Antidepressant Controlled Trial For Negative Symptoms In Schizophrenia (ACTIONS): a double-blind, placebo-controlled, randomised clinical trial

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 07/83/01. The contractual start date was in March 2010. The draft report began editorial review in September 2015 and was accepted for publication in December 2015. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

Thomas RE Barnes has received honoraria from Roche for speaking at educational meetings and has been a member of scientific advisory boards for Sunovion Pharmaceuticals Inc. and Otsuka Pharmaceutical Company Ltd/H. Lundbeck A/S in relation to antipsychotic medication. Carol Paton has undertaken consultancy work for Sunovion Pharmaceuticals Ltd and Eli Lilly and Company in relation to antipsychotic medication. Tom KJ Craig has received honoraria from Eli Lilly and Company and Sanofi for speaking at educational meetings. Patrick Keown has received an honorarium for speaking at an educational meeting and has been a member of a scientific advisory board for Otsuka Pharmaceutical Company Ltd/H. Lundbeck A/S in relation to antipsychotic medication, and has received support from Janssen Pharmaceutica to attend a conference. David Baldwin has received research funding from Pfizer Inc. and undertaken consultancy work for H. Lundbeck A/S. Glyn Lewis reports membership of Efficacy and Mechanism Evaluation board. Shôn Lewis reports a potential spin-out company, Clintouch.

Copyright statement

© Queen’s Printer and Controller of HMSO 2016. This work was produced by Barnes et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

Negative symptoms of schizophrenia

The negative symptoms of schizophrenia represent an important dimension of psychopathology and reflect the absence or diminution of normal behaviours and functions. They include deficiencies in emotional responsiveness (blunted or flat affect), poverty of speech, poor rapport, emotional and social withdrawal, anhedonia, apathy and avolition. Such phenomena have been recognised as part of a schizophrenic illness since early in the last century. Kraepelin1 described the more severe dissolution of personality seen in some people with schizophrenia, referring to ‘weakening of those emotional activities which permanently form the mainspring of volition’ (p. 74), with the process resulting in ‘emotional dullness, failure of mental activity, loss of mastery over volition, of endeavour, and of ability for independent action’ (p. 75). Bleuler2 noted that some patients ‘appear lazy and negligent because they no longer have the urge to do anything either of their own initiative or at the bidding of another’ (p. 70).

The introduction of antipsychotic medication in the 1950s revolutionised the treatment of schizophrenia. But the efficacy of these drugs in alleviating the positive symptoms of this illness tended to focus attention on such symptoms as the main treatment target, and it was not until 25 years or so later that attention once again turned to the role of negative or deficit symptoms. Studies have consistently reported worse functional outcomes in individuals with more prominent negative symptoms,3–5 demonstrating correlations between negative symptoms and impairments in occupational and social functioning in the community, as well as a reduced likelihood of living independently. 6,7 Cognitive dysfunction has also been repeatedly demonstrated to contribute to poor functional outcomes in schizophrenia. 8,9 While a degree of overlap between cognitive dysfunction and negative symptoms has been identified,10 they are generally conceptualised as separate domains. 11

Persistent negative symptoms are held to account for much of the long-term morbidity and poor functional outcome of patients with schizophrenia. In this regard, an important clinical distinction is between primary negative symptoms, which comprise an enduring deficit state, predict a poor prognosis and are stable over time, and secondary negative symptoms, which are consequent upon positive psychotic symptoms, depression or demoralisation, or medication side effects, such as bradykinesia as part of drug-induced parkinsonism. Secondary negative symptoms tend to be more common during acute psychotic episodes and would be expected to respond to treatment of the underlying cause. Around half to three-quarters of people with chronic schizophrenia objectively exhibit some negative symptoms,12 although the figure for persistent, primary negative symptoms is probably 15–20%. 13,14

While most research on negative symptoms assumes they constitute a single syndrome or therapeutic target, factor analysis has consistently generated two distinct domains. 15 The first may be conceptualised as expressive deficits: this comprises ‘alogia’ as a decrease in verbal output or verbal expressiveness and ‘flattened or blunted affect’ assessed by diminished facial emotional expression, poor eye contact, decreased spontaneous movement and lack of spontaneity. The second may be characterised as avolition/amotivation for daily life and social activities: this comprises ‘avolition’ as a subjective reduction in interests, desires and goals, and a behavioural reduction of self-initiated and purposeful acts, and ‘asociality’, manifest as a lack of self-initiated social interactions. Some would include ‘anhedonia’ in this domain. Anhedonia has been considered to reflect a diminished capacity for pleasure in all psychiatric and neurological conditions in which it is manifest, but recent, plausible arguments suggest that in schizophrenia it may reflect complex, abnormal psychological processes, such as low-pleasure beliefs and a reduction in the normal tendency to overestimate past and future pleasure, as well as dysfunctional behavioural processes such as reduced pleasure-seeking behaviour. 16,17 Cognitive impairments may play a role in these psychological and behavioural components of anhedonia.

