A pilot randomised controlled trial of community-led ANtipsychotic Drug REduction for Adults with Learning Disabilities

Primary objectives

The primary objective was to evaluate the impact of a blinded antipsychotic medication withdrawal programme for adults with LDs without psychosis compared with treatment as usual. More specifically, we wanted to determine whether or not reduction or withdrawal of antipsychotic medication prescribed for challenging behaviour without psychosis could be safely achieved without a corresponding increase in aggression, as indicated in previous non-blinded studies. The primary outcome (aggression) was to be assessed at baseline and 9 months (blinded) with levels of aggression compared between the intervention (reduced medication) and control (standard treatment) arms.

Secondary objectives

A secondary objective was to explore the potential non-efficacy-based barriers to drug reduction in clinical practice. We aimed to complete qualitative telephone interviews with principal investigators (PIs), carers and participants to explore their perceptions of involvement in the trial and medication usage, in addition to a final assessment of medication dosage at 12 months.

Inclusion criteria

Adults with a LD were eligible for the trial if they met all of the inclusion criteria and none of the exclusion criteria. Inclusion criteria were that the patient:

• was aged ≥ 18 years

• had a recognised LD as judged by administrative classification (e.g. on LD register, in receipt of an annual LD health check or in receipt of LD services)

• was currently prescribed risperidone or haloperidol for the treatment of challenging behaviour.

Exclusion criteria

Patients were excluded if:

• they had a current diagnosis of psychosis

• they had had a known recurrence of psychosis following previous drug reduction in the past 3 years

• the clinician primarily responsible for their care judged for any other reason that participation in a drug reduction programme may be contraindicated

• the research team were unable to identify an appropriate individual to complete outcome assessments.

Informed consent

It was expected that although some participants would have the capacity to give informed consent, there would also be a proportion judged by researchers to lack capacity. In such cases, consent from a personal legal representative was sought instead (failing that, from a professional legal representative). Assessment of capacity was made by members of the trial team or the research network, who were professionals with considerable experience in assessing capacity in this population. It was permitted for assessments of capacity and consent to be undertaken by the PI at site if necessary. Criteria for consent included presumption of capacity, an assessment of the individual’s understanding of the risks or benefits of taking part in the trial, their ability to retain this information and their ability to communicate their decision-making. Potential participants were given a plain language and pictorial participant information sheet at least 24 hours in advance of their meeting with the trial team that they could go through with a carer or legal representative (as appropriate) in their own time and at their own speed.

Potential participants with capacity

Upon meeting with the researcher, the trial and potential risks and benefits were explained verbally in simple terms. The researcher checked frequently that the potential participant understood the explanation. Once all questions had been answered individuals who indicated that they were happy to take part in the trial were asked to tick or initial each statement on the consent form as a means of indicating their consent and to sign the form. This process was witnessed and signed off by a carer who was independent of the research team. Participants could decide to withdraw their consent at any stage. A small sample of participants with capacity were also invited to take part in a qualitative interview at the end of the study. Capacity was assessed again at this time.

Potential participants who lacked capacity

When capacity was judged by an experienced researcher to be lacking, a similarly straightforward explanation of the trial and its potential risks and benefits was given verbally to a personal legal representative or, failing that, to a professional legal representative. Neither the personal nor the professional legal representative was connected with the conduct of the trial (e.g. the PI). That individual was asked to give consent on the participant’s behalf. Again, consent could be withdrawn at any stage. Legal representatives were kept informed of all material changes to the trial or participant’s condition to enable them to exercise their right of reviewing the person’s participation in the trial.

Carers of potential participants/principal investigators

The participant’s main carer was also asked to give separate consent to complete assessments designed to be completed by a third party and to consent to taking part in the qualitative interviews at the end of the trial if selected. PIs were also asked to consent to participate in the qualitative interviews.

Risks and expected benefits

Risk was considered against the recognised risks of long-term antipsychotic medication and, therefore, the potential benefits of withdrawal. Benefits would include reduction of cardiovascular risk, in particular stroke, reduction of musculoskeletal risk from tardive dyskinesia and other extrapyramidal side effects, reduction in acute life-threatening risk of malignant neuroleptic syndrome and a broad spectrum of psychosocial benefits from reduction of sedation, associated alertness and concentration and learning. Societal benefits would include increased contribution from adults with LDs who are not constrained by unnecessary medication and reduced expenditure/resource use on unnecessary treatment and medical complications of long-term antipsychotic medication use. However, withdrawal may be associated with the following risks.

• The emergence of tardive dyskinesia. Advice for PIs on the recognition, assessment and management approaches was included in the detailed treatment and safety package prepared by the trial team.

• Emergence of unrecognised psychiatric illness. There remained a slight possibility that, especially in the case of those on very long-term antipsychotics, the drugs masked an underlying mental illness. This, if present, was most likely to be an anxiety disorder. Advice for PIs on the recognition and assessment of psychiatric symptoms was included in the detailed treatment and safety package prepared by the trial team. A clinical algorithm (described in Interventions) was developed to support the primary care team to follow the appropriate treatment and care pathways. Clear guidance was available for predicted scenarios in which unblinding may be necessary, such as the emergence of psychotic symptoms.

• Deterioration in behaviour. A previous study39 showed that measurable behavioural deterioration was uncommon following drug reduction, but other studies have shown greater deterioration and that carer concern can be high. Advice on assessing a meaningful behaviour change was provided in the PI support package. As behavioural signs and psychiatric symptoms for this population are intertwined, the clinical algorithm referred to above also dealt with behaviour change.

Supporting secondary care services

In the case of individuals recruited through general practice and who had involvement with LD services, contact was made with these teams regarding confirmation of eligibility. At this time, a description was given of the study protocol, the PI support package (see Interventions) and the procedure for accessing the code break.

It was not expected that the study would have a considerable impact on the current well-developed specialist LD services. These services would most probably already be aware of many of the individuals involved in the study, and we estimated that the chance of severe deterioration would be small and would be distributed across at least six health boards. It was possible that the study might have increased the referrals to LD services because of a greater awareness of the issue of antipsychotic drug prescribing across primary care. Such referrals would be a positive outcome; LD teams are skilled in drug assessment, and regular review is a key component of good clinical care.

Supply of blinded medication

In order to achieve effective blinding, medication was encapsulated. Risperidone and haloperidol tablets of varying doses were encapsulated based on estimates of the likely numbers of participants recruited on each medication at the common doses. Encapsulated placebo medication that was identical in appearance to active medications was also produced. All participants experienced a change in the supply of their antipsychotic medication at the outset of the study. Although individuals started the trial on their usual dose of medication, it was important to ensure that the number of tablets that they took daily remained constant over the blinded period and that the effective dose could be reduced across dose reduction steps. In order to allow participants to familiarise themselves with their new medication, a run-in period was built into the programme for all participants, regardless of allocation and prior to any reduction.

Manufacturing estimates assumed that all participants would achieve at least a 50% reduction. In reality, the number of reduction steps achieved was likely to be much more variable, although this assumption allowed for a reasonable degree of flexibility. Manufacturing estimates included provision of medications to all participants up to 9 months, when the blind was lifted.

Investigational Medicinal Products were manufactured by St Mary’s Pharmaceutical Unit (SMPU) under its Manufacturer’s Authorisations for IMPs licence and dispensed using Nomad^® Clear 2 trays (Omnicell Ltd, Manchester, UK) in accordance with participant-specific prescriptions. SMPU was to do this under section 37 of the Medicines for Human Use (Clinical Trials) Regulations (2004)41 as ‘post (Qualified Person) certification labelling for safety purposes’. Nomad Clear 2 trays are disposable trays with separate compartments for days of the week as well as times of day – morning, midday, evening and bedtime. IMP was dispensed monthly for 9 months. Although participants were to take 28 days of medication at each stage, enough IMP was provided for 33 days (to allow for a +5-day window around the planned 28-day time frame between medication review visits) in case of any delays or issues in getting the prescription.

Participant-specific prescriptions were issued to SMPU by the research team following consent and randomisation. SMPU then dispensed and dispatched these directly to site, where they were formally received and kept secure by a practice nurse (or designated individual). Accountability documentation was completed on receipt of the IMP and returned to the study team, thus evidencing the ownership of the IMP. Prescriptions were handed out only to the patient or their carer, legal representative or researcher by authorised site staff. Again, accountability documentation was completed and returned to the study team at this time. This process was repeated for the IMP of each new month following the PI’s decision to allow the participant to progress through the trial.

When any new prescriptions of study drug were collected from site, unused IMP from previous stages had to be returned by the participant or their carer. Sites were then responsible for the destruction of any unused study medication in accordance with local procedure and following authorisation from the trial manager. Completion of accountability documentation at various time points allowed the study team to evidence the location of IMP throughout the trial.

Toxicity was not expected and use of all pro re nata (PRN) medication was permitted and recorded in study diaries during the trial. Drug reduction was unlikely to cause interaction with other drugs; however, it was recommended that participants taking warfarin underwent more frequent international normalised ratio tests. All concomitant medication was permitted and details of any medications that had been taken were collected by the research team. IMP was stored at ambient temperatures at site; therefore, no temperature monitoring was undertaken.

