Amisulpride augmentation in clozapine-unresponsive schizophrenia (AMICUS): a double-blind, placebo-controlled, randomised trial of clinical effectiveness and cost-effectiveness

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 08/116/12. The contractual start date was in April 2011. The draft report began editorial review in November 2015 and was accepted for publication in April 2016. 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 reports personal fees from Roche, Sunovion Pharmaceuticals Inc. and Otsuka Pharmaceutical/Lundbeck, outside the submitted work. Carol Paton reports personal fees from Sunovion Pharmaceuticals Inc. and Eli Lilly and Company, outside the submitted work. Linda Davies reports a grant from the National Institute for Health Research (NIHR) to the University of Manchester, during the conduct of the study. Patrick Keown reports personal fees from Otsuka Pharmaceutical/Lundbeck and other from Janssen-Cilag, outside the submitted work. Hemant Bagalkote reports personal fees from Janssen-Cilag, Lundbeck and Eli Lilly and Company, outside the submitted work. Peter M Haddad reports grants from the NIHR during the conduct of the study; personal fees for lecturing and/or consultancy (including attending advisory boards) from Allergan plc, Eli Lilly and Company, Galen Limited, Janssen-Cilag, Lundbeck, Otsuka Pharmaceutical, Quantum Pharmaceutical, Roche, Servier Laboratories, Sunovion Pharmaceuticals Inc., Takeda Pharmaceutical Company, Teva Pharmaceutical Industries Ltd; and received conference support from Janssen-Cilag, Lundbeck, Otsuka Pharmaceutical, Servier Laboratories and Sunovion Pharmaceuticals Inc., outside the submitted work. Mariwan Husni reports other from Otsuka Pharmaceutical, outside the submitted work.

Copyright statement

© Queen’s Printer and Controller of HMSO 2017. 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.

Clozapine augmentation with a second antipsychotic

For treatment-resistant schizophrenia, a common therapeutic approach is to use more than one antipsychotic, although a robust evidence base to justify this is lacking. Recent surveys of prescribing patterns in the USA suggest that about 15% of outpatients and up to 50% of inpatients with schizophrenia receive two or more antipsychotics. 8 In the UK in 2012, national clinical audit data from the Prescribing Observatory for Mental Health (POMH-UK) on over 5000 acute inpatients and over 3000 forensic patients prescribed antipsychotics revealed regularly prescribed combined antipsychotics in 14% and 17%, respectively. 9 A common reason given by clinical teams for prescribing such a combination was failure of the illness to respond to treatment with a single antipsychotic. In the POMH-UK audit, just under 5% of prescriptions for combined antipsychotics in acute adult wards represented the augmentation of clozapine with another antipsychotic, whereas in rehabilitation/complex needs services and forensic services the respective figures were in excess of 20%. These figures are in line with other reports of the prevalence of clozapine augmentation, ranging from 18% to 44% depending on the clinical setting and country. 10–12 In summary, it seems that around one-third of all clozapine-treated patients receive augmentation with another antipsychotic. 13 This is because it is one of the few therapeutic strategies available to clinicians for those people with schizophrenia that has proved to be poorly responsive to clozapine.

Clozapine is the only antipsychotic for which there is convincing evidence of efficacy in strictly defined treatment-resistant schizophrenia, but in such cases it has limited effectiveness, with 30–40% of patients showing an inadequate response to the drug. 14 In some patients, a range of potentially serious side effects such as seizures, sedation and tachycardia may prevent the optimal dose being reached. In addition, weight gain and metabolic side effects that increase the risk of diabetes and cardiovascular disease become apparent in many.

In an attempt to improve efficacy and limit tolerability problems, clinicians commonly augment clozapine with another antipsychotic, despite limited evidence on the potential risks and benefits of this practice. 15 In 2005, Kontaxakis et al. 16 identified 15 case studies, with a total of 33 patients, of adjunctive agents in clozapine-resistant schizophrenia, 10 of which involved a second antipsychotic (eight using risperidone, one using sulpiride and one using olanzapine). They concluded that various methodological shortcomings limited the impact of the findings. These authors came to a similar conclusion after conducting a critical review of 11 randomised controlled trials (RCTs; with a total of 270 participants) of clozapine augmentation in treatment-resistant schizophrenia,17 in only one of which was an antipsychotic (sulpiride) used as the adjunctive medication. Methodological weaknesses included the fact that only one trial adequately reported the dose and duration of the clozapine monotherapy phase. Similarly, Remington et al. 15 noted that the current body of evidence for clozapine augmentation consisted of data from a limited number of small open trials and case series reports. But they suggested that systematic research was warranted and argued for detailed cost–benefit analysis. Buckley et al. 11 agreed, stating that there was ‘a dearth of double-blind studies’, and concluded that these adjunctive therapies were worthy of further testing in carefully controlled clinical trials.

The subsequent publication of several new open studies and small RCTs testing the therapeutic value of augmenting clozapine with another antipsychotic prompted us to conduct a meta-analysis of eligible RCTs. 18 A systematic literature search identified eight open studies and four eligible RCTs in which clozapine was augmented with a second antipsychotic, with a total of 166 participants. The two RCTs that had lasted 10 weeks or more gave an odds ratio (OR) of response to treatment of 4.41 [95% confidence interval (CI) 1.38 to 14.07]. We concluded that for clozapine-refractory schizophrenia, augmentation of clozapine with another antipsychotic drug is worthy of an individual clinical trial, but this may need to be longer than the 4–6 weeks usually recommended for acute antipsychotic monotherapy, a view supported by Correll et al. 19 Mouaffak et al. 13 noted that the discrepant results of the published studies of clozapine augmentation with another antipsychotic; they identified methodological shortcomings that related to the heterogeneity of definitions of resistance to clozapine, choice of outcome measures, and the dose and duration of the adjunctive drugs, which they considered ‘a major limitation for drawing conclusions’. Clinical response in such studies has generally been defined as a 20% reduction in total score on the Brief Psychiatric Rating Scale (BPRS) or the Positive and Negative Syndrome Scale (PANSS). Both the BPRS and the PANSS assess a broad range of symptoms, including both positive symptoms (e.g. delusions, hallucinations and thought disorder) and negative symptoms (e.g. blunted affect and emotion, poverty of speech, lack of motivation, and social and emotional withdrawal).

