Cognitive remediation therapy for patients with bipolar disorder: A randomised proof‐of‐concept trial

Cognitive remediation therapy (CRT) may benefit people with bipolar disorder type I and II for whom cognitive impairment is a major contributor to disability. Extensive research has demonstrated CRT to improve cognition and psychosocial functioning in people with different diagnoses, but randomised trials of evidenced therapy programmes are lacking for bipolar disorders. The Cognitive Remediation in Bipolar (CRiB) study aimed to determine whether an established CRT programme is feasible and acceptable for people with bipolar disorders.


| INTRODUC TI ON
Bipolar disorder (BD) is common, complex, and often inadequately treated: calls have begun to pursue evidence-based therapies that target fundamental cognitive process 1 since two thirds of patients suffer cognitive deficits which are often pronounced and widespread (spanning memory, attention, and executive functions), persisting throughout periods of recovery and directly linked to psychosocial and occupational functioning impairments. 2 As well as functional impacts, a pattern of cognitive deterioration may increase episode recurrence, so improving cognitive abilities might enhance not only short-term wellbeing but potentially the future course of the disorder. 1,3,4 Cognitive Remediation Therapy (CRT) is a psychological intervention that has demonstrated reliable benefits for people with schizophrenia, improving core cognitive functioning, daily functioning, and quality-of-life as determined through extensive examination via clinical trials (efficacy, effectiveness, mechanistic) and comprehensive meta-analytic syntheses. eg5-7 There are well-documented similarities between the cognitive difficulties experienced by people with schizophrenia and bipolar disorder: they appear comparable in terms of impact and quality, although are on average less severe for patients with BD than schizophrenia and there is less evidence of a premorbid component. 8 Preliminary evidence for the potential benefits of CRT for BD is encouraging: an initial meta-analysis found analogous cognitive effects in trials including some participants with affective disorders (ES = 0.44) compared to those only assessing individuals with psychotic disorders (ES = 0.36). 7 A handful of small or non-randomised studies have ensued, recently synthesised in systematic reviews 3,4 reporting only two published randomised trials of CRT (one for those with psychotic bipolar disorder, 9 the other employing a short-term group therapy paradigm for a small sample of patients with BD 10 ) concluding that individual CRT provided more intensively or for a longer duration might demonstrate meaningful benefits.
We present here the findings of a randomised proof-of-concept trial (the "CRiB" study) comparing an evidence-derived, intensive, individual manualised CRT intervention with treatment-as-usual. The CRiB trial aimed to determine the potential for CRT development into a new evidence-based therapy for bipolar disorder (type I and II) by investigating four primary outcomes: whether an established therapist-led CRT programme is 1) feasible and 2) acceptable for this patient group and whether it yields benefits in comparison to usual care, in terms of 3) improved cognitive performance and 4) improved psychosocial functioning.

| Study design
This was a single-blind, proof-of-concept randomised trial conducted

| Participants
Participants were eligible if they had a diagnosis of BD, were currently euthymic, aged 18-65 years without meeting exclusion criteria (not being able to communicate in English, not being able to use a computer, have a substantial neurological illness, substanceuse, or personality disorder). No other eligibility restrictions were placed on somatic illnesses or medications taken either upon trial entry or medication changes throughout participation, in line with usual care. However, we did ask patients whether they were undergoing a psychological therapy at present or planning any treatment changes in the near future and if so, delayed screening for study eligibility. Current euthymia was defined as scoring <8 on the Hamilton Rating Scale for Depression (HAMD) 12

and Young Mania
Rating Scale (YMRS) 13 on two occasions one week apart assessing mood over the preceding month. Eligibility, including diagnosis, was assessed by trained researchers (DT, TM, RS) and validated by a psychiatrist (NY, KM, AHY). Participants were recruited via secondary and primary care services as well as off-and on-line advertisement from the community. Written informed consent was obtained from all individuals before participating, following full explanations of the procedures involved.

| Randomisation and masking
Participants were assigned to receive CRT in addition to treatmentas-usual (TAU) or TAU alone following completion of the baseline Conclusions: CRT is feasible for individuals with bipolar disorders and may enhance cognition and functioning. The reported effect sizes from this proof-of-concept trial encourage further investigation in a definitive trial.

