Journal of Affective Disorders 170 (2015) 78–84

Contents lists available at ScienceDirect

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

Research report

Cognitive behaviour therapy via the internet for depression: A useful strategy to reduce suicidal ideation Louise Mewton a,b,n, Gavin Andrews a,b a b

School of Psychiatry, University of New South Wales, Sydney, NSW, Australia Clinical Research Unit for Anxiety and Depression (CRUfAD), St. Vincent's Hospital, Sydney, NSW, Australia

art ic l e i nf o

a b s t r a c t

Article history: Received 1 May 2014 Received in revised form 25 August 2014 Accepted 25 August 2014 Available online 1 September 2014

Background: Depression is a major risk factor for suicide. Given the strong association between depression and suicide, treatment for depression should be a fundamental component of suicide prevention. Currently it is not. This study aims to demonstrate the usefulness of internet-delivered cognitive behavioural therapy (iCBT) for depression as a means of reducing suicide ideation. Methods: The sample comprised 484 patients who were prescribed iCBT for depression by their primary care physician. The outcomes of interest were major depression, as indexed by the PHQ-8, and suicidal ideation as measured by question 9 of the PHQ-9. Marginal models were used to appropriately analyse available data without biasing parameter estimates. Results: Following iCBT for depression, suicidal ideation and depression decreased in parallel over time. The prevalence of suicidal ideation reduced from 50% at baseline to 27% after treatment, whilst the prevalence of major depression reduced from 70% to 30%. Depression scores and suicidal ideation decreased after treatment regardless of demographic or clinical variables of interest. Limitations: This is a naturalistic study; randomisation and scientific control were not possible. Conclusions: The current study demonstrates the usefulness of iCBT for depression as a means of reducing suicidal ideation which can be implemented on a large scale without enacting major structural change at the societal level. These findings need to be replicated in randomised controlled trials. & 2014 Elsevier B.V. All rights reserved.

Keywords: Major depression Suicide Internet cognitive behavioural therapy

1. Introduction Depression is a major risk factor for suicide and related behaviours (Nock et al., 2008; Oquendo et al., 2004). Depression is implicated in 59–87% of suicides (Rihmer, 2001), and is associated with a 20-fold increase in the risk of suicide (Osby et al., 2001). Population attributable risk estimates indicate that if depression was eliminated, there would be 50–80% reduction in serious suicide attempts (Beautrais et al., 1996), similar to population attributable risk estimates between smoking and lung cancer (Goldney et al., 2003; Lilienfeld and Stolley, 1994). Whilst such a high attributable risk of depression for suicide has only been established in industrialised societies, and differs by the outcome in question (suicidal ideation, suicide attempts or completed suicide) (Pitman et al., 2012; Goldney et al., 2003), it remains clear that depression is the major risk factor for suicide and related behaviours. Given the strong association between depression and n Corresponding author at: Clinical Research Unit for Anxiety and Depression, Level 4, O’Brien Centre, St. Vincent's Hospital, 394–404 Victoria Street, Darlinghurst, NSW, Australia. Tel.: þ 61 2 8382 1437. E-mail address: [email protected] (L. Mewton).

http://dx.doi.org/10.1016/j.jad.2014.08.038 0165-0327/& 2014 Elsevier B.V. All rights reserved.

suicide, treatment for depression should be a fundamental component of suicide prevention (Mann et al., 2005). However, recent reviews of suicide prevention strategies have not focused on treatment for depression as a practical solution for suicide reduction (Yip et al., 2012; Hawton et al., 2012; Pitman et al., 2012). National public awareness campaigns targeting depression, as well as suicide specifically, have not demonstrated any detectable effect on the rates of suicidal acts (Mann et al., 2005). The question remains as to whether treatment for depression could be more successful. Internet cognitive behavioural therapy (iCBT) for depression has been shown to be efficacious, acceptable, scalable and cost effective (Andrews et al., 2010), and is therefore a candidate for suicide prevention at national level. Clinical trials of iCBT for depression frequently exclude, for methodological and ethical reasons, those who are actively suicidal (Zimmerman et al., 2005; Van der lem et al., 2011). As a result, it has been difficult to investigate the effects of iCBT for depression on suicidality under these restrictions. Previous studies have shown that iCBT for depression reduces suicidal ideation in individuals reporting high levels of psychological distress (Christensen et al., 2013), and that iCBT programs focusing specifically on suicidal ideation are also effective in reducing thoughts of suicide (van spijker et al., 2014).

L. Mewton, G. Andrews / Journal of Affective Disorders 170 (2015) 78–84

We recently conducted a clinical audit of 299 patients which, unlike previous studies, focused on those who were depressed and therefore completed an iCBT course for depression. Depression and suicidal ideation decreased significantly (Watts et al., 2012). We concluded that the continued exclusion of people from trials on the grounds of suicidal ideation was no longer defensible. The current study aims to replicate these findings in a larger sample of depressed individuals commencing iCBT for depression in primary care. In order to reduce the risk of suicidal ideation, systemic change at the societal level has often been advocated (Pitman et al., 2012; Pitman and Caine, 2012; Yip et al., 2012; Hawton et al., 2012). This study investigates iCBT in primary care as an alternative.

