A multicomponent positive psychology intervention for euthymic patients with bipolar disorder to improve mental well‐being and personal recovery: A pragmatic randomized controlled trial

Mental well‐being and personal recovery are important treatment targets for patients with bipolar disorder (BD). The goal of this study was to evaluate the effectiveness of an 8‐week group multicomponent positive psychology intervention (PPI) for euthymic patients with BD as an adjunct to treatment as usual (TAU) compared to TAU alone.


| INTRODUC TI ON
Bipolar disorder (BD) is a severe and recurrent mood disorder associated with substantial impairments in quality of life, 1 strongly increased suicide rates 2 and high economic costs. 3The lifetime prevalence of bipolar disorder type 1 (BD-I) has been estimated at 0.6%, and 0.4% for bipolar disorder type 2 (BD-II). 4Current treatment for euthymic (i.e.relatively stable) patients with BD mainly focuses on achieving and maintaining clinical and functional recovery and consists of psycho-education, enhancement of selfmanagement skills and psychotherapy if indicated, combined with pharmacotherapy. 5sides clinical and functional recovery, it is considered increasingly important to focus on the improvement of personal recovery and mental well-being in patients with BD. [6][7][8] Recovery from BD goes beyond being asymptomatic 9 and clinical recovery alone may misrepresent treatment success from the perspective of people living with BD. 10 Mental well-being comprises the presence of positive emotional, psychological and social functioning. 11The closely related concept of personal recovery has been defined as containing five underlying processes, namely connectedness, hope, identity, meaning and empowerment. 12[15][16][17] One form of intervention that specifically aims to improve mental well-being and personal recovery is positive psychology interventions (PPIs).Multicomponent PPIs aim to enhance a wide range of positive resources for clients, including positive relationships, meaning or positive emotions using different positive psychology interventions. 18,19PPIs have been shown to be effective in clinical samples. 20So far, however, only a few small studies have examined the effect of interventions specifically aimed at mental well-being or personal recovery for people with BD. 21,22 These studies confirmed the overall feasibility and acceptability of positive psychology and showed promising effects, but were underpowered 23 or did not explicitly use PPIs. 21The goal of the current study is to examine the effectiveness of a multicomponent PPI group intervention for euthymic patients with BD added to treatment as usual (TAU) in an adequately powered randomized controlled trial (RCT) up to 12 months after baseline.

| Trial design and procedure
A multicentre pragmatic parallel-group randomized controlled trial was conducted in six mental healthcare centres, which were located in the western, northern, eastern and central regions of the Netherlands.The experimental condition received TAU and the group-based positive psychology intervention as an adjunct, while the control condition received TAU only.Recruitment took place between September 2018 and October 2020.Initial screening of participants was conducted by the therapists in the centres and consisted of an assessment of current symptoms using the Clinical Global Impression Scale -Bipolar version (CGI-BP).After obtaining written informed consent, participants were contacted for an additional screening interview by the first author (JK).After sufficient participants were included in a treatment centre, participants were randomized, informed about their allocation and asked to complete the online baseline measurement.Randomization was conducted by the first author (JK) using block randomization by centre prepared using the website www.seale denve lope.com.Randomization was stratified according to type of diagnosis (BD-I or BD-II).Links to the online questionnaires were sent via email and participants received reminders to increase the response rate.The study included four measurement points: baseline (T0), 4 weeks mid-treatment (T1), post-treatment (T2), 6-month follow-up after baseline (T3) and 12month follow-up (T4).

| Inclusion/exclusion criteria
Participants met the following inclusion criteria: (1) diagnosis of BD-I or BD-II made by a clinician, and additionally checked using the MINI-international neuropsychiatric interview during screening, (2) between the ages of 18 and 65, (3) four or more supportive sessions in the previous year and (4) presence of subsyndromal symptoms at inclusion.The fourth criterion was checked using the CGI-BP.Participants were excluded if they were (1) in a depressive or (hypo)manic episode or (2) in treatment for addiction problems at the time of inclusion.Although initially patients with already optimal mental well-being levels (defined as flourishing on the Mental Health Continuum-Short Form 25 ) were also excluded, this criterion was dropped during the trial to increase inclusion rates.

