Results

The analysis presented here is reported according to the CONSORT flow diagram in Figure 1 and is based on the SAP version v2.0 (20 October 2022, Report Supplementary Material 6) which was in turn based on protocol version v11.0 (6 June 2022). Any analyses that were not pre-specified in the SAP are denoted as ‘unplanned’ or ‘ad hoc’. The main clinical outcomes from the PAX-BD study have previously been published.84

FIGURE 1 Consolidated Standards of Reporting Trials diagram illustrating flow of participants through the PAX-BD Trial. Reproduced with permission from McAllister-Williams et al.84 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 figure includes minor additions and formatting changes to the original text. (Figure not reproduced here; see the original article.)

Recruitment

• Number of sites: 21 opened, 14 recruiting into the pre-randomisation and 12 into the randomisation stage.
• Date first site opened: 28 November 2019.
• Date first participant randomised: 31 January 2020.
• The trial did not recruit to target and was closed to recruitment as of 15 June 2022.
• Date last recruited into pre-randomisation: 28 April 2022.
• Date last participant randomised: 26 May 2022.
• Date of last participant follow up: 7 September 2022.
• Date of final data set download: 5 and 6 December 2022.

Participant flow – Consolidated Standards of Reporting Trials flow diagram

Recruitment and participant flow through the trial is reported in the CONSORT diagram (see Figure 1).

Pre-randomisation stage

A total of 102 patients were assessed for entry into the trial. Fifty per cent of these (51 participants) were found to be eligible and entered pre randomisation. Demographic (see Table 28), illness characteristics (see Table 29) and medication history (see Table 30) can be found in Appendix 3.

The mean age of participants was around 47 years old (range 26–68), with slightly more men than women included. Most were non-smokers. The population studied was more highly educated than the general population with around half having achieved either an undergraduate or postgraduate university level. The majority of recruited participants were bipolar type I (73%) with a mean age of onset of BD was 28.3 years and duration of the current episode of 27.6 months. For just over half (51%) this was the only episode of BD in the last year, while eight (16%) participants would meet the criteria for rapid cycling bipolar disorder of four or more episodes in the last 12 months.85

On entry to the trial, 42 (82%) were on a mood stabiliser, 29 (57%) an antipsychotic, 29 (57%) an antidepressant, 25 (49%) an anxiolytic/hypnotic, 10 (20%) another psychotropic and 33 (65%) on a non-psychotropic. The specific drugs and doses are shown in the Appendix 3 in Table 30. The population appeared to be one with clinically difficult-to-treat BD as evidenced by 10 (20%) being on 2 mood stabilisers and the number on antidepressants (29; 57%), including five on two and several being on high doses and/or drugs, such as venlafaxine and tricyclic antidepressants (TCAs), believed to be associated with a high risk of precipitating mania.10

The criteria needed to meet the trial definition of TRBD, in terms of non-response, intolerance, declined/not clinically indicated in relation to the four guideline-recommended treatments for BD (olanzapine, quetiapine, lamotrigine and lurasidone),3,10 are shown in Appendix 3, Table 32. As can be seen, the most common drugs to which there was a recorded non-response were lamotrigine (41%) of pre-randomisation sample and quetiapine (39%). The most reported as associated with intolerance or declined/not clinically indicated were olanzapine (20% and 51%, respectively) and quetiapine (31% and 51%). The least commonly reported drug in all categories was lurasidone, consistent with this drug not being included in NICE guidelines.3 The number of NICE and BAP recommended medications to which there was an inadequate response in the current episode is shown in Table 33 in Appendix 3. The majority had failed to respond to one or none (65%), and only a small minority had failed to respond to three or four medications (6%).

Twelve participants recruited into the pre-randomisation stage (23.5%) did not progress to the randomisation. The time spent in pre-randomisation for the 12 participants who did not, and the 39 who did, progress to the randomised phase of the trial are shown in Figure 2 (with summary statistics for the latter in Appendix 3, Table 34). For those not randomised, the mean (SD) and median durations were 16.9 (16.6) and 8.6 weeks, respectively (range 2.7–52.4), while that for the randomised participants were 9.9 (7.8) and 8 weeks (range 3.9–43.9). There was little difference between the time in pre-randomisation between those randomised to pramipexole and those randomised to placebo (see Table 34, Appendix 3).

FIGURE 2 Time in pre-randomisation stage for participants who did not progress to randomisation stage (n = 12), and those that did (n = 39). (Figure not reproduced here; see the original article.)

