Journal of Affective Disorders 174 (2015) 485–492

Contents lists available at ScienceDirect

Journal of Affective Disorders journal homepage: www.elsevier.com/locate/jad

Research report

A web-based preventive intervention program for bipolar disorder: Outcome of a 12-months randomized controlled trial$ Caryl W. Barnes a,b,n, Dusan Hadzi-Pavlovic a,b, Kay Wilhelm a,b, Philip B. Mitchell a,b a b

School of Psychiatry, University of New South Wales, Sydney, NSW, Australia Black Dog Institute, Sydney, NSW, Australia

art ic l e i nf o

a b s t r a c t

Article history: Received 16 June 2014 Received in revised form 18 November 2014 Accepted 19 November 2014 Available online 26 November 2014

Background: The Internet is used to deliver information on many psychiatric disorders such as bipolar disorder. This paper reports on the results of a 12-months randomised controlled trial, which examined the efficacy of an Internet-based preventive program for bipolar disorder, adjunctive to usual pharmacological management. Methods: Participants were recruited by completing an online screening questionnaire accessed through the Black Dog Institute and Sentiens websites based in Australia. The treatment was predominantly psycho-educational with cognitive behavioral therapy optional elements. The attention control treatment comprised directing subjects to a variety of websites focused on ‘healthy living’. Time to recurrence was determined using Kaplan–Meier survival analysis. The main outcome measures were recurrence as defined by: (i) depressive and/or hypomanic symptomatology and functional capacity (using Beck Depression Inventory, Internal State Scale and Sheehan Disability Scale) and (ii) hospitalization. Results: Two-hundred-and-thirty-three subjects were randomized to the active or control treatment groups. There were no significant differences between the active and control treatment groups on any of the definitions of recurrence. Limitations: Reliance on an online self-report tool to confirm diagnosis and hospitalization rates may have potentially allowed for inclusion of individuals with other diagnoses such as borderline personality disorder. The ‘attention control’ treatment may have included more ‘active’ components than intended. Conclusions: This is the first report examining the efficacy of a randomized controlled web-based psychological intervention in a large sample of subjects with bipolar disorder. The potential reasons for failing to demonstrate a significant difference compared to the active control are discussed. & 2014 Elsevier B.V. All rights reserved.

Keywords: Bipolar disorder RCT Relapse prevention Psychoeducation Internet

1. Introduction Bipolar disorder is a chronic episodic condition characterized by frequent recurrences in many patients. It affects up to 4% of the adult population (Kessler et al., 1994,2005; Angst, 1998;Mitchell et al., 2013) and is highly disabling (Murray and Lopez, 1996; Murray et al., 2012). In recent years the field has become increasingly aware of the limitations of pharmacological treatment with the majority of people with bipolar disorder continuing to experience recurring episodes as well as inter-episodic sub-syndromal symptoms (Coryell et al., 1993; Gitlin et al., 1995; Judd et al., 2005). Furthermore, epidemiological studies report that many bipolar

☆ Trial Registration: Australian New Zealand Clinical Trials Registry (ANZCTR); Identifier-ACTRN12608000492358 n Corresponding author at: School of Psychiatry, University of New South Wales, Black Dog Institute Building, Prince of Wales Hospital, Hospital Rd, Randwick, NSW 2031, Australia. Tel.: þ 61 2 9382 4518; fax: þ61 2 9382 8207. E-mail address: [email protected] (C.W. Barnes).

http://dx.doi.org/10.1016/j.jad.2014.11.038 0165-0327/& 2014 Elsevier B.V. All rights reserved.

disorder patients are not receiving adequate medication for their condition (Mitchell et al., 2013) often due to poor adherence. A large body of evidence now exists demonstrating the efficacy of several forms of psychological intervention adjunctive to medications in patients with bipolar disorder (Miklowitz, 2008). The Internet provides one possible path for disseminating these targeted and standardized interventions, due to its accessibility, versatility and capacity to assist individuals who do not seek help, and provides a feasible means for increasing the rate of those with bipolar disorder receiving evidence-based psychological interventions. In addition to our prior description of this current web-based study (Barnes et al., 2007), there have also been published descriptions of three other similar interventions for bipolar disorder (Todd et al., 2012; Lauder et al., 2013; Smith et al., 2011; Poole et al., 2012). Two small studies have been published, first a quantitative outcome report to date of an Internet-based intervention for bipolar disorder was a pilot study of a web-based psychoeducational intervention, which used the quality of life instrument WHOQOL-BREF as the primary outcome measure