Piloting

Once the trial was open for recruitment, arrangements were piloted in primary care for 6 months in order to test the assumptions and practicalities of trial processes and recruitment. At the end of this period, any adjustments would be made as necessary and the study was to then continue until full recruitment.

Principal investigator visits/contact

Participants in both trial arms (intervention and control) had five appointments with the PI in total. The first four took place in the 2 weeks preceding the release of each new batch of blinded medication and were approximately 28 calendar days apart. The purpose of the appointments was for the PI to make an assessment of whether or not there were any concerns about the participant’s progression to the next stage of the trial. When face-to-face appointments could not be held, the PI could consult over the telephone. It was the responsibility of participating PIs to provide participants and the study team with details of each of these appointments. The PI provided appointment cards to the participant or their carer and were responsible for reminding participants of the appointment nearer the time of the visit to ensure attendance. The site was also responsible for rearranging any appointments as necessary. The appointment card contained the contact details of the PI, an emergency number for participants or carers to use should they need to and a reminder of the amount of medication the participant had been taking when they started the study. It was important that the PI was the first point of contact for participants or carers if they had any concerns. The fifth PI visit took place after the 9-month assessment and was for the PI to unblind the participant and their carer and reveal the treatment allocation. It was also the point at which a discussion would take place regarding the participant’s care from there on.

Practice nurse visits

Participants (or their carer/representative) in both trial arms (intervention and control) collected their prescribed study medication from the practice nurse monthly until the blind was broken at 9 months. At each of these visits, the practice nurse took receipt of any unused medication from the previous prescription before distributing any new medication. The practice nurse would then complete accountability paperwork before destroying any unused medication on confirmation from the trial manager.

Assessments and follow-up

Eligibility data were collected at screening. Full data were to be collected at baseline and post intervention, approximately 9 months from randomisation (Table 1). Data on medication and psychopathology [Modified Overt Aggression Scale (MOAS), Aberrant Behaviour Checklist (ABC), Psychiatric Assessment Schedule for Adults with Developmental Disability (PAS-ADD) checklist] and costs [Client Service Receipt Inventory (CSRI)] were to be obtained at 6 months and 12 months. All data collection was face-to-face, either at site or during home visits.

Details of outcomes and follow-up time points can be seen in Table 2 and were the same for both experimental and control groups.

Screening measure

Information collected included age, gender, current medication and psychiatric history. In addition, adaptive behaviour was assessed using the Adaptive Behaviour Scale (ABS)42 as a means also to estimate intelligence quotient (IQ). 43 Current mental health status was assessed using the PAS-ADD interviews. 44 The data gathered were used to confirm inclusion and exclusion criteria. If required, clinical review was undertaken for those exceeding thresholds for the ABS (a score that converts to an estimated IQ of > 70 using the method described by Moss and Hogg43) and/or the Mini PAS-ADD checklist (a score for section M, potentially indicative of psychotic disorder, of > 2).

Primary outcome measure

The primary outcome measure was aggression and was evaluated using the MOAS. 45 The MOAS rates four categories of aggression (verbal aggression, destruction of property, self-mutilation and physical aggression to others) each on a scale of 0–4 but then weighted by an ascending index of seriousness. The measurement to be used here was a non-inferiority comparison; therefore, a score difference of ≤ 3 was to be taken as clinically non-significant.

Secondary outcome measures

Secondary outcome measures at baseline, 6-month, 9-month and 12-month assessments were as follows.

• The ABC46 comprising 58 behaviours, each relating to one of five subscales, to assess other challenging behaviour.

• The PAS-ADD checklist47 to monitor mental health. The PAS-ADD checklist is a 25-item questionnaire designed for use primarily with care staff and families. The scoring system includes threshold scores, which, if exceeded, indicate the presence of a potential psychiatric problem in the scale’s three diagnostic domains (affective or neurotic disorder, possible organic condition and psychotic disorder). The proportions of people reaching threshold scores for possible mental ill health were to be compared.

• The Antipsychotic Side-effect Checklist (ASC)48 was used to measure adverse effects of psychotropic medication. The ASC comprises a list of the more common or clinically important side effects of antipsychotic treatment.

• The Dyskinesia Identification System Condensed User Scale (DISCUS)49 was used to assess movement disorders. A psychometrically derived DISCUS threshold of 5 was to be used.

• The CSRI was modified for use in those with LDs and has been used previously in LD research. 21,50 The CSRI was used to collect data on a comprehensive range of services used and support received by each individual in the study. The collection of these data facilitated the calculation of the cost of medication, health, social care and unpaid carer inputs incurred by trial participants because of challenging behaviour or mental ill health.

The primary and secondary outcomes relating to challenging behaviour and mental health were to be analysed for non-inferiority with other secondary outcomes, such as medication usage and adverse effects, to be analysed for difference. A description of all scales, the range of their possible values and their interpretation is given in Appendix 4.

Randomisation and unblinding

The offline password-protected randomisation programme was designed by the trial statistician and based on the method of minimisation. Allocations were balanced with respect to medication type (risperidone/haloperidol) and dose: low (< 4 mg for risperidone and < 5 mg for haloperidol) and high (at least 4 mg for risperidone and at least 5 mg for haloperidol). A random component, set at 80%, was used alongside the minimisation procedure to increase the integrity of the minimisation process (i.e. there was an 80% chance that the allocation would minimise the imbalance with respect to the aforementioned balancing variables).

Following consent and baseline assessments, participants were randomised to either the intervention arm (gradual reduction) or the control arm (treatment as usual) in a 1 : 1 ratio by a member of the study team. Any unblinding was performed only after authorisation from the chief investigator or (if not available) an authorised clinical reviewer who was an appropriately qualified clinician and member of the trial management team. In the event of an emergency, the treating clinician would have access to details of the participant’s baseline dose (i.e. the dose at which they entered the study) and so could treat accordingly.

Sample size

We originally aimed to randomise 310 participants (155 per group) in total, which would have provided 90% power to fit a one-sided 95% confidence interval (CI) around the between-group difference in mean MOAS scores at 9 months post randomisation. A sample size of 310 assumed a non-inferiority margin of 3 and a standard deviation (SD) of 8 (i.e. an effect size of 0.375) and had been adjusted to allow for 20% attrition.

Main analysis

The original proposed primary analysis focused on a comparison of MOAS scores at 9-month follow-up between the two trial arms. An analysis of covariance model, with baseline MOAS score and variables balanced on/stratified by at randomisation (medication type, dosage and recruitment source) controlled for as covariates, would have been fitted. Using the estimates from this model, a one-sided 95% CI of the adjusted mean difference in MOAS scores at 9-month follow-up (intervention – control) would be calculated. Non-inferiority would have been concluded if the limit of the CI was < 3 in all study populations [complete case (CC); full intention to treat (ITT), with multiple imputation used to impute missing outcome data; and a per protocol (PP), which would include participants who had outcome data available, had not withdrawn from trial treatment and, if they were allocated to the intervention group, had experienced at least one reduction].

A complier average causal effect analysis would have been performed as a secondary analysis of the primary outcome, to obtain an ITT estimate in the treatment adherent. If non-inferiority was concluded, a superiority analysis of the difference in MOAS scores between trial arms was planned in the CC and ITT populations, using a two-sided 90% CI.

All secondary analyses (antipsychotic medication use, other challenging behaviour, mental health, adverse effects, movement disorders) would have been conducted using the CC population, with those secondary outcomes assessed for non-inferiority (challenging behaviour and mental health) and adverse effects also being analysed using the PP population.

Potential moderators of the effect of the intervention on MOAS score (e.g. age, gender, medication type or adherence to intervention) would have been explored in multivariable analyses using interaction terms. It was also originally proposed to model aggression levels using mixed models to explore changes over time.

Cost-effectiveness analysis

The original proposed main cost-effectiveness analysis focused on the comparison of the two trial arms through the calculation of incremental cost-effectiveness ratios, defined as the difference between trial arms in mean costs divided by the difference in mean outcome (MOAS score) over 9 months.

It was proposed to conduct the main cost-effectiveness analyses from health and social care agencies and a wider societal perspective to include health and social care agencies and unpaid carers. To inform the cost-effectiveness analyses from these two perspectives, it was proposed that comprehensive data on health, social care and other services used by individuals were included in the study, using a tailored version of the CSRI. To estimate component costs, service use data were due to be combined with the unit costs for each service using long-run marginal opportunity cost principles. For services in which national figures were not available or not suitable, we proposed to calculate best estimates of long-run marginal cost; values and time spent by friends or relatives providing support were due to be estimated using the unit costs of a local authority care worker. Three main categories of costs due to be analysed were (1) medication costs; (2) medication costs, aggregated health and social care costs, consisting of inpatient admissions, outpatient appointments, accident and emergency department contacts and community-based health and social care contacts; and (3) medication costs, aggregated health and social care costs and cost of time spent caregiving by relatives and friends.

