009). However, outpatient care is argued to be both more effective and efficient in terms of therapy time and cost. Fairburn has described a comprehensive outpatient approach in his treatment manual for underweight, normal or overweight patients with eating disorders (Fairburn 2008). Such treatment is usually conducted where inpatient care for medical or psychiatric crises (e.g. hypokalaemia due to vomiting or active suicidal ideation) is available. Admissions are for brief periods sufficient for stabilisation prior to resumption of outpatient care.

It is thought that outpatient care preserves the patient's sense of autonomy and is perceived as collaborative. Thus, there is increased patient acceptability, and less risk of further psychological harm to a patient's already fragile self esteem. It may also be associated with less stigma, and the maintenance of usual social and work (including educational) activities.

Partial hospitalisation has the advantages of increased intensity of care (and supported or supervised meals) found in inpatient services. It also has an argued increased preservation of personal autonomy and life activities that are associated with outpatient care. However, it has higher fiscal cost than outpatient care.

Why it is important to do this review

Eating disorders have high social, medical and fiscal costs (Crow 2003). For example, in Australia they are the 12th leading cause of mental health hospitalisation expenses (Mathers 1999). It is agreed that the more intensive care of a hospital admission is mandatory where there is high medical or psychiatric risk (APA 2006; NICE 2004; RANZCP 2004). However, its advantage where patients may be 'safely' cared for as an outpatient is unproven. Outpatient care is also preferred by most people and has perceived advantages. These include being more collaborative in approach, with reduced associated stigma, maintenance of work/education and social relationships, and fiscal savings. An admission necessarily will lead to loss of usual social contacts and interruption to work or education. Cost savings associated with avoiding hospitalisation or reducing hospital stay, or both, to that needed for medical stabilisation alone could be very large. For example, one UK study found that outpatient care costs approximately 10% the cost of inpatient care (Katzman 2000).

However, there are few studies into the role of treatment settings in mental health and these are largely confined to anorexia nervosa. Previous systematic reviews have found results from studies of anorexia nervosa to be contradictory (Lock 2009; Meads 1999; Meads 2001). While some studies have found higher discharge weights to be associated with better long-term weight outcomes (Barren 1995; Gross 2000; Zipfel 2000), a more recent trial linked hospital admissions with poor weight outcomes (Gowers 2000). One very early trial also reported similar weight outcomes from in and out patient management (Crisp 1991). In addition, although hospitalisation is effective in the short-term, one 20-year retrospective study found it was not predictive of long-term recovery (Zipfel 2000).

The aim of this review is to investigate whether or not there is demonstrable benefit to inpatient compared with outpatient or day patient care in eating disorders, beyond that which is essential for treating acute medical and psychiatric emergencies. This review will extend the work of previous systematic reviews in anorexia nervosa (Meads 1999; Meads 2001; Lock 2009, and further strengthen the portfolio of Cochrane reviews in anorexia nervosa and other eating disorders (Hay 2001; Hay 2003; Bacaltchuk 2003; Pratt 2002; Perkins 2006; Claudino 2006; Hay 2009; Fisher 2010).

Objectives

  1. To assess the effects of treatment setting (inpatient, partial hospitalisation or outpatient) on reduction in symptoms and increase in remission rates in people with:

    1. anorexia nervosa and atypical anorexia nervosa; and

    2. bulimia nervosa and other eating disorders.

  2. To assess the effects of continuing hospitalisation until the person is restored to a normal weight in people with anorexia nervosa.

Methods

Criteria for considering studies for this review

Types of studies

Parallel design randomised controlled trials. We will include cross-over and cluster randomised trials.

Types of participants

Adults, adolescents and children with a diagnosis of acute anorexia nervosa, bulimia nervosa, binge eating disorder, or specified or unspecified eating disorder according to DSM-5 (APA 2013), or International Classification of Disease (ICD)-10 criteria (WHO 1992), or other internationally accepted diagnostic criteria such as the DSM-IV (APA 2000). We will analyse trials of weight restoration in anorexia nervosa separately to those of other eating disorders. We will stratify trials of other eating disorders by diagnosis (bulimia nervosa, binge eating disorder, other eating disorder) in analyses. We will extract and record data according to DSM-5 criteria. For example, trials of DSM-IV (APA 2000) anorexia nervosa and anorexia nervosa eating disorder not otherwise specified (EDNOS) type, where all are underweight, will now be trials of DSM-5 anorexia nervosa and trials of DSM-IV bulimia nervosa and EDNOS, where the binge eating frequency meets DSM-5 criteria for bulimia nervosa, will be treated as trials of bulimia nervosa.

