5 page paper follow instructions

© 2011 Th

Outcome of adolescent depression: 6 months after treatment

Michael S. Gordon , Bruce Tonge , Glenn A. Melvin

Michael S Adjunct L

Departme Road, Cla Monash U southernh

Bruce Ton

Objective: To evaluate the demographic and clinical factors that predicted depression in

adolescents at 6 months after treatment.

Method: A total of 130 adolescents (aged 12 to 18 years) who had been clinically referred

for treatment with a DSM-IV major depressive disorder, dysthymic disorder or minor depres-

sion were treated with either CBT, sertraline, their combination or supportive psychothera-

py, in two randomized clinical trials using the same assessment instruments. Assessments

in both studies were conducted at initial assessment, three months later at the conclu-

sion of treatment, and at 6-month follow up. The data of these two trials were pooled. The

primary outcome measures were the presence of a depressive disorder and the Reynolds

Adolescent Depression Scale (RADS) score at 6-month follow up.

Results: At baseline, only the Self Efficacy Questionnaire for Depressed Adolescents

(SEQ-DA) predicted depression at 6-month follow up. Individual measures following

3 months of acute treatment that predicted depression at 6-month follow up were SEQ-DA,

RADS, Revised Children ’ s Manifest Anxiety Scale-total score, Global Assessment of

Function, adolescent-rated Family Assessment Device General Functioning Subscale and

adolescent-rated Visual Analogue Scores of mood states.

Conclusion: Clinical variables as reported by the adolescent and identifi ed by the clini-

cian at baseline assessment and following 3 months of treatment predicted depression

at 6-month follow up. No demographic variables were predictive of depression at 6-month

follow up.

Key words: depression , adolescent , treatment outcome , prognosis.

Australian and New Zealand Journal of Psychiatry 2011; 45:232–239

While the duration of youth major depressive disorder

has been found to range from 2 to 520 weeks, the mean

duration is typically 20 to 35 weeks varying with the

sample under investigation [1,2]. Youth depression is a

relapsing and remitting condition that will reoccur in

over half of children and adolescents. In one study of

children with MDD, 72% had a recurrence of depression

e Royal Australian and New Zealand College of Psychiatrists

. Gordon (Correspondence), Acting Clinical Director and ecturer

nt of Child Psychiatry, Monash Medical Centre, 246 Clayton yton, Victoria 3168, and School of Psychology and Psychiatry, niversity, Victoria, Australia. Email: Michael.Gordon@ ealth.org.au

ge, Professor and Head

in the following 5 years [3], while in another study of

depressed adolescents the recurrence rate was 69% over

a 7-year follow up period [4]. These fi ndings suggest that

youth depression is typically a long-term, recurrent issue

for young people and their families.

Studies of the factors which affect the course of major

depressive disorder (MDD) have produced inconsistent

School of Psychology and Psychiatry, Monash University, Victoria, Australia

Glenn A. Melvin, Lecturer

Centre for Developmental Psychiatry and Psychology, School of Psy- chology and Psychiatry, Monash University, Notting Hill, Victoria, Australia

Received 5 September 2010; accepted 1 November 2010.

M. S. GORDON, B. TONGE, G. A. MELVIN 233

fi ndings. The factors affecting the course of dysthymic

disorder (DD) and minor depression (less than fi ve

depressive symptoms, for less than 1 year) in adolescents

are even less well understood than those affecting MDD.

A number of factors confound predictor studies, includ-

ing the use of varying diagnostic criteria, the duration

of follow up period, sampling differences, and assess-

ment measures [5]. Variability in fi ndings may refl ect

different criteria across studies, including defi ning a

response to treatment, patient selection, differences in

samples (e.g. gender, age), sample size, different assess-

ment measures, spontaneous remission, placebo response,

type of statistical analysis, lack of psychiatric control

groups, co-morbidity, life events, and family history of

psychiatric disorders [5,6].

Demographic variables which have been implicated in

affecting the course of youth depressive disorders include

female gender [2,7,8], lower SES [4,7], race [9] , and age

[10]. However, other studies have not replicated these

fi ndings [11].

Sanford et al. [10] reported that for adolescent MDD, a co-morbid anxiety disorder predicted persistence of

MDD at 12 months follow up. Lewinsohn et al. [12] found that in a community sample of adolescents, a past

history of an anxiety disorder was associated with a sta-

tistically signifi cant odds ratio of 2.4 of having a depres-

sive disorder in the future. In other studies, non-affective

co-morbidity did not alter the duration of the major

depressive episode [11,13]. In a British study of depressed

youth attending Child and Adolescent Mental Health

Clinics, predictors of outcome at 6 months were severity

of baseline depression, obsessive – compulsive disorder,

suicidal thinking and disappointing life events [14].