486

C.W. Barnes et al. / Journal of Affective Disorders 174 (2015) 485–492

(Smith et al., 2011). Although no significant difference was found between intervention and control (treatment as usual) groups on the total WHOQOL-BREF score, a marginal statistical (p¼ .05) improvement was found in the psychological quality of life subsection of this measure, but this did not persist after correction for multiple testing. The authors concluded that the program had a modest positive impact on the quality of life of individuals with bipolar disorder. The trial was relatively underpowered (n ¼50) and was only administered for a short duration (4 months). However, with over two-thirds in the intervention group completing more than 75% of the program, it would appear that online delivery of such psychosocial interventions might be acceptable to this population. The second by Proudfoot et al., which examined the comparative effectiveness of an 8-weeks online psychoeducation program for people newly diagnosed with bipolar disorder. Participants were randomized to either complete program alone, with a peer-supporter or to a control ‘attentional program’. This study found increased perception of control, decreased perception of stigma as well as significant decrease in depression and anxiety across all groups, with a small clinical difference in depressive symptoms and functional between supported and unsupported groups. (Proudfoot et al., 2012). This current paper reports on Phase I of a randomized controlled trial of an Internet-based psychoeducational program adjunctive to usual treatment for individuals with bipolar disorder. The sample was followed up over 12 months to determine the extent to which recurrence was averted (this Phase II follow-up is not included in this current report). The aim of this trial was to investigate the efficacy of an Internet-based intervention in reducing rates of recurrence in patients with bipolar disorder. An overview of this current trial and development of the treatment program included description of earlier pilot study validating online Bipolar Disorder Screening Questionnaire (BDSQ) have been described elsewhere (Barnes et al., 2007).

2. Methods The study was approved by the Human Research Ethics Committee of the University of New South Wales, Sydney, Australia. Informed consent was obtained on-line from all subjects prior to participation in the trial.

active and control treatments). Furthermore, all participants needed to have a current email address, access to a computer and the Internet, and to have an adequate understanding of English. Subjects were excluded if they were not currently taking psychotropic medication, were not under the care of a mental health physician during the trial, if they indicated that the diagnosis of bipolar disorder was a self-diagnosis (without medical confirmation), and if they did not meet the above inclusion criteria. All demographic and clinical data was collected on-line; there was no face-to-face contact between research staff and subjects. The diagnosis of DSM-IV bipolar disorder was determined by an algorithm using DSM-IV depressive and hypo/manic symptoms which were elicited by specifically developed online survey, the Bipolar Disorder Screening Questionnaire (BDSQ). This used questions similar to those used in the SCID (adapted as required for the utility of on-line usage) and was validated in an earlier pilot study which compared results BDSQ against clinician rated diagnosis with robust inter-rater reliability (Barnes et al., 2007). The comorbid presence of a range of anxiety and alcohol/substance abuse or dependence disorders was determined by subjects agreeing to a glossary-type description of each of these conditions, rather than specifying the presence of detailed operationalized DSM-IV criteria. Participants who met study inclusion criteria after completing the baseline questionnaires were sent an emailed invitation to enroll in the trial. Enrollment involved returning to a webpage via a link in the email to complete an online enrollment form. After submitting this enrollment form, participants were automatically randomized to either the study or control group. Randomization was performed using time sequencing software with each individual being allocated a unique eight digit number which then became their personal registration number for the study. An automatic password was also generated; this and the registration number were automatically emailed back to the participant with logon details on how to access their program. Although it was not possible to blind participants once they had commenced the study, subjects did not know to which program they had been randomized until logging onto their first session.

4. Study and control interventions 3. Study design This 12-months randomized controlled trial derived from of a research partnership between the University of New South Wales and Black Dog Institute in Sydney with a commercial e-health provider (Sentiens Pty Ltd) in Perth, Western Australia. For inclusion, participants needed to be at least 18 years of age and meet DSM-IV criteria for Bipolar I or II disorder. To examine the generalizability of this intervention to ‘real world’ bipolar disorder samples accessing the Internet, we pragmatically allowed for inclusion of those with current DSM-IV depressive or hypomanic episodes in addition to those with euthymia (defined as not fulfilling current DSM-IV mood episode criteria; see below for the approach for determining recurrence in those who were in an episode at baseline). For the same reason, subjects with current or past alcohol or drug abuse and those who met criteria for a current DSM-IV anxiety disorder were not excluded. All participants were required to be taking medication for bipolar disorder, previously had a clinical diagnosis of bipolar disorder confirmed by a mental health professional, and to be under the continuing care of a clinician. Subjects were also required to have had at least one prior episode (depressive or hypo/manic) in the preceding 2 years (to ensure a sufficient degree of active recurrent illness to demonstrate a difference between the

The study intervention – the Internet-based psychoeducational ‘Recovery Road for Bipolar Disorder’ program (now called ‘HealthSteps for Bipolar Disorder’) was adapted for use in the current study, and an ‘attention control’ program – ‘Virtual Highway for Bipolar Disorder’ – developed using the same web platform. The research partnership with Sentiens enabled all participants to have free access to the ‘active’ program or control to which they were randomized for 12 months. Control participants were then offered access to the ‘active’ program and ‘Recovery Road’ program after completion of the randomized controlled phase of the trial. The study and control Internet based programs both consisted of 20 sessions delivered automatically and sequentially over fixed intervals during the 12-months period. The first 8 sessions were released weekly, sessions 9 and 10 at 2-weeks interval, and sessions11–20 monthly. This scheduling was used by all of Sentiens other online programs and as a result could not be altered for this study. During each session the study group was given educational text-based material about bipolar disorder structured around five areas: ‘Dealing with Symptoms’; ‘Medication Issues’; ‘Psychological Approaches’; ‘Lifestyle and Relationships’; and ‘Staying Safe’, with the content for each area differing between sessions. It was estimated that completion of all outcome measures (which were done at start of each of 20 sessions) and in-built

