Second-generation antipsychotics for obsessive compulsive disorder (Review)

Transcription

1 Second-generation antipsychotics for obsessive compulsive disorder (Review) Komossa K, Depping AM, Meyer M, Kissling W, Leucht S This is a reprint of a Cochrane review, prepared and maintained by The Cochrane Collaboration and published in The Cochrane Library 2010, Issue 12

2 T A B L E O F C O N T E N T S HEADER ABSTRACT PLAIN LANGUAGE SUMMARY BACKGROUND OBJECTIVES METHODS RESULTS Figure Figure DISCUSSION AUTHORS CONCLUSIONS ACKNOWLEDGEMENTS REFERENCES CHARACTERISTICS OF STUDIES DATA AND ANALYSES Analysis 1.1. Comparison 1 Olanzapine added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 1 No clinically important response to treatment (as defined by the original study) Analysis 1.2. Comparison 1 Olanzapine added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 2 Mental state: 1a. YBOCS score at endpoint Analysis 1.3. Comparison 1 Olanzapine added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 3 Mental state: 1b. No clinically important change - (not 35% YBOCS reduction) Analysis 1.4. Comparison 1 Olanzapine added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 4 Mental state: 1c. No clinically important change - (not 25% YBOCS reduction) Analysis 1.5. Comparison 1 Olanzapine added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 5 Mental state: 2. Anxiety symptoms - HAA endpoint score Analysis 1.6. Comparison 1 Olanzapine added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 6 Mental state: 3. Depressive symptoms - HAD endpoint score Analysis 1.7. Comparison 1 Olanzapine added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 7 Leaving the study early Analysis 1.8. Comparison 1 Olanzapine added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 8 Adverse effects: 1. Extrapyramidal effects Analysis 1.9. Comparison 1 Olanzapine added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 9 Adverse effects: 2. Weight - change from baseline in kg Analysis 2.1. Comparison 2 Quetiapine added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 1 No clinically important response to treatment (as defined by the original study) Analysis 2.2. Comparison 2 Quetiapine added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 2 No clinically important change - CGI Analysis 2.3. Comparison 2 Quetiapine added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 3 Mental state: 1a. YBOCS score at endpoint Analysis 2.4. Comparison 2 Quetiapine added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 4 Mental state: 1b. No clinically important change - (not 35% YBOCS reduction) Analysis 2.5. Comparison 2 Quetiapine added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 5 Mental state: 1c. No clinically important change - (not 25% YBOCS reduction) Analysis 2.6. Comparison 2 Quetiapine added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 6 Mental state: 1d. Anxiety symptoms - HAA score Analysis 2.7. Comparison 2 Quetiapine added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 7 Mental state: 1e. Depressive symptoms Analysis 2.8. Comparison 2 Quetiapine added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 8 Leaving the study early Analysis 2.9. Comparison 2 Quetiapine added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 9 Adverse effects: 1. At least one adverse effect i

3 Analysis Comparison 2 Quetiapine added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 10 Adverse effects: 2. Extrapyramidal effects Analysis Comparison 2 Quetiapine added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 11 Adverse effects: 3a. Significant weight gain (as defined by the authors) Analysis Comparison 2 Quetiapine added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 12 Adverse effects: 3b. Weight - change from baseline in kg Analysis Comparison 2 Quetiapine added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 13 Adverse effects: 4. Sedation Analysis 3.1. Comparison 3 Risperidone added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 1 No clinically important response to treatment (as defined by the original study) Analysis 3.2. Comparison 3 Risperidone added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 2 No clinically important change - CGI Analysis 3.3. Comparison 3 Risperidone added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 3 Mental state: 1a. YBOCS score at endpoint Analysis 3.4. Comparison 3 Risperidone added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 4 Mental state: 1b. No clinically important change - (not 35% YBOCS reduction) Analysis 3.5. Comparison 3 Risperidone added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 5 Mental state: 2. Anxiety symptoms - HAA endpoint score Analysis 3.6. Comparison 3 Risperidone added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 6 Mental state: 3. Depressive symptoms - HAD endpoint score Analysis 3.7. Comparison 3 Risperidone added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 7 Leaving the study early Analysis 3.8. Comparison 3 Risperidone added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 8 Death - any reason Analysis 3.9. Comparison 3 Risperidone added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 9 Adverse effects: 1. At least one adverse effect Analysis Comparison 3 Risperidone added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 10 Adverse effects: 2. Extrapyramidal effects Analysis Comparison 3 Risperidone added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 11 Adverse effects: 3. Significant weight gain - as defined by the authors Analysis Comparison 3 Risperidone added to antidepressants versus placebo added to antidepressants - all data short term, Outcome 12 Adverse effects: 4. Sedation Analysis 4.1. Comparison 4 Risperidone added to antidepressants versus placebo added to antidepressants - all data short term - sensitivity analysis - cross-over excluded, Outcome 1 Leaving the study early HISTORY CONTRIBUTIONS OF AUTHORS DECLARATIONS OF INTEREST SOURCES OF SUPPORT DIFFERENCES BETWEEN PROTOCOL AND REVIEW INDEX TERMS ii

4 [Intervention Review] Second-generation antipsychotics for obsessive compulsive disorder Katja Komossa 1, Anna M Depping 1, Magdalena Meyer 1, Werner Kissling 1, Stefan Leucht 1 1 Klinik und Poliklinik für Psychiatrie und Psychotherapie, Technische Universität München Klinikum rechts der Isar, München, Germany Contact address: Katja Komossa, Klinik und Poliklinik für Psychiatrie und Psychotherapie, Technische Universität München Klinikum rechts der Isar, Moehlstrasse 26, München, 81675, Germany. k.komossa@lrz.tu-muenchen.de. Editorial group: Cochrane Depression, Anxiety and Neurosis Group. Publication status and date: New, published in Issue 12, Review content assessed as up-to-date: 9 November Citation: Komossa K, Depping AM, Meyer M, Kissling W, Leucht S. Second-generation antipsychotics for obsessive compulsive disorder. Cochrane Database of Systematic Reviews 2010, Issue 12. Art. No.: CD DOI: / CD pub2. Background A B S T R A C T Obsessive compulsive disorder (OCD) is a psychiatric disorder which has been shown to affect 2 to 3.5% of people during their lifetimes. Inadequate response occurs in 40% to 60% of people that are prescribed first line pharmaceutical treatments (selective serotonin reuptake inhibitors (SSRIs)). To date not much is known about the efficacy and adverse effects of second-generation antipsychotic drugs (SGAs) in people suffering from OCD. Objectives To evaluate the effects of SGAs (monotherapy or add on) compared with placebo or other forms of pharmaceutical treatment for people with OCD. Search methods The Cochrane Depression, Anxiety and Neurosis Group s controlled trial registers (CCDANCTR-Studies and CCDANCTR-References) were searched up to 21 July The author team ran complementary searches on ClinicalTrials.gov and contacted key authors and drug companies. Selection criteria We included double-blind randomised controlled trials (RCTs) comparing oral SGAs (monotherapy or add on) in adults with other forms of pharmaceutical treatment or placebo in people with primary OCD. Data collection and analysis We extracted data independently. For dichotomous data we calculated the odds ratio (OR) and their 95% confidence intervals () on an intention-to-treat basis based on a random-effects model. For continuous data, we calculated mean differences (MD), again based on a random-effects model. 1

