Antipsychotic drugs for anorexia nervosa
Abstract
This is a protocol for a Cochrane Review (Intervention). The objectives are as follows:
1. The primary objective of this review is to determine if antipsychotic drugs are clinically effective in improving weight gain and reducing the core psychopathology of AN when compared to placebo.
2. Secondary objectives are:
(a) To investigate whether different classes/types of antipsychotics (eg. atypical versus typicals) differ in efficacy or acceptability for AN;
(b) To compare the efficacy and acceptability of treatments (e.g. inpatient programs, specific psychotherapies) for AN combined with an antipsychotic drug versus the same intervention(s) alone (i.e. without any drug);
(c) To investigate if antipsychotics may be effective in ameliorating the core psychopathology of AN (e.g. distorted body image, morbid preoccupation with weight and shape) independently of the effect in weight restoration;
(d) To identify factors (clinical, methodological) which predict response to treatment with antipsychotic drugs.
Background
Anorexia nervosa (AN) is an eating disorder characterised by cognitive distortions related to overconcern with body weight and shape, serious restriction of food intake and other disturbed eating and compensatory behaviours (e.g. purging) that lead to severe weight loss. The current psychiatric classification manuals, Diagnostic and Statistical Manual, fourth edition, text‐revision (DSM‐IV‐TR) (APA 2000) and the International Classification of Diseases, revision 10 (ICD‐10) (WHO 1992), require the following features for the diagnosis of AN (acute phase) : (a) refusal to maintain weight within the normal range for height and age; (b) fear of weight gain; (c) body image disturbance and (d) absence of menstrual cycles or amenorrhoea in women (and loss of sexual interest in men). In ICD‐10, the impact that AN may have on growth and development (delay) in children or those in early adolescence is emphasised. In addition, two types of AN are described in DSM‐IV‐TR (APA 2006): the restricting and the binge‐purge type (detailed in Table 1).
| Criteria |
| 1. DSM‐IV‐TR Diagnostic criteria for anorexia nervosa: |
AN has an average point prevalence of 0.3% in adolescents and young females (Aalto‐Setala 2001; Hoek 2003; Favaro 2004). A more recent population‐based survey suggests that the lifetime prevalence of AN is more common in adult men than previously assumed: 0.3% in men versus 0.9% in women (just three times more frequent in women than in men) (Hudson 2006).
Many factors have been suggested to play a role in the pathogenesis of AN, including a genetic predisposition and a range of environmental risk factors (Jacobi 2004; Schmidt 2003; Fairburn 1999). It is not known, however, how they interact to develop and to maintain the disorder. Changes in neuroendocrine systems known to be involved in feeding, mood, anxiety and thought processes have been observed during the acute phase of AN (Kay 1988; Connan 2003). Although the observed fluctuations are considered to be a consequence of the starvation state and pathological eating behaviours, the finding that altered levels of the serotonin and dopamine neurotransmitters persist after weight restoration raises the possibility that they are involved in the aetiology of this disorder (Kaye 1991; Kaye 1999; Frank 2002). In line with this view, Kaye 1999 reported persistent low levels of cerebrospinal fluid (CSF) homovanillic acid (HVA) after recovery in patients with the restricting‐type AN. More recently, Frank 2005 found increased dopamine D2/D3 receptor binding in the anteroventral striatum (which includes the nucleus accumbens) through positron emission tomography in people with AN that could be associated with the harm avoidance and increased physical activity usually present in these people. These findings suggest that dopamine disturbances in AN could represent a trait‐related vulnerability of altered reward mechanisms (for instance, of altered hedonics of feeding behaviour). Thus, research on alterations in neurotransmitter systems is of considerable interest for developing specific targets for pharmacological interventions in the treatment of AN.
The treatment of AN is considered challenging. AN holds a mortality rate that is amongst the highest of all functional psychiatric disorders, five to ten per cent per decade of illness in outcome reviews (Steinhausen 2002; Nielsen 2001; Sullivan 1995). AN is frequently a chronic illness, with only one‐half of people achieving full recovery in the long‐term (Steinhausen 2002).