K E Y W O R D S
bipolar, cognitive enhancement, cognitive remediation, CRT, randomised trial assessment (week 0). A 1:1 block allocation design with random and variable block size was generated by an independent, validated web-based randomisation system including concealed sequence allocation and locked randomisation details after allocation, as detailed in the protocol. 11 Randomisation was stratified by the intensity of TAU as assessed by the Client Service Receipt Inventory (CSRI) service-use assessment 14 (intensive or non-intensive TAU, defined respectively as more or less than monthly appointments with a healthcare professional). Investigators responsible for obtaining informed consent and undertaking data collection (including symptoms, functioning, and cognitive outcome ratings; DT/ TM) were blinded to group allocation; the trial statistician (JH) was also blinded. To maintain investigator blinding, the unblinded trial coordinator (RS) carried out randomisation, conveyed allocation by telephone or in person to therapists and participants, requesting for participants to not disclose their group allocation to data collectors. Those delivering and receiving CRT intervention were unblinded out of necessity.

| Intervention
Over 12 weeks, CRT was delivered by trained postgraduate psychologists with supervision from clinical psychologists with >5 years CRT experience. The metacognition-informed, therapist-led, computerised CRT program "CIRCuiTS" emphasises strategy use and transfer having been established in populations diagnosed with psychoses 6,15,16 ; In addition to compensatory (or strategy-based) remediation approaches, all session formats also included a restorative remediation element (eg, "drill and practice" of cognitive tasks that were clearly distinct from cognitive outcome assessments). CRT flexibly utilised a combination of formats (in-person, telephone, and individual practice sessions) which were arranged under the responsibility of the therapist, accommodating the patients' needs. Target engagement comprised three sessions per week including at least one session guided by the therapist each week, with a target total of 20-30 hours. The minimum dose of 20 hours was proposed following evidence in schizophrenia research. 16 Appendix A contains additional details.

| Procedures
As detailed in the protocol, 11 all participants continued receiving usual care throughout the trial. After baseline assessment and randomisation (week 0), the 12-week intervention period comprised of CRT + TAU or TAU alone. After the intervention period, all outcomes (cognitive, functional, acceptability) were repeated at a post-intervention assessment (week 13) and a follow-up assessment 12 weeks later (week 25). 11 Each assessment was conducted in a private, quiet room and was only undertaken where participants had capacity to consent and complete the assessments.
Participants were instructed to sleep well (as much as possible) the night before assessments and to entirely avoid use of alcohol or non-prescribed drugs on the day of assessments. Cognitive tests were undertaken as close as possible to 1-4 PM to standardise daily fluctuations in alertness.

| Outcomes
As detailed in the protocol, the four main feasibility outcomes were considered primary with equal weighting. These, and the measures used to examine them, were: 1. Trial feasibility: recruitment (number consenting to participate) and completion (number completing the final outcome assessment) rates assessed acceptability of randomisation to TAU relative to CRT. No specific criteria were set a priori to determine sufficient feasibility (recruitment/ completion rates) to progress the research to a comprehensive future trial.  23 and perceived cognitive impairment (perceived deficits questionnaire [PDQ]). 24 A composite score of cognitive function (processing speed, working memory, verbal memory, verbal fluency, IQ, executive function domains) was calculated to examine global cognition. 11 Of these, the DSST was proposed as a potential primary outcome for a future efficacy CRT trial because it has shown high sensitivity to change over time. 19 Although IQ is commonly considered a stable trait the WASI assessment has been validated to show variability over time and is commonly used as a "current intellectual functioning" tool. 21 The cognitive tests and the UPSA (below) have known practice effects, thus it was expected that improvements would be observed over time across the sample and betweengroups comparisons addressed this to indicate relative performance between those randomised to TAU vs CRT. Additionally, symptoms of depression 12 and mania 13 were monitored and adverse events recorded throughout.

| Statistical analysis
The target sample size of 60 was selected due to recommendations for pilot trials to estimate efficacy. 28 All planned analyses adopted the intention-to-treat (ITT) approach. As detailed in the protocol 11 descriptive statistics summarized participant characteristics as appropriate (mean, standard deviation [SD], median, interquartile range [IQR], frequency and percentage [%]). Inferential statistics produced group differences and standardised effect sizes (SES) with confidence intervals (95% CI) computed using R version 3.6.1. 29

| Outcome 1-Trial feasibility
Descriptively, the rate of consenting eligible participants, assessment non-completion and study withdrawals were examined alongside reasons for each trial arm.

