Methods

Summary of trial design

The trial is a NIHR-funded Phase III multicentre, randomised, double-blind, placebo-controlled trial. Patients with TRBD were approached to participate. The trial included two stages: pre-randomisation stage and randomisation stage. The purpose of the pre-randomisation stage was to prepare participants for recruitment to the randomisation stage by adjusting antipsychotics (where required), commencing mood-stabilising medication (where required) and ensuring participant engagement with the TrueColours database and the central research assistant (RA) team. As part of the randomisation stage, participants were randomised and commenced taking the trial medication. Participants were randomised on a 1 : 1 basis to receive either pramipexole or matched placebo alongside their existing mood-stabilising medication.

The PAX-BD trial was designed to allow flexibility regarding face-to-face versus remote completion of trial tasks. The only exception to this was the final part of the consent discussion and the physical signing of the consent form at both pre-randomisation and randomisation stages which had to be done in person.

A central pharmacy was utilised as part of PAX-BD. The sponsor pharmacy team at Cumbria, Northumberland, Tyne and Wear (CNTW) NHS Trust was used for this purpose. This meant that site team were only responsible for completion of prescription for their participants and then securely e-mailing these to the central pharmacy team in a timely manner. The central team then dispensed and posted medication to all trial participants via Royal Mail.

In addition to a central pharmacy team, a central team of RAs were also involved in PAX-BD. The central RAs carried out the majority of participant follow-up contacts via telephone and text message, supported participants with the set up and completion of questionnaires via TrueColours (prompting participants when they were due), ensuring medication was received and taken appropriately [and escalating any concerns to the Chief Investigator (CI) and local clinician/principal investigator (PI) as applicable] and reporting any AEs reported to them via the regular telephone contacts.

The trial included a qualitative substudy and an economic evaluation.

PAX-BD included a 12-month internal pilot, as stated in the published study protocol.57 The funder ‘stop’ criteria were ≤ 50 participants randomised at 12 months (from the date the first participant was recruited to pre-randomisation) and with a ≤ 70% retention of those randomised at the primary outcome time point (12 weeks). These criteria were somewhat overtaken by the internal pilot period falling coincident with the ‘lockdowns’ related to the COVID pandemic. Consequently, the continuation of the study was discussed periodically with the funder on an ad hoc basis.

Patient flow through the trial, and an overview of the trial stages and phases are shown in Figure 24 [an a priori Consolidated Standards of Reporting Trials (CONSORT) diagram] and Figure 25 in Appendix 2.

Sites

The trial included 21 participating Secondary and Mental Health NHS Trusts across the UK. Site set up commenced in August 2019. Details of sites, date of opening and the number of participants recruited are shown in Table 25, Appendix 2.

Participants

The trial aimed to recruit participants with TRBD. This was defined as the failure of two medications recommended for BD by NICE or the British Association of Psychopharmacology (BAP).3,10 Eligibility criteria are described below.

Sample size calculation

PAX-BD required a pre-randomisation stage to allow for adjustment to antipsychotics and commencement of non-antipsychotic mood-stabilising medication (lithium, lamotrigine, valproate, carbamazepine), where necessary. A 30% dropout during this pre-randomisation stage was estimated based on the experience in the BALANCE study in BD run primarily in the UK.58 This trial required participants to be switched on to lithium and valproate over a 4- to 8-week period and then be randomised to have one or other or neither.

Dropout rates during the randomised phase were estimated on the basis of the CEQUEL study,59 which is the closest in nature to the PAX-BD trial on the basis of the nature of the patients (suffering from a depressive episode in the context of BD and with a clinical decision that a change in medication was required), the remote collection of data and the timings of the primary and final data collection points (12 and 52 weeks, respectively). Dropout rates in CEQUEL were 20% at 12 weeks and 50% at 52 weeks.59

The power calculation based on a two-sample t-test at 12 weeks detecting a 3-point difference in QIDS-SR between drug and placebo (at p < 0.05) with a QIDS-SR SD of 7 based on QIDS-SR data at the 12-week time point in CEQUEL.59 Three QIDS-SR points equated to Cohen’s d = 0.4 (which has previously been argued to represent a clinically meaningful effect size).60 For 90% power, 232 (116 per arm) participants were required to complete the trial, meaning a sample size of 290 at randomisation, assuming a 20% dropout rate at 12 weeks as described above. It also provided an 80% power for detection of a 3.3-point QIDS-SR difference between drug and placebo at 48 weeks, assuming the 50% dropout rate seen at 52 weeks in CEQUEL,59 was representative and using the above SD estimate as the most appropriate available.

Given the estimate of a 30% dropout during the pre-randomisation stage, it was estimated that a population of 414 participants would be required. The trial closed early and recruitment to the pre-randomisation stage ceased on 5 April 2022 with a sample size of 51 recruited. Participants already in the pre-randomisation stage were allowed to progress to the randomisation stage, if confirmed as eligible, until randomisation closed on 14 June 2022 with a sample size of 39 participants randomised.

Subsequently, a revised sample size calculation was conducted based on recent data, suggesting a more appropriate minimal clinical important difference (MCID) on the QIDS-SR of 4, rather than 3, points,61,62 and early observations that the dropout rate in the pre-randomisation stage was around 10%, and the dropout rate between randomisation and week 12 was 5%. This produced estimated required sample sizes of 155 participants to recruit to the pre-randomisation stage, 139 to randomise and 132 to reach the 12-week primary outcome time point for 90% power, or 118 recruited, 106 randomised and 100 reaching 12 weeks for an 80% power.

Recruitment

Recruitment for the trial took place in the aforementioned 21 NHS Trusts across the UK. Of the 21 sites listed, 12 of these randomised at least 1 participant into the randomisation stage. Interviews were also carried out with site staff involved in the trial as part of the qualitative element of the study.