We will not exclude because of co-morbidities.

Types of interventions

Experimental interventions
  1. Inpatient care for weight restoration in anorexia nervosa, provided by a specialist eating disorder service and health professionals.

  2. Inpatient care for bulimia nervosa and other eating disorders, provided by a specialist eating disorder service and health professionals.

Comparators
  1. Individual or group or family-based outpatient care (maximum two contacts per week) provided by specialist eating disorder health professionals.

  2. Individual or group outpatient care (maximum two contacts per week) that is provided by non-eating disorder specialist health professionals.

  3. Inpatient care from a specialist or a non-specialist eating disorder service for medical stabilisation that is time limited (maximum three weeks) and discharge before full weight restoration with planned outpatient follow-up.

  4. Wait-list (no active treatment for the eating disorder) group.

  5. Partial hospital or day patient care (more than two contacts per week and more than three hours per day and includes clinician-supervised meals).

As 1, 2 and 3 all include active outpatient psychotherapies, we will enter data to allow for analyses of these as a grouped comparator.

Types of outcome measures

Studies that meet the above inclusion criteria will be included regardless of whether they report on the following outcomes.

Primary outcomes
1. Clinical improvement

a. For anorexia nervosa: weight (body mass index (BMI), kg/m2) used as a proxy for physical health, corrected for age in adolescent and child samples, at end of treatment where groups are not significantly different in mean BMI at start of treatment. (Note: in the meta-view graph where standardised mean difference (SMD) is greater than zero, the active treatment is favoured. This differs from the convention of SMD less than zero favouring treatment.)

b. For bulimia nervosa, binge eating disorder, specified bulimia nervosa and binge eating disorder of low frequency, limited duration, or both: binge eating (as defined in DSM-IV (APA 2013)) frequency.

c. For specified atypical anorexia nervosa, purging disorder or unspecified eating disorder: global eating disorder symptom score as measured by the Eating Disorder Examination interview (Fairburn 1993) or questionnaire (Fairburn 1994).

2. Acceptability

To be measured by proportion of drop-outs. For this outcome, we will stratify the graph by type of eating disorder.

Secondary outcomes
3. Clinical response

a. For anorexia nervosa:

  • weight restoration to within the normal weight range for participant sample (e.g. BMI, 19 to 25 for female young adults, or greater than 85% of that expected for age/height), or;

  • recovery according to an agreed published definition (Kordy 2002) (e.g. the Morgan 1975 narrow scale of a good outcome, namely normal body weight (greater than 85% of mean for age, gender and height) with normal menstruation or an intermediate outcome, namely normal body weight (greater than 85% of mean for age, gender and height) with no menstruation, or;

  • the Morgan 1988 broader scale ratings (covering nutritional status, menstruation, mental state, psychosexual adjustment and socioeconomic status) of intermediate or better outcome. (In the meta-view graphs, this is depicted as those in a category less than two so that RR greater than 1 favours the active treatment.)

b. For bulimia nervosa, binge eating disorder, specified bulimia nervosa and binge eating disorder of low frequency, limited duration, or both: binge eating abstinence

4. Recovery

Global eating disorder symptom score as measured on the Eating Disorder Examination interview (Fairburn 1993), or questionnaire within one standard deviation above the community mean (Fairburn 1994).

5. Quality of life score

Measured by any validated questionnaire (e.g. the Short-Form-12) (Ware 1996) or eating disorder specific (e.g. the Engel quality of life instrument) (Engel 2005).

6. Depression symptom severity

Measured by any validated questionnaire (e.g. the Beck Depression Inventory) (Beck 1996).

7. Cost-effectiveness

Measured by the calculation of incremental cost-effectiveness ratios (ICER) - the additional costs of one intervention compared with another divided by the additional effects of one intervention compared with another.