C.W. Barnes et al. / Journal of Affective Disorders 174 (2015) 485–492

Recovery Road survey's and review of reading material would take average participant around 30–45 min per session. The study group also had access to ten ‘sessions’ of cognitive behavioral therapy provided as optional stand-alone homework sessions, which helped participants to identify early warning signs of a depressive or manic episode. These printable worksheets covered typical cognitive, emotional and behavioral changes for each type of bipolar episode as well as when in a ‘normal’ mood state. (It was not possible to track usage of this part of the program, as the underlying platform was not enabled for page tracking). These sessions also incorporated printable worksheets i.e. a weekly Mood Diaries with symptom checklists, Weekly Activity Diary, Thought Diaries The main interactive element of the study program was mood monitoring graphics and feedback based on participants answers to mood-monitoring questionnaires that tracked depressive, hypomanic symptoms as well as side effects and generally functioning. This feedback was text based using automated text responses based on symptom severity, encouraging participant to read text-based materials and to contact mental health professional if scoring in severe range on symptoms. For the ‘attention control’ control group, each session would feature a different ‘healthy lifestyle’ website focused on such topics as diet, meditation or exercise. Control participants had no access to specific bipolar disorder psychoeducational material, cognitive behavioral therapy or mood monitoring or automated feedback. For a fuller description of study and control programs earlier published paper on trial methodology (Barnes et al., 2007). Both control and treatment groups were supported throughout the trial through a ‘Case Management’ system. This included series of three automated email reminders (tagged as ‘session due to start’, ‘session started’, ‘session due to close’) to logon on at the scheduled times followed by one phone call follow up by a research assistance if participant had not logged in and a final automated email ‘session closed’ after session window of 7 days. No other phone call contact was made. This also included a ‘Red Flag’ system, which was triggered if a participant, reported significant symptoms based on outcomes measures (BDI-II and ISS). A BDI-II total score of 30 or more, and/or a positive reply to a question enquiring about suicidal plans triggered a depression ‘Red Flag’. Reaching cut off scores within the ISS subscales triggered a hypomania ‘Red Flag’. At the commencement of the study, participants were asked to consent to either a ‘Red Flag email’ being sent to them with the recommendation that they consult their mental health professional or to allow the research team to also send an email or letter to their treating mental health professional (mostly commonly a GP) directly. IT support was provided for all participants throughout the trial via [email protected] email only, no phone call support was available. It was made explicit that no clinical advice could be given and was for technical difficulties only. Various scripts were written prior to the start of the study to maintain consistency in response. The research officers liaised weekly with the main author to maintain this standardised response and to deal with any atypical email

487

contact. This was challenging at times, as participants would often have several queries within one email interspersed with significant amounts of self-disclosure. Care was taken to word such emails in an empathic way but to redirect participants to their normal mental health professionals for any clinical concerns.

5. Outcome measures The main primary outcome was time to first recurrence as defined by scores on: (i) the Beck Depression Inventory-II; (ii) the mood state discriminator function of the Internal State Scale (which assesses hypo/mania, depression and mixed states); and (iii) self-reported hospitalization. Symptom severity and functional impairment measures were adapted for on-line use with the permission of each scale author. A further primary outcome was impact on functioning as measured by the Sheehan Disability Scale. To ensure clinically meaningful outcome measures, which incorporated both symptom severity and functional impairment, we developed composite primary outcome measures based on these (see details below). Secondary outcomes based on compliance with medication as described in Australian New Zealand Clinical Trials Registry (ANZCTR) could not be measured due to substantial missing reported treatment data. The Beck Depression Inventory-II (BDI-II) (Beck et al., 1961) is a 21-item self-report scale developed to measure changes in subjective severity of depressive symptoms in adults over the preceding two weeks. The scale is summed with four ranges providing categorical indices of depression severity and indicates the presence or absence of a major depressive episode. The Internal State Scale (ISS) (Bauer et al., 1991) assesses mood symptoms over the preceding 24 h; it is comprised of 15 items and four principle components/subscales: ‘Activation’, ‘Depression Index’, ‘Well Being’ and ‘Perceived Conflict’. The two subscales, ‘Activation’ and ‘Well-Being’ can be used in combination using an algorithm to discriminate various mood states in bipolar disorder. Three of the subscales may also be used as a measure of self-report symptom severity of such episodes and can be used over time as repeated measures. The Sheehan Disability Scale (Sheehan et al., 1996) is a 5 item measure of impairment in function. The first 3 items can be summed to give a total score that reflects impairment in functioning across three domains: work/school, Social life and Family Life/ Home responsibilities. These items are presented as a visual analog scale of 0–10. The summing of this scale provides three cut-off points related to mild impairment (score 1–3), moderate impairment (score 4–6) and severe impairment (score 7–10). We used the DSM-IV definition of ‘recurrence’ in this study, i.e. the return of significant symptoms after a remission of at least 8 weeks (Judd et al., 2005). As detailed above, we defined symptomatic recurrence – for depressive, hypo/manic and mixed episodes – using combinations of the self-report measures (BDI-II, ISS and Sheehan) such that there was a gradation from low