Screening and consent

Patient identification

Suitable patients were opportunistically identified from secondary care services by clinicians at sites using patient clinic lists, databases and/or research registers. Clinicians were assisted by Clinical Studies Officers (CSOs) or other staff from the Clinical Research Networks (CRNs) where possible. The trial was also advertised to participants directly via a postal invitation letter and summary leaflet, recruitment posters and flyers in secondary care clinics, websites, social media and via patient support groups (e.g. the Bipolar Organisation).

Screening logs were completed by recruiting sites to record reasons for participants declining to take part in the trial and reasons for ineligibility.

As part of the screening process, participants were allocated a unique participant identifier. For those who went to be consented and confirmed as eligible, this unique identifier was then used to identify them throughout their duration in the trial.

Provision of trial information for pre-randomisation stage

Potentially eligible participants were approached by a member of the clinical team at a standard clinic visit and asked for verbal consent for a member of the trial team to get in touch regarding potential participation in the trial. The participants were given a participant information sheet (PIS, see Report Supplementary Material 1) to consider.

Potentially eligible participants were also approached via the provision of an invitation letter and summary sent by post or e-mail. Participants then had the option to contact either the clinical or research team, if they were interested in receiving further information about the trial. This was done via completion and return of a reply slip attached to the invitation letter. A copy of the full PIS was then provided by post, e-mail or at a standard clinic visit.

All participants were given a minimum of 24 hours to consider participation. A member of the trial team then contacted the patient to ask if they did wish to proceed, and if this was the case a screening contact was arranged. Routine care continued for all participants alongside trial participation.

Patient consent to pre-randomisation stage

The initial screening/consent discussion contact was arranged at a location convenient to the participant (e.g. in clinic when attending a standard clinic appointment, at the patient’s home or via telephone/teleconference or videoconference).

The need to use highly effective methods of contraception was discussed during this to ensure all women of child-bearing potential were aware of the risks of becoming pregnant while taking part in the trial.

Written informed consent was received by a General Medical Council (GMC) registered, medically trained doctor.

If the initial consent discussions (including pregnancy/contraception discussions) took place via telephone/teleconference or videoconference, written informed consent still had to be obtained in person (either at the site or at the patient’s home) no later than 72 hours after the initial consent discussions.

Participants who agreed to participate in the trial were also asked if they would be willing to take part in a qualitative interview.

Eligibility assessment and confirmation for pre-randomisation stage

Some material below is reproduced with permission from the study protocol by Azim et al.57 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) licence, which permits others to distribute, remix, adapt and build upon this work, in any medium and for any purpose, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The text below includes minor additions and formatting changes to the original text.

After consent, eligibility for the pre-randomisation stage was confirmed via the following:

• A urine sample pregnancy test for all female participants of child-bearing potential. The sample was either collected at the site, from the patient’s door by a member of the site team, delivered in person by the participant to the site, or posted back to the site team by the participant. Posting of the sample was used as a last resort as it was imperative that the samples were analysed ideally no later than 48 hours after collection to ensure validity of the results.
• Medical history and assessment of diagnosis/diagnoses according to Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-V) criteria. The assessment was completed either at a face-to-face visit or over the phone/via videoconference.
• Completion of Bipolar Demographics and Treatment Questionnaire (BDTQ) (via face-to-face visit or over the phone/via videoconference).
• QIDS-SR (paper) was self-completed by the participant using one of the following options:
  • face-to-face visit
  • a copy of the QIDS-SR posted or e-mailed, with answers then provided securely via e-mail or over the phone/via videoconference
  • made available via the trial website to allow participants access in advance of a screening telephone call/videoconference
  • questionnaire displayed via screen share during videoconference
  • completed over the phone with the questions read out to the patient by the site team as a last resort.

Eligibility was confirmed by the PI at site, or a GMC registered medically trained doctor.

For patients confirmed as eligible, a member of the site team contacted the trial RAs as soon as possible to provide them with the participant’s details.

Patients who did not wish to take part in the trial, or did not meet the eligibility criteria for pre-randomisation, continued with their standard treatment pathway and were considered as ‘screen failures’.

Eligibility: pre-randomisation stage

Inclusion criteria

1. Currently under the care of secondary care mental health services at screening with a plan for the patient to remain in secondary care throughout the period of the trial.
2. A decision made by the patient’s clinical team that a change in medication is indicated.
3. A current diagnosis of bipolar (type I or II), defined as in DSM-V, which is supported by the use of the Mini-International Neuropsychiatric Interview (MINI).63
4. Currently depressed, that is meeting DSM-V criteria for a major depressive episode (MDE) assessed via MINI and with a current QIDS-SR > 10.
5. Current episode of depression failed to have responded to adequate trials, or lack of tolerability or patient refusal, of two different NICE-recommended medications (quetiapine, olanzapine + fluoxetine, lamotrigine) or lurasidone. Adequacy of treatment trial defined using a custom-designed BDTQ – see Report Supplementary Material 5.
6. Aged 18 or over at the point of consent.
7. Willing and able to provide written informed consent prior to any trial procedures taking place.
8. In the opinion of the investigator, is able to follow the trial prescription instructions and is able to manage 8 weeks supply of trial medication without risk of overdose.
9. The patient, if female and of child-bearing potential, must have a negative pregnancy test [urine beta-human chorionic gonadotropin (β-hCG)].
10. Women of child-bearing potential are required to use a highly effective contraceptive method.

