Definitions
Cognitive behavioural therapy
Cognitive models of bipolar disorder suggest that dysfunctional thoughts and beliefs may be triggered by both positive and negative life events and influence mood and behaviour (Newman et al., 2003). CBT (Lam et al., 2010; Meyer & Hautzinger, 2012) is a form of talking therapy that focuses on the role thinking and behaviour has on emotions, and how they reciprocally influence each other. CBT for bipolar disorder typically consists of 12 to 20 individual sessions over a period of 6 months.
Group psychoeducation
Group psychoeducation (Castle et al., 2010; Colom et al., 2003b) is a relatively intensive intervention in which the patient attends weekly group sessions lasting from 90 minutes to 2 hours for up to 21 weeks. Each group session is designed to provide information on a key aspect of bipolar disorder with time allocated for group discussion on the chosen topic. The rationale for these groups is that by learning more about the symptoms, treatment and coping strategies relevant to bipolar disorder, service users will become more skilled in self-managing their condition.
Family-focused therapy
Family-focused therapy (Miklowitz et al., 2003) is a psychoeducational programme for individual families based on behavioural family therapy principles (Falloon et al., 1993), which have previously been applied effectively in the treatment of people with schizophrenia. In family-focused therapy the service user and family members are offered 21 sessions over a 9-month period. Therapy has three phases beginning with psychoeducation and relapse prevention followed by work on improving family communication and, finally, developing problem solving skills for both service user and family.
Self-management training groups
Self-management training groups are typically offered in a group format and facilitated by individuals with personal experience of bipolar disorder. They are informed by both cognitive therapy approaches and by a focus on relapse prevention (Copeland, 1994). Sessions have been delivered in a variety of ways from single intensive weekend courses to weekly group sessions. The focus of self-management training is for service users to learn more about how to avoid relapses by sharing coping skills in the group setting and developing relapse avoidance plans that are used after the group sessions are completed.
Relapse prevention/individual psychoeducation
Relapse prevention is informed by previous work in psychosis on coping strategy enhancement (Lobban et al., 2010; Perry et al., 1999). These approaches involve clinicians teaching service users to detect early changes in mood and to apply helpful strategies to avoid these early changes escalating in full episodes of mania or depression. Service users are typically offered six to 12 sessions over a period of 4 to 6 months. Enhanced relapse prevention (Lobban et al., 2010) has a stronger emphasis on facilitating self-management coping approaches (teaching the service user to use psychological techniques to manage their mood changes) in addition to accessing additional service support.
Interpersonal and social rhythm therapy
People with bipolar disorder often experience disrupted sleep patterns, and the circadian instability and appraisal model suggests that they are particularly sensitive to disturbances of 24 hour circadian rhythms, which trigger mood disturbances that themselves cause further circadian effect (Goodwin & Jamison, 2007).
Interpersonal and social rhythm therapy (Frank et al., 2005) is based on interpersonal therapy (Klerman et al., 1984) but adapted for bipolar disorder to try to help people develop more stable social rhythms. It focuses on two areas: (1) supporting service users to discuss experiences of change and loss associated with their bipolar disorder and how to deal with them; and (2) helping service users to learn to monitor their patterns of sleep and activity and stabilise these where required. Interpersonal and social rhythm therapy is an intensive psychological intervention of 39 to 40 individual therapy sessions over a period of 2 years.
