A contingency management intervention to reduce cannabis use and time to relapse in early psychosis: the CIRCLE RCT

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 09/144/50. The contractual start date was in January 2012. The draft report began editorial review in April 2018 and was accepted for publication in February 2019. 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

Michael King is a member of the Clinical Trial Units funded by the National Institute for Health Research and the Rapid Trials and Add-on Studies Board. Rumana Z Omar is a member of the Health Technology Assessment General Board. John Strang reports grants from Camurus (Lund, Sweden), Martindale Pharma (now Ethypharm UK, Woodburn Green, UK), Mundipharma (Mundipharma International Limited, Cambridge, UK) and Braeburn Pharma (Braeburn Inc., Plymouth Meeting, PA, USA), and other from Martindale Pharma and Braeburn Pharma outside the submitted work. In addition, John Strang has a patent Euro-Celtique SA issued and a patent pending with King’s College London.

Copyright statement

© Queen’s Printer and Controller of HMSO 2019. This work was produced by Johnson et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. 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.

2019 Queen’s Printer and Controller of HMSO

Overall design

The CIRCLE (Contingency Intervention for Reduction of Cannabis in Early Psychosis) trial was a rater-blind, multicentre RCT with two arms. The CM arm received a 12-week CM intervention, as well as a manualised psychoeducation intervention delivered by clinical staff, which represents an optimised version of TAU offered by EIPs in the management of cannabis misuse. The control arm received TAU only. Assessments were performed at the time of consent, at 12 weeks following trial inclusion (at the end of the intervention period) and at 18 months following trial inclusion. The primary outcome was time to relapse, operationalised as an admission to an acute mental health service. A Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) diagram for the trial is shown in Table 1. An initial internal pilot was successfully completed to demonstrate the feasibility of recruitment and of delivering the interventions. Subsequently, the main phase of the trial was approved and received ethics approval in amendment version 4 (approved on 23 July 2013).

Setting

Participants were recruited to the trial via EIP services throughout the Midlands and the south-east of England. Sites were added to the trial until the recruitment target (n = 544) was achieved. A total of 70 teams from 23 NHS trusts took part in the trial, representing a range of geographical and demographic characteristics. Willingness to participate and proximity to trial centres drove site selection, with almost all sites that were approached agreeing to participate.

Participants

The following inclusion and exclusion criteria were used to identify participants.

Measures and research data collection

Trial processes and interventions

Interventions

The optimised treatment-as-usual (OTAU) package was the context in which we assessed the impact of the CM intervention. The CM intervention involved the offer of voucher rewards for cannabis-free urines over 12 weeks to recent cannabis users with early psychosis. We first describe OTAU, which was delivered to both the CM arm and the control arm, and then the CM intervention to be received by the CM arm only.

Optimised treatment as usual (psychoeducation package)

Our accounts of the interventions include text reproduced from Johnson et al. 44 This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

To be confident that we measured the effects of CM, a psychoeducation intervention was delivered to both the CM arm and the control arm with the aim of achieving currently agreed good practice across both the CM arm and the control arm. Guidelines on EIP care recommends that psychoeducation interventions for cannabis should be an important component of routine care, but consultations with EIP managers and staff suggest that this is not often realised in practice, and we did not find routinely used resources for delivering such interventions. Our aim was to create a standardised version of good-quality psychoeducation to be delivered by staff working in EIP teams who were recruiting to the trial. A training manual for delivering the package, and supporting materials, were provided by the research team to clinicians delivering TAU. The intervention was designed to be sufficiently highly structured for staff without high-level clinical qualifications, such as support workers or assistant psychologists, to be able to deliver it competently following brief training.

The TAU was designed to be an individually tailored psychoeducation approach to cannabis use for generic EIP clinicians, which applies general psychoeducation approaches used in first-episode psychosis. 58 The psychoeducation package was developed by the research team through an iterative process, with frequent input from service users. It drew on the psychoeducation package offered in the control arm of a previous Melbourne pilot study of psychological intervention for cannabis use, the Cannabis and Psychosis trial;59 however, it was a novel package developed for this trial. The package comprised six modules that were delivered via a standard personal computer or laptop. Full delivery of all six modules was intended to take approximately 3 hours, normally offered over six regularly programmed sessions of 30 minutes’ duration. The package included a Portable Document Format (PDF) package for clinicians to work through with their clients, which presents information regarding the effects of cannabis, and motivational materials and strategies for harm reduction or for abstaining from cannabis. The package also included video materials, short quizzes, audio files and further information and written records of the modules for the service user to keep. The primary goal of the materials was to deliver information to meet psychoeducation goals, not to act as a psychological intervention. The clinician’s main aim was harm minimisation, with an acknowledgement that, in a young person with psychosis, cannabis abstinence may be required to ensure that no harm is done. The content was based on MI principles, relapse prevention and harm reduction strategies.

The psychoeducation package presented current information on the potential advantages and disadvantages of cannabis use and of cannabis abstinence. To help the participant to make an informed decision about continued use, EIP staff discussed the positives and negatives of cannabis use by exploring its impact on seven areas: family, finance, activity/engagement in work or education, mental health, physical health, legality and social group/friendships. Finally, staff discussed setting goals regarding the young person’s future use of cannabis in the context of harm minimisation, as well as strategies for achieving their goals and avoiding relapse into patterns of cannabis use that compromise those goals.