| Outcome measures
The primary outcome well-being was assessed using the 14-item Mental Health Continuum-Short Form (MHC-SF). 25Besides the total mental well-being score, the MHC-SF contains three subscales: emotional (3 items), social (5 items) and psychological well-being (6 items).Personal recovery was measured with the 15-item Questionnaire about the Process of Recovery (QPR) 26 ; Social role participation was measured with the 12-item Short-Social Role Participation Questionnaire (S-SRPQ) 27 containing the two subscales satisfaction with role performance (6 items) and experienced difficulties with role performance (6 items).Depressive symptoms were assessed using the 16-item Quick Inventory of Depressive Symptomatology (QIDS-SR). 28Anxiety symptoms were assessed using the 7-item anxiety subscale of the Hospital Anxiety and Depression Scale (HADS-A). 29Manic symptoms were assessed using the 5-item Altman Self-Rating Mania Scale (ASRM). 30Positive emotion regulation was assessed with the 17item Responses to Positive Affect Questionnaire (RPA). 31The RPA contains three subscales: (1) dampening (8 items), (2) emotionfocused positive rumination (5 items) and (3) self-focused positive rumination (4 items).Positive emotions were measured with the positive affect subscale of the Positive and Negative Affect Schedule (PANAS). 32Self-compassion was measured with the 12item Self-Compassion Scale-Short Form (SCS-SF). 33The experience of meaningful intrapersonal relationships was measured with the 9-item positive relations with others subscale of Ryff's Scales of Psychological Well-Being (PWBS). 34Internal consistency of all included scales was adequate to excellent (all Cronbach's α values >0.75).At 12-month follow-up, semi-structured telephone interviews using the life chart method were conducted by trained interviewers blinded to group allocation to retrospectively assess recurrence of depression and mania during the study period. 24,35riods of symptoms were graphically depicted over time from −1 to −3 for depressive symptoms and from +1 to +3 for manic symptoms, representing mild, moderate and severe symptoms, respectively.

| Positive psychology intervention
The intervention Living well with bipolar disorder, specifically developed for this trial, 24 consisted of eight sessions, was given in groups of 5-8 patients, and was facilitated by two trained therapists from the respective centres.To ensure a standardized delivery of the intervention, participating therapists received 1-day training workshop and a treatment manual. 24The workshop covered the central concepts of the intervention, including background, goals and exercises.The therapists were clinical psychologist, healthcare psychologist or nurse specialist.The intervention and manual were based on the well-being theory by Seligman 36 and Ryff's theory of psychological well-being. 37Patients also received the Dutch positive psychology book 'Dit is jouw leven' (English: This is your life), which contained information about positive psychology topics and exercises.The following topics were addressed in the intervention: (self)compassion, positive emotions, coping with fear of relapse, personal strengths, personal goals and optimism and positive relationships.Participants were also asked to complete homework exercises as preparation for each session.

| Treatment as usual
The comparison group received TAU according to the Dutch treatment guidelines for BD.TAU for people with BD in an euthymic phase in the Netherlands consists of supportive individual or group sessions (usually 2-12 sessions per year) with a psychologist or psychiatric nurse, maintenance pharmacotherapy, group psychoeducation and enhancement of self-management skills.If indicated, TAU is complemented with individual or group psychotherapy, including CBT and interpersonal approaches.

| Treatment integrity
Therapists who facilitated the intervention were asked to complete treatment logbooks after each session.For each meeting, it was noted whether participants were present and whether exercises could be carried out.If this was not the case, they were asked to list the reason(s) why.

