Culturally adapted psychoeducation for bipolar disorder: a multicentre, assessor-masked, randomised controlled trial

Summary

Background

Adjunctive psychoeducation can reduce recurrence of mood episodes in bipolar disorder (BD), but evidence from low-resource settings is scarce. We evaluated whether culturally adapted psychoeducation (CaPE) plus treatment-as-usual (TAU) prolongs time to recurrence in adults with BD in remission in Pakistan, a lower-middle-income country.

Methods

In this multi-centre, assessor-masked randomised controlled trial, adults with remitted bipolar I or II disorder at seven urban centres in Pakistan were randomly assigned to CaPE plus TAU or TAU alone. CaPE comprised 12 weekly individual sessions delivered by master's-level psychologists. The primary outcome was time to recurrence of any mood episode over 48 weeks, assessed with the Longitudinal Interval Follow-up Evaluation. Secondary outcomes included recurrence of depressive and manic/hypomanic episodes and clinical and quality-of-life measures. The trial is registered with ClinicalTrials.gov (NCT05223959).

Findings

Between June 2022 and March 2025, 303 participants were randomly assigned (148 TAU, 155 CaPE plus TAU); 12-month data were available for 284 of 303 (93.7%). Over 48 weeks, CaPE prolonged time to recurrence of any mood episode versus TAU (adjusted hazard ratio 0.54, 95% CI 0.42–0.69; p < 0.001). At 12 months, CaPE improved recovery (measured with the Bipolar Recovery Questionnaire, on which higher scores indicate greater recovery; adjusted difference 144.8, 95% CI 51.5–238.1; p = 0.002), knowledge and attitudes (78.2, 22.6–133.8; p = 0.006), medication adherence (−0.4, −0.8 to −0.1; p = 0.013), and depressive symptoms (−5.3, −7.8 to −2.7; p < 0.0001). All adjusted estimates account for age, sex assigned at birth, marital status, weeks since diagnosis, and weeks in treatment.

Interpretation

Individual culturally adapted psychoeducation added to routine care reduced recurrence risk and improved clinical outcomes in Pakistani adults with BD, and could optimise maintenance treatment in Pakistan and other low-resource settings.

Funding

Canadian Institutes of Health Research.

Research in context.

Evidence before this study

In a component network meta-analysis of 39 randomised controlled trials (RCTs) of psychosocial interventions for BD, trials of psychoeducational interventions demonstrated reduced relapse risk in BD compared with treatment as usual. Notably, none of the included psychoeducation trials were conducted in low-resource settings and/or low and lower-middle income countries. A key benchmark in high-income settings was the PARADES RCT, a pragmatic multi-centre study in the UK that compared structured group psychoeducation with an active peer-support intervention in remitted BD. That study showed no overall superiority of psychoeducation versus active peer support, although psychoeducation was more acceptable and appeared to benefit participants earlier in illness course. A pilot study of individually delivered Culturally Adapted Psychoeducation (CaPE) suggested benefits for adherence, illness knowledge, quality of life, and relapse prevention in Pakistani adults with BD.

Added value of this study

This study extends findings of the preliminary CaPE trial in 303 adults with BD in remission from Pakistan. CaPE was individually delivered, culturally adapted, and designed for a setting in which routine psychosocial support for BD is limited. Compared with treatment as usual, CaPE reduced risk of recurrence of any mood episode over 48 weeks. The intervention also improved recovery, illness-related knowledge and attitudes, medication adherence, and depressive symptoms, with high follow-up completion across 12 months.

Implications of all the available evidence

Taken together with previous psychoeducation trials in BD, these findings suggest that psychoeducation can be effective when adapted to the cultural and service context in which it is delivered. In settings where stigma, language, and limited psychotherapy infrastructure make group or specialist models difficult to implement, individual culturally adapted psychoeducation could strengthen maintenance treatment for BD.

Introduction

Bipolar disorder (BD) is a chronic and highly recurrent mood disorder in which the prevention of recurrence, preservation of functioning, and recovery of quality of life are central goals of maintenance treatment.¹ Although pharmacotherapy remains the cornerstone of care, many patients continue to experience residual symptoms, repeated mood episodes, impaired social and occupational functioning, and high rates of all-cause mortality despite receiving medication.1, 2 Relapse prevention therefore usually requires more than pharmacological management alone. Adjunctive psychosocial interventions including psychoeducation can help patients understand their illness, recognise early warning signs, improve treatment adherence, regulate daily routines, and respond more effectively to stressors that might otherwise precipitate recurrence.2, 3 In practice, psychoeducation combines structured information about the illness with practical self-management strategies, including understanding symptoms, identifying triggers, maintaining medication adherence, recognising early signs of relapse, improving regularity of sleep and activity, and involving family or carers when appropriate. Systematic reviews and network meta-analysis suggest that psychoeducation can reduce recurrence risk and improve clinical outcomes, but much of this evidence comes from high-income settings and from interventions delivered in specialist services with infrastructure that is not readily available in low-resource environments.2, 3, 4 Evidence from low- and middle-income countries (LMICs) has remained limited to small single-centre trials with short follow-up.4, 5

The need for contextually appropriate psychosocial interventions is urgent in LMICs including Pakistan. In these settings, the treatment gap for mental disorders remains wide, access to specialist psychotherapy is limited, and stigma, family dynamics, explanatory models of illness, and language can all shape whether patients engage with continuing care.6, 7 In BD, these barriers are especially important because long-term outcomes depend on symptom reduction and sustained treatment engagement, accurate understanding of the illness, and timely action when early signs of relapse emerge. A psychoeducation programme that is culturally and linguistically adapted, delivered individually, and feasible for the existing resources might therefore have value in Pakistan.

In this context, we previously developed and piloted culturally adapted psychoeducation (CaPE) for 34 adults with remitted BD in Karachi, Pakistan.⁵ The pilot randomised controlled trial (RCT) suggested that CaPE was acceptable to the target population, a larger RCT was feasible, and preliminary clinical data supported improved treatment adherence, greater knowledge and more adaptive attitudes towards BD, lower mood symptoms, and better quality of life than treatment-as-usual (TAU).⁵ Here we report on a multi-centre RCT that tested whether those early signals would translate into a clinically important delay in recurrence in a larger, geographically broader sample recruited across Pakistan. We evaluated whether CaPE plus treatment as usual prolonged time to recurrence of any mood episode over 48 weeks compared with TAU alone.