The updated National Institute for Health and Care Excellence (NICE) 2009 guideline for the treatment of schizophrenia20 supported the augmentation of clozapine with a second antipsychotic in patients with an inadequate response to clozapine alone; advice that was in accord with our own meta-analysis. 18 After the publication of our meta-analysis, data became available for four further, short-term, clozapine augmentation RCTs, one each for risperidone21 and haloperidol,22 and two using aripiprazole,23,24 a medicine that has a different pharmacology to other antipsychotic drugs in that it is a D₂ partial agonist. All four studies were reported as negative, although the trial by Chang et al. 23 showed a statistically significant advantage for augmentation with aripiprazole in regard to a reduction in negative symptom score.

By 2012, Sommer et al. 25 were able to identify 10 RCTs of augmentation of clozapine with a second antipsychotic and found only modest or absent efficacy. Also in 2012, Taylor et al. 26 conducted a meta-analysis of 14 such studies. These investigators were more positive about this strategy for clozapine-refractory illness, concluding from their findings that augmentation of clozapine with a second antipsychotic was modestly superior to placebo and equally well tolerated. Looking at effect size by treatment duration, they found no significant relationship between duration of treatment and reduction in symptoms. In other words, there was no confirmation of the earlier suggestion that longer trials might produce relatively more robust outcomes for adding a second antipsychotic.

Amisulpride augmentation of clozapine

Despite the lack of robust clinical evidence on the potential risks and benefits of clozapine augmentation with amisulpride, this is a strategy commonly used by clinicians in the NHS. The POMH-UK clinical data from 20129 revealed that amisulpride was the antipsychotic most commonly prescribed in association with clozapine.

Amisulpride has been tested in case reports and case series36–38 and open studies of clozapine augmentation. 39–41 Augmentation with amisulpride was found to be well tolerated, and clinical responses (again defined as a ≥ 20% reduction on PANSS total score) in the open studies by Munro et al. 40 and Ziegenbein et al. 39 occurred in around 70% of patients. Hotham et al. 42 augmented clozapine with amisulpride in six forensic (high-secure) patients and reported a reduction in aggression and violence. There has been one previous pilot, double-blind, placebo-controlled RCT of clozapine augmentation with amisulpride43 for 6 weeks in 16 patients, with established schizophrenia, who were partially responsive to clozapine. The primary outcome measures, such as reduction in BPRS total score, failed to show a significant improvement, which the investigators attributed to the study’s lack of power.

Study aims

The main aims of the study were:

• to test the benefits, costs and risks of augmenting clozapine with amisulpride compared with placebo

• to add to the clinical and economic evidence base for clozapine augmentation with a second-generation antipsychotic

• to examine the potential benefits, costs and risks of clozapine augmentation in treatment-resistant schizophrenia.

The key aim of the economic analysis was to assess whether or not augmenting clozapine with amisulpride was likely to be cost-effective. To address this aim, two sets of analyses were conducted. The first set used the trial data set (within-trial analyses) and the 3-month, scheduled, follow-up period to:

• estimate the use of health- and social-care services to manage participants who were treated with clozapine and placebo and those who had clozapine plus amisulpride

• estimate the direct costs of health- and social-care services to manage participants who were treated with clozapine and placebo and those who had clozapine plus amisulpride

• estimate the overall health status of participants who had clozapine and placebo and those who were treated with clozapine plus amisulpride

• estimate the utility and quality-adjusted life-years (QALYs) of participants who were treated with clozapine and placebo and those who had clozapine plus amisulpride

• estimate the net costs, QALYs and incremental cost per QALY gained associated with augmenting clozapine with amisulpride.

The second set of analyses used an economic model to explore the following questions over the longer term:

• What are the likely net costs, QALYs and incremental cost per QALY gained, if symptom response of augmenting clozapine with amisulpride increases/decreases in the long term?

• What are the likely net costs, QALYs and incremental cost per QALY gained, if augmenting clozapine with amisulpride is associated with increased/decreased side effects in the long term?

Trial design

The study was a randomised, double-blind, placebo-controlled trial lasting 12 weeks. Eligible patients, all receiving continuing clozapine treatment, were randomised 1 : 1 to receive augmentation with either amisulpride or placebo. Researchers, participants and care providers were blinded to patients’ treatment randomisation until the end of their participation by the use of identically matched placebo medicine. The trial recruited from November 2011 to December 2014.

Sample size

In previous augmentation studies of clozapine with a second antipsychotic lasting ≥ 10 weeks, 35%28 and 50%44 of the actively augmented arm responded compared with 10% in those augmented with placebo. Response was defined as a > 20% improvement on the BPRS or PANSS. To detect a more conservative response of 30% in our amisulpride arm and 10% in the placebo arm, with 90% power and an alpha of 0.05, we required 92 participants per group (two-sided). Using existing continuous data, detecting an effect size of 0.5 BPRS score standard deviations (SDs; compared with the final placebo group mean BPRS score) would require 85 participants per arm. This held whether we used the results from Josiassen et al. 28 [mean BPRS: 44.7 (placebo) vs. 40.1 (active arm), SD 9.2] or Shiloh et al. 44 [mean BPRS: 41.2 (placebo) vs. 35.1 (active), SD 12.2], both using 90% power and a two-sided alpha of 0.05.

The plan was to randomise 230 patients (including 20% inflation for drop-outs) on clozapine 1 : 1 to receive augmentation with placebo or amisulpride. Clinical protocols for the use of clozapine involve regular attendance for haematological testing and most patients either are inpatients or attend clozapine clinics. Once stabilised on treatment, the attrition rate is low. Small studies of clozapine augmentation report completion rates of > 80%, so we estimated there would be 184 study completers (i.e. 92 completers in each treatment group).

Our sample size calculation was based on results from previous studies. 28,44 These studies were comparable with the proposed study in terms of length of follow-up, intervention used and proposed primary outcome. From these we based our sample size on 30% of participants in the amisulpride arm of the trial achieving a > 20% improvement on the BPRS or on the PANSS, which is slightly lower than those shown in the literature. 28,44 Therefore, it was expected that there would be a large difference in the percentage achieving such a threshold level of improvement on the BPRS or on the PANSS between the amisulpride and placebo groups at the end of follow-up, and so there was no statistical justification for increasing the sample size.

As part of the amisulpride augmentation in clozapine-unresponsive schizophrenia (AMICUS) study rescue plan submitted to the Health Technology Assessment (HTA) programme in March 2013, the sample size was recalculated with the same assumptions but with 80% power. This gave 72 participants in each arm (144 in total); accounting for 20% drop-out, the revised target was 90 participants in each group (180 participants in total). This was a pragmatic decision to determine the sample size needed while maintaining power at an acceptably high level.