Costs were proposed to cover the period from baseline to 6 months (the end of the full treatment withdrawal period) and 6–9 months (3 months following the full treatment withdrawal period). The MOAS score was to be used as the primary measures of effectiveness in a series of cost-effectiveness analyses. As cost data are likely to be skewed, and to explore if unobserved difference in service use at baseline between the allocation groups may result in differences in cost between treatment groups, regression analysis using bootstrapping was proposed, adjusting for baseline covariates (MOAS score, baseline costs and variables balanced on/stratified by at randomisation: medication type, dosage and recruitment source).

A series of cost-effectiveness analyses was to be conducted by combining outcomes with costs from health and social care agencies and unpaid carers in turn. In the event that the experimental reduction group had lower costs and better outcomes than its comparator, it would have been interpreted as the dominant treatment, and if the experimental reduction group had higher costs and worse outcomes than the comparator treatment, the experimental reduction group would have been dominated by the comparator. If the experimental reduction group was both more effective and more costly than its comparator, the nature of the trade-offs to be made would have been made using cost-effectiveness acceptability curves (CEACs). To generate the CEAC and non-parametric bootstrapping of the costs and effectiveness, data would have been used to generate the joint distribution of incremental mean costs and incremental effects. The CEAC shows the likelihood of one treatment arm being seen as cost-effective relative to another treatment arm given different (implicit monetary) values placed on incremental outcome improvements.

We originally planned to use one-way sensitivity analyses to examine robustness of the findings to (1) changes in the unit costs of informal support, (2) analyses based on all randomised participants whose 9-month follow-up MOAS score is known (CC population) and (3) analyses based on all randomised participants (ITT population).

Qualitative study

We undertook qualitative telephone interviews with a proportion of carers, PIs and participants who took part in the trial. One of the main purposes of these interviews was to gain insight into the non-efficacy-based barriers to drug reduction in clinical practice, as well as attributions of behavioural changes in relation to potential reduction of medication. The interviews were scheduled to take place during the unblinded phase of the trial between the 9- and 12-month time points and were to ascertain (1) views about participating in the study, (2) reasons for any partial or full reinstatement of medication after unblinding and (3) views about antipsychotic medication use to treat or control challenging behaviour for the participant, in particular, and the patient group in general. PI interviews also focused on PI views of the support package and views about how the patient and carer(s) managed during the trial period. Interviews were expected to take up to 30 minutes.

We aimed to interview up to 60 carers and the corresponding PI. It was hoped that both parties would agree to take part in these paired interviews, but we accepted that this was not guaranteed. The sample was to be selected, purposefully incorporating participants from both trial arms and from across the geographical recruitment areas.

We also hoped to interview a proportion of participants of the ANDREA-LD trial. Those taking part would be required to have the capacity to provide consent for a face-to-face interview. Interview topics for participants focused on (1) the reasons of participants for participating in the trial, (2) how they felt they managed during the trial period and (3) their views about taking medicines to help with their behaviour.

Carers and participants who agreed to take part in an interview were offered a £10 high street shopping voucher to thank them for their time and considered views. PIs who participated in interviews were offered £50. With the participants’ consent, all interviews were audio-recorded, transcribed and anonymised.

Qualitative analysis

It was proposed that data from the transcribed anonymised telephone interviews would be subject to thematic analysis as described by Braun and Clarke. 51 Thematic analysis allows researchers to take an initial inductive approach towards the data set. Following familiarisation with the data, researchers index data according to a priori and emerging themes. A priori themes are informed by the research literature on the topic of antipsychotic medication for people with LDs. Analysis is facilitated by use of the computer-assisted qualitative data analysis software package NVivo version 10 (QSR International, Warrington, UK). Data from each data set (participants, carers and PIs) would be analysed separately and then comparisons made across data sets.

Recruitment via community learning disability teams

Despite expansion of recruitment in primary care to areas in England, it was apparent that targets would not be achieved in the predicted time frames by relying on this route. We gained approval from the funders to expand recruitment to CLDTs, with LD psychiatrists acting as PIs and hospital-based pharmacies taking on the role of the practice nurses and dispensing trial medication.

Twenty LD psychiatrists from six trusts in Wales and England were then recruited to act as sites in the trial. An additional 13 hospital pharmacies were recruited in order to dispense trial medication.

Other alterations

Evidence from screening logs showed that the number of potential participants receiving haloperidol was much lower than expected. For this reason, the decision was taken not to manufacture blinded haloperidol medication but to continue to recruit only those taking risperidone. With these changes in place, the randomisation programme was also altered so that allocations were stratified by recruitment source (general practice/CLDT) and balanced with respect to medication dose only: low (< 4 mg for risperidone) and high (at least 4 mg for risperidone).

Exploratory pilot study methods

In November 2015, the decision was taken to close the trial to recruitment because of the difficulties described. At this point, 22 participants had been recruited into the trial. The study team submitted a close-down plan to the funders and it was agreed that all randomised participants would continue to receive the intervention and follow-up to 9 months. This meant that the trial would be complete by the end of June 2016 and would be reported as an exploratory pilot study (as defined by the Health Technology Assessment programme). As such, there were a few key alterations to the methods previously described, which are detailed in Table 3.

Exploratory study analysis

As the required sample size would not be achieved, we planned to focus on estimating feasibility outcomes. With a particular interest in recruitment and retention, we planned to estimate the following:

• the number and proportion of primary care practices/CLDTs that progressed through the various stages from initial approach to recruitment of participants

• the number and proportion of recruited participants who progressed through the various stages of the study.

We also compared trial arms regarding the following clinical outcomes:

• MOAS at 6 and 9 months post randomisation

• level of psychotropic medication use, assessed at the 6-month and 9-month post-randomisation assessments

• ABC at 6 and 9 months post randomisation

• PAS-ADD checklist at 6 and 9 months post randomisation

• ASC at 9 months post randomisation

• DISCUS at 9 months post randomisation

• use of PRN medication over the study period

• use of other interventions to manage challenging behaviour at 9 months post randomisation, including:

  • physical intervention/restraint

  • seclusion

  • PRN medication

• costs and service utilisation at 6 and 9 months post randomisation.

The analysis of recruitment and retention outcomes was descriptive, with frequencies and percentages reported both overall and split by recruitment route (primary care or CLDTs). Pre-randomisation variables, including those related to the participant, their recruitment route and their starting medication, were used to explore the association between participant characteristics and their progression through the study post randomisation (to stage 4 of the intervention).

Clinical outcomes were compared between arms using appropriate regression models (linear or logistic, depending on type of variable). Data transformations were made, when required, to fulfil regression assumptions. The analyses were also adjusted for the corresponding clinical score at baseline (if measured), as well as variables that were balanced on at randomisation [dose of antipsychotic medication (< 4 mg or ≥ 4 mg) and recruitment route (primary care or secondary care)].

Three analysis sets were considered:

1. The ITT population, which comprised all randomised participants. When participants had either withdrawn from the study intervention (but remained in the trial for follow-up assessments) or not provided follow-up assessments at 6 or 9 months post randomisation, it was assumed that they returned to their original starting dose.

2. The modified intention-to-treat population (MITT), which comprised all randomised participants whose follow-up data were known.

3. The PP population, which comprised all randomised participants who had progressed to stage 4 of the intervention (regardless of the trial arm to which the participant was randomised).

The analysis of the level of psychotropic medication at 6 and 9 months was conducted in the ITT, MITT and PP populations. The remaining clinical outcomes analyses were conducted in the MITT and PP populations only.

As it was the original primary outcome, further exploratory analyses were performed with the MOAS score:

• The MOAS score at 9 months post randomisation was fitted with a two-sided 90% CI in order to reflect the original primary analysis intended for this study.

• Individual trajectories for the MOAS scores at baseline, 6 months and 9 months post randomisation were plotted and described, with particular attention paid to individuals whose MOAS scores changed by at least 4 points (a change considered to be clinically meaningful).

Recruitment of sites

Research-active general practices in south Wales and south-west England were recruited into the study in the first instance. Initially, the number of practices interested was quite low; therefore, we expanded our approach to include practices that were not known to be research active. In total, approaches were made to 351 practices in four health boards in south Wales and 127 in eight Clinical Commissioning Groups in south-west England using a variety of methods, including e-mails, mailshots, articles in GP magazines/circulars and follow-up phone calls. The result was an active decline response from 204 practices across Wales and 17 practices in England and no response from the rest.

When recruitment later moved to CLDTs, we approached 30 LD psychiatrists also in south Wales and south-west England. While the decision to participate in the research could be made at a practice level by GPs, this decision required buy-in not only from the whole health board or trust at secondary care level but also from the LD services and teams. In one area, meetings were held with a member of the trial team and representatives from the whole LD directorate before a decision was made to participate. Concerns centred mainly around the buy-in from wider care teams and how any escalations in behaviour would be handled and communicated between relevant parties. Of the 30 LD psychiatrists approached, 20 across six health boards/trusts agreed to take part in the trial.

In primary care, a site was considered to be the general practice, and in secondary care a site referred to the LD directorate of each health board or trust. Owing to the changing nature of the NHS, establishing which was the appropriate body to obtain approval from was more complex than expected. In one area of England, provision of LD services was provided on behalf of the NHS by a community interest company. Research was not something commonly dealt with by the company; therefore, the permissions process was unclear and took time to establish. This had the knock-on effect of delaying the recruitment of investigators and, thus, of participants.