Pharmacological treatment of negative symptoms

The introduction of the second-generation antipsychotics was accompanied by claims that even primary negative symptoms might respond to this new group of antipsychotics, but evidence for an effect independent of an improvement in positive symptoms or medication side effects remains limited,18,19 and recent effectiveness data20,21 provide little support for a robust effect in this regard, with the possible exception of clozapine. 22–24

Antipsychotic medications aside, there are currently no approved pharmacological treatments for negative symptoms of schizophrenia. A range of potential pharmacological interventions has been tested in small studies, usually as augmentation of antipsychotic medication. The drugs used include antidepressants, principally selective serotonin reuptake inhibitors (SSRIs) and mirtazapine,25,26 dehydroepiandrosterone, selegiline hydrochloride (a selective monoamine oxidase inhibitor), cholinesterase inhibitors, omega-3 fatty acids27 and Ginkgo biloba extracts. In addition, a range of glutamatergic agents has been tried, such as glycine, D-cycloserine and glutathione, as well as alpha-7 nicotinic agonists, modulators of metabotropic glutamate receptors, and minocycline – an antibiotic that may exert differential control over NMDA (N-methyl-D-aspartate) receptor signalling.

Prevalence of the use of antidepressants with antipsychotics in schizophrenia

In current clinical practice, antidepressants are commonly used in combination with antipsychotics. For example, audit data from the UK Prescribing Observatory for Mental Health revealed that of 3885 community patients with schizophrenia or related disorder (International Classification of Diseases, Tenth Revision,28 diagnostic category F20–29) who were prescribed depot/long-acting antipsychotic injections in the community, 640 (16.5%) were also prescribed an antidepressant. 29 Furthermore, of 1502 patients in the same diagnostic category under the care of forensic services, 287 (19%) were also prescribed an antidepressant. 30 The clinical indications for such a combination are likely to be comorbid depression or negative symptoms, but the magnitude of use specifically to treat the latter is not known.

Efficacy of antidepressant augmentation for negative symptoms

Currently available treatments for negative symptoms have only modest benefits, with the result that negative symptoms continue to disproportionately limit patient recovery. 19,31,32 The notion that adjunctive antidepressant medication may treat the negative symptoms of schizophrenia has been mooted for almost 20 years. 33–35 Clinical trials have provided evidence that the combined administration of an antipsychotic drug and an ‘add-on’ SSRI – such as fluoxetine, fluvoxamine, paroxetine or citalopram – can improve negative symptoms and some affective disorders associated with schizophrenia, without exacerbating extrapyramidal side effects, in patients in whom such problems have proved persistent and who have been, being unresponsive to antipsychotic monotherapy. 36–41 For example, the efficacy and safety of paroxetine augmentation has been demonstrated in small pilot studies42 and a double-blind placebo-controlled study. 41 Fluvoxamine and fluoxetine have also shown some beneficial effects in placebo-controlled, double-blind studies,36,38,43 in particular, fluvoxamine improved affective blunting. Fluoxetine in combination with a depot antipsychotic caused an overall improvement in negative symptoms,38 although it failed to do so in combination with clozapine. 31

Such benefit may not be limited to SSRI antidepressants: one double-blind, placebo-controlled trial tested the augmentation of clozapine with mirtazapine (which has antagonist properties at serotonin receptors 5-HT2A, 5-HT3 and alpha-2 adrenergic receptors as well as an indirect 5-HT1A agonist effect) for the treatment of negative symptoms of schizophrenia. 44 A significant reduction in negative symptom scores in the mirtazapine group was reported, with a significant improvement on the avolition/apathy and anhedonia/asociality subscales of the Scale for the Assessment of Negative Symptoms. Zocalli et al. 45 considered that there was a potential role for mirtazapine as an augmentation strategy in the treatment of negative symptoms of schizophrenia, without negative effects on the metabolism of different antipsychotics. Berk et al. 46 conducted a 6-week, randomised, placebo-controlled trial of mirtazapine added to haloperidol in a sample of inpatients with schizophrenia. Negative symptom scores were again significantly reduced in the mirtazapine group compared with the placebo group by the end of the trial period. The lack of any difference between the treatment groups on depression ratings led the investigators to conclude that the improvement in negative symptoms was not an artefact of mood improvement. Having conducted a meta-analysis of data from five placebo-controlled double-blind randomised controlled trials (RCTs) of mirtazapine, including the Berk et al. 46 study, Vidal et al. 47 concluded that the findings supported the hypothesis that mirtazapine augmentation of antipsychotic medication could improve negative symptoms in schizophrenia, but considered that additional and larger studies, with more methodological rigour, were needed.

In another double-blind, placebo-controlled study, Shoja-Shafti48 reported that augmentation of haloperidol with the antidepressant nortriptyline, a serotonin and noradrenaline reuptake inhibitor with anticholinergic effects, produced a significant reduction in negative symptom scores. However, there is some evidence that adding a noradrenaline reuptake inhibitor (e.g. maprotiline) antidepressant to an antipsychotic may not improve schizophrenic negative symptoms. 49

The mechanism by which adding an antidepressant to antipsychotic medication may cause a reduction in negative symptoms is unclear. The explanations postulated have included a non-specific antidepressant effect, generally increased noradrenergic drive and, more specifically, alpha-2 antagonism. How serotonergic agents might specifically exert any such benefit also remains unknown,50 not least because both SSRI antidepressants and 5-HT2 antagonists (most second-generation antipsychotics) have claims for improving negative symptoms in schizophrenia. 51 Nevertheless, the clinical evidence has been enough to prompt investment in the development of novel antipsychotic drugs that selectively combine SSRI functionality with D2 dopamine receptor antagonism. 52 However, it is possible that SSRI effectiveness might be influenced by the pharmacological action of the antipsychotic medication.