2. Methods 2.1. Participants The sample comprised 484 consecutive patients who were prescribed iCBT for depression (http://www.thiswayup.org.au/clinic) between October 2010 and September 2012. Patients with depression were prescribed the iCBT course by a range of primary care clinicians including general practitioners, psychologists, psychiatrists, nurses and social workers. Prescribing clinicians were advised that patients were unlikely to benefit if they were: 1) actively suicidal; 2) presenting comorbid drug or alcohol dependence, schizophrenia, or bipolar disorder; or 3) using atypical antipsychotics or benzodiazepines. Clinicians were free to ignore this advice and formal exclusion criteria were not implemented at any stage. Data were confined to the limited number of measures taken to inform practitioners about the progress of their patients. As such, limited data was collected regarding suicidality, whilst medication use and comorbid disorders and were not assessed. The depression course was prescribed by clinicians, and all clinicians were encouraged to maintain contact with their patient for the duration of the course. This study was approved as part of the quality assurance activities undertaken by the Patient Safety and Quality Unit at St. Vincent's Hospital, Sydney, with whom a copy of this manuscript has been lodged. The ongoing quality assurance studies have been approved by Dr Brett Gardiner, the Director of Clinical Governance at St Vincent's Hospital and the person to whom the Head of the Human Research Ethics Committee and the Head of the Patient Safety and Quality Unit reports. No data were collected for research purposes. Prior to enrolment in any of the treatment programs, all individuals are informed that data will be collected and used as per the following: ‘By participating in This Way Up clinic, you acknowledge that your data will be pooled, analysed and periodically published in scientific articles to enhance scientific knowledge in anxiety and depression. In any publication, information will be provided in such a way that you cannot be identified’. All patients provided electronic informed consent that their pooled data could be used for these purposes. Prior to analysis, all data was de-identified. 2.2. Intervention/procedure The iCBT course for depression consisted of six fully automated, unassisted online lessons involving components such as psychoeducation, behavioural activation, cognitive restructuring, problem solving, graded exposure, relapse prevention, and assertiveness skills. Content was presented in the form of an illustrated story in which the main character gains mastery over their symptoms of depression with the help of a clinician. The patient followed the character's journey to recovery across the six lessons. At the end of each lesson the patient downloaded “homework” tasks which reinforced the content of the lesson. The efficacy of this iCBT depression course has

79

been established previously in two registered randomized controlled trials, Cohen's d¼ 0.73, 1.20 respectively, mean number needed to treat of two (NNT¼ 2) (Perini et al., 2009; Titov et al., 2010). 2.3. Outcome measures At baseline, limited demographic information (email address, age, gender) was collected for each patient and rurality was imputed from the location of the prescribing clinician. Data was also collected on the profession of the prescribing clinician. Prior to each lesson, the patient completed the Kessler Psychological Distress Scale (K10) (Kessler et al., 2003) as a means of tracking patient progress during the iCBT course. Clinicians were automatically notified if their patient's K10 score increased by half a standard deviation or more. Before and after treatment, patients also completed the Patient Health Questionnaire (PHQ-9), a brief 9-item measure of depression severity (Kroenke et al., 2001). The nine items assess DSM-IV Criterion A for major depressive disorder (MDD). Patients rate each item in terms of the frequency of symptoms over the past two weeks, on a four point scale (0 ¼not at all, 1¼ several days, 2 ¼more than half of the days, 3 ¼nearly every day). Suicidal ideation was measured by question nine from the PHQ-9 which asks about the frequency of suicidal ideation (“thoughts that you would be better off dead, or of hurting yourself in some way”) in the previous two weeks using the above four point scale. Data for this question were dichotomised into suicidal ideation negative (those with a score of 0) and suicidal ideation positive (scores ranging from 1–3) for use in generalised estimating equations. All questionnaire data and demographic information were collected online. In order to analyse suicidal ideation and depression independently, depression was indexed by the PHQ-8 (Kroenke et al., 2009). The questions in the PHQ-8 are identical to those asked in the PHQ-9, with the exclusion of question 9, which asks about suicidal ideation. The PHQ-8 is comparable to the PHQ-9 in terms of diagnosing DSM-IV depressive disorders and has been standardised and validated in clinical and epidemiological samples (Kroenke et al., 2009; Kroenke et al., 2001). Scores can range from 0 to 24, with higher levels representing higher symptom severity. A cut off score of Z10 on the PHQ-8 has 70% sensitivity and 98% specificity for major depression (21). PHQ-8 severity cut points for MDD have been established as follows: 0–9 ¼well or sub-threshold, 10–14 ¼mild, 15–19 ¼ moderate, and 20–24¼ severe depression (Kroenke et al., 2009). 2.4. Statistical analysis The level of adherence to the iCBT programme was good (Hilvert-Bruce et al., 2012), given that levels of adherence in open access research settings are typically very low, ranging from around 1% (Christensen et al., 2004) to 10% (Klein et al., 2011). However, it was necessary to consider analytic methods that appropriately make use of all the available data without biasing parameter estimates. Univariate analyses were initially conducted to investigate the baseline demographic and clinical variables associated with nonadherence. There were no statistically significant differences between those who did and did not complete the iCBT depression course (results available upon request), indicating that it was reasonable to assume that data was missing according to a missing at random (MAR) mechanism (Kenward et al., 1994). These preliminary analyses provided support for the implementation of marginal multilevel modelling for obtaining regression estimates. Marginal models estimate the regression coefficients in repeated measures studies with unbalanced data using maximum likelihood estimation, making