Research in AN faces many obstacles, such as difficulties in recruiting sufficient numbers and retaining participants to the completion of treatment (Halmi 2005). As a result, approaches to AN recommended in recognised guidelines largely rely on uncontrolled trials and on the consensus of expert opinion (APA 2000; NCCMH 2004; RANZCP 2004). Both the medical risks of being at low weight and the role of starvation in perpetuating the disorder make early nutritional rehabilitation a priority, with the aim to restore weight and menstrual cycles. Usually this is achieved with a multidisciplinary approach, which can be delivered in many settings (inpatient, outpatient, day care). Psychosocial interventions are considered a key part of the treatment, although for adults with AN no particular approach has emerged as being superior to others in randomised clinical trials (RCTs), most of which were conducted unblinded (Treasure 2005, Hay 2003). For children and adolescents, research suggests family‐based approaches are the treatment of choice (Russell 1987; Eisler 1997; Eisler 2000; Fairburn 2003). A new report comparing two cohorts of female inpatients suggest a reduction in the mortality of AN related to the introduction of specialist care units (Lindblat 2006).
Although not considered a first‐line treatment in AN, pharmacotherapy is frequently used as an adjunctive intervention. The focus of pharmacotherapy in AN depends on the phase of illness (Powers 2004). In the acute phase, when people are underweight and fulfilling all the diagnosis criteria, drugs are expected to favour weight gain and to reduce symptoms of AN (weight phobia, body image disturbance, excessive preoccupation with food, weight and shape, excessive exercise, eating rituals, bulimic and purging symptoms). Associated psychopathology (depressive, anxious and other obsessive‐compulsive symptoms not related to body, eating or food) are also target symptoms of the treatment. In the maintenance phase, after weight and menses have been restored and core symptoms improved, drugs are expected to prevent relapse (deterioration of overall symptomatology). In both stages non‐specific aims of the treatment include improvement in functioning and quality of life.
Most drug trials in AN have focused on the acute phase of the illness, but little evidence of efficacy of drugs tested was found (Treasure 2005; Claudino 2006). Fluoxetine is the only drug to have been tested in the maintenance phase of treatment, with conflicting results reported in two randomised, placebo‐controlled trials (Kaye 2001; Walsh 2006).
The rationale for using neuroleptic drugs is based on their possible action on symptoms of AN that could be considered to be "psychotic‐like" (Powers 2004), such as: (a) the fixed, false belief that they are obese when they are in fact emaciated; an unrealistic interpretation of weight that seems to reach delusional proportions; (b) the body image disturbance or distortion, where people perceive themselves to be larger than they are, and (c) possible hallucinations in some cases, such as a voice telling them " not to eat ". Others (Mondraty 2005) conceptualise the morbid preoccupations in AN as "ruminative thinking", with intrusive thoughts (usually ego‐syntonic and not resisted) and rituals related to eating (e.g., compulsive calorie‐counting) and body (e.g., body checking), a conceptualisation that suggests an obsessional character of the symptoms. If an antipsychotic drug can help lessen the intensity of these symptoms, the anxiety and resistance to gain weight may decrease and people with AN may improve. Besides these psychopathological reasons for using antipsychotics to treat AN, the biological findings discussed earlier (neurotransmitter dysregulations) and the weight gain inducing properties of many neuroleptic drugs (e.g., atypical antipsychotics) lend some support to the evaluation of these medications in AN (Taylor 2000).
The first trials of antipsychotics for AN tested chlorpromazine against regular inpatient treatment without drugs. Although higher weight gain occurred in the short‐term, in the chlorpromazine group significant adverse effects were reported and no difference in long‐term outcome was observed (Dally 1960; Dally 1966). Other early small trials of antipsychotics performed for AN tested pimozide (Vandereycken 1982) and sulpiride (Vandereycken 1984) against placebo and against behaviour therapy (Weizman 1985), but did not find any clear benefit of these drugs on eating behaviours or weight gain.
In recent years, interest in using antipsychotics to treat AN has been renewed with the emergence of the atypical drugs with their broader range of targeted symptoms and different spectrum of side‐effects. Several case reports of using atypical antipsychotics for AN have been published (Bosanac 2005). Olanzapine and risperidone were the most frequently used drugs in uncontrolled studies, usually in low to moderate doses. In most case studies, people with AN were severely underweight cases with treatment resistant eating disorder symptoms, important co‐morbid psychopathology and histories of previous unsuccessful treatment. Antipsychotics were usually added to inpatient treatment (Montecchi 1998; Bianco 2000; Gaskill 2001; Powers 2002; Malina 2003; Barbarich 2004). In general, case reports and open trials suggested clinical improvement (compared to baseline) in weight gain, core eating disorder symptoms (some of delusional features), compulsive activity, co‐morbid symptoms and social withdrawal. However, some studies with a comparison control group failed to find a significant difference in weight gain between groups (Gaskill 2001; Barbarich 2004). In one RCT , the atypical antipsychotic amisulpride was compared to fluoxetine and clomipramine, without a placebo arm (Ruggiero 2001). Amisulpride showed the greatest increase in weight from baseline, but there were no statistical differences between groups on end‐of‐treatment mean weight or core AN symptoms.