Systematic reviews focusing on the effects of the combination of antipsychotic and antidepressant drug treatment for the management of negative symptoms have been relatively circumspect in their conclusions, partly reflecting that most of the relevant studies have been characterised by small sample sizes and failure to control for change in secondary negative symptoms of schizophrenia. Rummel et al. 53 reviewed five eligible RCTs testing the combination of an antidepressant (amitriptyline, mianserin hydrochloride, trazodone hydrochloride, paroxetine, fluoxetine or fluvoxamine) added to antipsychotic medication against an antipsychotic alone for the treatment of prominent negative symptoms in schizophrenia and/or schizophrenia-like psychoses. Significant differences in favour of the combination therapy were seen in core negative symptoms: affective flattening, alogia and avolition.

Sepehry et al. 54 conducted a further meta-analysis of 11 randomised, placebo-controlled, double-blind trials (five involving fluoxetine, two involving sertraline, two involving fluvoxamine, one involving paroxetine and one involving citalopram) specifically comparing the addition of SSRI antidepressants to antipsychotics versus adding placebo for negative symptoms in people with schizophrenia spectrum disorder. They included studies excluded by Rummel et al. 53 on methodological grounds, for example, where the study sample was not characterised by predominant negative symptoms. The total sample size for the 11 studies was still relatively small (393 patients). Sepehry et al. 54 concluded that the findings of the trials, which lasted between 4 and 16 weeks, provided no global support for the addition of a SSRI for the treatment of negative symptoms of schizophrenia, which had shown a poor response to antipsychotics alone, although a mild therapeutic effect could exist in patients with more ‘chronic’ illness. Lecrubier et al. 55 commented that these results suggested that a treatment period longer than 1–3 months may be needed before there is substantial benefit, and considered that the conclusion should not be extrapolated to all people with schizophrenia.

A subsequent meta-analysis by Singh et al. 56 included data from 23 randomised, placebo-controlled trials comparing the effect of antidepressants and placebo on the negative symptoms of chronic schizophrenia. These investigators concluded that antidepressants prescribed along with antipsychotics were more effective in treating the negative symptoms of schizophrenia than antipsychotics alone. They reported a small-to-medium effect size (−0.48; p < 0.05), comparable with a later meta-analysis of 26 relevant RCTs by Fusar-Poli et al. 57 which yielded similar, respective figures (−0.349; p = 0.001). Nevertheless, a Cochrane review,58 although agreeing that a combination of antipsychotics and antidepressants may be effective in treating negative symptoms of schizophrenia, considered that the research information available was currently too limited to allow any firm conclusions to be drawn, and that large, pragmatic, well-designed and well-reported long-term trials were justified.

Selection of an antidepressant for augmentation

The question investigated in this study was whether or not antidepressants benefit negative symptoms in schizophrenia independent of any beneficial impact on depressive symptoms. For that reason, we excluded people who had a definite comorbid depressive illness. Although there is some controlled evidence that antidepressants are useful in treating depression when this fulfils DSM-IV (Diagnostic and Statistical Manual of Mental Disorders-Fourth Edition59) criteria for major depressive disorder superimposed on schizophrenia,60 the use of a combination of an antidepressant and antipsychotic agent has not been found universally helpful for depression occurring in the course of schizophrenia. 61–63 Glick et al. 64 suggested that such adjunctive medication is not generally helpful in established schizophrenia for depressive features, as they usually represent demoralisation. However, antidepressants are widely used and if people with depression had been included this could have reduced the acceptability of the randomisation to placebo or antidepressant, and potentially confused the picture. Similar reasoning argued for using an antidepressant medication that did not have any anticholinergic activity. Such activity could ameliorate extrapyramidal symptoms and so lead to a spurious, apparently beneficial effect on negative symptoms. These considerations, along with other considerations given below, guided the selection of a SSRI antidepressant for testing in this study.

When adding one drug to another it is important to consider any potential for interactions that could lead to adverse consequences for the patient. Drug interactions can be either pharmacokinetic (where one drug interferes with the way the body handles the other, usually by increasing or decreasing metabolism of the day) or pharmacodynamic (where one drug enhances or opposes the pharmacological action of the other). When added to an antipsychotic, almost all antidepressant drugs can precipitate at least one of these types of interaction and patients may experience an increased side-effect burden as a result.

With respect to pharmacokinetic interactions, for example, the SSRI antidepressants fluoxetine and paroxetine are both potent inhibitors of the hepatic cytochrome P450 (CYP) enzymes CYP2D6 and 3A4, which metabolise many antipsychotic drugs, most notably clozapine. When such enzyme activity is inhibited, the capacity to metabolise clozapine is significantly reduced and this can result in the patient experiencing serious clozapine-related side effects, such as seizures. 42,65,66

With respect to pharmacodynamic interactions, most antipsychotic drugs have variable affinity for Histamine H1, alpha-1-adrenergic and acetylcholine receptors leading, to side effects such as sedation, postural hypotension and constipation, respectively. 66 Tricyclic antidepressants also have affinity for these receptors and, for example, when combined with risperidone would be predicted to exacerbate postural hypotension, or when combined with olanzapine, sedation and constipation. Unacceptable side effects are an important cause of non-adherence to prescribed medication. 67