Overall, fi ndings are inconsistent in relation to predic-

tors such as demographic factors and co-morbid psychiat-

ric diagnoses, but more consistent in relation to predictors

such as severity of depression at baseline, adolescent

suicidal behaviours/thoughts, and parental depression.

These inconsistent fi ndings highlight the need for

further studies into predictors of outcome of adolescent

depression. The aim of this study was to explore demo-

graphic and clinical variables which might predict cate-

gorical depressive disorders or dimensional depressive

symptoms in adolescents at six-month follow up after the

treatment for depression.

Method

Subjects

The 130 adolescents (12 to 18 years inclusive) included in this

study were referred for treatment of depression by school guidance

offi cers, medical practitioners, community health centres and Child

and Adolescent Mental Health Clinics to a community-based Adoles-

cent Mental Health Research Clinic. They were suffering with a

primary DSM-IV diagnosis of major depressive disorder (MDD),

dysthymic disorder (DD), depressive disorder not otherwise specifi ed

(DDNOS), or adjustment disorder with depressed mood (ADDM) and

were enrolled in one of two randomized controlled trials for the treat-

ment of adolescent depression; the Berriga House (BH) (N � 57) and

Time for a Future (TFF) (N � 73) studies. The TFF cohort and meth-

odology has been previously described [15]. In brief, the TFF sample

consisted of 73 adolescents, aged 12 to 18 years inclusive, who were

clinically referred for the treatment of depression, between July 2000

and December 2002. The 57 BH depressed adolescents were treated

between February 1997 and July 2002. The TFF and BH projects had

identical exclusion criteria of major physical illness, epilepsy, bipolar

affective disorder, intellectual disability impairing capacity to partici-

pate in psychotherapy, psychosis, organic brain syndrome, primary

substance abuse, actively suicidal or suffering with a psychiatric dis-

turbance requiring acute hospital admission, and current antidepres-

sant treatment. The BH and TFF studies employed the same assessment

instruments and assessment intervals. The BH study was a group com-

parison design with two active treatments (adolescent-focused CBT

(CBT-A) and adolescent- and parent-focused CBT (CBT-A&P)) and a

control treatment (supportive psychotherapy). Treatment with adoles-

cent-focused CBT (CBT-A) comprised 14 adolescent sessions and 7

parent sessions. Treatment with adolescent- and parent-focused CBT

(CBT-A&P) comprised 14 adolescent sessions and 14 parent sessions.

In both the BH and TFF studies depressed adolescents were treated

acutely for their depression for 3 months, with a follow up at 6 months

following the end of acute treatment. The similar design and identical

assessment tools permits the measures in each of the two studies to be

pooled for analysis; see Figure 1. The pooled cohort comprised 85 girls

and 45 boys (female-to-male ratio of 1.9:1). The mean age in this TFF/

BH cohort was 15.1 years (boys 14.8 years, girls 15.2 years) with the

modal age 16 years. Of the total, 119 (92%) of the adolescents were

born in Australia, three were born in England, two in New Zealand,

and one each in Iran, China, Japan, Lebanon, Serbia and the Nether-

lands. Using DSM-IV criteria [16] 72 (55.4%) were diagnosed with

MDD, 30 (23.1%) with DD, 27 (20.8%) with DDNOS, and one ado-

lescent suffered with ADDM.

Analyses

Multiple predictor measures were taken at baseline assessment (T1)

and immediately following 3 months of acute treatment (T2). Primary

categorical and continuous outcome measures determined a priori were

the presence of a depressive disorder 6 months following acute treatment

(T3) and the Reynolds Adolescent Depression Scale [17] score at T3.

Measures

The Schedule for Affective Disorders and Schizophrenia for School

Aged Children Lifetime was used to determine current depressive

diagnosis according to DSM-IV criteria [18]. The Reynold ’ s Adoles-

cent Depression Scale (RADS) [17] is a reliable and valid adolescent

self-report questionnaire used to gauge depressive symptoms along

a dimensional scale with higher scores indicating a higher level of

234 PREDICTORS OF ADOLESCENT DEPRESSION

Month

Pre-Treatment Assessment (n = 130)

Treatment (n = 130) (cognitive behavioural therapy, sertraline, cognitive behavioural therapy & sertraline,

supportive psychotherapy)

Post-Treatment Assessment (n = 111)

6-Month Follow Up Assessment (n = 96)

Time 1

Time 2

Time 3

0

0.5

3.5

6.5

Figure 1. Program overview.

depression. The clinician assessed global functioning of the adolescent

with the Global Assessment of Functioning (GAF) which is axis V of

the DSM-IV [16]. The Self-Effi cacy Questionnaire for Depressed Ado-

lescents (SEQ-DA), a 12-item self-report scale questionnaire rated the

adolescent ’ s perceived ability to cope with depressive symptoms [19].