| Statistical analyses
Statistical analyses were conducted following the Consolidated Standards of Reporting Trials (CONSORT) statement. 38A twosided significance level of α = 0.05 was used for all analyses.Data were processed and analysed using R, including the packages esc for the calculation of effect sizes and nlme for mixed-effects models.Based on the a priori sample size calculation, 24 we originally planned to include 112 participants to achieve 80% power for detecting at least a moderate effect on the primary outcome well-being at post-treatment (T2).Due to stagnating inclusions rates, primarily caused by Covid-19 pandemic regulations, inclusion was stopped after 97 patients had started in their respective condition and completed the intervention for those in the experimental group.Post hoc power analyses using the same assumptions indicated the study still had 75% power to avoid a type II error.Intention-to-treat (ITT) analyses were conducted using linear mixed models (LMMs) with restricted maximum-likelihood estimation and with group, time and group-by-time interactions to evaluate the effect of the intervention on primary and secondary outcomes.The covariance structure was set to compound symmetry since a more complex unstructured variance-covariance matrix did not show a significantly better fit for the primary outcome of mental well-being (p = 0.32).A random intercept model was used and baseline scores were included as covariate to adjust for any baseline imbalances.As the additional variance of the centre level was negligible (ICC = 0.01), only two-model levels are presented.Between-group effect sizes in estimated marginal means were expressed as Cohen's d with 95% confidence intervals.Effect sizes >0.80 were considered large, effects >0.50 as moderate and effect sizes >0.20 as small. 39Per-protocol analyses were conducted to examine whether similar results were found when including completers only in the analyses.Participants were considered as completers if they attended at least six of eight sessions.Post hoc analyses with participants who completed respective questionnaires were conducted to test for between-group differences for each measurement point using analyses of covariance (ANCOVAs) including baseline scores as covariate.To test the robustness of the results, we performed outlier checks on all outcomes and re-ran the analyses excluding the outlier values for those outcomes for which outliers were identified.Outliers were defined as scoring lower or higher than 3 standard deviations from the sample mean at the respective time point.Also, within-group dependent samples t-tests were conducted to determine whether possible effects at post-treatment were maintained until 6-and 12-month follow-up.Finally, the reliable change index (RCI) for the primary outcome mental well-being was calculated according to the Jacobson and Truax 40 method using Cronbach's α of the total sample at baseline as the measure of reliability.The RCI indicates whether changes in an outcome are beyond changes that would be expected due to measurement error.Individual change scores in the primary outcome of mental well-being were calculated as the difference between baseline and post-treatment and baseline and 6-month follow-up.Patients who improved by more than the RCI are considered to have achieved a reliable change.
Binary recurrence data from the 12-month retrospective life chart interviews were analysed with Kaplan-Meier survival analyses to compare recurrence-free survival time between the intervention and control groups.Cox proportional hazard models were used to determine the effect of group allocation on time to recurrencefree survival and to calculate hazard ratios.The following changes in symptoms were coded as recurrence: a change from no symptoms to moderate or severe symptoms, and a change from mild symptoms to moderate or severe symptoms.

| RE SULTS
A total of 120 patients were assessed for eligibility, of which 12 participants were excluded at screening and 108 were randomized.

| Baseline characteristics
Baseline characteristics of the sample are summarized in Table 1.
Participants' mean age was 47.4 (SD = 10.3) and the sample was predominantly female (71.1%).More than half of the sample had low or intermediate education (55.7%) and more than half of the sample was not working (52.6%).Of all participants, 53.6% were diagnosed with BD-I and 42.3% with BD-II.Half of the sample (50.6%) followed another psychological training in the context of their BD in the past 5 years.

| Treatment integrity
In total, nine intervention groups were carried out in six different treatment centres.From 15 possible exercises, four of the nine groups completed all exercises.The average number of exercises completed was 14.1, with 12 being the lowest number of exercises completed in one group.Reasons that exercises could not be carried out included that there was no sufficient time, that the exercise was not clear or that the exercise was too burdensome.

| Primary outcome
Effects on primary and secondary outcomes can be found in Table 2. Per-protocol analyses with completers only did not re-

| Secondary outcomes
Significant group-by-time interaction effects were also found for the separate MHC-SF subscales of social (p < 0.01) and psychological well-being (p < 0.01).Small-to-moderate effects for the two subscales were found at post-treatment, 6-month and 12-month follow-ups ranging from 0.38 to 0.77.The interaction effect for personal recovery was significant (p = 0.04), with a moderate effect at post-treatment (d = 0.76) and 12-month follow-up (d = 0.64).In addition, significant interaction effects were found for the RPA subscale emotion-focused positive rumination (p = 0.012) and PANAS positive affect scores (p = 0.049).All these effects were in favour of the intervention group.Per-protocol analyses with completers only did not reveal different results compared to ITT analyses for primary and secondary outcomes.
Between-group post hoc tests were all non-significant at midtreatment, but were significant for the MHC-SF subscale emotional well-being at post-treatment (p = 0.04) and 6-month follow-up (p = 0.04) and psychological well-being at post-treatment (p = 0.02) and for social well-being at post-treatment (p = 0.03) and 6-month follow-up (p < 0.01).QPR scores were also significantly higher in the intervention group at post-treatment (p < 0.01), but the difference fell short of significance at 6-and 12-month follow-up.A significant between-group effect was also found for emotion-focused positive rumination at post-treatment (p = 0.02).Well-being, personal recovery, as well as manic and depressive symptoms per group and over time are summarized in Figure 2.

| Effect maintenance
Within-group tests showed no significant difference in MHC-SF scores at post-treatment and 6-month follow-up and 12-month follow-up (p's > 0.20) for the intervention group.Similarly, no significant changes in the intervention group were found for the three subscales of the MHC-SF and QPR scores (p's > 0.10).

| Recurrence
Kaplan-Meier survival curves for time to first recurrence of depression and mania are shown in Figure 3. Visual inspection suggests that the intervention did not have an effect on recurrence of mania, but that participants in the intervention group were more likely to experience an earlier recurrence of depression.However, the effect of group obtained from the Cox proportional hazards model was not significant, neither for recurrence of depression (b = 0.39, p = 0.24) nor mania (b = −0.08,p = 0.86).For depression, the hazard ratio for the intervention group was 1.48 (95% CI = 0.77-2.84),and 0.92 (95% CI = 0.37-2.32)for mania.