80

L. Mewton, G. Andrews / Journal of Affective Disorders 170 (2015) 78–84

use of the incomplete data in a way that does not bias the parameter estimates (West et al., 2006). To investigate reductions in depression from pre- to post-treatment, a marginal linear model was implemented using the MIXED procedure with a repeated statement in SPSS Version 19. Measurement occasion (pre–post) was treated as a categorical variable and an unstructured (UN) covariance structure was specified to model the relationship amongst observations at different time points. Generalised estimating equations (GEE) were used to evaluate reductions in suicidal ideation across time. GEE is a semi-parametric method used to implement marginal models when the outcome variable is not normally distributed (Liang and Zeger, 1986). The GENLIN procedure with a repeated statement was implemented using SPSS Version 19. A binomial distribution with a logit link function was specified. An unstructured covariance structure was used to model the within-subject dependencies. To estimate the effect of iCBT for depression on suicidal ideation and depression severity, the main effect of measurement occasion (baseline and post-treatment) was first established for each outcome. For depression, the main effects of baseline covariates (sex, age, clinician's location, clinician's profession and baseline suicidal ideation) were investigated separately whilst controlling for measurement occasion and the time by covariate interaction. Estimated marginal means, which indicate the mean response for each categorical variable adjusted for any other variables in the model (i.e., time and the time by covariate interaction) were also calculated. Baseline and post-treatment suicidal ideation (dichotomised as absent or present) was also entered separately as a time varying factor. This model was used to examine whether suicidality changed in parallel with depression symptoms within individuals over time. Similar analyses were conducted with suicidal ideation (dichotomised as present or absent) as the dependent variable. The main effects of baseline covariates (sex, age, clinician’s location, clinician's profession, clinician contact and baseline depression severity) were investigated separately whilst controlling for measurement occasion and the time by covariate interaction. Interactions between suicidal ideation and each of the variables of interest were also calculated as above. The estimated probability of reporting suicidal ideation, which adjusts for all variables in the model, was also calculated. Scores on the PHQ-8 were also entered separately as a timevarying continuous variable, again to determine whether suicidal ideation and depression changed in parallel within individuals over time. In order to contextualise the results from the marginal model analysis, analyses based on last observation carried forward were also conducted to establish treatment effects using a different missing data strategy. Whilst marginal models are considered the most appropriate analytic method in pre–post designs, even when faced with substantial missing data (Salim et al., 2008), and last observation carried forward has been shown to bias parameter estimates (Siddiqui et al., 2009), we report the outcomes of both analyses to strengthen the findings. Under the last observation carried forward model, it is assumed that all patients who failed to adhere gained no benefit from the intervention, with baseline PHQ-8 and suicidal ideation scores remaining constant from pre- to post-treatment. This additional analysis was conducted to strengthen the findings and

establish a conservative, lower boundary estimate of the treatment effects on both suicidal ideation and depression.

3. Results 3.1. Baseline characteristics The sample was 60.3% female (n¼292), the mean age was 41.9 years (range 18–83 years). Almost half of the prescribing clinicians were general practitioners (45.2%; n¼ 219), whilst 29.5% were psychologists (n¼143) and 25.2% were other health professionals (n¼122). Over half of the clinicians were based in rural areas of Australia (55.6%; n¼ 269). The mean PHQ-8 score at baseline was 13.2 (range 0–24) and 28.5% were classified as sub-threshold (0–9; n¼138), 31.4% as mild (10–14; n¼152), 21.7% as moderate (14–19; n¼105) and 18.4% as severe (20–24; n¼89). At baseline, 50.4% reported no suicidal ideation in the past two weeks (n¼244), whilst 29.8% reported that they thought about suicide “several days” (n¼144), 13.2% reported they thought about suicide “more than half the days” (n¼ 64) and 6.6% reported they thought about suicide “nearly every day” (n¼ 32). Nearly 60% of the sample (58.7%) of the patients reported some clinician contact whilst undertaking the course. 3.2. Patient adherence In total, 56.8% (n¼ 275) of the patients completed the six lesson depression course and provided post-treatment data for the PHQ-9. Of those who did not complete the six lesson course, 12.2% (n¼59) completed one lesson, 7.0% (n¼34) completed a total of two lessons, 9.5% (n¼46) completed a total of three lessons, 7.0% (n¼34) completed a total of four lessons and 7.6% (n¼37) completed a total of five lessons. 3.3. Reductions in K10 score by number of lessons completed In order to assess the benefits gained by those patients who did not complete the six lesson course, effect size change on the K10 by the number of lessons completed was calculated (Table 1). As can be seen, there was a dose–response relationship between effect size change on the K10 and number of lessons completed, with those completing more lessons achieving a greater effect size change on the K10. Those who completed only two lessons demonstrate a small effect size change, but those who completed three or more lessons demonstrated moderate to large effect size changes on the K10. 3.4. Reductions in depression and suicidal ideation: marginal model Table 2 displays the overall treatment effects on depression scores and suicidal ideation under both the marginal and last observation carried forward models. Under the marginal model, scores on the PHQ-8 reduced on an average by 5.34 points, whilst the prevalence of any suicidal ideation reduced from 50% at baseline to 27% post-intervention. Reductions in depression and suicidal ideation were statistically significant and effect sizes (Cohen's d for

Table 1 Effect of size reductions in the Kessler Psychological Distress Scale (K10) by the number of lessons completed.