Methods

Study design and participants

A multi-centre, assessor-masked, parallel-group RCT in psychiatric outpatient services in 7 urban centres in Pakistan: Karachi, Lahore, Multan, Rawalpindi, Peshawar, Hyderabad, and Quetta, Pakistan. The trial was designed as an individual-level superiority study comparing CaPE plus TAU with TAU alone, and it followed a previously published protocol.⁸ Recruitment was undertaken in teaching and non-teaching hospitals across the participating centres, with the aim of assembling a geographically diverse sample that reflected routine BD care in Pakistan.

Eligible participants were adults aged 18 years or older with Diagnostic and Statistical Manual of Mental Disorders, 5th Edition (DSM-5)⁹ bipolar I or bipolar II disorder confirmed by Structured Clinical Interview for DSM-5 (SCID-5),¹⁰ who were currently in remission and euthymic at study entry. Euthymia was defined as a 17-item Hamilton Depression Rating Scale (HDRS-17)¹¹ score of <8 and a Young Mania Rating Scale (YMRS)¹² score of <8. All eligible participants provided informed consent, were stable on a psychiatric medication regimen for at least 3 months, and lived within the trial catchment area to support ongoing follow-up. Exclusion criteria were active substance use disorder, current recurrence of mania, hypomania, depression, or mixed affective symptoms, active suicidal ideation, and unstable residential arrangements that would make continued participation unlikely. Potential participants were identified by treating psychiatrists in participating services and then approached by research staff, who provided study information, answered questions, and obtained written informed consent before confirming eligibility and completing baseline assessment. Active suicidal ideation was assessed clinically during the eligibility interview and, where present, was exclusionary; substance use disorders were identified using the SCID-5; and unstable residential arrangements were defined as the absence of a fixed address or a planned relocation outside the study catchment area during follow-up. Data on race and ethnicity were not collected. All participants were recruited from routine bipolar disorder care clinics within Pakistan, where these categories are not routinely recorded and the population within the catchment areas is relatively racially homogeneous; preferred language (Urdu or English) and recruitment site were recorded instead.

The protocol was approved by the National Bioethics Committee of Pakistan (Ref: No. 4-87/NBC-688/21/669) and the Centre for Addiction and Mental Health Research Ethics Board in Toronto, Canada (Ref: 138/2021). The trial was conducted in accordance with the principles of the Declaration of Helsinki. The trial was registered at ClinicalTrials.gov (NCT05223959) on January 25, 2022. Procedures for participant safety, referral for urgent psychiatric assessment, and oversight by trial steering and data monitoring structures were prespecified in the protocol.⁸

Community Engagement and Involvement (CEI) was integrated from trial development to implementation through a structured CEI group comprising individuals with lived experiences of bipolar disorder, carers and mental health professionals. CEI members reviewed participant-facing materials, assessment tools, and the intervention manual and provided recommendations on translation of patient material, language simplification, session burden, cultural relevance, family involvement, and local support resources, which informed refinement of trial procedures and materials. They contributed to identifying key facilitators and barriers to improve participants' engagement and retention. They also informed the development of lay summaries and community awareness activities to improve accessibility and acceptability of the trial. We also engaged with the wider community through quarterly awareness sessions and stakeholder events across sites, reaching more than 10,000 people, and by disseminating study-specific newsletters.

Randomisation and masking

After baseline assessment, participants were randomly assigned in a 1:1 ratio to CaPE plus TAU or TAU alone. Randomisation used a computer-generated permuted block sequence with variable block sizes, stratified by site, bipolar disorder type (I versus II), and self-reported sex-assigned at birth. Allocation occurred after completion of baseline assessment. Participants, therapists, and treating clinicians could not be masked to treatment assignment due to the nature of the intervention. Outcome assessors were masked throughout follow-up, were located separately from treatment providers where possible, and participants were asked not to reveal treatment allocation during assessments. To protect masking, outcome assessments were conducted by trained analysts who were not involved in delivering the intervention and, wherever site logistics allowed, were based separately from the therapists; assessors had no access to therapy records or session schedules, and participants were reminded at the start of each assessment not to disclose their group allocation.

Procedures

CaPE is a manualised, one-to-one psychoeducation intervention specifically adapted for Pakistani adults with BD. It consists of 12 weekly individual sessions of approximately 60 min, delivered by trained master's-level clinical psychologists.⁸ Each session combines a brief structured presentation with a practical exercise tailored to the session topic, such as constructing a life chart, listing relapse triggers, or planning actions to take when early warning signs of relapse occur. The intervention was adapted from the Barcelona Bipolar Disorders psychoeducation programme by the trial team in accordance with methods used in our previous work on cultural adaptation of psychosocial interventions in Pakistan and the UK.13, 14, 15, 16 The cultural adaptation was central to the design of CaPE and included: use of Urdu language, culturally familiar idioms rather than technical jargon, attention to local explanatory models of mental illness, use of locally meaningful examples and folk narratives, and incorporation of examples drawn from religious teachings when relevant and acceptable. The intervention was adapted primarily in the mode of delivery and therapeutic engagement rather than by changing its core clinical components. Family or carers could be involved, with the participant's agreement, particularly in sessions concerning medication interruption and the recognition and management of relapse. The choice of individual rather than group delivery reflected local experience suggesting that stigma, privacy concerns, and discomfort with public discussion of mental illness can substantially reduce engagement with group-based formats in Pakistan.

The 12-session structure covers core domains of BD self-management: concept and causes of illness; symptoms of mania, hypomania, depression, and mixed states; illness course and prognosis, including substance misuse; pharmacological treatment and common medication classes; alternative therapies; risks of interrupting treatment; recognition of early relapse signs; steps to take when relapse is detected; regularity of habits and routines; stress control; problem-solving strategies; and a final integrative session.⁸ All therapists received two months of training before delivering the intervention and weekly supervision from local co-investigators to support fidelity. The same therapist delivered all sessions to a given participant whenever possible to maintain continuity. Fidelity was assessed using a structured fidelity checklist capturing content and structural adherence to the intervention manual. For content adherence, a session-specific checklist was developed, capturing all core components (e.g., psychoeducation on BD, symptom domains, life charting, psychotropic medication, and stigma-related content). Each component was coded as present (1) or absent (0), and adherence scores were calculated as the proportion of components delivered. Structural adherence was assessed in terms of whether sessions were delivered in the sequence outlined in the manual and was coded as adherent or non-adherent at the session level. Two independent raters coded the randomly selected sessions. To ensure representation across therapists, recordings included all sessions delivered by each therapist for at least one participant yielding a total of 12 participants’ sessions for fidelity coding.