Gaining permissions to use community LD services in south Wales was also challenging. The Abertawe Bro Morgannwg University Health Board LD directorate acts as the provider of LD services across Bridgend, Cardiff, Merthyr, Neath Port Talbot, Rhondda Cynon Taf and Swansea. This meant that permission to recruit participants through the LD clinics in these areas was gained through the Abertawe Bro Morgannwg University Health Board but the permission to use hospital pharmacies that were local to those clinics was gained through each individual health board.

Recruitment of participants depended on ensuring that approvals were in place for an investigator and a local hospital pharmacy at the same time. This had an impact on recruitment in one health board, as approval had been granted to recruit through a LD clinic by one health board but the health board in charge of the local hospital pharmacy delayed granting permission for some months while the board reviewed the request. Securing NHS costs added to the delay in gaining permission to recruit in certain areas, as the allocation process had become more complex with this change in study recruitment.

Recruitment of investigators

Along with permission from research and development boards, it was necessary for investigators to have undertaken Good Clinical Practice (GCP) training before undertaking research activities. This proved to be an obstacle for many, particularly in secondary care settings in which clinician involvement in research was less common. GCP training is specific to clinical trials and typically takes up to 3 hours to complete – a time commitment that was not always easy for clinicians to accommodate. Despite having 16 investigators based in primary care sites and 20 in CLDTs, only 14 of these recruited any participants into the trial: three GPs and 11 community LD psychiatrists. Low recruitment rates on the part of PIs are explored in more detail through interviews that are reported in Chapter 6.

Of the 11 psychiatrists who recruited participants, three were specialist registrars in LDs. The specialist registrars were able to provide invaluable support for the trial in that they took on the role of investigator, obtained GCP training and were able to see a number of participants on behalf of other clinicians who might not have had the time to get involved in research. Specialist registrars worked on rotation, however, and if they moved to different health boards or trusts there were difficulties in maintaining care of participants as part of the trial. Fortunately, we were able to overcome this in the current trial and arrange for other clinicians to take on the role of investigator.

Recruitment of participants

Details of the eligibility criteria were clearly laid out in the trial protocol, and practices and LD teams were able to identify sufficient numbers of potentially eligible patients. However, the number of potentially eligible patients who were actually approached with details about the trial was fairly low, particularly among primary care clinicians.

The study team drafted an audit tool that provided search terms and Read codes for general practices to help them identify those who might be eligible to take part. Once patients had been identified, GPs appeared reluctant to directly approach these patients or their carers about the study and to invite them to take part. It is not clear what the reasons for this were, as we did not gain any feedback through interviews with primary care clinicians (this is discussed in more detail in Chapter 6). Anecdotal feedback suggests that it may have been in part because of the nature of the consent procedure as well as concern about taking on decisions for care that are normally the domain of secondary care LD clinicians. However, following the change in recruitment from primary care to CLDTs, primary care practices became participant ID (identification) centres, rather than full sites, in an attempt to mitigate this issue. When clinicians had identified patients who might potentially be eligible, it was then possible to refer them to the CLDT clinicians who would be able to discuss the study in more detail. This potential was not subsequently utilised by GPs and did not result in participants coming from leads in primary care.

Consent procedure

There is understandable anxiety over the capacity of individuals with intellectual impairment to participate in clinical trials. Within the drafting of the Mental Capacity Act 200552 specific provision was made relating to the care, treatment and decisions on behalf of people who lack capacity, including participation in clinical trials. Further, important information is contained in the Medicines for Human Use (Clinical Trials) Regulations (2004). 41 As a Clinical Trial of an Investigational Medicinal Product (CTIMP), the ANDREA-LD trial was required to adhere to the latter regulations (which supersede the Mental Capacity Act 2005), meaning that, in the case of those who lacked capacity, consent would need to be given by a personal or professional legal representative of the participant. Although clinicians (particularly CLDT psychiatrists) are well versed in the Mental Capacity Act 2005,52 a lack of experience in research means that many are not familiar with differences in the consent procedure as specified under the Medicines for Human Use (Clinical Trials) Regulations (2004). 41

It was expected that many potential participants in the ANDREA-LD trial would probably lack capacity to give informed consent, and so clear study-specific guidance explaining the Regulations was drafted. The guidance specified that the following factors needed to be taken into consideration.

• First, no patient would be treated as unable to make a decision unless all practical steps to help them to do so had been taken without success.

• Second, even if an individual lacked capacity, their opinion would still be taken into consideration.

• Third, should they lack capacity, the research team would, in accordance with due legal process, consult with relevant parties to clarify whether or not their participation was in the patient’s best interest.

Even with this guidance and reassurance, it became apparent that the consent process for the trial would be more challenging than expected.

Not only were clinicians apprehensive regarding the consent regulations, but wider care teams, carers and support staff were as well. The impact of this meant that it took much longer than hoped to explain the consent procedure to individuals and also to identify someone willing to act as a legal representative when the participant lacked capacity. On average, anecdotal evidence suggests it was taking between 2 and 3 weeks just to complete phone calls and arrange a meeting in order to complete the consent procedure. This was especially the case in residential settings, where carers, who would be well placed to provide consent, were required to refer this decision to managers and seniors, whose involvement often meant that a best interests meeting was called with wider care teams. With added layers of referral, it was often difficult to identify an individual who would be willing to provide any necessary consent.

Dosing

After extensive discussions, the decision was made by the trial team to include patients on any dose of risperidone, provided it was not being prescribed for psychosis and that the participant showed no evidence of psychosis. The rationale for this was that, if the withdrawal programme was to be shown to be effective and safe, it needed to work for individuals who would be on varying doses of medication. If the patient was deemed clinically eligible to enter the trial on all other criteria, their dose of risperidone should not be a restricting factor. By not limiting the entry criteria in this way, the pool of potentially eligible participants was also increased. We hypothesised, based on clinical experience, that doses of risperidone would be relatively low in the target population given that it was being prescribed for challenging behaviour as opposed to psychosis, for which larger doses are clinically indicated. The trial team, therefore, felt that there would probably be a limited number of dose combinations to cover as part of the trial. Careful consideration was given to the practicalities of how participants would receive their individually tailored medication regime.

Blinding medication

Blinding clinicians, carers and participants to treatment allocation presented a practical challenge. How could varying medication strengths be potentially tapered off without revealing which arm the participant had been randomised to? The decision was made to overencapsulate all trial medication using size 0 Swedish orange hypromellose capsules so that all of the strengths of risperidone and the placebo looked the same. The number of pills taken each day would also need to remain the same throughout the trial. To achieve this, the trial statistician created an algorithm that was used at baseline to create a unique dosing schedule for each participant (Table 4). Based on the participants’ prescription upon entering the trial, the algorithm calculated how many capsules each individual would need to take on a daily basis to ensure that the blind would remain intact. As a result of manufacturing constraints, only specific tablet strengths of risperidone were used in the trial (0.5 mg, 1 mg, 2 mg). The algorithm aimed to make the reductions as equal as possible using the minimum number of capsules as possible, based on up to four possible drug reduction stages within a 6-month period. (Note: it was not possible to reduce medication in four stages for participants on very low doses of risperidone, e.g. 0.5 mg daily.) If the participant had been allocated to the reduction arm, as the dose of active medication was reduced, a placebo capsule was introduced into the treatment regime. This allowed a constant number of capsules to be taken throughout the trial but with a variation in dose as necessary to accommodate the reducing regime.

Using the Nomad trays with separate compartments for days of the week as well as times of day meant that participant-specific doses of IMP could be safely dispensed and delivered. The trial team was able to specify which tablets (and thus which strength) should be taken at which time while maintaining the blind.

Because of the change in appearance of participants’ normal medication, a run-in period was implemented, which allowed individuals to get accustomed to using the Nomad trays and taking the slightly larger than normal capsules. For the vast majority, the change in the appearance of the medication was not a problem. Only one participant had difficulty taking the trial medication and had to be excluded. It was also important for carers to become accustomed to the new medication during this period, as they would be responsible for ensuring that participants took their medication as prescribed. Written and verbal information was provided to carers on how to handle study medication and the importance of using the Nomad trays correctly. It was important to ensure that carers fully understood how we were using the capsules and trays to blind the medication and for them to be clear on how to raise any concerns they had. In the case of participants residing in a staffed house, it was particularly important that everyone involved in providing that person with their medication knew about the trial.

A number of individuals required more tailored IMP deliveries, which included aligning IMP dispensing with that of other prescriptions the participant may be taking in order that medication administration record sheets could be completed more easily. These are sheets that serve as a record of the drugs administered to a patient at a residential setting. Although Nomad trays were used along with European Union Good Manufacturing Practice annex 13-compliant labelling, two residential settings delayed the progression of participants onto study medication by insisting that an extra label, signed off by the clinician, was included. Individual requests such as these became extremely time-consuming and difficult for the study team to manage.