8. Remission or recovery (as defined in 1a, 1b or 2 above) sustained over one-year follow-up.

Note: where outcomes are combined across eating disorders (i.e. 2, 3, 4, 5 and 6), we will stratify graphs by type of eating disorder.

Timing of outcome assessment

Outcome assessments will be done at end of treatment and one-year follow-up.

Search methods for identification of studies

Electronic searches

1. CCDAN's Specialized Register (CCDANCTR)

The Cochrane Depression, Anxiety and Neurosis Group (CCDAN) maintains two clinical trials registers at their editorial base in Bristol, UK, a references register and a studies based register. The CCDANCTR-References Register contains over 31,500 reports of randomised controlled trials in depression, anxiety and neurosis. Approximately 65% of these references have been tagged to individual, coded trials. The coded trials are held in the CCDANCTR-Studies Register and records are linked between the two registers through the use of unique Study ID tags. Coding of trials is based on the EU-Psi coding manual. Please contact the CCDAN Trials Search Coordinator for further details. Reports of trials for inclusion in the Group's registers are collated from routine (weekly), generic searches of MEDLINE (1950-), EMBASE (1974-) and PsycINFO (1967-); quarterly searches of the Cochrane Central Register of Controlled Trials (CENTRAL) and review specific searches of additional databases. Reports of trials will also be sourced from international trials registers c/o the World Health Organization's trials portal (ICTRP), drug companies, the handsearching of key journals, conference proceedings and other (non-Cochrane) systematic reviews and meta-analyses. Details of CCDAN's generic search strategies can be found on the Group's website.

  • The CCDANCTR-Studies Register will be searched on condition alone ("anorexia nervosa" or bulimia or "eating disorder*") and records will be screened to identify relevant studies.

  • The CCDANCTR-References Register will be searched for additional untagged/uncoded records using a more sensitive set of free-text terms (focusing on the treatment setting): (anorexi* or bulimi* or "eating disorder*" or EDNOS) and (setting or inpatient* or in-patient* or outpatient* or out-patient* or hospital* or admission* or confinement or clinic or clinics or "clinical management" or "clinical support" or specialist or "specialized treatment" or "specialised treatment" or supervi* or "day care" or "day centre*" or "day center*" or "day unit*" or "day treatment*" or "community mental health" or "mental health service*" or residential or referral or referred or "patient care" or (weight and restor*) or feed* or re-feed* or refeed*)

2. International Trial Registers

The WHO International Clinical Trials Registry Platform (ICTRP) and ClinicalTrials.gov will be searched on condition alone ("anorexia nervosa" or bulimia or "eating disorder*").
Records will be screened to identify relevant unpublished and/or ongoing studies.

Searching other resources

We will inspect the reference lists of all papers and reviews (identified from a separate systematic review search) for further relevant studies.

We will conduct a citation search (on the Web of Science) to identify additional reports citing any of the included studies.

We will send personal letters to lead researchers and trialists in the field of eating disorders.

Data collection and analysis

Selection of studies

We will evaluate all studies according to the inclusion criteria listed above. One review author (PH) will inspect abstracts and three review authors (PH, AC and CS) will independently select studies based on abstracts and reading full-text articles. If the abstract indicates that it was a trial of treatment setting for eating disorders, three review authors (PH, AC and CS) will review the full article to determine, first, if the trial was randomised and second, if it was a trial of treatment setting for eating disorders meeting our inclusion criteria. Consensus between authors will be reached through discussion.

We will translate non-English texts.

Data extraction and management

Authorship will not be concealed at the point of data collection. Two review authors (PH, AC) will extract data independently to include documentation of the country or specific cultural aspects (or both) of the treatment setting. A third review author (CS) will adjudicate where there are discrepancies. We will extract data on features of each trial, quality appraisal, interventions and outcomes found according to a pre-designed data extraction form (see Appendix 1). We will enter agreed data into a spreadsheet program, and into the Review Manager 5 software program (RevMan 2012).

Main planned comparisons:
  1. Inpatient care for weight restoration in anorexia nervosa provided by a specialist eating disorder service and health professionals; and

  2. Inpatient care in bulimia nervosa and other eating disorders by a specialist eating disorder service and health professionals.

