Statement and interpretation of results
The primary outcome in the trial was negative with p = 0.0865, potentially reflecting the small sample size included in the analysis (n = 36) due to the trial being stopped prior to recruitment of the target sample size. Nevertheless, it does appear that pramipexole may well be having a positive effect on mood in participants with TRBD, in line with the two previously published pilot studies,28,29 and clinical experience.27 Numerically, the improvement in QIDS-SR score between baseline and 12 weeks averaged 4.4 points in the pramipexole arm. This magnitude of change would be regarded as clinically significant,61,62 and was over double that seen in the placebo arm. Over the course of follow-up in the trial, participants taking pramipexole also experienced twice as many depression-free weeks as those on placebo, though this was not statistically significant. However, secondary and post hoc analyses revealed a statistically significantly greater reduction in QIDS-SR score at 36 weeks of follow-up and significantly greater rates of response (46% vs. 6%; p = 0.026) and remission (31% vs. 0%; p = 0.030) for participants taking pramipexole at the point of their exit from the trial.
A positive effect of pramipexole is also supported by the statistically significant improvement in work and social function at 36 and 48 weeks and HRQoL over the 48-week follow-up period compared with placebo. While not statistically significant the increase in ability to experience pleasure at 6 weeks, as shown by the SHAPS scores being 2.04 (95% CI −0.11 to 4.20; d = −0.76) lower, and improvement in anxiety symptoms at 36 weeks, shown by GAD-7 scores being 3.44 (95% CI −1.06 to 6.99; d = −0.67) lower, for participants in the pramipexole arm is consistent with the change in work and social function and quality of life. It is tempting to speculate that the early increase in pleasure may be a mediating factor related to improvement in mood. It is also perhaps not surprising if psychosocial function and quality-of-life measures take longer to change subsequent to an improvement in mood. Essentially, the relative time scales of the changes seen on the various measures have face validity.
While the overall tolerability and acceptability of pramipexole were good, its use is not without challenges. High rates of impulse control problems have been reported in patients with Parkinson’s disease treated with pramipexole and other dopamine agonists.111 While there were no significant differences in ratings of impulsivity between the two treatment arms, 33% of participants in the pramipexole arm reported an AE related to such problems, compared with 19% in the placebo arm. Interpretation of these data is complicated by the association between impulsivity with bipolar disorder per se,112 and the overlap between such problems and hypomania/mania.
However, the biggest issue in the PAX-BD trial was the occurrence of hypomanic and manic symptoms which resulted in the only medication-related SAE and AEs in 8 of the 16 participants in the pramipexole arm compared with 6 out of 20 in the placebo arm. This increase in hypomanic/manic symptoms occurred early, as shown by the weekly ratings illustrated in Figure 10, and by the rate being significantly higher at 12 weeks in the pramipexole arm.
The use of an antipsychotic in combination with pramipexole is contentious. While it has been reported that the combination can be used successfully to treat BD in clinical practice,24 the commissioned call for a study of pramipexole for the treatment of TRBD by the NIHR HTA panel specifically required participants to not be taking an antipsychotic. This, most probably, was due to the view that it was illogical to combine pramipexole (a dopamine agonist) with an antipsychotic (a dopamine antagonist). However, this takes a simplistic view of the drug’s pharmacology. Pramipexole is a relatively ‘clean’ dopamine agonist, but it has activity at three dopamine receptors: D2, D3 and D4. The binding affinity of pramipexole is highest at the D3 receptor (Ki = 0.5 nM) – around an order of magnitude greater than at D2 (Ki = 3.9 nM) and D4 (Ki = 5.1 nM) receptors.31–33 The pharmacology of antipsychotics varies, but in general, they have a higher affinity for D2 receptors compared with D3 (see Table 23 in Appendix 1). Genetically modified mice that do not express D3 receptors have been reported to exhibit depressive and anxious features.70 In addition, a wealth of data using highly selective D3 antagonists in animal models has demonstrated that D3 receptors play a critical role in reward processes.71,72,113,114 Importantly, levels of baseline reward processing are associated with pramipexole response in those with depression.115 Further, PET imaging data in humans suggest D3 receptor expression may be related to motivation for rewards.37 Such findings support a hypothesis that the mechanism of action of pramipexole in BD may be mediated via D3 receptors. On the basis of these arguments, and the high rate of use of antipsychotics in BD in clinical practice (data from POMH – reporting 80% usage in 6025 patients in 2018) hampering recruitment, the NIHR HTA panel agreed to a PAX-BD protocol amendment to allow participants to be randomised still taking an antipsychotic if this was with certain limits based on the drugs pharmacology (see Table 24, Appendix 1). While the numbers of those participants in the pramipexole arm who were and who were not taking an antipsychotic are small, it appears that coadministration had little impact on the positive effects on depressive symptoms, but probably reduced the risk of manic symptoms. It is noteworthy that the one participant experiencing a SAE, and seven who had an AE, related to mania while on pramipexole were not taking an antipsychotic. Conversely, only one participant on an antipsychotic experienced a hypomanic AE when taking pramipexole.