8.1.2. Studies considered51
Fifty-five trials of psychological and psychosocial interventions met the inclusion criteria for this review: BALL2006 (Ball et al., 2006), BARROS2012 (De Barros Pellegrinelli et al., 2012; De Barros Pellegrinelli et al., 2013), BAUER2006a (Bauer et al., 2006a; Bauer et al., 2006b), BERNHARD2009 (Bernhard, 2009), BORDBAR2009 (Bordbar, 2009), CASTLE2010 (Castle et al., 2007; Castle et al., 2010), CLARKIN1998 (Clarkin et al., 1998), COCHRAN1984 (Cochran, 1984), COLOM2003a (Colom et al., 2003b), COLOM2003b (Colom et al., 2003a; Colom et al., 2009; Miklowitz, 2009), COSTA2012 (Costa et al., 2012), DIJK2013 (Van Dijk et al., 2013), DOGAN2003 (Dogan & Sabanciogullari, 2003), DSOUZA2010 (D’Souza et al., 2010), EKER2012 (Eker & Harkin, 2012), FAGIOLINI2009 (Fagiolini et al., 2009; Kupfer et al., 2009), FRANK1999a (Frank et al., 1999), GENT1991 (Van Gent & Zwart, 1991), GLICK1993 (Glick et al., 1993), GOMES2011 (Gomes et al., 2011), JAVADPOUR2013 (Javadpour et al., 2013), JONES2014 (Jones et al., 2014), KESSING2013 (Kessing et al., 2013), KILBOURNE2008 (Kilbourne et al., 2008), KILBOURNE2012 (Kilbourne et al., 2012), LAHERA2013 (Lahera et al., 2013), LAM2000 (Lam et al., 2000), LAM2003 (Lam et al., 2003), LOBBAN2010 (Lobban et al., 2010), MADIGAN2012 (Madigan et al., 2012), MEYER2012 (Meyer & Hautzinger, 2012), MIKLOWITZ2000 (Miklowitz et al., 2000), MIKLOWITZ2007b (Miklowitz et al., 2007b), MILLER2004 (Miller et al., 2004), PARIKH2012 (Parikh et al., 2012), PERICH2013 (Perich et al., 2013), PERLICK2010 (Perlick et al., 2010), PERRY1999 (Perry et al., 1999), PROUDFOOT2012 (Proudfoot et al., 2012), REA2003 (Rea et al., 2003), REINARES2008 (Reinares et al., 2008; Reinares et al., 2004), SAJATOVIC2009 (Sajatovic et al., 2009), SCHMITZ2002 (Schmitz et al., 2002), SCHWANNAUER2007 (Schwannauer, 2007), SCOTT2001 (Scott et al., 2001), SCOTT2006 (Scott et al., 2006), SIMON2005 (Simon et al., 2005), SMITH2011 (Smith et al., 2011b), SWARTZ2012 (Swartz et al., 2012), TODD2014 (Todd et al., 2014), TORRENT2013 (Torrent et al., 2013), WEISS2007 (Weiss et al., 2007), WEISS2009 (Weiss et al., 2009), WILLIAMS2008 (Williams et al., 2008), ZARETSKY2008 (Zaretsky et al., 2008).
A further five trials were excluded; three because a minority of participants had bipolar disorder and it was not possible to obtain disaggregated data: JACKSON2008 (Jackson et al., 2008), PICKETTSCHENK2008 (Pickett-Schenk et al., 2008) and STARING2010 (Staring et al., 2010); one because on closer inspection it did not appear to be randomised: COSTA2011 (Costa et al., 2011); and one because the GDG determined it was not relevant to the UK: DASHTBOZORGI2009 (Dashtbozorgi et al., 2009).
Two ongoing studies were also identified: PRASKO2013 (Prasko et al., 2013) and GINDRE2009 (Gindre et al., 2009).
Of the 55 included studies, two were unpublished (BERNHARD2009, SCHWANNAUER2007) and the other 53 were published between 1984 and 2014. Seven were not included in the meta-analysis because the authors did not report useable outcomes, which remained unavailable after contacting authors: CLARKIN1998, BARROS2012, EKER2012, FAGIOLINI2009, GLICK1993, PARIKH2012, and WEISS2007.
Study characteristics
Included studies randomised 6,010 participants, ranging from 19 to 441 per study (a summary of study characteristics can be found in Appendix 23). Studies were conducted in North America (k = 22), England and Ireland (k = 12), Europe (k = 11), Australia (k = 5), Brazil (k = 3), and Iran (k = 2). Participants were recruited from an outpatient (k = 23) or inpatient setting (k = 12), GP practice (k = 2), community mental health team (k = 2), or via advertising combined with referral (k = 16). In 52 studies a diagnostic interview was used to establish the presence of a bipolar disorder, in one study participants themselves reported if they had a bipolar disorder, another confirmed the diagnosis through a mood questionnaire, while one study only reported that bipolar disorder was an inclusion criteria.
The median of mean age of participants was 40 years (range of 26 to 55 years), 58% were female and 81% had bipolar I disorder. Four studies included participants in a depressed episode at baseline (MIKLOWITZ2007b, SCHMITZ2002, SWARTZ2012, DIJK2013), six studies had a mix of participants in depressed or manic episode (BAUER2006a, CLARKIN1998, FRANK1999a, GLICK1993, MILLER2004, SAJATOVIC2009) and 32 studies included euthymic participants. Twelve studies (FAGIOLINI2009, KILBOURNE2012, KILBOURNE2008, MIKLOWITZ2000, PERLICK2010, PROUDFOOT2012, SCOTT2001, SCOTT2006, SIMON2005, TODD2014, WEISS2009, WEISS2007) included a mix of euthymic and symptomatic participants at baseline, while two (PROUDFOOT2012, TODD2014) provided disaggregated data.