Exclusion criteria

1. DSM-V defined severe substance use disorder.
2. Current psychotic symptoms as assessed using the MINI.
3. History of retinal disease.
4. Current cardiovascular symptoms or significant concerns around cardiovascular disease.
5. History of renal disease.
6. Any known sensitivity to trial drug including its excipients.
7. Current pregnancy or planned pregnancy during the trial period, or breastfeeding.
8. Starting specific psychotherapy from 4 weeks before randomisation through to week 12 post randomisation.
9. Currently taking part in another clinical trial that would interfere with the outcomes of PAX-BD [site team to check with the CI and Trial Management Group (TMG) if in doubt].
10. Confirmed diagnosis with potential confounding factors such as Parkinson’s disease, restless leg syndrome.
11. Clinical concern of previous impulse control behaviours including harmful alcohol or drug use, binge eating, gambling or sexual behaviours, or regarding significant suicidal risks.

Updates to eligibility criteria

A few minor updates were made to the eligibility criteria, mostly by way of clarification, during the course of the study. These are listed below (changes underlined and in italics):

• Inclusion criteria clarification: ‘Current episode of depression failed to have responded to adequate trials, or lack of tolerability or patient declined/clinically inappropriate, of two different NICE recommended medications (quetiapine, olanzapine (with or without fluoxetine), lamotrigine) or lurasidone. Adequacy of treatment trial defined using a custom designed “Bipolar Demographics and Treatment Questionnaire” (BDTQ)’.
• Exclusion criteria clarification: ‘History of renal disease (for example within the last 6 months eGFR is less than 50 ml/min/1.73 m² or there is a concern that eGFR is deteriorating and may be expected to fall below 50 during the course of the study)’.
• Exclusion criteria clarification: ‘Confirmed diagnosis with potential confounding factors such as Parkinson’s disease, restless leg syndrome (where restless legs syndrome has been formally diagnosed by a sleep clinic)’.
• Exclusion criteria relating to impulse control concerns were updated to simply read: ‘Significant clinical concern regarding impulse control behaviours’.

All updates received required regulatory approvals prior to implementation at sites.

Progression to randomisation stage

Provision of trial information for randomisation stage

The RAs confirmed to the appropriate site team, once a patient was potentially eligible for randomisation. The site team then provided the participant with the randomisation information sheet (RIS, see Report Supplementary Material 2), at standard clinic visit, via post or e-mail, and arranged a screening and randomisation contact (again observing the 24-hour minimum time frame for the participant to read the RIS).

Patient consent to be randomised

As with the pre-randomisation stage, this contact could take place at a local clinic, at the patient’s home, via telephone/teleconference or videoconference.

Again, during the informed consent discussions, female participants of child-bearing potential were reminded of the risks of becoming pregnant while taking part in the trial.

Again, as with the pre-randomisation stage, the informed consent discussion could take place via telephone/teleconference or videoconference, but written informed consent still had to be obtained in person, no later than 72 hours after the initial consent discussions.

Willingness to take part in a qualitative telephone interview was also discussed again.

Eligibility assessment and confirmation for randomisation stage

After consent to the randomisation stage the following assessments were completed in order to confirm eligibility:

• A urine sample pregnancy test for all female participants of child-bearing potential.
• QIDS-SR.

The same options for completion were available as that for the pre-randomisation stage.

Eligibility was confirmed by the PI at site, or another GMC registered medically trained doctor.

These participants remained under the care of their usual treating team outside of the trial.

Eligibility: randomisation stage

Some material below is reproduced with permission from the study protocol by Azim et al.57 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) licence, which permits others to distribute, remix, adapt and build upon this work, in any medium and for any purpose, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The text below includes minor additions and formatting changes to the original text.

Inclusion criteria

1. Currently depressed, that is meeting DSM-V,64 criteria for a MDE and with a current QIDS-SR > 10.
2. A minimum of two telephone calls with a trial RA and two online weekly symptom ratings have been completed during the pre-randomisation stage.
3. On mood-stabilising medication (lithium, valproate, carbamazepine, lamotrigine).
4. All regular psychotropic medication, including mood stabilisers, at a stable dose for a minimum of 4 weeks.
5. The patient, if female and of child-bearing potential, must have a negative pregnancy test (urine β-hCG).
6. Women of child-bearing potential are required to use a highly effective contraceptive method during the post-randomisation stage of the trial.
7. Willing and able to confirm written informed consent at the point of randomisation, after the pre-randomisation period.

Exclusion criteria

1. Psychotic symptoms over the preceding 4 weeks.
2. Any known sensitivity to trial drug including its excipients.
3. Any deterioration in physical or mental health since pre-randomisation that means there is a clinical concern to proceed with the study.
4. On an antipsychotic at the point of randomisation.
5. Current or planned pregnancy during the trial period, or breastfeeding.
6. Starting specific psychotherapy from 4 weeks before randomisation through to week 12 post randomisation.
7. Currently taking part in another clinical trial that would interfere with the outcomes of PAX-BD (site team to check with the CI and TMG if in doubt).
8. Confirmed diagnosis with potential confounding factors such as Parkinson’s disease, restless leg syndrome.
9. Clinical concern of previous impulse control behaviours including harmful alcohol or drug use, binge eating, gambling or sexual behaviours or regarding significant suicidal risks.
10. Any trial team’s concern regarding the patient’s ability to remain engaged in the trial collecting self-ratings of their symptoms.

Updates to eligibility criteria

Changes were made to the eligibility criteria based on screening data (as this became available) and feedback received from sites. The major change made to the eligibility criteria allowed the continued use of specific antipsychotics at specific doses. Changes were as follows (changes underlined and in italics):

• A new randomisation stage inclusion criterion added: ‘If on an antipsychotic this must be one listed, and at a dose of no more than the maximum stated’. Full details of medications/doses allowed were provided in the protocol.
• An exclusion criterion was removed: ‘On an antipsychotic at the point of randomisation’.
• An inclusion criterion was further clarified: ‘All regular psychotropic medication, including antipsychotics and mood stabilisers, at a stable dose for a minimum of four weeks. Additionally, if a participant is on lamotrigine, quetiapine, olanzapine or lurasidone then this must have been at the current dose or higher for a minimum of three months’.