Progression through the trial

Participant safety and careful monitoring of behaviour was of utmost importance in the trial design. The programme of reducing medication was devised to allow PIs to have regular contact with participants and to provide the opportunity to delay any potential reductions at any point if there was concern about the individual. This meant that trial medication could only be given out on a monthly basis once the PI had confirmed how the participant was to progress (Figure 2). To accommodate this, monthly prescriptions were dispensed and delivered to site within a 10 working day time frame. PIs therefore made their monthly contact with participants 2 weeks from the start of a new medication stage to allow adequate time for a new batch of trial medication to be dispensed and delivered. The PI’s decision regarding progression to the next study stage was translated into an IMP order by the study team and sent through to the pharmaceutical unit for dispensing. This pattern of working required investigators to be prompt in their communication with the trial team and to see participants within the specified time frame. When this was done, the system worked well; however, it also meant that the trial team needed to provide constant oversight to investigators on a real-time basis, which increased the burden on the study team’s workload and monitoring.

FIGURE 2.

Flow diagram for PI visits and IMP collection.

Dispensing medication

Another challenge was to ensure that instructions for taking medication while maintaining the blind were clear. To do this, the study team chose to deliver medication using the Nomad dosing system. Some participants and carers would have been used to receiving medication in this type of dosing tray, which is designed to make it more straightforward for individuals to know how much medication to take and when.

Once IMP had been manufactured at SMPU in Cardiff, it then had to be dispensed into the Nomad trays in accordance with the participant-specific prescriptions. The trial team made extensive investigations into who would be able to carry out the dispensing and then how to get the IMP to participants in different geographical areas throughout south Wales and south-west England. After discussions with various parties, the option chosen was to use SMPU not only to manufacture the IMP but to also supply to the patient-specific orders in a Nomad system under section 37 of the Medicines for Human Use (Clinical Trials) Regulations (2004)41 as the process of ‘post (Qualified Person) certification labelling for safety purposes’. SMPU was then able to dispatch IMP orders as necessary directly to specified sites or pharmacies via courier to be received by a designated member of staff.

When recruiting in primary care settings, each practice nurse’s role was to take receipt of participants’ medication and ensure that it was handed out according to the protocol. They also completed accountability records to evidence the whereabouts of the IMP. Enough IMP for 33 days was delivered at each stage, which meant that there was usually unused IMP that needed to be returned and destroyed. Any unused IMP from a previous stage could have been at a higher dose for participants in the reduction arm. It was important that the correct blinded medication was used each time.

Once recruitment in CLDTs began, it was apparent that the set-up used in primary care would not translate to CLDTs, as LD psychiatry clinics were not equipped to handle, store and dispose of medication as required by GCP. To resolve this, alternative solutions were examined, including the use of community pharmacies and delivering IMP via the post. The priority was to minimise the disruption experienced by the participant or their carer in obtaining medication on a frequent basis. This had, of course, to be balanced with what would be practical for the study team to set up and maintain and what would fit with regulatory requirements. The benefit of using community pharmacies would mean an IMP pick-up location local to each participant. However, the drawback was that it was not feasible for the study team to set up and maintain a potentially very large number of pharmacies in this way (we would not know which pharmacy would be most appropriate to use until the participant had been recruited). It would not be possible to ensure that agreements, GCP training and initiation were all in place prior to issuing study medication. The use of postal services to supply IMP was also not feasible because of regulatory issues and ensuring that any unused IMP was returned.

The approach taken was to use hospital-based pharmacies. Larger hospitals with ready GCP-trained clinical trials pharmacies were used to take receipt of IMP orders from SMPU and to issue medication to participants or their carers. Participants often had other health-care reasons to visit local hospitals; therefore, picking up trial medication from these locations was not overly burdensome for participants or their carers. For those who did find this a challenge, the study team was able to co-ordinate medication delivery in person by various members of the research team and supporting research network. However, this was possible for only a small number of participants. Researchers did not have the capacity to do this for all, as it involved special journeys from the office to the hospital pharmacy, on to the participant’s home, back to the pharmacy and then back to the office.

Pre-randomisation characteristics of study participants

Participants were well balanced with respect to variables collected pre randomisation (including at screening and during baseline visits). Fifteen participants were male (68%) and the mean age was 43 years (SD 13.4 years). There were no self-referrals. The majority of participants were consented into the study by a legal representative (n = 18, or 82% of those randomised). Ten of the randomised participants had a diagnosis of autistic spectrum disorder (45%) and none had ever been diagnosed with attention deficit hyperactivity disorder (ADHD). Ten had been diagnosed with epilepsy (45%), with all 10 currently on medication for epilepsy, four having had a seizure in the last year and seven having had a seizure in the past 5 years.

Regarding strategies used by carers to manage challenging behaviour, three participants were managed using physical intervention (14%), six were managed using seclusion (29%) and nine using PRN medication (43%). Most participants were in regular contact with their LD team (n = 21, or 95%), and when confidence handling challenging behaviour was self-rated (by carer) on a scale of 1 to 10 (1 being least confident and 10 being most confident), the median score was 10 [interquartile range (IQR) 8–10] (Table 7).

Participants were on a median risperidone dose of 1.5 mg prior to randomisation (IQR 1.0–2.0 mg), with two participants on a dose of 4 mg or higher (9%). The majority of participants were given risperidone twice a day (n = 17, or 77% of those randomised), and among those whose data were known/available 20 participants had been on risperidone for more than 1 year (Table 8).

The mean ABS total score was 162 (SD 61.7), with no participants having an ABS-derived IQ that was 70 or higher. Median scores were 0 on all domains of the Mini PAS-ADD interview, for both 4-week and 2-year recall periods, and mental health thresholds were rarely triggered. One participant triggered the threshold for having a depressive disorder in the last 2 years, one for an anxiety disorder in the last 2 years, and three for psychosis in the last 2 years. No thresholds were triggered for disorders in the past 4 weeks (Table 9).

Clinical scores were generally low at baseline. The median MOAS total score was 1 (IQR 0–2). Median scores on the ABC were highest for the irritability (median 4, IQR 1–8) and hyperactivity subscales (median 5, IQR 2–13). The median scores on the PAS-ADD checklist were 0, and no participants met thresholds for any conditions/disorders. At least one antipsychotic side effect was reported by 16 participants (73%), with a median total number of side effects of 1 (IQR 0–3). No participants met the threshold for a possible movement disorder prior to randomisation (Table 10).

Randomisation characteristics

Twenty-two participants were randomised in total, with 11 allocated to each arm. As previously described, the majority of participants were on a total daily dose of risperidone of < 4 mg, and the majority were recruited from CLDTs. The arms were well balanced with respect to these key variables (Table 11).

Aggression (Modified Overt Aggression Scale total score)

At 6 months post randomisation, MOAS total scores remained low but were higher in those randomised to the intervention arm, with differences more discernible for the MITT population {control mean 3.0 [standard error (SE) 1.86], intervention mean 4.5 (SE 4.5)} than for the PP population [control mean 3.5 (SE 2.13), intervention mean 3.6 (SE 1.85)]. The adjusted mean differences for both populations indicated higher MOAS total scores for those randomised to the intervention arm, although 95% CIs were wide and included zero.

Modified Overt Aggression Scale total scores were higher at 9 months post randomisation than they were at 6 months, remaining higher in those randomised to the intervention arm in both MITT [control, mean 3.7 (SE 3.55); intervention, mean 7.7 (SE 3.51)] and PP [control, mean 4.3 (SE 4.14); intervention, mean 9.3 (SE 4.94)] populations. The adjusted mean differences both indicated higher MOAS total scores for those randomised to the intervention arm, although as for data at 6 months, 95% CIs were wide and included zero (Table 14).

Reflecting on our originally planned primary analysis (i.e. between-group comparison of the MOAS total score at 9 months post randomisation, with a two-sided 90% CI fitted and the upper limit of the CI inspected for non-inferiority), Table 15 and Figure 4 demonstrate that the upper limit of the CI in both the MITT and the PP populations (and for unadjusted and adjusted analyses) includes the stated non-inferiority margin. Therefore, we were unable to conclude non-inferiority on the basis of this study.

FIGURE 4.

Forest plot illustrating between-group mean differences on the MOAS at 9 months post randomisation, with two-sided 90% CIs. Estimates/CIs are on the ln [x + 1] scale. Original non-inferiority margin of 3 has been translated onto this scale (1.39).

Figure 5 plots MOAS total scores for each individual at each time point, separating participants by their trial arm and whether or not they progressed onto stage 4. A change in the MOAS total score of 4 was deemed clinically important. 20 For the majority of participants, change in MOAS total scores over time was slight. However, five participants experienced a change from baseline in MOAS total score of at least 4. Two of these participants had been allocated to the intervention arm and had progressed to stage 4, one had been allocated to the control arm and had progressed to stage 4 and two had been allocated to the intervention arm and had not progressed to stage 4 (with both of these participants experiencing an increase in their MOAS total score at 6 months and then a decrease between 6 and 9 months).

FIGURE 5.

Individual MOAS total scores over time by trial arm and whether or not the participant progressed onto stage 4 of the intervention. (a) Control, did not progress to stage 4; (b) intervention, did not progress to stage 4; (c) control, did progress to stage 4; and (d) intervention, did progress to stage 4.