Versus:
  1. Individual or group or family-based outpatient care (maximum two contacts per week) provided by specialist eating disorder health professionals; or

  2. Individual or group outpatient care (maximum two contacts per week) provided by non-eating disorder specialist health professionals; or

  3. Inpatient care from a specialist or a non-specialist eating disorder service for medical stabilisation that is time limited (maximum three weeks) and discharge before full weight restoration with planned outpatient follow-up;

  4. Wait-list (no active treatment for the eating disorder) group;

  5. Partial hospital or day patient care (more than two contacts per week and more than three hours per day and includes clinician supervised meals).

As 1, 2 and 3 above all include active outpatient psychotherapies data will be entered to allow for analyses of these as a grouped comparator.

Assessment of risk of bias in included studies

For each included study, two review authors (PH, AC) will independently complete The Cochrane Collaboration's tool for assessing risk of bias (Higgins 2011, Section 8.5.1) with any disagreements to be resolved through consultation with a third review author (CS).

We will allocate each domain one of three possible categories for each of the included studies: 'low risk of bias', 'high risk of bias' and 'unclear risk of bias'.

We will also document the following trial features.

  1. Adequate method of sequence generation:

    1. appropriate method of randomisation used;

    2. method of randomisation not described;

    3. randomised method described but not randomised (e.g. every alternate participant given the control treatment).

  2. Adequacy of allocation sequence concealment ('allocation concealment'):

    1. indicates adequate concealment;

    2. indicates uncertainty about whether allocation was adequately concealed;

    3. indicates the allocation was definitely not adequately concealed.

  3. Blinding of participants and personnel:

    1. blinding of both outcome assessor and participant (double-blind);

    2. blinding of outcome assessor only (single-blind);

    3. blinding not done.

  4. Incomplete outcome data (attrition bias):

    1. attrition less than 20%;

    2. attrition 20% to 49%;

    3. attrition 50% or greater.

  5. Selective outcome reporting:

    1. complete outcomes reported analyses;

    2. outcomes reported on subset (e.g. only those participants who were treated according to protocol or other selective outcome reporting, e.g. selective omission of outcomes from reports).

  6. Other sources of bias: the comparability of groups after randomisation with regards to the following putative demographic and illness severity confounding factors: age, gender, body weight, severity of illness at study inception (using measures applied at outcome assessment):

    1. Groups comparable at baseline on demographic and illness severity;

    2. Uncertain - comparability not assessed; and

    3. Groups not comparable at baseline.

Although imbalance may occur by chance, it may also occur where there is inadequate randomisation (or exclusion of participants after randomisation) or inadequate allocation concealment (Higgins 2011 Section 8.14.1.2) and it is thus relevant in assessing bias for group comparability to be assessed and reported.

Measures of treatment effect

We will conduct risk ratio (RR) analyses for dichotomous outcome data and SMD analyses for continuous outcome data, together with 95% confidence intervals (CI). We chose RRs, as they are less likely to overestimate a treatment effect (Higgins 2011 Section 9.2.2.3).

Unit of analysis issues

Cluster-randomised trials

If we find cluster-randomised trials, we will proceed according to the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011 Section 16.3.3) (i.e. to extract a direct estimate of the required effect measure from an analysis that properly accounts for the cluster design (according to statistical advice of author SL)). We will then meta-analyse the effect estimates and their standard errors from correct analyses of cluster-randomised trials using the generic inverse-variance method in Review Manager 5 (RevMan 2012).

Cross-over trials

If we find cross-over trials, we will follow the procedure according to the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011 Section 16.4.4). If carry-over or period effects are not thought to be a problem, then we will analyse continuous data from a two-period, two-intervention cross-over trial with a paired t-test. This will evaluate the value of 'measurement on experimental intervention (E)' minus 'measurement on control intervention (C)' separately for each participant. We will include the effect estimate in a meta-analysis using the generic inverse-variance method in Review Manager 5 (RevMan 2012).