This amendment had the support of the PPI Group, Trial Steering Committee (TSC), funder and received full regulatory approvals via Substantial Amendment 11 dated 18 February 2021.

As a result of the above and the evidence outlined in Chapter 1, PAX-BD allowed randomisation of participants if they were taking any of the antipsychotics listed in Table 24 in Appendix 1 at a dose no greater than the maximum stated dose. No other concomitant treatment with antipsychotics was allowed at the time of randomisation. Participants recruited to the pre-randomisation stage of the trial who were on antipsychotics not included in Table 24, or at doses above those allowed, had to have these treatments adjusted to be consistent with what was allowed as defined in Table 24. This was managed by the participant’s own local clinical team. To preclude the possibility that any change in depression severity could be attributable to ongoing, or newly initiated, treatment with an antipsychotic, antipsychotics must have been used by the participant at a stable dose for a minimum of 4 weeks. Additionally, if a participant is on lamotrigine, quetiapine, olanzapine or lurasidone, then this must have been at the current dose or higher for a minimum of 3 months. If the participant has received ECT, the last session must have been > 28 days prior.

Additional minor changes to eligibility, on top of those described above for the pre-randomisation stage and specifically related to the criteria for the randomisation stage, were:

• An inclusion criterion was added: ‘Been in pre-randomisation stage for a minimum of 23 calendar days’.
• An exclusion criterion was added: ‘Electroconvulsive therapy (ECT) in the last 28 days’.
• An exclusion criterion was further clarified: ‘Any study team’s concern regarding the patient’s ability to remain engaged in the study collecting self-ratings of their symptoms and undertake all study procedures’.

Randomisation

Eligible participants were randomised anonymously on a 1 : 1 ratio using a central, secure, 24-hour web-based system called Sealed Envelope, to receive either pramipexole or matched placebo in addition to mood stabilisers. Randomisation was carried out by trained members of the research team at each site who had been delegated this task via the delegation log.

Randomisation took place as soon as possible and no more than 2 weeks after a participant had been confirmed as eligible. If randomisation did not take place within 4 weeks of confirmation of eligibility, eligibility had to be reconfirmed, including a repeat pregnancy test for women of child-bearing potential, before randomisation could take place.

Participant allocation

A non-deterministic minimisation algorithm was used to produce treatment groups balanced for important prognostic factors. The first 10 participants were allocated randomly without minimisation to avoid predictability. Subsequently, the minimisation algorithm was applied with an allocation ratio that was not fully deterministic: an 80% bias in favour of allocations that minimise the imbalance. The randomisation algorithm minimised for nine variables related to prognosis at baseline:

• bipolar I or bipolar II (based on DSM-V criteria)
• severity of depression at randomisation (QIDS-SR) – three categories: moderate 11–15, severe 16–20 and very severe > 20
• age – two categories: 18–50 and > 50
• biological sex – two categories: male and female
• site region – eight categories: North, Midlands and East, London, South East, South West, Scotland, Wales and Northern Ireland
• concurrent mood stabiliser – six categories: lithium, valproate, lamotrigine, carbamazepine and multiple mood stabilisers and no mood stabiliser.
• concurrent antidepressant (Y/N)
• on an antipsychotic at randomisation (Y/N) [This was changed from ‘Withdrawn from an antipsychotic during the pre-randomisation stage (Y/N)’ as part of Amendment 5 (see Table 27, Appendix 1).]
• number of mood episodes in the past year – two categories: < 4 and ≥ 4.

Blinding

PAX-BD was a double-blind trial. The Investigative Medicinal Product (IMP) and matching placebo were delivered blinded by MODEPHARMA to CNTW NHS Foundation Trust. Trial medication was dispensed at St Nicholas Hospital, then distributed directly to participants via secure post (in batches).

As the trial was double blind, neither the participants nor the local site teams were aware of which arm a participant had been allocated to.

Unblinding

While participating in the trial, participants could be unblinded if this was required due to a clinical emergency.

While it was not planned that all participants would be routinely unblinded at the point they had completed the trial, in the lead up to their final trial visit participants were asked to consider if they would want to continue taking pramipexole outside of the trial. Those that indicated they would like to continue taking pramipexole outside of the trial as part of their final visit were then unblinded after the final visit assessments had been completed. This then allowed the local site team to make provisions for ongoing open-label access to pramipexole to be put in place outside of the trial.

In both scenarios unblinding was carried out by the site PI, or another member of the team delegated this responsibility via the delegation log, by accessing the 24-hour randomisation system Sealed Envelope.

The CI and a team of delegated psychiatrists were also available via an out-of-hours on-call system called ‘Out of Hours Research Mental Health’. The on-call psychiatrists had access to a trial-specific unblinding standard operating procedure (SOP) and were available as a back-up to aid sites or clinicians with emergency unblinding in the event that a delegated member of the site team was not available to perform this.

Details of trial interventions

Participants were randomised to one of two groups:

• pramipexole
• matched placebo.

For both groups, this was in addition to ‘mood stabilising’ medication (defined as at least on of lithium, lamotrigine, valproate or carbamazepine).

Delivery of trial interventions

Participants received trial medication for self-administration for up to 52 weeks (dependent upon when they joined the trial in relation to the early closure) – this was split into four phases:

• dose titration 0.25 mg increased every 3 days to maximum 2.5 mg depending on tolerability (weeks 1–4)
• fixed dose (weeks 5–12)
• flexible dosing 0.25–2.5 mg/day depending on response and tolerability (week 13 up to week 48/end of trial)
• tapering or continuing IMP (depending on clinical decision).