Completed Completed Completed Completed Completed

2 3 4 5 6

lessons lessons lessons lessons lessons

N

Mean K10 score at baseline (SE)

Mean K10 score at final lesson completed (SE)

t-test

Correlation (r)

Effect size (d)

34 46 34 37 275

27.29 28.61 30.18 29.52 28.96

25.29 24.41 24.41 21.52 19.67

t(33)¼ 2.06, p ¼0.047 t(45)¼ 4.70, po 0.001 t(33)¼ 4.68, p o0.001 t(36)¼ 6.28, p o0.001 t(274) ¼ 21.87, p o 0.001

0.84 0.72 0.69 0.68 0.59

0.20 0.52 0.63 0.83 1.19

(3.04) (2.71) (3.01) (2.99) (2.69)

(3.20) (2.93) (3.03) (3.05) (2.87)

L. Mewton, G. Andrews / Journal of Affective Disorders 170 (2015) 78–84

81

Table 2 Reductions in depression and suicidal ideation after treatment, results from both marginal and last observation carried forward models. Marginal model

PHQ-8 score

Suicidal ideation

LOCF model

Baseline mean (SE)

Post-treatment mean (SE)

t-test

Effect size (Cohen's d)

Baseline mean (SE)

Post-treatment mean (SE)

t-test

Effect size (Cohen's d)

13.21 (0.26)

7.87 (0.32)

t(308) ¼17.67, p o 0.001

0.90

13.21 (0.26)

10.26 (0.30)

t(483) ¼ 13.76, p o 0.001

0.51

Baseline Post-treatment proportion (SE) proportion (SE)

OR (95% CI)

Effect size (Cramer's phi)

Baseline Post-treatment proportion (SE) proportion (SE)

OR (95% CI)

Effect size (Cramer's phi)

0.50 (0.02)

2.71 (2.13–3.46), p o 0.001

0.37

0.50 (0.02)

1.71 (1.49–1.96), p o 0.001

0.35

0.27 (0.03)

0.37 (0.02)

Table 3 Estimated marginal PHQ-8 means and results from the marginal model with PHQ-8 scores as the outcome variable.

Time Sex Male Female Age Clinician's location Rural Urban Clinician's profession GP Other Psychologist Clinician contact No Yes Baseline suicidal ideation Not at all Some of the days Most of the days Nearly every day n

Estimated marginal mean PHQ-8 score

Main effects

Interactions with time

Baseline (SE)

Post-treatment (SE)

Beta (SE)

df

t

13.21 (0.26)

7.87 (0.32)

5.34 (0.30)

308

17.67

12.21 (0.42) 13.87 (0.34) –

7.65 (0.50) 8.01 (0.41) –

 0.36 (0.65) [Ref]  0.01 (0.02)

325

13.37 (0.40) 13.01 (0.35)

8.15 (0.43) 7.52 (0.48)

13.27 (0.39) 13.26 (0.53) 13.08 (0.49)

Beta (SE)

df

t

–

–

–

 0.56

 1.29 (0.65)

325

 2.10

n

326

0.42

 0.05 (0.02)

306

 2.34

n

 0.62 (0.64) [Ref]

325

 0.98

0.26 (0.61)

307

0.43

7.51 (0.46) 8.23 (0.64) 8.19 (0.60)

 0.68 (0.75) 0.04 (0.88) [Ref]

323 325

 0.91 0.05

0.88 (0.72) 0.15 (0.84)

306 307

1.22 0.17

12.93 (0.50) 13.24 (0.35)

7.91 (0.56) 7.84 (0.39)

0.07 [Ref]

315

0.10

-0.39

301

 0.58

10.57 (0.32) 14.40 (0.42) 17.80 (0.63) 18.84 (0.88)

5.70 (0.42) 9.26 (0.54) 10.70 (0.86) 12.08 (1.14)

[Ref] 3.56 (0.69) 4.99 (0.95) 6.37 (1.21)

303 304 303

5.19 nn 5.23 nn 5.25 nn

0.27 (0.69) 2.23 (0.96) 1.90 (1.22)

305 307 304

0.39 2.33 1.56

nn

n

po 0.05. p o0.001.

nn

depression and Cramer's phi for suicidal ideation) were moderate to large (NB: a Cramer's phi of 0.30 or greater is classified as large). The prevalence of probable depression as diagnosed by a score of 10 or greater on the PHQ-8 reduced from 71% to 28%. 3.5. Reductions in depression and suicidal ideation: last observation carried forward Under the last observation carried forward model, scores on the PHQ-8 reduced on an average by 2.95 points, whilst the prevalence of suicidal ideation decreased from 50% at baseline to 37% posttreatment. In this more conservative analysis, the reductions in depression and suicidal ideation were statistically significant and the effect sizes remained moderate to large. The prevalence of probable depression reduced from 71% to 47%. 3.6. Post-treatment outcomes: depression Table 3 displays the results from the marginal model with PHQ-8 scores as the continuous outcome variable. In each model, the main effect of time was statistically significant (po0.001 in all cases), indicating that depression scores reduced regardless of demographic and clinical characteristics. For overall depression scores on the PHQ-8, the main effect of baseline suicidal ideation was also statistically significant. When compared with those who did not