Studies with multiple treatment groups

If we find multiple intervention groups, we will follow the procedure recommended by the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011 Section 16.5.4) and first determine which intervention groups are relevant to this systematic review, and second determine which intervention groups are relevant to a particular pair-wise comparison in the meta-analysis. All intervention groups of a multi-intervention will be described in the 'Characteristics of included studies' table. We will describe interventions relevant to this review and thus those potentially used in analyses in detail. If more than two intervention groups are relevant, we will, if possible, combine groups to create a single pair-wise comparison (e.g. if a trial has one inpatient condition and two outpatient conditions we will plan to combine the two outpatient conditions if appropriate). If this cannot be done, we will determine if it is possible to split the 'shared' group into two or more groups with smaller sample size, and include two or more (reasonably independent) comparisons. If this cannot be performed, we will select one pair of interventions and exclude the others.

Dealing with missing data

We will contact study authors to request information not available in the published study, including information needed for data analyses, subgroup and sensitivity analyses, quality evaluation of the trials, and to obtain the results of unpublished or partly published trials. We will contact study authors on a maximum of two occasions. We will ask study authors to respond on each occasion within three months of the request. We will do calculations of unpublished data, such as the standard deviation, where there is sufficient information (published or unpublished) provided as per Chapter 7.7 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).

We will record when and what information was supplied by the study authors in the 'Risk of bias' sections of the 'Characteristics of included studies' tables.

Where there are missing statistics, we will seek this first from the study authors or, if possible, we will calculate it using methods that enable this (e.g. if standard deviations are missing it may be possible to calculate these from the CI values, standard errors or t values). As the number of trials will be few, we will take statistical advice on imputation and only attempt this if the majority of trials in the meta-analysis have completed statistics (Higgins 2011 Section 16.2.3.1).

The only data that will be imputed for categorical outcome variables are where outcomes on weight restoration or binge/purging abstinence (or both) are not available. In these first instances, we will assume that if there is no follow-up information the participant(s) concerned did not attain normal weight for age and height or did not achieve abstinence (or both).

In the case of continuous data, we will only include data on those whose results are known. We will consider the potential impact of the missing data on the results in the interpretation of the results of the review with reference to the amount of missing data, the pooled estimate of the treatment effect and the variability of the outcomes, which will also be considered as a potential source of heterogeneity. We will perform a sensitivity analysis for continuous data using the method described in Higgins 2011 (Section 16.2.3), which is to assume a fixed difference between the actual mean for the missing data and the mean assumed by the analysis.

Assessment of heterogeneity

We will assess heterogeneity using the Chi2 test (P value < 0.10) and the observed value of the I2 statistic. The observed value of the I2 statistic depends on (i) the magnitude and direction of effects and (ii) the strength of evidence for heterogeneity (e.g. P value from the Chi2 test, or a CI for the I2 statistic). However, interpretation of the I2 statistic can be misleading, since the importance of inconsistency depends on several factors. Higgins 2011 has provided a guide to interpretation with overlapping bands as follows:

  • 0% to 40%: might not be important;

  • 30% to 60%: may represent moderate heterogeneity;

  • 50% to 90%: may represent substantial heterogeneity;

  • 75% to 100%: considerable heterogeneity.

If the I2 statistic is moderate to high and the direction and magnitude of treatment effects suggests important heterogeneity, we will investigate the putative source of the heterogeneity (see below) (Higgins 2011).

Assessment of reporting biases

We will investigate systematic differences between reported and unreported findings by inspection of funnel plots (in meta-analyses of 10 or more trials) and statistical tests for funnel plot asymmetry of primary continuous outcome variable(s). However, it is acknowledged that an asymmetrical funnel plot is not necessarily indicative of publication bias and that publication bias does not necessarily cause asymmetry.

We will attempt to minimise putative duplicate publication bias by checking with authors for suspected duplicate publication. We will attempt to minimise location, language and citation bias by comprehensive and systematic searches that are as broad as possible and include non-English language trials.

Data synthesis

We will favour the random-effects model when there are different interventions and where it is unlikely that the same treatment effect is being estimated. Thus, we will use the random-effects model in all analyses.

Subgroup analysis and investigation of heterogeneity

It is planned that we will combine data for all eating disorders in a single meta-analysis of the same outcome for the same intervention comparison. We will enter data such that the following subgroup analyses can be conducted.