5 Main results We included 11 RCTs with 396 participants on three SGAs. All trials investigated the effects of adding these SGAs to antidepressants (usually SSRIs). The duration of all trials was less than six months. Only 13% of the participants left the trials early. Most trials were limited in terms of quality aspects. Two trials examined olanzapine and found no difference in the primary outcome (response to treatment) and most other efficacyrelated outcomes but it was associated with more weight gain than monotherapy with antidepressants. Quetiapine combined with antidepressants was also not any more efficacious than placebo combined with antidepressants in terms of the primary outcome, but there was a significant superiority in the mean Yale-Brown Obsessive Compulsive Scale (Y-BOCS) score at endpoint (MD -2.28, 95% to -0.52). There were also some beneficial effects of quetiapine in terms of anxiety or depressive symptoms. Risperidone was more efficacious than placebo in terms of the primary outcome (number of participants without a significant response) (OR 0.17, 95% 0.04 to 0.66) and in the reduction of anxiety and depression (MD -7.60, 95% to -2.83). Authors conclusions The available data of the effects of olanzapine in OCD are too limited to draw any conclusions. There is some evidence that adding quetiapine or risperidone to antidepressants increases efficacy, but this must be weighed against less tolerability and limited data. P L A I N L A N G U A G E S U M M A R Y Second-generation antipsychotic drugs for obsessive compulsive disorder This review found some trials comparing the effects of adding second-generation antipsychotic drugs or placebo to antidepressants in obsessive compulsive disorder. There were only 11 trials on three second-generation antipsychotic drugs (olanzapine, quetiapine and risperidone). While not much can be said about olanzapine, quetiapine and risperidone showed some efficacy benefit, but also adverse effects. B A C K G R O U N D Description of the condition Obsessive compulsive disorder (OCD) is a psychiatric disorder which affects approximately 2.3% of Americans over the course of their lifetimes (Ruscio 2008). Figures are similar for Germany where prevalence has been reported to be 2% to 3% (Zaudig 2002). Other authors have reported prevalence between 2% and 3.5%, without indicating specific countries (Maia 2008). The criteria for the disorder include two components: obsessions and compulsions. Obsessions are recurrent thoughts that an individual experiences as intrusive, inappropriate and as a source of distress. Compulsions are repetitive behaviours that afflicted individuals are driven to carry out in a rigid manner, although they are aware of the inappropriateness of their behaviour. Typical compulsions include compulsions to wash or to count. After carrying out the compulsions, distress is briefly relieved, but soon restarts anew. The degree of suffering and disability of people affected by OCD is considerable. The classification of the Diagnostic and Statistical Manual of Diseases, fourth revision (DSIV-R) (APA 1994), considers OCD to be a subset of anxiety disorder, while in the International Classification of Diseases (ICD-10) (WHO 1992) it is described as a separate disorder, and we followed this classification. Furthermore, many other psychiatric conditions are accompanied by obsessive compulsive symptoms (e.g. personality disorders, depression and schizophrenia), but in this review we focused exclusively on trials in participants with a primary diagnosis of OCD. Description of the intervention While selective serotonin reuptake inhibitors (SSRIs) or cognitive behavioural therapy (CBT) are efficacious treatments for OCD, 2

6 the rates of treatment resistance are high. Pallanti 2006 reported that 40% to 60% of people with OCD do not respond sufficiently to treatment with an SSRI. Therefore, there is a need for other agents to alleviate the symptoms of these disorders. Second-generation ( atypical ) antipsychotic drugs (SGAs) were introduced for the treatment of schizophrenia in the 1990s. This drug class includes: amisulpride, aripiprazole, clozapine, olanzapine, quetiapine, paliperidone, risperidone, sertindole, ziprasidone or zotepine. Their main advantage compared with conventional antipsychotic drugs such as haloperidol is that they induce fewer extrapyramidal side effects, including the disabling, often irreversible and stigmatising condition, tardive dyskinesia (Correll 2004). Superior tolerability and potential effects on depressive and negative symptoms make them ideal candidates as sole or adjunctive agents for the treatment of disorders other than schizophrenia. New substances such as iloperidone and asenapine will be considered in the update of this review. In routine clinical practice, SGAs are already frequently being used for depression (Ravindran 2007). There is evidence that 70% of all prescriptions of SGAs are given for indications other than schizophrenia (Sajatovic 2003). The manufacturers of these drugs also aggressively try to expand the indications of their compounds. However, atypical antipsychotic drugs are very costly. For example, the estimated costs of atypical antipsychotics totalled $11.7 billion in the United States in 2005 (Vital Signs 2007). Furthermore, some SGAs are associated with severe adverse events such as weight gain (Allison 1999), extrapyramidal symptoms or agranulocytosis (decrease in the number of certain types of white blood cells). How the intervention might work There is no absolute definition of what an atypical or secondgeneration antipsychotic is. They were initially said to differ from typical antipsychotics in that they did not cause movement disorders (catalepsy) in rats at clinically effective doses (Arnt 1998). SGA drugs block central dopamine receptors, most of them also block serotonin receptors and many other neuroreceptors (Arnt 1998); but how these drugs may work in OCD is as yet unknown. Inadequate response occurs in 40% to 60% of people after first line pharmaceutical treatment with SSRIs (Bloch 2006). In refractory OCD there are data suggesting potential therapeutic benefits for the use of some atypical antipsychotics (Goodwin 2009). Why it is important to do this review Given the frequent off-label use in practice, the uncertain efficacy, the side effects and the high costs of SGA drugs in OCD, a systematic review is important. As far as we know there is one published systematic review on this topic (Bloch 2006), but given the attempts of the pharmaceutical industry to find other indications for their compounds, there is a need for an up-to-date systematic review. Furthermore, the previous review addressed only efficacy, but the efficacy of a drug can only be evaluated in the context of its side effects. Ipser 2006 evaluated the efficacy of pharmacological augmentation strategies in treatment-resistant anxiety disorders, which covered OCD. However, any trials in non treatmentresistant participants were excluded. Ipser also included other additional anxiety disorders and medications, and therefore had a much broader scope. Its primary analysis pooled the results of all antipsychotic drugs whereas in our review the focus is on the effects of the different agents. O B J E C T I V E S To assess the effects and side effects of SGAs alone or as adjunctive therapy compared with other psychopharmacological treatments for people with OCD. M E T H O D S Criteria for considering studies for this review Types of studies We included all double-blind randomised controlled trials, meaning that at least the participants, the assessors and the treatment team had to be blinded (double blind is a conventional definition of this situation in drug trials). Randomised cross-over trials were eligible but only data up to the point of first cross-over were used because of the instability of the problem behaviours and the likely carryover effects of the treatments. We also included cluster randomised trials that met certain criteria (please see Unit of analysis issues). Types of participants We included studies in which people with a primary diagnosis of OCD according to DSIII/DSIV (300.3) (APA 1980; APA 1987; APA 1994) or ICD-10 (F 42) (WHO 1992) aged 18 years and older were included. We did not include children and adolescents as they are the subject of another Cochrane review (Hawkridge 2005). There were no limits in terms of setting. We excluded trials in participants with a primary or secondary diagnosis of another axis I or axis II disorder, if these made up for more than 20% of the participants. We did not exclude any OCD trials in participants with a serious concomitant medical illness. 3