The aim of this systematic review is to evaluate the efficacy and acceptability of antipsychotic drugs for the treatment of AN. As RCTs in AN are few in number, and usually of small size (Treasure 2005), we anticipate that combining data in meta‐analysis from available studies will increase power to detect a treatment effect.
Objectives
1. The primary objective of this review is to determine if antipsychotic drugs are clinically effective in improving weight gain and reducing the core psychopathology of AN when compared to placebo.
2. Secondary objectives are:
(a) To investigate whether different classes/types of antipsychotics (eg. atypical versus typicals) differ in efficacy or acceptability for AN;
(b) To compare the efficacy and acceptability of treatments (e.g. inpatient programs, specific psychotherapies) for AN combined with an antipsychotic drug versus the same intervention(s) alone (i.e. without any drug);
(c) To investigate if antipsychotics may be effective in ameliorating the core psychopathology of AN (e.g. distorted body image, morbid preoccupation with weight and shape) independently of the effect in weight restoration;
(d) To identify factors (clinical, methodological) which predict response to treatment with antipsychotic drugs.
Methods
Criteria for considering studies for this review
Types of studies
RCTs and quasi‐randomised controlled trials (quasi‐RCTs). Studies with a cross‐over design will be included.
Types of participants
People diagnosed with anorexia nervosa by any defined criteria such as Russell's (Russell 1970), ICD (e.g., ICD‐10) (WHO 1992) or DSM criteria (e.g., DSM‐IV‐TR) (APA 2006).
Other criteria:
Participants can be recruited through primary, secondary and tertiary health‐care services. Age and gender will not be used as exclusion criteria. Trials that include patients with psychiatric co morbidities will also be included.
Types of interventions
Interventions testing any antipsychotic in comparison with placebo or other antipsychotic drug, either as monotherapy or adjunctive intervention in the treatment of AN.
Trials testing any antipsychotic added to "routine treatment for AN" (combination of interventions) versus only "routine treatment for AN (without any drug)" will also be included.
Treatment should last at least four weeks and may be conducted on an outpatient, inpatient or day‐care basis.
Types of outcome measures
Primary outcome measures
1. Weight gain or weight restoration measured as follows:
(a) number of participants achieving target weight or weight within a normal range (e.g. BMI >18.5 kg/m2 or weight >85% of average for age, gender and height)
(b) end‐of‐treatment mean absolute weight or body mass index (BMI) (where groups were not significantly different in mean weight/BMI at start of treatment);
(c) any other consistent measure of change in weight (e.g. rate of weight gain).
Secondary outcome measures
1. Eating disorder symptoms severity reduction, evaluated by any recognised and validated eating disorder questionnaire or interview (e.g., Eating Disorders Examination (EDE)(Cooper 1989); Eating Disorders Inventory (EDI)(Garner 1983) and others). The following outcome measures will be considered:
(a) number of participants achieving scores within the "normal" range (when a cut‐off point has been previously determined);
(b) mean scores on the scales, questionnaires or interviews.
2. Recovery evaluated through the number of participants presenting:
(a) Morgan 1975 narrow criteria of good or intermediate outcome:
i) A good outcome, namely normal body weight (>85% of average for age, gender and height) with normal menstruation.
ii) Intermediate outcome, namely normal body weight (>85% of average for age, gender and height) with no menstruation.
(b) Morgan 1988 broader scale ratings of average outcome.
3. Psychiatric symptoms severity reduction, measured by mean scores on any recognised and validated rating scale that measures:
(a) level of psychotic symptoms;
(b) level of anxiety symptoms, including obsessive‐compulsive ones;
(c) level of depression symptoms.
Examples of these include Yale‐Brown Obsessive‐Compulsive Scale (YBOCS)(Goodman 1989) and the Positive and Negative Syndrome Scale for Schizophrenia (PANSS)(Kay 1988).