  1. Diagnostic subgroups of DSM-5 (APA 2013), bulimia nervosa, binge eating disorder and other specified or unspecified eating disorders.

  2. Children and adolescents (i.e. under 17 years of age).

  3. Adults and older adolescents (i.e. 17 years of age or older).

  4. Family-based inpatient programmes and non-family-based inpatient programmes.

  5. Cognitive behavioural therapy versus other individual psychotherapies used.

We will address identified heterogeneity by first checking again that the data are correct. If data are correct and if there is a large degree of inconsistency in results, we will consider conducting a sensitivity analysis by sequentially removing trials by sample size starting with the smallest until there are only three trials or heterogeneity is reduced to a non-significant (P value ≥ 0.1) level, or both.

Sensitivity analysis

We will perform sensitivity analyses whereby the following types of trials will be excluded.

  1. Trials with attrition rates of greater than 50%.

  2. Trials where care is not multidisciplinary across all intervention groups.

  3. Non-blinded trials.

Where there is a large amount of missing continuous data then we will perform a sensitivity analysis using the method described in Higgins 2011 (Section 16.2.3), which is to assume a fixed difference between the actual mean for the missing data and the mean assumed by the analysis.

Summary of findings tables

Summary of findings tables will be developed to summarise the key findings of the review, for all relevant populations. For the given comparisons, we will select up to seven of the most important outcomes (both desirable and non-desirable) and present the findings for these relating to standardised effect size estimates (and 95% CIs) to illustrate comparative risk, the number of studies and participants, and the quality of evidence based on standards of the GRADE working group (see Balshem 2011).

Acknowledgements

CRG Funding Acknowledgement
The National Institute for Health Research (NIHR) is the largest single funder of the Cochrane Depression, Anxiety and Neurosis Group.

Disclaimer
The views and opinions expressed herein are those of the authors and do not necessarily reflect those of the NIHR, National Health Service or the Department of Health.

Appendices

Appendix 1. Data extraction form

INPATIENT TREATMENT FOR EATING DISORDERS

 

Trial ID    

           

Date Form completed              

Other comments                       Extractors initials    

Title:   Author(s):                    Year (Issue)   Volume   Pages    Country   Language  

Inclusion criteria   Men                   Women               AN      BN      BED       EDNOS     

Diagnosis DSM IV     Diagnosis  ICD10            Diagnosis other (specify).                

Describe Inpatient program/care - Type of programme and psychotherapies applied including intensity and numbers of sessions and if pharmacotherapy was included and if so what medications.        

Describe comparison program/care  

1.  

2.  

3.  

4.      

Interventions multidisciplinary YES/NO      

Inpatient non-family-based programme YES/NO                  

Outpatient non-family-based programme YES/NO  

Inpatient family-based programme YES/NO             

Outpatient family-based programme YES/NO  

Partial hospitalisation family based YES/NO           

Partial hospitalisation not family based YES/NO

Interventions evidence based YES/NO      

Inpatient non-family-based programme YES/NO                  

Outpatient non-family-based programme YES/NO  

Inpatient family-based programme YES/NO             

Outpatient family-based programme YES/NO  

Partial hospitalisation family based YES/NO           

Partial hospitalisation not family based YES/NO

Interventions incorporating re-feeding YES/NO      

Inpatient non-family-based programme YES/NO                  

Outpatient non-family-based programme YES/NO  

Inpatient family-based programme YES/NO             

Outpatient family-based programme YES/NO  

Partial hospitalisation family based YES/NO           

Partial hospitalisation not family based YES/NO  

Interventions not incorporating re-feeding

Inpatient non-family-based programme YES/NO                  

Outpatient non-family-based programme YES/NO  

Inpatient family-based programme YES/NO             

Outpatient family-based programme YES/NO  

Partial hospitalisation family based YES/NO           

Partial hospitalisation not family based YES/NO

 

Selection bias

1.      How was the allocation sequence generated? Specify method & circle below:

 

Adequate                                                                    Inadequate

Computer generated sequence                                    Case number

Random number tables                                                 Date of birth

Lot drawing                                                               Date of admission

Coin tossing                                                              Alternation

Shuffling cards                                                              Other (specify)

Throwing dice

 

OR     NOT STATED

 

Was the allocation sequence adequately generated?