Considering the choice of augmenting antidepressant, among the SSRIs no individual agent has a convincingly stronger evidence base for treating negative symptoms54,58 and therefore safety and pharmacokinetic considerations are key. As discussed above, some SSRIs are potent inhibitors of the hepatic CYP enzymes involved in the metabolism of certain antipsychotics, leading to variable increases in plasma antipsychotic drug concentrations. Of particular concern is that some patients with enduring negative symptoms are likely to be receiving clozapine; increases in clozapine plasma levels are associated with potentially serious side effects such as seizures and serotonergic syndrome. 42,65 The risk of a pharmacokinetic interaction is lowest with the SSRI citalopram. At the time the Antidepressant Controlled Trial For Negative Symptoms In Schizophrenia (ACTIONS) began, the only published evidence for citalopram as an adjuvant treatment in patients with established schizophrenia and persistent negative symptoms39 was positive, being associated with a significant advantage over placebo on ‘subjective well-being’.

As the patients recruited to our study could be receiving any antipsychotic drug, it was important to select an augmenting antidepressant that was both well tolerated and had minimal potential to interact with any antipsychotic in any way. Citalopram best fits these criteria, being generally well tolerated, having little effect on hepatic metabolising enzymes and having a relatively low affinity for alpha-1-adrenergic and muscarinic receptors. 66

Study aims

The main objectives were, first, to test the benefits of citalopram (a SSRI antidepressant) for people with schizophrenia and negative symptoms in terms of improved quality of life and reduction of negative symptoms, as well as recording the relative risks and costs of this augmentation of antipsychotic medication. Second, the study was designed to extend the current evidence that indicates that SSRI augmentation in the treatment of schizophrenia acts directly on negative symptoms, has only limited efficacy in treating depressive symptoms, and does not have a detrimental effect on positive symptoms or extrapyramidal side effects.

Trial design

The study was a multicentre, double-blind, individually randomised, placebo-controlled, parallel-arm RCT with a 1-year follow-up. Following the baseline assessment, follow-up assessments occurred after 12, 36 and 48 weeks. The first participant was recruited in September 2011 and recruitment finished at the end of September 2013.

Sample size

Calculation of the target sample size used the pragmatic, Health Technology Assessment-funded Cost Utility of the Latest Antipsychotic Drugs in Schizophrenia Study (CUtLASS 1) of first- versus second-generation antipsychotic treatment for broadly defined, treatment-resistant schizophrenia21,68 as a precedent. In CUtLASS 1, the within-group standard deviation (SD) for the Heinrich’s Quality of Life Scale69,70 (QLS) change score at 12 months was 13. There is often difficulty in defining a clinically significant difference in scores. There is some justification for considering a difference of over 4.5 points or 0.35 SDs as an important change,71,72 which is slightly less than the CUtLASS 1 assumption of a 5-point increase being the criterion for a clinically important improvement. In order to detect a difference of 0.35 SDs with 85% power at 5% significance (two-sided), data on 148 individuals in each group were required. In CUtLASS 1 there was 83% follow-up at 3 and 12 months; assuming the same attrition as CUtLASS 1, the recruitment target was 358 individuals. The same calculations would also apply to scores on the Positive and Negative Syndrome Scale (PANSS), the other primary outcome. In CUtLASS 1 the SD for the PANSS negative subscale was 6, so 0.35 SDs would correspond to approximately 2 points.

Primary outcomes

The primary outcome was quality of life measured at 12 and 48 weeks, assessed using an observer-rated scale, the QLS,69 and negative symptoms at 12 weeks, measured on the negative symptoms subscale of the PANSS. 73 We also used the PANSS negative symptom subscale revised by Marder et al. ,74 which consists of the following PANSS items: blunted affect, emotional withdrawal, poor rapport, passive/apathetic social withdrawal, motor retardation, active social avoidance and lack of spontaneity/flow of conversation. Furthermore, we derived assessments for the two distinct domains of negative symptom psychopathology,15,75,76 ‘expressive deficits’ and ‘avolition for daily life and social activities’, as separable components of the negative syndrome that might have a differential response to treatment interventions. 77 Two subscales were generated: an expressive deficits subscale (comprising three PANSS items: lack of spontaneity and flow of conversation, poor rapport and blunted affect) and an avolition subscale (comprising three PANSS items: disturbance of volition, emotional withdrawal and passive/apathetic emotional withdrawal). These subscales are composed of items also included in the two respective subdomains of negative symptoms (‘expressive deficits’ and ‘social amotivation’) yielded by an exploratory factor analysis of PANSS items by Liemburg et al. 78

Secondary outcomes

Mental state symptoms were assessed using the PANSS total and positive symptoms subscale. 73 Depression was assessed using the Calgary Depression Rating Scale for Schizophrenia (CDSS). 79 Social function was assessed by the Birchwood Social Function Scale80 and the level of engagement with clinical services by the Service Engagement Scale. 81 Antipsychotic side effects were assessed comprehensively by a combination of scales. General side effects were systematically evaluated using the Antipsychotic Non-Neurological Side Effects Scale (ANNSERS),82 which covers a range of aversive subjective experiences as well the cardiovascular, gastrointestinal, endocrine/metabolic, autonomic, genitourinary, sexual and central nervous system side effects of antipsychotic medication. Additional items were added to address recognised SSRI antidepressant side effects, including serotonergic symptoms, such as nausea and dizziness. These additional SSRI antidepressant side effect items were collated as the ANNSERS-compiled (ANNSERS-c) rating scale. The motor, extrapyramidal side effects were rated with scales for each of the distinct motor syndromes: drug-induced parkinsonism was rated using the Extrapyramidal Side Effects Scale,83,84 akathisia using the Barnes Akathisia Rating Scale85 and tardive dyskinesia using the Abnormal Involuntary Movements Scale. 86 Medication adherence was tested using the self-reported Morisky Compliance scale. 71