Reasons for lack of continuation to randomisation were varied. One participant tragically died during the pre-randomisation stage. One was not yet meeting eligibility criteria for randomisation at the point when the study had to be closed down. Three no longer met eligibility criteria for randomisation (two due to no longer meeting criteria for a current episode of depression and one due to a relapse in alcohol use). Three were not able to tolerate withdrawal from an antipsychotic and/or commencement of a non-antipsychotic mood stabiliser. Two withdrew because of COVID infections. Two withdrew for other/personal reasons.

In terms of demographics, there was a higher proportion of women in those who did not progress to randomisation (75%) compared with those who did (46%) and more only had a basic level of education (25% vs. 10%), although numbers in each group are small making comparisons potentially unreliable. However, in all other respects, the participants who did or did not progress to randomisation appeared very similar with regard to their demographics and illness characteristics. It is, however, noteworthy that a higher proportion of participants not randomised were on an antipsychotic (75% vs. 51%) but not on a (non-antipsychotic) mood stabiliser (58% vs. 5%). This is to be expected. Any participant not on a mood stabiliser at recruitment was required to commence one to be eligible for randomisation, with those that struggled to tolerate the introduction of such medication not progressing. Similarly, at the start of the trial, a requirement for randomisation was to not be taking an antipsychotic, so any participant recruited on one had to have this withdrawn, and if this led to clinical problems, the participant potentially had to be withdrawn from the study.

Trial analysis populations

Of the 39 participants who were randomised, 54% were male and 26% were smokers. The primary analysis for this trial was carried out on those participants in the ITT population who provided a 12-week score on the primary outcome measure: the QIDS-SR (n = 36). Three participants withdrew from the trial prior to the 12-week time point: two in the pramipexole (who withdrew at 3 and 6 weeks) and one in the placebo treatment arm (who withdrew at 6 weeks). As can be seen from Tables 28–30 in Appendix 3, there were few differences in participant demographics, illness characteristics or medication histories. Baseline scores in the various rating scales used in the trial were also similar between all those randomised and those who contributed data to the primary outcome analysis (see Table 31, Appendix 3).

One of the eligibility criteria for randomisation was a QIDS-SR score of 10 or more. All participants randomised were assessed as meeting this criterion. However, at the point of ‘baseline’, the point at which trial medication was started, three participants no longer met this criterion (two with scores of 9 and one with a score of 3). A review of these participants' scores before and after the eligibility check and baseline QIDS-SR assessment is shown in Table 35 in Appendix 3. All three had scores of < 10 before the randomisation eligibility check, between 10 and 31 days prior to starting medication, though two scored 10, 4–5 days after starting medication (both of these had a baseline QIDS-SR of 9). It is possible there was some ‘inflation’ of QIDS-SR scores to enable participants to enter randomisation. However, the variability in scores may simply reflect fluctuations in symptom severity or be within the measurement error of the scale. However, a ‘per protocol’ population (n = 33) was analysed, excluding participants whose baseline QIDS-SR score was < 10.

Randomisation and minimisation factors

Participant characteristics with regard to the minimisation factors are shown in Table 36, Appendix 3. Close balance over minimisation factors was not achieved due to the trial closing early, though most factors were similar between treatment arms.

Treatment at the point of randomisation

As per the eligibility criteria, all participants who were randomised were taking a mood stabiliser. The most commonly taken were lithium and valproate (see Appendix 3, Table 30). Two participants had a mood stabiliser started, one had the dose of their drug increased and one was switched from one mood stabiliser to another, prior to randomisation. However, the vast majority (90%) were on the same medication at the same dose as at the point of recruitment to pre randomisation.

At the point of randomisation, 18 (46%) participants were taking an antipsychotic (with 2 participants taking 2), with the details of these shown in Appendix 3, Table 30. Ten of the participants in the ITT population were recruited to the pre-randomisation stage, of whom three had been randomised, prior to the protocol amendment allowing for participants to continue on antipsychotics into the randomisation stage. Three of the randomised participants discontinued an antipsychotic during the pre-randomisation stage. Of those participants taking an antipsychotic at the point of randomisation, 11 (61%) remained on the same dose as at the point of recruitment. Five had their dose lowered to comply with the maximum dose allowed in the amended protocol, and one had their dose increased.

Fifty-six per cent of participants were taking an antidepressant at the point of randomisation (drugs and doses shown in Table 30, Appendix 3).