| Outcome 2-CRT intervention acceptability
Descriptive statistics examined intervention acceptability (CRT satisfaction questionnaire) 17 and feasibility (number of participants completing 20 therapy sessions, total hours engaged with therapy, reasons for non-engagement). A health and social care perspective was adopted in the cost-effectiveness. The economic costs of CRT and TAU services were calculated; service-use during the trial was measured by the CSRI 14 and combined with standard unit costs of these services 30 added to intervention costs (the resources required to deliver CIRCuiTS, eg, therapist time, overheads). These were combined with quality-adjusted life years derived from the EQ-5D-3L 18 using area under the curve methods to compare groups. 11

| Outcomes 3 and 4-putative effects of CRT on cognition and function
Intervention effect sizes (ES) were estimated using linear mixed models for each outcome measurement at both post-treatment (week 13) and follow-up (week 25) with outcome timepoint as dependent variable plus fixed effects of trial arm (CRT vs TAU) and baseline measures of outcome. Standardised effect sizes (SES) and 95% CI were calculated by dividing the group mean differences by standard deviation of the baseline outcome score. The following convention was used to interpret effect size: small 0.2-0.49, moderate 0.5-0.79, large ≥0.8. As a pilot study, estimation is focused on ES but includes pvalues as an aid to interpretation. Missing data were dealt with using a maximum likelihood (ML) approach with the missing at random (MAR) assumption. Baseline predictors of missingness were sought, to be included in the analysis model if an association was found at a liberal P < .2 criterion. The study was monitored by a Trial Steering Committee and the trial was pre-registered in the ISRCTN registry (ID32290525; http://www.isrctn.com/ISRCT N3229 0525).

| Changes to the protocol
Compared to the published protocol, minor changes since trial commencement are as follows: 1. Diagnostic subtype: only patients with bipolar disorder type I were eligible. From November 2016 (recruitment month 9, after n = 25), eligibility was broadened to include individuals with BD type II in response to lack of evidence suggesting cognitive differences between diagnostic subtypes. 2 2. Sub-analyses: We had planned to analyse differences in cognition over the menstrual cycle but many patients were not currently menstruating and therefore these analyses were not included.

Functional capacity:
A practical test of functioning was added to the outcome assessment protocol prior to trial commencement, to assess function using an applied measure. The UPSA 25 is a recommended outcome measure in cognitive trials due to its moderate correlations with cognitive performance and psychosocial functioning. 31  Clinical characteristics are displayed in Appendix Table A1.

| Recruitment rate
A mean of 2.3 participants per month were recruited and consented into the trial (SD = 0.97, range 0-5). For a sample size of 60, 226 potential participants were identified, giving a recruitment rate of 36% (95% CI [29%, 44%]). The most common reason for non-participation was inability to commit to appointments at the research site multiple times per week for 12 weeks for CRT, most frequently due to long travel distances required to attend.

| Completion rate
Of the 60 individuals randomised, seven participants (12%, 95% CI [5%, 22%]) were lost to follow-up due to withdrawal (n = 6) or F I G U R E 1 CONSORT diagram. *26 participants received at least some CRT therapy. The two participants who did not complete any therapy plus two who completed < 10 hours CRT had the same therapist who did not engage with the study. One further participant was identified subsequently as meeting an exclusion criterion for the study (10-19 hours therapy). These five individuals were subject to sensitivity analysis for cognitive and functional outcomes because they were considered to not have been treated in full adherence to the protocol, unlike other participants (including those who withdrew, who were not excluded in per-protocol analyses) contact loss (n = 1). The odds of dropout were non-significantly higher in TAU (n = 5) than CRT (n = 2) groups; OR = 2.31, P = .44.
Withdrawal immediately after randomization occurred for one CRT participant (life events) and two TAU participants (one unsatisfied with randomization to TAU; one life events). Loss at follow-up during the intervention period occurred for one CRT participant (illness) and three TAU participants (one sought other cognitive intervention; one illness; one contact loss). All other participants attended all assessments: unless specified, all analyses include all participants.
No demographic or clinical baseline predictors of missing data were identified (all P > .25).  participants were allocated the same therapist: this therapist was subsequently found to not comply with study protocols (including adherence to treatment guidelines and timely communications with participants), which was thus considered a protocol deviation. Of these four participants, two did not begin therapy and two completed <10 hours. These five participants whose participation in the trial were considered to deviate from the original protocol were subject to sensitivity analysis (N = 55 is referred to as "per-protocol," N = 60 as "ITT analysis").