report suicidal ideation, those reporting any level of suicidal ideation also reported higher depression scores regardless of measurement occasion. The main effects of age, sex, rurality, clinician's profession and clinician contact were not statistically significant. There were statistically significant interactions between measurement occasion and sex, with a larger reduction in depression scores amongst females compared with males. In addition, there was a statistically significant interaction between measurement occasion and age (continuous), indicating that as age increased, the reduction in PHQ-8 scores across measurement occasions also increased. When compared with those who reported no suicidal ideation, those who reported that they thought about suicide “most of the days” had larger reductions in depression scores across time but this effect was not evident in any of the other suicidal ideation response options. There were no other statistically significant interactions. The main effect of time-varying suicidal ideation with depression scores was statistically significant [β ¼ 6.94, t(740.5)¼  17.32, po0.001], indicating that within individuals, suicidal ideation and depression decreased in parallel. 3.7. Post-treatment outcomes: suicidal ideation Table 4 displays the results from generalised estimating equations with suicidal ideation (dichotomised as absent or present) as the binary outcome variable. For each model, the main effect of

82

L. Mewton, G. Andrews / Journal of Affective Disorders 170 (2015) 78–84

Table 4 Estimated probability of reporting suicidal ideation before and after treatment and results from generalised estimating equations with suicidal ideation as the outcome variable.

Time Sex Male Female Age Clinician's location Rural Urban Clinician's profession GP Other Psychologist Clinician contact No Yes Baseline depression severity Sub-threshold Mild Moderate Severe n

Estimated probability of reporting suicidal ideation

Main effects

Baseline (SE)

Chi-square

Post-treatment (SE)

Interactions with time OR (95% CI) nn

Chi-square

OR (95% CI)

–

–

0.50 (0.02)

0.27 (0.03)

65.26

2.71 (2.13–3.46)

0.52 (0.03) 0.47 (0.03) –

0.28 (0.03) 0.25 (0.04) –

0.42 [Ref] 0.79

0.85 (0.51–1.41)

0.01

0.98 (0.60–1.61)

1.01 (0.99–1.03)

1.86

0.99 (0.97–1.01)

0.52 (0.02) 0.48 (0.02)

0.24 (0.02) 0.26 (0.02)

0.43 [Ref]

0.88 (0.70–1.10)

0.01

0.99 (0.69–1.44)

0.49 (0.03) 0.54 (0.05) 0.47 (0.04)

0.24 (0.04) 0.33 (0.05) 0.25 (0.05)

0.004 1.37 [Ref]

0.98 (0.53–1.81) 1.50 (0.76–3.00)

0.12 0.13

1.11 (0.62–1.98) 0.89 (0.46–1.71)

0.45 (0.04) 0.50 (0.03)

0.30 (0.05) 0.25 (0.03)

0.83 [Ref]

1.29 (0.75–2.24)

3.10

0.62 (0.37–1.06)

0.19 (0.03) 0.47 (0.04) 0.69 (0.05) 0.79 (0.04)

0.09 0.26 0.32 0.49

[Ref] 8.44 11.49 23.15

– 3.63 (1.52–8.66) n 4.95 (1.96–12.49) nn 9.95 (3.90–25.37) nn

– 0.02 1.88 1.60

– 1.07 (0.46–2.50) 1.90 (0.76–4.75) 2.19 (0.61–4.20)

(0.03) (0.04) (0.06) (0.07)

po 0.05. p o0.001.

nn

measurement occasion was statistically significant (po 0.05), indicating that suicidal ideation decreased regardless of baseline characteristics. Suicidal ideation was reported by 50% of the sample at baseline (estimated probability ¼0.50) and 27% of the population after treatment (estimated probability ¼0.27). Suicidal ideation was 2.71 times more likely at baseline compared with post-treatment. The main effect of baseline depression severity was also statistically significant. When compared to those who were sub-threshold on the PHQ-8, those who were mild, moderate or severe were more likely to report suicidal ideation regardless of measurement occasion. None of the other main effects were statistically significant. There were no statistically significant interactions. The main effect of time-varying depression and suicidal ideation was statistically significant, confirming that suicidal ideation and depression reduced in parallel (OR ¼1.29; 95% CI: 1.22–1.36, p¼ 0.02).

4. Discussion The current study demonstrated that in a large sample of patients from rural and urban Australia suicidal ideation was common. After treatment for depression, the prevalence of suicidal ideation reduced from 50% to 27%. The prevalence of probable major depression according to the PHQ-8 reduced from 71% at baseline to 28% after treatment. Reductions in depression were greater amongst females when compared with males, and reductions in depression increased as age increased. Reductions in suicidal ideation were also greater amongst those who were more severely depressed when compared to those with sub-threshold depression. Following iCBT for depression, suicidal ideation and depression decreased in parallel over time. The current study provides preliminary support for the usefulness of iCBT for depression as a prevention strategy for suicidal ideation, a strategy which can be implemented on a large scale without enacting major structural change at the societal level. It is anticipated that the current findings will place treatment for depression on the suicide prevention agenda and promote further research specifically designed to clarify the role of psychotherapy