7 Types of interventions SGAs could been given as a monotherapy or as adjunctive therapy compared with placebo or antidepressants. There were no limits in terms of study duration. 1. Experimental treatments included one of the following SGA drugs at any dose and any oral mode of administration: amisulpride, aripiprazole, clozapine, olanzapine, quetiapine, paliperidone, risperidone, sertindole, ziprasidone, zotepine. 2. Comparator substances were either placebo, a benzodiazepine or any of the following antidepressants, all at any dose and oral mode of administration: tricyclic/heterocyclic antidepressants (TCAs), SSRIs (fluoxetine, paroxetine, sertraline, fluvoxamine, citalopram, escitalopram), serotonin norepinephrine reuptake inhibitors (SNRIs) (venlafaxine, duloxetine, milnacipran), monoamine oxidase inhibitors (MAOIs), or newer agents (mirtazapine, bupropion, reboxetine), or St John s Wort. Treatment could have been given either as monotherapy or as augmentation therapy. We excluded treatment with first-generation antipsychotics, which due to their overall higher risk of extrapyramidal side effects are no longer the drugs of choice. We excluded head-to-head comparisons of SGA drugs, unless there was also a concurrent placebo group. Psychological treatment was allowed as long as it was provided to both the treatment and control group. We excluded trials with only non-pharmacological treatments as comparator. There were no limits in terms of duration of treatment. We categorized outcomes as short term (up to six months), medium term (7 to 12 months) and long term (longer than 12 months). Types of outcome measures Primary outcomes Our primary outcome measure was failure to respond to treatment. We operationally defined this as failure to demonstrate a reduction in OCD symptom severity of at least 25% as measured by a validated measures such as the Yale-Brown Obsessive Compulsive Scale (Y-BOCS) (Goodman 1989), or any other scale to rate OCD, or at least up to much improved on the Cinical Global Impressions Scale (CGI) (Guy 1976). We used the criteria applied by the trial authors only if none of these criteria had been indicated. Secondary outcomes 1. Yale-Brown Obsessive Compulsive Scale (Y-BOCS) scores at the end of the trials. 2. Anxiety symptoms were addressed by change on Hamilton Anxiety Rating Scale (HAA) (Hamilton 1959). 3. Depressive symptoms were addressed by the Montgomery Åsberg Depression Rating Scale (MADRS) (Montgomery 1979) or Hamilton Depression Rating Scale (HAD) (Hamilton 1960). 4. Premature trial discontinuation due to any reason, or due to inefficacy of treatment, or due to adverse events. 5. Adverse events (for example, use of anti-parkinson medication, tardive dyskinesia, weight gain, sedation, prolactin increase). 6. Service use - numbers of participants rehospitalised for psychiatric reasons. We classified the outcomes as short term (up to six months), medium term (seven to 12 months) and long term (longer than 12 months). Search methods for identification of studies Electronic searches With the assistance of the Cochrane Collaboration Depression, Anxiety and Neurosis Group s (CCDAN) Trials Search Coordinator (TSC) we searched the Group s Controlled Trials Registers (CCDANCTR-References and CCDANCTR-Studies) up to 21 July The CCDANCTR-References Register contains more than 25,500 references to completed or ongoing trials in depression, anxiety and neurosis, of which 16,500 have been tagged to individually coded trials held in the CCDANCTR-Studies Register. The CCDANCTR-References Register is updated weekly with the results of searches of MEDLINE (1966-), EMBASE (1980-) and PsycINFO (1974-). CENTRAL is searched every three months (with each new release of this database on The Cochrane Library). In addition to this, reports of trials are also identified from international trials registers c/o the World Health Organisation s trials portal (ICTRP) ( drug companies, the hand-searching of key journals, conference proceedings and other (non-cochrane) systematic reviews and meta-analyses. Details of CCDAN s generic search strategies can be found in the Specialized Register section of the Cochrane Depression, Anxiety and Neurosis Group s module text. The search terms used were: ((obsess* or compul* or OCD) and (amisulprid* or aripiprazol* or clozapin* or olanzapin* or quetiapin* or paliperidon* or risperidon* or sertindol* or ziprasidon* or zotepin* or RGH-188 or cariprazine or atypical antipsychotic* or second-generation antipsychotic* or second generation antipsychotic* )) We also searched using search terms for intervention and condition e.g. amisulpride AND OCD, aripiprazole AND OCD. No language restrictions were applied. Searching other resources KK and AMD inspected the references of all identified trials and of previous reviews for more trials. 4

8 KK and SL contacted the first author of each included trial for missing information and for the existence of further trials. KK and SL contacted the manufacturers of atypical antipsychotic drugs and ask them about further relevant trials and for missing information on identified trials. Data collection and analysis Selection of studies Authors KK and AMD independently inspected citations identified from the search. We ordered all potentially relevant reports that we identified for reassessment. Where difficulties or disputes arose we asked authors SL and WK for assistance/adjudication and where it was not possible to decide, we ordered full texts for assessment. We repeated this process for the assessment of the full texts. Where it was not possible to resolve disagreements, we contacted the authors of the papers for clarification. Data extraction and management Authors KK and AMD independently extracted data from included trials. Again, we discussed any disagreement, documented decisions and, if necessary, we contacted authors of trials for clarification. For any remaining problems SL and WK assisted to clarify issues and we documented the final decisions. We extracted data onto standard, simple forms. Assessment of risk of bias in included studies Again, working independently, KK and AMD assessed risk of bias using the tool described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2008). This tool considers how the sequence was generated, how allocation was concealed, the integrity of blinding at outcome, the completeness of outcome data, selective reporting and other potential sources of bias. We did not include trials where sequence generation was assessed as being at a high risk of bias, where allocation was clearly not concealed, and where double blinding had not taken place. If disputes arose as to which category a trial was to be allocated, again we resolved this by discussion, after working with a third review author (SL or WK). Measures of treatment effect 1. Binary data We calculated the odds ratio (OR) and its 95% confidence interval (). 2. Continuous data We calculated the mean differences as it preserves the original units and is therefore easier to interpret. 2.1 Change versus endpoint data We used change data only when endpoint data were not available. 2.2 Skewed data Continuous data on clinical and social outcomes are often not normally distributed. To avoid the pitfall of applying parametric tests to non-parametric data, we applied the following standards to all data before inclusion: (a) data from trials of, for example, at least 200 participants were entered in the analysis irrespective of the following rules, because skewed data pose less of a problem in large trials. (b) Endpoint data: when a scale started from the finite number zero, we subtracted the lowest possible value from the mean, and divided this by the standard deviation. If this value was lower than one, it strongly suggested a skew and we excluded the trial. If this ratio was higher than one but below two, there was a suggestion of skew. We entered the trial and tested whether its inclusion or exclusion would change the results substantially. If the ratio was larger than two we included the trial, because skew is less likely (Altman 1996; Higgins 2008). (c) When continuous data were presented on a scale which included a possibility of negative values (such as change data), it was difficult to tell whether data were skewed or not. We entered the trial, because change data tend to be less skewed and because excluding trials would also lead to bias, as not all the available information would have been used. Unit of analysis issues 1. Cluster trials Trials increasingly employ cluster randomisation (such as randomisation by clinician or practice), but analysis and pooling of clustered data pose problems. Firstly, authors often fail to account for intra-class correlation in clustered trials, leading to a unit of analysis error (Divine 1992) whereby P values are spuriously low, confidence intervals unduly narrow and statistical significance overestimated. This causes type I errors (Bland 1997; Gulliford 1999). Where clustering was not accounted for in primary trials, we presented data in a table, with a (*) symbol to indicate the presence of a probable unit of analysis error. We sought to contact the lead authors of trials to obtain the intra-class correlation co-efficient of their clustered data and adjusted for this by using accepted methods (Gulliford 1999). Where clustering was incorporated, we presented the data as if from a parallel-group randomised trial, but adjusted for the clustering effect. Additionally, we would have excluded such trials in a sensitivity analysis. 5