Contingency management

The CM intervention offered financial incentives contingent on urinalysis results indicating cannabis abstinence. The intervention voucher schedule and rules were adapted from Budney et al. ,60,61 who offered a voucher-based CM intervention for the treatment of cannabis dependence in the general population. The intervention comprised 12 once-weekly urinalysis sessions and was delivered by clinicians supported by the CIRCLE trial research team. At each session, the participant was required to provide a urine sample.

In week 1 of the intervention, details of the intervention were explained to the participant and they were asked to sign an ‘abstinence contract’ indicating that they understood and accepted its rules, and agreed to abide by the test results. In the first week, participants received a £5 voucher for attending and providing a urine specimen, independent of the drug test results, which provided a ‘baseline’ result. From week 2 to week 12, participants were rewarded if their urinalysis results demonstrated abstinence from cannabis use. In the pilot, the value of vouchers increased by £2 each week, contingent on consecutive negative specimens. A bonus voucher to the value of £10 was earned each time two consecutive specimens suggesting abstinence were provided. However, based on feedback from clinicians and participants gained during the qualitative substudy, changes were made to this reward schedule between the pilot and the main trial to simplify it. In the main trial, from week 2 to week 12, assuming participants provided negative samples, the voucher value rose every 2 weeks by £5. No bonus voucher was offered for passing consecutive weeks. In both the pilot and the main trial, participants who abstained from cannabis use for the full duration of the intervention earned £240 (Table 2). This change to the reward schedule made it easier for clinicians to administer and easier for clinicians and participants to understand. Removing the bonus for two consecutive clean urines made it simpler for participants and clinicians to know what reward a participant should receive for passing. In addition, the new reward schedule meant clinicians needed to stock fewer denominations of vouchers. This change to the intervention was included in the trial protocol version 5 (dated 17 June 2013), approved in substantial amendment 4 (approved 23 July 2013).

TABLE 2 - Reward schedule for the CM intervention

Number of negative urines	Reward (£)
1. Attending session	5
2. Negative urine	10
3. Negative urine	10
4. Negative urine	15
5. Negative urine	15
6. Negative urine	20
7. Negative urine	20
8. Negative urine	25
9. Negative urine	25
10. Negative urine	30
11. Negative urine	30
12. Negative urine	35

Failure to attend intervention sessions, specimens suggesting cannabis use or failure to submit a scheduled specimen reset the value of vouchers back to the initial £5. If the participant attended the following week and provided a negative sample, they were rewarded with £10. In the subsequent week, if the participant provided a second consecutive negative sample, the voucher values resumed from the highest previous level of reward. Vouchers were for major supermarkets, including Tesco (Welwyn Garden City, UK), Sainsbury’s (London, UK) and Asda (Leeds, UK), depending on a participant’s preference. They could be swapped in store for other shop vouchers, if preferred.

Urinalysis was performed using a small bench-top analyser (Kaiwood CHR-110; Kaiwood Technology Co. Ltd, Hsin-Shi, Taiwan) capable of providing rapid test results of drug misuse via the concentration of the drug in a urine sample. To perform the analysis, the clinician pipetted a fixed amount of urine into a buffer solution tube to give a 7 : 1 serial dilution. This allowed a standard 50 ng/ml marijuana test cassette placed in the analyser to provide a urine cannabis concentration reading of between 0 ng/ml and 350 ng/ml. Guidelines were provided to all staff delivering the intervention to allow interpretation of the test results. The guidelines set out upper and lower thresholds for urinary tetrahydrocannabinol (THC) based on recommendations from the suppliers of the urinalysis equipment, SureScreen Diagnostics (Derby, UK). Results greater than the upper threshold (350 ng/ml) clearly indicated use within the last week, and so participants with these results were not rewarded. Results below the lower threshold indicated urinary THC concentration of < 50 ng/ml, the accepted standard for detecting urinary THC; participants with these results were rewarded. Within these thresholds, urinary THC was expected to fall each week until it reached the lower THC threshold, which should be reached within 1 month. A participant with a drop in urinary THC would receive the reward. A participant with a result similar to, or higher than, the previous week would fail. A temperature strip on the side of the specimen cup allowed staff to check whether or not the sample had been tampered with.

Participants could arrange in advance to miss scheduled sessions (‘holiday week’) and still receive the reward for that week if they had a valid commitment that prevented them from attending. They could do this on a maximum of two occasions for 1 week only each time. They were still expected to show evidence of abstinence at the following session to receive a reward for the holiday week. If a participant missed the following week or provided a positive sample, no financial incentive was received for the holiday week. Holiday weeks normally needed to be arranged with the clinician performing the intervention no later than at the previous scheduled appointment. However, a dispensation of this rule was offered for unforeseeable events, such as illness, transport problems or being offered a work shift. The intervention was suspended for a maximum of 1 month if a participant relapsed or otherwise lost capacity to consent to participate. If capacity was not regained in 1 month, the intervention did not continue. If a participant failed to attend on multiple consecutive weeks or if contact was lost with the participant entirely, each missed week was counted as a failure to attend. Figure 1 is a decision flow chart for the CM intervention.

FIGURE 1.

Decision flow chart for the CM intervention.