Participants assigned to TAU received routine psychiatric care as determined by their treating physicians. In Pakistan, routine care for BD generally consists of pharmacotherapy delivered through medical, psychiatric, and family medicine services, with very limited routine access to structured psychotherapy.⁸ Research staff recorded the nature and intensity of usual care received.

Outcome measures

The primary outcome was time (in weeks) from randomisation to recurrence of any mood episode over 48 weeks post-randomisation. Recurrence was assessed with the Longitudinal Interval Follow-up Evaluation (LIFE), a semi-structured interview designed to characterise the longitudinal course of psychiatric disorders and treatments over the preceding interval.¹⁷ The LIFE generates weekly psychopathology ratings aligned to DSM-5 syndromal definitions and is well suited to time-to-event analyses in BD. This study used standardised 4-week monthly intervals to retrospectively assess symptoms and functioning across the follow-up timepoints (3, 6, and 12 months). Secondary time-to-event analyses examined recurrence of major depressive episodes and manic or hypomanic episodes.

Secondary outcomes were measured at baseline and 3, 6, and 12 months. They included recovery with the Bipolar Recovery Questionnaire (BRQ), knowledge and attitudes with the Bipolar Knowledge and Attitudes Questionnaire (BKAQ), selected illness-perception domains from the Revised Illness Perception Questionnaire (IPQ-R), health-related quality of life measured by the EuroQoL EQ-5D-3L index score and EQ-5D visual analogue scale (VAS), medication adherence with the 4-item Morisky Medication Adherence Scale (MMAS-4), depressive symptoms with the HDRS-17, and manic symptoms with the YMRS.18, 19, 20, 21 All scales were translated and back-translated into Urdu, have been used previously in Pakistan, and delivered by trained blinded research analysts either in person or, when necessary, via secure videoconferencing.

Statistical analysis

We specified the trial objective using the estimand framework.²² The primary estimand targeted the treatment-policy effect of offering CaPE plus TAU versus TAU alone on time to recurrence of any mood episode over 48 weeks in all randomised participants (population), with the treatment-policy strategy applied to intercurrent events. Intercurrent events including discontinuation of CaPE sessions (n = 5), changes to pharmacological treatment (n = 17), and receipt of additional care during follow-up (n = 28), were handled by analysing participants as randomised and continuing to ascertain recurrence irrespective of these events; participants therefore contributed follow-up time regardless of any intercurrent event, consistent with the intention-to-treat principle. The summary measure was the hazard ratio for time to recurrence.

The primary endpoint of time to recurrence was analysed under the intention-to-treat principle with Cox proportional-hazards models, with treatment assignment as the main predictor and adjustment for a prespecified set of covariates comprising age, sex assigned at birth, marital status, weeks since BD diagnosis, and weeks in treatment at baseline. This covariate set was defined a priori in the statistical analysis plan (SAP), which was developed alongside the protocol and finalised before database lock and unblinding. Bipolar disorder type (I versus II) was examined in prespecified exploratory analyses rather than included as an adjustment covariate, because stratification variables do not require adjustment in the primary model. The survival-curve outputs report adjusted hazard ratios for recurrence of any mood episode, major depressive episode, and manic or hypomanic episode. The proportional-hazards assumption was assessed for each Cox model using the Schoenfeld residual test, evaluated both globally and for the treatment term. When this assumption was violated, we report, as prespecified, the adjusted hazard ratio and 95% CI from a Cox model with robust (sandwich) standard errors as the primary estimate; robust standard errors do not alter the hazard-ratio point estimate but provide valid variance estimation when the proportional-hazards assumption does not hold. For comparison and transparency, we also report the corresponding estimate from the standard Cox model (model-based standard errors) and the Schoenfeld test statistics in the Results section. The covariates were prespecified in the SAP as clinically relevant prognostic factors for recurrence in bipolar disorder (older age, male sex, being unmarried, and longer illness and treatment duration are associated with course and outcome). The randomisation stratification factors were accommodated as follows: site was included as a random effect in the mixed-effects models, sex assigned at birth was included as an adjustment covariate, and bipolar disorder type was examined in a prespecified exploratory analysis rather than forced into the primary model, because balance on stratification factors is ensured by design and their inclusion as fixed covariates is not required for valid inference.

Continuous secondary outcomes were analysed using linear mixed-effects models fitted by restricted maximum likelihood, with fixed effects for treatment, time (baseline, 3, 6, and 12 months), and the treatment-by-time interaction, and adjustment for the same covariates as the Cox model. Rather than modelling change from baseline, the baseline assessment was included as a level of the repeated time factor, so that the treatment-by-time interaction estimates the adjusted between-group difference at each follow-up while accounting for the baseline distribution of the outcome; this constrained longitudinal approach is appropriate for randomised trials and avoids conditioning on a baseline that is itself subject to sampling variation.²³ Random intercepts were included for site and for participants nested within site, with an unstructured covariance matrix. Treatment effects are reported as adjusted between-group differences with 95% confidence intervals. Missing data were addressed separately for outcome and covariate data. For the continuous secondary outcomes, follow-up assessments were missing for a small proportion of participants at each timepoint (5.0% at 3 months, 5.6% at 6 months, and 6.3% at 12 months; see Results); because linear mixed-effects models fitted by restricted maximum likelihood yield valid estimates under a missing-at-random assumption, all observed data at each timepoint were used without imputation, and no participant was excluded solely for missing a follow-up assessment. For the primary time-to-event outcome, participants without a recorded recurrence were censored at their last completed assessment, so that all randomised participants contributed follow-up time to the Cox models. There were no missing data on any of the prespecified baseline covariates, so all 303 randomised participants contributed to the adjusted Cox and linear mixed-effects models without need for complete-case exclusion or imputation of covariate data. In this specification the baseline assessment was treated as a repeated outcome measurement rather than as a covariate; because randomisation makes the expected baseline means equal between groups, the treatment effect at baseline is constrained towards zero and the treatment-by-time interaction provides the adjusted between-group differences at 3, 6, and 12 months. No single follow-up timepoint was prespecified as the primary timepoint for the secondary outcomes; these outcomes were exploratory and were evaluated across all timepoints through the treatment-by-time interaction and interpreted as a consistent pattern over follow-up rather than at any single privileged timepoint.

