Trial design
The study design is a parallel-group randomised controlled trial, with 1:1 allocation. Participants with distressing persecutory beliefs who meet the inclusion criteria (see below) will be independently randomised to receive either the SlowMo intervention added to TAU, or TAU. Independent randomisation (centrally administered independently of the trial team by the King’s Clinical Trials Unit (CTU)) will use an online system generating randomly varying permuted blocks, stratified by site and baseline paranoia severity. Stratification by paranoia severity will use a median split of ≥ 62 (Green Paranoid Thoughts Scale (GPTS) part B [36] based on data from [10]). Research workers will be blind to therapy allocation, to facilitate completion of unbiased and objective assessments. Adherence to the blindness procedure will be supported by the research coordinator and therapists having responsibility for the randomisation process and informing participants of randomisation outcome. Further, the blinding procedure will be explained to participants and they will be reminded not to inform research workers of therapy allocation. Breaks in blinding will be monitored and recorded. Embedded within the design will be measures to elucidate how the treatment works. For reporting the trial, the CONSORT (Consolidated Standards of Reporting Trials; http://www.consort-statement.org/) Statement will be followed, with consideration of the mHealth evidence reporting and assessment (mERA) [37] and CONSORT-EHEALTH Checklists [38]. For the protocol, the SPIRIT (Standard Protocol Items: Recommendations for Interventional Trials [39]) Checklist and Figure are provided in this paper see: Additional file 1.
Participants
The inclusion criteria are as follows: aged 18 years and over; persistent (3 + months) distressing paranoia (as assessed using the Schedules for Clinical Assessment in Neuropsychiatry (SCAN, [40]) and scoring > 29 on the GPTS, part B, persecutory subscale [36]; diagnosis of schizophrenia-spectrum psychosis (F20-29, ICD-10 [41]); capacity to provide informed consent; sufficient grasp of English to participate in informed consent process, assessments and interventions.
Criteria for exclusion are as follows: profound visual and/or hearing impairment; inability to engage in the assessment procedure; currently in receipt of other psychological therapy for paranoia; primary diagnosis of substance abuse disorder, personality disorder, organic syndrome or learning disability.
Mobile ownership is not a criterion for participation, as android smartphones with the SlowMo mobile app will be provided.
Participants will be recruited from mental health services across three main trial sites in England with the same procedures followed at each site: South London and Maudsley NHS Foundation Trust, Sussex Partnership NHS Foundation Trust and Oxford Health NHS Foundation Trust. Up to six additional Patient Identification Centres, comprising NHS trusts geographically near to the main recruitment trust sites, will be used as required.
Trial flowchart
Figure 1 illustrates the trial/recruitment flowchart.
Planned trial interventions
SlowMo therapy consists of eight individual, face-to-face sessions, of 60–90 min, delivered by trained therapists within a 12-week timeframe, assisted by a web-based app hosted on a touchscreen laptop, with interactive personal accounts and tasks. Initial sessions involve building the meta-cognitive skill of noticing thoughts and thinking habits (visualised as bubbles spinning faster or more slowly). People learn that everyone thinks fast at times, and this can be useful. However, thinking slowly can be helpful in dealing with stress and fears about other people. This key principle frames the sessions in which people are supported to try out tips to slow down for a moment, e.g. by considering the impact of mood and past experiences on concerns and by looking for safer alternative explanations. There is an emphasis throughout the intervention on practising the skills inside and outside sessions. Participants build confidence in managing paranoia, feeling safer in their daily life and working towards a valued goal. The overall session structure is fixed, but individual content is personalised throughout as participants record their individual worries, ways of feeling safer, key learning from each session, and a message for the week ahead. All of the personalised session content is synchronised with a native mobile app installed on a standard android smartphone to assist therapy generalisation into daily life. This allows people to notice their fears and thinking habits, and supports them to slow down for a moment, by providing strategies, encouraging them to audio- or text-record helpful new information and to generate safer thoughts. Recorded information is stored in a format whereby, when experiencing recurrent concerns, people can readily access what was previously useful. Optional notifications are available if people wish the app to check-in with them. The app is specifically designed for offline use, to minimise concerns about privacy and security. Participants are not given standardised instructions about when to use the app, rather the emphasis is on tailoring usage according to what is most helpful for the individual. Use of the app is monitored objectively through data input and system analytics. Please see Fig. 2 for an overview of the main SlowMo screens.