Ethics approval

Ethics approval for the trial was received on 16 March 2012 from the National Research Ethics Service Committee London – South East (Research Ethics Committee reference number 11/LO/1939). Written informed consent was obtained from all participants in the trial. The original consent forms are stored at the collaborating universities (University College London, King’s College London, the University of Sussex, and the University of Warwick) and copies were kept in the patients’ clinical notes. Data were also stored at the collaborating universities.

Sample size

The proposed sample size for the trial was 544 participants. This figure was based on data suggesting a usual relapse rate for cannabis users of around 50% over the trial time frame. 6,12 A 15% decrease in this relapse rate as a result of the intervention was agreed to be of substantial clinical benefit. Using a power of 90% and a significance level of 5%, a total sample size of 460 participants was estimated to be required. This sample size was based on an analysis of time to relapse and would allow detection of a 37% decrease in the hazard of relapse [hazard ratio (HR) of 0.63] in the CM arm using a Cox proportional hazards model. This sample size was calculated using Stata^® version 11 (StataCorp LP, College Station, TX, USA). The sample size was inflated by a factor of 1.06, assuming that each person delivering the intervention sees an average of four service user participants in the trial, and an intraclass correlation coefficient (ICC) of 0.02 for clinician clustering, which gives a total sample size of 488 participants. Finally, the sample size was inflated by 10% to account for attrition for the primary outcome, giving a total sample size of 544 participants.

Statistical methods

All analyses were carried out by the treatment allocated, using all available data (complete case).

The continuous variables were summarised using mean [standard deviation (SD)] or median [interquartile range (IQR)]. Categorical variables were presented as frequencies and percentages.

Clustering by psychoeducation clinician was accounted for in all modelling. Those participants who declined psychoeducation or for whom the psychoeducation clinician was not known were assigned to a cluster of one (i.e. just that participant). Unadjusted ICCs were calculated for all outcomes except the primary outcome because estimation of the ICC for a survival outcome is not straightforward. 62 We estimated approximate ICCs using log-transformed observed event times and also by using the censoring indicator and treating the outcome as binary. We also obtained an approximate estimate of the correlation by fitting marginal generalised estimating equation models.

Logistic regression was used to determine whether or not those who had no secondary outcome data available at 3 months, 18 months or both had different baseline characteristics to those who remained in the trial. This was repeated for each outcome to determine baseline predictors of missingness.

Kaplan–Meier survival curves by randomised groups were used to examine the primary outcome (time to relapse) descriptively. Cox proportional hazards modelling was used to compare the CM arm and the control arm, adjusting for severity of cannabis use (the stratification variable; 1–3 times per week vs. ≥ 4 times per week) at baseline and whether or not the participant was part of the pilot trial. We carried out a number of supportive analyses for the primary outcome:

• including significant baseline predictors of missingness

• excluding those participants who have no secondary outcome data (for 12 weeks and 18 months separately) because those who drop out or die may be different from those who remain in the trial and/or survive

• including those in the main trial only as some minor changes were made to the protocol at the end of the pilot before the main trial commenced, which might mean that the participants behave differently to those in the pilot trial

• controlling additionally for the number of psychoeducation sessions attended (which was offered in both arms of the trial)

• controlling additionally for the number of admissions in the 6 months prior to baseline

• controlling for the same factors as the primary analysis, but using centre as the clustering variable instead of therapist.

Secondary outcomes were analysed separately at 3 and 18 months. Models were adjusted for severity of cannabis use at baseline and whether or not the participant was part of the pilot trial. For the dichotomous outcomes (cannabis-positive urine, engaged in work or study), logistic regression was used. The continuous outcomes (positive and negative symptoms from PANSS), had non-normal residuals and were, therefore, log-transformed and analysed using linear regression models. For count outcomes (number of cannabis days and number of admissions over follow-up), zero inflated Poisson was applied regression as there were excess zeros in all of these outcomes, except number of cannabis days at 12 weeks which utilised Poisson regression. As with other models, these controlled for severity of cannabis use at baseline and whether or not participants were part of the pilot trial.

All secondary outcomes were also analysed after adjusting for predictors of missingness. An additional supportive analysis compared the number of admissions between trial groups. The analysis included participants who were discharged from psychiatric services within the 18-month follow-up and assumed participants did not have any admissions beyond 18 months.

Robust standard errors63 were used in all regression models to account for clustering of participants by psychoeducation clinician.

Results from all supportive and secondary analyses are presented as estimates and 95% confidence intervals (CIs), as specified in the statistical analysis plan. No p-values are presented to avoid the problem of multiple significance testing. All analyses were carried out using Stata^® version 14 (StataCorp LP, College Station, TX, USA).

Qualitative substudy

Following the pilot phase of the trial, a qualitative substudy was performed to inform the main trial by exploring the acceptability and feasibility of the trial design and interventions from the perspectives of both participants and clinicians. The full text of the trial design and results can be found in Appendix 3 and was included in version 1 of the trial protocol (24 October 2011). Following the main trial, further qualitative interviews were performed that were included in protocol version 7 (16 August 2016) and approved in substantial amendment 6 (approval received 22 November 2016). The results of that second qualitative data collection are still forthcoming. In the post-pilot qualitative data collection, data were collected from three groups: one-to-one interviews with 11 participants in the intervention arm, five focus groups with clinicians in EIP teams participating in the CIRCLE trial and one carer of a participant in the intervention arm. Results showed that the intervention had good acceptability among all groups, and all three groups viewed it positively. Participants described it as helpful and beneficial, and said that they would recommend it to others. Recommended changes to trial procedures that emerged from the focus groups with EIP staff included (1) simplifying the CM reward schedule; (2) adding a temperature strip to sample cups to prevent participants submitting adulterated urine samples; and (3) having a dedicated EIP team member to deliver the intervention rather than care co-ordinators, such as a support worker/assistant psychologist. As previously mentioned, these recommendations were adopted ahead of the main trial.