| Putative functional effects of CRT
At baseline, 27 (55%) of participants were categorised as having moderate to severe impairments in functioning while 33 (45%) had no or mild functional impairment, according to validated FAST cut-off scores 32 (see Table 2). Similar to cognition, between-group effect sizes indicate larger improvements in CRT than TAU participants on functional measures ( Table 2). Linear mixed models (see

| Affective symptoms
Symptoms were similar between groups (see Appendix Table A1)

| Adverse events
No serious adverse events occurred. Four adverse events were recorded, each for different participants, related to study procedures (experience of distress at feeling they performed poorly on cognitive tests, n = 2) or bipolar illness (depressive episode, n = 2). No suicide attempts or self-harm were disclosed to the study team.

TA B L E 2 Cognitive and functional measures at all time points
Test

| D ISCUSS I ON
These results indicate high feasibility and acceptability of individual, therapist-led CRT using the established CIRCuiTS programme, as a potential treatment to enhance cognition and functioning for BD, with promising effect sizes on cognitive and functional outcomes.

| Feasibility and acceptability
The ADePT (Decision-making after Pilot and feasibility Trials) process aims to stimulate systematic identification and consideration of problems and solutions arising from feasibility trials and enhance the decision-making process in progressing to future studies. 33 While a priori progression criteria appear optimal, ascertaining the potential for progression post-hoc is often more appropriate to maximise comparability between the present trial findings and the most relevant, high-quality research published to date. For example, cognitive trial recommendations 31 and the only randomised individual CRT trial for BD 9 were not available by the time of CRiB protocol finalisation.
ADePT considers issues related to whether it is feasible to achieve adequate power in a future efficacy trial based on sample size calculations using effect sizes indicated from pilot trial outcomes: We report mostly medium ES in line with previous work. 9 Achieving adequate power also relies on the feasibility of meeting recruitment targets by identifying sufficient participants who are eligible and willing to consent (in CRiB, similar to previous trials 9, 10,15 ), the number of patients engaged with intervention and completing the trial (CRiB being equivalent or better compared to previous similar studies 9,14,34 with its 12% loss to follow-up rate and high therapy completion rate) between the number of required sites (multisite differences not assessed in the single-site CRiB study). In addition, we did not experience issues with key aspects of the trial design (randomisation, blinding) or cost and duration of intervention (with health economic analyses indicating non-significant service use/QALY differences between CRT and TAU groups). Outcome selection was also in accordance with taskforce recommendations 31 ; thus taken together, our findings suggest an efficacy trial of CRT for BD is both feasible and justified in terms of its putative benefits to patients.
The CRT satisfaction questionnaire indicated high acceptability, which we aim to examine more comprehensively in qualitative analyses for understanding and potentially refining the CIRCuiTS treatment manual tailored to people with BD. Five non-completers' participation deviated from the trial protocol (one due to ineligibility, four whose therapist did not adhere to protocols). Sensitivity analyses removing these participants indicate somewhat higher ES after they were removed. The protocol deviations alone suggest that patients with active substance-use disorders should be advised to seek help for this before undertaking CRT and emphasises the need to prioritise therapist adherence to protocols when designing future trials (ie, carefully consider therapist selection to ensure their adherence to study and treatment protocols, and engagement with participants). After excluding the protocol-deviation participants, the therapy completion rate was 93% (7% study withdrawals due to illness or life events) which indicates very high feasibility of CIRCuiTS.
Minimal adverse events also imply safety. Future efficacy studies should examine the long-term cost-effectiveness of CIRCuiTS vs TAU in the months and years following therapy completion, as is