Primary outcomes

The primary outcome measure was the proportion of patients with a criterion response threshold of a 20% reduction in total PANSS score. This is a commonly used criterion for therapeutic response in schizophrenia trials, which allows for comparison with other published trials18 and inclusion of the study results in any appropriate systematic review or meta-analysis. The PANSS45,46 is a 30-item rating scale designed to provide a comprehensive assessment of psychopathology in adult patients with schizophrenia.

Secondary outcomes

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 sections The addition of study sites to Changes to study medication procedures.

Patient and public involvement

A service user group contributed to the design of the study at the proposal stage. In particular, feedback from qualitative interviews with service users highlighted the relevance of a systematic assessment of side effects, including subjective experience. Once funded, two experts by experience, were involved in the trial throughout: one a service user and the other a parent and carer of a service user. These individuals sat on the Trial Steering Committee and the Trial Management Group, respectively, and, as a result, contributed to both trial oversight and the strategic direction of the trial, including ways to improve the recruitment rate.

Both of the experts by experience who were members of the trial’s oversight committees made additional contributions. These included shaping the content of the website and newsletters, and communicating the importance of the study to health-care professionals through the Clinical Research Network. The last, in particular, was well received and influenced clinician engagement in the trial.

One further issue that was felt to particularly necessitate the perspective of service users was whether or not to introduce remuneration for completing assessments, and we sought the opinion of a wider network of service users through support from the Clinical Research Network on this issue before making a decision.

Participants

People aged 18–65 years with schizophrenia that had proved to be unresponsive, at a criterion level of persistent symptom severity (as used by Honer et al. 30) and impaired social function, to an adequate trial of clozapine monotherapy in terms of dosage, duration and adherence, and who were competent and willing to provide written, informed consent.

Pharmacological intervention

The pharmacological intervention was the augmentation of clozapine treatment with another second-generation antipsychotic, amisulpride, or placebo: 400 mg of amisulpride or two placebo capsules for the first 4 weeks, with the option of titrating up to 800 mg of amisulpride or four placebo capsules for the remaining 8 weeks. The amisulpride and placebo tablets were encapsulated to look identical. A fully automated online (and telephone) randomisation service was provided by the Clinical Trials Research Unit, University of Sheffield, Sheffield, UK. In addition, a 24-hour unblinding service was provided by Emergency Scientific and Medical Services (ESMS Global; Medical Toxicology Information Service Ltd, London, UK).

Data analysis

Health economic analysis

Economic model

The economic within-trial sensitivity analyses explored aspects of the trial design and participants that could affect generalisability. A key issue for the within-trial economic evaluation is the short length of follow-up and whether or not any savings in health-care costs or gains in QALYs may be sustained over a longer 12-month period.

An economic model was constructed using the same target population, perspective, health benefit measure and costs as the within-trial analyses. The intervention arm of the model was defined as clozapine with amisulpride augmentation. The control arm of the model was defined as clozapine with placebo augmentation.

A focused literature search identified no full economic evaluations of amisulpride-augmentation treatment or any clozapine-augmentation treatment that could be used to inform the economic model design or provide data to estimate model parameters. The searches were conducted in September 2015 and were restricted to the previous 10 years (restricted because of relevance in terms of resource-use and health-care service changes, as well as modelling techniques). The databases searched were: NHS Economic Evaluation Database (NHS EED), the HTA database, EMBASE, MEDLINE, PsycINFO and EconLit. The search strategy included treatment terms, for example, ‘clozapine unresponsive’ and ‘clozapine augmentation’ and economic evaluation terms derived from the NHS EED search strategy. Two clinical effectiveness reviews were found in a search of The Cochrane Library for clozapine-augmented therapy. 76,77 These reviews revealed that there were insufficient data about the long-term effects of any augmentation on clinical outcomes or side effects. The reviews included a few studies with a follow-up of up to 1 year, which showed that some patients take longer to respond (responding between 6 months and 1 year into treatment).

Given the paucity of evidence for long-term use of clozapine augmentation, as well as the complexities of the disease area, it was concluded that there is little to gain from building a full economic model. The evidence collected in this clinical and economic trial is the most robust evidence available to inform the model structure and economic parameter estimates. Accordingly, to explore what happens after the 12-week trial period, a partial model was constructed. The model explored the impact of varying the probabilities that participants would experience improvement in symptoms (as measured by the PANSS) and that participants would experience side effects. The analyses were based on the assumption that these probabilities are key drivers of costs and QALYs.

The economic model used a simple Markov structure to model the impact of symptom inprovement and side effects over 12 months, split into four 3-month cycles (Figure 1). The initial cycle was assigned the probabilities, costs and QALYs for amisuplride and clozapine, to mirror the outcome of the trial. The subsequent cycles used pooled trial data to estimate the baseline values. For the initial cycle the mean costs/QALYs for clozapine were used. The net saving/QALYs gain from the primary analysis were applied to the clozapine values to estimate the mean costs and QALYs for the initial cycle of the amisulpride choice.

FIGURE 1.

The Markov model.

All data were entered into the model as distributions. Beta distributions were used for the probability data. Gamma distributions were used for the cost and QALY data. Monte Carlo simulation in TreeAge software (TreeAge Pro Health Care Module 2015, TreeAge Software Inc., Williamstown, MA, USA) (10,000 iterations) was used to estimate the net costs, QALYs and ICER of amisulpride augmentation. Probabilistic sensitivity analysis was used to estimate the likelihood that amisulpride augmentation was cost-effective compared with clozapine and placebo.

Side effects

The data in Table 7 regarding side effect assessment using the ANNSERS-E show that, over the course of the study, the mean ANNSERS-E total score was 3 points higher in those participants assigned to amisulpride than in those in the placebo group. The information in Table 1b shows a greater frequency of cardiac symptoms in the amisulpride group, when checked 7–10 days after starting study medication. The data in Table 5 on plasma prolactin concentration (including following logarithmic transformation, i.e. ln prolactin) reveal that those participants in the amisulpride group had, on average, a much higher level during the study. Plasma prolactin concentration is 24% lower (95% CI 12% to 49%) in the placebo group than in the amisulpride group.

Table 8 shows the data relating to the nature and frequency of adverse events reported during the trial. There were 65 adverse events in 31 participants; more of these events were in the amisulpride intervention group than in the placebo group (47 events vs. 18 events, respectively). Almost half of adverse events in the amisulpride group were in the cardiac, general or endocrine systems. With regard to the nature of the cardiac adverse events in the amisulpride-treated group, dizziness and breathlessness were the most common symptoms, each reported by six participants, with postural dizziness, irregular heartbeat and tachycardia each reported by two participants.