Health economic analysis is limited by the small sample size, but as stated above, there were significant advantages in favour of pramipexole in HRQoL over the 48 weeks of follow-up in the trial. With regard to health and social care, pramipexole was associated with a cost saving of £52 at 12 weeks and £798 at 48 weeks, with NMB of around £200 and £1500, respectively. This implies that the treatment is cost-effective. There was a 90% chance that the threshold for cost-effectiveness of £30,000/QALY gained was met. From a societal perspective, the situation was slightly different with pramipexole being more expensive than the control arm, with greater costs of £794 at 12 weeks and £1170 at 48 weeks. As a result, there was < 50% probability that the cost-effectiveness threshold would be met.
An extensive qualitative review of the PAX-BD trial was conducted, particularly focusing on the barriers and facilitators to both recruitment and conduct of the trial. Essentially this demonstrated the feasibility of the largely remote methodology used in the trial. This trial design was put into place pre-COVID-19 and required little adjustment following commencement of the pandemic. Dropout rates in the trial were similar or lower than in previous studies in similar populations of patients. In the pre-randomisation stage, the dropout rate was 24%, identical to the run-in phase prior to randomisation in the CEQUEL study,59 and similar to the 28% in the BALANCE study,58 which also required potential changes in participants medication prior to randomisation. The dropout rate prior to the primary outcome time point was low at 7.7%, below that seen over the same time period in the CEQUEL study (18.8%). This may have reflected the frequent contact with the study RAs. Completion rates of the participant-rated weekly questionnaires were high at around 80%, again similar to that seen in the CEQUEL study.59 Again, this may have been related to the frequent contact between trial participants and the central trial RAs based in Newcastle. A concern regarding this was whether such frequent contact might increase the placebo response rate. However, there was relatively little evidence of much improvement in symptoms in the placebo arm. The trial required a large number of sites to attempt to recruit to target. One issue that became clear during the running of the trial was that there was a vast gulf in experience in research between the sites taking part in the trial. Some were highly expert in taking part in CTIMPs. For others, they had little or no such experience which contributed to the high rate of protocol deviations. By the end of the trial, 174 deviations had been recorded across the 15 sites that had recruited, with the number per site ranging from 0 to 48.
Comparison with the literature
The potential role of pramipexole as a treatment for depressive episodes in bipolar disorder is supported by several lines of investigation. Pre-clinical investigations have demonstrated that pramipexole can reverse stress-induced suppression of sucrose intake in rats,11 a model of anhedonia which is a core diagnostic symptom of depression. Similarly, antidepressant-like effects of pramipexole have been observed in other animal models of depression or models known to be responsive to drugs with antidepressant efficacy, such as the forced swim test,12,13 social interaction test13 and olfactory bulbectomised rats.14 It has also been shown to increase hippocampal neurogenesis,13 an effect believed to be common to antidepressants.15,16 Pramipexole has extensive evidence for efficacy in Parkinson’s disease,17 for which it has a marketing licence. A meta-analysis of controlled trials of pramipexole for the treatment of motor symptoms in patients with Parkinson’s disease reported an odds ratio of 2.41 (95% CI 1.78 to 4.13; p < 0.001) for improvement in depression severity as measured by a single item on the UPDRS.18 This led to a 12-week randomised double-blind placebo-controlled trial of pramipexole in patients with Parkinson’s disease and significant depressive symptoms.19 A significantly greater decrease in Beck Depressive Inventory score was seen from pramipexole compared to placebo (difference 1.9, 95% CI 0.5 to 3.4; p = 0.01). Path analysis suggested that the effect was a direct one on depressive symptoms rather than secondary via effects on motor symptoms.