Higher scores refl ect a more adaptive self-evaluation of coping with

daily activities and feelings. The Revised-Children ’ s Manifest Anxiety

Scale (RCMAS) total score was used to measure adolescent self-report

trait anxiety [20 – 22]. This study also used a 100-point Visual Analogue

Scale (VAS) for the assessment of self-reported depressed mood (Sad

Thermometer) and irritable mood (Cranky Thermometer). Higher

scores refl ect higher levels of depression or irritability respectively,

referring to current sadness (SAD 1 ), the worst sadness in the last

2 weeks (SAD 2 ), and the worst sadness in their lifetime (SAD

3 ) and

current irritability (Cranky 1 ), the worst in the last 2 weeks (Cranky

2 ),

and the worst irritability in their lifetime (Cranky 3 ).

Each parent independently completed the Beck Depression Inven-

tory II (BDI-II) and the Abbreviated Dyadic Adjustment Scale (ADAS).

The BDI-II is a self-report questionnaire assessing the severity of

depressive symptoms with higher scores refl ecting more severe depres-

sion [23]. The ADAS is a seven item abridged appraisal of parental

satisfaction in the relationship with their life partner, which is derived

from a subset of items in the Dyadic Adjustment Scale [24,25]. The

clinician-rated Global Assessment of Relational Functioning (GARF)

and the self-report General Functioning Subscale of the Family Assess-

ment Device (FAD-GF) were used to gauge family functioning. The

GARF is a marker of family functioning as described in DSM-IV [16,

p. 758]. Like the GAF, higher scores represent more adaptive function-

ing, and representative values are given as a rating guide. The FAD-GF,

a 12-item version of the McMaster Family Assessment Device (FAD)

was used to assess both parents and the young person ’ s perception of

family functioning [26]. The possible scoring range is between 1

(healthy) and 4 (unhealthy family functioning). A score on the FAD-GF

of greater than 2 cut-off has been used in other adolescent MDD stud-

ies to identify dysfunctional family dynamics [27,28].

Design

The TFF was a group comparison design in which depressed ado-

lescents were randomly allocated to cognitive behavioural therapy

(CBT), sertraline pharmacotherapy, or combined sertraline and cogni-

tive behavioural therapy. The BH study was a group comparison design

where depressed adolescents were randomized to adolescent-focused

individual CBT, adolescent and parent-focused CBT, or non-directive,

supportive psychotherapy. In both the BH and TFF studies the CBT

was offered by trained psychologists or mental health practitioners.

Treatments comprised weekly sessions scheduled over 3 months. The

TFF study design, manual-based CBT and fl exible dose design of

sertraline pharmacotherapy has been reported elsewhere [15]. Sup-

portive psychotherapy involved 3 months of non-directive psycho-

education and support, where the trained therapist used a manual to

structure discussion of the patient ’ s illness and provided information,

and educational videos.

Of the combined 130 TFF/BH cohort that began treatment at T1, 111

were assessed at T2, and 96 were assessed at T3. A total of 44 variables

were identifi ed as potential predictors of depressive disorders and

depressive symptoms at T3. These potential predictors included six

demographic variables (age, gender, father ’ s occupation, residing in met-

ropolitan versus rural area, marital status of the mother, and education

type of the adolescent), and 12 adolescent variables at T1, and 12 vari-

ables at T2 (FAD-GF, RCMAS, RADS, SEQ-DA, GAF, SAD 1 , SAD

2 ,

SAD 3 , Cranky

1 , Cranky

2 , Cranky

3 , and the adolescent ’ s depressive diag-

nosis). The seven potential parent and family predictors were GARF

scores, mother FAD-GF, mother ADAS, mother BDI score, father FAD-

GF, father ADAS, and father BDI score, each assessed at T1 and T2.

Statistical analysis

The presence or absence of a depressive disorder at T3 was modelled

as a function of individual demographic variables in a logistic regres-

M. S. GORDON, B. TONGE, G. A. MELVIN 235

sion. The signifi cant demographic predictor variables were retained, and

adolescent and parent clinical variables from T1 and T2 were singularly

entered in a logistic regression analysis. The independent variables that

were found to be signifi cant predictors in a binary regression were then

all entered into the model in a combined regression analysis.

The RADS score at T3 was modelled as a function of individual

demographic variables in a linear regression. The signifi cant demo-

graphic predictor variables were retained, and adolescent and parent

clinical variables from T1 and T2 were assessed as the independent

variables. The independent variables that were found to be signifi cant

predictors in a linear regression were then all entered in the model, in

a combined regression analysis.

Data were analysed using the Statistical Package for the Social Sci-

ences Version 11 [29]. The studies received ethical approval from the

Southern Health Human Research Ethics Committee. Written informed

consent was obtained from the parents and young person.

Results

The mean scores of the adolescent and parent continuous variables

at T1 are given in Table 1. A total of 81 adolescents (62.3%) at T1

suffered a co-morbid diagnosis including generalized anxiety disorder

(12.3%), oppositional defi ant disorder (4.6%), post-traumatic stress

disorder (5.3%), and social phobia (2.3%).