4. Global clinical improvement, assessed using validated scales as the Clinical Global Impression‐Improvement item (CGI‐I) or a similar scale, and evaluated through number of patients presenting scores of "much improved" or "improved". In trials that employ multiple measures of improvement, data from scales that have also been used by other included trials in the review will be selected in order to aggregate data in meta‐analysis.
5. Acceptability of the treatment measured by:
(a) proportion of drop‐outs due to any reason or post‐randomisation exclusions (although there is a possibility of bias in this measure);
(b) proportion of drop‐outs due to adverse effects.
Search methods for identification of studies
Electronic searches
1. The Cochrane Depression, Anxiety & Neurosis Review Group Register (CCDANCTR‐references and CCDANCTR‐studies) will be searched using the following terms:
CCDANCTR‐Studies
Diagnosis = "Eating Disorders" or "Anorexia Nervosa"
and
Intervention = (Aripiprazole or Clozapine or Ziprasidone or Haloperidol or Thioridazine or Thiothixene or Fluphenazine or Chlorpromazine or Risperidone or Quetiapine or Perphenazine or Trifluoperazine or Olanzapine or Zuclopenthioxol or Droperidol or Pimozide or Sulpiride or Antipsychotic* or Neuroleptic )
CCDANCTR‐References
Keyword = "Eating Disorders" or "Anorexia Nervosa"
and
Free Text = (Aripiprazole or Clozapine or Ziprasidone or Haloperidol or Thioridazine or Thiothixene or Fluphenazine or Chlorpromazine or Risperidone or Quetiapine or Perphenazine or Trifluoperazine or Olanzapine or Zuclopenthioxol or Droperidol or Pimozide or Sulpiride or Antipsychotic* or Anti‐Psychotic* or "Anti Psychotic*" or Neuroleptic )
2. The Cochrane Controlled Trials Register (CENTRAL/CCTR) on the Cochrane Library will be searched using the following terms:
clozapine or olanzapine or pirenzepine or sertindole or imidazoles or indoles or risperidone or quetiapine or ziprasidone or piperazines or thiazoles or sulpiride or zotepine or zotepine or amisulpride or sulpride or deniban or solian or socian or sulamid or DAN or abilify or OPC or OPC or clozaril or alemoxan or elcrit or froidir or klozapol or leponex or W or LX ‐ or HF or olasek or ziprexa or zyprexa or LY‐ or seroquel or ICI or ZD or ZH or risperdal or belivon or risperin or rispolept or rispolin or neripros or rizodal or R or serdolect or serlect or zoleptil or aiglonyl or alimoral or ansium or arminol or betamaks or calmoflorine or championyl or darleton or depex or depreal or desisulpid or digton or dixibon or dobren or dogmatil or dogmatyl or dolmatil or dresent or eclorion or eglonyl or enimon or equilid or esipride or fardalan or fidelan or intrasil or lebopride or mariastel or meresa or mirbanil or neogama or neoride or noneston or norestran or normum or nufarol or nylipark or omaha or omiryl or ozedeprin or paratil or prosulpin or psicocen or quastil or quiridil or restful or stamoneurol or sulp or sulparex or sulpiphar or sulpir or sulpirid or sulpiryd or sulpitil or sulpivert or sulpril or suprium or synedil or tepavil or tepazepam or valirem or neogamma or zemorcon or zeprid or zymocomb or geodon or zoleptil or Benperidol or anquil or benperidols or frenactil or glianimon or psichoben or Chlorpromazine or aminazins or ancholactil or biscasil or bukatel or chlorazin or chlorpromazin or clordezalin or fenactil or fleksin or kloproman or klorproman or largactil or largatrex or megaphen or Nevropromazine or Plegomazin or Propaphenin or prozil or prozin or repazine or solidon or tardyl or thorazine or zuledin or Flupentixol or depixol or fluanxel or fluanxol or Fluphenazine or anatensol or cardilac or cenilene or dapotum or decafen or decazate or decentan or eutimox or fludecate or flufenazin or lyogen or lyoridin or lyorodin or mirenil or modecate or moditen or omca or pacinol or permitil or prolixin or prolongatum or sevinol or siqualone or fludecasin or Haloperidol or alased or aloperidin or avant or bioperidolo or buteridol or cereen or decaldol or decanoate or dozic or duraperidol or fortunan or haldol or haloneural or haloper or haloperin or norodol or novo‐peridol or peridol or sedaperidol or serenace or serenelfi or sevium or sigaperidol or sylador or vesadol or vesalium or zafrionil or levomepromazine or methotrimeprazine or levium or levomepromazin or levoprome or levozin or neurocil or norzinan or novo‐meprazine or nozinan