for depression in suicide prevention. Cuijpers et al. (2012) recently reviewed a comprehensive database of 1344 studies on the psychological treatment of depression in adult populations. Of these 1344 studies, only three reported the effects on suicidal ideation or suicide risk. The effects of psychotherapy for depression on suicidality resulted in a mean effect size of g¼0.12 (95% CI: 0.20–0.44) which was not significantly different from zero. The authors concluded that there was insufficient evidence to evaluate the effects of psychotherapy of depression on suicidal behaviours and suicide risk in adult populations (Cuijpers et al., 2012). Previous studies of adolescents and geriatric patients, however, have shown that psychotherapy for depression is effective in reducing suicidal ideation and risk (March et al., 2004; Bruce et al., 2004; Brent et al., 2008; Goodyer et al., 2008). Pharmacotherapy for depression has also been shown to reduce suicide risk in geriatric patients and adults but the evidence for adolescents is equivocal (Gibbons et al., 2012). A large-scale observational study of outpatient claim data indicated that, regardless of treatment for depression (i.e., psychotherapy or pharmacotherapy), the incidence of suicide attempts reduces in the months after commencing treatment (Simon and Savarino, 2007). In the context of these findings, the current study sought to clarify the relationship between suicidal ideation and psychological treatment for depression. Whilst previous research has indicated that face-to-face psychotherapy for depression is effective in reducing suicidal ideation in adolescent and geriatric patients (March et al., 2004, Bruce et al., 2004; Brent et al., 2008; Goodyer et al., 2008), the current study demonstrates that internetdelivered treatments may be similarly effective in the depressed adult population. Future research should focus on replicating these findings in rigorously designed randomised controlled trials, devised specifically to assess whether or not iCBT for depression is capable of reducing suicidal ideation and behaviours. The strengths of this study include a large, practice-based sample which is representative of patients seeking iCBT treatment for depression from rural and urban Australia, as well as the use of appropriate statistical modelling which was capable of handling missing data without biasing parameter estimates. Additional analyses based on traditional missing data strategies (last observation carried forward) confirmed the study findings, demonstrating large treatment

L. Mewton, G. Andrews / Journal of Affective Disorders 170 (2015) 78–84

effects even under a highly conservative missing data strategy. The assessment of depression and suicidal ideation was necessarily brief but appropriate given the study setting. The PHQ-8 has been standardised and validated in clinical and epidemiological samples with excellent diagnostic specificity and sensitivity (Kroenke et al., 2009). Question nine of the PHQ-9 (“Thoughts that you would be better off dead or hurting yourself not at all, several days, more than half the days, nearly every day”) was designed to operationalise the ninth symptom of Criterion A for Major Depressive Disorder in the Diagnostic and Statistical Manual of Mental Disorders (“Recurrent thoughts of death (not just fear of dying), recurrent suicidal ideation without a specific plan, or a suicide attempt or a specific plan for committing suicide”). Whilst there are many ways to operationalise this symptom, the above method was favoured by the developers of the PHQ-9, which included the architects of the modern DSM system. In a study of 84,418 patient records, a one category increase in the frequency of suicidal ideation according to question nine of the PHQ-9 was associated with over 90% increase in the risk of suicide attempt and suicide death in the year after assessment (Simon et al., 2013). Question nine of the PHQ-9 is therefore a strong predictor of suicidal attempts and death. Despite these strengths, these results also need to be interpreted within the context of some limitations. This practice-based study was conducted for quality assurance purposes which meant that randomisation and scientific control were not possible. The lack of a control sample means that treatment effects could be attributable to regression to the mean, spontaneous remission or placebo effects, rather than the intervention per se. The fact that benefits were observed amongst patients at different levels of baseline risk indicates that regression to the mean may not underlie treatment effects. Given the cyclical nature of major depression, spontaneous remission would be expected amongst a small proportion of the patients in the current study. Spontaneous remission, however, is a rare phenomenon especially across relatively short periods of time. In previous clinical trials of the same depression course (Perini et al., 2009; Titov et al., 2010), 14.5% of those in the waitlist control condition recovered prior to receiving treatment whilst 23.6% worsened. Given the rarity of spontaneous remission, it is unlikely to have accounted for the large treatment effects shown. Study attrition and missing data also need to be considered when interpreting the current findings. Despite the fact that a large proportion of the sample did not complete the six lesson course, most of those who did not complete all six lessons still gained moderate to large effect size changes in terms of psychological distress. Consistent with previous clinical audits of the same iCBT treatment, patients who do not complete the full course therefore gained considerable benefits from the treatment prior to dropping out (Hilvert-Bruce et al., 2012; Mewton et al., 2012). These findings indicate that most of the patients are being exposed to a level of therapy that results in clinically significant change, at least in terms of psychological distress. Unfortunately, the impact of the treatment on suicidality could not be established in those who did not complete all six lessons of the course. Whilst our statistical methods were designed to account for this missing data, the high rates of study attrition remain a significant limitation. Furthermore, it was not possible to establish whether treatment effects were sustained over time due to the lack of follow up data. Finally, the lack of formal exclusion criteria meant that patients may have been using adjunctive treatments which contributed to the magnitude of treatment effects. Comorbid psychiatric illnesses were also not controlled for. Whilst the limitations outlined above may be critical within the context of an efficacy trial, they are endemic to practice-based research. The efficacy of the current iCBT course has been supported in two previous RCTs which have maximised the internal validity of the results (Perini et al., 2009; Titov et al., 2010) but in these trials, as with most trials

83

of treatments for depression those who were actively suicidal were excluded from participating for methodological and ethical reasons. It was therefore necessary to address the effects of psychological therapy for depression on suicidality outside of a clinical trial framework. To date, suicide prevention research has tended to focus on the epidemiology and social risk factors for suicide at the expense of intervention studies (Huisman et al., 2010). As a result, relatively less is known about interventions that reduce major risk factors for suicide such as major depression. In a well-cited review, Mann et al. (2005) concluded that there were only two suicide prevention strategies with unequivocal support: restriction of access to lethal means and educating physicians in the detection and diagnosis of depression. The current results provide further support for the notion that psychological therapy, in this case iCBT, for depression may also be effective in reducing suicidal ideation. Pirkis (2014/In press) have similarly shown that a partnership between primary care and specially trained psychological therapists is effective in reducing suicidal ideation. Future research needs to focus on reinforcing these findings in well-designed clinical trials.