A total of 48 of the 111 adolescents (43.2%) assessed at T2 suffered

with a diagnosed depressive disorder. At T3, 26 of the 96 adolescents

assessed (27.1%) suffered with a depressive disorder. See Table 2 for

changes to the adolescent continuous variables from T1 across to T2

and T3.

Repeated measures analysis of variance (ANOVA)

A repeated measures ANOVA was undertaken to assess whether the

six different treatment arms are equally effective interventions for

depression in adolescence. The predictor analyses proposed are predi-

cated on an assumption of therapeutic equivalence between each of the

six treatment arms. If any of the treatments vary in effi cacy then this

will skew the predictors towards more heavily weighting those vari-

ables that change more in the effi cacious treatment arms. Repeated

measures ANOVA were performed on the RADS at T1, T2 and T3. In

this mixed model analysis, outcome factors assessed included time

(RADS scores at T1, T2 and T3), treatment (the six different treat-

ments), and gender of the participants. Interactions of factors were

also assessed between factors, including time × treatment, time × gen- der, and time × treatment × gender. Time was the only signifi cant factor in this repeated-measures ANOVA; F(2,136) � 58.991, p �

0.001, partial η � 0.465. Time was a signifi cant factor between pre- and T3 (p � 0.001) and between pre-/post, and post/T3 (p � 0.013).

However, a time × treatment interaction was not signifi cant p � 0.392; suggesting that based on RADS scores, there is no evidence that

response varies between treatments.

Repeated measures ANOVAs were performed on the RCMAS

at T1, T2 and T3. Time was the only signifi cant factor; F(2,144) �

47.268, p � 0.001, partial η � 0.396.

Logistic regression: predictors at T1

A total of 96 adolescents had complete datasets for T1, T2 and T3.

A binary logistic regression was conducted on the six demographic and

19 clinical variables (13 adolescent variables and six parent variables),

which were thought to predict a depressive disorder at T3. The variables

were individually entered in the regression equation with depressive

disorder at T3 as the dependent variable. None of the demographic

variables were found to be predictors of a depressive disorder at T3.

Demographic variables were excluded from subsequent binary regres-

sion modelling. At T1 the presence of MDD, as opposed to DD, pre-

dicted depressive disorder at T3 (odds ratio (OR) 12.35, p � 0.018,

95%CI 1.54 – 98.95). In univariate analysis, the only continuous variable

that predicted depressive disorder at T3, was the SEQ-DA score at T1

(OR 0.928, p � 0.036, 95%CI 0.866 – 0.995). No other adolescent or

parent variable at T1 predicted depressive disorder at T3.

Logistic regression: predictors at T2

Nineteen clinical variables (12 adolescent variables and 7 parent

variables) were entered in single model binary regression analyses.

Table 1. Mean scores and standard deviation for continuous variables at T1

T1 variables Mean score sd

Adolescent variables

RADS 82.18 14.62

SEQ-DA 34.4 6.68

FAD-GF 2.43 0.52

RCMAS total score 63.84 10.12

GAF 57.86 7.46

Parental measures

GARF scores 65.7 12.32

Mother FAD-GF 2.13 0.45

Mother BDI 10.3 8.58

Mother ADAS 24.22 5.09

Father FAD-GF 2.05 0.41

Father BDI 7.07 6.7

Father ADAS 24.05 5.46

Table 2. Changes to continuous adolescent variables across the study (mean scores with standard

deviation in brackets)

Changes to

mean scores

over study T1 T2 T3

GAF 57.86 (7.46) 69.12 (9.76) 71.15 (10.53)

RADS 82.18 (14.62) 66.33 (16.64) 59.44 (17.37)

RCMAS total

score

63.84 (10.12) 53.81 (12.91) 48.84 (14.83)

SEQ-DA 34.4 (6.68) 41.7 (7.54) 44.22 (8.06)

Adolescent

FAD-GF

2.43 (0.52) 2.25 (0.55) 2.15 (0.56)

236 PREDICTORS OF ADOLESCENT DEPRESSION

The six clinical variables identifi ed at T2 that predicted the diagnosis

of a depressive disorder at T3 in a univariate regression were the diag-

nosis of MDD or MDD partial remission, a lower GAF score, an ele-

vated RADS score, an elevated FAD-GF score as assessed by the young

person, a lower SEQ-DA score, and a higher RCMAS score (see Table 3).

The T2 variables which were non-predictors of depressive disorder at

T3, include the VAS measures (SAD 1 , SAD

2 , SAD

3 , Cranky

1 , Cranky

2 ,

Cranky 3 ), all the parent measures (FAD-GF, ADAS and self-report BDI

scores) and the clinician-rated GARF score.