The development work has been done by Evolyst Ltd., a user-centred and evidence-based health care software development company. The design and development of the app has been informed by the British Standards Institute quality criteria and code of practice for health care apps (BSI; [42]). SlowMo uses a proprietary software platform developed using an Azure-based WCF (Windows Communication Foundation) Web Service, acting as an Application Programming Interface (API) to a Model View Controller (MVC) Asp.Net Web application; and a Xamarin.Android-based mobile application, allowing for use of the full Microsoft Stack and negating interoperability issues. SlowMo has currently been developed as a standalone product, given the lack of consensus on operating systems across the NHS trusts, and current interoperability issues.
TAU is care delivered to both randomised groups, according to national and local service protocols and best practice guidelines (specifically, NICE guidance on community mental health treatment for people with psychosis and the standards of community care required by the national regulators). Participation will not alter usual treatment decisions about medication and additional psychosocial interventions which remain the responsibility of the clinical team. A modified version of the Client Service Receipt Inventory [43] will be used to measure service use. Antipsychotic medication data will be extracted from medical records and dosages converted into chlorpromazine equivalents.
Assessments and follow-up
Assessment of efficacy
Participants will complete a range of self-report and interview-based measures to assess the impact of the interventions on primary and secondary outcomes, the hypothesised mediators, and other key processes implicated in paranoia and response to therapy. Assessments will be completed at baseline, 12 and 24 weeks. Every effort will be made to ensure that data collection and completeness is optimised throughout the trial, and to minimise attrition/loss to follow-up. Please refer to Fig. 3 (SPIRIT Figure) for details of assessment at each visit. Assessments will be audio-taped (after first establishing consent) to allow evaluation of adherence to the research protocol and assessment ratings.
The primary outcome is paranoia severity measured by the GPTS [36] over 24 weeks. The GPTS comprises two scales assessing thinking relevant to paranoia: ideas of social reference and persecution, rated over the preceding month. Each item is scored on a five-point Likert scale from 1 (‘not at all’) to 5 (‘totally’). A total score can be calculated ranging from 32 to 160, with higher scores reflecting higher levels of paranoia. Two 16-item subscales assess ideas of social reference (part A) and persecution (part B) relevant to paranoia.
Other paranoia outcomes:
-
1.
The Psychotic Symptom Rating Scales-Delusions (PSYRATS-Delusions; [44]), consisting of six items which assess the following dimensions of delusions: amount of preoccupation with delusions, duration of preoccupation with delusions, conviction, amount of distress, intensity of distress and disruption to life caused by beliefs
-
2.
The persecutory delusions and ideas of reference items from the Scales for Assessment of Positive Symptoms (SAPS; [45]), a semi-structured interview designed to assess the positive symptoms of psychosis
Hypothesised mediators are measured by changes in fast thinking assessed by:
-
1.
Possibility of Being Mistaken (taken from the Maudsley Assessment of Delusions Schedule (MADS; [46]); Alternative Explanations from the Explanations of Experiences interview [47]). These are commonly used published methods of assessing lack of belief flexibility relating to delusional beliefs
-
2.
Jumping to Conclusions (JTC) Beads Data-gathering Task [48] versions 85:15 and 60:40.
Please refer to Fig. 3 for details of secondary outcomes and other key processes hypothesised as moderators; these include published and established measures of wellbeing [49], quality of life [50], self and other schemas [51], service use [43], worry [52], cognitive tests [45, 53–55], other paranoia measures [56] and measures of beliefs about mental health problems and perceived relationship with carers [57, 58].
Safety and adverse event assessment and monitoring and stopping rules
The occurrence of adverse events (AEs) will be monitored actively and systematically, following SPIRIT guidance for reporting of harms. AEs include: deaths; self-harm; serious violent incidents; complaints about therapy; and referrals to crisis care or admission to psychiatric hospital during therapy. A standard method of reporting will be employed, categorising events by severity (five grades, A–E). Subject to the approval by the independent chairperson of the Data Monitoring and Ethics Committee (DMEC, see below, ‘Research governance’), investigators will also determine whether an event is temporally related to the intervention, and whether it is unexpected or unexplained given the participant’s clinical course, previous conditions and history, and concomitant treatments. Following [59], the event will then be rated within five categories from ‘not related’ to ‘related’. Any associations between AEs and the SlowMo hardware or software will also be recorded. At each meeting of the DMEC, or at any time at the request of the DMEC chairperson, a full report of AEs will be reviewed. The DMEC will be responsible for investigating further, if there are concerns about unexpectedly high rates of AEs. This may involve the DMEC members being unblinded to the trial condition or seeking further data on AEs. If there are any ethical or safety reasons why the trial should be prematurely ended, they will advise the Trial Steering Committee (TSC) accordingly.