| DISCUSS ION
This is the first full RCT that examined the effect of an 8-week PPI as an addition to TAU to improve mental well-being and personal recovery in euthymic patients with BD.Our findings indicate that a multicomponent group PPI as an adjunct to TAU was more effective in improving mental well-being and personal recovery compared   to TAU alone.Between-group effects on mental well-being were still significant at 6 months, and fell just short of significance at 12-month follow-up.Effects within the positive psychology group could be sustained until 12-month follow-up.Since the study was conducted in a naturalistic setting, the findings suggest that this PPI may be a valuable addition to psychological interventions in relatively euthymic patients with BD in a real-world outpatient clinical setting with a specific focus on improving mental well-being and personal recovery.

| Main findings
The finding that mental well-being was significantly improved in the positive psychology group is in line with previous meta-analyses suggesting that PPIs are also beneficial in clinical settings. 20However, studies specifically examining the effect of PPIs on BD still remain scarce.One recent pilot study examined the effect of a 4-week PPI in BD and found moderate-to-large between-group effects for outcomes of positive emotions, gratitude and optimism, but used a relatively small sample. 23The current findings, in a more rigorous RCT, showed that mental well-being can indeed be improved in this population.In comparison with patients receiving TAU, participants receiving the positive psychology group intervention had substantially higher levels of positive psychological and social functioning.
A significant moderate effect in favour of the PPI group was also found for personal recovery at post-treatment.This finding is in line with one previous RCT that found that recovery-focused cognitive behavioural therapy, compared with TAU, significantly improved personal recovery in BD. 21The effects on personal recovery between groups at 6-and 12-month follow-up, however, fell short of significance.Although the within-group effect on personal recovery within the PPI group could be sustained until 12-month follow-up, we also found a relatively strong improvement in personal recovery in the control group at 6-month follow-up.These findings suggest that PPIs may accelerate the process of personal recovery of patients with BD.This supports previous accounts from the recovery field, arguing for the potential of positive psychology to improve personal recovery. 41The current findings are relevant as patients with BD describe severe negative consequences in relation to personal aspects of life because of their disorder, including loss of control, identity, social relationships, hope and purpose, or the intentional avoidance of pleasure because of fear of relapse.They often describe their life as disrupted or 'derailed' by the disorder. 42Therefore, leading a meaningful and joyful life, despite the presence of the limitations caused by the illness, is a difficult but vital adaptive challenge. 43 terms of sample, baseline mean scores on the primary outcome of mental well-being were comparable to previous studies including participants with BD in the Netherlands, one RCT study (N = 144) 44 and a cross-sectional survey study (n = 107), 45 but considerably lower compared to the general population. 46In terms of gender and education, the sample in our study seemed to include somewhat more women and higher educated people compared with

TA B L E 2 (Continued)
people with BD in the Netherlands in general. 47It is therefore advisable to not fully generalize the findings of the current RCT to the overall population of people living with BD.
Although well-being significantly improved in the intervention group, in the interviews we did not find that the PPI protected from recurrence of mood episodes.In line with this, no significant effects between groups were found for the secondary outcomes such as depression and anxiety symptoms.An explanation may be that BD is a recurrent illness with relatively strong symptoms and substantial impairments even in euthymic phases.Therefore, a certain level of symptomatology is usually present most of the time, even in euthymic phases. 48Also, since the disorder has a strong biological basis, improved well-being might not equally decrease the risk for psychopathology in BD as has been shown in previous studies in the general population for example. 49The finding that mental wellbeing did improve while psychopathology and time to recurrence of depression or mania did not also supports the idea that mental wellbeing and psychopathology are related, but distinct dimensions of mental health 11 and stresses the importance to assess well-being in addition to psychopathology in clinical settings.
Interestingly, the between-group difference in (hypo)manic symptoms measured with the ASRM approached significance at The finding that social role participation did not significantly improve between groups could be explained by the relatively short study period and by the severe impact the course of the disorder might have on social functioning.Patients with BD often report severe disruptions in social functioning, including loss of work, early retirement or negative impact on relationships due to decreased capacities and missing purpose. 42Improving functioning in these societal roles may require additional interventions focusing on promoting societal participation.
Significant group-by-time interaction effects and moderateeffect sizes at post-treatment in favour of the intervention group were also found for the process measures of positive emotions and positive rumination.Positive emotions have been found to be related to improved functioning in BD. 50The finding that positive rumination was improved at post-treatment is relevant since this emotion regulation strategy has been found to be uniquely associated with personal recovery in BD, even after controlling for symptomatology and social role participation. 51However, research also suggests that extreme valuing of happiness is related to worsened illness course in BD 52 and that positive rumination might be associated with higher lifetime frequency of mania 53 and diagnoses of mania or hypomania. 54Although high levels of positive rumination might thus be associated with negative consequences, moderate levels are pivotal for recovery.Therefore, it remains important for patients and therapists to continuously balance these positive feelings.