Most of the adverse events reported were mild and resolved. Almost one-third of adverse events in the amisulpride group were either ‘probably’ or ‘definitely’ related to the study medication, compared with a little over one-tenth in the placebo group. Sixty per cent of participants in the amisulpride group had at least one adverse event, compared with 30% in the control group; however, serious adverse events were rare, with one person experiencing a serious adverse event in the amisulpride group and two people experiencing such an event in the placebo group.

Health economics

Within-trial analysis

Covariates for the analyses of cost and quality-adjusted life-years and imputation of missing data

The EQ-5D utility score at baseline was correlated with all the clinical measures except the SES and SOFAS (Pearson’s correlation coefficient, p ≤ 0.05; see Appendix 1, Table 22, for the Pearson correlation coefficients of each analysis). The following variables were selected as baseline covariates for the analyses to estimate nets costs and QALYs. The covariates used in the analysis of QALYs were:

• PANSS

• whether or not the participant was white British

• whether or not the participant lived alone

• age

• baseline EQ-5D visual analogue scale (VAS) score.

The PANSS was included to take into account the relationship between utility and psychosis symptoms (Pearson’s correlation coefficient = –0.304; p = 0.014). In addition, a complete set of baseline data for the PANSS measure was available, and correlated with the clinical measures that were associated with the EQ-5D utility score. Age was included to reflect the fact that health and, therefore, utility typically declines with age and was identified as an important covariate in previous studies of antipsychotic treatment. 60,67 The binary variables to reflect whether or not the participants lived alone or were white British were included as proxies for social isolation. Additionally, whether or not a person is white British may act as a proxy for socioeconomic status, given that most of the trial participants were not in employment at the time of the trial.

The covariates used in the analysis of costs were:

• the costs in the 3 months prior to baseline

• study site

• whether or not the participant was white British

• whether or not the participant lived alone

• EQ-5D utility measure.

Past studies indicate that previous cost is a strong predictor of future cost. 60 Study site was included to reflect the fact that the total cost of treatment also depends on the availability and mix of services in different localities, particularly the relative provision of hospital compared with primary and community services. Whether or not a person lives alone or is white British may indicate a greater need for formal health and social services because of a lack of support, lower access to health, and socioeconomic deprivation. The EQ-5D was included as an overall measure of health and health need.

In addition to the baseline measures described in the preceding paragraph, to impute missing costs the clinical measures with complete data at baseline and follow-up, for participants who completed a follow-up assessment, were included in the imputation models. The variables with missing observations that were imputed were inpatient, outpatient, community, primary care and medication costs and total costs for each assessment.

EQ-5D health status, utility and quality-adjusted life-years

There were no missing EQ-5D observations for participants who completed the 12-week follow-up assessment (n = 57). Tables 9 and 10 present the EQ-5D data for these participants. Baseline and follow-up data, where available, are reported in Appendix 1, Tables 20 and 21. Table 9 summarises the proportion of people with no problems on each of the EQ-5D health domains. Table 10 presents the mean VAS scores, which represent how participants rate their own health. Table 10 also includes the utility scores, estimated from the EQ-5D domains and published population weights. The utility scores are anchored at 0 (dead) and 1 (full health), with some states also valued as worse than death. There are no clear differences between the amisulpride and placebo groups at baseline or follow-up. As noted above, there were statistically significant correlations between the EQ-5D utility score and clinical measures of outcome, including measures of side effects. These indicated the utility score increased with decreased symptoms or lower levels of side effects. The results of the effectiveness analyses described above indicated that amisulpride may be associated with improvements in symptoms, but also increased side effect burden. This may explain the lack of difference in utility scores, which is a measure of overall health.

TABLE 9 - EQ-5D scores and utility values at each assessment: participants with complete utility data

EQ-5D health states	Treatment group
	Amisulpride (n = 32/35)		Placebo (n = 25/33)
	n/N	%	n/N	%
Baseline
No problem with mobility	25/32	78	18/25	72
No problem with self-care	27/32	84	19/25	76
No problem with usual activities	24/32	75	14/25	56
No problem with pain/discomfort	17/32	53	17/25	68
No problem with anxiety/depression	12/32	38	10/25	40
12-week assessment
No problem with mobility	25/32	78	19/25	76
No problem with self-care	25/32	78	19/25	76
No problem with usual activities	23/32	72	13/25	52
No problem with pain/discomfort	18/32	56	16/25	64
No problem with anxiety/depression	11/32	34	11/25	44

TABLE 10 - EQ-5D VAS scores and utility values at each assessment: participants with complete utility data

Assessment period	Treatment group
	Amisulpride (n = 32/35)		Placebo (n = 25/33)
	Mean (SD)	95% CI	Mean (SD)	95% CI
EQ-5D VAS scores
Baseline	58 (26)	52 to 56	59 (22)	53 to 66
12-week assessment	58 (23)	52 to 63	53 (23)	46 to 59
EQ-5D utility values
Baseline	0.64 (0.37)	0.50 to 0.77	0.68 (0.38)	0.53 to 0.83
12-week assessment	0.64 (0.37)	0.51 to 0.77	0.65 (0.36)	0.50 to 0.79

Table 11 summarises the QALY measure for participants who completed the 12-week follow-up assessment and for the sample of participants for whom complete cost and QALY data at baseline and follow-up were available. Again, there is no evidence of differences in QALYs between the two groups.

TABLE 11 - Quality-adjusted life-years: baseline to 12-week follow-up, unadjusted for baseline covariates

QALY	Treatment group
	Amisulpride			Placebo
	Sample size (n/N)	Mean (SD)	95% CI	Sample size (n/N)	Mean (SD)	95% CI
Participants for whom complete QALY and cost data at baseline and follow-up were available	18/35	0.178 (0.141)	0.127 to 0.228	15/33	0.188 (0.131)	0.141 to 0.235
Participants who completed the 12-week follow-up assessments	32/35	0.179 (0.106)	0.142 to 0.217	25/33	0.176 (0.108)	0.137 to 0.214

Service use and costs

Table 12 summarises the use of primary and community-based services and hospital-based services, for baseline and follow-up, for those participants with complete data about the services used at both baseline and follow-up.