Study medication received

Of the 39 participants randomised to IMP, 18 received pramipexole and 21 placebo. The dose of medication achieved at the end of the titration phase and used for the fixed-dose period was 2.18 mg/day (SD: 0.58; range: 0.75–2.50) for the pramipexole arm and 2.25 mg/day (SD: 0.55; range: 0.50–2.50) for the placebo arm. By the end of the flexible-dose phase, the doses were 2.02 mg (SD: 0.64; range: 0.25–2.50) for pramipexole and 1.66 mg (SD: 0.55; range: 0.25–2.50) for placebo, indicating a degree of dose reduction during this phase of the trial, interestingly more so in the placebo arm. The maximum doses achieved by participants were similar in both treatment arms: 2.50 mg (range: 0.75–2.50) versus 2.50 mg (range: 2.25–2.50) for pramipexole and placebo, respectively. It is noteworthy that the lowest maximum dose for any participant on pramipexole was 0.75 mg, but 2.25 mg on placebo.

Note: all doses are based on salt, rather than base, weights.

Missed doses

Fifteen participants (38%) recorded in their diaries that they had missed at least one dose of trial medication, nine (50%) in the pramipexole group and six (29%) in the placebo group. Details of the number of missed doses recorded by these individuals are reported in Table 37 in Appendix 3.

Duration of trial follow up

The initial protocol detailed follow-up for 48 weeks post randomisation followed by up to a 4-week taper of trial medication. However, because of the early closure of the trial, a protocol amendment allowed for participants to be withdrawn from the trial once the last participant randomised had reached the 12-week primary outcome time point while maximising the number reaching 24-, 36- and 48-week time points when there was extra data collection (see Chapter 2 and Table 26).

The duration of follow-up for the participants in each of the treatment arms, defined as the time they started taking the trial medication to week 48 adjusted for withdrawal and early closure of the trial, is shown in Figure 3 (with summary statistics in Appendix 3, Table 38). As can be seen, the duration of follow-up was similar between groups.

FIGURE 3 Duration of follow-up by treatment arm. (Figure not reproduced here; see the original article.)

Outcome data quality and completeness

There was a high rate of completion by participants of the online weekly rated scales (78% for the QIDS-SR across both treatment arms) with full details provided in Appendix 3, Table 39. Return rates for other scales are reported in Appendix 3, Table 40, where it can be seen that completion rates were, if anything, even higher.

For the primary analysis, the QIDS-SR score completed during week 12 after the commencement of trial medication was used where available. Where this score was missing, the score closest in time to 12 weeks within the range of 10–14 weeks was used as pre-specified in the protocol and SAP. If data collected before and after were equidistant from the 12-week time point, then the later data within the compliance window were used. Note that in the ITT analysis, the nearest score to 12 weeks for one participant was recorded at 9 weeks and 3 days (i.e. 4 days outside of the 10–14-week compliance window). Summary statistics for the timing of the 12-week primary outcome measure can be found in Appendix 3, Table 41.

Descriptive analysis of the primary outcome measures

The distribution of QIDS-SR scores at baseline by treatment arm is shown in Appendix 3, Figure 26. Scores at baseline and the primary end point (12 weeks) are summarised in Table 1. Slight differences in summary measures between treatment arms are observed at baseline, probably reflecting the small sample size due to the early closure of the trial. The raw data show an apparent difference between the arms with lower scores (i.e. less depressive symptoms) in the pramipexole arm. Data for baseline and 12, 24, 36 and 48 weeks are shown as a box plot in Figure 4.

FIGURE 4 Box plots of QIDS-SR scores at baseline, 12, 24, 36 and 48 weeks (n = 36). Reproduced with permission from McAllister-Williams et al.84 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. (Figure not reproduced here; see the original article.)

The change in QIDS-SR scores from baseline to 12 weeks is presented in Table 2. As can be seen, the decrease in QIDS-SR score in the pramipexole arm was over twice as great as in the placebo arm (means of −4.4 and −2.1, respectively). The reduction of over four QIDS-SR points in the pramipexole arm would be judged as being a clinically meaningful reduction.65,66

Efficacy analysis – primary outcome

Univariable analysis

A two-tailed t-test was used to compare the two treatment arms. The null hypothesis was that there was no difference between means at the 12-week primary outcome time point. The two-sided t-test revealed p = 0.0303, with lower average QIDS-SR scores in the pramipexole compared to the placebo group (10.7 vs. 15.2), so the null hypothesis was rejected. This could have resulted from a baseline difference between groups. As a result, this univariate analysis was repeated for the change in QIDS-SR scores between the baseline and the 12-week primary end point. In this case, the two-sided t-test resulted in a p-value of 0.1689, so the null hypothesis of there being no difference between treatment arms is accepted.