Note that these weights, and all pramipexole/placebo weights described in this report, are salt weights: 0.25 mg salt is equivalent to 0.18 mg pramipexole base and 1 mg salt equivalent to 0.7 mg pramipexole base weight.

All participants initially followed a 4-week titration schedule starting at 0.25 mg/day in a single dose usually at night for 3 days. The dose was then increased by 0.25 mg/day every 3 days. The target dose was 2.5 mg/day, but titration was based on tolerability and response.

During the fixed dose phase, the dose attained by the end of the dose titration phase was continued with mood stabilisers ideally kept stable throughout the titration and fixed dose phases (unless clinical need dictated otherwise).

During the flexible dosing stage, pramipexole was flexibly dosed between 0.25 and 2.5 mg/day (determined by response and tolerability). Decisions around dose alterations were based on weekly mood scores [QIDS-SR and Altman Self-rating Scale of Mania (ASRM)] and scores from the side-effect items of the Treatment Satisfaction Questionnaire for Medication (TSQM). Patient’s mood and response, and tolerability, were categorised every 4 weeks by the RAs to allow them to provide advice to the local PI regarding the dose of trial medication to prescribe going forward. This advice was provided via a Flexible Dosing Report. It was not mandatory that this advice was followed and the clinical team, together with the patient, could agree to over-ride this if considered necessary; however, this decision had to be clearly documented within the Flexible Dosing Report which was returned to the RAs for their information.

Mood stabilisers could also be adjusted during this phase, but while the participant remained on trial medication, it was recommended that they also remain on at least one of the four recommended mood stabilisers.

At the end of participation (either end of trial or early withdrawal), medication was tapered. Dose reductions mirrored titration, in that participants decreased their dose by 0.25 mg every 3 days. This was to reduce the risk of developing dopamine withdrawal syndrome which can be caused by abrupt withdrawal of dopaminergic therapy.

Prescriptions for trial medication were written by a member of the local site team delegated to do so on the site delegation log. Prescriptions were written, signed and then a scanned copy was securely e-mailed to the central sponsor pharmacy team.

Participants were provided with ‘diaries’ (see Report Supplementary Material 3 and Report Supplementary Material 4) that provided guidance regarding the study medication and allowed them to enter details of the medication they took and any AEs they experience, to aid discussion with the study RAs.

Outcomes

The schedule of events (see Table 26) and patient flow diagrams (see Figures 24 and 25) can be found in Appendix 2.

Data collection

Data were collected via a mixture of methods as part of PAX-BD including via site teams, through scheduled contacts with trained trial RAs and via participant self-reported questionnaires. All participants were allocated a unique trial identification number that was used to identify them across all documentation and databases throughout their participation.

The primary outcome and majority of secondary outcomes were collected via participant self-reported questionnaires using the online TrueColours system.

Data collected by site teams (as part of consent and eligibility checks) and by the RAs as part of their regular contacts with the participants were recorded using electronic case report forms (eCRFs) in the clinical data management system MACRO™, Ennov, Raleigh, USA. Randomisation and allocation of IMP kits were managed via the Randomisation system Sealed Envelope.

Participants could not be identified via the eCRFs as no identifiable data were held in any of these databases (participants were only identified via the unique trial identification number allocated).

Primary outcome

The primary outcome for the trial was assessed by measuring depression symptoms at 12 weeks post randomisation through participant self-completion of the Quick Inventory for Depressive Symptomology questionnaires (QIDS-SR) covarying for QIDS-SR score at baseline. Improvement in scoring was assessed through comparison of the QIDs questionnaires from baseline to 12 weeks. After 12 weeks, the original design was to follow up participants weekly up to 48 weeks; however, with the early closure of the trial, this was reduced for some participants.

Secondary outcomes

The following secondary outcomes were collected:

Secondary efficacy outcomes:

• The impact of pramipexole treatment on mood and anxiety symptoms was examined for the duration of trial involvement, and pleasure symptoms over 12 weeks. This was assessed using the QIDS-SR65 and the Generalised Anxiety Disorder-7 (GAD-7)66 scores, rated weekly from baseline, and the Snaith–Hamilton Pleasure Scale (SHAPS)67 at baseline and weeks 6 and 12.
• Depression response was defined as a QIDS-SR score reduction of ≥ 50% from the baseline measure at randomisation to week 12 and similarly for remission of depressive symptoms defined as a QIDS-SR score ≤ 5 at 12 weeks.
• The impact of pramipexole on psychosocial function was examined for the duration of trial involvement. This was assessed using GAD-7 scores, rated weekly compared to baseline, and the work and social adjustment scale (WSAS),68 at weeks 6, 12, 24, 36 and 48, compared to baseline.

Secondary safety and acceptability outcomes:

• The trial examined known possible side effects of pramipexole: the risk of switching to mania and occurrence of psychosis or impulse control disorders. This was assessed using the ASRM69 weekly.
• Rates of impulsivity during treatment with pramipexole were examined, using the Questionnaire for Impulsive–Compulsive Disorders in Parkinson’s Disease – Rating Scale (QUIP-RS)60 at baseline and weeks 6, 12 and 4 weekly thereafter.
• Side effects and overall acceptability of pramipexole treatment were examined using the TSQM,58 at weeks 6, 12 and then 4 weekly thereafter.
• Tolerability of pramipexole was examined by reporting rates of AEs, SAEs and SUSARs describing severity, seriousness, causality and expectedness.
• Adherence to medication to which participants are randomised was examined using dose taken as reported during RA contacts and from central trial medication accountability and reconciliation records.