Demographics

The EIP clinicians reported that they had approached 2402 service users to inform them of the trial and seek consent to being contacted by a CIRCLE trial researcher. Following this, initial meetings were held with 2014 service users by the CIRCLE trial researchers, during which the trial was fully explained to potential participants. A total of 551 service users gave informed consent to participate and were then assessed and randomised into the trial.

Table 3 presents baseline characteristics of study participants. More than 85% of participants were male in both randomised arms, with a mean age of aound 25 years (SD 4 years) in both arms. More than half of participants were white and one-quarter were black. A total of 43% of each arm lived with their parents; only 5% of control and 6% of CM arm participants were married or cohabiting. Around one-third of participants were diagnosed with schizophrenia or schizoaffective disorder and half had other types of psychosis. In many participating services, clinicians had significant reservations about making use of schizophrenia as a diagnostic category. One-quarter of participants were engaged in ‘any work or study’ but a large majority had held open-market employment at some point, understood as work that the employee has applied for through an open job market.

Most participants were using cannabis more than three times per week. The PANSS positive symptoms median scores were 12 (IQR 9–17) in the psychoeducation group and 13 (IQR 9–16) in the CM group. The PANSS negative symptoms median scores were 14 in both groups (see Table 5). The rates of alcohol misuse or dependence and of reports of using substances other than cannabis were high (e.g. 47% of the control arm and 52% of the CM arm members reported using cocaine; 36% of the control arm members and 32% of the CM arm members met the criteria for alcohol dependence).

At 3 months, demographics and outcome measures were compared alongside delivery of the intervention. At this point, among the sample participating in the interview, almost one-third of participants were engaged in work or study. A total of 72% in the psychoeducation group and 70% in the CM group had cannabis-positive urine, compared with 80% and 79% at baseline, respectively. The median PANSS positive symptom score was 11 (IQR 8–16) in the control arm and 10 (IQR 8–14) in the CM arm. For PANSS negative symptoms, the median scores were 14 (IQR 11–18) and 12 (IQR 9–17) for the psychoeducation and CM groups, respectively. The number of days using cannabis in the previous 3 months was slightly lower in the CM group than in the psychoeducation group. The median number of psychoeducation sessions attended was higher in the CM group than in the psychoeducation group [6 sessions (IQR 1–6) and 4 sessions (IQR 0–6), respectively] (Table 4). Participants in the CM and control arms attended a median of six sessions and four sessions, respectively, of the TAU psychoeducation intervention. However, 86 participants in the control arm declined the intervention or attended no sessions compared with 63 participants in the CM arm. A mean of £64 in voucher rewards was obtained in the CM treatment and the median number of CM sessions attended was 9 (IQR 3–12 sessions). A total of 46 participants declined the CM intervention, despite their initial consent to randomisation, or did not attend any sessions.

At 18 months, 33% of the psychoeducation group were engaged in work or study compared with 29% in the CM group. A total of 61% of participants in the control arm and 57% of participants in the CM arm had cannabis-positive urine. The PANSS positive and negative median score and the median number of days using cannabis in the previous 6 months were similar between arms. One-third of participants in both arms were admitted to hospital or a crisis house, were seen by a crisis resolution team or attended an acute day treatment service (Table 5).

Outcomes

There was no significant difference in time to admission between the randomised groups (HR 1.03, 95% CI 0.76 to 1.40) (Figure 3). Results from the supportive analyses were similar (see Table 4). The odds of at least one admission over 18 months’ follow-up are slightly higher for the CM group than for the psychoeducation group (OR 1.02, 95% CI 0.70 to 1.48), but approximately 33% of participants in both groups were admitted. The ICC estimates obtained from using the log-transformed event times using the analysis-of-variance approach and the marginal models were similar and produced an ICC estimate of 0.01 after adjusting for randomisation group, severity of cannabis use (the stratification variable) at baseline and whether or not the participant was part of the pilot trial. The binary outcome analysis produced a near-zero estimate of the between-cluster variance. In addition, we fitted the primary analysis model without robust standard errors. The estimates of the standard errors and p-values for the covariates were similar to those obtained when robust standard errors were used, suggesting that the ICC is small and consistent with the estimates of 0.01.

FIGURE 3.

Kaplan–Meier survival curve by randomised group for the primary outcome, time to relapse.