Such findings, together with hypothesised roles for a hypo-dopaminergic state underlying BD,20 have led to studies of pramipexole in BD. The use of pramipexole in TRBD is supported by naturalistic and open trial data.21–27 More robustly, two small pilot RCTs, conducted around two decades ago, have examined pramipexole in BD.28,29
Goldberg and colleagues undertook a RCT of 22 patients with BD.28 Participants were mainly suffering from bipolar disorder type I, similar to the population in PAX-BD (72% bipolar type I). All patients had failed to respond to at least two adequate trials of standard antidepressants during the current episode, a slight difference from the way TRBD was defined in PAX-BD. They were treated with a mean of 1.7 mg/day (SD = 1.3) pramipexole per day (slightly lower than the 2.18 mg/day used in the fixed-dose period in PAX-BD) in combination with lithium or anticonvulsant mood stabilisers. The mean 17-item HDRS at baseline in the Goldberg et al.’s study was around 20, which is equivalent to 16 points on the QIDS-SR scale,116 and similar to that in the PAX-BD study (15.1 in the pramipexole arm and 17.3 in the placebo arm). Mean improvements in HDRS scores at 6 weeks were 48% for pramipexole and 21% for placebo (p = 0.05). This equates to a 9-point improvement in HDRS score for participants treated with pramipexole, equivalent to 7 QIDS-SR points and so a greater improvement than seen after 12 weeks in PAX-BD (4.1 QIDS-SR points).
Zarate and colleagues’ RCT included 21 patients with bipolar type II.29 Patients had failed at least one trial of a standard antidepressant, which is a very different criterion to that used in PAX-BD and also somewhat contrary to the notion that antidepressants are ineffective as treatments for BD.3 Study medication was given in combination with lithium or valproate. Similar to Goldberg et al., the mean dose of pramipexole was 1.7 mg/day (range 0.375–4.5 mg/day). The baseline severity of illness was slightly greater than in the Goldberg et al. and PAX-BD study being rated as scoring around 32 on the MADRS, equivalent to 18 on QIDS-SR.117 The reduction in severity of depressive symptoms at 6 weeks was also greater than the Goldberg and PAX-BD studies being around 15 MADRS points, equating to 9 QIDS-SR.
As can be seen, while the reduction in depressive symptoms in the pramipexole arm of the PAX-BD study was of a clinically meaningful magnitude, it was somewhat smaller than that seen in the two previous small RCTs in BD. This may reflect differences in study populations (UK vs. USA; how TRBD was defined), or possibly that some of the participants in PAX-BD were also taking an antipsychotic, though there did not appear to be any difference in reductions in depression in those who were and were not. While the primary outcome was after 12 weeks in the PAX-BD study compared with 6 weeks in the two previous RCTs, dose titration was somewhat slower in PAX-BD. It was noteworthy that the reduction in depressive symptoms appeared to continue to build beyond 12 weeks, with the biggest effect seen 36 weeks after randomisation. This raises questions as to what in fact might be the optimal time point to examine for effects of pramipexole in BD.
It is a challenge to ascribe changes in mood in bipolar disorder to medication, given the nature of the spontaneous fluctuations seen in the disorder. This is a particular issue when considering whether hypomanic/manic symptoms result from the use of pramipexole or are just coincidental. Zarate and colleagues found that while one participant in the pramipexole group experienced a manic relapse, two did so in the placebo arm. Goldberg et al. described one dropout due to a manic relapse but found that formal ratings of manic symptoms did not differ between treatment groups. These findings are in stark contrast to the PAX-BD study where we did observe a significantly greater increase in scores on the ASRM in the pramipexole arm, and the proportion of participants experiencing AEs related to mania was also higher compared with those in the placebo group (44% vs. 29%). This higher rate of problems with hypomania/mania compared with the two previous RCTs may relate to the higher dose used in PAX-BD, or differences in the population being studied.