TABLE 12 - Service use by participants with complete data at baseline and follow-up

Type of service	Treatment group
	Amisulpride (n = 30)	Placebo (n = 22)
	Mean (SD)	Mean (SD)
Number of visits to community services in 3 months prior to baseline	3.20 (8.42)	7.64 (16.48)
Number of visits to community services baseline to follow-up	2.90 (8.05)	7.18 (15.62)
Number of GP visits in 3 months prior to baseline	0.93 (1.46)	1.59 (2.30)
Number of GP visits baseline to follow-up	0.67 (1.12)	1.23 (1.54)
Number of other primary visits in 3 months prior to baseline	5.57 (9.34)	4.64 (5.04)
Number of other primary visits baseline to follow-up	6.87 (9.66)	4.32 (6.99)
Total mental health outpatient visits in 3 months prior to baseline	0.27 (0.69)	1.82 (3.70)
Total mental health outpatient visits baseline to follow-up	0.63 (1.52)	1.05 (2.80)
Total physical health outpatient visits in 3 months prior to baseline	0.07 (0.25)	0.00 (0.00)
Total physical health outpatient visits baseline to follow-up	0.03 (0.18)	0.00 (0.00)
Length of stay for physical health admission to hospital in 3 months prior to baseline, days	0.00 (0.00)	0.00 (0.00)
Length of stay for physical health admission to hospital baseline to follow-up, days	0.03 (0.18)	0.00 (0.00)
Length of stay for mental health admission to hospital in 3 months prior to baseline, days	15.00 (34.11)	12.27 (31.61)
Length of stay for mental health admission to hospital baseline to follow-up, days	13.83 (36.24)	17.95 (39.14)

Table 13 reports the costs for the different types of services used. Overall, inpatient admissions for mental health care were the highest cost component, although a small proportion of participants had a mental health hospital admission (3 months to baseline – five participants in the amisulpride treatment group and three participants in the placebo group; baseline to follow-up – four participants in the amisulpride treatment group and five participants in the placebo group). Table 14 presents the total costs for participants with complete cost data, as well as the imputed cost for participants who completed the week 12 follow-up assessment. Overall, there are no clear differences in total cost between the amisulpride and placebo groups. The total costs are characterised by high levels of variance and overlap in the 95% CIs.

TABLE 13 - Costs (£, 2014) of services used by participants with complete data at baseline and follow-up

Type of service	Treatment group
	Amisulpride (n = 18), mean (SD)	Placebo (n = 15), mean (SD)
Cost of community services in 3 months prior to baseline	373 (552)	744 (2295)
Cost of community services baseline to follow-up	251 (436)	841 (2282)
Cost of primary care in 3 months prior to baseline	492 (654)	367 (436)
Cost of primary care baseline to follow-up	568 (839)	198 (274)
Cost of daily medications in 3 months prior to baseline	246 (142)	426 (716)
Cost of daily medications baseline to follow-up	309 (204)	232 (139)
Amisulpride baseline to follow-up	27 (5)	0 (0)
Cost of mental health outpatient visits in 3 months prior to baseline	53 (126)	278 (473)
Cost of mental health outpatient visits baseline to follow-up	166 (283)	191 (414)
Cost of physical health outpatient visits in 3 months prior to baseline	0 (0)	0 (0)
Cost of physical health outpatient visits baseline to follow-up	0 (0)	0 (0)
Cost of physical health admission to hospital in 3 months prior to baseline	6 (26)	0 (0)
Cost of physical health admission to hospital baseline to follow-up	0 (0)	0 (0)
Cost of mental health admission to hospital in 3 months prior to baseline	3690 (8490)	1476 (5717)
Cost of mental health admission to hospital baseline to follow-up	2193 (6460)	3378 (8916)

TABLE 14 - Total costs (£, 2014) of health and social care

Total costs	Treatment group
	Amisulpride			Placebo
	Sample size (n/N)	Mean (SD)	95% CI	Sample size (n/N)	Mean (SD)	95% CI
Participants with complete cost data at baseline and follow-up
3 months prior to baseline	18/35	4854 (10,844)	950 to 8759	15/33	3291 (8425)	258 to 6325
Baseline to 12-week follow-up assessment	18/35	3520 (8336)	519 to 6522	15/33	4840 (12,874)	205 to 9476
Participants who completed 12-week follow-up assessments: missing costs imputed
3 months prior to baseline	32/35	5014 (9063)	1801 to 8226	25/33	4616 (8329)	1664 to 7568
Baseline to 12-week follow-up assessment	32/35	3988 (8103)	1116 to 6860	25/33	5330 (10,252)	1697 to 8964

Net costs, quality-adjusted life-years and cost-effectiveness of amisulpride augmentation

Table 15 reports the net costs and QALYs after adjusting for differences in baseline covariates. The data suggest a net saving for amisulpride and a small net gain in QALYs. It is important to note that the net cost and QALY data are associated with wide CIs that cross zero, which suggests that the differences are not statistically significant.

TABLE 15 - Net costs (£, 2014) and QALYs of amisulpride augmentation, adjusted for baseline covariates

SE, standard error.

Analysis	Net cost		Net QALY
	Mean (SE)	95% CI	Mean (SE)	95% CI
Participants who completed 12-week follow-up assessments: missing costs imputed	–£1788 (£1182)	–£4165 to £590	0.0009 (0.014)	–0.026 to 0.028

Figures 3 and 4 show the cost-effectiveness plane and CEAC. Table 16 reports the bootstrap simulations of net cost and QALYs, as well as the results of the cost-effectiveness acceptability analyses. These results show similar net costs and QALYs to the analysis adjusted for covariates. However, in this case, the 95th percentiles of cost indicate a net saving associated with augmentation of clozapine with amisulpride. The cost-effectiveness plane illustrates the scatter of the 10,000 simulated net costs, all of which indicate a net saving. Figure 3 also indicates the distribution of the 10,000 simulated net QALYs, which are evenly spread between net QALY gain and net QALY loss. Overall, there is a net saving with no clear difference in QALYs, with a mean net benefit of £1806 (SD £373, 95th percentiles £1088 to £2530). In this simulation, the probability that amisulpride augmentation is cost-effective is 1.

FIGURE 3.

Cost-effectiveness plane of net costs and QALYs: primary analysis.

FIGURE 4.

Cost-effectiveness acceptability curve, probability that amisulpride augmentation is cost-effective: primary analysis.

TABLE 16 - Primary analysis of the cost-effectiveness of amisulpride augmentation: bootstrapped data

SE, standard error.