In addition, there was a range of health economic outcome measures. In line with the recommendations from the National Institute for Health and Care Excellence,87 the European Quality of Life-5 Dimensions, a widely used, generic, multiattribute utility scale, was completed for each participant at baseline, 12, 36 and 48 weeks, to measure patients’ general health-related quality of life. 88 A new measure of capabilities, the ICEpop CAPability measure for Adults (ICECAP-A),89 was also used alongside this as it was anticipated that this measure would become an important outcome measure for economic analysis by the end of the trial. The capability approach is a fairly new concept that defines well-being in terms of an individual’s ability to ‘do’ and ‘be’ the things that are important in life. It provides a broader framework for benefit measure in economic evaluations than the current gold standard, health-related quality of life. In addition, a short specific questionnaire was also completed to collect resource-use information in the trial from a social perspective.

Protocol changes

In addition to wording changes to clarify procedures, a number of amendments were made to the protocol during the trial, as described in the following sections.

Participants

The study was conducted at 15 clinical sites in the UK. The first participant was recruited in September 2011 and recruitment finished at the end of September 2013. The participants were people under the care of secondary mental health services, with an established schizophrenic illness characterised by persistent negative symptoms at a criterion level of severity, despite treatment with antipsychotic medication: a negative subscale score of 20 or more41,94,95 on the PANSS with at least three of the seven items on the negative symptom subscale rated ≥ 3. Eligibility for the study was determined by the following, additional, inclusion and exclusion criteria.

Pharmacological intervention

Eligible participants were randomised 1 : 1 to treatment with either placebo (one capsule) or 20 mg of citalopram per day for 48 weeks, but at 4 weeks a participant’s clinician had the option to increase the dose of citalopram to 40 mg per day or two placebo capsules for the remainder of the study; if there were problems with tolerability the clinician could reduce the dose back to 20 mg per day (or one placebo capsule). Medication was supplied as identical capsules containing either 20 mg of citalopram or placebo, packaged into monthly packs of 28 tablets. The 20-mg dose or the starting placebo dose was prescribed as one capsule each morning; the 40-mg dose or increased placebo option was prescribed as two capsules each morning. Participation in the study did not restrict the therapeutic options of the clinical team in terms of additional medication (with the exception of additional mood stabilisers or other antidepressants, which were not allowed) or psychosocial interventions. A fully automated online (and telephone) randomisation service was provided by the Bristol Randomised Trials Collaboration. In addition, a 24-hour unblinding service was provided by the Medical Toxicology Information Service of Guy’s and St Thomas’ NHS Foundation Trust.

Patient and public involvement

During development of the application for funding, a North London, service-user led group dedicated to facilitating research (SUNLOWS) provided advice on the trial design. Both the Trial Management Group and Trial Steering Committee had a mental health service user as a member. In addition, this individual was involved in the development of documents for participants, such as the information sheet and newsletters.

Data analysis

The analysis and reporting of this trial were undertaken in accordance with Consolidated Standards of Reporting Trials guidelines. 97 All analyses were conducted in Stata 12.1 (StataCorp LP, College Station, TX, USA), following a pre-defined analysis plan agreed with the Trial Steering Committee. All program files were stored to maintain a record of any generated variables and analyses, along with a copy of the regional data set. All analyses remained blind to the intervention group until they were completed.

Descriptive statistics of the key clinical and sociodemographic variables were used to review the data. The baseline comparability of the intervention and control groups was examined in terms of QLS and PANSS negative subscale score. These and other potential prognostic factors were examined over the course of the study, including PANSS total score, CDSS score, Birchwood Social Function Scale score, Service Engagement Scale score, ANNSERS total score, the scores on the extrapyramidal side effects rating scales and medication adherence. In addition, health and social care resource use and sociodemographic information were compared between the two groups.

Baseline data are reported for the intervention and control group. The data are presented in tables with mean and SDs for continuous outcome measures, and proportions and range for binary outcomes. For the primary outcome, linear regression was used to determine mean difference between treatment groups. Differences between groups are presented as mean differences with associated 95% confidence interval (CI). Baseline values for variables were included as covariates in regression models and adjusted according to analysis of covariance. All analyses were conducted under the intention-to-treat principle.

Primary outcomes

The results of the data analysis presented in Table 2 indicate that there was no difference between the two treatment arms in terms of QLS score at the primary follow-up time of 12 weeks. Those participants who were assigned to the citalopram group had slightly higher QLS scores compared with those patients who were randomised to placebo. This difference was reversed by adjustment for baseline PANSS score but the 95% CIs are wide and the difference is neither statistically nor clinically relevant. QLS scores were higher in both groups at 48 weeks post intervention. Similarly to the week-12 results, there was no difference between groups in terms of quality of life.

However, applying the criterion for a clinically significant improvement (a change of ≥ 5 points on QLS total score), a higher proportion of participants in the citalopram arm met this responder criterion (Table 3), although this was not deemed statistically important.