Despite the major concern regarding impulse control problems in patients being treated with dopamine agonists, particularly pramipexole,118 there was no mention of such problems occurring in the two previous RCTs, perhaps due to the short duration of follow-up. However, no difference was seen in the formal ratings of impulse control behaviours between pramipexole and placebo in the PAX-BD study even over 48 weeks. This may reflect an issue with the use of the QUIP-RS scale, which was designed for use in patients with Parkinson’s disease, in individuals with BD due to the high rate of inherent impulse control problems in BD110 and the overlap with hypomanic/manic symptoms. It could also be that patients with Parkinson’s disease are more prone to developing impulse control problems with dopamine agonists than those with BD. It would, however, be foolhardy to not consider this risk when using pramipexole clinically, especially given the observation in PAX-BD of the higher rate of AEs associated with pramipexole than in the placebo arm (33% vs. 19%).
Commissioned research in bipolar depression
In 2015, the HTA Panel put out a call for ‘Promising pharmacological therapy for treatment-resistant bipolar depression’ (15/33). There was, and remains, a significant clinical need for new medication options for BD, with NICE,3 listing just three options, two associated with a significant side-effect burden (olanzapine and quetiapine) and the other with a limited effect size and slow onset of action (lamotrigine). The intervention described in the details of the HTA call was stated as ‘One or more promising pharmacological therapies, such as pramipexole’. Experts in the clinical management and trial of BD within the NIHR Mental Health Translational Research Collaboration (MH-TRC) discussed the call and concluded that pramipexole was not the most sensible medication to investigate. This was due to the complexity of titration and optimal dose selection, the high rate of impulsivity seen with the drug in Parkinson’s disease and the theoretical risks of a dopamine agonist precipitating mania/hypomania and possibly psychosis, particularly in a population vulnerable to such symptoms. Consequently, a multicentre collaborative group of experts submitted a proposal to undertake a trial of an alternative medication, lurasidone, a second-generation antipsychotic. Lurasidone is currently licensed in the UK for the treatment of schizophrenia, but not BD. Short-term data support the view that it may be an effective treatment for BD as monotherapy, or in combination with a mood stabiliser.50,51 Additionally, its side-effect profile suggests benefits over those seen with quetiapine and olanzapine.119 However, the published studies were not conducted in patients with TRBD, and there are very limited data regarding its longer-term efficacy.120 As a result, lurasidone is not included in NICE guidelines for the management of BD,3 though it is in the BAP guidelines (ref: note that lurasidone was included in the PAX-BD trial as one of the medications included in the criteria to determine TRBD, but it was the least used of the four medications included: < 20% reported either a failure to respond or tolerate the drug, compared with around double this for quetiapine). Critically, a positive finding for the long-term efficacy, safety and cost-effectiveness of lurasidone would have been relatively straightforward to implement in the NHS, with mental health services having expertise in the use of similar drugs and lurasidone itself, in schizophrenia. Unfortunately, the applicants were unable to persuade the HTA panel to fund a trial of lurasidone, with the feedback being that the case was not made that lurasidone was a novel option compared with other antipsychotics. This view is somewhat counter to the evidence that while there is a class effect of antipsychotics in treating bipolar mania, this is most definitely not the case for the treatment of BD, with some antipsychotics even exacerbating depressive symptoms.10
In 2016, the HTA panel were more explicit in their commissioned call 16/154 ‘Pramipexole for treatment-resistant bipolar depression’. Experts within the MH-TRC discussed the call. There was concern around whether a large, multicentre, trial could be conducted safely and effectively, and that even if there were positive findings that pramipexole would be challenging to use within routine secondary care mental health services. This was particularly the case regarding the requirement specified in the HTA call that the drug must be investigated without coadministration of an antipsychotic. As described above, while this appeared to superficially make sense, it ignores the precise pharmacology of pramipexole and most antipsychotics. However, it was agreed that a large, well-designed, trial of pramipexole could potentially provide proof of concept regarding the use of dopamine D3 agonists in the treatment of BD. As a result, the PAX-BD trial application was submitted and funded. Ultimately, the combination of the eligibility requirement for participants not to be on an antipsychotic and the occurrence of the COVID-19 pandemic made recruitment to the trial extremely challenging. Following approval from HTA to significantly modify the antipsychotic requirement (see above), recruitment to the trial increased, but there was limited time prior to the end of the original contract with NIHR and the trial was closed down with the reason being stated as one of the criteria for closure that was to be used in the review of all studies across the NIHR portfolio. However, since the decision was taken before this formal process began, there was no opportunity to appeal the decision.