Those in the CM arm who had a full 18 months’ follow-up had a slightly higher rate ratio for number of admissions than the controls [incidence rate ratio (IRR) 1.08, 95% CI 0.75 to 1.54]; this changed little when assuming that those who were discharged from services had no admissions during follow-up or when including predictors of missingness. Those randomised to CM had slightly lower odds of cannabis-positive urine at 3 months and 18 months than those randomised to OTAU (OR 0.86, 95% CI 0.56 to 1.34; and OR 0.84, 95% CI 0.49 to 1.41, respectively). However, those in the CM arm also had lower odds of paid work or study at both 3 months and 18 months. For the log-transformed PANSS positive outcome at 3 months, the CM score is, on average, 7% lower than that of the psychoeducation group (95% CI –14% to 0%). Results for the log-transformed PANSS negative score at 3 months are similar. Illicit substance use other than cannabis was very low at both follow-ups [3-month median days of use = 0 (IQR 0–1), in both groups, 18-month median days of use = 0 in both groups (IQR 0–2 in control; IQR 0–1 in experimental)]. On average, the number of alcohol-using days was 4 (control IQR 0–12 days; experimental IQR 0–15 days) in both groups at 3 months, and 6 days (IQR 0–24 days) in both groups at 18 months.

Mean quality-of-life utility scores of the two treatment groups were compared at 18 months using unpaired t-tests. No significant difference was observed between the two groups using EQ-5D utility scores (95% CI –0.02 to –0.078; p = 0.25). The result was the same when using SF-12 utility scores (95% CI –0.022 to –0.037; p = 0.6).

Economic evaluation

In analyses adjusted for baseline cost, the adjusted mean difference between the two arms was –£4277 (95% CI –£12,662 to £4107; p = 0.32). The null hypothesis of no difference in costs between the two treatment groups cannot be rejected. These results remained the same when the period of 3 to 12 months (including some imputed costs) was excluded, or regardless of the time period tested, as Table 8 shows.

Table 9 shows that costs for inpatient stays were lower for the CM arm (£10,342 vs. £13,247) than the OTAU arm, with large SDs (£34,029 and £35,747, respectively), whereas costs for other items were broadly similar. In the broader perspective, reported benefits received were the largest additional cost.

Outcomes

The EQ-5D tariff over the 18-month follow-up period is detailed in Table 10 and shown graphically in Figure 4. Utility scores are similar between groups over the follow-up period, with the CM arm having a modestly higher mean raw tariff score at baseline (CM, 0.77; OTAU, 0.76) and all the way through to 18 months (CM, 0.83; OTAU, 0.80). When QALYs are calculated, the mean for the CM arm was higher (CM, 1.21; OTAU, 1.15), but this difference was not significant when adjusting for baseline EQ-5D score (adjusted mean difference 0.053, 95% CI –0.0069 to 0.11; p = 0.083). These differences were smaller when looking at other quality-of-life measures. Figure 4 shows the EQ-5D-3L utility scores over time. Figure 5 shows reported EQ-5D visual analogue scale (VAS) scores over time. The adjusted mean difference for the EQ-5D VAS was 1.2 (95% CI –3.0 to 5.3; p = 0.58). Figure 6 shows SF-6D (Short Form questionnaire-6 Dimensions) utility scores, also showing little difference. Again, the adjusted mean difference had a positive sign (0.015), and again this difference was not significantly different from zero (95% CI –0.018 to 0.048; p = 0.37).

TABLE 10 - The EQ-5D-3L tariff, VAS score and QALYs over the follow-up period

VAS, visual analogue scale.

Quality-of-life measure	Randomised group				Adjusted mean difference	95% CI	p-value
	Experimental		Control
	Mean	SD	Mean	SD
EQ-5D-3L tariff
Baseline (n = 548)	0.77	0.26	0.76	0.25	–	–	–
3 months (n = 372)	0.82	0.24	0.78	0.26	–	–	–
18 months (n = 275)	0.83	0.18	0.8	0.23	–	–	–
QALYs – EQ-5D-3L
3 months (n = 371)	0.18	0.05	0.18	0.05	0.0046	–0.01072 to 0.01	0.09
18 months (n = 236)	1.2	0.26	1.2	0.31	0.053	–0.1169 to 0.11	0.083
EQ-5D VAS
Baseline (n = 528)	59	19	59	21	–	–	–
3 months (n = 354)	65	20	65	19	0.32	–3.4 to 4.0	0.87
18 months (n = 277)	67	17	66	19	1.2	–3.0 to 5.3	0.58
SF-12 tariff
Baseline (n = 549)	0.68	0.12	0.68	0.13	–	–	–
3 months (n = 370)	0.71	0.14	0.71	0.14	–	–	–
18 months (n = 277)	0.72	0.12	0.71	0.13	–	–	–
QALYs – SF-12
3 months (n = 370)	0.16	0.025	0.16	0.026	0.00051	–0.0023 to 0.0034	0.72
18 months (n = 235)	1	0.15	1	0.16	0.015	–0.018 to 0.048	0.37

FIGURE 4.

The EQ-5D-3L utility scores by randomised group over time.

FIGURE 5.

The EQ-5D VAS utility scores over time.

FIGURE 6.

The SF-6D utility scores over time.

Cost-effectiveness analysis

Figure 7 shows the cost-effectiveness plane showing bootstrap cost and QALY (EQ-5D) pairs at the 18-month follow-up. There are 10,000 bootstrapped samples shown, with 50% and 95% confidence ellipses surrounding the bootstrapped pairs. The central dot shows the adjusted mean cost difference of –£4277 and the adjusted mean QALY difference of 0.053. This represents an ICER of –£80,772. The bootstrapped replications of costs and outcomes fall with the majority in the south-east quadrant (81.8%), with very few in the north-west (0.6%), and more in the north-east (14.1%) than in the south-west quadrant (3.5%).