Logistic regression: multivariate analysis at T2

A binary regression analysis with all the T2 continuous variables

(GAF score, SEQ-DA, RADS, RCMAS, and the young person ’ s FAD-

GF) were combined using the enter method. Only the GAF score at

T2 and the adolescent FAD-GF at T2 remained as signifi cant predic-

tors of a depressive disorder at T3 (p � 0.008, OR � 0.899; 95%CI

0.831 to 0.973, and p � 0.049, OR � 3.344; 95%CI 1.003 to 11.151

respectively).

Linear regression modelling

The RADS at T3 was used as the continuous dependent variable in

linear regression analyses using the enter method. Regression models

were initially undertaken using only gender and age. Female gender (but

not age) was found to be a signifi cant predictor of T3 RADS score, B �

11.125, 95%CI; 4.01 to 18.24. None of the other four demographic vari-

ables were signifi cant predictors of RADS score at T3. Further, regression

models were undertaken retaining gender as a co-variate, and then singly

entering the other clinical continuous variables. The fi ve variables at T1

which predicted T3 RADS score were the RADS score (B � 0.407; Beta

� 0.369, p � 0.001, 95%CI for B 0.176 to 0.638), SEQ-DA (B �

� 0.858; Beta � � 0.341, p � 0.001, 95%CI for B � 1.345 to � 0.371),

adolescent FAD-GF (B � 6.761; Beta � 0.206, p � 0.039, 95%CI for

B 0.362 to 13.16), and the SAD thermometers SAD 1 (B � 0.148; Beta

0.226, p � 0.028, 95%CI for B 0.016 to 0.281) and SAD 2 (B � 0.148;

Beta � 0.217, p � 0.033, 95%CI for B 0.012 to 0.284). None of the

parent variables at T1 were found to predict RADS at T3.

Linear regression at T2

Gender was retained as a co-variate, and the 12 adolescent and seven

parent continuous measures were entered individually as independent

variables. The continuous variables at T2 which were found to be sig-

nifi cant predictors of RADS scores at T3 were the GAF, SEQ-DA,

RADS, RCMAS, the young person ’ s FAD-GF, and the VAS of SAD 1 ,

SAD 2 , Cranky

2 and Cranky

3 (see Table 4).

Linear regression analyses: multivariate predictors at T1

The clinically signifi cant variables at T1, and then separately at T2

were entered into a combined regression analysis. Gender and T1 vari-

ables of RADS, SEQ-DA, the adolescent ’ s FAD-GF, SAD 1 , and SAD

2

were entered in a regression analysis using the enter method. No base-

line (T1) continuous variable remained signifi cant when all the signifi -

cant variables from the previous analyses were combined.

Linear regression analyses: multivariate predictors at T2

Gender was combined with the T2 variables of RADS GAF, SEQ-

DA, RCMAS, adolescent ’ s FAD-GF, SAD 1 , SAD

2 , Cranky

2 , and

Cranky 3 in a regression analysis using the enter method. In this com-

bined regression analysis, the remaining signifi cant predictors were

gender (B � 9.34; Beta � 0.253, p � 0.005, 95%CI for B 2.915 to

15.764) and T2 SEQ-DA (B � � 0.734; Beta � � 0.334, p � 0.012,

95%CI for B; � 1.3 to � 0.168).

Discussion

As prior studies investigating predictors of adolescent

depression have shown inconsistent fi ndings, the aim

of this study was to investigate predictors in a sample of

130 adolescents with depressive disorders. Binary regres-

sion analyses indicated that at baseline (T1) only the

presence of major depression and self-effi cacy (SEQ-DA

score) were predictors of a depressive disorder at T3. In

linear regression analysis, the T1 variables of adolescent

FAD-GF, RADS, and SEQ-DA were all predictors of

RADS scores at long-term follow up (T3).

Regression analyses indicated that after 3 months of

treatment, individual measures that were in part predic-

tive of whether the adolescent would be depressed at

T3 included a self-report measures of adolescent self-

effi cacy (SEQ-DA), adolescent depression (RADS),

adolescent anxiety (RCMAS), a family functioning

(FAD-GF), the residual diagnosis of a depressive disor-

der (MDD or MDD partial remission), and a clinician ’ s

assessment of adolescent functioning (GAF). Clinically,

Table 3. Logistic regression and the prediction of a depressive disorder at T3; signifi cance, odds ratios

and confi dence intervals

At T1

Signifi cance

(p value)

Odds

Ratio 95%CI

• Initial diagnosis of MDD

0.018 12.353 1.542 – 98.951

• SEQ-DA 0.036 0.928 0.866 – 0.995

At T2

• MDD • MDD partial

remission

0.003

0.002

35.000

8.000

3.470 – 353.014

2.122 – 30.155

• GAF score 0.001 0.900 0.849 – 0.954 • SEQ-DA 0.002 0.888 0.825 – 0.956 • Young person’s

FAD score

0.012 3.224 1.289 – 8.061

• RADS 0.009 1.041 1.01 – 1.073 • RCMAS 0.011 1.052 1.011 – 1.094

M. S. GORDON, B. TONGE, G. A. MELVIN 237

it would make sense that self-reported markers of self-

effi cacy, depression, anxiety, perceptions of the family,

clinician assessment and residual depression would pre-

dict depressive disorder at long-term follow up. There is

good face-validity to these fi ndings. The same variables

that were predictors at T2 in binary regression were

also predictors in a linear regression; T2 VAS variables

of sad mood and irritability (SAD 1 , SAD

2 , Cranky

2 , and

Cranky 3 ) were all predictors.