Role of funding source None.

Conflict of interest None.

Acknowledgements None.

References Andrews, G., Cuijpers, P., Craske, M.G., Mcevoy, P., Titov, N., 2010. Computer therapy for the anxiety and depressive disorders is effective, acceptable and practical health care: a meta-analysis. PLoS One 5, e13196. Beautrais, A.L., Joyce, P.R., Mulder, R.T., Fergusson, D.M., Nightingale, S.K., 1996. Prevalence and comorbidity of mental disorders in persons making serious suicide attempts: a case control study. Am. J. Psychiatry 153, 1009–1014. Brent, D., Emslie, G., Clarke, G., Wagner, K.D., Asarnow, J.R., Keller, M., Vitiello, B., Ritz, L., Iyengar, S., Abebe, K., 2008. Switching to another SSRI or to venlafaxine with or without cognitive behavioral therapy for adolescents with SSRIresistant depression. JAMA: J. Am. Med. Assoc. 299, 901–913. Bruce, M.L., Ten Have, T.R., Reynolds III, C.F., Katz, I.I., Schulberg, H.C., Mulsant, B.H., Brown, G.K., Mcavay, G.J., Pearson, J.L., Alexopoulos, G.S., 2004. Reducing suicidal ideation and depressive symptoms in depressed older primary care patients. JAMA: J. Am. Med. Assoc. 291, 1081–1091. Christensen, H., Farrer, L., Batterham, P.J., Mackinnon, A., Griffiths, K.M., Donker, T., 2013. The effect of a web-based depression intervention on suicide ideation: secondary outcome from a randomised controlled trial in a helpline. BMJ Open, 3. Christensen, H., Griffiths, K.M., Korten, A.E., Brittliffe, K., Groves, C., 2004. A comparison of changes in anxiety and depression symptoms of spontaneous users and trial participants of a cognitive behavior therapy website. J. Med. Internet Res. 6, e46. Cuijpers, P., De Beurs, D.P., Van Spijker, B.A.J., Berking, M., Andersson, G., Kerkhof, A.J.F.M., 2012. The effects of psychotherapy for adult depression on suicidality and hopelessness: a systematic review and meta-analysis. J. Affect. Disord. 144, 189–190. Gibbons, R.D., Brown, C.H., Hur, K., Davis, J.M., Mann, J.J., 2012. Suicidal thoughts and behavior with antidepressant treatment: reanalysis of the randomized placebo-controlled studies of fluoxetine and venlafaxine. Arch. Gen. Psychiatry 69, 580–587. Goldney, R.D., Grande, E.D., Fisher, L.J., Wilson, D., 2003. Population attributable risk of major depression for suicidal ideation in a random and representative community sample. J. Affect. Disord. 74, 267–272. Goodyer, I.M., Dubicka, B., Wilkinson, P., Kelvin, R., Roberts, C., Byford, S., Breen, S., Ford, C., Barrett, B., Leech, A., 2008. A Randomised Controlled Trial of Cognitive Behaviour Therapy in Adolescents with Major Depression Treated by Selective Serotonin Reuptake Inhibitors: The ADAPT Trial. Health Technol. Assess. 12, ix-60. Hawton, K., Saunders, K.E.A., O’connor, R.C., 2012. Self-harm and suicide in adolescents. Lancet 379, 2373–2382.