The target sample size was based on pilot-trial data.⁵ Assuming a conservative mean time to recurrence of 21.4 weeks in the TAU group, a sample of 240 (120 per group) provided 0.91 power to detect a hazard ratio of 1.60 (corresponding to a mean time to recurrence of about 34.2 weeks with CaPE) at a two-sided significance level of 0.05; the minimum detectable hazard ratio at 0.80 power was 1.49. To allow for up to 20% attrition, we aimed to recruit 300 participants.⁸ Two-sided p values below 0.05 were interpreted as statistically significant for the primary survival outcome. As prespecified in the SAP, secondary outcomes were regarded as exploratory and supportive, and no formal adjustment for multiplicity was applied.

Data management and trial oversight

Trial data were collected on paper forms and were entered and uploaded to a secure, access-controlled study database; all electronic data were stored on encrypted, password-protected systems. Trial conduct was overseen by an independently chaired Trial Steering Committee that included the principal investigator, an independent statistician, local co-investigators, and a patient representative; an independent statistician masked to allocation undertook periodic data-quality checks throughout the trial. Patient and public involvement was embedded throughout through a structured Community Engagement and Involvement group of people with lived experience of bipolar disorder, carers, and mental-health professionals, as described above. Adverse events were reported to the Centre for Addiction and Mental Health Research Ethics Board in accordance with the protocol.

Role of the funding source

The study was funded by the Canadian Institutes of Health Research (Grant Number: PJT-460586). The funding agency had no role in study design, data collection, or analysis.

Results

Between June 20, 2022, and March 31, 2025, 303 participants were randomly assigned: 148 to TAU and 155 to CaPE plus TAU (Fig. 1). For the repeated secondary outcome measures, post-baseline data were available for 288 of 303 participants (95.0%) at 3 months, 286 (94.4%) at 6 months, and 284 (93.7%) at 12 months. Month-12 outcome data were available for 143 of 148 participants in the TAU group and 141 of 155 in the CaPE + TAU group.

Fig. 1.

CONSORT flow diagram. Flow of participants through the trial. Of 1228 adults assessed for eligibility, 925 were excluded (age <18 years, n = 105; active hypomanic or manic episode, n = 79; active depressive episode, n = 195; unstable residence, n = 241; active substance use disorder, n = 44; active suicidal ideation, n = 55; not available for the entire study, n = 193; declined to participate, n = 6; unable to complete baseline assessment, n = 7), leaving 303 who were randomly assigned (1:1) to CaPE plus TAU (n = 155) or TAU alone (n = 148). In the CaPE plus TAU group, 151 received the allocated intervention and 4 did not; in the TAU group, all 148 received usual care. The primary time-to-event analysis followed the intention-to-treat principle and included all 303 randomised participants (155 and 148), with participants who did not have a recorded recurrence censored at their last completed assessment; 141 (CaPE plus TAU) and 143 (TAU) completed the month-12 assessment.

Baseline characteristics were broadly balanced between groups (Table 1). The mean age of the total sample was 36.1 years (SD 11.6), and 94 of 303 participants (31.0%) were female. More than half of participants were aged 25–44 years (177 of 303; 58.4%), 186 of 303 (61.4%) were married, and 149 of 303 (49.2%) were employed. Educational attainment was heterogeneous: the median years of education was 8.0 (IQR 5.0–12.0), 63 of 303 participants (20.8%) had no formal education, and 37 of 303 (12.2%) had a graduate degree. The average time since BD diagnosis was about 336 weeks and the average time in treatment at baseline was about 324 weeks. Bipolar I disorder was the diagnosis for 185 of 303 participants (61.1%) and bipolar II disorder for 118 of 303 (38.9%). Baseline psychotropic treatment reflected routine practice in Pakistani outpatient care: 186 of 303 participants (61.4%) were receiving antipsychotics, 64 of 303 (21.1%) antidepressants, 47 of 303 (15.5%) mood stabilisers, and 34 of 303 (11.2%) hypnotic or anxiolytic medication, noting that medication categories were not mutually exclusive.

Table 1.

Baseline characteristics of the 303 randomised participants, by treatment group.

Characteristic	Treatment as usual (n = 148)	CaPE plus treatment as usual (n = 155)	Total (N = 303)
Age (years), mean (SD)	36.8 (11.6)	35.4 (11.6)	36.1 (11.6)
Age group, n (%)
18–24 years	23 (15.5%)	30 (19.4%)	53 (17.5%)
25–44 years	88 (59.5%)	89 (57.4%)	177 (58.4%)
45–59 years	31 (20.9%)	32 (20.6%)	63 (20.8%)
60–74 years	6 (4.1%)	4 (2.6%)	10 (3.3%)
Sex, n (%)
Female	43 (29.1%)	51 (32.9%)	94 (31.0%)
Male	105 (70.9%)	104 (67.1%)	209 (69.0%)
Education (years), median (IQR)	8.0 (2.8–10.0)	10.0 (5.0–12.0)	8.0 (5.0–12.0)
Highest level of education, n (%)
No formal education	36 (24.3%)	27 (17.4%)	63 (20.8%)
Up to primary	21 (14.2%)	20 (12.9%)	41 (13.5%)
Middle	21 (14.2%)	28 (18.1%)	49 (16.2%)
Matriculation	37 (25.0%)	33 (21.3%)	70 (23.1%)
Intermediate	14 (9.5%)	29 (18.7%)	43 (14.2%)
Graduate	19 (12.8%)	18 (11.6%)	37 (12.2%)
Marital status, n (%)
Single	44 (29.7%)	54 (34.8%)	98 (32.3%)
Married	97 (65.5%)	89 (57.4%)	186 (61.4%)
Divorced or widowed	7 (4.8%)	12 (7.8%)	19 (6.3%)
Family system, n (%)
Nuclear	82 (55.4%)	77 (49.7%)	159 (52.5%)
Joint	66 (44.6%)	78 (50.3%)	144 (47.5%)
Number of household members, median (IQR)	6.0 (5.0–8.0)	7.0 (5.0–10.0)	6.0 (5.0–9.0)
Monthly household income (PKR), mean (SD)	45 760 (31 520)	51 635 (35 815)	48 766 (33 858)
Occupation, n (%)
Employed	81 (54.7%)	68 (43.8%)	149 (49.2%)
Homemaker	40 (27.0%)	44 (28.4%)	84 (27.7%)
Unemployed	27 (18.3%)	43 (27.8%)	70 (23.1%)
Bipolar disorder diagnosis, n (%)
Bipolar I	96 (64.9%)	89 (57.4%)	185 (61.1%)
Bipolar II	52 (35.1%)	66 (42.6%)	118 (38.9%)
Time since bipolar diagnosis (weeks), mean (SD)	336 (299)	336 (319)	336 (309)
Time in treatment (weeks), mean (SD)	324 (291)	324 (314)	324 (302)
Previous psychiatric admissions, median (IQR)	1.0 (0.0–2.0)	1.0 (0.0–2.0)	1.0 (0.0–2.0)
Baseline psychotropic medication, n (%)
Antipsychotic	92 (62.2%)	94 (60.6%)	186 (61.4%)
Antidepressant	28 (18.9%)	36 (23.2%)	64 (21.1%)
Mood stabiliser	20 (13.5%)	27 (17.4%)	47 (15.5%)
Hypnotic or anxiolytic	13 (8.8%)	21 (13.5%)	34 (11.2%)

Data are mean (SD), median [IQR], or n (%); percentages are calculated using the column denominator (treatment as usual, n = 148; CaPE plus treatment as usual, n = 155; total, N = 303). Medication categories were not mutually exclusive. In line with CONSORT guidance, no significance tests of differences between the randomised groups at baseline are presented.