The results from the PAX-BD study should be treated with caution due to the small sample size, including those that were significant. However, despite the lack of a statistically significant effect of pramipexole on the primary outcome measure, the direction of effect coupled with a number of statistically significant secondary outcome measures are in line with previous small pilot studies,28,29 and clinical anecdotes suggesting that pramipexole may have a positive effect in TRBD.27 The trial also suggests that the drug would be challenging to implement in routine secondary care. Issues were found with the complexity of dosing the drug (20 AEs related to this reported by 9 participants – likely an underestimate due to missed and incorrect doses not being reported by participants), and with high rates of problems with hypomania/mania (particularly if participants were not taking an antipsychotic). Such findings are consistent with the concerns aired within the MH-TRC when the HTA call first came out, suggesting that it probably was not the most sensible medication for a NIHR HTA Panel commissioned call. However, once funded, the closure of the PAX-BD trial at the point where recruitment was occurring means that it is a lost opportunity for definitive evidence regarding the efficacy and safety of a D3 agonist in TRBD, which would have been of importance to highly specialised affective disorder clinics and from a mechanistic perspective.
Limitations and strengths
Limitations
There are a number of issues with the PAX-BD study which limit interpretation of the data. Most notable is the small sample size, with only 36 participants contributing primary outcome data. In addition, due to the early study closure, not all participants were able to continue follow-up to 48 weeks. This resulted in particularly small sample sizes at the 36- and 48-week assessment time points, of relevance since this is the time where some of the significant findings were reported.
The small sample size is likely to have also contributed to the baseline differences in QIDS-SR and HRQoL outcome measures. While analysis adjusted for these differences, there remains a concern that it may have biased the results. However, it should be noted that the baseline severity was lower in the pramipexole arm. In general, larger improvements in symptoms tend to occur in participants with higher baseline starting points and hence more ‘room for improvement’. So it is possible that the baseline differences seen in PAX-BD, rather than contributing to the greater effect seen in the pramipexole arm, actually contributed to this effect not being greater.
The study was designed to investigate the efficacy of pramipexole in TRBD. There is some debate and relatively little consensus as to how TRBD is defined.121 Definitions are complicated by whether lack of response to medications is only counted in the current episode of illness, or over a person’s lifetime. Most commonly, as done in PAX-BD and the two previous RCTs of pramipexole for BD,28,29 non-response is only considered for the current episode. In clinical practice; however, if a person had a previous episode of depression which did not respond to a drug, generally this is rarely tried again in a new episode, not least because it can be challenging for a patient to re-try a medication that they perceive themselves as having failed on. In addition, the challenge of finding an effective drug for a patient is not just limited by non-response to trials of certain medications, but also lack of tolerability. PAX-BD took a pragmatic, clinical approach to defining TRBD. When assessing for criteria of two ‘failed’ treatments in the current episode, in addition to non-response, lack of tolerability and patient refusal or clinical contraindication to a drug were also allowed. Clinical contraindication included a medication that a patient’s depression had been unresponsive to in a previous episode, therefore ruling it out in the current one. This, together with allowing a patient to reject a particular treatment due to the risks associated with it, more approximates clinical reality than a definition of TRBD based solely on non-response in the current episode. However, it does mean that the population included in PAX-BD is likely to be somewhat different from those included in other studies with more conventional criteria of TRBD. Indeed, 24% of participants randomised in PAX-BD had not failed to respond to one of the four recommended drugs in the current episode, and 41% had failed to respond to just one. We do not have data for the lifetime number of failed treatments, but it could be that the PAX-BD population was at least partially as much intolerant of treatment as non-responsive to it.