Yes (low risk of bias)

No (high risk of bias)

Unclear (uncertain risk of bias)

 

Were participants cluster randomised?

Yes

No

Unclear

2. Was there adequate concealment of allocation sequence? Specify method & circle:

Adequate (low risk of bias)

Central and/or computerised randomisation

Coded boxes

Envelopes:   Sealed / Opaque   Sequentially numbered

 

Inadequate (high risk of bias)

Open allocation sequence

Procedures based on inadequate generation

         

Unclear (uncertain risk of bias)

 

 

3 Outcome of randomisation.

A. Groups comparable at baseline on demographic features and illness severity

B. Uncertain -comparability not assessed

C. Groups not comparable at baseline

 

 

Performance bias

Measure used to attempt or ensure blinding of trial participants and key personnel from knowledge of which intervention a participant had received.

Was effectiveness of blinding assessed? YES/NO Specify method.

Patient

Adequate

Inadequate

Feasible

Unfeasible

Unclear           

NA            

Therapist            

Adequate

Inadequate

Feasible

Unfeasible

Unclear           

NA

Outcome assessor    

Adequate

Inadequate

Feasible

Unfeasible

Unclear           

NA

Incomplete data outcome

Were reasons for attrition/exclusions reported? YES/NO

 

Flow diagram          N

Potentially eligible       

Excluded before randomisation

Number randomised in each group Control   Experiment  

Post randomised exclusions in

each group

No. analysed          

No. analysed at 12 months follow-up                 

 

Circle

1. Low risk - attrition < 20 %   Some risk - attrition 20% to 49% High risk ≥ 50%

2. Intention to treat analyses  Completers only analyses

 Follow-up points

Three months             YES/NO              

12 months YES/NO

Six months              YES/NO                  

> 12 months (specify) YES/NO

  

OUTCOME RESULTS

                                                                        Inpatient                    Control                                                                                    group                           group

 

 
 Primary outcome measure

Event                    n

or mean (SD)

Event                   n

or mean (SD)      

1BMI  
2Binge eating frequency  
3Global ED symptom score  
4Weight restoration  
5Morgan recovered  
6No ED behaviour  
7Binge eating abstinence 1 month  
 Secondary outcomes  
1Quality of life generic  
2Quality of life ED specific  
3Depression  
4Psychological symptoms general  
5Patient satisfaction  
6Adaptive function  
7Morgan broad  
8Non-completers (any reason)  
9Non-completers (adverse event)  
     

 

 

SUBGROUP ANALYSIS  1.

Diagnostic subgroups of eating disorders, anorexia nervosa, bulimia nervosa, and subtypes of EDNOS

  ANonly                                           Inpatient                    Control                     

 

 Primary outcome measure

Event                    n

or mean (SD)

Event                   n

or mean (SD)      

1BMI  
3Global ED symptom score  
4Weight restoration  
5Morgan recovered  
6No ED behaviour  
 Secondary outcomes  
1Quality of life generic  
2Quality of life ED specific  
3Depression  
4Psychological symptoms general  
5Patient satisfaction  
6Adaptive function  
7Morgan broad  
8Non-completers (any reason)  
9Non-completers (adverse event)  

 

 

      BN only                                                                             Inpatient Control                   

 

 
 Primary outcome measure

Event                    n

or mean (SD)

Event                   n

or mean (SD)      

2Binge eating frequency  
3Global ED symptom score  
6No ED behaviour  
7Binge eating abstinence 1 month  
 Secondary outcomes  
1Quality of life generic  
2Quality of life ED specific  
3Depression  
4Psychological symptoms general  
5Patient satisfaction  
6Adaptive function  
8Non-completers (any reason)  
9Non-completers (adverse event)  

    BED only                                                                            Inpatient                    Control                                                                                   

 

 
 Primary outcome measure

Event                    n

or mean (SD)

Event                   n

or mean (SD)      

2Binge eating frequency  
3Global ED symptom score  
6No ED behaviour  
7Binge eating abstinence 1 month  
 Secondary outcomes  
1Quality of life generic  
2Quality of life ED specific  
3Depression  
4Psychological symptoms general  
5Patient satisfaction  
6Adaptive function  
8Non-completers (any reason)  
9Non-completers (adverse event)  
     