Multivariable analysis

Analysis of covariance conducted according to the SAP demonstrated a non-statistically significant difference in QIDS-SR score at 12 weeks, adjusted for the effects of baseline QIDS-SR, 2.9 points higher for participants randomised to placebo [unadjusted mean (SD) for pramipexole: 10.7 (5.4) and placebo: 15.2 (6.4); p = 0.087, 95% confidence interval (CI) −0.4 to 6.3; d = −0.72]. The unadjusted results are presented in Table 1 and the results of the ANCOVA analysis are presented in Table 3.

Further to the above ANCOVA analysis on the primary outcome at 12 weeks, it was intended to consider further adjustment for the minimisation factors used during randomisation. However, due to the trial closing early and the reduced sample size, this work could not be fully completed.

The first stage of assessing the minimisation factors for inclusion, as specified in the SAP, was to consider each individually to assess their influence on QIDS-SR at 12 weeks while adjusting for baseline scores following the earlier analysis showing this to be significant. Any factors that appeared potentially important (with p < 0.1) would then have been considered for inclusion in the forward stepwise selection procedure to identify the most parsimonious final model. The results of this can be seen in Appendix 3, Table 42 which reports the results of the analysis of each variable, adjusting for baseline QIDS-SR, alone and when also including the treatment arm in the regression. None of the minimisation variables appeared to have a statistically significant influence on the difference in QIDS-SR score at 12 weeks between treatment arms, and the difference between treatment arms remained non-significant and generally close to that in the uncorrected ANCOVA as reported above (see Table 3).

A further analysis, similar to that conducted to assess each minimisation factor, was employed to assess the influence of GAD-7 baseline scores (see Table 42 in Appendix 3). Baseline GAD-7 score was found to be related to the primary outcome measure when adjusted for baseline QIDS-SR scores. However, when added to the above regression model, the difference between treatment arms remained not statistically significant. Note that at baseline, 12-, 24-, 36- and 48-week time points there was a statistically significant correlation between QIDS-SR and GAD-7 scores with correlation coefficients of between 0.4 and 0.7 (see Appendix 3, Table 43).

The next step undertaken was to assess the baseline demographic variables for relationships with the primary outcome and to consider their inclusion in the final parsimonious model. The procedure for this was identical to that used to assess the minimisation factors. The results of this assessment can be found in Table 44, Appendix 3. As with the minimisation factors, none of the baseline variables assessed impacted the primary outcome.

As described in the Chapter 2, sensitivity and additional analyses were performed.

Ad hoc end-point analyses

On the basis of observation of Figures 4 and 6, it appeared that the effects of pramipexole had not plateaued at week 12, perhaps as a result of the 4-week up-titration phase. As a result, ad hoc analyses of QIDS-SR scores at 24, 36 and 48 weeks were carried out, including those participants still in the trial at these times. Each of these time-points’ data were determined in the same manner as the primary end point with a ± 2-week compliance window with the exception of week 48 where this was specified in the protocol as −3/+ 1 weeks. The results can be found in Table 4. There are diminishing numbers of participants at each time point.

The primary QIDS-SR outcome at 12 weeks for participants who had a baseline QIDS-SR score of 10 + at the point of starting medication were almost identical to that of the ITT population (see Table 45 in Appendix 3).

Secondary analysis of depressive symptomatology

Observer rated depressive symptom severity

The MADRS and QIDS-C were completed via the phone by the central trial RAs and administered at baseline and week 12, with scores reported in Tables 46 and 47 in Appendix 3, with the MADRS data shown in Figure 5.

FIGURE 5 MADRS scores at baseline and week 12. (Figure not reproduced here; see the original article.)

The non-statistically significant difference in the change in scores from baseline to week 12 for MADRS was a 2.59-point larger reduction (95% CI −4.54 to 9.72; d = 0.65) for the pramipexole arm. Similarly, for QIDS-C the non-significant change in scores was 2.30 points (95% CI −1.72 to 6.32; d = 0.34) greater in the pramipexole arm.