Table 1 Combinations of outcome measures and cut off scores used to create three symptomatic and functional thresholds to define ‘recurrence’. Level of threshold used to define recurrence

(Hypo)Manic episode

Mixed episode

Depressive episode

Low threshold Medium threshold High threshold

ISS M ISS Mþ SDS 4–6 ISS Mþ SDS Z 7

ISS Mi ISS Mi þSDS 4–6 ISS Mi þSDS Z 7

ISS D ISS Dþ BDI 19–29þSDS 4–6 ISS Dþ BDI Z 30þ SDS Z7

ISS¼ Internal Stage scale M ¼Mania; ISS Subscales: Activation Z 155 þWell-Being Z125 Mi ¼Mixed; ISS Subscales: Activation Z 155þ Well-Being o 125 D¼ Depressed; ISS Subscales: Activation o155þ Well-Being o 125 BDI ¼Beck Depression Inventory-II SDS ¼ Sheehan Disability Scale

488

C.W. Barnes et al. / Journal of Affective Disorders 174 (2015) 485–492

threshold (to capture low symptom severity and low functional impairment) to high threshold (to capture high symptom severity and high functional impairment) (See Table 1.) This use of a gradation of definitions of recurrence based on severity was undertaken to ensure that arbitrary low or high thresholds did not distort any apparent outcomes. The higher threshold definitions incorporated increased levels of functional impairment as reported by the Sheehan Scale. Kaplan—Meier survival analyses were performed to explore for differences between treatment groups for recurrence into ‘any’ bipolar disorder episode or specific mood episodes (depressive, hypo/manic, mixed) as defined above, or hospitalization.

sequentially into a Cox regression. Survival function curves were produced for the “treatment group” variable and for significant covariates. Analyses were repeated for time to relapse into ‘any’ episode and depression, hypomania/mania or mixed episode. As is often the case with correlated covariates, not all covariates significant in the univariate models remained significant when included in the multivariate models. When that occurred, a forward stepwise (conditional Logistic Regression (LR) criterion) model was fitted, in order to clarify the relative importance of the variables. No participant was withdrawn if they missed some assessment points at a particular time.

7. Results 6. Statistical methods The statistical software package SSPS v17 was used to perform analyses of the baseline data. Study and control groups were examined for any differences between demographic (age and gender), current and historical clinical variables. Linear regression was performed to explore the relationship between a dependent continuous variable and one or more independent variables, i.e. mean current age, age onset mood episode and mean age at diagnosis. A series of binary logistic regressions were performed when appropriate to further explore the impact of set of predictors (mean current age, mean age onset of mood episodes and mean age of diagnosis) on a dependent categorical variable (e.g. presence of co-morbid anxiety disorder). One-way between group analyses of covariance (ANCOVA) were performed to explore differences between groups whilst statistically controlling for an additional (continuous) variable. Two-way between group analysis of covariance (ANCOVA) were performed when it was felt that that an additional variable or moderator was influencing the effect of other independent variables. Kaplan–Meier survival analysis was repeated using each set of criteria for defining ‘recurrence’ as detailed previously. A series of univariate Cox regressions were performed using the covariates with only those with a significance level r.1 retained entered

Of the 559 individuals who logged on to the study website, 374 completed the baseline assessment. Seventy-four subjects were excluded, as they did not fulfill the study inclusion criteria. Of the 300 who satisfied the inclusion criteria and were invited via email to enrol in the study, 233 (77.7%) went on to complete the enrollment and consent process, and thereby comprised the intention to treat (ITT) sample. These subjects were then randomised into the study (n ¼113) and control (n ¼ 120) groups. (See CONSORT diagram Fig. 1) The subject mean age was 39.0 710.8 (SD) years. Seventy-two percent (n¼ 168) were female. Educational and marital status was reported by 154 subjects: eighty-four (55%) were married or in a de-facto relationship, fifty-one (33%) were single and nineteen (12%) divorced. Forty percent were in full-time employment (n ¼62), thirty-five percent were unemployed or retired (n ¼54) and twenty-five percent were in part-time work or studying (n ¼38). Regarding the highest educational level attained, fifteen (10%) had gained an advanced diploma/diploma and seventy-eight (51%) had obtained a university degree. The majority of subjects met DSM-IV criteria for Bipolar I disorder (87.6%, n ¼204) with the remainder (n ¼29, 12.4%) meeting criteria for Bipolar II Disorder. The mean age of onset of bipolar disorder was 18.8 78.0 (SD) years, with the average self-reported

Fig. 1. CONSORT diagram.