The data analysis shown in Table 4 indicates that there was no difference between the two treatment groups in mean PANSS negative subscale score at the primary follow-up time of 12 weeks.

The participants who had been randomised to citalopram treatment had slightly higher PANSS negative subscale scores at 12 weeks than those patients who were randomised to placebo. This difference was similar after adjustment for baseline PANSS negative subscale score, indicating no difference between citalopram and placebo in terms of negative symptoms.

A clinically significant improvement in the PANSS negative subscale score is indicated by a reduction of ≥ 2 points. Patients in both groups recorded a clinically meaningful reduction in PANSS subscale score at 12-week follow-up compared with baseline, although the proportion was higher in the citalopram arm (see Table 3). There was no evidence that the proportion of patients who ‘improved’ in terms of QLS or PANSS negative symptoms differed significantly between the two groups.

Secondary outcomes

The data presented in Tables 5 and 6 indicate that, overall, there was no difference in secondary outcomes between the intervention and control groups. There is a suggestion that lower levels of avolition (mean difference –1.3, 95% CI –2.5 to –0.09) were recorded at 12-week follow-up for participants in the citalopram group compared with the placebo group, although this difference between the treatment groups was not maintained at follow-up at 36 weeks (mean difference –0.2, 95% CI –1.7 to 1.3) or at 48 weeks (mean difference –1.3, 95% CI –3.0 to 0.4).

Similarly, there is a suggestion that lower mean ANNSERS scores were recorded for the citalopram-treated participants at 12-week follow-up, reflecting fewer and/or less-severe side effects (mean difference –4.3, 95% CI –7.9 to –0.83). Examination of this outcome at the follow-up assessment time points showed that this difference between the treatment groups was maintained at 36 weeks (mean difference –4.5, 95% CI –8.3 to –0.7) but absent at 48 weeks (mean difference –5.2, 95% CI –12.1 to 1.5).

The ANNSERS outcome was further defined as the proportion of participants who had experienced a ‘severe’ side effect, that is, a rating of 3 on at least one of the 34 items contributing to this score. The proportion of participants who recorded at least one ‘severe’ symptom was similar between groups (n = 7 vs. n = 9 in the placebo and citalopram groups, respectively).

Health economics analysis

The final economic analysis was based on the 30 study participants randomised to citalopram and the 32 participants randomised to placebo. Appendix 3 shows the number of patients with available health economics data by information category at each assessment point.

Table 8 summarises the baseline characteristics of the participants relevant for the economic analysis. Participants in the citalopram arm had a somewhat lower duration of illness compared with the placebo arm (14 years vs. 18 years, respectively). Considerably more patients in the placebo arm also lived in an inpatient mental health facility at baseline (15.63% vs. 3.33%, respectively). Only five patients had any kind of employment at any point in time over the trial period.

Baseline costs are important for recognising potential differences between the two arms in health-care utilisation prior to randomisation. At baseline, costs tended to be higher in the placebo arm but differences compared with the treatment arm were not statistically significant (for all p > 0.05). However, the mental health and non-mental health inpatient costs and the cost of absenteeism in the two arms were considerably imbalanced at baseline (see Table 8).

Appendix 4 lists the types of the health-care, social care and voluntary sector services used by any of the patients in the ACTIONS study and their respective UK national-level unit costs with source information. The participants with the highest costs at baseline (and throughout the study) were those who were living in an inpatient mental health setting and those who used additional drug and/or alcohol addiction services.

There was a slight trend towards improvement in quality of life and well-being measured by the EQ-5D-3L and the ICECAP-A, respectively, in both arms as shown in Tables 9 and 10. This trend was observable both in the available case and imputed complete data sets. There was no evidence of a significant difference in terms of improvement between the citalopram and the placebo arms in neither the EQ-5D-3L (p = 0.250) nor ICECAP-A (p = 0.248).

After adjusting for baseline differences, there was no statistically significant difference between the arms in any of the cost categories, except non-mental health inpatient costs. This borderline significant difference in non-mental health inpatient costs (p = 0.05) is, however, a result of having only two patients in the available sample (both in the citalopram arm) with positive resource use in this category, and cannot be seen as evidence of difference because of the small overall sample size (Table 11).

Primary outcomes: quality of life and negative symptoms

No differences emerged between the two treatment groups on the quality-of-life measure. In relation to negative symptoms, there were no differences between the treatment groups on overall negative symptoms scores but there was a suggestion that, after 12 weeks of treatment, those participants treated with citalopram had lower levels of avolition/amotivation. A differential improvement of separate negative symptom domains in response to antidepressant medication has been reported before. 103 Had ACTIONS recruited to target, the intention was to include depression scores as a covariate to determine whether or not any difference between groups at follow-up in negative symptoms could be explained by amelioration of depressive features in the citalopram group. The limited power of the study does not allow us to do this for the avolition finding using the originally planned methodology. However, the treatment groups did not differ significantly on depression scores at follow-up, so there is no indication that differences found in avolition can be explained by the antidepressant effect of citalopram. This finding of a possible preferential effect of citalopram on the avolition/amotivation negative symptom domain has, perhaps, two possible implications. First, it suggests that citalopram may be worthy of further investigation as an adjunctive medication in patients with established schizophrenia and negative symptoms. A key rationale for such treatment in clinical practice is to render a patient more disposed to, and more able to engage with, psychosocial interventions and thus increase the effectiveness of rehabilitation programmes. Avolition/amotivation is recognised as a critical barrier to such psychosocial rehabilitation and to achieving better functional outcomes,104–107 and is a therefore a critical treatment target in this context. Second, this finding supports the view that future studies examining the treatment response of negative symptoms in schizophrenia should use rating scales that allow for a valid and separate evaluation of avolition/amotivation. 106,107