Depression response

At 12 weeks, 4 of 16 in the pramipexole arm met response criteria compared to 3 of 20 in the placebo arm. Fisher’s exact test shows no significant difference in these rates. The response was also examined at the point participants exited the trial or began tapering down trial medication. This was planned to be at 48 weeks. However, the trial was closed early and, as a result, many participants exited early, with this ranging between 16 and 48 weeks. At this time, rates were 6 of 13 (46%) in the pramipexole arm compared to 1 of 16 (6%) in the placebo arm. Fisher’s exact test showed this to be a statistically significant difference (p = 0.026). Full details of response rates are provided in Table 48 in Appendix 3.

Remission of depressive symptoms

Remission was defined as a QIDS-SR score ≤ 5. There were 2 of 16 participants in remission in the pramipexole arm compared to 3 of 20 in the placebo arm. Fisher's exact test shows no significant difference between treatment arms. At exit from the trial, 4 of 13 (31%) were in remission in the pramipexole arm compared to 0 of 16 (0%) in the placebo arm. Fisher’s exact test showed a statistically significant difference (p = 0.03) in these proportions, in favour of pramipexole. Full details of the remission data are presented in Appendix 3, Table 49.

Weekly ratings of depression symptoms

During the trial, participants completed weekly QIDS-SR ratings (see Figure 6) allowing for an examination of the proportion of time participants were free of significant depressive symptoms (defined as scores of QIDS-SR ≤ 5). A table showing the proportion of weeks when participants were free of depression in each arm is reported in Table 50 (see Appendix 3). The average proportions of weeks free of depressive symptoms reported were approximately twice as large for the pramipexole treatment arm (15.4%) as for placebo (7.1%), though a t-test showed differences were not statistically significant.

FIGURE 6 Weekly QIDS-SR scores from randomisation to a maximum of 48 weeks (n = 36). Reproduced with permission from McAllister-Williams et al.84 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 figure includes minor additions and formatting changes to the original text. (Figure not reproduced here; see the original article.)

Overall summary of effect of pramipexole on depressive symptomatology

• No significant difference between pramipexole and placebo was detected for the primary outcome, although this may be a result of the small sample size. Pramipexole treatment was associated with a clinically meaningful reduction in QIDS-SR score from baseline to week 12, double the reduction seen with placebo. The effect size for the difference in reduction in depressive symptoms varied between 0.72, 0.65 and 0.34 for QIDS-SR, MADRS and QIDS-C, respectively.
• Similarly, participants in the pramipexole group had around twice as many weeks relatively free of symptoms during the trial as those in the placebo group, but this was not a statistically significant difference.
• Further analyses of depressive symptoms suggested that the 12-week primary outcome time point may have been too soon to assess the effectiveness of the treatment, with significant differences on baseline-adjusted ANCOVA of QIDS-SR scores at 36 weeks. There were also significantly greater numbers of participants meeting remission and response criteria at their exit from the trial.
• No difference between treatment arms was seen on the two observer rating scale scores (MADRS and QIDS-C) conducted at baseline and at week 12.
• The difference between treatment groups appeared largely unaffected by minimisation factors and baseline characteristics of participants.

Additional secondary outcomes related to efficacy

Generalised anxiety

The severity of generalised anxiety was assessed weekly using the GAD-7 self-report scale from baseline to week 48. Summary statistics of the data can be found in Table 51 in Appendix 3. Scores over time are plotted in Figure 27, Appendix 3, together with a box plot of the data at baseline, 12, 24, 36 and 48 weeks in Figure 7. As can be seen, there was no suggestion that pramipexole worsened anxiety.

FIGURE 7 Box plot of GAD-7 scores at baseline, 12, 24, 36 and 48 weeks (n = 36). Reproduced with permission from McAllister-Williams et al.84 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 figure includes minor additions and formatting changes to the original text. (Figure not reproduced here; see the original article.)

Data were analysed as done for the QIDS-SR data, at 12, 24, 36 and 48 weeks using an ANCOVA covarying for baseline GAD-7 score. This demonstrated a non-significant difference between treatment arms at 36 weeks (p = 0.0564; Table 5). Although not significant, this finding is in line with the QIDS-SR scores where a statistically significant benefit of pramipexole was seen at 36 weeks on ANCOVA analysis (see Table 3). As noted previously, and detailed in Table 43, Appendix 3, there were significant correlations between QIDS-SR and GAD-7 scores at all time points.