or sinogan or tisercin or tisercinetta or veractil or Pericyazine or neulactil or perphenazine or decentan or peratsin or perdenasin or terfluoperazine or trilafon or trilafan or triptafen or fentazin or Pimozide or antalon or norofren or orap or pirium or Prochlorperazine or buccastem or bukatel or chlorperazinum or compazine or cotrazine or emetiral or klometil or mitil or nu‐prochlor or prochlorperazin or scripto‐metic or stemetil or tementil or trinigrin or ultrazac or ultrazinol or vertigon or Promazine or prazine or promazin or protactyl or prozyl or sinophenin or sparine or talofen or Thioridazine or aldazine or elperil or flaracantyl or mallorol or mefurine or meleril or mellaril or melleretten or mellerettes or melleril or orsanil or ridazine or stalleril or thioridazin or tirodil or visergil or Trifluperazine or discimer or eskazine or foille or iremo or jatroneural or jatrosom or modalina or oxyperazine or parmodalin or parstelin or sedofren or sporalon or stelabid or stelazine or stelbid or stelium or stilizan or terfluoperazine or terflurazine or terfluzin or terfluzine or trifluoperazin or trisedyl or Pipotiazine or lonseren or piportil or piportyl or Zuclopenthixol or ciatyl or cisordinol or clopixol or sordinol
and
Eating Disorder* or Anorexia Nervosa
Please see Table 2; Table 3; Table 4 for Medline, PsycINFO and EMBASE search strategies
| Search terms |
| Search Strategy: |
| Search terms |
| Search Strategy: |
| Search terms |
| Search Strategy: |
EMBASE
4. Social Sciences and the Science Citation Index will be searched for all included papers.
5. Searches for ongoing and unpublished trials will be conducted at databases such as metaRegister of Controlled Trials (http://www.controlled‐trials.com) and clinicalstudyresults.org database (http://www.clinicalstudyresults.org).
Handsearching
International Journal of Eating Disorders (all issues)
European Eating Disorders Review (all issues)
Reference Lists
The reference lists of relevant articles and book chapters on treatment or pharmacological treatment of AN will be searched.
Personal communication
Personal letters and e‐mail will be sent to the corresponding authors of papers in the field of AN pharmacological treatment in the US, UK, Europe, NZ and Australia. Researchers will be asked for information on unpublished or ongoing trials. A list of all available antipsychotics will be drawn up and the pharmaceutical companies that developed the drugs will be contacted and requested for information on any unpublished or ongoing trials.
Data collection and analysis
Selection of trials
The eligibility of all abstracts identified by the search will be examined by two authors (AMC and JB) working independently. Authorship of studies will not be concealed. Trials under consideration for inclusion will also be independently assessed for their appropriateness and methodological quality. Any unresolved disagreements, will be discussed with a third author (PH, MSL, US or JT). Excluded studies will be listed with reasons for their exclusion. Authors of identified unpublished trials will be contacted.
Assessment of methodological quality
Studies will be independently assessed by the authors (AC and JB) for their methodological quality using the following criteria (Higgins 2006): methods used for generation of the sequence of allocation and adequacy of methods used for their concealment, the blindness of participants, carers and outcome assessors to treatment allocation, description of proportion of participants who did not complete follow‐up (dropouts and withdrawals) for the main outcomes, and analysis by intention‐to‐treat (ITT).
Studies will be ranked by quality of allocation concealment (A, B, or C):
A = adequate;
B = unclear and
C = inadequate.
Data extraction and management
The two authors (AC and JB) will independently extract the data using a standardised data extraction sheet, which will be cross‐checked to ensure reliability. Disagreements will be discussed and the decisions documented. Besides information on the quality of studies described above (assessment of methodological quality), the following data will be extracted from studies: participant characteristics including duration of illness, comorbidities, baseline nutritional status, interventions tested (including combination of treatments offered, length of trial, setting, drug doses) and outcome measures.