The design of this study and its fi ndings are similar to

the published papers by Birmaher et al. [30], Brent et al. [31], Vostanis et al. [32] and Clarke et al. [33]. Birmaher et al. treated 107 adolescents suffering with MDD with short-term psychotherapy (CBT, systemic behavioural

family therapy or non-directive supportive therapy) and

followed them up at 2 years [30]. Like this study, Birma-

her et al. had a naturalistic follow up phase, where over half the cohort received other support and treatment out-

side the study design [30]. They found that the child

FAD-GF, Children ’ s Global Assessment Scale and self-

reported BDI following psychological treatment were

predictors of recovery from MDD over the 2-year follow

up period [30], which were comparable to the RADS,

GAF and adolescent FAD-GF fi ndings in this study.

In a companion paper, Brent et al. [31] using the same cohort as Birmaher et al. [30] examined the 2-year out- come following brief psychotherapy for those adoles-

cents with subsyndromal depression or MDD or who

were no longer depressed following psychotherapy. Sim-

ilar measures were employed to this study, and they found

a trend for a greater rate of recovery in the subsyndromal

compared with the still depressed group. Those with sub-

syndromal depression were not at higher risk of recur-

rence than those who were non-depressed following

treatment [31]. Predictors of time to recovery (recur-

rence) were MDD at the end of treatment and impaired

functioning on the FAD-GF, although it was not clear

whether this measure was reported on by the adolescent

or parent [31].

Vostanis et al. [32] assessed 58 children and adoles- cents nine months after a randomized control trial of out-

patient CBT or non-focused therapy. In binary regression,

with depression at LTFU being the outcome measure,

school functioning, additional outpatient appointments,

child ’ s expectancy of treatment and self-esteem were the

predictors of depression at 9 months. In a multiple regres-

sion analysis using the Mood and Feelings Questionnaire

as the dependent variable, child self-esteem was found to

be a predictor [32]. A comparison of the Vostanis [32]

study with this study highlights the problem of the dif-

fi culty comparing studies because of the use of different

measures, cohort differences, differences in treatments

and in the handling of the naturalistic phase with the

result that the different studies of predictors yield differ-

ent results.

Clarke et al. [33] found differences in the predictor fi ndings depending on the methodology within their

study CBT, or 14-session adolescent and parent CBT, or

a waiting list control condition for adolescent depression

(MDD or DD). The variables at baseline were assessed

in a discriminant function analysis to see if they predicted

(i) those who recovered from their depressive disorder,

and (ii) adolescent-reported BDI [33]. They reported dif-

ferent predictor variables when they used a self-report

measure (BDI) as the dependent variable, compared to

when they employed a dichotomous diagnostic outcome

as the dependent variable. Recovery from depression was

predicted by four baseline variables of lower BDI scores,

lower anxiety scores, increased level of enjoyment and

pleasant activities and a higher level of rational thoughts

on the Subjective Probability Questionnaire [33].

Clarke et al. [33] reported that their ‘ results support the thesis that defi nitions of depressive symptomatology

and/or disorder which differ with respect to assessment

(self-report versus semi-structured diagnostic interview)

as well as classifi cation methods (categorical diagnoses

versus continuous scales) will often yield vastly different

associations with variables of interest ’ (p.351). This is in

Table 4. Linear regression: dependent variable is RADS at T3

Independent variables

gender as co-variate B Beta Signifi cance, p 95%CI for B

T2 GAF � 0.596 � 0.293 0.003 � 0.989 to � 0.203 T2 SEQ-DA � 1.197 � 0.538 � 0.001 � 1.584 to � 0.81 T2 adolescent FAD 10.764 0.334 0.001 4.386 to 17.143

T2 RCMAS total score 0.614 0.483 � 0.001 0.379 to 0.848

T2 RADS score 0.578 0.555 � 0.001 0.395 to 0.761

T2 SAD 1 0.212 0.243 0.024 0.029 to 0.395

T2 SAD 2 0.177 0.296 0.008 0.048 to 0.306

T2 Cranky 2 0.215 0.346 0.001 0.93 to 0.337

T2 Cranky 3 0.152 0.074 0.043 0.005 to 0.3

238 PREDICTORS OF ADOLESCENT DEPRESSION

keeping with the fi ndings of this study, in which fi ve

baseline variables were predictors of RADS score at T3,

but were not predictive of depressive disorder at T3.