C.W. Barnes et al. / Journal of Affective Disorders 174 (2015) 485–492

489

Table 2 Comparison of demographic and clinical characteristics of study and control groups at baseline. N ¼ 233 (%)

Gender Age Diagnosis Onset Course Admission history

Current status (DSM-IV)

Co-morbidity

Location

a

Female Male Mean Age BPAD I BPAD II Mean age onset mood swings Mean age diagnosis Rapid cycling Mixed episodes Mean hospitalized mania Mean hospitalized depression Mean no. times admitted mania Mean no.times admitted depression Euthymic Current mixed episode Current depressive episode Current hypo/manic episode Mean no. co-morbid anxiety disorders Panic disorder Social phobia GAD Agoraphobia Specific phobia PTSD OCD Past drug abuse Current drug abuse Past alcohol abuse Current alcohol abuse Australia/New Zealand North America South America Europe Asia

Study group 113(48.5)

Control group 120(51.5)

83(73.5) 30(26.5) 40.9a 100(88.5) 13(11.5) 18.8(7.5) 34.8(10.3) 60 59 49 59 2.6(n¼56) 5.8(n¼64) 24(22.4) 19(17.8) 46(43.0) 18(16.8) 1.9(1.6) 1(.9) 49(43.4) 65(57.5) 42(37.2) 18(15.9) 27(23.9) 14(12.4) 1(.9) 13(11.5) 9(8.0.) 28(24.8) 96(85) 8(7.1) 2(1.8) 5(4.4) 1(.9)

85(70.8) 35 (29.2) 37.6 (9.9)a 104(86.7) 16(13.3) 18.7(8.5) 32.6(10.1) 63 60 43 52 2.1(n¼ 51) 4.4(n¼63) 46(21.5) 40(18.7) 90(42.1) 20(18.7) 1.7(1.5) 3(2.5) 47(39.2) 61(50.8) 35(29.2) 13(10.8) 31(25.9) 17(14.2) 4(3.3) 17(14.2) 4(3.4) 28(23.4) 100(83.3) 17(14.2) 0 3(2.5) 0

p o .05

age at diagnosis being 33.7710.3 (SD) years. More than one half (53%) of subjects reported a history of rapid cycling and 52% (n ¼119) mixed episodes. At the commencement of the study, only 31% were euthymic; the majority were in a depressive episode (28.3%), 18% were hypomanic and 22.7% met criteria for current mixed episode. Forty percent (n ¼92) had been previously admitted for a manic episode and 47% (n ¼109) had been hospitalized for depression. Overall, the group reported high levels of co-morbid anxiety disorders: generalized anxiety disorder (GAD; 54%, n ¼126), social phobia (41%, n ¼96) agoraphobia (33%, n ¼71) and post-traumatic stress disorder (PTSD; 25%, n¼ 58). Baseline demographic and clinical characteristics of those subjects randomized to the two treatment groups are detailed in Table 2. The only significant difference between groups was for age: study group 40.9 years, control group 37.6 (p o.05). Two-hundred-and-fourteen participants completed the first treatment session and no significant between-group differences were found on mood state as defined by the ISS or BDI-II. Only 22.4% (n ¼24) of the active treatment group and 22% (n ¼22) of the control group were euthymic at Session 1 as defined by the ISS, with the majority of both groups being in a depressive episode as defined by this measure (study group 43%, n ¼46; control group 41.1%, n ¼44). Similarly, no significant difference was seen between the groups in terms of BDI-II scores. The mean total scores on the BDI-II were: active treatment group 21.5 (SD 13.9); control group 21.8 (SD 13.6) indicating that for both groups the majority had depression scores in the moderate to severe range. Furthermore, there were no significant differences between the two groups on the Sheehan Scale. The active treatment group had

a mean score of 15.6 (SD 9.4; n ¼107) and the control group 16.2 (SD 8.8; n ¼107). Both groups had mean scores greater than 7, indicating severe functional impairment. The completion rate at the end of the 12-months RCT was 75% (n ¼85) for the study group and 69% (n ¼83) for the control group. This is the number of participants who completed the majority of the 20 sessions and logged on and completed the final assessment point at 12 months after initial enrollment. A large number of participants never reached criteria for achieving ‘wellness’ (as defined by the three symptomatic and functional severity thresholds as defined in Table 1) throughout the 12-months study. These individuals were therefore excluded from the survival analyses. When the low threshold for definition of recurrence was used, 54.9% (n ¼128) were considered ‘never well’ compared to 30% (n ¼70) for the medium threshold and 20.2% (n ¼47) for the high threshold. The main outcome measure was time to recurrence. As the symptomatic and functional threshold for ‘recurrence’ increased, as would be expected, the mean and median survival times to recurrence for any mood episode increased. For the ‘low’ symptomatic and functional definition of recurrence, the mean survival time was 15.6 weeks (95% CI 12.6–18.6). As detailed in Table 3, the between-group hazard ratio for recurrence of ‘any’ subtype of bipolar episode using this low threshold was .86 (95% CI .56–1.3; p¼ .48). The mean survival time using the ‘medium’ threshold definition was 27.0 weeks (95% CI 3.9–30.2) with between-group hazard ratio for recurrence of ‘any’ subtype of bipolar episode using this medium threshold being .82 (95% CI .55–1.22; p¼ .33). When the ‘high’ threshold to define recurrence was used, the

490

C.W. Barnes et al. / Journal of Affective Disorders 174 (2015) 485–492

Table 3 Between-group hazard ratios for each definition of ‘recurrence’ by type of bipolar disorder episode. Type of bipolar disorder recurrence

Low threshold for recurrence Medium threshold for recurrence High threshold for recurrence Admission to hospital