Individual participants will have the right to withdraw from the trial at any time. In addition the therapist, in collaboration with the participant and relevant clinical team, may decide to stop the therapy if it is directly associated with a worsening of mental state. Reasons for withdrawal from the study will be recorded. For the final reports of the trial, the numbers, types and severity of AEs by trial condition, as well as discontinuations, will be reported, using descriptive statistics (since there are no pre-specified hypotheses concerning AEs or harms, and, given the expected low frequency of AEs, the data will not be suitable for an intention-to-treat (ITT) statistical analysis).
The trial may be prematurely discontinued by the sponsor or chief investigator on the basis of new safety information or for other reasons given by the DMEC, the TSC, the regulatory authority or the Ethics Committee concerned. The trial may also be prematurely discontinued due to lack of recruitment or upon advice from the TSC, which will advise on whether to continue or discontinue the study and make a recommendation to the sponsor. If the study is prematurely discontinued, active participants will be informed and no further participant data will be collected.
Accessibility, usability and acceptability assessments
Given the novelty of the digital therapy platform, its accessibility, usability and acceptability will be assessed in the SlowMo arm. This will be done through assessment of current mobile use and confidence at the beginning of therapy, monitoring of connectivity for the web app, system analytics data on the use of the platform, the User Experience Survey (adapted from [35]), and a service-user led qualitative interview with a sub-sample of those receiving SlowMo (n = 20).
Therapy adherence assessments
In the SlowMo arm, therapy adherence will be assessed from the number and duration of sessions attended, and system analytics data on mobile app use. Therapy delivery will be evaluated in terms of fidelity to the treatment manual.
Data management and security
All data will be anonymised at source. All personal data will be kept in a locked filing cabinet in a locked office and will be kept separate from all the research data. Therapy files will be kept in a secure office in the clinic and will not be accessible to the staff collecting the research outcome data. Data will be entered on a computerised database, held centrally and managed by King’s College London CTU, by research assistants using a secure network connection. Audio-recording equipment will be used to record assessments to check fidelity to assessment protocols and to ensure interrater reliability. The therapy sessions will be audio-recorded (with participant consent) for monitoring the intervention in terms of fidelity and competence. These audio files, named with a unique participant identifier, will be stored as computer files on secure NHS/university servers.
Security and privacy of information stored on the app has been considered throughout its development. If informed consent is provided, app data will only be synched during therapy sessions, over secure connections and stored on a password-protected, secure database. Data transferred will only contain a name (chosen by the person) and a Unique Device Identifier (UDID) which is generated automatically by the therapy platform, and will match the anonymised participant number. Participants can also opt to use the app in a fully offline mode. Participants will have the opportunity, if they wish, to password protect the handset with a pin number or password. During the informed consent process potential participants will be made fully aware of the data collected by the platform, and how data will be stored and used. Access to this privacy and security information is also available from the settings menu of the app, which consenting participants can access at any time.
Data quality
Data quality will be ensured by close monitoring and routine auditing for accuracy throughout the data collection period. In order to ensure the accuracy of the data entered into the database, the main outcome measure entry will be checked for every participant by comparing the paper record with that on the database. An error rate of no more than 5% is acceptable. This will be done once all possible assessments for each time point have been completed. If the error rate is higher than 5%, advice will be sought from the trial statistician and methodologist regarding further data checking.
Sample size
Recruitment of 360 participants will be split equally across sites. We have powered the study conservatively to detect a clinically meaningful 10-point reduction in the primary outcome measure (GTPS [36]); based on a standard deviation of 25, this is a 0.4 effect size [10]. We account for: clustering in the SlowMo arm with an intraclass correlation coefficient (ICC) = 0.01 with 10 therapists (no clustering in the TAU arm), 1:1 allocation, 0.05 significance level. Calculations used Clsampsi in Stata. A simple two-tailed t test with 150 people per group gives 90% power to detect an effect size of 0.4, and 80% for 0.35. In practice, power will be increased by using multiple regression. To allow for conservatively high 20% attrition we will recruit 360 patients at baseline split equally across three sites (120 per site, 60 per arm per site). For the mediational analyses, a sample of N = 300 has > 80% power to detect a proportion mediated of 40%, and > 70% power to detect a proportion mediated of 30%, corresponding to findings in our pilot work [12] (calculated using PowerMediation in R).