8.1.3. Clinical evidence review
Evidence from each important outcome and overall quality of evidence are presented in Appendices 23 to 26.
Risk of bias
No trials were at high risk of bias for sequence generation (not truly random), however, the method of randomisation was unclear (not reported) in 15 trials. Allocation concealment was unclear in 25 trials and low risk in 30 trials. All trials were at high risk of bias for blinding for participants and providers per se. Nine trials had no assessors and 31 reporting assessor-rated outcomes used a blind assessor and were at low risk of bias for blinding, but eight studies did not have blind assessors, which was a reason for a high risk of bias. For six studies, blinding of assessors remained unclear. For incomplete outcome data, almost half (k = 25) of the trials were at low risk of bias and the other half (k = 23) were at high risk of bias because of the high amount of dropouts or because dropouts were excluded from the analyses.
There was a risk of outcome reporting bias in 22 trials. Only 11 studies were prospectively registered, but 23 others were assessed to be at low risk of bias because authors provided missing data or confirmed that all outcomes were published. Risk of publication bias could not be assessed by means of funnel plots because of the small number of studies per intervention.
Overall quality of the evidence
Most evidence was of low or very low quality. Nearly all results were downgraded at least one level owing to imprecision because the analyses included few participants or events, and/or the boundaries of the confidence interval crossed the decision-making threshold. Also, risk of bias in studies and reporting bias had a negative influence on some of the outcomes. Some outcomes were also downgraded for inconsistency when there was evidence of statistical heterogeneity.
Post-treatment data were mostly of low to very low quality. Only relapse data on individual interventions, hospitalisation data on collaborative care and discontinuation on interpersonal and social rhythm therapy were of moderate quality.
Studies also reported controlled comparisons at follow-up, but most outcomes were of very low quality, except for most hospitalisation and relapse outcomes with regards to the comparisons of individual and group psychological interventions, and family psychoeducation with treatment as usual.
Effects of interventions
Across nine comparisons, results of the meta-analyses suggest that psychological interventions may be associated with symptomatic improvement, reduced relapse and hospitalisation. The majority of these moderate to low quality outcomes are summarised per comparison and presented in (post-treatment) and (follow-up), and additional outcomes are presented in Appendix 26. Reasons for downgrading are given per outcome in the tables52.
Outcomes at post-treatment.
Individual psychological interventions
The search identified RCTs of face-to-face psychoeducation and interactive online psychoeducation (DOGAN2003, JAVADPOUR2013, LOBBAN2010, PERRY1999, PROUDFOOT2012, SMITH2011, TODD2014), CBT (BALL2006, JONES2014, LAM2000, LAM2003, MIKLOWITZ2007b, SCOTT2001, SCOTT2006, ZARETSKY2008) and medication adherence therapy (COCHRAN1984). Eleven trials started with euthymic participants at baseline, and four had a mix of participants in an acute episode and euthymic (PROUDFOOT2012, SCOTT2001, SCOTT2006, TODD2014).
At post-treatment, seven trials (N = 637) reported low quality evidence that individual psychological interventions when compared with treatment as usual, produced a small effect in symptoms of depression (see ). Six trials (N = 365) reported moderate quality evidence that individual psychological interventions reduced the risk of relapse. One trial with few events was inconclusive regarding the risk of hospitalisation.
At follow-up, seven trials (N = 446) reported moderate quality evidence that individual psychological interventions were associated with a long-term reduction in the risk of relapse (see ). In three studies (N = 214) there was a reduction in the risk of hospitalisations, but the estimate was imprecise.
One study (N = 76) compared individual CBT with supportive therapy for depression (MEYER2012). At follow-up, there was very low quality evidence favouring supportive therapy for symptoms, but the effect on relapse was not conclusive (see ).
Group psychological interventions
The search identified trials of group interventions including psychoeducation, (CASTLE2010, COLOM2003A, COLOM2003B, SAJATOVIC2009, TORRENT2013) CBT (BERNHARD2009, COSTA2012, GOMES2011), mindfulness (PERICH2013, WILLIAMS2008), social cognition and interaction training (LAHERA2013), and dialectical behaviour therapy (DIJK2013). Interventions were compared with treatment as usual, except for two studies that compared psychoeducation with attention control (COLOM2003A, COLOM2003B). In ten trials, participants were euthymic at baseline (BERNHARD2009, CASTLE2010, COLOM2003A, COLOM2003B, COSTA2012, GOMES2011, LAHERA2013, PERICH2013, TORRENT2013, WILLIAMS2008) and two studies included participants experiencing an acute episode (SAJATOVIC2009, DIJK2013).