During the course of the study, an amendment to the protocol allowed for participants to be randomised to study medication while being prescribed an antipsychotic. Around half the randomised participants ended up being coprescribed an antipsychotic. While observation and analysis of the data suggest that this did not reduce the effectiveness of pramipexole, the numbers of participants in each subgroup are such that it is not possible to definitively say that antipsychotic coprescription does not limit the effectiveness of pramipexole. This coprescription also complicates the comparison of the PAX-BD results with those from the two previous small RCTs in which none of the participants were taking an antipsychotic.28,29
Key to achieving a response to medication is adherence to taking it as prescribed. Due to the need to run the study as remotely as possible, opportunities for objective assessments of medication adherence (e.g. plasma drug concentrations) were not possible. Instead, we have had to rely on participant self-report to the study RAs. It could therefore be that response to pramipexole was limited by less than ideally medication adherence rates.
The vast majority of outcome measures used in the study were participant self-reported, as opposed to observer-rated. While this does limit comparisons, for example, with the two previous RCTs of pramipexole in BD,28,29 there are an increasing number of studies using self-report scales to assess mood, for example the CEQUEL study in a similar patient population.59 Computer-administered self-report ratings can help avoid baseline score inflation seen in clinician-rated scores when there is a severity criterion for inclusion into studies,122 and there are instances where self-rated mood is more sensitive to change with treatment than observer-rated scales.123 Nevertheless, it should be noted that there was less evidence of a benefit of pramipexole on the observer-rated MADRS at 12 weeks than on the self-rated QIDS-SR in PAX-BD. This might be in part accounted for the MADRS being administered by phone rather than face to face.
It is important to highlight that the population included in the PAX-BD study is not representative of all patients suffering from TRBD. Many patients currently experiencing an episode of BD, indeed probably the majority, are being managed in primary (rather than secondary) care in the UK, if they are receiving any care at all. Indeed, it is estimated that as many as 10% of patients prescribed antidepressant medication in primary care for depression or anxiety actually have undiagnosed bipolar disoder.124 The eligibility requirement that patients needed to be in secondary care to take part in PAX-BD therefore did not allow for a truly representative sample of all patients receiving care. This criterion was applied due to the complexity of using pramipexole and the need for participants to be carefully monitored for the occurrence of serious adverse effects (such as mania and impulse control problems) and that they could receive timely specialist intervention if needed.
The plan to conduct PAX-BD largely remotely proved to be a good one when the COVID-19 pandemic developed. However, the requirements for outcome measures to be completed online by participants potentially caused a bias in participant involvement. While a few participants did not have access to a computer, tablet or smartphone, and hence had to receive help to complete rate scales from site staff, the reality was that this was too labour intensive to recruit many people in this situation. As a result, the vast majority of participants were computer literate and had their own devices on which they could complete the outcome measures. This may be a factor that led to the population as a whole having a low rate of the most basic level of education. The issue was even more acute for potential participants who were not fluent in English. It would have been prohibitively expensive to have included the ability to complete the various questionnaires in a range of languages, and would have required extensive validation as well. As a result, the study population was unlikely to be fully representative of the UK population, though details of participant’s ethnicity were not collected.
Strengths
The PAX-BD does have a number of strengths as well. Most obvious among these is that the study provides by far the longest follow-up of participants randomised to pramipexole or placebo of any RCT conducted to date (48 weeks vs. 6). As such, the study does provide valuable information regarding the longer-term use of the drug. In addition, relatively high doses of study medication were achieved (higher than the two previous RCTs). In addition, the PAX-BD study is the first RCT in BD to explore the effect of pramipexole on anxiety, pleasure, psychosocial function, HRQoL and cost-effectiveness as well as mood.
As described above, the definition of TRBD was clinically pragmatic and hence the population studies were made up of those for which there is current clinical equipoise as to how they should be treated.
While the sample size was small, the treatment arms were remarkably well balanced with regard to demographics, illness characteristics and medication usage, with the exception of baseline QIDS-SR and HRQoL differences.
Participants in the study had a high rate of completion of online self-related symptom severity scales, comparable with previous studies. In addition, dropout rates were similar, or lower, than seen in previous studies.
In addition to the main quantitative findings, the PAX-BD study was able to acquire qualitative information to help inform the conduct of future trials in this therapeutic area in terms of facilitators and barriers to recruitment and retention. In addition, there has been an extensive review of the PPI involvement in the study, again leading to recommendations for future studies.