9 If cluster trials were appropriately analysed taking into account intra-class correlation co-efficient and relevant data documented in the report, synthesis with other trials would be possible using the generic inverse variance technique. Dealing with missing data 1. Missing participants 2. Cross-over trials A major concern of cross-over trials is the potential for a carryover effect. It occurs if an effect (e.g. pharmacological, physiological or psychological) of the treatment in the first phase is carried over to the second phase. As a consequence, on entry to the second phase the participants can differ systematically from their initial state despite a wash-out phase. For the same reason, cross-over trials are not appropriate if the condition of interest is unstable (Elbourne 2002). As both effects are very likely in a depressive disorder, randomised cross-over trials were eligible but only data up to the point of first cross-over. We did not consider data from the following (second) period of the cross-over trial for analysis. We described cross-over trials to be at risk of bias and we excluded cross-over trials in a sensitivity analysis. Dichotomous data We analysed all data on the basis of the intention-to-treat (ITT) principle: dropouts were always included in these analyses. Where participants had withdrawn from the trial before the endpoint, it was assumed their condition would have remained unchanged if they had stayed in the trial. This is conservative for outcomes related to response to treatment (because these participants will be considered to have not responded to treatment). It is not conservative for adverse events, but we think that for the adverse events of interest in our review (see Secondary outcomes) a worst case scenario is clinically unlikely. When there were missing data and the method of last observation carried forward (LOCF) had been used to do an ITT analysis, then we used the LOCF data, with due consideration of the potential bias and uncertainty introduced. We did not perform a worst case and best case ITT analysis. 3. Trials with multiple treatment groups 1. Multiple dose groups We expected that some trials would address the effects of different doses of the same compound compared with the competitor (e.g. substance A 200 mg/day, substance A 400 mg/day and placebo). In the case of dichotomous outcomes we intended to sum up the sample sizes and the number of people with events across both groups. For continuous outcomes we intended to combine means and standard deviations using the methods described in chapter 7 (section ) of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2008). 2. Multiple medications We expected that some other trials would combine several interventions with one comparison group (e.g. a three-arm trial comparing amisulpride, olanzapine and placebo). In this case we intended to analyse the effects of the amisulpride group and the olanzapine group against placebo separately, but would have divided up the total number of participants in the placebo group. In the case of continuous outcomes, the total number of participants in the placebo group would again be divided up, but the means and standard deviations left unchanged (see chapter 16, section in Higgins 2008). Continuous data: Concerning continuous data, the Handbook recommends avoiding imputations of continuous data and suggests rather that the data must be used in the form in which they have been presented by the original authors. Whenever ITT data had been presented by the authors, they were preferred to per protocol/completer datasets. Furthermore, we acknowledge that all methods of imputation to deal with missing data introduce uncertainty about the reliability of the results. This depends on the degree of missingness, the pooled estimate of the treatment effect and the variability of the outcomes. We considered variation in the degree of missing data as a potential source of heterogeneity. 2. Missing statistics When only the standard error (SE) or P values were reported, we calculated standard deviations (SDs) according to Altman ( Altman 1996). In the absence of supplemental data after requests to the authors, we calculated the SDs according to a validated imputation method (Furukawa 2006). We examined the validity of these imputations in the sensitivity analyses, as described in Cipriani Assessment of heterogeneity We assessed heterogeneity on the basis of the Handbook s recommendations (I 2 values of 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, Higgins 2008). In addition to the I 2 value 6

10 (Higgins 2003), we presented the χ 2 and its P value and considered the direction and magnitude of the treatment effects. As in a meta-analysis with few trials, where the χ 2 test is underpowered to detect heterogeneity should it exist, a P value of 0.10 is used as a threshold of statistical significance. See: Characteristics of included studies; Characteristics of excluded studies; Characteristics of ongoing studies. Please see Characteristics of included studies, Characteristics of excluded studies and Characteristics of ongoing studies for more information. Assessment of reporting biases Reporting biases arise when the dissemination of research findings is influenced by the nature and direction of results. These are described in section 10.1 of the Handbook (Higgins 2008). We are aware that funnel plots may be useful in investigating reporting biases but are of limited power to detect small-trial effects. We did not use funnel plots for outcomes where there were 10 or fewer trials, or where all trials were of similar sizes. Data synthesis We employed the random-effects model for analysis (Der- Simonian 1986). We understand that there is no closed argument for the preferred use of fixed-effect or random-effects models. The random-effects method incorporates an assumption that the different trials are estimating different, yet related, intervention effects. This does seem true to us as we are a priori expecting some clinical heterogeneity between the patients in the different trials. Therefore, we chose the random-effects model for all analyses. Subgroup analysis and investigation of heterogeneity We planned a subgroup analysis based on whether or not participants with a history of treatment resistance were included, which may have affected response to treatment. We also analysed whether the response of inpatients (and day hospital patients) differed from that of outpatients. When we found high inconsistency of effects by the measures explained above and clear reasons explaining the heterogeneity were present, we presented the data separately. If not, we commented on the heterogeneity of the data. Sensitivity analysis We conducted the following sensitivity analyses to investigate the degree to which the effect sizes depended on the assumptions made by the reviewers: we excluded cluster trials, cross-over trials, included trials with children, imputed statistics and risk of bias assessment. R E S U L T S Results of the search The search of the CCDAN Controlled Trials Register up until July 2010 yielded 31 reports of which we examined 27. From additional handsearching and searches of clinical trials.gov we identified a further six reports for closer inspection. After excluding 11 trials, we included 17 publications on 11 trials and 3 comparisons: olanzapine and antidepressants versus placebo and antidepressants (two); quetiapine and antidepressants versus placebo and antidepressants (five) and risperidone and antidepressants versus placebo and antidepressants (four). We identified three ongoing trials (Diniz 2008; NIMH 2006; USF ). Included studies The 11 included trials (Bystritsky 2004; Carey 2005; Denys 2004; Erzegovesi 2004; Fineberg 2005; Hollander 2003; Kordon 2008; Li 2004; McDougle 2000; Shapira 2003; Vulink 2009) randomised 396 participants (see Characteristics of included studies for additional information). All of the publications were reported in English. Eight trials were sponsored by pharmaceutical companies and three trials did not report sponsoring. 1. Duration of trials All of the included trials were short-term trials with a duration of 6 to 16 weeks. We did not identify any medium- or long-term trials. 2. Setting One trial reported the setting as inpatient, one included in- and outpatients, four trials reported the setting as outpatient and five trials did not report on the setting. 3. Participants In all trials the participants were diagnosed with OCD according to DSIV. All trials, except for one (Vulink 2009), included participants resistant to previous treatment with, for example, an SSRI. Vulink 2009 also included drug-free participants. The mean ages were in the late thirties. Description of studies 4. Trial size The trial sizes ranged from 16 to 76 participants. 7