 

 

  EDNOS                                                                      Inpatient                    Control                       

 

 
 Primary outcome measure

Event                    n

or mean (SD)

Event                   n

or mean (SD)      

1BMI  
2Binge eating frequency  
3Global ED symptom score  
4Weight restoration  
5Morgan recovered  
6No ED behaviour  
7Binge eating abstinence 1 month  
 Secondary outcomes  
1Quality of life generic  
2Quality of life ED specific  
3Depression  
4Psychological symptoms general  
5Patient satisfaction  
6Adaptive function  
7Morgan broad  
8Non-completers (any reason)  
9Non-completers (adverse event)  
     

 

SUBGROUP ANALYSIS  2. Children and adolescents

     Children and adolescents                          Inpatient               Control 
 Primary outcome measureEvent/mean (SD) nEvent /mean (SD) n                     
1BMI  
2Binge eating frequency  
3Global ED symptom score  
4Weight restoration  
5Morgan recovered  
6No ED behaviour  
7Binge eating abstinence 1 month  
 Secondary outcomes  
1Quality of life generic  
2Quality of life ED specific  
3Depression  
4Psychological symptoms general  
5Patient satisfaction  
6Adaptive function  
7Morgan broad  
8Non-completers (any reason)  
9Non-completers (adverse event)  
   Adults                                                          Inpatient                    Control                          
1BMI  
2Binge eating frequency  
3Global ED symptom score  
4Weight restoration  
5Morgan recovered  
6No ED behaviour  
7Binge eating abstinence 1 month  
 Secondary outcomes  
1Quality of life generic  
2Quality of life ED specific  
3Depression  
4Psychological symptoms general  
5Patient satisfaction  
6Adaptive function  
7Morgan broad  
8Non-completers (any reason)  
9Non-completers (adverse event)  
     

SUBGROUP ANALYSIS  3. Family based vs. non family based

     family-based programme                               Inpatient               Control 
 Primary outcome measureEvent/mean (SD) nEvent /mean (SD) n                     
1BMI  
2Binge eating frequency  
3Global ED symptom score  
4Weight restoration  
5Morgan recovered  
6No ED behaviour  
7Binge eating abstinence 1 month  
 Secondary outcomes  
1Quality of life generic  
2Quality of life ED specific  
3Depression  
4Psychological symptoms general  
5Patient satisfaction  
6Adaptive function  
7Morgan broad  
8Non-completers (any reason)  
9Non-completers (adverse event)  
   Non-family-based programme                                     Inpatient                      Control                          
1BMI  
2Binge eating frequency  
3Global ED symptom score  
4Weight restoration  
5Morgan recovered  
6No ED behaviour  
7Binge eating abstinence 1 month  
 Secondary outcomes  
1Quality of life generic  
2Quality of life ED specific  
3Depression  
4Psychological symptoms general  
5Patient satisfaction  
6Adaptive function  
7Morgan broad  
8Non-completers (any reason)  
9Non-completers (adverse event)  
     

 

Contributions of authors

PH is lead author and responsible for all stages of the review.

AC and CM are experienced Cochrane review authors and will assist at all stages including data checking and quality appraisal ratings.

SM is a content expert and will assist in trial selection and appraisal, data interpretation and preparation of final manuscript.

Sl is a biostatistician who will contribute to statistical analyses and data interpretation.

Declarations of interest

PH: receives sessional fees and lecture fees from the Australian Medical Council, Therapeutic Guidelines publication, and New South Wales Institute of Psychiatry and royalties from Hogrefe and Huber, McGraw Hill Education and Blackwell Scientific Publications, and she has received research grants from the National Health and Medical Research Council (NHMRC) (2010 to 2013) and Australian Research Council (ARC) (2010 to 2011). She is Deputy Chair of the National Eating Disorders Collaboration in Australia (2012 to 2013).

ST has received royalties from McGraw Hill and Hogrefe and Huber, as well as research support from the NHMRC.

AC, CM, SM, SI: none known.

Sources of support

Internal sources

  • None, Not specified.

External sources

  • None, Not specified.

Ancillary