FIGURE 7.

The EQ-5D-3L cost-effectiveness plane.

When this is compared with a cost-effectiveness plane using QALYs generated by the SF-6D (Figure 8), we see less variation in QALY differences, with fewer, but still a majority of, replications to the east of the plot. The relevant mean QALY difference with the SF-6D is 0.015, and the ICER is –£284,937. Both plots suggest that CM may be a cost-effective alternative to OTAU. This is also shown in Figure 9.

FIGURE 8.

The SF-6D cost-effectiveness plane.

FIGURE 9.

Cost-effectiveness acceptability curves by QALY method.

The above analysis shows the probability that CM is more cost-effective than OTAU in terms of QALYs at the 18-month follow-up. The fact that 69.5% of bootstrap replications were in the south-east quadrant suggest that CM is more effective and less costly than OTAU. Again, there were few replications in the north-west quadrant (2.9%), and more replications in the south-west (15.8%) than in the north-east (11.7%). There is a high probability (> 85%) that CM is cost-effective regardless of willingness to pay for one QALY, and irrespective of whether or not QALYs are calculated using EQ-5D or SF-6D. The probability is slightly higher using QALYs generated via the EQ-5D.

An analysis using QALYs at 3 months was conducted, with adjusted mean differences not statistically different from 0 when using EQ-5D or SF-12 tariffs. An analysis of self-reported cannabis-free days was also conducted. From 12 to 18 months, the mean numbers of cannabis-free days in each arm were similar: 107 (SD 66) for CM and 107 (SD 67) for OTAU. The adjusted mean difference was 4.1 days (95% CI –10.7 to 18.9; p = 0.58).

A sensitivity analysis including a broader selection of criminal justice costs, benefits payments and productivity losses did not alter the findings from the base-case analysis. The total mean costs in each arm were £21,793 (SD £30,417) for CM and £24,432 (SD £35,179) for OTAU. The adjusted mean difference was –£4896 (95% CI –£13,000 to £3100; p = 0.23). Another sensitivity analysis using multiple imputation by chained equations of individual cost elements did not alter the findings from the base-case analysis. This analysis assumed that cost elements were missing at random. They did shrink the adjusted mean difference in total costs between the two arms from –£4475 in the base case to –£1075. The CIs remained wide (95% CI –£6191 to £4042, p = 0.68). These results are shown in Table 11.

TABLE 11 - Economic evaluation sensitivity analyses

Sensitivity analysis	Randomised group				Adjusted mean difference (£)	95% CI (£)	p-value
	Experimental (£)		Control (£)
	Mean	SD	Mean	SD
Broader perspective of total costs	21,793	30,417	24,432	35,179	–4896	–13,000 to 3100	0.23
Multiple imputation of missing cost elements	17,928	34,076	17,845	33,584	–1075	–6191 to 4042	0.68

Main findings

For the primary outcome, there was no evidence of a difference in time to relapse when the CM intervention was delivered in addition to OTAU. Most secondary outcomes, including cannabis use and occupation, also showed no substantial evidence of an effect on direct comparison between the CM and control arms. However, positive symptoms were slightly lower in the CM arm than the control arm at 3 months. The cost-effectiveness analysis indicated a > 85% likelihood that the intervention was cost-effective, mainly resulting from higher mean inpatient costs for the control arm than the CM arm. Interpretation of this and the improvement in positive symptoms is, however, complex and there was strikingly little difference between the arms on primary and cannabis-use-related outcomes. Thus, our trial is unfortunately another trial in a growing list of carefully designed intervention trials that have shown no clear effect on cannabis use in the context of psychosis.

Other noteworthy findings included a lower relapse rate than anticipated when reviewing relevant literature and carrying out the power calculation: only around one-third in each group required admission to acute care, possibly indicating that a comparatively stable group were recruited to the trial. Although a majority still had a cannabis-positive urine, the number of reported cannabis-using days in the previous 6 months fell from > 100 days to 26 days in each group by the 6-month follow-up, suggesting that, whether through the natural history of the condition or the effects of the psychoeducation package, cannabis use declined in both groups over this period. Engagement with the psychoeducation package was good in both groups, with six and four sessions attended, respectively. Success in completing the CM programme was limited for most, with a mean of £64 per person obtained in voucher rewards from a potential sum of £240. There was no evidence of participants replacing their cannabis use with use of other illicit substances.

Limitations

A number of important limitations constrain interpretation of the findings.

First, we have demonstrated the likely ineffectiveness of a CM intervention at the threshold set for the trial (although with the caveat that health economic findings suggested that cost-effectiveness was probable), which was intended to be substantial enough to be motivating without being coercive and thus ethically problematic. The threshold set was supported by clinicians, service users who were consulted via the patient and public involvement consultation, and experts in the field. However, there is presumably some threshold at which the incentive would change behaviour, at least in the short term – for example, a reward for abstinence of £1M per week would presumably be a more powerful incentive for abstinence, albeit an unaffordable and ethically unacceptable one. The question thus remains of whether or not there is a level of reward that would shape behaviour more convincingly without appearing ethically and economically unacceptable, and, if so, whether or not effects would be enduring and outcomes would be improved in general.