11 5. Interventions In all of the trials antipsychotic (olanzapine, quetiapine and risperidone) and placebo treatment was given as adjunctive therapy to antidepressive treatment with mainly SSRIs, but also SNRIs or clomipramine. The doses of olanzapine ranged from 15 mg to 20 mg/day (start dose 2.5 mg) in one trial (Bystritsky 2004) and 5 mg to 10 mg/day (Shapira 2003) in the other trial. Flexible doses of quetiapine ranged from 25 mg to 600 mg/day. Risperidone doses could be given flexibly from 0.5 mg to 3 mg and 1 mg to 6 mg/ day in two trials (Hollander 2003; McDougle 2000); fixed-dose regimens were used in the other two risperidone versus placebo trials (Li mg/day, Erzegovesi mg/day). 6. Outcomes 6.1 Response to treatment We prespecified at least a 25% reduction in Y-BOCS total score from baseline as a relevant cut off to define response or at least much improved on CGI. At least 25% Y-BOCS total score reduction data were presented by Bystritsky 2004 and Fineberg Kordon 2008 described at least 35% Y-BOCS reduction. The authors Denys 2004, Erzegovesi 2004, and Vulink 2009 used at least 35% Y-BOCS reduction and CGI of one or two as response criteria. McDougle 2000 defined at least 35% Y-BOCS reduction and a final score of 16 or less for response. Shapira 2003 used a response criteria of at least 25% Y-BOCS reduction and OCD symptoms of less than moderate CGI severity and a Y-BOCS score < 16. Carey 2005 and Hollander 2003 reported on at least 25% Y-BOCS reduction and CGI of one or two for response. One trial did not predefine response criteria (Li 2004). 6.2 Rating scales Details of scales that provided usable data are shown below. Reasons for exclusion of data from other instruments are given under Outcomes in the Characteristics of included studies section Global state scales Clinical Global Impression Scale - CGI Scale This scale (Guy 1976) is used to assess both the severity of the illness and clinical improvement by comparing the conditions of the person standardised against other people with the same diagnosis. A seven-point scoring system is used with low scores showing decreased severity and/or overall improvement. Four trials reported data on CGI - improvement (Carey 2005; Denys 2004; McDougle 2000; Vulink 2009) Mental state scales Yale-Brown Obsessive Compulsive Scale - Y-BOCS The Y-BOCS (Goodman 1989) is a 21-item rating scale where higher scores indicate a higher degree of symptoms. Nine trials reported on Y-BOCS change or endpoint data Hamilton Anxiety Rating Scale - HAA The Hamilton Anxiety Scale (Hamilton 1959) is a rating scale which consists of 14 items. Each item is rated on a five-point scale, ranging from zero (not present) to four (severe). Four trials reported on the Hamilton Anxiety Rating Scale (Bystritsky 2004; Denys 2004; McDougle 2000; Vulink 2009) Montgomery Åsberg Depression Rating Scale - MADRS The MADRS (Montgomery 1979) is a 10-item rating scale, where symptoms are assessed in an interview. Scores can range from 0 to 60, where higher scores indicate a higher degree of depression. One trial provided data on MADRS score (Carey 2005) Hamilton Depression Rating Scale - HAD The HAD (Hamilton 1960) is a well-established 17-item scale for the measurement of depression and is sensitive to change. Five trials provided data on HAD (Bystritsky 2004; Denys 2004; Kordon 2008; McDougle 2000; Vulink 2009). 6.3 Leaving the trial early All trials reported data on participants leaving the trial early. 6.4 Adverse effects Adverse effects were mainly recorded in open interviews. Continuous data were provided for weight in terms of change from baseline. Excluded studies We excluded 11 out of 22 trials of potential interest for the following reasons: two trials were not randomised, two were singleblinded trials, six were open-label trials, and one used another intervention. 8

12 Awaiting assessment We did not categorise any study as awaiting classification. Ongoing studies We identified three ongoing trials (see Characteristics of ongoing studies table). Risk of bias in included studies For details, please refer to the risk of bias tables (Figure 1; Figure 2). Figure 1. Methodological quality graph: review authors judgements about each methodological quality item presented as percentages across all included studies. 9

13 Figure 2. Methodological quality summary: review authors judgements about each methodological quality item for each included study. 10

14 Allocation Three trials (Bystritsky 2004; Carey 2005; Erzegovesi 2004) described the method of allocation concealment. Four trials gave a few details on the randomisation method (Bystritsky 2004; Carey 2005; Erzegovesi 2004; McDougle 2000). For the other trials information was so limited that in seven included trials it remained unclear whether there was a risk of bias. Blinding All of the included trials were double-blinded. Denys 2004, Erzegovesi 2004 and Vulink 2009 described using identical capsules for blinding. The other trials did not provide any information on the blinding procedure. No trial examined whether blinding was effective. We found that the side-effect profiles of the examined compounds are quite different, which may have made blinding difficult. We therefore conclude that the risk of bias for objective outcomes (e.g. weight change or laboratory values) was low but there was a risk of bias for subjective outcomes. Other potential sources of bias In one trial (Li 2004), due to the cross-over design, only very short-term data (two weeks) could be evaluated. Seven of the ten included trials were sponsored by the manufacturer of the antipsychotic drug used (Bystritsky 2004; Carey 2005; Denys 2004; Fineberg 2005; Hollander 2003; Li 2004; Vulink 2009). There is evidence that pharmaceutical companies sometimes highlight the benefits of their compounds and tend to suppress their disadvantages (Heres 2006); therefore, risk of bias in these trials was considered as unclear. One trial did not report on the sponsoring (Erzegovesi 2004) and two trials (McDougle 2000; Shapira 2003) had neutral sponsors. Effects of interventions 1. Comparison 1. Olanzapine and antidepressants versus placebo and antidepressants Incomplete outcome data Overall, the number of participants leaving the trial early was low to moderate. In one trial the overall rate of participants leaving the trial early was 30.7% (Bystritsky 2004). Two trials reported an attrition rate of 25% (Kordon 2008; Li 2004). Vulink 2009 reported an attrition rate of 20.5% for the quetiapine group and 5.4% for the placebo group. Hollander 2003 and Shapira 2003 reported a rate of leaving the trial early between 10% and 20%. In the other trials the attrition lay below 10% (Carey 2005; Denys 2004; Fineberg 2005; McDougle 2000). All but two trials applied the last-observation-carried-forward (LOCF) method to account for participants leaving the trial early, which is an imperfect method. It assumes that a participant who left the trial prematurely would not have had a change of his condition if he had stayed in the trial. This assumption can obviously be wrong. Mixed effects models analysis was additionally applied in one trial (Li 2004). McDougle 2000 analysed the trial completers. Selective reporting In terms of selective reporting there was a high risk of bias in all of the included trials. Five trials did not fully report on primary outcome data (Bystritsky 2004; Erzegovesi 2004; Fineberg 2005; Hollander 2003; Shapira 2003). Nine trials did not fully address treatment associated side effects (Bystritsky 2004; Denys 2004; Erzegovesi 2004; Fineberg 2005; Hollander 2003; Kordon 2008; Li 2004; McDougle 2000; Shapira 2003) and one trial (Vulink 2009 reported only on the most prevalent side effects. 1.1 No clinically important response to treatment (as defined by the original trial) Analysis 1.1 There was no significant difference (n = 70, two RCTs, odds ratio (OR) 0.28; 95% confidence interval () 0.01 to 6.45), but the I² value of 74% and the significant heterogeneity test (P = 0.05) indicated a substantial heterogeneity. Bystritsky 2004 (n = 26, OR 0.04; 95% 0.00 to 0.87) indicated a benefit for olanzapine, but Shapira 2003 did not (n = 44, OR 1.00; 95% 0.30 to 3.33). The mean doses of olanzapine in Bystritsky 2004 were much higher (11.2 mg/day, almost double) than the mean doses in Shapira 2003 (6.1 mg/day) which might have contributed to the heterogeneity of data. 1.2 Mental state Analysis Y-BOCS - score at endpoint There was no significant difference (n = 70, two RCTs, mean difference (MD) -2.96; 95% to 1.22) No clinically important change - (less than 35% YBOCS reduction) Analysis 1.3 There was no significant difference (n = 44, one RCT, OR 0.76; 95% 0.17 to 3.29). 11