As with this study, Clarke et al. [33] also reported on the importance of gender as being a univariate predictor

for change on the pre-to-post BDI, but gender was not a

predictor of recovery. This study was consistent with the

fi nding of Wilkinson et al. who reported that the most effi cient predictor of depression was overall depressive

symptoms [14].

Limitations

This study pooled the diagnoses of major depression, DD

and minor depression. Parker has made the argument

that, at least for adults, specifi c treatments for each

depressive disorder sub-type are indicated [34]. Con-

versely, a community study of psychosocial functioning

and co-morbid substance use in adults and adolescents

found that depressive symptoms exist as a continuum

rather than a categorical threshold [35]. This dimensional

concept of depression would validate pooling MDD, DD

and minor depression for the purposes of statistical anal-

ysis. However in the logistic regression analysis an MDD

diagnosis was predictive but not DD, perhaps indicating

that the treatment offered might have been have been

more benefi cial to those young people with a more

chronic and milder depression presenting as a DD.

Conclusion

The clinical implication of this study is that self-reported

questionnaires completed by the young person are useful

and reliable markers which can predict the course of their

mental health 6 months after treatment. One of the more

reliable markers was the SEQ-DA. Further research is

required to establish a cut-off score for the SEQ-DA;

however, these fi ndings point to the importance of the

development of a sense of self-effi cacy in teenagers who

have experienced depression. Further work is underway

to compare a longer follow up period into adult life, to

determine those variables which continue to predict

depression.

Declaration of interest: Funding for Time for a Future

was through the Department of Human Services Victoria

Mental Health Branch, Australian Rotary Health Research

Fund, Department of Human Services Barwon Region

Innovative Practice Grant, and the Commonwealth

Department of Health and Aged Care. Berriga House was

funded through an NHMRC project grant (grant 990154).

The sertraline used in Time for a Future was provided

by Pfi zer Pharmaceuticals. The authors alone are respon-

sible for the content and writing of the paper.

References

Kaminski KM, Garber J. Depressive spectrum disorders in high-1.

risk adolescents: episode duration and predictors of time to recov-

ery. J Am Acad Child Adolesc Psychiatry 2002; 41:410 – 418.

Lewinsohn PM, Clarke GN, Seeley JR, Rhode P, Major depression 2.

in community adolescents: age at onset, episode duration, and

time to recurrence. J Am Acad Child Adolesc Psychiatry 1994; 33: 809 – 818.

Kovacs M, Feinberg TL, Crouse-Novak M, Paulauskas SL, Pollack M, 3.

Finkelstein R, Depressive disorders in childhood. II: a longitudinal

study of the risk for a subsequent major depression. Arch Gen Psychiatry 1984; 41:643 – 649. Rao U, Ryan ND, Birmaher B, Dahl RE, Williamson DE, Kaufman J, 4.

Rao R, Nelson B. Unipolar depression in adolescents: clinical out-

come in adulthood. J Am Acad Child Adolesc Psychiatry 1995; 34:566 – 578.

Birmaher B, Arbelaez C, Brent D. Course and outcome of child 5.

and adolescent major depressive disorder. Child Adolesc Psychiatr Clin N Am 2002; 11:619 – 637. Nierenberg AA. Predictors of response to antidepressants. General 6.

principles and clinical implications. Psychiatr Clin North Am 1995; 26:345 – 352.

McCauley E, Myers K, Mitchell J, Calderon R, Schloredt K, 7.

Treder R. Depression in young people: initial presentation and

clinical course. J Am Acad Child Adolesc Psychiatry 1993; 32:714 – 722. Dunn V, Goodyer IM. Longitudinal investigation into childhood- 8.

and adolescent-onset depression psychiatric outcomes in early

adulthood. Br J Psychiatry 2006; 188:216 – 222. Emslie GJ, Rush AJ, Weinberg WA, Gullion CM, Rintelmann J, 9.

Hughes CW. Recurrence of major depressive disorder in hospital-

ized children and adolescents. J Am Acad Child Adolesc Psychiatry 1997; 36:785 – 792.

Sanford M, Szatmari P, Spinner M, Munroe-Blum H, Jamieson E, 10.

Walsh C, Jones D. Predicting the one-year course of adolescent

major depression. J Am Acad Child Adolesc Psychiatry 1995; 34:1618 – 1628.

Kovacs M, Obrosky S, Gastonis C, Richards C. First-episode major 11.

depressive and dysthymic disorder in childhood: clinical and

sociodemographic factors in recovery. J Am Acad Child Adolesc Psychiatry 1997; 36:777 – 784. Lewinsohn PM, Roberts RE, Seeley JR, Rohde P, Gotlib IH, 12.