Feelings of pleasure

The SHAPS was completed at baseline, 6 and 12 weeks and is shown in Figure 8 with summary statistics reported in Table 52 (see Appendix 3).

FIGURE 8 Baseline and weeks 6 and 12 SHAPS scores (n = 36). Reproduced with permission from McAllister-Williams et al.84 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 figure includes minor additions and formatting changes to the original text. (Figure not reproduced here; see the original article.)

Analysis of the SHAPS scores found a non-significant difference between treatment arm at 6 weeks with pramipexole being associated with an increase of 2.04 (95% CI −0.11 to 4.20; d = −0.76) in experiences of pleasure (Table 6).

Psychosocial function

Potential effects of pramipexole on psychosocial function were assessed using the WSAS assessed at baseline, 6, 12, 24, 36 and 48 weeks, with the data shown in Figure 9 and summary statistics in Table 53 in Appendix 3.

FIGURE 9 Baseline and 6-, 12-, 24-, 36- and 48-week WSAS scores (n = 36). Reproduced with permission from McAllister-Williams et al.84 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 figure includes minor additions and formatting changes to the original text. (Figure not reproduced here; see the original article.)

Statistical comparison of the effect of the treatment arm on WSAS is shown in Table 7. Statistically significant advantage of pramipexole over placebo was seen at weeks 36 with 5.36 (95% CI 0.38 to 10.34; d = 0.80) and 48 with 9.63 (95% CI 0.25 to 19.02; d = −0.98) higher scores on ANCOVA adjusting for baseline scores.

Secondary outcomes regarding expected adverse effects of pramipexole

Pramipexole is known to be associated with an increased risk of manic symptoms and impulse control problems.86 These symptoms were systematically assessed.

Manic and hypomanic symptoms

Manic and hypomanic symptoms were assessed weekly using the self-reported ASRM. Weekly scores are shown in Figure 10.

FIGURE 10 Weekly ASRM scores from randomisation to maximum of week 48 (n = 36). Reproduced with permission from McAllister-Williams et al.84 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 figure includes minor additions and formatting changes to the original text. (Figure not reproduced here; see the original article.)

Being free of manic symptoms is defined as ASRM ≤ 5, with similar proportions of time seen in both treatment arms: pramipexole (88%) and placebo (96%; see Table 54 in Appendix 3), with a t-test showing no statistically significant difference.

The ASRM data were also analysed in line with that of other scales, assessing scores at 12, 24, 36 and 48 weeks, with this shown in Figure 28, Appendix 3 and summary statistics in Table 55 in Appendix 3. ANCOVA analysis of these data revealed statistically significant lower scores in the placebo group −1.22 (95% CI −2.42 to −0.02; d = 0.85) at week 12, adjusting for baseline score. However, there were no statistically significant differences between treatment arms at any other time point (Table 8).

Manic and hypomanic symptoms were also assessed using the observer-rated YMRS at baseline and week 12 by the trial RAs via phone. Summary statistics can be found in Table 56 in Appendix 3. ANCOVA revealed no significant differences between treatment arms.

Impulsive and compulsive behaviour

Impulsivity was assessed using the QUIP-RS at baseline and then weeks 6, 12 and 4-weekly thereafter to week 48. Scores are shown in Figure 11, with summary data in Appendix 3, Table 57. Data at 12, 24, 36 and 48 weeks were analysed as for other scales, which did not indicate any differences between treatment arms (Table 9).

FIGURE 11 Questionnaire for Impulsive–Compulsive Disorders in Parkinson’s Disease – Rating Scale box plots of baseline, 12, 24, 36 and 48 weeks (n = 36). Reproduced with permission from McAllister-Williams et al.84 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 figure includes minor additions and formatting changes to the original text. (Figure not reproduced here; see the original article.)

Impact of cotreatment with an antipsychotic

During the trial, a protocol amendment was approved allowing participants to remain on antipsychotics at randomisation. Post hoc analyses were conducted to assess the impact of this on both the efficacy of pramipexole in reducing depressive symptoms and the risk of inducing hypomania or mania. Sixteen participants from the pramipexole arm were included in the analysis, nine of whom were on an antipsychotic and seven who were not.