Health economic outcomes:

• The quality of life, well-being, health and social care and broader societal costs of participants randomised to either pramipexole or placebo were examined. The incremental cost-effectiveness of pramipexole in comparison to placebo was established, primarily over 12 weeks with secondary analysis exploring the cost-effectiveness over longer follow-up based on available data from the trial. The Health Economics Questionnaire (HEQ) was used to capture information on health and social services utilisation and broader societal costs (e.g. lost productivity, informal care). In addition, the EuroQol-5 Dimensions, five-level version (EQ-5D-5L),10 was used to capture health-related quality of life (HRQoL) and the ICEpop CAPability measure for Adults (ICECAP-A),33,70 and Oxford CAPabilities questionnaire-Mental Health (OxCAP-MH),71,72 instruments to capture broader well-being information. All questionnaires, with the exception of the HEQ, were completed at the start of pre-randomisation, baseline and at week 12, and during further follow-up at 24, 36 and 48 weeks for those who reached these time points. The HEQ was completed at baseline and during further follow-up at 24, 36 and 48 weeks for those who reached these time points.

Comparison analysis outcomes:

• Content validity was increased by using gold-standard observer rated scales to measure participant scores on mania and depression to compare participant scores with previous studies in the field. This was achieved by using the Young Mania Self-Rating Scale (YMRS),73 to capture presence of manic symptoms, and the MADRS,74 and Quick Inventory of Depressive Symptoms, Clinician-Rated (QIDS-C),60 to capture severity of depressive episodes. The scales were administered at baseline and week 12.

Adverse event reporting

Adverse events for this trial were recorded by the trial RAs on the MACRO database from the date of consent to the pre-randomisation stage until completion of the final follow-up assessment.

Line listings of all AEs for a participant were sent to the site PI or delegated clinician for assessment 4 weekly during the pre-randomisation stage, then at weeks 2, 8 and 12, and then 4 weekly until the participant’s final trial assessment during the randomisation stage. The PI or delegated clinician assessed each AE for seriousness, severity and also for causality if the AE occurred after the participant started taking the trial medication.

Adverse events and adverse reactions that met the definition of serious were reported as SAEs or serious adverse reactions (SARs). SAEs occurring from the date of consent to the pre-randomisation stage until the participant’s final trial assessment were recorded as part of the trial. After this point, only any potential SARs that came to the attention of the site team had to continue to be reported up until the point of trial closure.

Serious adverse events occurring before the date the participant consented to the randomisation stage were recorded on the MACRO database only, whereas those occurring after randomisation/commencement of IMP also required completion of a SAE form.

All potential SARs were assessed for expectedness using the MHRA-approved reference safety information (RSI). For PAX-BD the RSI was contained in section 4.8 of the summary of product characteristics for pramipexole 0.7 mg Aurobindo Pharma – Milpharm Ltd (South Ruislip, UK), 5 September 2018.

Adverse events were coded using the Medical Dictionary for Regulatory Activities terminology (MedDRA^®, MedDRA, Herndon, USA).

Compliance and withdrawal

Compliance

Participants took trial medication daily; at the highest dose, they were able to tolerate (up to a maximum of 2.5 mg daily) throughout their involvement in the trial. IMP compliance was checked via RA contacts and recorded centrally on eCRFs by the RAs in order to allow for ongoing accountability monitoring.

Unused medication and bottles were returned by the trial team (local CSO or equivalent) to the central CNTW pharmacy, or where this was not possible by the participant using the prepaid envelope provided. All returned medication was monitored by the Newcastle Clinical Trials Unit (NCTU) Trial Managers and compared against the central pharmacy accountability records. Approval for secure destruction was confirmed and documented by CNTW pharmacy. Pharmacy staff on the delegation log completed paperwork to document details of the IMP/placebo returned, documented when it had been destroyed, and retained copies of these records. The Trial Managers performed and monitored IMP accountability for all sites.

Withdrawal

Participants had the right to withdraw from the trial at any time without having to give a reason. Investigator sites tried to ascertain the reason for withdrawal and documented this on the Withdrawal Form and patient medical notes if available. Participants that withdrew were provided with an end-of-trial information sheet which reminded them how to safely taper down off the trial medication.

The Investigator could also discontinue a participant from the trial at any time if they considered it necessary for any reason.

Participants who withdrew from the trial were not replaced and were not unblinded, unless there was a clinical need.

Data collected prior to withdrawal was retained.

Definition of end of trial

The definition of end of trial was the last patient, last visit date, which took place on 29 October 2022.

Data management

Three systems were used as part of PAX-BD. MACRO was used for the trial database, Sealed Envelope for the randomisation system and TrueColours was used to collect electronic patient-reported outcomes. The MACRO database was built by the trial Data Manager with input from the TMG. Sealed Envelope was configured by the Data Manager with input from the TMG and built by Sealed Envelope. TrueColours was built by the TrueColours team with input from the central RA team.

Participants used TrueColours to complete trial questionnaires at defined time points during the trial. The central RAs supported participants with this and reminded them that these needed to be completed as part of the RA contacts that took place regularly throughout participation.

Sites used the Sealed Envelope system to randomise participants and then to subsequently allocate further IMP kits to participants as they progressed through the trial. The trial protocol provided sites with a dispensing schedule to follow so that each participant was provided with more IMP in a timely manner to prevent any gap in supply. At the appropriate time point, the site team would use Sealed Envelope to allocate more IMP kits based on the participant’s current dose. Allocation of kits via Sealed Envelope had to be done by a GMC-registered clinician allocated this task via the delegation log and this was to ensure medical oversight of the number of bottles generated by Sealed Envelope as part of this process.