Hops H. Adolescent psychopathology: II. Psychosocial risk factors

for depression. J Abnorm Psychol 1994; 103:302 – 315. Rohde P, Lewinsohn PM, Seeley JR. Comorbidity of unipolar 13.

depression: II. Comorbidity with other mental disorders in adoles-

cents and adults. J Abnorm Child Psychol 1991; 100:214 – 222. Wilkinson P, Dubicka B, Kelvin R, Roberts C, Goodyer I. Treated 14.

depression in adolescents: predictors of outcome at 28 weeks.

Br J Psychiatry 2009; 194:334 – 341. Melvin GA, Tonge BJ, King NJ, Heyne D, Gordon MS, Klimkeit E. 15.

A comparison of cognitive – behavioral therapy, sertraline, and their

combination for adolescent depression. J Am Acad Child Adolesc Psychiatry 2006; 45:1151 – 1161. American Psychiatric Association. 16. Diagnostic and statistical manual of mental disorders . 4th edn. Washington, DC: American Psychiatric Association, 1994.

M. S. GORDON, B. TONGE, G. A. MELVIN 239

Reynolds WM. 17. Professional Manual for the Reynolds Adolescent Depression Scale (RADS) . Odessa, FL: Psychological Assessment Resources, 1987.

Ambrosini PJ. Historical development and present status of the 18.

schedule for affective disorders and schizophrenia for school-age

children (K-SADS). J Am Acad Child Adolesc Psychiatry 2000; 39:49 – 58.

Tonge B, King N, Klimkeit E, Melvin G, Heyne D, Gordon M. 19.

The self-effi cacy questionnaire for depression in adolescents

(SEQ-DA). Development and psychometric evaluation. Eur Child Adolesc Psychiatry 2005; 14:357 – 363. Reynolds CR, Richmond BO. What I think and feel: a revised 20.

measure of children ’ s manifest anxiety. J Abnorm Child Psychol 1978; 6:271 – 280.

Reynolds CR, Richmond BO. 21. Revised Children ’ s Manifest Anxiety Scale: Manual. Los Angeles: Western Psychological Services, 1985.

Castaneda A, McCandles R, Palermo DS. The children ’ s form of 22.

the manifest anxiety scale. Child Dev 1956; 27:317 – 326. Dozois DJ, Dobson KS, Ahnberg JL. A psychometric evaluation of 23.

the Beck depression inventory – II. Psychol Assess 1998; 10:83 – 89. Sharpley CF, Rogers HJ. Preliminary validation of the Abbreviated 24.

Spanier Dyadic Adjustment Scale: some psychometric data regard-

ing a screening test of marital adjustment. Educ Psychol Meas 1984; 44:1045 – 1049.

Sharpley CF, Cross DG. A psychometric evaluation of the Spanier 25.

Dyadic Adjustment Scale. J Marriage Fam 1982; 44:739 – 741. Byles J, Byrne C, Boyle MH, Offord DR. Ontario child health 26.

study: reliability and validity of the general functioning subscale

of the McMaster Family Assessment Device. Fam Process 1988; 27:97 – 104.

Tamplin A, Goodyer IM. Family functioning in adolescents at high 27.

and low risk for major depressive disorder. Eur Child Adolesc Psychiatry 2001; 10:170 – 179. Tamplin A, Goodyer IM, Herbert J. Family functioning and parent 28.

general health in families of adolescents with major depressive

disorder. J Affect Disord 1998; 48:1 – 13. SPSS 11 for Windows, Rel. 11.0. 2001. Chicago: SPSS. 29.

Birmaher B, Brent DA, Kolko D, Baugher M, Bridge J, Holder D, 30.

Iyengar S, Ulloa RE. Clinical outcomes after short-term psycho-

therapy for adolescents with major depressive disorder. Arch Gen Psychiatry 2000; 57:29 – 36. Brent DA, Birmaher B, Kolko D, Baugher M, Bridge J. Subsyn-31.

dromal depression in adolescents after a brief psychotherapy trial:

course and outcome. J Affect Disord 2001; 63:51 – 58. Vostanis P, Feehan C, Grattan E, Bickerton W-L. A randomised 32.

controlled out-patient trial of cognitive – behavioural treatment

for children and adolescents with depression: 9 month follow-up.

J Affect Disord 1996; 40:105 – 116. Clarke G, Hops H, Lewinsohn PM, Andrews J, Seeley JR, Williams 33.

J. Cognitive – behavioral group treatment of adolescent depression:

prediction of outcome. Behav Ther 1992; 23:341 – 354. Parker G. Critique of the guidelines for the treatment of depres-34.

sion: fl aws in the construction. Aust N Z J Psychiatry 2004; 38: 885 – 890.

Lewinsohn PM, Solomon A, Seeley JR, Zeiss A. Clinical implica-35.

tions of ‘ subthreshold ’ depressive symptoms. J Abnorm Psychol 2000; 109:345 – 351.