The primary Cox model included all 303 randomised participants, with those who did not have a recorded recurrence censored at their last completed assessment. CaPE was associated with a significant increase in the time to recurrence of any mood episode over 48 weeks compared with TAU (Fig. 2). A recurrence was recorded in 127 of 155 participants (81.9%) in the CaPE plus TAU arm and 138 of 148 (93.2%) in the TAU arm. The Schoenfeld residual test indicated a departure from the proportional-hazards assumption (global χ² = 17.8, p = 0.01; treatment term: χ² = 13.6, p = <0.001); as prespecified, we therefore report the adjusted hazard ratio from a Cox model with robust standard errors as the primary estimate, which was 0.54 (95% CI 0.42–0.69; p < 0.001), corresponding to an estimated 46% lower risk of recurrence relative to TAU. For comparison, the standard Cox model (model-based standard errors) gave an identical point estimate (HR 0.54, 95% CI [0.42–0.70]; p [<0.001]), indicating that the treatment-effect estimate was not sensitive to the choice of variance estimator, and supporting our use of robust standard errors to obtain valid inference given the non-proportional hazards. Visual inspection of the adjusted survival curves suggested early separation between groups within the first 12 weeks after randomisation, with sustained separation throughout follow-up. Regarding BD subtype, bipolar II disorder was associated with a significantly lower risk of recurrence of any mood episode (HR 0.62, 95% CI 0.48–0.79; p < 0.001) than bipolar I disorder (Supplementary Figure S1).

Fig. 2.

Adjusted survival curves for time to recurrence of any mood episode (primary outcome). Adjusted survival curves showing the probability of remaining free of any mood episode over 48 weeks, by treatment group, with the number of participants at risk shown below the plot. The hazard ratio was estimated from a Cox proportional-hazards model adjusted for age, sex assigned at birth, marital status, weeks since bipolar disorder diagnosis, and weeks in treatment at baseline. Because the proportional-hazards assumption was violated (Schoenfeld residual test), the hazard ratio and 95% CI were estimated with robust (sandwich) standard errors: adjusted HR 0.54 (95% CI 0.42–0.69; p < 0.001), corresponding to a 46% lower risk of recurrence with CaPE plus TAU than with TAU. All 303 randomised participants were included, with those without a recorded recurrence censored at their last completed assessment.

Secondary time-to-event analyses were consistent with the primary outcome. CaPE reduced the risk of major depressive episode recurrence (adjusted HR 0.58, 95% CI 0.45–0.75; p < 0.0001; Supplementary Figure S2) and the risk of manic or hypomanic episode recurrence (adjusted HR 0.65, 95% CI 0.50–0.85; p = 0.002; Supplementary Figure S3).

Effects on continuous secondary outcomes were also generally favourable to CaPE (Table 2). Recovery, assessed with the BRQ, was higher in the intervention group at every post-baseline assessment. The adjusted between-group difference was 161.8 points (95% CI 69.1–254.5; p < 0.0001) at 3 months, 179.6 points (86.6–272.5; p < 0.0001) at 6 months, and 144.8 points (51.5–238.1; p = 0.002) at 12 months. Knowledge and attitudes towards bipolar disorder, measured with the BKAQ, also improved consistently with CaPE, with adjusted differences of 71.7 points (16.5–126.9; p = 0.011) at 3 months, 81.2 points (25.8–136.6; p = 0.004) at 6 months, and 78.2 points (22.6–133.8; p = 0.006) at 12 months.

Table 2.

Adjusted mean scores by treatment group and time for selected secondary outcomes.