Future research
Whether or not pramipexole could be implementable as a treatment for TRBD in routine secondary care, there remains a need for definitive evidence regarding its efficacy, safety and cost-effectiveness. It is likely that pramipexole will continue to be used in highly specialist mood disorder services, such as those in the USA that included patients in the case series described by Fawcett et al.27 However, definitive evidence would aid our understanding of the pharmacological mechanisms that are able to treat BD. In particular, evidence regarding pramipexole would help clarify whether dopamine D3 agonists are of potential use, opening avenues for further drug development. The PAX-BD has established a methodology that allows the conduct of such trials in the UK NHS in a semiremote fashion.
While the PAX-BD study was ended early, the NIHR/Medical Research Council (MRC) Efficacy and Mechanism Evaluation (EME) programme-funded PAX-D study – a RCT of pramipexole in unipolar depression – continues. It will be of interest to compare its findings with those of PAX-BD. There are commonalities in study design and, for example, the SHAPS tool to examine pleasure was included in PAX-BD to enable this cross-comparison between the treatment of bipolar and unipolar depression with pramipexole. Increases in the ability to experience pleasure were seen after just 6 weeks of treatment, prior to improvements in mood and psychosocial function. It is of interest to speculate that this increase in pleasure may be of mechanistic relevance to the improvement in mood. PAX-D may provide further evidence in this regard, and further research in this area with pramipexole and other pro-dopaminergic treatments is warranted.
Given that there are now three completed RCTs of pramipexole in BD, in the absence of funding for a single large definitive trial, a meta-analysis of the data from the three studies would be of value.
The study raised questions regarding the use of antipsychotics in combination with pramipexole. This warrants further investigation, partly because of the inherent value of such information when pramipexole is being used in highly specialist centres, but also because this would further help clarify the relative roles of dopamine D2 and D3 receptors in the mechanism of action of pramipexole, which could inform future drug development.
Running similar studies in the future – lessons learnt
Insights to inform future trial design, from the qualitative data obtained in the trial from participants and HCPs, together with those from the central trial team, are reported below and in more detail in Appendix 7.
Sample size estimates for future studies
PAX-BD is able to provide evidence to help inform sample size calculations for future similar studies.
Examining the QIDS-SR at 12 weeks reveals an overall SD (combined groups) of 6.6, and a difference between arms (pramipexole arm with lower mean) of 8.6. This gives an effect size of 8.6/6.6 = 1.3. A power calculation reveals that 14 participants per arm (28 total) would need to provide 12-week data to detect this effect size with 90% power at a 0.05 significance level. However, this assumes that 8.6 is effectively the MCID. As discussed above, there is evidence to suggest that this is in the order of 4 points when using the QIDS-SR.61,62 This was used for a revised sample size calculation when we were applying to the funder for a costed extension to the study (which was turned down) of needing 100 participants to complete 12 weeks for 80% power or 132 for 90% power. Adjusting this using a SD of 6.6 (rather than 7 as in the original calculation) would mean a sample of approximately 90 needing to complete 12 weeks for 80% power, or approximately 120 for 90% power. The current study would not have had to have continued for a great length of time further for this sample size to have been achieved, illustrating the lost opportunity of the early closure.
Insights from the qualitative analysis
The qualitative data obtained in the trial were examined to identify potential lessons for future studies in similar populations of patients and/or using similar methodology.
Organisational issues
Undertaking research, particularly something as complex as a CTIMP, alongside busy clinical services is a challenge. HCPs identified the importance of the research being seen as clinically important, having full staffing complements and protected time for research. The requirement to inform patients about ongoing trials to facilitate informed choices about treatment was also emphasised.
Given that specialist mental health care is increasingly provided on an episodic basis, with patients discharged back to their general practitioner (GP) in between, the ability to recruit participants from primary care has become more critical, as has the need to be able to advertise studies directly to patients. This can bypass barriers of patient gatekeeping and reliance on staff with low availability, as well as providing access to depressed patients who may otherwise remain without treatment or referral to secondary care. It is also hoped that developments in electronic health records and the ability to search these, in both primary and secondary care, would also greatly facilitate accurate participant identification.
Trial design
The dose titration of the trial medication was complex, and some participants would have benefited from additional support, for example more information, ready-filled dosette boxes and extended discussions about medication with RAs and/or CSOs as well as the involvement of carers in such conversations.