Statistic	Participants completing the 12-week follow-up assessments: missing costs imputed
Net (2014) cost (SE) [95th percentiles], £	–1816 (369) [–2540 to –1092]
Net QALY (SE) [95th percentiles]	0.0009 (0.004) [–0.007 to 0.008]
Incremental cost/QALY gained	Amisulpride dominates (net cost saving and no difference in QALYs)
Net benefit (SE) [95th percentiles], £; WTPT = £15,000	1806 (£373) [1088 to 2530]
Probability amisulpride cost-effective if WTPT = £15,000	1.00

Table 17 presents the results of the sensitivity analyses. Overall, these are similar to the results of the primary analysis, which supports a conclusion that amisulpride augmentation of clozapine is likely to be cost-effective. However, this conclusion only applies to the short treatment and follow-up period for the 57 participants who completed a follow-up assessment. In addition, excluding inpatient costs reduces the potential savings associated with amisulpride. If the configuration of available services is an important driver of whether or not a person can access and use inpatient care, the savings estimated in the primary analysis are likely to be overestimated.

TABLE 17 - Sensitivity analyses, adjusted for baseline covariates, bootstrapped data, within-trial analyses

SE, standard error.

Analysis	Net (2014) cost, £ (SE)	95th percentiles, £	Net QALY, (SE)	95th percentiles	Net (2014) benefit, £ (SD)	95th percentiles, £	Probability that amisulpride is cost-effective (WTPT = £15,000)
Sensitivity analyses
Participants with complete cost data	–2011 (1761)	–5463 to 1441	–0.003 (0.027)	–0.057 to 0.050	1671 (1863)	–1016 to 5793	0.82
Excludes inpatient costs	–329 (132)	–587 to 71	0.0009 (0.004)	–0.007 to 0.008	327 (145)	52 to 629	0.99
Excludes costs of daily medications	–1864 (368)	–2586 to –1142	0.0009 (0.004)	–0.007 to 0.008	1855 (372)	1129 to 2575	1.00
Cost per person with a 20% response on the PANSS	–1816 (369)	–2540 to –1092	0.22 (0.19) people with a 20% response	0.155 to 0.586; people with a 20% response	3362 (2824); WTPT = £15,000 to gain one person with a response	–1842 to 882; WTPT = £15,000 to gain one person with a response	0.89
Cost per person with no serious side effects/adverse events	–1816 (369)	–2540 to –1092	–0.39 (0.28)	–0.94 to 0.16	–7563 (4058)	–15,422 to 522	0.04

Because amisulpride augmentation was associated with a higher rate of side effects and adverse events, there was no evidence that it was cost-effective when the cost per serious side effect (any severe non-neurological side effect, any EPS or any serious adverse event) was used as the measure of cost-effectiveness.

The economic model was used to explore the potential of amisulpride augmentation to be cost-effective over the longer time frame of 1 year. The results are discussed in the Economic model section below.

Efficacy

This trial is only the second double-blind, placebo-controlled RCT testing amisulpride augmentation of clozapine in patients with schizophrenia that has shown an insufficient response to clozapine treatment. It has a much larger sample size than the first RCT43 but, nevertheless, it under-recruited, and the power of any analysis to detect significant differences between the active and placebo groups is necessarily limited. Looking at the primary outcome measure, the participants in the amisulpride group had higher odds of achieving the response criterion of a 20% or greater reduction in the total PANSS score by the end of the 12-week study, although this was not statistically significant. This advantage was not evident at 6 weeks, reinforcing the possibility that an adequate trial of clozapine augmentation with a second antipsychotic may be at least 10–12 weeks,18 that is, longer than the 4–6 weeks considered adequate for acute treatment of a psychotic episode. There was some evidence of a greater reduction in the PANSS negative subscale score by 12 weeks in those participants assigned to amisulpride than in those of the placebo group. This finding is in line with earlier reports of a greater improvement in negative symptoms than positive symptoms in randomised studies where clozapine augmentation with a second antipsychotic for treatment-refractory schizophrenia had proved to be beneficial,23,28 as well as some limited evidence for improvement in negative symptoms with amisulpride monotherapy. 79–83

The clinical relevance of a reduction in total PANSS score of 20% or more as a response threshold for people with treatment-refractory schizophrenia may have its limitations. First, it involves the interpretation of scores on a scale rarely used in clinical practice. Second, Leucht et al. 84 showed that even a 25% percentage reduction in the PANSS total score only reflected a reduction of the CGI scale score by one severity step. Thus, given the marked heterogeneity of the clinical presentation of treatment-refractory schizophrenia, future studies might consider adding an individualised response criterion, based on the change in severity of each participant’s critical target symptoms. This last point is reinforced by the finding of a heterogeneous clinical picture in relation to the key symptoms and behaviours presenting in this study sample. Although persistent positive symptoms were the most common symptoms at baseline judged to be of clinical significance by the mental health professionals providing care, some participants presented other such target symptoms, including anxiety and negative symptoms in the avolition/amotivation domain.

Side effects

Amisulpride was chosen for this study because of the robust evidence for tolerability benefits, particularly a low risk of compounding characteristic clozapine side effects such as sedation, weight gain and other metabolic problems and a low risk of EPSs. Amisulpride is more likely than most second-generation antipsychotics to cause hyperprolactinaemia,85 but causes little or no weight gain, has a similarly low risk of causing diabetes and lipid abnormalities and a relatively low liability for EPSs. 86,87 With regard to cardiac side effects, QT interval prolongation and the potentially fatal arrhythmia, torsade de pointes, are not uncommon with overdose,88 but the risk at therapeutic dosages is rather uncertain. 89,90

In our clinical trial, the results of the ANNSERS-E assessments revealed a greater side effect burden with amisulpride augmentation; that is, over time, the mean ANNSERS-E total score for those participants in the amisulpride-treated group was 3 points higher than that for the placebo group. However, the separate assessment of EPSs, such as akathisia and parkinsonism, revealed that these were not treatment-emergent problems with amisulpride augmentation of clozapine, despite previous studies reporting a high rate of tremor, bradykinesia and akathisia with this drug combination. 43,76

Among the participants in the amisulpride group, 60% experienced at least one adverse event over the course of the study, compared with 30% of participants in the placebo group. However, these adverse events were predominantly mild and the vast majority resolved. Serious adverse events were rare. Cardiac symptoms were a relatively common prompt for an adverse event report, occurring more commonly in the amisulpride group. Furthermore, an additional check for any emerging cardiac symptoms in the 7–10 days after starting study medication revealed that shortness of breath and dizziness were more common in the amisulpride treatment group. Amisulpride augmentation was also associated with endocrine effects; the most common was raised plasma prolactin concentration, an expected side effect that also provides some indirect but reassuring evidence of adherence to the study medication.