Outcome and timepoint	Treatment as usual		CaPE plus treatment as usual		Adjusted between-group difference (95% CI)	p value
	n	Mean (SD)	n	Mean (SD)
Bipolar Recovery Questionnaire (BRQ) total score
Baseline	148	2030.9 (328.1)	155	2016.1 (351.7)
3 months	143	2087.9 (434.2)	145	2246.6 (322.4)	161.8 (69.1–254.5)	<0.0001
6 months	145	2092.8 (471.0)	141	2269.1 (341.7)	179.6 (86.6–272.5)	<0.0001
12 months	143	2163.8 (500.6)	141	2308.2 (420.1)	144.8 (51.5–238.1)	0.002
Bipolar Knowledge and Attitudes Questionnaire (BKAQ) total score
Baseline	148	1263.5 (290.7)	155	1293.7 (268.4)
3 months	143	1275.9 (211.7)	145	1347.6 (220.5)	71.7 (16.5–126.9)	0.011
6 months	145	1283.2 (221.1)	141	1364.6 (218.7)	81.2 (25.8–136.6)	0.004
12 months	143	1306.8 (239.1)	141	1385.1 (220.4)	78.2 (22.6–133.8)	0.006
IPQ-R: timeline
Baseline	148	18.4 (6.3)	155	18.5 (6.4)
3 months	143	19.0 (4.8)	145	20.0 (2.7)	1.0 (−0.2 to 2.3)	0.10
6 months	145	19.2 (5.8)	141	20.1 (4.6)	0.9 (−0.3 to 2.2)	0.14
12 months	143	19.2 (6.6)	141	20.1 (4.5)	0.9 (−0.4 to 2.1)	0.17
IPQ-R: consequences
Baseline	148	19.7 (6.1)	155	20.0 (5.8)
3 months	143	20.9 (4.1)	145	21.8 (3.4)	0.9 (−0.0 to 1.8)	0.063
6 months	145	21.2 (3.3)	141	22.0 (3.3)	0.8 (−0.2 to 1.7)	0.12
12 months	143	21.6 (2.2)	141	22.0 (2.5)	0.4 (−0.6 to 1.3)	0.43
IPQ-R: personal control
Baseline	148	20.9 (5.1)	155	21.2 (4.8)
3 months	143	21.7 (4.8)	145	22.1 (3.7)	0.4 (−0.7 to 1.5)	0.47
6 months	145	22.5 (4.9)	141	23.0 (4.4)	0.5 (−0.6 to 1.6)	0.35
12 months	143	23.1 (5.4)	141	24.1 (4.6)	1.1 (−0.0 to 2.2)	0.052
IPQ-R: treatment control
Baseline	148	18.6 (3.6)	155	18.6 (3.6)
3 months	143	19.2 (2.9)	145	19.4 (2.7)	0.2 (−0.6 to 1.0)	0.63
6 months	145	19.3 (3.9)	141	19.9 (3.1)	0.5 (−0.3 to 1.3)	0.19
12 months	143	19.5 (4.2)	141	20.2 (3.3)	0.7 (−0.1 to 1.5)	0.074
IPQ-R: illness coherence
Baseline	148	14.1 (3.8)	155	14.3 (4.0)
3 months	143	14.8 (3.3)	145	15.6 (2.9)	0.9 (0.1–1.7)	0.038
6 months	145	15.1 (3.7)	141	16.1 (3.6)	1.0 (0.2–1.8)	0.017
12 months	143	15.4 (3.4)	141	16.5 (3.7)	1.1 (0.3–1.9)	0.010
IPQ-R: cyclical timeline
Baseline	148	15.0 (2.4)	155	14.9 (2.5)
3 months	143	15.1 (2.1)	145	15.1 (1.9)	0.0 (−0.5 to 0.6)	0.89
6 months	145	15.1 (2.6)	141	15.4 (2.1)	0.3 (−0.3 to 0.9)	0.29
12 months	143	15.0 (3.0)	141	15.3 (2.4)	0.3 (−0.3 to 0.9)	0.31
IPQ-R: emotional representation
Baseline	148	20.9 (5.0)	155	20.4 (5.1)
3 months	143	21.1 (4.2)	145	21.5 (4.2)	0.4 (−0.7 to 1.5)	0.49
6 months	145	21.4 (5.2)	141	21.7 (4.6)	0.3 (−0.8 to 1.4)	0.55
12 months	143	21.8 (4.9)	141	22.0 (4.6)	0.2 (−0.9 to 1.3)	0.73
EQ-5D index score
Baseline	148	0.8 (0.2)	155	0.8 (0.2)
3 months	143	0.8 (0.2)	145	0.9 (0.2)	0.1 (0.0–0.1)	0.004
6 months	145	0.9 (0.2)	141	0.9 (0.1)	0.1 (0.0–0.1)	0.021
12 months	143	0.9 (0.2)	141	0.9 (0.1)	0.0 (−0.0 to 0.1)	0.088
EQ-5D visual analogue scale
Baseline	148	62.3 (12.3)	155	66.7 (15.9)
3 months	143	67.8 (13.9)	145	72.5 (13.0)	4.7 (1.6–7.8)	0.003
6 months	145	68.8 (13.1)	141	73.5 (14.3)	4.7 (1.6–7.8)	0.003
12 months	143	70.8 (12.8)	141	75.4 (11.7)	4.6 (1.4–7.7)	0.004
Morisky Medication Adherence Scale (MMAS-4) score
Baseline	148	2.7 (1.4)	155	2.6 (1.4)	−0.0 (−0.3 to 0.3)
3 months	143	2.5 (1.5)	145	2.1 (1.4)	−0.4 (−0.8 to −0.1)	0.010
6 months	145	2.5 (1.7)	141	2.0 (1.4)	−0.4 (−0.8 to −0.1)	0.011
12 months	143	2.4 (1.6)	141	1.9 (1.5)	−0.4 (−0.8 to −0.1)	0.013
Hamilton Depression Rating Scale (HDRS-17) score
Baseline	148	4.2 (1.9)	155	4.3 (2.0)
3 months	143	15.9 (10.3)	145	11.1 (10.1)	−5.0 (−7.5 to −2.5)	<0.0001
6 months	145	16.8 (10.9)	141	11.1 (9.8)	−5.8 (−8.4 to −3.3)	<0.0001
12 months	143	17.8 (12.2)	141	12.7 (11.7)	−5.3 (−7.8 to −2.7)	<0.0001
Young Mania Rating Scale (YMRS) score
Baseline	148	2.5 (1.9)	155	2.6 (1.9)
3 months	143	9.2 (7.0)	145	6.8 (6.3)	−2.4 (−4.2 to −0.5)	0.011
6 months	145	11.0 (9.6)	141	7.8 (8.0)	−3.2 (−5.0 to −1.3)	<0.0001
12 months	143	11.0 (7.9)	141	9.3 (8.2)	−1.8 (−3.6 to 0.1)	0.062

Values in the treatment-group columns are observed means (SD) at each timepoint. Adjusted between-group differences (CaPE plus TAU minus TAU) and p values are derived from linear mixed-effects models with fixed effects for treatment, time, and the treatment-by-time interaction, adjusted for age, sex assigned at birth, marital status, weeks since bipolar disorder diagnosis, and weeks in treatment at baseline, with random intercepts for site and for participants nested within site. For the BRQ, BKAQ, IPQ-R subscales, EQ-5D index, and EQ-5D VAS, higher scores indicate better outcomes; for the MMAS-4, HDRS-17, and YMRS, lower scores indicate better outcomes, so a negative between-group difference favours CaPE. n at each timepoint is given in the table. The mixed-effects models used restricted maximum likelihood under a missing-at-random assumption, with no imputation.

Among illness-perception domains, illness coherence was the most responsive to intervention. The adjusted advantage for CaPE on the IPQ-R illness-coherence subscale was 0.9 points (95% CI 0.1–1.7; p = 0.038) at 3 months, 1.0 point (0.2–1.8; p = 0.017) at 6 months, and 1.1 points (0.3–1.9; p = 0.010) at 12 months. By contrast, other IPQ-R domains showed little clear evidence of sustained between-group differences. Treatment-control beliefs, personal control, timeline beliefs, cyclical timeline beliefs, perceived consequences, and emotional representation were numerically favourable to CaPE at some assessments but were not consistently different at month 12.