Any

Mixed

Depression

Hypo/mania

.86(.56–1.3) p¼ .48 .82(.55–1.22) p¼ .33 .91(.59–1.39) p¼ .65 1.05(.47–2.34) p ¼ .90

# .88(.37–2.08) p ¼.77 1.17(.58–2.34) p ¼ .67 n/a

.82(.50–1.3) p ¼.43 .82(.46–1.7) p ¼.49 .70(.34, 1.46) p ¼ .35

.90(.35–2.34) p ¼ .84 .78(.38–1.6) p¼ .51 .88(.38–2.04) p¼ .76

#only 3 events (mixed episode relapse) recorded over 12 months, Survival analysis not performed.

mean survival time was 32.8 weeks (30.1–35.8) and the between group hazard ratio for ‘any’ type of bipolar episode recurrence was .91 (95% CI .59–1.39; p¼ .65). No significant differences in survival times were seen between the active treatment and control groups for ‘any mood’ type of recurrence, or specifically for depressive or hypo/manic recurrences, using all three threshold definitions of recurrence. Too few individuals experienced recurrence into a mixed episode (n ¼3) during the 12 months for survival analysis to be performed. Finally, when hospitalization was used as a measure of recurrence, 10.3% (n¼ 24) of all subjects had a recurrence using this definition, with the mean survival time to ‘any’ bipolar episode being 48.0 weeks (95% CI 46.8–49.3). There was no difference between the study and control groups on this measure, with the between-group hazard ratio being 1.05 (95% CI .47–2.34; p ¼.90) (also see Table 3).

8. Discussion This is the first report formally examining the efficacy of a randomized controlled web-based psychological intervention – adjunctive to usual medications – in a large sample of subjects with bipolar disorder. We found no evidence for a greater treatment effect of this on-line psycho-educational and cognitive therapy treatment program compared to ‘attention control’ therapy arm. This is in contrast to published meta-analyses that have demonstrated that face-to-face psychological interventions (cognitive-behavioral, psychoeducational and interpersonal) are more effective than control treatments in reducing rates of recurrence (Miklowitz, 2008). Several groups have described similar web-based interventions for bipolar disorder (Todd et al., 2012; Smith et al., 2011; Lauder et al., 2013). Only the intervention developed by Smith et al. (2011),(Poole et al., 2012) has been evaluated in terms of formal outcome measures – albeit only in a pilot study – reporting no significant difference in the primary (the WHOQOL-BREF total score) or secondary outcome measures, with the only positive result being some modest improvement on a subsection of the primary outcome measure Possible reasons for the failure of our current study to demonstrate a significant between-group difference include: (i) a limited effect of the active study intervention, (ii) a possible active effect of the supposed ‘control’ intervention, (iii) the high morbidity of the study population (particularly with respect to depressive symptoms), (iv) the limitations of the outcome measures employed, (v) the heterogeneity of the participants due to the intentional broad inclusion criteria, and (vi) difficulties in creating sub-groups receiving the active intervention based on use of extra components of the program, namely CBT sessions. These possibilities will be discussed in turn below. Were there sufficient differences in potentially active components between the study and control groups? The study group received evidence-based psychoeducational sessions with some

cognitive behavioral components, while the control group was merely directed to ‘attention placebo’ healthy lifestyle and similar websites. It was important to create a credible ‘control’ program which would ensure ongoing partcipation for the duration of 12 months. Although healthy lifestyle interventions are an important part of wellbeing for individuals with bipolar disorder, both groups had access to such information freely outside of the study. It is possible however that enclsuon of such a focus in ‘control’ group did make this more of an ‘active’ interevtion and thereby limit the ability to distingush difference between groups. Both programs contained a shared ‘mood monitoring’ feature. The ‘study’ program provided feedback (in terms of graphs and reports) which the control group did not receive. It may have been possible, however, that the process of regularly answering a battery of scales on mood/functioning may have been sufficient for the control group to learn to identify changes in mood which led to help seeking or behavioral change. This combined with the elements of ‘Case Management’ (which both groups received) may have become unintended ‘active’ interventions in their own right. Scott et al. (2006) have postulated that the phase of the illness rather than the intervention itself is the strongest predictor of outcome. That group found that only subjects who had been euthymic for more than a year prior to the initation of the psychologcial intervention had significantly lower risk of relapse than those recieving treatment as usual(Scott et al., 2006). Could the active study program have been made ‘more active’? We deliberately constructed the active treatment program such that there was no clinician input, apart from the Case Management incorporated to ensure safety for those with significant suicidal ideation. This lack of clinical input was deliberately undertaken to develop an Internet-based intervention which could be potentially and widely disseminated in the future without the constraining expense of clinical involvement. However, it has been reported from internet-based interventions for other mental illnesses, such as major depressive disorder, that clinician involvement may be an important factor in determining outcome (Palmqvist et al., 2007; Titov et al., 2008). A number of meta-analyses and reviews of face-to-face psychological interventions (Miklowitz, 2008; Scott et al., 2006) for bipolar disorder have led to recommendations concerning the specific phases of bipolar disorder at which an intervention should be employed, and which type of interventions are most effective during particular episode types. The content of the active treatment group in our study was predominantly psychoeducational, with a limited CBT component. As has been discussed above, the study sample had a depressive predominance. The bipolar disorder psychological treatment intervention literature would perhaps suggest a greater likelihood of the benefit of a cognitive behavioral and interpersonal coping strategies rather than a psychoeducational focus for such a depressive population. (Miklowitz, 2008) Often when a novel treatment appears, it is often the chronic, severe and/or treatment refractory individuals who are attracted to such novel therapies. Was this the case in this group? Our study population was characterized by high rates of: prior admissions for