Eight trials (N = 423) reported very low quality evidence of a small effect on depression outcomes (see ). Furthermore, the two studies comparing psychoeducation with attention control (N = 170) found a reduction in depression and mania relapses. In three trials (N = 205) the effect estimate on the number of hospitalisation was very imprecise.
Long-term results in five studies (N = 333) reported low quality evidence of a reduction in depression relapses (see ). Also, four studies (N = 274) reported a reduction of relapses into mixed episodes. However, the effect on depression symptoms and hospitalisation was inconclusive.
Family psychoeducation
Two trials included an intervention on psychoeducation for service users and their family members (DSOUZA2010, MILLER2004) and in five trials psychoeducation was only for family members (BORDBAR2009, MADIGAN2012, PERLICK2010, REINARES2008, GENT1991). Five trials started with euthymic participants at baseline (BORDBAR2009, DSOUZA2010, MADIGAN2012, REINARES2008, GENT1991), one trial had a mix of participants in an acute episode and euthymic (PERLICK2010) and another included only participants in an acute episode (MILLER2004).
In comparison with treatment as usual, one trial (N = 43) found low quality evidence of medium effect in depression symptoms favouring family psychoeducation at post-treatment (see ).
At follow-up, three trials (N = 228) reported low quality evidence of a reduction in the risk of relapse (see ). One trial (N = 113) reported a reduction in the risk of mania relapses, but the effect on depression relapses was inconclusive. One study (N = 57) reported a very large effect on reduction of the number of hospitalisations, but effect estimates were imprecise with only nine events in the study.
Family-focused therapy
Trials of family-focused therapy included participants who were euthymic (REA2003), either in an acute episode and euthymic (MIKLOWITZ2000), only depressed (MIKLOWITZ2007b) or in any type of episode (MILLER2004).
Post-treatment data were of low quality. One study (N = 79) found a medium effect favouring family-focused therapy when compared with treatment as usual on depression symptoms (see ). Furthermore, a study (N = 53) comparing family-focused therapy with psychoeducation found little difference with regard to relapse, but the estimate was imprecise.
The follow-up evidence was of very low quality and found little difference in effects on depression symptoms, relapse and response, but the estimates were imprecise (see ). The evidence suggested family-focused therapy reduced the risk of hospitalisation.
Interpersonal and social rhythm therapy
There were three trials of interpersonal and social rhythm therapy with participants in an acute episode at baseline (FRANK1999a, MIKLOWITZ2007b, SWARTZ2012). At post-treatment, very low quality from one study was inconclusive with regard to symptoms of depression, relapse and response (see ). At follow-up, one trial (N = 41) reported that interpersonal and social rhythm therapy reduced the risk of relapse, but the results were imprecise (see ).
Collaborative care
Two trials of collaborative care started with euthymic participants (BAUER2006a, KESSING2013) and three trials recruited participants in an acute episode (KILBOURNE2012, KILBOURNE2008, SIMON2005).
In comparison with treatment as usual, two trials (N = 123) reported low quality evidence of a small effect favouring collaborative care in depression and mania symptoms at post-treatment, but the effect estimate was imprecise (see ). One trial (N = 234) found no difference in the risk of relapse. However, two trials (N = 572) reported moderate quality evidence suggesting collaborative care reduced the risk of hospitalisation at post-treatment. At follow-up, there was very low quality evidence from one trial suggesting a medium effect favouring collaborative care on symptoms of depression (see ).
Integrated group therapy and group drug counselling
One study (N = 61) included euthymic or depressed participants and compared integrated group therapy with group drug counselling (WEISS2009). Based on very low quality evidence, there was no conclusive evidence of difference between groups at post-treatment (see ) or follow-up (see ).
Integrated cognitive and interpersonal therapy
One trial compared a group of participants that were randomised to integrated cognitive and interpersonal therapy or treatment as usual (SCHWANNAUER2007). Participants in the intervention group could choose to follow individual or group integrated cognitive and interpersonal therapy. Outcome data were presented for the whole intervention group versus treatment as usual.
The trial reported low quality evidence of a medium effect favouring the intervention on depression symptoms at post-treatment (see ).
8.1.4. Health economics evidence
Systematic literature review
The systematic search of the economic literature undertaken for the guideline identified two eligible studies on psychological and psychosocial interventions for adults with bipolar disorder (Lam et al., 2005b; Scott et al., 2009). References to included studies and evidence tables for all economic evaluations included in the systematic literature review are provided in Appendix 32. Completed methodology checklists of the studies are provided in Appendix 31. Economic evidence profiles of studies considered during guideline development (that is, studies that fully or partly met the applicability and quality criteria) are presented in Appendix 33.