Attempts will be made to obtain information of interest that is not included in the published reports (for quality evaluation of trials or results) through letters and e‐mails to corresponding authors of studies. If there is no reply after four weeks, a follow‐up e‐mail will be sent to the corresponding author and another co‐author (if available). In cases where there is no reply after another four weeks, only published data will be considered.
Because of the rarity of trials in this area and methodological problems associated with conducting trials in AN, data on efficacy measures will not be included if the non‐completion rates are 50% or higher in trials. Data on acceptability of treatment (completion rates) will be considered for all included trials.
Measures of treatment effect
Data will be entered into Review Manager 4.2 (RevMan 2003).
Relative risk (RR) analyses will be conducted for dichotomous data. For continuous data, standardised mean difference (SMD) analyses will be conducted for outcomes measured by scales/questionnaires (as different scales are often used to assess psychopathology, to allow for possible heterogeneity in outcome measures); for outcomes such as weight and BMI, weighted mean difference (WMD) analyses will be performed (as the same units of measurement are expected to be used across studies).
Ninety‐five per cent confidence intervals (CI) will be reported for both dichotomous (RR) and continuous (SMD or WMD) variables, based on the random‐effects model (DerSimonian 1986), as this takes into account any differences between studies even if heterogeneity is not statistically significant.
In trials with multiple intervention groups, data from the active treatments will be combined, where possible, for comparison with placebo (procedure described in the Cochrane Handbook for Systematic Reviews of Interventions) (Higgins 2006). Studies with a cross‐over design will be included, but only results of the first phase of treatment will be considered for analyses, if available.
Dealing with missing data
If no information is available (either from the report or from the authors) on missing data, both dichotomous and continuous outcomes data will be considered for analyses as presented in studies (or informed by authors) ‐ "on treatment" analyses, based on the number of participants for whom data is available. However, data of dichotomous outcomes measures will be further tested, in sensitivity analyses, against results analysed according to intention‐to‐treat (ITT) principles ‐ considering all randomised participants in the treatment group to which they have been initially assigned, irrespective of how much treatment they received or if they completed treatment. For the ITT analyses, assumptions will be made based on the "worst result" for the outcome (a negative outcome), such as "participant did not achieve target weight or weight within normal range", and impact in results discussed.
Assessment of heterogeneity
Heterogeneity between studies will be assessed by graphical inspection of results (evaluation of forest plots), and by chi‐square (statistically significant at 10% probability level). In addition, I‐squared tests (I2) for homogeneity will be conducted. I2 test describes the percentage of total variation across studies that is due to heterogeneity rather than chance alone, and a value greater than 50% is considered substantial heterogeneity (Higgins 2003).
If significant heterogeneity is found, possible reasons for heterogeneity will be investigated by stratified analysis of continuous outcomes (when possible) (Deeks 2001), considering the subgroups described under "subgroup analyses" below, and trials with differential methodological quality (adequate randomisation procedure versus inadequate or unclear randomisation procedure). Differences in dichotomous outcomes will be determined by whether the confidence intervals for the effect sizes of the subgroups overlap. If a reason (or more) for heterogeneity is found, the authors will judge if it is reasonable to aggregate data.
Assessment of reporting bias
Wherever the number of trials allow, funnel plots will be used to assess the possibility of publication bias.
Subgroup analyses
Subgroup analyses are planned to compare:
1. Different classes/types of antipsychotics (typical versus atypical drugs), as atypicals may have a different profile of action;
2. Treatment setting(in or outpatient treatment), as drug treatment may have a clearer effect in a less intensive setting of treatment (combined to less interventions);
3. Treatment duration (less than or equal to 3 months or > 3 months), as a minimum period of time to restore weight may be necessary;
4. Age (mean age less than or equal to 18 or > 18 years), as younger people with a less chronic illness may show a better response.
Sensitivity analyses
Sensitivity analyses are planned as follows:
1. To compare "intention to treat principles" to analyse dichotomous data, based on assumptions of "worst case scenario" causes any substantial change in results (compared to "on treatment" analyses based on available data from studies);
2. If heterogeneity among studies is found (by means of stratified analysis if possible, as discussed above).
3. To compare the effect of study quality, those rated A versus B/C.
| Criteria |
| 1. DSM‐IV‐TR Diagnostic criteria for anorexia nervosa: |
| Search terms |
| Search Strategy: |
| Search terms |
| Search Strategy: |
| Search terms |
| Search Strategy: |