The remote nature of the trial required multiple questionnaires to be completed at regular intervals by participants. Many found that breaking the task up over several days, rather than completing all those required at one sitting, was easier to manage.
Some HCPs indicated that the option of face-to-face site training and participant visits would add a personal touch. This may have improved understanding of the trial, connectivity to the trial and engagement from both research personnel and PAX-BD participants, all of which can have positive implications for recruitment and retention.
Trial processes
A key element to the conduct of the trial was for flexibility to allow sites to use local knowledge to optimise participant identification, approach and recruitment.
Collaboration between the central team and site staff was highlighted as an important facilitator to recruitment and trial conduct. Sites reported benefitting from the CI and CI delegate interacting with their staff and presenting to colleagues at local site meetings. Central RAs also expressed they would value more site involvement with trial participants in terms of encouraging engagement with safety monitoring and highlighting the importance of this. In future trials, introductions or handovers with all parties could be arranged to bolster the sense of collaboration and continuity in care.
Insights from central trial team
Remote design
Remote designs can alleviate burden from busy site teams and optimise funding using central RAs. They are likewise appealing to participants with busy schedules or have limited scope for in-person visits. However, RAs observed that for some participants, completing trial tasks in their own time seemed to feel like an encroachment on their personal lives. A hybrid design with the option of face-to-face appointments with research staff could offer a more optimal solution and help minimise participants lost to follow-up.
The use of technology as the only means of data collection is inadvertently exclusionary. In addition, the choice of a digital platform is also vital; issues with systems can be frustrating and demoralising for participants, as well as tarnishing trial credibility and stagnating data collection. Comprehensive, accessible and established procedures for offline options (such as paper questionnaires and staff-assisted completion) are also required to provide a contingency in case of technical failures, as well as expanding inclusivity of the trial.
Trial management
Adverse events for the trial were collected by the RAs during weekly assessments with participants and recorded centrally. It was discovered at the end of the trial that some disparities existed between the data collected by the central team and those collected by clinicians locally. Future trial design should build in ways of confirming local data collected in clinical consultations with those collected centrally to ensure the quality and completeness of data collected. Many sites needed support in completing routine research activities such as the upkeep of an Investigator Site File (ISF) and paper archiving. This would be facilitated by utilisation of e-ISFs and e-archiving. An annual formal review of local site arrangements for upkeep of the ISF provided, planned archiving arrangements and provision of staffing would facilitate adherence to research requirements.
Central pharmacy
A central pharmacy had a series of positives over dispensing at a local level. Feedback from sites was positive as it took any potential burden away from the local sites which was particularly pertinent during COVID-19. As such, it meant that only one pharmacy organisation needed to enact amendments, whereas multiple organisations would be required to do so if prescribing at a local level. Key feedback for the trial and for setting up any future research would be to involve the pharmacy from the very beginning of the trial in all areas to create a robust trial design and to reduce the need for amendments later on.
There were, however, some drawbacks to a central pharmacy, namely around storage of stock, risk assessments and the need for clear trial roles. This trial was fortunate, in that the storage requirements for IMP did not require temperature or other special environmental considerations which would have otherwise rendered the central storage impossible due to the quantity of IMP. Delivery of IMP was disrupted by the COVID-19 pandemic as well as postal strikes. The central trial RAs were essential in keeping open lines of communication with participants during the weekly calls to ensure that IMP had arrived as well as the empty bottles being posted back to the pharmacy. Future studies should consider contingencies for IMP delivery when undertaking their initial risk assessments.
Equality, diversity and inclusion
Unfortunately, data on participant’s ethnicity were not collected and so it is not possible to comment on the ethnic diversity of the population included. However, it would have been challenging to take part in the study for a participant who was not fluent in English. This may have limited the diversity of the study population.
Similarly, we did not collect data on neurodiversity, so are unable to comment on this.
In terms of age and gender, the mix of the study population was similar to the population with BD in general and those included in similar studies.
Data on the participants show that they were probably more highly educated than the general population. This may have been driven by the remote collection of outcome data online by participants. While there was not a requirement for participants to have a computer/tablet/smartphone and be computer literate, it was challenging for those who did have this to take part. This remains a challenge for remotely conducted studies.
Patient and public involvement
There was extensive PPI in PAX-BD, and this is described in detail in Chapter 6 and Appendix 6.