Lam and colleagues (2005a) undertook an economic analysis to assess the cost effectiveness of CBT added to TAU versus TAU alone for adult outpatients with bipolar I disorder in the UK. The analysis was conducted alongside a RCT (LAM2003). CBT consisted of 14 sessions on average for 6 months and two booster sessions for the following 6 months. TAU was defined as use of mood stabilisers at a recommended level and regular psychiatric outpatient follow-up. The analysis adopted a NHS and social care perspective. Costs included inpatient care (psychiatric and general), outpatient care, day hospitals, accident and emergency departments, community mental health care, day centres, medication, staff (psychiatrists, GPs, psychologists, social workers, counsellors, other therapists), residential care and support groups. The primary measure of outcome was the mean number of days in an acute bipolar episode per person. Clinical and resource use data were taken from the RCT; resource use data were based on self-reports and hospital records. Unit costs were derived from national sources. The study considered two time horizons, 12 and 30 months.
CBT added to TAU was significantly more effective than TAU alone over both 12 and 30 months. The mean number of days in an acute episode was 26.6 (SD 46.0) per person for CBT added to TAU and 88.4 (SD 108.9) per person for TAU alone over 12 months; over 30 months these figures became 95.3 (SD 152.1) per person for CBT added to TAU and 201.0 (SD 95.3) per person for TAU alone (p < 0.05 in both time horizons). Regarding costs, no statistically significant differences were observed between the two interventions: over 12 months, the mean cost per person was £4,383 (SD £5,264) for CBT added to TAU and £5,356 (SD £6,599) for TAU alone; over 30 months, the mean cost per person was £10,352 (SD £13,464) for CBT added to TAU and £11,724 (SD £12,061) for TAU alone (1999-2000 prices). Therefore CBT added to TAU was the dominant option, as it was significantly more effective than TAU alone and it resulted in lower total costs (it has to be noted, though, that cost differences between CBT added to TAU and TAU alone were not statistically significant). Probabilistic analysis showed that the probability of CBT added to TAU being cost effective at a zero willingness to pay per additional day free from bipolar episodes (that is, the probability of CBT added to TAU being cost-saving) was 0.85 at 12 months and 0.80 at 30 months. When the willingness to pay per additional day free from bipolar episodes was £10, the probability of CBT added to TAU being cost effective became 0.90 at 12 months and 0.85 at 30 months.
The study by Lam and colleagues (2005b) is directly applicable to the NHS and is characterised by minor limitations.
Scott and colleagues (2009) also conducted an economic analysis alongside a RCT (COLOM2003A) to assess the cost effectiveness of group psychoeducation versus unstructured group support, both added to TAU, for adults with bipolar disorder type I or II in Spain. Group psychoeducation consisted of up to 21 sessions over 6 months. TAU comprised administration of mood stabilisers. People participating in the trial had to be euthymic for at least 6 months before entering the study. The perspective of the analysis was that of the Spanish healthcare system. Costs consisted of inpatient, outpatient and emergency visit costs, costs of medication and lab testing, and costs of group and individual psychological therapy. The primary outcomes of the analysis were the percentage of people experiencing at least one relapse, the mean number of relapses per person, and the mean number of days in an acute episode per person over the time horizon of the analysis, which was 5.5 years (6 months of intervention plus 5 years’ follow-up). Effectiveness and cost data were taken from the RCT. Resource use was based on self-reports and hospital records. Unit costs were based on hospital prices and other published sources.
Group psychoeducation was significantly better than unstructured group support in two out of the three primary outcomes. Although the percentage of people experiencing at least one relapse was not statistically different between the two groups (85% versus 95%, respectively, p > 0.05), the mean number of relapses per person was significantly lower for group psychoeducation (3.86, SD 4.18) compared with unstructured group support (8.37, SD 6.02; p < 0.05); the mean number of days in acute episode was also significantly lower for group psychoeducation (154.73) compared with unstructured group support (586.45; p = 0.01). The mean cost per person was €17,582 (SD €16,395) for group psychoeducation and €20,909 (SD €17,392) for unstructured group support (p > 0.05, cost year not reported but likely 2006). Thus, group psychoeducation was the dominant option, as it was significantly more effective than unstructured group support at no extra cost.
The study by Scott and colleagues (2009) is partially applicable to the UK context as it was conducted in Spain, and is characterised by minor limitations.
Economic evidence statement
There is limited economic evidence suggesting that psychological and psychosocial interventions may be cost-effective treatment options for adults with bipolar disorder. This evidence comes from one directly applicable and one partially applicable study and is characterised by minor methodological limitations.