Second, the inclusion of an active control makes it more difficult to interpret results. The relatively good engagement with the psychoeducation package and fall in cannabis-using days across the sample could be interpreted as suggesting benefit from psychoeducation. Although the OTAU intervention was intended to reduce the impact of variations in TAU, it became apparent that it was a much better-developed and more ambitious psychoeducation intervention than was available in any of the participating teams.

Third, our initial intention preceding the pilot was for the CM package to be delivered by care co-ordinators in teams. It became apparent early in the trial that very few were prepared to do so, partly because of time pressures and partly because some feared disruption to therapeutic relationships. Delivery of the intervention thus tended to become part of the responsibilities of additional workers in the teams such as support workers and assistant psychologists. Protocols were, in general, well adhered to, but it is possible that effectiveness might have been greater with more integration into routine appointments.

Fourth, although the use of routine data meant excellent completion of the primary outcome and a fully powered trial, the drop-out rate was greater than anticipated on the interview measures, potentially introducing bias.

Fifth, using acute service use as a proxy for relapse is a pragmatic choice because of the very high completion rate for this measure, but it is limited in that some relapses are likely to be contained without acute service use, especially in the context of EIP care for which a relatively intensive response is often feasible. The health economic results suggest a missed effect is a possibility.

Sixth, although we approached all EIP patients who were identified to trial researchers as potentially eligible by their clinicians and who agreed to be contacted by a trial researcher, it may be that self-selection by patients could have introduced bias. However, this was a pragmatic RCT and it is likely that the trial sample is a good reflection of the range of characteristics of the EIP service users who would enter treatment if the intervention was offered through EIP or specialist drug treatment services. Some of the more common reasons patients gave for declining participation in the CIRCLE trial included that they were ‘not interested’ or ‘not trying to quit at the moment’ or they denied that they were using cannabis. Similar self-selection seems probable if the intervention was to be introduced as part of routine care.

Seventh, the qualitative substudy found that, although clinicians generally viewed the interventions positively, with most seeing them as acceptable, a small number raised ethical concerns about paying people to stop smoking cannabis, which they should not be doing anyway. Such concerns, if widely held among patients, carers, clinicians and the public, are a potential barrier to adopting CM in routine care. However, most of the clinicians who raised such concerns said that they would still engage their clients in the treatment if it helped. In addition, there was no evidence of these concerns being shared by participants or carers, who viewed the intervention positively, although we were able to interview only one carer. Public acceptability of CM, however, is currently unclear as few studies have been published on the topic. One of the studies that did explore this issue, Promberger et al. ,71 reported that, in the UK, reward-based incentive schemes were viewed favourably for treatment adherence in mental illness but were viewed slightly negatively for drug addiction. These results do not have a straightforward implication for the CIRCLE trial as it trialled CM for drug misuse in severe mental illness; however, it would be interesting to investigate this issue further.

Eighth, we did not systematically record the type of cannabis participants were using. It may be that users of high-potency cannabis would have shown more of an effect in the primary outcome if the intervention had been effective at reducing use.

Finally, we focused only on cannabis, but baseline data indicated that the participants had a considerable history of alcohol and substance use disorders. It may be that an effective intervention would need to address the use of additional substances.

Implications for research and practice

The lack of clear effects in the trial means that immediate implications for practice are limited, although the health economic data may suggest that the intervention should not be discounted entirely and may be effective with a different design or reward level.

Regarding further research, it continues to be very challenging to address the substantial problem of cannabis use in psychosis effectively. Regarding further use of CM, effectiveness at a different threshold or with a different threshold is possible. Investigation of stakeholder perspectives on the threshold tested in our trial is desirable before considering this (a PhD linked to the study will include relevant data of this type).

Beyond this, the persistent failure of a variety of approaches to reduce substance misuse in psychosis requires the research community to take stock and reconsider the theoretical basis underpinning approaches tested and the potential for change. Most approaches tried thus far have been relatively narrowly focused on changing cannabis use, with a variety of psychological approaches taken to do this. Behaviour-change models have been deployed relatively rarely and it is worth considering if these might enhance the approaches taken. A more personalised approach taking account of specific individual triggers might also be of value.

However, a more substantial change in approach is also worth considering. Several studies72,73 have noted that young people who have psychosis and problematic cannabis use are often multiply disadvantaged, and often have restricted social networks in which their main sources of social contact are other cannabis users. This may make cannabis use hard to modify without a broader approach to the social context, for example to support isolated young people with comorbid cannabis use to find new forms of social contact and activity, perhaps alongside psychological approaches to reducing substance misuse. The evidence of multiple substance use should also be noted, suggesting that, although cannabis has been identified as the principal problematic substance in EIP contexts, interventions may also need to address other substance use.

Patient and public involvement

Patient and public involvement groups were consulted during the development of the trial, and were involved in the development of the CM treatment and the psychoeducation package. In the course of designing the trial and applying for funding, we presented our plans and modified them in the light of comments to the London Early Intervention Research Network (by clinicians and a carer representative) and to the SURF (Service User Research Forum) at University College London. The psychoeducation package used in both arms of the trial was based on the Cannabis and Psychosis Trial conducted in Melbourne, Australia by Edwards et al. :59 service users from a lived experience group called the Service User Research Forum had been invited on multiple occasions to comment on this package. In the set-up phase of the trial, we again consulted the London Early Intervention Research Network and the SURF on the content of our intervention and on trial plans. Four service users at Camden and Islington Early Intervention Services were also asked to comment in some detail on the CM and psychoeducation plans during the set-up phase of the trial. Further modifications based on qualitative interviews with patients and on discussions with staff were made to the CM schedule between the pilot and the full trial. There were two service user members on the Trial Steering Committee. Had we been designing the package now, it is probable that service user involvement would be more extensive.