C.W. Barnes et al. / Journal of Affective Disorders 174 (2015) 485–492

mania or depression; high rates of baseline depression; and prevalent comorbidity with anxiety and/or substance use disorders – all suggesting that this Internet-recruited sample group was relatively clinically severe, thereby potentially limiting our capacity to demonstrate an effect of any ‘active’ intervention.

9. Limitations There were several methodological limitations of our study design that need to be acknowledged. First, the inclusion of depressed and hypomanic subjects at baseline may have made it more difficult to demonstrate an effect using survival analysis; there were relatively few euthymic at baseline (only 30–40%) and only a minority of those originally unwell were to later remit then relapse. This may have led to reduced statistical power to demonstrate any ‘true’ differences. The proportion that were ‘never well’ was 20.2–54.9%, depending on criteria used to define recurrence. Although not assessed, it is possible that such symptom burden may have impacted on cognitive capacity, which in turn may have influenced the potential benefits of the intervention. Despite this, however, there were not even any trends towards statistical significance between groups, suggesting that there was unlikely to have been any true differences in effect. The high levels of missing treatment data (therefore leading to an inability to examine treatment intervention as a measure of recurrence), we were dependent upon both rates of hospitalization and selfreports of bipolar symptoms and functional capacity as indices of recurrence. Despite the limitations of the self-report measures, these were consistent with the lack of differences in rates of hospitalization (with about 10% of the total sample being hospitalized during the course of the study). The study design of relying on self-reported symptoms to establish the diagnosis of bipolar diagnosis may have led to the inclusion of participants whose primary diagnosis was other than bipolar disorder, e.g. borderline personality disorder. Though this issue is an inherent limitation for on-line studies of this condition, we attempted to minimize the potential for such diagnostic error by means of the screening tool pilot validation study, and also excluding those who indicated self-diagnosis without medical confirmation of bipolar disorder. We would note that both underand over-diagnosis occur even after extensive face-to-face assessments (Ruggero et al., 2011; Mitchell, 2012)

10. Conclusions As stated above, this paper comprises the first report formally examining the efficacy of a randomized controlled web-based psychological intervention in a large sample of subjects with bipolar disorder. While the finding of this particular trial was negative, there is no doubt that this form of treatment delivery will continue to grow in importance in mental health service provision due to the accessibility and utility of the Internet. Future studies need to consider both the strengths and limitations of current web-based interventions for bipolar disorder, such as that described in this report.

Role of funding source This research was supported by the University of New South Wales, Faculty of Medicine PhD Scholarship for Dr Barnes and a program grant from the Australian National Health and Medical Research Council (No. 1037196).

Conflict of interest Dr Barnes received no benefit (financial or otherwise) from Sentiens – the commercial provider of the Internet-based program used in this trial – during or

491

after this investigation. Dr Barnes has not accepted remuneration from pharmaceutical companies for over 3 years, and has not been a member of an industry advisory committee in that time. Dr Mitchell has not accepted remuneration from pharmaceutical companies for over 5 years, and has not been a member of an industry advisory committee in that time. Mr Hadzi-Pavlovic and Dr Wilhelm have no conflicts of interest and no disclosures. This research was supported by a University of New South Wales Faculty of Medicine PhD scholarship for Dr Barnes and a program grant from the Australian National Health and Medical Research Council (No. 1037196). Preliminary data was also presented in a poster as part of proceedings for 7th International Conference on Bipolar Disorders in Pittsburgh in 2007. An overview of the aims and methodology of the study was published as a paper in Disease Management and Health Outcomes in 2007 (Barnes et al., 2007).

Acknowledgments We thank the research staff at Sentiens; in particular Michael Smith, Gavin Pinto, Stacey Bosley and Dr Robyn Harvey who were allowed to assist in the development and management of this trial without cost to the research team. We also thank the patients who participated in this study.