15 1.2.3 No clinically important change - (less than 25% YBOCS reduction) Analysis 1.4 There was no significant difference (n = 70, two RCTs, OR 0.28; 95% 0.01 to 6.45), but the I² value of 74% and a significant heterogeneity test Chi² 3.90 (P = 0.05) indicated a substantial heterogeneity. Bystritsky 2004 (n = 26, OR 0.04; 95% 0.00 to 0.87) found olanzapine to be superior while Shapira 2003 (n = 44, OR 1.00; 95% 0.30 to 3.33) did not Anxiety symptoms - HAA endpoint Analysis 1.5 There was no significant difference (n = 26, one RCT, MD -5.80; 95% to 0.62), but there is suggestion of skew Depressive symptoms - HAD endpoint Analysis 1.6 There was no significant difference (n = 26, one RCT, MD -0.90; 95% to 4.11), but there is suggestion of skew. 1.3 Leaving the trial early Analysis Due to any reason There was no significant difference (n = 70, two RCTs, OR 0.80; 95% 0.06 to 10.57), but the I² value of 75% and a Chi ² of 4.04 (P = 0.04) indicated a substantial heterogeneity. Bystritsky 2004, (n = 26, OR 0.21; 95% 0.03 to 1.36) indicated a tendency but no significant difference to favour olanzapine, whereas Shapira 2003 (n = 44, OR 2.94; 95% 0.50 to 17.14) showed a tendency, but no significant difference to favour the placebo group. As these are the same trials as in outcome 1.1, differences in doses might have contributed to the heterogeneity of data Due to adverse effects There was no significant difference (n = 70, two RCTs, OR 1.70; 95% 0.30 to 9.50) Extrapyramidal effects (EPS) In Bystritsky 2004 no participant suffered from EPS Weight - change from baseline in kg Analysis 1.9 Olanzapine produced significantly more weight gain than placebo (n = 44, one RCT, OR 2.30; 95% 0.80 to 3.80). 1.5 Publication bias Due to the small number of included trials, we did not perform a funnel plot analysis. 1.6 Subgroup analysis and investigation of heterogeneity There was heterogeneity in response data and data on leaving the trial early. Although not entirely clear, a possible reason for the heterogeneity is the difference in olanzapine doses between Bystritsky 2004 (11.2 mg/day, almost double) and Shapira 2003 (6.1 mg/ day). The reasons for the preplanned subgroup analyses did not apply, because both trials were conducted in non-responders and because the setting was not clear. 1.7 Sensitivity analysis The reasons given for the preplanned sensitivity analyses did not apply and therefore they were not undertaken. 2. Comparison 2. Quetiapine and antidepressants versus placebo and antidepressants 2.1 Non-response to treatment Analysis No clinically important response to treatment (as defined by the original trial) There was no significant difference between the addition of quetiapine or placebo to antidepressants (n = 219, five RCTs, OR 0.53; 95% 0.27 to 1.05) Due to inefficacy There was a significant difference in favour of olanzapine (n = 70, two RCTs, OR 0.10; 95% 0.01 to 0.98, P = 0.05). 1.4 Adverse effects Analysis Global state - no clinically important change Analysis 2.2 There was no significant difference (n = 198, four RCTs, OR 0.86; 95% 0.35 to 2.10) but the I² value of 49% indicated some heterogeneity. Carey 2005 (n = 42, OR 1.78; 95% 0.52 to 6.04) tended to favour the placebo group, whereas Denys 2004 (n = 40, OR 0.17; 95% 0.03 to 0.92) and Vulink 2009 (n = 12

16 76, OR 0.73; 95% 0.29 to 1.85) tended to favour quetiapine. Clear reasons explaining the heterogeneity were not identified. 2.2 Mental state Analysis Y-BOCS score at endpoint There was a significant difference in favour of quetiapine (n = 209, five RCTs, MD -2.28; 95% to -0.52) No clinically important change - (less than 35% YBOCS reduction) Analysis 2.4 There was a significant difference in favour of quetiapine (n = 80, two RCTs, OR 0.27; 95% 0.09 to 0.87) No clinically important change - (less than 25% YBOCS reduction) Analysis 2.5 There was no significant difference (n = 103, three RCTs, OR 0.37; 95% 0.06 to 2.46), but the data were again heterogeneous with an I² of 72% and a significant heterogeneity statistic (P = 0.03). Again, Carey 2005 tended to favour placebo augmentation (n = 42, one RCT, OR 1.48; 95% 0.43 to 5.05), while Denys 2004 (n = 40, OR 0.09; 95% 0.02 to 0.50) and Fineberg 2005 (n = 21, one RCT, OR 0.30; 95% 0.03 to 3.45) showed a superiority of quetiapine. Clear reasons explaining the heterogeneity were not identified Anxiety symptoms - HAA score at endpoint Analysis 2.6 There was a significant difference in favour of quetiapine (n = 109, two RCTs, MD -2.42; 95% to -0.28), but there is suggestion of skew in both included trials Depressive symptoms - score at endpoint Analysis HAD There was no significant difference (n = 147, three RCTs, MD ; 95% to 0.10), but there is suggestion of skew in both included trials. 2.3 Leaving the trial early Analysis Due to any reason Significantly more participants in the quetiapine group than in the placebo group left the trials prematurely due to various reasons (n = 219, five RCTs, OR 3.38; 95% 1.32 to 8.67) Due to adverse effects Significantly more participants in the quetiapine group than in the placebo group left the trials prematurely due to adverse effects (n = 219, five RCTs, OR 4.48; 95% 1.43 to 14.04) Due to inefficacy There was no significant difference (n = 219, five RCTs, OR 0.95; 95% 0.09 to 10.36). 2.4 Adverse effects Analysis At least one adverse effect There was a significant difference in favour of placebo (n = 39, one RCT, OR 32.22; 95% 1.71 to ) Extrapyramidal effects - tremor Analysis 2.10 There was no significant difference (n = 76, one RCT, OR 0.49; 95% 0.16 to 1.52) Weight Analysis MADRS There was no significant difference (n = 62, two RCTs, MD 0.22; 95% to 3.31), but there is suggestion of skew in both included trials Significant weight gain Quetiapine produced significantly more weight gain than placebo (n = 117, two RCTs, OR 4.14; 95% 1.59 to 10.81). 13