Weekly QIDS-SR scores are shown in Figure 12. This suggests that the effect of pramipexole on reducing QIDS-SR scores was similar in those participants who were, and who were not, taking antipsychotics. Due to the small numbers in each group (see Table 58, Appendix 3), limited statistical analysis was possible. Small coefficients for the difference between arms indicate no difference in the reduction in QIDS-SR scores between those who were on, and not on, antipsychotics. A box plot for baseline, 12-, 24-, 36- and 48-week QIDS-SR scores for participants on and not on antipsychotics in the pramipexole arm is shown in Figure 29 in Appendix 3.

FIGURE 12 Quick Inventory of Depressive Symptoms, Self-Rated scores weekly from randomisation to a maximum of 48 weeks. (Figure not reproduced here; see the original article.)

Weekly ASRM scores are shown in Figure 13, with summary statistics in Table 59, Appendix 3. This appeared to suggest higher ASRM scores in those participants who were not on an antipsychotic. Again, due to the small numbers involved, analysis was limited, but there was no apparent significant difference between treatment arms (Table 10). A box plot for baseline, 12-, 24-, 36- and 48-week ASRM scores for participants on and not on antipsychotics in the pramipexole arm is shown in Figure 30, Appendix 3.

FIGURE 13 Altman Self-rating Scale of Mania scores weekly from randomisation to a maximum of 48 weeks, for participants on and not on antipsychotics in the pramipexole arm. (Figure not reproduced here; see the original article.)

Overall safety and acceptability of treatment

Overall acceptability of treatment was assessed using the TSQM at weeks 6, 12 and then 4-weekly to week 48 (Figure 14 and summary data shown in Table 60, Appendix 3).

FIGURE 14 Treatment Satisfaction Questionnaire for Medication box plots for weeks 12, 24, 36 and 48 (n = 36). Reproduced with permission from McAllister-Williams et al.84 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 figure includes minor additions and formatting changes to the original text. (Figure not reproduced here; see the original article.)

Treatment Satisfaction Questionnaire for Medication scores were generally higher in the pramipexole arm (indicating better tolerability and acceptance of the treatment), though analysis of the data at 12, 24, 36 and 48 weeks (Table 11) found no difference between treatment arms. A figure of TSQM scores at weeks 6-, 12-, and then at 4-week intervals to week 48 is shown in Figure 31, Appendix 3. No adjustment for baseline and small sample size makes it difficult to establish the significance of observed differences.

Safety events

Adverse events

After removing entries that were errors or otherwise denoted as unintended entries, a total of 361 AEs remained. Of these, 71 were reported in pre-randomisation prior to any trial medication being taken, and 290 in the randomisation stage.

Adverse events in pre-randomisation stage

Thirty-five events (50%) were reported in five unique participants who did not enter the randomised phase of the trial. The remaining 36 (50%) were reported in 13 unique participants who were randomised, the severities of which are shown in Table 61 in Appendix 3. Line listings of all AEs reported during pre-randomisation stage are line listed in Table 62 in Appendix 3.

Adverse events in randomised phase of the trial

Two hundred and ninety AEs were reported by 36 (92%) of the 39 participants; 128 were reported by 17/18 unique participants in the pramipexole arm and 162 by 19/21 in the placebo arm. The severity, causality and status by the time of exit from the trial of reported AEs are shown in Table 12. Two hundred and sixty-five (91%) required no action to address the AEs, 22 (8%) had treatment interrupted/dose reduced or discontinued and 3 (1%) had IMP withdrawn. Listings of all AEs reported during randomisation are line listed in Table 63 in Appendix 3. The AEs that occurred during the randomisation are further summarised in Medical Dictionary for Regulatory Activities terminology (MedDRA^®, MedDRA, Herndon, USA) categories in Table 64 in Appendix 3. The most common categories of AEs were those related to the psychiatric, gastrointestinal (particularly nausea) and the nervous system. Overall, for those rated as being definitely related to the IMP, there were few differences between rates of mild symptoms between treatment arms across all categories of AEs. There were, however, more moderately severe AEs in the pramipexole arm related to psychiatric and gastrointestinal symptoms. These findings are as expected. The British National Formulary87 listed psychiatric and neurological symptoms (behaviour abnormal; confusion; dizziness; drowsiness; fatigue; hallucination; headache; movement disorders; psychiatric disorders; sleep disorders; vision disorders) and symptoms related to the gastrointestinal tract (appetite abnormal; constipation; nausea; vomiting; weight changes) as common or very common side effects of pramipexole.