84

L. Mewton, G. Andrews / Journal of Affective Disorders 170 (2015) 78–84

Hilvert-Bruce, Z., Rossouw, P.J., Wong, N., Sunderland, M., Andrews, G., 2012. Adherence as a determinant of effectiveness of internet cognitive behavioural therapy for anxiety and depressive disorders. Behav. Res. Ther. 50, 463–468. Huisman, A., Pirkis, J., Robinson, J., 2010. Intervention studies in suicide prevention research. Crisis: J. Crisis Interv. Suicide Prev. 31, 281–284. Kenward, M., Lesaffre, E., Molenberghs, G., 1994. An application of maximum likelihood and generalized estimating equations to the analysis of ordinal data from a longitudinal study with cases missing at random. Biometrics 50, 945–953. Kessler, R.C., Barker, P.R., Colpe, L.J., Epstein, J.F., Gfroerer, J.C., Hiripi, E., Howes, M.J., Normand, S.-L.T., Manderscheid, R.W., Walters, E.E., 2003. Screening for serious mental illness in the general population. Arch. Gen. Psychiatry 60, 184–189. Klein, B., Meyer, D., Austin, D.W., Kyrios, M., 2011. Anxiety Online—a virtual clinic: preliminary outcomes following completion of five fully automated treatment programs for anxiety disorders and symptoms. J. Med. Internet Res., 13. Kroenke, K., Spitzer, R.L., Williams, J.B.W., 2001. The PHQ-9: validity of a brief depression severity measure. J. Gen. Intern. Med. 16, 606–613. Kroenke, K., Strine, T.W., Spitzer, R.L., Williams, J.B.W., Berry, J.T., Mokdad, A.H., 2009. The PHQ-8 as a measure of current depression in the general population. J. Affect. Disord. 114, 163–173. Liang, K.Y., Zeger, S.L., 1986. Longitudinal data analysis using generalized linear models. Biometrika 73, 13–22. Lilienfeld, D.E., Stolley, P.D., 1994. Foundations of Epidemiology. Oxford University Press, New York, USA. Mann, J.J., Apter, A., Bertolote, J., Beautrais, A., Currier, D., Haas, A., Hegerl, U., Lonnqvist, J., Malone, K., Marusic, A., 2005. Suicide prevention strategies. JAMA: J. Am. Med. Assoc. 294, 2064–2074. March, J., Silva, S., Petrycki, S., Curry, J., Wells, K., Fairbank, J., Burns, B., Domino, M., Mcnulty, S., Vitiello, B., 2004. Treatment for Adolescents With Depression Study (TADS) Team. Fluoxetine, cognitive-behavioral therapy, and their combination for adolescents with depression: Treatment for Adolescents With Depression Study (TADS) randomized controlled trial. JAMA 292, 807–820. Mewton, L., Wong, N., Andrews, G., 2012. The effectiveness of internet cognitive behavioural therapy for generalized anxiety disorder in clinical practice. Depress. Anxiety 59, 843–849. Nock, M.K., Borges, G., Bromet, E.J., Alonso, J., Angermeyer, M., Beautrais, A., Bruffaerts, R., Chiu, W.T., De Girolamo, G., Gluzman, S., 2008. Cross-national prevalence and risk factors for suicidal ideation, plans and attempts. Br. J. Psychiatry 192, 98–105. Oquendo, M.A., Galfalvy, H., Russo, S., Ellis, S.P., Grunebaum, M.F., Burke, A., Mann, J.J., 2004. Prospective study of clinical predictors of suicidal acts after a major depressive episode in patients with major depressive disorder or bipolar disorder. Am. J. Psychiatry 161, 1433–1441. Osby, U., Brandt, L., Correia, N., Ekbom, A., Sparen, P., 2001. Excess mortality in bipolar and unipolar disorder in Sweden. Arch. Gen. Psychiatry 58, 844–850.

Perini, S., Titov, N., Andrews, G., 2009. Clinician-assisted Internet-based treatment is effective for depression: randomized controlled trial. Australas. Psychiatry 43, 571–578. Pirkis, J., 2014. Suicide prevention in a primary casr context: a pilot study. J. Ment. Health (In press). Pitman, A., Caine, E., 2012. The role of the high-risk approach in suicide prevention. Br. J. Psychiatry 201, 175–177. Pitman, A., Krysinska, K., Osborn, D., King, M., 2012. Suicide in young men. The Lancet 379, 2383–2392. Rihmer, Z., 2001. Can better recognition and treatment of depression reduce suicide rates? A brief review. Eur. Psychiatry 16, 406–409. Salim, A., Mackinnon, A., Christensen, H., Griffiths, K., 2008. Comparison of data analysis strategies for intent-to-treat analysis in pre-test–post-test designs with substantial dropout rates. Psychiatry Res. 160, 335–345. Siddiqui, O., Hung, H.M.J., O’neill, R., 2009. MMRM vs. LOCF: a comprehensive comparison based on simulation study and 25 NDA datasets. J. Biopharm. Stat. 19, 227–246. Simon, G.E., Rutter, C.M., Peterson, D., Oliver, M., Whiteside, U., Operskalski, B., Ludman, E.J., 2013. Does response on the PHQ-9 depression questionnaire predict subsequent suicide attempt or suicide death? Psychiatr. Serv. 64, 1195–1202. Simon, G.E., Savarino, J., 2007. Suicide attempts among patients starting depression treatment with medications or psychotherapy. Am. J. Psychiatry 164, 1029–1034. Titov, N., Andrews, G., Davies, M., Mcintyre, K., Robinson, E., Solley, K., 2010. Internet treatment for depression: a randomized controlled trial comparing clinician vs. technician assistance. PLoS One 5, e10939. Van der lem, R., Van der wee, N., Van Veen, T., Zitman, F., 2011. The generalizability of antidepressant efficacy trials to routine psychiatric out-patient practice. Psychol. Med. 41, 1353–1363. van spijker, B.A., Van Straten, A., Kerkhof, A.J., 2014. Effectiveness of online self-help for suicidal thoughts: results of a randomised controlled trial. PLoS One 9, e90118. Watts, S., Newby, J., Mewton, L., Andrews, G., 2012. A clinical audit of changes in suicide ideas with internet treatment for depression. Br. Med. J. Open 2, e001558. West, B., Welch, K.B., Galecki, A.T., 2006. Linear Mixed Models: Practical Guide Using Statistical Software. Chapman & Hall/CRC, Boca Raton, Florida. Yip, P.S.F., Caine, P., Yousuf, S., Chang, S., Wu, K.C., Chen, Y., 2012. Means restriction for suicide prevention. Lancet 379, 2393–2399. Zimmerman, M., Chelminski, I., Posternak, M.A., 2005. Generalizability of antidepressant efficacy trials: differences between depressed psychiatric outpatients who would or would not qualify for an efficacy trial. Am. J. Psychiatry 162, 1370–1372.