17 Weight - change from baseline in kg Analysis 2.12 There was a significant difference in favour of placebo (n = 76, one RCT, MD 3.40; 95% 2.15 to 4.65) Global state - no clinically important change Analysis 3.2 There was a significant difference in favour of risperidone (n = 36, one RCT, OR 0.04; 95% 0.00 to 0.85) Sedation Analysis 2.13 Quetiapine produced significantly more sedation than placebo (n = 196, four RCTs, OR 5.91; 95% 2.87 to 12.18). 2.5 Publication bias Due to the small number of included trials we did not perform a funnel plot analysis. 2.6 Subgroup analysis and investigation of heterogeneity There was some degree of heterogeneity concerning response to treatment, but we did not identify clear reasons explaining this heterogeneity. As the number of trials and participants is still low, some of this heterogeneity may be due to chance alone. Four of the included studies (Carey 2005; Denys 2004; Fineberg 2005; Kordon 2008) reported data on treatment-resistant participants and one study (Vulink 2009) reported data on non treatmentresistant participants. Due to our decision to use a random-effects model we did not use the test for differences between subgroups, but we presented the results for each study separately Carey 2005 (n = 42, one RCT, OR 1.21; 95% 0.36 to 4.09), Denys 2004 (n = 40, one RCT, OR 0.17; 95% 0.03 to 0.92), Fineberg 2005 (n = 21, one RCT, OR 0.30; 95% 0.03 to 3.45), Kordon 2008 (n = 40, one RCT, OR 0.41; 95% 0.09 to 1.95) and Vulink 2009 (n = 76, one RCT, OR 0.53; 95% 0.21 to 1.31). 2.7 Sensitivity analysis The reasons given for the preplanned sensitivity analyses did not apply and were therefore not undertaken. 3. Comparison 3. Risperidone and antidepressants versus placebo and antidepressants 3.2 Mental state Y-BOCS - score at endpoint Analysis 3.3 There was no significant difference in favour of risperidone (n = 91, three RCTs, MD -3.35; 95% to 1.55), but the data were heterogeneous with an I² of 59% and a significant heterogeneity statistic (P = 0.09). In contrast to the other two studies, Erzegovesi 2004 described data on a mixed sample of responders and nonresponders to former antidepressant treatment (see Analysis 3.7), but clear reasons explaining the heterogeneity were not identified No clinically important change - (less than 35% Y- BOCS reduction) Analysis 3.4 There was no significant difference (n = 40, one RCT, OR 0.33; 95% 0.06 to 1.97) Anxiety symptoms - HAA endpoint score Analysis 3.5 There was a significant difference in favour of risperidone (n = 36, one RCT, MD -6.30; 95% to -2.43) Depressive symptoms - HAD endpoint score Analysis 3.6 There was a significant difference in favour of risperidone (n = 36, one RCT, MD -7.60; 95% to -2.83), but there is suggestion of skew. 3.3 Leaving the trial early Analysis Non-response to treatment Analysis Due to any reason There was no significant difference (n = 103, four RCTs, OR 0.71; 95% 0.19 to 2.72) No clinically important response to treatment (as defined by the original trial) There was a significant difference in favour of risperidone (n = 92, three RCTs, OR 0.17; 95% 0.04 to 0.66) Due to adverse effects More participants in the risperidone group left the trials prematurely due to adverse effects (n = 103, four RCTs, OR 2.54; 95% 0.10 to 66.59). 14

18 3.3.3 Due to inefficacy There was no significant difference (n = 103, four RCTs, OR 0.22; 95% 0.02 to 3.22). 3.6 Publication bias Due to the small number of included trials we did not carry out a funnel plot analysis. 3.4 Death - any reason One trial reported on this outcome and there were no deaths (Erzegovesi 2004). 3.5 Adverse effects At least one adverse effect Analysis 3.9 There was no significant difference (n = 91, three RCTs, OR 2.21; 95% 0.30 to 16.26) Extrapyramidal effects Analysis Use of antiparkinson medication McDougle 2000 reported that antiparkinson medication was not used in either group. 3.7 Subgroup analysis and investigation of heterogeneity The data were rather homogeneous. Two of the included studies (Hollander 2003; McDougle 2000) reported data on treatmentresistant participants and one study (Erzegovesi 2004) described data on a mixed sample of responders and non-responders to former antidepressant treatment. Due to our decision to use a random-effects model we did not use the test for differences between subgroups, but we presented the results for each study separately Hollander 2003 (n = 16, one RCT, OR 0.11; 95% 0.00 to 2.51): McDougle 2000 (n = 36, one RCT, OR 0.04; 95% 0.00 to 0.70) and Erzegovesi 2004 (n = 40, one RCT, OR 0.33; 95% 0.06 to 1.97). A subgroup analysis of the primary outcome, response to treatment, on the basis of whether in- or outpatients were included was not done, because one of the two relevant trials did not report on setting (McDougle 2000). 3.8 Sensitivity analysis Excluding the trial Li 2004 for the reason of cross-over design did not change the results to a marked extent. D I S C U S S I O N Extrapyramidal symptoms Two trials (Erzegovesi 2004; Hollander 2003) described that extrapyramidal symptoms did not occur in either group Abnormal movement/acute dystonia Li 2004 reported on no occurrence of abnormal movement during risperidone or placebo treatment Significant weight gain - as defined by the authors Analysis 3.11 There was no occurrence of significant weight gain reported by one trial (Erzegovesi 2004) Sedation Analysis 3.12 Significantly more participants in the risperidone groups felt sedated (n = 91, three RCTs, OR 7.35; 95% 2.07 to 26.11). Summary of main results This review is based on the data from 11 short-term trials with 396 participants. We only identified randomised controlled trials on three SGA drugs - olanzapine, quetiapine and risperidone. All trials examined the effects of antipsychotics as an adjunct to antidepressants in obsessive compulsive disorder. Olanzapine Only two trials with 70 participants compared olanzapine and antidepressants to placebo and antidepressants in OCD. Overall, there was no significant difference in terms of the primary outcome, response to treatment, nor in any other efficacy-related outcome such as mean Y-BOCS score at endpoint, anxiety, depressive symptoms or leaving the trials early due to inefficacy. In addition, some of these results were significantly heterogeneous, because the results of Bystritsky 2004 found olanzapine to be superior while Shapira 2003 did not. Almost twice as high doses in the Bystritsky 2004 trial are the most likely explanation for this heterogeneity. Future trials may therefore consider that low-dose strategies of 15