Mechanism of action

The rationale for the choice of an augmenting antipsychotic in patients on clozapine includes a complementary receptor profile, that is, potent D₂ dopamine receptor blockade. 8,41,91 Amisulpride was chosen for this study because it fulfils this criterion in that it preferentially binds to dopamine D₂ and D₃ receptors in limbic rather than striatal brain structures92,93 and has low affinity for other dopamine receptor subtypes, although it also has affinity for a range of other receptors, including serotonergic, histaminergic and adrenergic receptors.

The limited benefit seen with amisulpride in this study supports the emerging notion that the rationale of potent D₂ blockade for an augmenting antipsychotic to treat clozapine-unresponsive illness may be simplistic. Treatment-refractory schizophrenia may have a more complex pathophysiology than illness showing a good therapeutic response to standard antipsychotic therapy; the underlying pathophysiology may even be non-dopaminergic. For example, dopamine synthesis capacity is lower in those patients with a treatment-resistant illness (indeed, no different from healthy controls) than in those with a responsive illness. 94,95 Vayısoğlu and Anıl Yağcıoğlu96 speculated that treatment-resistant illness may benefit from a multisite receptor effect rather than a stronger antidopaminergic effect. 97

Economic analyses

Overall, the economic analyses indicate that amisulpride augmentation has the potential to be cost-effective in the short term and, possibly, the longer term. The within-trial analysis indicated that amisulpride augmentation is associated with a net saving of £1816 (SD £369, 95th percentiles –£2540 to £1092) and no clear difference in QALYs. Although the extent of any savings is uncertain, because of the wide variation in costs, the cost-effectiveness acceptability analysis indicated that amisulpride augmentation can be cost-effective. The probability that amisulpride was cost-effective estimated from the bootstrap simulation was high with p-values of 0.82–1.00. An economic model extrapolated the results over a longer time frame of 1 year. This found similar net savings, but increased the variance associated with both net costs and QALYs. For the base case, the net saving was –£1935 (SD £13,735, 95th percentiles –£33,908 to £25,968). The 95th percentiles cross zero, indicating less certainty that there is a net saving for amisulpride augmentation; the probability that amisulpride is cost-effective is reduced to 0.54–0.58. However, the extent to which amisulpride augmentation represents value for money from the patients’ perspective depends on their preferences and willingness to accept a potentially higher side effect burden for reduced symptoms.

Recruitment

From early on, the AMICUS trial did not achieve the rate of recruitment that was anticipated and necessary in order to reach the original target of 230 participants. A rescue plan was agreed with the HTA programme, as funders, which included a lower recruitment target of 180 participants based on a reduced power of 80%. However, subsequently the recruitment rate decreased, partly as a result of loss of study momentum following a temporary closure of the study at the original end date. By the close of the extended recruitment period, we had achieved only 38% of the revised target. Negotiation of the hurdles of research governance, regulation, and NHS permissions, contracts and costs allocation meant that opening a single study site often took many months, undoubtedly impacting on our accrual. Nevertheless, even after sites were open to recruitment, referrals to the study were slow, despite feasibility work prior to this trial indicating that an adequate number of patients would fulfil the eligibility criteria.

The slow rate of participant recruitment was discussed by the Trial Management Group, during teleconferences with researchers at the study sites, and by the Trial Steering Committee. Particular consideration was given to barriers to recruitment and how these might be overcome. Some changes were made, such as the addition of a small remuneration for participants, following feedback that other clinical studies were offering this. Another amendment provided the option of an additional prescription of trial medication to allow time for unblinding at the end of the study follow-up period and, therefore, continuity of medication for the participant, an issue that had been raised by clinicians at sites.

Much of the discussion on the slow rate of recruitment revolved around the eligibility criteria and whether or not these were appropriate and pragmatic. Fifteen per cent of participants were excluded during final eligibility screening because their level of residual symptoms and/or impaired social function failed to reach the severity threshold necessary for inclusion. Given that these individuals had been identified by their clinical team as having shown a poor response to clozapine, we considered whether or not it would be pragmatic to include them in the trial. However, as the aim of the trial was to establish the risks and benefits of this augmentation strategy in patients with an indubitable treatment-refractory illness, we decided this was not appropriate. To modify these inclusion criteria would have resulted in a larger sample size and thus an increase in the power of the study, but it would have rendered any effectiveness findings more difficult to translate into recommendations for the management of treatment-refractory schizophrenia in clinical practice.

Another reason for the exclusion of participants was concern about ECG abnormalities or a prolonged corrected QT interval (> 450 milliseconds, following Medicines and Healthcare products Regulatory Agency guidance). Eleven per cent of participants were excluded on these criteria during final eligibility screening, but several further patients did not reach the consent stage as cardiac abnormalities had already been identified. We are aware that, subsequent to not being eligible to enter the trial, a number of these patients were prescribed amisulpride augmentation, raising the question of whether or not they should have been included in this trial as representative of routine clinical practice. However, the exclusion of patients with cardiac contraindications, including long QT syndromes, from exposure to potentially high-dose antipsychotic medication was in line with recommendations for best practice safety monitoring54 and, therefore, it was not acceptable for this exclusion criterion to be removed, despite our awareness that the same conservative approach was not systematically adopted in clinical practice. The greater number of cardiac adverse events seen in participants in the amisulpride arm during the trial reassured us that we had made the right decision.

Feedback from researchers at the sites indicated that many clinical staff felt unable to review their case load and identify potential participants, citing competing clinical priorities, concerns about how introducing a trial to the patient might impact on their therapeutic relationship, or demonstrating a lack understanding of the clinical equipoise of the research question. 101,102 On numerous occasions, clinicians expressed the view that it would be unethical for a patient to be given placebo rather than amisulpride. Furthermore, researchers found some clinical services commonly employed amisulpride augmentation as a pharmacological option for clozapine-refractory schizophrenia, limiting the number of eligible patients for the trial.

As referrals had to come from a member of a patient’s clinical team, recruitment tended to occur in the services in which we encountered sympathetic clinical staff. Thus, many eligible patients in other areas missed the opportunity to participate. We opened additional trial sites over time in order to access more eligible patients, and worked hard to promote the trial with talks to clinical teams, newsletters, and researcher recruitment training days that covered engagement of both patients and clinical staff. However, recruitment failed to achieve the necessary rate for us to attain our target sample size. Indeed, 7 of the 23 sites failed to randomise a single participant.

Conclusions

The risk–benefit balance of amisulpride augmentation of clozapine for schizophrenia that has shown an insufficient response to a trial of clozapine monotherapy is worthy of further investigation in larger studies. This is the first study to provide detailed cost and QALY data about amisulpride augmentation of clozapine treatment. The economic analysis demonstrated that amisulpride augmentation has the potential to be cost-effective in the short term and, possibly, the longer term, but larger prospective RCTs of amisulpride augmentation that were methodologically sound and of adequate duration would be necessary to establish this. Whether or not such trials are feasible remains uncertain, given the continuing challenge of recruitment in mental health studies in the NHS. Nevertheless, this trial provides important process, clinical and economic data to inform the design of future studies.