Health-related quality of life showed a mixed but generally positive pattern. EQ-5D-3L index scores were significantly higher in the CaPE group at 3 months (difference 0.1, 95% CI 0.0–0.1; p = 0.004) and 6 months (0.1, 0.0–0.1; p = 0.021), but the month-12 difference was not statistically significant (0.0, −0.0 to 0.1; p = 0.088). EQ-5D VAS scores were consistently higher in the intervention group throughout follow-up, with adjusted differences of 4.7 points (1.6–7.8; p = 0.003) at 3 months, 4.7 points (1.6–7.8; p = 0.003) at 6 months, and 4.6 points (1.4–7.7; p = 0.004) at 12 months.

Medication adherence improved with CaPE, as reflected by lower MMAS-4 scores. The adjusted difference was −0.4 points (95% CI −0.8 to −0.1; p = 0.010) at 3 months, −0.4 (−0.8 to −0.1; p = 0.011) at 6 months, and −0.4 (−0.8 to −0.1; p = 0.013) at 12 months. Depressive symptoms also showed robust improvement. Compared with TAU, CaPE was associated with lower HDRS scores at all post-baseline assessments: −5.0 points (95% CI −7.5 to −2.5; p < 0.0001) at 3 months, −5.8 (−8.4 to −3.3; p < 0.0001) at 6 months, and −5.3 (−7.8 to −2.7; p < 0.0001) at 12 months.

Manic symptoms measured with the YMRS improved initially but the difference narrowed over time. The adjusted mean difference was −2.4 points (95% CI −4.2 to −0.5; p = 0.011) at 3 months and −3.2 (−5.0 to −1.3; p < 0.0001) at 6 months. By 12 months, the adjusted difference was −1.8 points (−3.6 to 0.1; p = 0.062), which did not meet thresholds for statistical significance. Overall, the pattern across continuous outcomes suggests that CaPE had its strongest and most consistent effects on recovery, knowledge and attitudes, medication adherence, and depressive symptoms, with more modest effects on broader illness-perception domains and manic symptom severity by the end of follow-up.

Therapist fidelity to the content of the intervention manual was 88%, with strong inter-rater reliability (κ = 0.85). Structure fidelity was 84%, with substantial inter-rater reliability (κ = 0.79).

Six adverse events were recorded during the 48-week follow-up: four of 148 (2.7%) in the TAU group (one leg fracture, one episode of paralysis, one surgical procedure, and one hospitalisation due to high fever) and two of 155 (1.3%) in the CaPE plus TAU group (one incident of cannabis use and one hospitalisation due to high fever; Supplementary Table S1). Four of these events met criteria for a serious adverse event as defined by the International Council for Harmonisation Good Clinical Practice (ICH-GCP): three in the TAU group and one in the CaPE plus TAU group, and none were judged by site investigators to be related to the CaPE intervention.

Discussion

In this multi-centre, assessor-masked randomised trial, CaPE added to TAU substantially reduced the risk of recurrence of mood episodes in adults with BD in remission in Pakistan. The effect on the primary endpoint was clinically important and statistically robust, with an adjusted hazard ratio of 0.54 for recurrence of any mood episode over 48 weeks. The intervention also reduced the risks of both depressive and manic or hypomanic recurrence and improved a range of outcomes, most notably personal recovery, illness-related knowledge and attitudes, medication adherence, and depressive symptom burden. Taken together, these findings suggest that CaPE is effective as an adjunctive maintenance psychosocial intervention for remitted patients with BD in Pakistan when delivered at scale across multiple routine-care sites.⁵ Consistent with the time-to-event analysis, bipolar II disorder was associated with a lower risk of recurrence than bipolar I disorder; this difference warrants cautious interpretation, because subtype was not a randomised comparison, but it suggests that subtype may help identify patients at greatest risk who could be prioritised for adjunctive psychoeducation.

The present trial adds important evidence to the literature on psychosocial interventions for BD. Reviews and meta-analyses have generally supported psychoeducation as one of the more promising adjunctive psychosocial interventions for relapse prevention, but the evidence base has been shaped largely by studies undertaken in high-income countries and often in specialist academic services.2, 3, 4 The UK PARADES trial remains one of the most important pragmatic multi-centre studies in this field.²³ In that study, structured group psychoeducation did not show overall superiority to an active comparator of optimised unstructured peer support, although psychoeducation was more acceptable and may have conferred greater benefit in participants earlier in the course of bipolar illness.²⁴ Our findings do not contradict PARADES so much as extend them to a different demographic and clinical context. Whereas PARADES compared psychoeducation with another structured psychosocial package in a well-resourced setting, our study compared a culturally adapted individual psychoeducation programme with usual care in a setting where structured psychological care is rarely available in routine practice. That contrast is likely to matter for effect size interpretation. Structured peer support of the kind used as the comparator in PARADES is not routinely available in Pakistan and other low resource settings, so usual care rather than an active psychosocial comparator reflects the realistic alternative to CaPE in this context.

Several features of the intervention and setting plausibly contributed to the favourable results. First, CaPE was individually delivered. As stated earlier, in Pakistan concerns about privacy and discomfort discussing mental illness in front of others can reduce engagement with group-based therapy.⁸ Individual delivery might therefore have improved uptake, candour, and personal relevance. Second, the content was explicitly culturally adapted. The use of Urdu, culturally familiar idioms, attention to local beliefs about illness, practical exercises, and optional involvement of family members could have strengthened both comprehension and behavioural implementation.

The pattern of secondary outcomes is informative and clinically plausible. The strongest effects were observed for recovery, knowledge and attitudes, medication adherence, and depressive symptoms. These are the domains that would be expected to shift if patients become more able to understand BD, recognise early warning signs of relapse, maintain pharmacological treatment, and use structured coping strategies in daily life. The improvement in illness coherence on the IPQ-R supports this interpretation. Although illness perceptions other than coherence changed little, a clearer and more coherent model of illness might be especially relevant for BD self-management because it underpins decision making when symptoms begin to recur. The sustained adherence benefit is also noteworthy. Non-adherence is a major contributor to relapse in BD, and sessions focussing on medication, interruption of treatment, and relapse action plans are likely to have been particularly salient.

The stronger and more durable effect on depressive outcomes than on manic symptoms by month 12 also deserves comment. Depressive symptoms often persist between full syndromal episodes and are a major driver of disability in BD.²⁵ A psychoeducation programme that improves routine structure, stress management, insight, and self-monitoring might therefore have especially visible effects on depressive burden. The more modest month-12 effect on YMRS scores does not negate the recurrence findings, because prevention of manic or hypomanic episodes was still evident in the time-to-event analysis. Rather, it suggests that symptom-severity differences between groups may narrow over time once some episodes have already been prevented or treated, or that the study had less power to detect smaller between-group differences on mania ratings at later follow-up.