19 olanzapine may not be effective. In addition, Bystritsky 2004 was sponsored by the manufacturers of olanzapine, whereas Shapira 2003 mentions a neutral sponsorship, but this alone is not a strong reason to explain heterogeneity. Overall 21% of participants left the trial early due to any reason, with similar percentages in the olanzapine (20%) and placebo (23%) groups. This suggests similar acceptability of olanzapine and placebo augmentation for people with OCD. The reporting on tolerability was very limited. The few available data revealed no difference in overall tolerability (leaving the trial early due to adverse events) and extrapyramidal side effects, but they did confirm the well-known weight increasing effects of olanzapine which was considerable (2.3 kg within six weeks). In summary, there are insufficient data to suggest any conclusion about olanzapine s effects in this condition, and more trials are clearly needed. Quetiapine Five trials with only 219 participants compared adjunctive quetiapine in mean doses between 200 and 600 mg/day with placebo adjunctive to antidepressants in obsessive compulsive disorder. Overall, there was no significant difference in terms of the primary outcome, response to treatment. However, the Y-BOCS score at endpoint and at least a 35% Y-BOCS score reduction indicated a benefit for quetiapine. There were also some possible beneficial effects of quetiapine on anxiety; however, the results may have been skewed, limiting interpretation. Data on leaving the trial early due to any reason (18% in the quetiapine group and 6% in the placebo group) and due to adverse events favoured the placebo group. This suggests that the overall acceptability (leaving early due to any reason) and the overall tolerability (leaving early due to adverse events) of adding quetiapine to antidepressants may be worse than in monotherapy with antidepressants. Side effects were incompletely reported. Nevertheless, there are some data suggesting that adding quetiapine leads to more sedation and more weight gain (3.4 kg within 10 weeks of treatment). Based on very limited data, response to quetiapine did not differ depending on whether treatment-resistant people (four RCTs) were included or not (one RCT). In summary, there might be an efficacy benefit of quetiapine augmentation which, however, needs to be weighed against additional side effects in terms of sedation and weight gain. More, particularly longer-term, trials are needed to draw firm conclusions. of anxiety and depression. The latter data were, however, possibly skewed, limiting interpretation. There was no difference in terms of leaving the trial early for any reason (risperidone 14%, placebo 18%), suggesting that the overall acceptability of antidepressant monotherapy and augmentation with risperidone may be similar. Again, side effects were insufficiently reported. For example, only one trial reported on weight gain. Nevertheless, based on limited data there was no difference in the occurrence of EPS in these trials that used relatively low average risperidone doses between 0.5 mg/day and 2.2 mg/day. The interpretation of our subgroup and sensitivity analyses is very limited due to the small number of trials and participants included. In summary, low-dose risperidone added to antidepressants may be efficacious, but seems to be associated with more sedation. The available amount of data are small and therefore not robust. More trials, in particular ones that improve reporting on side effects, are needed to draw clearer conclusions. Overall completeness and applicability of evidence We searched for 10 different SGAs in the treatment (monotherapy or augmentation) of OCD compared with placebo, antidepressants or benzodiazepines. We found data on only three different SGAs compared with placebo added to antidepressants. We did not find any comparisons between SGAs and antidepressants as a monotherapy. Therefore the evidence has to be considered as incomplete. Quality of the evidence All trials were randomised and double-blinded, but the details of the methods used were rarely presented. Therefore it is unclear whether randomisation and blinding were really appropriately carried out. All trials provided short-term data. In the largest comparison (quetiapine versus placebo), we could evaluate data on no more than 219 participants. Additionally, the reporting on side effects was very limited. In particular, reporting on EPS could have been more detailed. Overall the review authors consider the quality of the evidence to be rather low. Risperidone Risperidone plus antidepressants for OCD was compared with placebo and antidepressants in four trials with only 103 participants. The primary outcome was response to treatment, but no data on the mean Y-BOCS score indicated a benefit for risperidone augmentation. There were superior effects of risperidone in terms Potential biases in the review process We predefined some adverse effects outcomes which we considered to be most important, but there are also other side effects of antipsychotic drugs which we did not address. Furthermore, although our search strategy was quite comprehensive, a possible publication bias cannot be excluded with certainty. 16

20 Agreements and disagreements with other studies or reviews An earlier systematic review (Bloch 2006) pooled the results of all first-generation and second-generation antipsychotics and found a significant efficacy benefit of antipsychotics compared with placebo in OCD. There are two differences compared to our approach: first, we only included SGAs. Secondly, we addressed each SGA drug separately. We think that these compounds are very different, especially in terms of side effects questioning the appropriateness of pooling them. Therefore, our results are less clear cut, but some efficacy benefit was found for quetiapine and risperidone. In contrast to our review, side effects were not addressed in Bloch 2006, limiting its conclusions. One limitation of this review is the lack of SGA treatment compared with psychotherapy for treatment-resistant OCD. 2. For managers/policy makers The evidence for the use of SGA augmentation therapy for OCD is very limited, bearing in mind that some of the SGAs are still rather expensive compounds. Data that are relevant for policy makers such as service use due to hospitalisation were not available. We are thus unable to make any recommendations for decision makers. Implications for research 1. General A U T H O R S C O N C L U S I O N S Implications for practice 1. For clinicians and patients First of all, it is important for clinicians and patients to know that the evidence on the effects of SGA drugs in OCD is very limited. All the available evidence relates to the effects of SGAs added to antidepressants; nothing is known on monotherapy with these antipsychotic drugs. Too few data are available to make any judgement of the effects of olanzapine in OCD and more trials are clearly needed. Somewhat more, but also very limited data suggest a benefit of quetiapine augmentation in some aspects of efficacy, but this must be weighed with worse overall tolerability, more sedation and more weight gain compared with monotherapy using antidepressants. Similarly, adding risperidone to antidepressants was found to be efficacious, but this led to a decrease of overall tolerability and more sedation. Differences in side effects were too poorly described for a comprehensive assessment. On the basis of these findings, no strong recommendations for the use of SGAs in the treatment of OCD can be given. The outcome reporting in trials assessing the effects of SGA drugs in OCD is insufficient, especially the reporting on side effects. Strict adherence to the CONSORT statement (Moher 2001) would make such trials much more informative. 2. Specific There is plenty of room and need for further trials on the effects of SGA drugs in OCD. Trials on SGAs other than olanzapine, quetiapine and risperidone are needed. There is also a lack of longterm trials for this indication. Furthermore, trials investigating the effects of monotherapy with these compounds rather than as augmentation strategies would be of interest. A C K N O W L E D G E M E N T S We would like to thank all members of the Cochrane Depression, Anxiety and Neurosis editorial team for their assistance. We would also like to thank C. Li for his help with the Chinese trials. We thank C. McDougle, S. Erzegovesi, N. Fineberg, N. Shapira and the pharmaceutical companies Astra Zeneca, Pfizer and Sanofiaventis for providing additional information. 17