Secondary outcomes: safety and tolerability of citalopram

Detailed and comprehensive assessment of medication side effects throughout the follow-up period did not yield any evidence for a greater side-effect burden with adjunctive citalopram compared with placebo. Indeed, the findings suggest a slightly lower general side-effect burden for the citalopram-treated participants at 12 and 36 weeks’ follow-up. This suggests that citalopram is generally well tolerated in people with schizophrenia on antipsychotic medication, but it is also a plausible speculation that the lower side-effect score in the citalopram-treated group may partly reflect some alleviation of anxiety symptoms that participants had attributed to their medication108 and thus reported as side effects.

There was no difference between the treatment groups in terms of duration of QTc interval at any time point. No participant had a QTc duration of more than 500 milliseconds and only one had an increase in the QTc of more than 60 milliseconds, thresholds that are generally taken as an indication to stop medication because of the risk of potentially fatal cardiac arrhythmias such as Torsades de pointes. 109,110 Thus, there was little in the limited QTc interval data collected in this study to suggest that citalopram has any increased risk for QTc interval prolongation in individuals prescribed concomitant antipsychotic medication.

Health economics analysis

It appears from the health economics analysis that there is no statistically significant difference between the two arms over the 48-week follow-up period in either the health economics outcomes or costs. Given the small sample size and large instance of losses to follow-up and withdrawals, the data were not adequately powered to discern any small difference between the two groups that might exist. Nevertheless, citalopram is a relatively low-cost intervention, with a mean medication cost of £27.62 per patient in the treatment arm. Safety concerns and the additional clinical tests required, however, may make the use of this medication considerably more expensive and less suitable for the target population. This study highlights the importance of balancing mental health trial arms at baseline according to usual accommodation, collecting baseline cost information and correcting for any imbalance in the respective cost categories in the health economic analysis.

One of the major limitations of this economic analysis is the small sample size. Particular costs could be under-represented; for example, the cost of living in sheltered/supported accommodation, nursing homes and inpatient mental health facilities. The considerable differences in mental health inpatient and non-mental health inpatient costs, favouring the citalopram arm in the first instance, are the result of the small number of cases with actual resource use in these cost categories (citalopram arm n = 1, placebo arm n = 14; citalopram arm n = 0, placebo arm n = 10; for mental health inpatient and non-mental health inpatient costs, respectively) and the existing baseline imbalance between the trial arms. We could account for baseline differences, but not for the small case numbers, using regression analysis.

Resource-use data collection was based on patient recall, checked where possible against patient records by the investigators. Owing to the large number of patients lost to follow-up and withdrawals, and some other study hurdles, the current data on resource use are prone to error, likely causing an underestimation of real-life costs for this patient cohort; thus, the current cost estimates should be used with caution.

Some further limitations of the analysis include the lack of carer perspective. Although investigating this was also one of the original intentions of the trial, because of difficulties in recruiting and following up patients in addition to identifying primary informal carers, these data could not be collected. Such data could be particularly useful in identifying further benefits or detriments of the treatment.

Recruitment

From the beginning of this trial, we encountered various barriers to recruiting participants to time and target, as appears to be the case for many trials in the UK and particularly so for those conducted in mental health services. 111 Negotiation of the hurdles of research governance, regulation and NHS permissions, contracts and costs allocation meant that opening the study sites was a slower process than expected, resulting in the waste of many valuable months for recruitment. Once sites were open to recruitment, referrals to the study were patchy and often ebbed to non-existent, despite feasibility work prior to this trial indicating that a large proportion of patients would be eligible. These were not refusals on the part of eligible patients but rather that patients were not being given an opportunity to take part. Referrals must come from a member of the patient’s clinical team, and clinical teams could have competing clinical priorities, concerns about how introducing a trial to the patient might impact on their therapeutic relationship, a lack understanding of the clinical equipoise of the research question, and/or safety concerns (see Urgent safety measures). The result was that recruitment occurred in focused areas where sympathetic clinical staff were found, and many eligible patients in other areas missed the opportunity to participate. Additional trial sites were opened over time in order to access more eligible patients and the trial was promoted using talks to clinical teams, newsletters and researcher recruitment training days that covered engagement of both patients and clinical staff. Adopting more pragmatic eligibility criteria was also considered, but was not implemented as the majority of these criteria reflected safety requirements, compromise on the clearly defined target population was unwanted, and the screening log information did not identify any particular criterion whose modification would have led to a substantial increase in participant numbers.

Conclusions

This trial under-recruited, falling well short of the target of 358 participants, which reflects the range of barriers to recruitment that were encountered; in particular, the impact of a MHRA warning about cardiac problems with citalopram which was issued shortly after the study commenced recruitment. This warning contraindicated the combination of citalopram and an antipsychotic agent, and required the implementation of urgent safety measures in the trial. Given the relatively small sample size achieved, the power of any statistical analysis to detect clinical or statistically meaningful significant differences between the citalopram and placebo groups was limited.