Contributions of authors

Sonia Johnson is Professor of Social and Community Psychiatry in the Division of Psychiatry, University College London, and chief investigator for the CIRCLE trial. She managed the primary research centre, UCL, and designed the study with contributions from Steven Marwaha, David Fowler, Tom Craig, Louise Marston, Rumana Omar, Michael King, Tim Weaver, John Strang and Paul McCrone. She also drafted the final report with Luke Sheridan Rains.

Luke Sheridan Rains has been the trial manager for the CIRCLE trial since the completion of the feasibility study. He also drafted the final report with Sonia Johnson.

Steven Marwaha is a Reader in Psychiatry in the Mental Health and Wellbeing Department in the University of Warwick, and lead investigator for the CIRCLE trial in the Midlands. He contributed to study design and managed the other research centres with David Fowler and Thomas Craig.

John Strang is a Professor in Addictions, Head of the Addictions Department and Director of the National Addiction Centre at the Institute of Psychiatry, Psychology, and Neuroscience, King’s College London. John Strang is an expert in substance misuse and has substantial experience of CM trials in the UK. He provided expertise to the study design and delivery.

Thomas Craig is Professor of Social Psychiatry in the Department of Health Service and Population Research at King’s College London. He is the lead investigator for the CIRCLE trial in south London. He contributed to study design and managed the other research centres with Steven Marwaha and David Fowler.

Tim Weaver is Associate Professor in Mental Health in the department of Mental Health, Social Work and Interprofessional Learning, Middlesex University, London. He has significant experience of CM trials in the UK and is an expert in qualitative research methods. He provided expertise to the study design and delivery.

Paul McCrone is Professor of Health Economics in the Department of Health Services and Population Research, King’s College London. He contributed to study design, authored the cost-effectiveness analysis and, with Jonathan Spencer, performed the analysis.

Michael King is Professor of Primary Care Psychiatry and joint director of PRIMENT Clinical Trials Unit in the Division of Psychiatry, University College London. He contributed to study design, statistical analyses, and the interpretation of the findings through the PRIMENT clinical trial unit.

David Fowler is Professor in Psychology in the University of Sussex and Lead Investigator for the CIRCLE trial in Sussex and the surrounding areas. He contributed to study design and managed the other research centres with Steven Marwaha and Thomas Craig.

Stephen Pilling is Professor of Clinical Psychology and Clinical Effectiveness in the Division of Psychology and Language Sciences in University College London.

Louise Marston is Senior Research Statistician in the Department of Primary Care and Population Health, University College London. She contributed to study design. She also authored and performed the statistical analyses with Rumana Z Omar.

Rumana Z Omar is a Professor of Medical Statistics at the Department of Statistical Science, University College London. She contributed to study design, authored and performed the statistical analyses with Louise Marston.

Meghan Craig was the trial manager for the CIRCLE trial during the feasibility study. She managed the pilot phase of the trial and developed the psychoeducation package with Mark Hinton.

Jonathan Spencer is a Health Economist based in the Department of Health Services and Population Research, King’s College London. He performed the cost-effectiveness analysis with Paul McCrone.

Mark Hinton is a Clinical Psychologist with expertise in substance misuse and CM. He was instrumental in conceiving and developing the CIRCLE trial, and was largely responsible for the development of the contingency management intervention. He developed the psychoeducation package with Meghan Craig.

Publication

Sheridan Rains L, Marston L, Hinton M, Marwaha S, Craig T, Fowler D, et al. Clinical and cost-effectiveness of contingency management for cannabis use in early psychosis: the CIRCLE randomised clinical trial. BMC Med 2019;17:161.

Data-sharing statement

All data requests should be submitted to the corresponding author for consideration. Access to available anonymised data may be granted following review and appropriate agreements being in place.

Patient data

This work uses data provided by patients and collected by the NHS as part of their care and support. Using patient data is vital to improve health and care for everyone. There is huge potential to make better use of information from people’s patient records, to understand more about disease, develop new treatments, monitor safety, and plan NHS services. Patient data should be kept safe and secure, to protect everyone’s privacy, and it’s important that there are safeguards to make sure that it is stored and used responsibly. Everyone should be able to find out about how patient data are used. #datasaveslives You can find out more about the background to this citation here: https://understandingpatientdata.org.uk/data-citation.

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 and Social Care. 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 and Social Care.

Introduction

The pilot phase of the trial successfully recruited 62 participants and featured 11 EIP teams. Following the pilot study, a qualitative study was performed to explore the usefulness and acceptability of the intervention from the perspective of participants and of clinicians, and potential mechanisms for change.

Methods

Results

Results from the qualitative interviews are presented by group.

1. Demographics

(PDF download)

2. Positive and Negative Syndrome Scale

(PDF download)

3. SF-12

(PDF download)

4. Client Service Receipt Inventory (CSRI)

(PDF download)

5. Timeline Follow Back (TLFB)

(PDF download)

6. Structured Clinical Interview for the DSM-IV

(PDF download)