The central pharmacy team used the Sealed Envelope database to manage the IMP kits including moving kits between the two central storage locations and making them live for allocation. They also used the system to return unused kits that were allocated in error by site teams, so that they were available to allocate again.

Data were entered into the MACRO database by both the site team and central RA team, with site teams responsible for entering screening, consent, eligibility and end-of-trial details, while the RA was responsible for recording trial contacts, AEs and IMP compliance.

Serious adverse event data were reported by the site team via secure e-mail. This was input centrally into the MACRO database.

Essential data will be retained for a period of at least 5 years following the close of trial in line with sponsor policy and the latest Directive on good clinical practice (GCP).75 Data were handled, digitalised and stored in accordance with the Data Protection Act 1998,76 and the Data Protection Act 2018.77 In line with General Data Protection Regulations, the sponsor acted as the data controller for this trial, and NCTU as data processor.

Data monitoring, quality control and assurance

Monitoring of trial conduct at sites, central RAs and central Pharmacy teams was conducted using a mixture of both onsite and remote monitoring activities to ensure a comprehensive oversight was maintained.

Primary outcome data were participant-reported through TrueColours and could not be edited after questionnaire submission. Similarly, the majority of secondary outcome data and health economics data were participant-reported through TrueColours.

Data recorded on MACRO were monitored throughout the trial by the NCTU Data Manager. Any discrepancies were queried with sites using the MACRO data clarification request function. Any changes made to the data were documented in the audit trial. This included details of who made the change, when the change was made and why the change was made. Source data verification was completed by the NCTU Trial Managers either remotely or during on-site monitoring visits. Data from MACRO were also compared with data contained within the randomisation system (Sealed Envelope) to ensure consistency. Recruitment and retention rates were reviewed regularly at TMG meetings using a data quality report.

Safety was monitored by the Data Monitoring Committee (DMC). Overall supervision of trial progress and conduct was provided by the TSC. The TSC and DMC met at the start of the trial and through the trial. Both committees met regularly throughout the trial.

Analysis

Statistical analysis

Analysis was undertaken following data lock after the closure of the trial and carried out initially on an intention-to-treat (ITT) population. Consideration was to be given to per-protocol analyses, should a population defined in this way deviate sufficiently from the ITT population. This was ultimately found not to be the case.

Analysis of the primary outcome measure of QIDS-SR at week 12 used ANCOVA or related regression methods to examine the difference between the trial arms with adjustment for baseline covariates as far as permitted by the achieved sample size. A two-sided significance level of p < 0.05 was used throughout. Unadjusted analysis using the t-test, or related regression or ANCOVA methods adjusting only for the trial arm, together with other unadjusted analyses, were also undertaken.

Sensitivity analyses were planned, as detailed in statistical analysis plan (SAP), examining the primary outcome, on a per-protocol cohort of participants who had a baseline QIDS-SR score of 10+ at the point of starting medication (n = 33). Based on the observations of the weekly QIDS-SR scores changing gradually over time, additional analyses were conducted of participants remaining in the trial at 24, 36 and 48 weeks – the time points when additional assessments were conducted as defined in the study protocol.

As a result of the early closure of the trial, many of those contributing 12-week data would not be able to supply data for later time points. It was decided to redefine the 48-week time point as an end-of-trial time point, using the last available data after 16 weeks for those affected by the early closure.

Secondary outcomes, other than those collected specifically for the health economic analysis, were analysed in a manner analogous to that for the primary outcome described above.

Weekly repeated measures scores for QIDS-SR were examined to ascertain patient compliance in completion. The final achieved sample size prevented using the originally planned mixed-effects linear regression model accounting for the repeated measures over time. For similar reasons, a number of other originally planned analyses were also unable to be performed.

Safety and tolerability assessments included examination of TSQM and QUIP-RS scores. Additionally, tolerability was assessed through an examination of AEs, SAEs and SUSARs.

Effect sizes (Cohen’s d) were calculated from the difference in mean scores divided by the pooled SDs where relevant.

Full details of the planned analyses are detailed in the SAP. As a result of the early closure of the trial, further ad hoc analyses and analysis populations were considered. Where appropriate, these are described in the results chapter (see Chapter 3).

All analyses were conducted with STATA 16 (StataCorp. Stata Statistical Software: Release 16. College Station, TX, USA: StataCorp LLC; 2019).

Sources of bias

It was a requirement of the trial eligibility criteria that participants must be and remain for the duration of their involvement in the trial, in secondary care mental health services. This was because of the complexity of the use of pramipexole and the risks of serious adverse effects including mania, psychosis and impulse control behaviours which potentially could require urgent specialist intervention. However, this means that the trial population did not include patients with BD being managed only in primary care, who probably constitute the majority of such patients. Further, participant identification was very dependent on on-site clinicians and they may well have biased selection in ways that we are unaware of.

The trial required participants to conduct repeated self-assessments of symptoms using online tools. While there were a few participants included who neither had access to a computer, tablet or smartphone, nor were computer literate, such situations made engagement in the trial much more difficult for participants and site trial staff. This may have been a factor contributing to the recruited population of participants having relatively high levels of education compared with the general population. In addition, all online assessment tools were only available in English. To have had all tools available in multiple languages not only would have been prohibitively expensive, but also would have required extensive validation testing before the tools could have been used. The consequence is that those for whom English was not their first language, or anybody with limited literacy, would have found participation challenging if not impossible. This further limits the generalisability of the results.

Health economic analysis

A health economics analysis was included in the PAX-BD trial in order to examine the quality of life, well-being, health and social care and broader societal costs of participants randomised to either pramipexole or placebo.