The trial found no difference between treatment groups on the primary outcome measures over a follow-up period of up to 1 year. The question of whether or not statistically significant differences would have been identified with the original sample size is moot, as is whether or not further trials of such an augmentation strategy for persistent negative symptoms in people with established schizophrenia are warranted. However, the secondary analysis of the data in this relatively small trial suggested a modest, positive effect on the negative symptom subdomain, avolition/amotivation, which is recognised as a critical barrier to psychosocial rehabilitation and to achieving better social and community functional outcomes. If there are future studies of this treatment strategy, rating scales that allow for evaluation of avolition/amotivation as a distinct and separate negative symptom domain should be used.

With regard to health economic measures, there were no statistically significant differences between the citalopram and placebo arms but a considerably larger sample size may be needed to draw robust conclusions on the cost-effectiveness of the use of antidepressants in combination with antipsychotics to treat persistent negative symptoms in people with schizophrenia who do not fulfil the criteria for major depression. Our study highlights the importance of balancing the treatment arms at baseline in such trials according to usual accommodation, collecting baseline cost information and correcting for any imbalance in a respective health economic analysis.

Regarding choice of antidepressant, information on the relative merits of available antidepressants in this treatment context is limited, so the rationale for originally choosing citalopram for this study largely reflected its relatively favourable safety and tolerability profile. The publication of some limited data on QTc prolongation with citalopram resulted in the MHRA safety warning and brought our choice of antidepressant into question. However, the contraindication of the prescription of the combination of citalopram and an antipsychotic drug was based on evidence that antipsychotic medication also has the potential for QTc prolongation, rather than any data clarifying the impact on QTc interval of this particular combination. Overall, comprehensive assessment of side-effect burden in this trial did not identify any serious safety or tolerability issues for citalopram as an adjunct to continuing antipsychotic medication; citalopram did not increase the side-effect burden or have any evident effect on QTc. However, if adjunctive citalopram were to be tested in future trials, appropriate cardiac monitoring should be in place to further assess the safety of this drug combination.

The viability of adjunctive antidepressant treatment as a long-term prescribing strategy for the treatment of negative symptoms in schizophrenia may merit further investigation but the ability to recruit an adequate sample size remains a barrier to any trial conducted in a similar clinical setting. Leaving aside the exceptional issue of the necessary urgent safety measures, the various strategies we adopted to overcome the barriers to recruitment identified did not allow the target sample size to be achieved, and so we are unable to derive any recommendations for ensuring an adequate recruitment rate in future similar trials.

Contributions of authors

Thomas RE Barnes (Professor of Clinical Psychiatry) was the principal grant applicant, developed the trial protocol and was responsible for the overall conduct of the trial as chief investigator. He led the interpretation of the results and drafting of the final report.

Verity C Leeson (Clinical Trials Manager) contributed to the development of the original trial protocol and its amendments, and was responsible for the day-to-day management of the trial. She contributed to the interpretation of the results and preparation of the final report.

Carol Paton (Chief Pharmacist) cowrote the original grant application, was a principal investigator, contributed to the trial protocol and its amendments, and to the interpretation of the results for the final report.

Céire Costelloe (Career Development Fellow) led the statistical analysis and contributed to the interpretation and write-up of the results for the final report.

Judit Simon (Professor of Health Economics) contributed to the original grant application and the development of the trial protocol. She co-led the health economics analysis and write-up.

Noemi Kiss (Health Economics Researcher) co-led the health economics analysis and write-up.

David Osborn (Professor of Psychiatric Epidemiology) cowrote the original grant application, was a principal investigator and commented on earlier drafts of the final report.

Helen Killaspy (Professor of Rehabilitation Psychiatry) cowrote the original grant application, was a principal investigator and commented on earlier drafts of the final report.

Tom KJ Craig (Professor of Social and Community Psychiatry) cowrote the original grant application, was a principal investigator and commented on earlier drafts of the final report.

Shôn Lewis (Professor of Adult Psychiatry) cowrote the original grant application, was a principal investigator and commented on earlier drafts of the final report.

Patrick Keown (Consultant Psychiatrist) was a principal investigator and commented on earlier drafts of the final report.

Shajahan Ismail (Consultant Psychiatrist) was a principal investigator and commented on earlier drafts of the final report.

Mike Crawford (Professor of Mental Health Research) was a principal investigator and commented on earlier drafts of the final report.

David Baldwin (Professor of Psychiatry) cowrote the original grant application, was a principal investigator and commented on earlier drafts of the final report.

Glyn Lewis (Professor of Psychiatric Epidemiology) cowrote the original grant application, was a principal investigator and commented on earlier drafts of the final report.

John Geddes (Professor of Epidemiological Psychiatry) cowrote the original grant application, was a principal investigator and commented on earlier drafts of the final report.

Manoj Kumar (Consultant Psychiatrist) was a principal investigator and commented on earlier drafts of the final report.

Rudresh Pathak (Consultant Psychiatrist) was a principal investigator and commented on earlier drafts of the final report.

Simon Taylor (Consultant Psychiatrist) was a principal investigator and commented on earlier drafts of the final report.

Publication

Barnes TRE, Paton C. Uncertainties page: do antidepressants improve negative symptoms in schizophrenia? BMJ 2011;342:d3371.

Data sharing statement

All data are available on request from the corresponding author.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health.