Psychotropic medication use

The average total daily dose of risperidone was lower in those randomised to the intervention arm than in those randomised to the control group at both 6 and 9 months post randomisation and in both ITT and MITT populations. In the MITT population (i.e. in those for whom this outcome was available), the mean was 1.6 mg lower in the intervention arm (SE 0.24 mg), whereas in the ITT population (i.e. assuming risperidone use had returned to baseline levels in those for whom this outcome was not available) the mean was 0.6 mg lower (SE 0.78 mg). As one participant withdrew from trial treatment soon after progressing to stage 4 (i.e. there were no additional withdrawals before the 6- and 9-month time points), the differences in level of psychotropic medication use at 6 and 9 months were identical (Table 16).

Other challenging behaviour (Aberrant Behaviour Checklist subscales)

At 6 months post randomisation, mean ABC subscale scores were higher in those randomised to the intervention arm, with differences generally more discernible for the PP population than for the MITT population. While descriptively the means were consistently higher for the intervention arm, the adjusted mean difference was negative (indicating a lower adjusted mean in the intervention arm) for the irritability subscale (PP population) and inappropriate speech subscale (MITT and PP populations). However, other than for the stereotypy subscale (both MITT and PP populations), all 95% CIs included zero.

Although mean scores on the irritability and stereotypy subscales remained higher in the intervention arm at 9 months post randomisation, mean lethargy, hyperactivity and inappropriate speech scores were lower in the intervention arm. The 95% CIs of the adjusted mean differences for all ABC subscales included zero at 9 months (Table 17).

Mental health (Psychiatric Assessment Schedule for Adults with Developmental Disability checklist)

Mean scores on the subscales of the PAS-ADD checklist were higher in the intervention group at both 6 and 9 months post randomisation and for both MITT and PP populations. Owing to a high frequency of zero scores from participants, linear regression analyses were not possible for all outcomes. Similar to the analysis of ABC subscales, although descriptively the mean scores were consistently higher for the intervention arm, for two analyses (affective/neurotic disorder at 9 months for the PP population and psychotic disorder at 9 months in the MITT population) the adjusted mean difference (when an adjustment was made for the recruitment route and corresponding score at baseline) indicated that these were lower in the intervention arm. However, all 95% CIs of the adjusted mean differences included zero (Table 18).

At 6 months post randomisation, the threshold for a possible organic disorder was triggered by one participant in the intervention arm. At 9 months post randomisation, the threshold for an affective or neurotic disorder was triggered by one participant in the intervention arm and the threshold for a psychotic disorder was triggered by two participants (one per arm) (Table 19).

Adverse effects of psychotropic medication (Antipsychotic Side-effect Checklist)

In the MITT population, at least one side effect of psychotropic medication was reported by seven control participants (100%) and eight intervention participants (80%) when they were asked at 9 months post randomisation. In the PP population (i.e. in those who had progressed to stage 4 and for whom outcome data were available), all participants reported at least one side effect (Table 20).

The mean number of psychotropic medication side effects reported at 9 months post randomisation was higher in the control arm [MITT, mean 2.6 (SE 0.72)] than the intervention arm [MITT, mean 1.4 (SE 0.31)], with mean numbers similar for the PP population. However, the 95% CI of the adjusted mean difference included zero for both analyses (Table 21).

Movement disorders (Dyskinesia Identification System Condensed User Scale)

At 9 months post randomisation, the mean total DISCUS score was higher in the intervention arm [MITT, mean 2.1 (SE 1.67)] than in the control arm [MITT, mean 0.0 (SE 0.00)], while the mean for intervention participants who progressed to stage 4 was higher again [PP, mean 3.0 (SE 2.35)] (Table 22). One participant in the intervention arm met the threshold for a possible movement disorder (Table 23).

Managing challenging behaviour

At 9 months post randomisation, carers reported using physical intervention to manage challenging behaviour for two participants (both allocated to the intervention arm), although only one of these had progressed through to stage 4. They reported using seclusion for three participants (both allocated to the control arm) and PRN medication for nine participants (four allocated to the control arm and five to the intervention arm), although for the intervention arm only six of these participants had progressed to stage 4 (three per arm) (Table 24).

The mean score on the scale measuring confidence handling challenging behaviour was higher for the intervention arm [MITT, mean 9.3 (SE 0.26)] than for the control arm [MITT, mean 8.7 (SE 0.36)], with scores similar for the PP population. However, the 95% CIs of the adjusted mean differences included zero (Table 25).

Use of pro re nata medication

Use of PRN medication post randomisation was captured using carer-reported diaries. However, as indicated by Table 26, the majority of participants did not return these diaries. For those who did, it would appear that PRN use was generally higher in those allocated to the intervention arm and appeared to increase over time. This is also illustrated in Figure 6.

FIGURE 6.

Pro re nata use following randomisation.

Safety reporting

During the course of the trial, there were four adverse events reported and one serious adverse event (SAE). The SAE was categorised as ‘an event which required intervention to prevent outcomes such as hospitalisation’. There was reported deterioration in the participant’s mental health with increased agitation, tearfulness and depressive thoughts as well as reports of hearing a voice. Deterioration was such that crisis intervention was required to avoid hospital admission. It was suspected that these symptoms were not a side effect of the trial medication but a recurrence of symptoms masked by the risperidone and, hence, a serious adverse reaction to the withdrawal (i.e. the intervention being trialled during the study) as opposed to the medication itself. This participant withdrew from the trial intervention and the blind was broken. On entry into the trial, the participant had been taking 2.5 mg of risperidone daily and progressed through three stages of reduction so that, when the blind was broken at the time of the SAE, they were taking 0.5 mg daily.

Interviews with carers

In total, 16 carers (five parents and 11 professional carers) were invited to be interviewed by ANDREA-LD research staff. Interviews were conducted at the carer’s home, in a private room within the residential care home or by telephone.

Interviews with patient participants

With the support of residential care staff in three cases and a parent carer in the remaining case, a research assistant approached ANDREA-LD patient participants who had been identified as having mental capacity to consent for themselves during the screening phase of the study. Consent to participate in the qualitative study was undertaken. Interviews were conducted in a private room (an office or lounge) within the residential setting in three cases and in an office within a day centre in the remaining case. All interviews were conducted by a research assistant. A residential carer was present during three of the interviews; a member of day service staff was present during the fourth interview. Carers were asked to support the participants, but to refrain from answering on behalf of the participants.

Telephone interviews with clinicians

Eleven secondary care clinicians were contacted to take part in a telephone interview.

Interviews with carers

All 16 (five parent and 11 professional) carers who had been invited to participate in the study agreed to take part. The five parents interviewed had given advice as legal representative that their son or daughter could participate in the ANDREA-LD study. Interviews lasted between 12 and 31 minutes with an average time of 19 minutes.

Interviews with patient participants

Four patient participants consented to take part in an interview, all of whom had participated in the ANDREA-LD study, although two had withdrawn at some point during the study. Interviews lasted between 6 minutes and 18 minutes with an average of 11 minutes, although the study team had expected that they might be longer than this.

Reasons for participating in the trial

Clinicians discussed both personal and professional reasons for their interest in the study. Some clinicians talked about an ongoing interest in the study question of whether or not antipsychotics could be safely withdrawn for some patients. Most talked about their involvement in the study with colleagues and the wider team. Some clinicians had not participated in research studies or CTIMPs before and said that experience was part of the attraction of involvement. There were thought to be benefits to the wider team in terms of learning about research procedures and awareness raising about prescribing of antipsychotics:

Research is something that has always been part of my career and I think really interesting, in fact I found that in the last maybe, probably, 8–10 years I’ve been a bit out of touch with proper research and this gave me a chance to just be back, even if it was from a, within the background not as a main, not as a main, one of the main researchers, but that gave me a chance to get back in touch with research which I have always been very interested.

Clinician 4

There was also the attraction of being able to contribute to an important clinical question:

Obviously it’s a contentious area and we need a definitive answer one way or the other in terms of whether um, people with challenging behaviour do benefit from longer term use of antipsychotic medication, um I’m very aware of people becoming put on antipsychotic medication because of bad episodes, remain on it for many years later without it being adequately reviewed. I wanted to support high-quality research, firm believer that we need applied research, and jobbing psychiatrists being part of research, not just um, people with, with academic roles.

Clinician 6

Concerns about participating in the trial

Most concerns related to issues of recruitment and consenting arrangements. In relation to recruitment, the clinicians reported that many carers had been reluctant to be involved in the study or that carers had concerns about the rapidity of the withdrawal:

I guess really the uptake from some patients or their carers, yeah uh I think people generally get anxious when they think they’re going to take some medication away or may take some medications away that sort of . . . yeah it’s challenging.

So was there a little bit of figuring out how to manage that?

Yeah how to manage it and reassure people but I think that has affected the uptake.

Clinician 3

There were also concerns about the exclusion and inclusion criteria of the study. Although some clinicians seemed to consider that any patient of theirs on either risperidone or haloperidol was a potential candidate for the study, other clinicians appeared to apply the criteria more selectively and approached only participants and carers who they thought were ready to appropriately withdraw:

I didn’t have any real concerns for the patients we selected either because I think they were quite carefully selected and chosen the people who were on a medication that we thought perhaps they could do without.