Statistical analysis
We will report all participant flow, and analyses will be conducted on the ITT population: all participants will be randomised regardless of non-compliance with protocol or withdrawal from the study. Analyses will post-date final follow-up assessments, with due consideration of potential biases from loss to follow-up. The primary analysis will test for a treatment effect on the primary and secondary clinical outcomes. Random effects regression models allowing for clustering by both participants and therapists will be fitted to the repeated measures, controlling for treatment site, baseline paranoia severity and the corresponding baseline assessment for the outcome under investigation. We will allow for missing outcome data under the Missing At Random assumption [60]; we may also use inverse probability weighting to adjust for non-adherence to allocated treatment and other intermediate outcomes as predictors of future loss to follow-up [61]. Secondary analyses will test treatment-effect mechanisms, moderation and process/adherence effects using modern causal inference methods [62, 63]. The trial outcomes will comprise two parallel series of longitudinal data: one for the putative mediators (M) and one for the clinical outcomes (Y).
For the mechanistic analysis, to test for a treatment effect on the putative mediator, we will replace the clinical outcome with the mechanistic variable as the dependent variable in the random-effect models. If we separately demonstrate a treatment effect on both the putative mediator and on the clinical outcome, we will evaluate mediation in these parallel longitudinal data sets through the use of parallel growth curve and latent change models [64, 65]. These models preserve the basic mediation model by replacing observed variables with latent constructs – the growth factors driving the temporal responses, M1 to Mp and Y1 to Yp. Importantly, the mediational structure only applies to the slope growth or change factors since randomised treatments are independent of the intercept growth factors (baseline values). Growth curve and latent change models can be estimated by maximum likelihood and other methods using the software package Mplus [66]. The application of these methods to mechanism evaluation within EME (Efficacy and Mechanism Evaluation) trials is illustrated in [62]
The aim of these analyses is to demonstrate that the effect of treatment on the growth (change) in the clinical outcome (Y) is explained (caused) by its effect on the growth (change) in the mediator. The major challenge to a valid inference is that there may be confounding of the mediator and outcome. We will begin by allowing for baseline values of the mediator and of the clinical outcome, as in the analyses of the successful EME Worry Intervention Trial [10]. We will then check the sensitivity of the results to the possibility of hidden confounding (unmeasured variables) through the use of instrumental variable methods [62, 63].
Research governance and patient and public involvement (PPI)
King’s College London is the research sponsor and the South London and Maudsley NHS Foundations Trust is co-sponsor. The trial has received a favourable ethical opinion from Camberwell St. Giles Research Ethics Committee (REC) (REC Reference: 16/LO/1862; IRAS: 206680). Any changes to the study protocol will be submitted to the REC and then communicated to all relevant parties (including the DMEC, TSC and study funders). The trial will be conducted in compliance with the principles of the Declaration of Helsinki [67], the Medical Research Council Guidelines for Good Clinical Practice [68] and in accordance with all applicable regulatory requirements including but not limited to the Research Governance Framework and the Mental Capacity Act 2005 [69]. The chief investigator (CI) will have overall responsibility for the trial data set and will permit trial-related monitoring, audits and REC review by providing the sponsor(s), and REC direct access to source data and other documents as required. A dedicated trial coordinator post will assist in the day-to-day management of the project reporting to the CI. A Trial Management Committee (TMC) will meet monthly: its membership will include the investigators and the trial coordinator and site coordinators. It will be chaired by the CI and will manage the day-to-day running of the study and oversee the preparation of reports to the TSC and DMEC. The TSC will meet at least annually and will include in its membership a lay member and access to consultation with a patient and public involvement (PPI) advisory group. The TSC’s purpose is to provide independent overall supervision of the trial, approving the protocol and amendments, and monitoring progress, through audits of recruitment and data completion rates and adherence to the protocol. It will provide independent advice on all aspects of the trial. A DMEC will be convened and will meet at least annually and report to the TSC. It will have access to all trial data and will receive regular reports on AEs. Membership of the DMEC will be fully independent of the trial team and will comprise two independent clinician researchers, one of whom will act as chair, and a statistician who will be independent of the applicants and of the TSC. The DMEC chair will be notified of any serious AEs as they occur, and with the DMEC will consider whether any interim analyses are warranted, review data and advise the TSC on any ethical or safety reasons why the trial should be prematurely ended. The PPI Advisory Group will advise on and contribute to recruitment, qualitative data collection and dissemination activities throughout the trial.