| Limitations
First, the TAU-only control group used in the current study was not an active control group, meaning that it remains unclear whether the PPI might be more effective in improving well-being and personal recovery than an active comparator.The goal of this trial was to determine whether the positive psychology treatment is a (cost)-effective addition to TAU, therefore this control group was chosen.For future studies, it might be interesting to compare the effect of this positive psychology treatment on mental wellbeing and personal recovery as compared to, for example, cognitive behavioural therapy to determine the unique efficacy of PPIs.
Second, adherence to the intervention was not extensively assessed.Therefore, it also remains unclear how many participants completed homework exercises as prescribed, and per-protocol Figure 1.
veal different results compared to ITT analyses for the primary outcome.A significant group-by-time interaction effect was found for the primary outcome mental well-being (p < 0.01), indicating that mean scores on the MHC-SF were significantly different over time in the intervention group compared to TAU.Moderate effects were found at post-treatment (d = 0.77), 6-month followup (d = 0.72) and 12-month follow-up (d = 0.60) in favour of the PPI group.Between-group post hoc tests were not significant for mid-treatment (p = 0.38) but were significant for post-treatment (p < 0.01) and 6-month follow-up (p = 0.01).The betweengroup difference at 12-month follow-up fell short of significance (p = 0.08).
Outlier checks revealed three outliers for ASRM scores: one outlier at post-treatment (score of 12) and two outliers at 6month follow-up (scores of 11 and 17, respectively).One outlier was identified for the dampening subscale of the RPA at posttreatment (score of 31).Removing the two outlier values for the ASRM at 6-month follow-up resulted in the between-group post hoc test at 6-month follow-up approaching significance (F = 3.22, p = 0.08), with higher scores in the intervention group.Removing only the case that scored 17 on the ASRM at 6-month follow-up similarly resulted in the between-group post hoc test at 6-month follow-up being close to significant (F = 3.55, p = 0.06).Removal of any other outlier did not substantially influence the effects.

F I G U R E 1
Flow of participants.
Abbreviations: ASRM, Altman Self-Rating Mania Scale; CI, Confidence Interval; FU, Follow-up; HADS-A, Hospital Anxiety and Depression Scale-Anxiety Subscale; M, Mean; MHC-SF, Mental Health Continuum-Short Form; PANAS, Positive and Negative Affect Schedule; PP, Positive Psychology; PWBS, Psychological Well-being Scale; QIDS-SR, Quick Inventory of Depressive Symptomatology-Self-Report; QPR, Questionnaire About the Process of Recovery; RPA, Responses to Positive Affect Questionnaire; SCS-SF, Self-compassion Scale-Shot Form; SE, Standard Error; S-SRPQ, Short-Social Role Participation Questionnaire; TAU, Treatment as Usual.a Higher scores on the primary and all secondary outcomes should be interpreted as being indicative of higher levels of the corresponding construct or process.*p < 0.05, **p < 0.01.

F I G U R E 2
Well-being (MHC-SF), personal recovery (QPR), depression (QIDS-SR) and manic symptoms (ASRM) per time point and group.6-monthfollow-up after removal of one outlier, which suggests that PPIs might lead to increases in (hypo)manic symptoms.In the literature, however, a cut-off score of 6 or higher is used to indicate a high probability of (hypo)manic conditions,30 from which the mean of the intervention group in the current study is still far away.Also, the life chart interviews at 12-month follow-up did not show any indication that people in the intervention group experienced earlier or more frequent recurrence of mania.

F I G U R E 3
Kaplan-Meier survival curves of time to first depressive (Plot A) or manic (Plot B) recurrence over the study period of 12 months for TAU (dashed line) versus positive psychology (solid line).
Estimated marginal means per group and measurement point and effects on primary and secondary outcomes.
Abbreviations: PP, Positive Psychology; SD, Standard Deviation; TAU, Treatment as Usual.a One missing response.b This category includes participants with paid work and voluntary work, students and who are self-employed.TA B L E 1 Sample characteristics at baseline.TA B L E 2