Clinician 7

A couple of clinicians (6 and 7) also raised the issue that they thought people with autism were a distinct population who might not be suitable for the trial:

I think the population we looked at was correct, I think the only difficulty that I saw was, we do have patients with autism and sometimes they’re on small doses of antipsychotics and trying to, although it’s not for a psychosis, it’s for other reasons, so I think they may have been included. And when they’re taken off, even a small dose of antipsychotics they had difficulties I think, so that maybe something, that they’re a slightly separate population to people without autism.

Clinician 7

In relation to capacity, there were some concerns about whether or not participants with some capacity really had the opportunity to make an informed choice about their participation and contribute to the study. Furthermore, clinicians reported that obtaining consent from carers (personal legal representatives) had been challenging:

We did spend quite a lot of time trying to find out who was the proper person to discuss informed consent and talk about informed consent, you know in the paperwork it was very clear that the carers could give full consent, but the reality was that the carers were not happy, were quite uneasy with that.

Clinician 3

Experience of the study

Some clinicians who had no prior experience of being involved in CTIMPs reported that the study set-up stage had been time-intensive and, in their opinion, unnecessarily complicated:

I think what initially before actually doing something there was a whole I think there was just so much paper work and so many versions so actually that was a bit confusing um but I think once we’d done the first one and actually walked the process so to speak um it just became clearer. But the paperwork actually didn’t help, there was quite a lot.

Clinician 1

Sometimes problems about documentation seemed to spill into the main study period:

I’ll tell you what was a bit of a faff was having to fax over the sheets to the study centre once they were filled in because the fax never seemed to get there.

Clinician 2

However, generally, once the study was running, study processes had gone relatively smoothly and clinicians were generally very grateful of the support they received from the research team:

Yes once recruited the person I felt really supported and in fact I felt that, you know, in a way it was very straightforward just to do the follow-ups yes that was absolutely fine, yeah there were no problems at all there.

Clinician 4

Only one clinician felt that contact with and support from the study team had been rather lacking:

Maybe looking back, if there had been more frequent, for example telecontact, around the study itself. For other studies we have kind of given um, better kind of um, feedback on the study as well.

Clinician 5

Changes in behaviour and reasons for withdrawal from the study

Clinicians were not as forthcoming with comments about behaviour changes as the carers had been during interviews. One clinician stated that the clinicians relied on carers to report changes in the behaviour of study participants to them. Mainly, they thought that behaviours were stable and, when there had been changes, these were generally explained by other influences within the participant’s life such as an acute infection, admission to hospital or changes with other medication. As with the carer interviews, some positive changes in behaviour were reported as well, including regaining personality.

Some clinicians also expressed frustration that some of their patients had been withdrawn from the ANDREA-LD study because of concerns about behaviour changes from carers only to find that the participant was on the full dose of medication:

I did rely a lot on carers or family members to tell me about the behaviour. I always told the carers and the person ‘please do let me know if you see any significant changes, anything that will concern’. I made a point every single time, um but sometimes with the carers whether they could read more into something or not, because there was one person who unfortunately pulled out, they, they pulled out because they saw a change of behaviour and the person was still on there, I think it was still on their first month, therefore he wasn’t yet on the active medication study, he was still just on the same dose but with the placebo tablet.

Clinician 4

Clinicians also reported instances of patients who had withdrawn from the study, but perhaps for good reason. Even so, there was still some expressed regret that the patient had to do it before they finished the trial:

The second one is the one we had to break the blind, and that’s where um, in retrospect the family and the carers absolutely accurately um, guessed when he came off his final dose of medication. So they knew. They could see, and again that’s a young autistic man, going back to this autism business, they could see when that last dose went, he had that background agitation returned and of course he was aggressive or whatever, um unfortunately it was, the blind was broken.

Clinician 6 (participant in intervention arm and interview conducted at stage 2)

On other occasions clinicians reported that unblinding and withdrawal from the study was completely necessary, and there was no question that they wanted the participant withdrawn from the trial:

[Participant 17] we had to unblind, so there were concerns about significant change in behaviour that seemed to correlate with his involvement in the trial and potential reduction of his medication.

Clinician 7 (participant in intervention arm and was withdrawn at stage 3)

Management of participants after the completion of the study

How participants were managed, or were expected to be managed, after their ANDREA-LD involvement differed depending on the participant’s and carer’s experience of behaviours during the study period. For some, this meant continued and complete withdrawal from the study medication, but for others it meant a return to a small dose of the medication, primarily to abate concerns from family members or carers about observed changes in the participant’s behaviour:

One of your participants has finished in the study, any changes in um, prescribing medication since they finished?

No, I mean if they had been withdrawn, they have withdrawn, we have not put them back, but those who have been on the, who have not been on the placebo arm and they want back on the medication we went back on the medication for a short period to try and start again the process of withdrawing the medication.

Clinician 5

When the blind was broken, um for the one participant, um if you can, reflecting on how care kind of continued from there.

Yeah, he went back on the meds.

Yeah, and that was, that was just . . .

Small dose. That was at the request of his parents, well he’s in supported accommodation, the staff would’ve hung on in there, but the parents were saying, you know, ‘we noticed a difference, he’s less settled, he’s hit us, which he hasn’t done for months, um he’s not how he was’ and the carers said ‘look, you know we’ve noticed it as well’ so it was a bit of a no brainer in that he hadn’t had any side effects from the medication, we just put him on, back on the lowest possible dose and sure enough it took the edge off him again.

Clinician 7 (participant in intervention arm and was withdrawn stage 3)

Views about the use of antipsychotic medication in the wider learning disability population

There was a general view that antipsychotic medication has its uses in managing challenging behaviour within this population but is probably overused. There was also an awareness that some patients were prescribed these medications initially for a short period, or had their dose increased initially for a short period, but a lack of review meant that the patient remained on the medication for unnecessarily long periods or at unnecessary strengths:

I truly believe that when antipsychotics are prescribed for challenging behaviour, if they were prescribed on low dosages, sometime it can make a really, really important difference because that is what you see in real practice. But I think that we need to be able to evidence that really, really well and look at what are the side effects for the person.

Clinician 4

I wouldn’t prescribe medication just purely for behaviour problems, I would always definitely make sure that is part of the behaviour management protocol, so there are steps in place.

Clinician 5

There was also a feeling that the clinicians who were participating in the ANDREA-LD study were perhaps more consistent with nationally agreed standards in relation to antipsychotic prescribing than clinicians working on the boundaries of evidence-based care:

We’ve done audit on it, you know, it is an area that’s been looked at time and time again, which goes back to the fundamental problem with ascertainment, that we’ve primarily gone to the good services who have already been doing this and perhaps weren’t able to get so rapidly that the areas where practice might not have been so um, in keeping with national standards.

Clinician 6

There was also a sense of disappointment that, as a result of recruitment problems, the ANDREA-LD study would not be able to deliver an answer to the question of whether or not these medications can be safely withdrawn in people with LDs:

I think this is a really important subject, I think there has been some research in the past but obviously there has not really been a trial like this where you’re looking at um, a placebo control double-blind trial, because um, prospective, I think it’s a really good, and I think unfortunately we didn’t get that many people engaged in it which has caused a bit of an issue.

Clinician 7

Recommendation

We would like to see the Royal College of Psychiatrists, potentially working with the Royal College of General Practitioners, establish a working group with relevant stakeholders to provide guidance on antipsychotic drug withdrawal. This guidance should provide an auditable standard for drug withdrawal including recommendations for levels of behavioural and other service support and accessible information for people with a LD and their carers. Development of guidance for carers on this issue would also be beneficial. As Sheehan and Hassiotis note in their recent systematic review,40 a systems approach to this complex issue is required.

Recommendation

Future studies are needed to explore interventions that could reduce unnecessary antipsychotic drug use in people with a LD. They should involve interventions that, in addition to the process of reduction, address the concerns of professionals and carers by having extra support or alternative (i.e. non-pharmacological) interventions available for perceived or actual deterioration. It may be that reduction should be non-blinded and these alternative support mechanisms should instead be compared with treatment as usual.

Recommendation

We would like to see measures put in place to improve recruitment to studies in people with a LD and for consideration to be given to how adults with a LD might also be included in general population trials. These measures may include a commitment from the Royal College of Psychiatrists for GCP training (including covering recruitment with reduced or no capacity) to be mandatory for its members and for Clinical Commissioning Groups and other health groups, and for performance-related measures linked to recruitment to be considered for LD studies.

1. Concerns of behaviour escalation or new behavioural/mental health concerns outside normal range of behaviour

It is quite likely that carers or families will raise concerns relating to the individual for whom they care in relation to alteration of behaviour. In this first stage of assessment you can ask a few simple questions to signpost the next step in the flow chart:

Does this seem to be a new condition rather than worsening of usual behaviour?

If it is not worsening of usual behaviour, it is worth deciding if there is an underlying physical or psychiatric condition. Completely new patterns of behaviour are unusual so detailed questioning will ensure that it is definitely a new behaviour. If it does seem a new behaviour, then it is best to follow the advice on physical illness (likely to be common) or psychotic symptoms (likely to be rarer) in the flow chart. You may be directed to which step by the nature of the behaviour and other symptomatology.