The analysis of the data in our relatively small trial found that, even among patients with a clozapine-refractory illness, the addition of amisulpride was associated with a greater chance of improvement to a criterion level of overall symptom reduction within 12 weeks and some suggestion of improvement in negative symptoms. However, despite amisulpride being chosen for its favourable tolerability and safety profile, when combined with clozapine treatment in this study it was associated with a greater side effect burden, including cardiac side effects. This may partly reflect the thorough assessment of side effects in this study, which was more systematic and comprehensive than is generally conducted in clinical trials of antipsychotics. 103 These findings have implications for the nature and frequency of safety and tolerability monitoring of clozapine augmentation with a second antipsychotic in both clinical and research settings.

Trial sites and principal investigators

Avon and Wiltshire Mental Health Partnership NHS Trust (Dr Tim Amos).

Birmingham and Solihull Mental Health NHS Foundation Trust (Dr Domingo Gonzalez).

Bradford District Care Trust (Dr Khalid Iqbal).

Camden and Islington NHS Foundation Trust (Professor David Osborn).

Central and North West London NHS Foundation Trust (Dr Mariwan Husni).

Derbyshire Healthcare NHS Foundation Trust (Dr Vineet Singh).

Devon Partnership NHS Trust (Dr Christine Brown).

Greater Manchester West Mental Health NHS Foundation Trust (Professor Peter M Haddad).

Kent and Medway NHS and Social Care Partnership Trust (Dr Ahmed Ibrahim Ismail).

Leeds and York Partnership NHS Foundation Trust (Dr Ranga Rattahali).

Lincolnshire Partnership NHS Foundation Trust (Dr Rameez Zafar).

Manchester Mental Health and Social Care Trust (Dr Zachary Fitzgerald).

Northumberland, Tyne and Wear NHS Foundation Trust (Dr Patrick Keown).

North Essex Partnership University NHS Foundation Trust (Dr Pavel Fridrich).

Nottinghamshire Healthcare NHS Foundation Trust (Dr Hemant Bagalkote).

Oxleas NHS Foundation Trust (Ms Carol Paton).

Plymouth Teaching Primary Care Trust (Dr Jocelyn Dawe).

Somerset Partnership NHS Foundation Trust (Dr Fatin Hussein).

South Staffordshire and Shropshire Healthcare NHS Foundation Trust (Dr Manoj Kumar).

Southern Health NHS Foundation Trust (Dr Amanda Taylor).

Tees, Esk and Wear Valley NHS Foundation Trust (Dr Raj Kumar).

West London Mental Health NHS Trust (Professor Thomas RE Barnes).

Study oversight committees

We thank all the members of the study oversight committees for their valued contributions.

Trial Steering Committee: Professor Stefan Priebe (independent chairperson), Professor Eileen Joyce, Ms Jenny Trite (expert by experience), Professor Tom Jamieson-Craig and Professor David Osborn.

DMEC: Professor Mohammed Abou-Saleh (chairperson), Dr Stuart Cox, Dr Louise Marston and Ms Ulrike Naumann.

Trial Management Group: Professor Thomas RE Barnes (chairperson), Dr Verity C Leeson, Mrs Carol Paton, Professor David Osborn, Mr Kavi Gakhal, Dr Louise Marston, Ms Megan Lawrence, Mr Roger Oliver (expert by experience), Mr George Salaminios and Dr Dilveer Sually.

Contributions of authors

Thomas RE Barnes (Professor of Clinical Psychiatry, Centre for Mental Health, Imperial College London) was the principal grant applicant, contributed to the development of 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, Centre for Mental Health, Imperial College London) 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, Oxleas NHS Foundation Trust) 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.

Louise Marston (Senior Trial Statistician, Department of Primary Care and Population Health and PRIMENT Clinical Trials Unit, University College) cowrote the original grant application, and the development and subsequent amendments of the trial protocol. She led the statistical analysis, and contributed to the interpretation and write-up of the results for the final report.

Linda Davies (Professor of Health Economics, Institute of Population Health, University of Manchester) contributed to the original grant application and the development of the trial protocol. She co-led the health economics analysis and write-up.

William Whittaker (Research Fellow, Institute of Population Health, University of Manchester) co-led the health economics analysis and write-up.

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

Raj Kumar (Consultant Psychiatrist, Tees, Esk and Wear Valley NHS Foundation Trust) was a principal investigator and commented on earlier drafts of the final report.

Patrick Keown (Consultant Psychiatrist, Northumberland, Tyne and Wear NHS Foundation Trust and Honorary Senior Clinical Lecturer, Newcastle University) was a principal investigator and commented on earlier drafts of the final report.

Rameez Zafar (Consultant Psychiatrist, Lincolnshire Partnership NHS Foundation Trust) was a principal investigator and commented on earlier drafts of the final report.

Khalid Iqbal (Consultant Psychiatrist, Bradford District Care Trust) was a principal investigator and commented on earlier drafts of the final report.

Vineet Singh (Consultant Psychiatrist, Derbyshire Healthcare NHS Foundation Trust) was a principal investigator and commented on earlier drafts of the final report.

Pavel Fridrich (Consultant Psychiatrist, North Essex Partnership University NHS Foundation Trust) was a principal investigator and commented on earlier drafts of the final report.

Zachary Fitzgerald (Consultant Psychiatrist, Manchester Mental Health and Social Care NHS Trust) was a principal investigator and commented on earlier drafts of the final report.

Hemant Bagalkote (Consultant Psychiatrist, Nottinghamshire Healthcare NHS Foundation Trust) was a principal investigator and commented on earlier drafts of the final report.

Peter M Haddad (Consultant Psychiatrist, Greater Manchester West Mental Health NHS Foundation Trust and Honorary Reader, University of Manchester) cowrote the original grant application, was a principal investigator and commented on earlier drafts of the final report.

Mariwan Husni (Consultant Psychiatrist, Central and North West London NHS Foundation Trust) was a principal investigator and commented on earlier drafts of the final report.

Tim Amos (Consultant Forensic Psychiatrist, Avon and Wiltshire Mental Health Partnership NHS Trust and Senior Lecturer in Forensic Psychiatry, University of Bristol) was a principal investigator and commented on earlier drafts of the final report.

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.