| Putative effects of CRT
Despite a relatively small sample size, large effects on cognition and functioning were observed, enduring for three months after the end of therapy. These are similar to effects observed in CRT for schizophrenia 5,6 and equivalent 9 or greater than 10,34 previous BD trials. As reported in recent systematic reviews 3,4 a handful of studies investigating related interventions for BD.
A distinct intervention focusing on functional rather than cognitive remediation has been supported in a randomised trial finding small benefits in functioning compared to psychoeducation treatment. 34 Specific CRT programmes have been tested in 4 studies.
Initially, one open-label study reported cognitive and clinical improvements particularly for participants less cognitively impaired at baseline (N = 18). 36 Conclusions are limited by the single-arm study design but this promoted CRT's potential for this population. Two subsequent group-CRT studies followed; one naturalistic study reporting non-significant cognitive benefits (N = 26) 37 and a randomised study finding benefits to subjective but not objective cognitive tests. 10 The group CRT studies conclude that more intensive, therapist-led individual CRT emphasizing F I G U R E 2 cognitive outcomes after CRT and TAU. Effect size (time x group interactions) at week 13 and week 25 compared to pretreatment cognitive performance. * The Hotel test was reverse-scored, meaning that a negative effect size represents greater improvement. we report compared to previous BD studies may be attributable to this. This speculation is supported further by the reports from the only other randomised trial examining individual CRT for individuals with BD, whose CRT intervention consisted of intensive training with multiple weekly sessions over 24 weeks. 9 This randomised trial was for patients with psychotic bipolar disorder, representing a slightly different population to CRiB, but reported similar medium/high effect sizes to those we present here. 9 Together, the extant studies support the use of time-intensive individual cognitive remediation for people with BD but a focus is needed on therapy components in upcoming studies.
Some cognitive improvements were not evident in our study; notably the participants' perceived deficits, which contrasts somewhat with CRT satisfaction responses and objective performance tests, but possibly reflects an increased awareness of individuals' own cognitive weaknesses after CRT (especially given the metacognitive focus of the intervention

| Strengths and limitations
Notable strengths of this study include trial preregistration, a low proportion of missing data, the use of intention-to-treat analyses, stringent randomisation, and blinding of assessment researchers.
The ability to determine benefits of CRT are limited by the feasibility nature of the study and correspondingly small sample size. The sample size prevented a more in-depth examination of the effects of CRT; we were unable to adjust for factors known to influence cognition such as various medications (psychotropic and somatic), physical illness, and sleep quality. We stratified randomisation based on the intensity of usual care patients received at baseline, which balanced TAU between groups; however, there is at present little justification for taking this decision in a future trial and stratifying based on cognitive impairment would likely be more beneficial, not least for ensuring that baseline impairment is well balanced between randomisation allocations. In this trial, CRT/TAU groups showed little difference on cognitive or functional variables except that verbal recall performance was better in the CRT group at baseline (ES = 0.6 and was one of the few domains which did not show a relative improvement in CRT compared with TAU participants at follow-up). This occurrence can happen in small trials and is addressed through adjusting for baseline performance in linear mixed models, but one would expect poor performance to precede higher effect sizes based on the notion of "room for improvement." 31 This is an assumption made in recommendations for trials to recruit only cognitively impaired individuals; namely the ISBD taskforce recommendations which were not available at the time of CRiB initiation; however, more evidence to support or refute this is certainly warranted. It is also worth noting that cognitive assessments (and the UPSA) have known practise effects which were addressed through between-group comparisons and baseline adjustment in linear mixed models. The results of this RCT require replication in larger, more definitive trials with multiple sites (since single studies such as CRiB limit generalisability of findings). Our participants were predominantly white, British and educated; more diverse samples are needed in future trials to better represent BD populations and improve generalisability of results. We recommend that these trials also seek to establish the core therapeutic components of CRT, the role of metacognitive skill development and whether there is a "target" group of patients with BD who would confer the most meaningful and enduring benefits from CRT, which would obviate unnecessary resource-intensive studies. In the current study, outcome assessors were instructed to disclose any incidents of inadvertent unblinding and an alternative assessor undertook these outcome assessments. Because procedures for recording and addressing incidents of unblinding were not formalised, it cannot be confirmed whether any assessments were undertaken by assessors who might have suspected which group the participant had been allocated to. Future trials should therefore formally assess whether, despite blinding, inadvertent beliefs about which trial arm participants were allocated could influence trial outcomes.

| Future implications
While current clinical practice commonly uses pharmacological treatments targeting acute affective symptoms, clinical evidence 39 and our engagement work indicates that service users desire additional psychological therapies which may enhance quality-of-life more broadly.
CIRCuiTS may be a good candidate to address these needs. Previous examination of CRT for bipolar disorder is limited and reports mixed findings, often from non-randomising studies or using less evidencebased programmes. 1 The CRiB study sought to fill this evidence gap in setting a precedent for future work to establish whether this intervention has real potential for people with BD via efficacy and mechanistic trials. Considering the present findings, we urge future CRT studies to use robust methodology and manualised interventions to advance comprehension of the putative benefits of CRT and accumulate goodquality evidence to inform clinical practice.