If it is the usual pattern of behaviour then the next step is to assess its severity.

If this is worsening of usual behaviour – has the individual displayed behaviour as concerning as this before starting in the study?

If the answer is ‘yes’, then it is likely that this is simply a fluctuation in normal patterns as people with challenging behaviour tend to have ups and downs in the normal course of their condition.

In such situations it is unlikely to be a direct result of the study. Such fluctuations are best approached by watchful waiting.

If the behaviour is of a greater degree or the current degree causes great concern, then the most likely short-term reason for deterioration is environmental stressors-so please follow the advice in the environmental stressor box. If there is no such stressor then checking for a physical condition or new psychiatric symptoms would be worthwhile.

2. Environmental stressors

If carers do report a change in the level of a person’s challenging behaviour, it’s important to try and identify what the reasons for this might be. More often than not, these will be linked with some kind of change in the person’s environment rather than any change in the person themselves. Some good questions to put to carers include:

• Has there been any change in the person’s routine (e.g. have they been provided with their normal pattern of day activity, been able to go out as often as they normally do, been able to see people that they normally see)?

• Has there been any change in the people working with the person? (For example have there been a lot of relief staff on duty who don’t know the person well and not fully briefed regarding key aspects of their care?; has a particular carer of whom the person is particularly found been away from for some reason?; have the person’s parents been away on holiday and unable to visit?; has the service manager been off on leave?)

• Are all carers currently implementing agreed support plans (particularly behavioural support plans)? How do carers check that this is the case?

• Have there been any reductions in levels of support to the person that impact on the ability to implement these plans?

• Are the reported increases well-known ones that happen at a particular time of year? (Christmas and birthdays are often difficult times for people.)

• Has the person been exposed to any recent trigger (e.g. having to go to particular places that they don’t like or asked to do things that they’re not keen on; have they been left alone for lengthy periods)?

This is not an exhaustive list – but it gives you an idea of things to probe for. Carers will often not be aware that these sorts of issues may in themselves explain why the person may have been a little more difficult lately. The solution here is clearly to prompt the carers to address the environmental issues identified rather than to treat the person. Allowing some time for this to occur and the behaviour to settle is the best strategy here therefore.

Escalation of challenging behaviour and non-psychotic psychiatric symptoms

Escalation of challenging behaviour is especially linked to anxiety symptoms and may particularly be associated with environmental stressors. Addressing these stressors, as outlined above, may resolve the problem. If this does not resolve the issue please refer to the What to do if Challenging Behaviour Escalates section below.

3. Physical illnesses

Individuals with a LD are especially prone to developing physical illness but, given communication difficulties, behavioural difficulties may be the presenting feature. Sources of pain [such as ear infections, headaches, abdominal pain (because of urinary tract infection, constipation)] may be especially difficult to pick up and should be actively sought out.

4. Suggestion of psychotic symptoms

The participants in this study will have been screened for the presence of serious mental illness so, in theory, few if any should experience a relapse of psychosis or evolution of a new illness. However, the presence of mental illness in people with LD can be masked by a variety of factors including intellectual function and communication barriers. Therefore, some people whose medication was thought to have been originally prescribed for ‘challenging behaviours’ may have in fact had an undiagnosed psychotic illness, such as schizophrenia or a bipolar disorder.

If a patient in the study starts to develop an unusual pattern of speech or behaviour, it is important to remain vigilant for potential psychosis. In people with mild LD, the presentation would usually be similar to that seen in the general population, that is, abnormal beliefs out of keeping with their social situation which are not amenable to reason or abnormal perceptions which are not explained by obvious environmental stimuli. However, when the person lacks a level of intellectual function or emotional development commensurate with a chronological age, their ability to express their symptoms coherently may be severely diminished. Here, a carer’s account of unusual behaviours or communication which is out of keeping with their pre-morbid presentation may be critical in highlighting the evolution of serious mental illness.

In these circumstances it is appropriate to contact your local CLDT to ask for an urgent psychiatric assessment. They may be able to reassure you, or could, if necessary, ask for the medication code to be broken to identify the current prescription and withdrawal of the patient from the study.

5. Tardive dyskinesia

Tardive dyskinesia is characterised by chewing or sucking movements, grimacing and choreoathetoid movements particularly affecting the face, but also potentially the limbs and, most importantly, the muscles responsible for swallowing and respiration. While this syndrome can occasionally be seen in drug-naïve patients, it is more common in those who have taken antipsychotic medication for many years. However, neither dose nor duration of treatment is the sole determinant, and it is also more common in women, the elderly and patients with diffuse brain pathology. Almost half of the cases arise when drugs are being reduced or discontinued.

The cause is uncertain but could be super sensitivity to dopamine, resulting from prolonged dopaminergic blockade, and arises with many antipsychotic agents. Many treatments have been tried, but none is universally effective, hence the need to avoid prolonged unnecessary prescription of antipsychotic medication, especially at higher doses.

Should any patient develop symptoms suggestive of tardive dyskinesia during the study, it is recommended that you contact your local learning disability team for further psychiatric evaluation, including, if necessary a breaking of the code to identify the current prescription and withdrawal of the patient from the study. This advice does not preclude you from seeking other expert opinion, such as a consultant neurologist if you feel this to be a more appropriate course of action.

Definitions and reporting procedures

It is the responsibility of the Investigator to report all adverse events to South East Wales Trials Unit within 24 hours of becoming aware of it. Any queries concerning adverse event reporting should be directed to the Trial Manager in the first instance.

How to report

Complete an ANDREA-LD SAE form as follows and fax to South East Wales Trials Unit on XXXX.

1. The first report should be marked as ‘Initial’. Any other types of report (‘follow-up’ or ‘final’ will be prompted by the study team.

2. Complete the ‘Report Date’ and ‘Details of Subject affected by Event’.

3. Complete ‘Details of Event’. Note, an adverse event is considered serious if it:

   • results in death

   • is life-threatening (Note: the term ‘life-threatening’ in the definition of serious refers to an event in which the participant was at risk of death at the time of the event; it does not refer to an event which hypothetically might have caused death if it were more severe.)

   • required hospitalisation or prolongation of existing hospitalisation (Note: Hospitalisation is defined as an inpatient admission, regardless of the length of stay, even if the hospitalisation is a precautionary measure, for continued observation. Pre-planned hospitalisation, e.g. for pre-existing conditions which have not worsened or elective procedures does not constitute an adverse event.)

   • results in persistent or significant disability or incapacity

   • consists of a congenital anomaly or birth defect

   • any other medically important condition (Note: other events that may not result in death are not life-threatening, or do not require hospitalisation may be considered as a SAE when, based upon appropriate medical judgement, the event may jeopardise the participant and may require medical or surgical intervention to prevent one of the outcomes listed above).

4. Complete ‘Details of Investigational Medicinal Product(s)’ section with as much detail as possible. Note, for the question ‘Is the SAE related to the IMP?’, consider the following:

   • Most adverse events and drug reactions that occur in this trial, whether they are serious or not, may be due to drug reduction. They will not be toxicity related effects. The assignment of the causality should be made using the definitions in the table below.

1. Complete ‘Details of other treatment’ and ‘Further information relevant to assessment’ sections with as much detail as possible.

2. Sign the form and send a copy as described at the end of the form. Place the original in the Trial Site File and add a copy to the patient’s medical notes.

Emergency

If, having worked through the management flow chart above, you still have significant concerns or feel that risks remain despite amber intervention or there was an inability to access referral, you may feel it appropriate to request emergency unblinding of treatment allocation.

In such a situation, you must report an adverse event as described above by completing a Serious Adverse Event (SAE) Form (this can be found in your Site File) and submitting it to the Trial Manager. The Chief Investigator or Clinical Reviewer will then contact you to discuss the issues surrounding the situation and, if necessary, confirm that the participants treatment allocation needs to be unblinded.

Routine

Breaking the code (blind) at the 9-month visit.

When each patient reaches their 9-month visit, it will be time for the blind to be broken, thus revealing whether they were in the reduction arm or were receiving treatment as normal.

The ANDREA-LD team will reveal the allocation to you along with details of the dose of medication the patient is currently taking. It will be your responsibility to have a discussion with the patient and their carer (or representative) in order to relay this information.

For those still receiving medication, any ongoing treatment and dosage could be based on the dose the patient finished on depending on the clinical judgement of their clinician and the patient’s needs at the time.

Once the blind has been broken and the patient’s treatment allocation and drug dosage is known, we would like you to share this information with any secondary services involved in their care. Knowledge of how the patient progressed in the trial could be of importance in future decision-making regarding treatment.

ADDRESS

South East Wales Trial Unit

7th Floor Neuadd Meirionnydd

Cardiff University

Heath Park

Cardiff

CF14 4YS

Fax

FAO ANDREA-LD Trial

XXXX

Telephone

Trial Manager – XXXX

Data Manager – XXXX

Trial Administrator – XXXX

E-mail – XXXX

Clinical support

Professor Michael Kerr – XXXX

E-mail – XXXX

Dr Glyn Jones – XXXX

E-mail – XXXX

Secretary – Sandra Jones, XXXX