The aim was to establish the incremental cost-effectiveness of pramipexole in comparison to placebo primarily over 12 weeks with secondary analysis exploring the cost-effectiveness over longer follow-up based on available data from the trial. In order to achieve this, a HEQ was utilised to capture information on health and social services use and broader societal costs. In addition, the EQ-5D-5L,78 captured HRQoL and the ICECAP-A,79,80 and OxCAP-MH,81,82 instruments captured broader well-being information.

Qualitative substudy

Qualitative interviews were conducted by the trial RAs and qualitative researchers. Interviews included both site staff and participants.

Staff interviews included PIs, clinicians and CRN staff (such as CSOs/Research nurses/trial co-ordinators) at participating sites that were open to recruitment.

The purpose of these interviews was to discuss any barriers and facilitators to recruitment and retention that sites experienced while taking part in PAX-BD. Verbal consent was obtained from staff over the phone prior to the interview.

For the participant interviews, it was intended that a selection of participants that were recruited to pre-randomisation and a selection of participants that were randomised would be interviewed.

The purpose of the interviews in pre-randomisation was to review how many participants were on antipsychotics and on mood stabilisers at the point of consent, the time taken to adjust antipsychotics and their success rates, and the time taken to establish participants on an alternate mood stabiliser and its success rate.

The purpose of interviewing those randomised was to explore retention rates and reasons for discontinuation where applicable. These interviews would primarily include participants still taking part in the trial at the week 12 primary outcome point.

Verbal consent was obtained from participants over the phone prior to the interview. Interviews were directed by two themes: barriers and facilitators. Interviews were transcribed and transcripts were coded and reviewed by clinical and non-clinical staff. Then, the interpretation of the transcripts was guided by an a priori framework of subthemes (see Appendix 5, A priori framework for analysis of barriers and facilitators to recruitment and retention to PAX-BD trial) from a meta-analysis of studies in depression.83

Ethics and governance

The sponsor for the trial was CNTW NHS Foundation Trust (reference RES-17-031). Clinical Trial Authorisation was received from the MHRA on 5 August 2019 (Eudract Reference 2018-002869-18). A Favourable Opinion was granted by the North East – Newcastle and North Tyneside 2 Research Ethics Committee on 4 September 2019 (REC reference 19/NE/0233). Approval was also sought and received from the Health Research Authority (HRA) on 4 September 2019. Local approval was subsequently granted by each participating trust.

All required approvals were sought prior to any protocol amendments being implemented throughout the duration of the trial.

Participant expenses

Trial participants received three £25 gift vouchers as a token of thanks for their participation. These were distributed at 12, 36 and 52 weeks/final trial visit. Following the decision to close the trial early, these timings may have differed slightly by participant. Participants received a total of £75 in gift vouchers.

Trial management and oversight

A TMG met regularly throughout the trial. The TMG was responsible for the day-to-day management of the trial and for monitoring the conduct and progress of the trial. The TMG comprised the CI, co-applicants, statisticians, health economists, qualitative researchers, project manager, CRN representative, sponsor representative, pharmacy representative, patient representative and NCTU trial management team [Trial Manager, Senior Trial Manager, Clinical Trial Administrator, and Data (base) Manager].

Both a DMC and a TSC were established as part of the PAX-BD trial.

The role of the TSC was to act as the oversight body on behalf of the Sponsor/Funder. The TSC comprised an independent clinical Chairperson, two further independent clinicians, an independent statistician, an independent health economist, and an independent lay member. The TSC met at the start of the trial and then three further times throughout the duration of the trial. A joint TSC/DMC was also held at the end of the trial to go over the initial analysis and results.

The role of the DMC was to monitor the key trial outcome measures including safety and efficacy and monitor overall trial conduct. The DMC comprised an independent clinical Chairperson, an independent Psychiatrist, an Independent Statistician and an Independent PPI member. The DMC met at the start of the trial and then three further times throughout the duration of the trial. As mentioned above, a joint TSC/DMC meeting was also held at the end of the trial.

Patient and public involvement

Patients and carers, including from a Bipolar Support Group, contributed to the selection of the outcome measures used in the trial and to the trial planning. A trial PPI group was set up and it met numerous times during the trial. The group was involved in protocol development, reviewed patient-facing documents and inputted into discussions regarding ethical issues around recruiting and retaining participants in the trial. The group was also involved in training the central RAs before the trial commenced. The PPI group lead was invited to all TMG and core team catch-up meetings as an equal partner, with PPI as a standing agenda item.

Changes to trial design

Significant Protocol amendments made during the trial are listed in Table 27 within Appendix 2. All amendments received HRA approval, favourable Research Ethics Committee opinion and MHRA approval as applicable prior to implementation.

Impact of COVID-19

In line with the COVID-19 national guidelines for pausing recruitment and prioritising resources to pandemic essential clinical trials, the trial submitted amendment 5 to formally pause the trial to recruitment as of 23 March 2020. In order to enact necessary changes and introduce greater flexibility for face-to-face versus remote activities, amendment 7 was given approval on 17 August 2020. Approval from the MHRA to lift the pause of recruitment was given on 28 August 2020.

The amendments to the protocol allowed for greater flexibility in trial processes and comprised:

• removal of blood pressure and pulse measurements as agreed by the DMC. Additional updates to safety measures for monitoring
• updating of the protocol and consent procedures to allow for remote discussions and completion of questionnaires
• introduction of doorstep collection by CSOs for used medication and empty bottles, urine samples, provision of diaries and collection of completed diaries
• update to methods of provision of trial information and completion of additional forms by post
• update to pharmacy processes regarding methods of returning empty bottles.

Additional administrative changes took place to reflect the changes to practice across trial documents.

Trial registration and protocol availability

PAX-BD was registered via International Standard Randomised Controlled Trials Number under ISRCTN72151939 on 28 August 2019.

The trial protocol was published in BMC Psychiatry in 2021.57