The trial has several strengths. It recruited a large sample across seven centres, improving generalisability within Pakistan. The primary endpoint was clinically meaningful and assessed with a recognised longitudinal instrument designed for recurrent mood disorders.¹⁷ Follow-up completion exceeded 93% at 12 months, reducing the likelihood that the results are explained by differential attrition. The intervention itself was manualised, linked to a clear theoretical and clinical rationale, and designed from the outset with scalability in mind. CaPE was delivered by master's-level psychologists, a cadre that is relatively more scalable in Pakistan than highly specialised psychotherapists or psychiatrists. This task-sharing logic is important because an intervention can only influence population-level care if it can be delivered by a realistically available workforce. These strengths distinguish the study from much of the earlier LMIC literature, which has often been constrained by single-centre recruitment, smaller samples, and short follow-up.⁴

Several limitations should also be acknowledged. Participants and clinicians could not be masked to allocation, which is an inherent challenge in psychosocial intervention trials and could have influenced behaviour or self-reported outcomes. The early separation of the survival curves, including a relatively steep early rise in recurrence in the TAU arm, could partly reflect expectancy effects among participants aware that they had not been allocated to the active intervention, and we cannot exclude a contribution of disappointment or reduced engagement in the control group. TAU was pragmatic but also makes the control condition heterogeneous. Follow-up retention was high for a 12-month psychosocial trial in bipolar disorder, which may indicate a particularly motivated and engaged sample and could limit generalisability to routine populations with poorer engagement. In addition, the mixed-effects models examined multiple secondary outcomes, so those findings should be interpreted as supportive rather than as independent definitive tests. Because eligibility required euthymia, stable medication, outpatient residence, and the absence of active suicidal ideation, unstable housing, or a substance use disorder, the findings may not generalise to acutely unwell patients, those with active suicidality or substance use, or people who are homeless. We did not systematically collect data on co-occurring neurodevelopmental conditions, physical or intellectual disability, or gender identity, and the extent to which the sample reflects the wider population of people with bipolar disorder in these regions of Pakistan is uncertain; future studies should characterise and include these understudied groups. Race and ethnicity were not recorded, and validated measures of functioning or disability were not systematically collected, and future work should address these gaps in the data. The recruited sample was drawn from urban outpatient psychiatric services in Pakistan and may not represent people with bipolar disorder in rural areas or those not engaged with care, so caution is needed when generalising across sociocultural and socioeconomic contexts, where unmeasured structural drivers rather than individual characteristics are likely to shape outcomes. Notwithstanding these limitations, the trial deliberately recruited a linguistically and educationally diverse sample, including participants with no formal education, and the culturally adapted, task-shared model was designed to widen access to effective maintenance care in an under-served population.

Despite these caveats, the clinical signal in this study is strong. In a context where pharmacotherapy often constitutes most routine care in BD, adding a structured, culturally appropriate psychoeducation programme appears to improve both relapse prevention and several intermediate outcomes that are likely to matter to patients and health services. The implications extend beyond Pakistan. Many health systems serving South Asian populations or other ethnic minority communities face similar problems of stigma, linguistic mismatch, low access to formal psychotherapy, and underuse of non-medical mental health professionals.²⁶ CaPE offers a model for how evidence-based psychoeducation can be adapted without losing the clinical core.

Future work should address several issues. First, a health economic evaluation will help determine whether the clinical gains seen here translate into acceptable value for money. In low-resource settings, cost-effectiveness is essential for policy adoption. Second, process-evaluation data should clarify which elements of CaPE were most important for engagement and benefit, and whether family involvement or specific session topics were especially influential. Third, longer follow-up would be valuable to establish whether the advantages seen at 48 weeks are sustained and whether booster sessions are needed.

Overall, this trial suggests that culturally adapted individual psychoeducation can be an effective adjunct to routine maintenance treatment for BD in Pakistan. By reducing recurrence risk and improving recovery-oriented outcomes, CaPE has the potential to enhance BD care in settings where psychological treatments are scarce yet the need for sustainable, person-centred maintenance interventions is substantial.

Contributors

MIH conceived the study. MIH, MA, ABK, MU, NC, IBC, MOH, FN, NH, BHM, CdO, JZ, WW, TK, and SX contributed to study design. MA, ABK, MU, AHR, MS, MRB, AAM, SM, KMSS, AT, and IBC coordinated recruitment, intervention delivery, and site oversight. MIH, MA, ABK, MU contributed to manuscript writing. WW, RT, and SS contributed to statistical analysis and interpretation of results. MA and ABK contributed to data curation and interpretation of results. MA also served as trial manager. All authors approved the final manuscript. MIH, IBC, NH, RT, and WW directly accessed and verified the underlying trial data. MIH was responsible for the decision to submit the manuscript for publication.

Data sharing statement

Deidentified participant data together with the study protocol and statistical analysis code will be available from the corresponding author on reasonable request after publication, subject to review of a methodologically sound proposal, applicable ethics approvals, and a signed data access agreement.

Declaration of generative AI and AI-assisted technologies in the writing process

AI-assisted technology was used to proof-read the language and formatting of the manuscript during preparation for submission.

Declaration of interests

MIH reports grants from the Canadian Institutes of Health Research and University of Toronto and salary support from the University of Toronto. MIH provides consultancy to Neurocentrx and Otsuka Pharmaceuticals and has led contracted research for COMPASS Pathways Ltd. NH has previously served as a Trustee of the Pakistan Institute of Living and Learning (PILL), Abaseen Foundation UK, Lancashire Mind UK, and Manchester Global Foundation (MGF). NH has received research funding from NIHR, UKRI, CIHR, SMRI, and other research funding bodies, and he is an NIHR Senior Investigator. BHM reports support for the present manuscript from the University of Toronto (Labatt Family Chair in Biology of Depression in Late Life), the Centre for Addiction and Mental Health (CAMH) (compensation), and The Royal Ottawa Hospital (compensation); grants from Brain Canada, the Canadian Institutes of Health Research, the Patient-Centred Outcomes Research Institute, the US National Institute of Health, and the Krembil Foundation; an honorarium from Columbia University; is a member of the Board of Trustees of The Royal Ottawa Hospital; and reports use of software in clinical trials from Capital Solutions Design LLC (trial funded by CAMH Foundation) and HAPPYneuron (trial funded by Brain Canada). JZ is a Director of the CAMH Foundation. SX reports a seed grant from the University of Toronto Collaborative Centre for Climate, Health & Sustainable Care.