References Angst, J., 1998. The emerging epidemiology of hypomania and bipolar-II disorder. J. Affect. Disord. 50, 143–151. Barnes, C., Harvey, R., Smith, M., Mitchell, P., Wilhelm, K., 2007. Evaluation of an online relapse prevention program for bipolar disorder: an overview of the aims and methodology of a randomised controlled trial. J. Dis. Manag. Health Outcomes 15, 215–224. Bauer, M., Crits-Christoph, P., Ball, W., Dewees, E., Mcallister, T., Alahi, P., Cacciola, J., Whybrow, P., 1991. Independant assessment of manic and depressive symptoms by self-rating. Arch. Gen. Psychiatry 48, 807–812. BECK, A., WARD, C., MENDELSON, M., MOCK, J., ERBAUGH, J., 1961. An inventory measuring depression. Arch. Gen. Psychiatry 4, 561–571. Coryell, W., Scheftner, W., M, M.K., Endicott, J., Maser, J., Klerman, G., 1993. The enduring psychosocial consequences of mania and depression. Am. J. Psychiatry 150, 720–727. Gitlin, M., Swendsen, J., Heller, T., Hammen, C., 1995. Relapse and impairment in bipolar disorder. Am. J. Psychiatry 152 (1653-1640). Judd, L., Akiskal, H., Schettler, P., Endicott, J., Leon, A., Solomon, D., Coryell, W., Maser, J., Keller, M., 2005. Psychosocial disability in the course of bipolar I and II disorders: a prospective, comparative, longitudinal study. Arch. Gen. Psychiatry 62, 1322–1330. Kessler, R., Berglund, P., Demler, O., Jin, R., Merikangas, K., Walters, E., 2005. Lifetime prevalence and age-of-onset distributions of DSM-IV disorders in the National Comorbidity Survey Replication. Arch. Gen. Psychiatry 62, 593–602. Kessler, R., Mcgonagle, K., Zhoa, S., 1994. Lifetime and 12-months prevalence of DSM-III-R psychiatric disorders in the United States: results from the National Co-Morbidity Survey. Arch. Gen. Psychiatry 51, 8–19. Lauder, S., Chester, A., Castle, D., Dodd, S., Berk, L., Klein, B., Austin, D., Gilbert, M., Chamberlain, J., Murray, G., White, C., Piterman, L., Berk, M., 2013. Development of an online intervention for bipolar disorder. Psychol. Health Med. 18, 155–165. Miklowitz, D., 2008. Adjunctive psychotherapy for bipolar disorder: state of the evidence. Am. J. Psychiatry, 1–12. Mitchell, P., 2012. Bipolar disorder: the shift to overdiagnosis. Can. J. Psychiatry 57, 659–665. Mitchell, P., Johnston, A.K., Frankland, A., Slade, T., Green, M.J., Roberts, G., Wright, A., Corry, J., Hadzi-Pavlovic, D., 2013. Bipolar disorder in a national survey using the World Mental Health Version of the Composite International Diagnostic Interview: the impact of differing diagnostic algorithms. Acta Psychiatr. Scand. 127, 381–393. Murray, C.J.L., Lopez, A., 1996. The Global Burden of Disease: The World Health Organization. Harvard University Press, Cambridge, MA. Murray, C.L., Vos, T., Lozano, R., Naghavi, M., Flaxman, A., Michaud, C., Ezzati, M., Shibuya, K., Salomon, J.A., Abdalla, S., 2012. Disability-adjusted life years (DALYs) for 291 diseases and injuries in 21 regions, 1990–2010: a systematic analysis for the Global Burden of Disease Study 2010. Lancet 380, 2197–2223. Palmqvist, B., Carlbring, P., Andersson, G., 2007. Internet-delivered treatments with or without therapist input: does the therapist factor have implications for efficacy and cost? Expert Rev. Pharmacoecon. Outcomes Res. 7, 291–297. Poole, R., Simpson, S., Smith, D., 2012. Internet-based psychoeducation for bipolar disorder: a qualitative analysis of feasibility, acceptability and impact. BMC Psychiatry, 12. Proudfoot, J., Parker, G., Manicavasagar, V., Hadzi-Pavlovic, D., Whitton, A., Nicholas, J., Smith, M., Burckhardt, R., 2012. Effects of adjunctive peer support on perceptions of illness control and understanding in an online psychoeducation program for bipolar disorder: a randomised controlled trial. J. Affect. Disord. 142, 98–105. Ruggero, C.J., Zimmerman, M., Chelminski, I., Young, D., 2011. Borderline personality disorder and the misdiagnosis of bipolar disorder. J. Psychiatr. Res. 44, 405–408. Scott, J., Colom, F., Vieta, E., 2006. A meta-analysis of relaspe rates with adjunctive psychological therapies compared to usual psychiatric treatment for bipolar disorders. Int. J. Neuropsychopharmacol., 1–7.

492

C.W. Barnes et al. / Journal of Affective Disorders 174 (2015) 485–492

Sheehan, D., Harnett-Sheehan, K., Raj, B., 1996. A measurement of disability. Int. Clin. Psychopharmacol. 11, 89–95. Smith, D., Griffths, E., Poole, R., Florio, A.D., Barnes, E., Kelly, M., Craddock, N., Hood, K., Simpson, S., 2011. Beating Bipolar: exploratory trial of a nove; internet-based psycho educational treatment for bipolar disorder. Bipolar Disord. 13, 571–577. Titov, N., Andrews, G., Schwencke, G., Choi, I., 2008. Results of three randomized controlled trials of internet based treatment programs for social phobia: The

SHYNESS Program. In: Joyce, P. (Ed.), The Australian Society for Psychiatric Research (ASPR), A76. Informa Healthcare, Newcastle (December 2008). Todd, N., Solis-Trapala, I., Jones, S., Lobban, F., 2012. An online randomised controlled trial to assess the feasibility, acceptability and potential effectiveness of ‘Living with Bipolar’: a web-based self-management intervention for bipolar disorder: trial design and protocol. Contemp. Clin. Trials 33, 679–688.