Serious adverse events

There were six SAEs reported by six unique participants. One SAE (suicide by drug overdose) was reported in the pre-randomisation stage in January 2022. Of the remaining five, all were reported by different participants, two in the pramipexole arm and three in the placebo arm. Details of all SAEs occurring during the randomisation stage are presented in Table 65 in Appendix 3. Of those participants experiencing a SAE, three (one randomised to pramipexole and two to placebo) completed their 12-week primary outcome measures. Of the other two, one withdrew (from the pramipexole arm) 7 weeks after randomisation and the other (in the placebo arm) died after 8 weeks.

Of the two participants in the pramipexole arm who experienced a SAE, for one the event was judged to be unrelated to trial medication. The other participant with a SAE judged as related to study mediation experienced an episode of mania that led to hospitalisation.

Other safety events

For the period from the date of first randomisation to the last patient’s last visit, there were no reported SUSARs, SARs or Urgent Safety Measures. One suspected SUSAR was downgraded to an AE.

Safety events related to hypomania, mania and impulsivity

Given the known risk of hypomania and mania with treatment with pramipexole,86 which might be expected to be high in patients with bipolar disorder, and observations of ASRM scores (see Figure 10), all AEs, SAEs and withdrawals that related to this were examined, particularly in relation to whether or not the participant was taking an antipsychotic in addition to trial medication. Table 13 details this review.

It is interesting that there was one SAE and two withdrawals (one of whom was the same participant who experienced the SAE) related to manic symptoms in the pramipexole arm, but no similar situations in the placebo arm. In addition, AEs related to hypomanic/manic symptoms were more common for participants in the pramipexole arm [11 AEs reported by 8 participants (44%)] compared with the placebo arm [12 AEs reported by 6 participants (29%)].

It is noteworthy that of the 18 unique participants involved who experienced one or more events related to hypomania or mania, only 3 were taking an antipsychotic. Of these, two were on relatively low doses (risperidone 1 mg and quetiapine 100 mg).

There was less evidence of significant issues with regard to impulsivity when reviewing the routinely collected QUIP-RS scores (see Figure 11 and Table 9). However, there were 11 AEs related to impulse control behaviours reported by 6 unique participants in the pramipexole arm (i.e. 33% of participants in the arm) compared with and 8 from 4 unique participants in the placebo arm (19% of participants), hinting at a possibly higher rate with the active drug.

Summary of key findings

Efficacy of pramipexole

• Pramipexole was associated with clinically meaningful reduction in QIDS-SR score between baseline and 12 weeks, twice that seen in the placebo arm. It was also associated with twice the number of depression-free weeks during follow-up compared with placebo. However, neither of these changes were statistically significant possibly due to the small sample size. There were, however, statistically significant greater reductions in QIDS-SR scores at 36 weeks of follow-up, and higher rates of response and remission at exit from the trial, compared with placebo.
• While pramipexole can be activating, there was no increase in GAD-7 scores over the course of the trial. Indeed, there was a near significant reduction in scores at 36 weeks compared with placebo, in line with the QIDS-SR finding.
• There was an increased ability to experience pleasure at 6 weeks in the pramipexole compared with placebo group which was not statistically significant.
• There was a statistically significant benefit of pramipexole on work and social function at 36 and 48 weeks of follow-up.

Safety of tolerability of pramipexole

• Overall tolerability and acceptability of pramipexole were good, with this being rated (non-significantly) higher in the pramipexole arm.
• Rates of AEs were similar between treatment arms, with 91% not requiring any action and only one in the pramipexole arm being adjudged as severe. There were two SAEs in the pramipexole arm, one adjudged as unrelated to medication and the other an occurrence of a manic episode.
• Pramipexole treatment was associated with significantly higher rates of hypomanic symptoms at 12 weeks compared with placebo. This effect was not seen at later time points.
• There was no apparent difference in the rates of impulse control problems on structured testing between pramipexole and placebo treatment arms, but a higher proportion of participants in the pramipexole compared with the placebo arm (33% vs. 19%) experienced at least one AE related to impulse control problems.
• Comparing participants who were and who were not taking an antipsychotic in combination with trial medication, it appears that the reduction in depressive symptoms with pramipexole is not affected, but the severity of hypomanic symptoms may be reduced by coadministration with an antipsychotic, at least from observations of AE rates.
• Of 18 participants who withdrew from follow-up and/or experienced an AE or SAE related to mania, only 3 were taking an antipsychotic (compared with around half of the ITT population who were). Of these, two were taking doses that might be considered low.