Cognitive Behaviour Therapy, 2014

Vol. 43, No. 4, 275–288, http://dx.doi.org/10.1080/16506073.2014.947316

CBT for Children with Depressive Symptoms:

A Meta-Analysis

Alexandra Arnberg1 and Lars-Göran Öst1,2

1
Department of Psychology, Stockholm University, Stockholm, Sweden; 2Department of
Clinical Neuroscience, Karolinska Institutet, Stockholm, Sweden
Abstract. Pediatric depression entails a higher risk for psychiatric disorders, somatic complaints,
suicide, and functional impairment later in life. Cognitive behavior therapy (CBT) is recommended
for the treatment of depression in children, yet research is based primarily on adolescents. The present
meta-analysis investigated the efficacy of CBT in children aged 8–12 years with regard to depressive
symptoms. We included randomized controlled trials of CBT with participants who had an average
age of #12 years and were diagnosed with either depression or reported elevated depressive
symptoms. The search resulted in 10 randomized controlled trials with 267 participants in
intervention and 256 in comparison groups. The mean age of participants was 10.5 years. The
weighted between-group effect size for CBT was moderate, Cohen’s d ¼ 0.66. CBT outperformed
both attention placebo and wait-list, although there was a significant heterogeneity among studies
with regard to effect sizes. The weighted within-group effect size for CBT was large, d ¼ 1.02. Earlier
publication year, older participants, and more treatment sessions were associated with a larger effect
size. In conclusion, the efficacy of CBT in the treatment of pediatric depression symptoms was
supported. Differences in efficacy, methodological shortcomings, and lack of follow-up data limit the
present study and indicate areas in need of improvement. Key words: youth; depression; cognitive
behavioral therapy; psychotherapy; review.

Received 1 June 2014; Accepted 18 July 2014

Correspondence address: Lars-Göran Öst, Department of Psychology, Stockholm University, S-106 91

Stockholm, Sweden. Tel: þ46 737121285. Email: ost@psychology.su.se

Depression in children is a serious condition the risk of new episodes in adulthood

that entails an increased risk of psychological (Costello et al., 2002; Weissman et al., 1999).
and physiological ill-health in the future, Thus, there is a large need of knowledge about
suicide, and social adjustment problems treatment of children already at an early age.
(Jonsson et al., 2011; Klein, Lewinsohn, & The proportion of children suffering from
Seeley, 1997). Depressive symptoms during depression increases with age. The lifetime
childhood have been shown to be associated prevalence of depression in preschool children
with anxiety symptoms, low self-esteem, is about 1%, in school children about 3%, and
externalizing behavior, alcohol and drug use, in adolescents about 6% (Costello, Erkanli, &
and low school achievements (Harter, 1990; Angold, 2006). Up to puberty, the prevalence
King, Ollendick, & Gullone, 1991; Rowlison is the same in boys and girls, but thereafter it is
& Felner, 1988). Without treatment, both significantly higher in girls (Weissman, War-
depression and subclinical depressive symp- ner, Wickramaratne, Moreau, & Olfson,
toms may progress into chronic depression or 1997). Lifetime risk of depression in the
more severe conditions (DuBois, Felner, adult population is 5 –25% for women and
Bartels, & Silverman, 1995; Jonsson et al., 3 –10% for men (Swedish Council on Tech-
2011). The risk of developing other affective nology Assessment in Health Care [SBU],
disorders is especially high, rather than 2004).
psychiatric problems in general (Fonagy, In contrast to adolescent and adult
Target, Cottrell, Phillips, & Kurtz, 2005), depression, depressive disorders during child-
and depression in the adolescence increases hood may be expressed in unspecific symptoms

q 2014 Swedish Association for Behaviour Therapy

276 Arnberg and Öst COGNITIVE BEHAVIOUR THERAPY

rather than symptoms specific to the proto- context of the child has been shown to be an
typical clinical presentation according to the important component in order to understand
diagnostic features: young children with emergence and maintenance of depression in
depression complain about somatic symp- children, and the child’s parents are often
toms, e.g., stomach pain, chest pain, and participating in the treatment to various
headache. Worry, fear of the dark, irritability, degrees (Abela & Hankin, 2008). Current
feelings of discomfort, and separation anxiety research focuses on implementing a develop-
are also common (Birmaher et al., 1996). mentally sensitive psychotherapy for
Expressions of depression often touch many depressed children (Kovacs & Lopez-Duran,
other conditions, e.g., anxiety disorders 2012; Luby, Lenze, & Tillman, 2012). For
(Cummings, Caporino, & Kendall, 2014). example, an ongoing study of Contextual
That children show signs of low mood through Emotional Regulation Therapy is based on the
grief reactions and sadness is a part of normal proposition that dysphoric mood is the most
development and does not mean that major salient feature of clinical depression, which
depression is at hand. starts as a response to an initiating stressful
Comorbidity is common for depressive event or process (Kovacs & Lopez-Duran,
disorders in children, and the characteristic 2012). Whether or not the dysphoria develops
symptoms are found in many child and into a disorder is suggested to depend mainly
adolescent psychiatric diagnoses (Fonagy on how the affected youngster responds to the
et al., 2005). In children with the diagnosis of emotion and the kind of support that is
major depression, 40 – 70% fulfill criteria for available. Regarding psychological treatment
another psychiatric disorder and at least 20% of children with depressive symptoms, CBT is
has three or more comorbid disorders (Bir- the only treatment for which there are several
maher et al., 1996). Comorbidity has impli- studies of its efficacy, whereas interpersonal
cations for assessment and treatment. In the therapy also may have an effect for adolescents
case of anxiety, measures of anxiety and with depressive symptoms (David-Ferdon &
depression are highly correlated and symp- Kaslow, 2008; SBU, 2004; Watanabe, Hunot,
toms of anxiety tend to predate depressive Omori, Churchill, & Furukawa, 2007),
symptoms (Cummings et al., 2014). Comor- although the long-term effects have not been
bidity with conduct disorder is strongly investigated closely enough (SBU, 2004).
associated with problems in all sorts of social Importantly, however, guidelines and rec-
relationships and is linked to an increased risk ommendations for adolescents and younger
of suicide (Angold & Costello, 2001). children are based primarily on treatment
Thirty years ago, it was considered improb- results for adolescents (e.g., SBU, 2004)
able that children could suffer from depression despite the fact that it is reasonable to believe
(Fonagy et al., 2005). After the first random- that the effect is not equal for younger
ized controlled trial (RCT; Butler, Mietzitis, children: symptoms, cognitive development,
Friedman, & Cole, 1980), the second one was and the included treatment components differ
not published until seven years later (Stark, between children and adolescents (Compton
Reynolds, & Kaslow, 1987). Treatment studies et al., 2004; Michael & Crowley, 2002;
in children with depression lag far behind Watanabe et al., 2007; Weisz, McCarty, &
those for adults. Historically, the treatment of Valeri, 2006). A recent meta-analysis con-
children has applied those methods that have cluded that CBT and interpersonal therapy are
been shown effective for adults, primarily probably efficacious when compared to no
medication and cognitive behavior therapy treatment, but the authors qualified their
(CBT; Kaslow & Thompson, 1998). However, findings by noting that it was likely that the
a gradual methodological development has effect seen in their analysis was restricted to
taken place with focus on a treatment adjusted children 12 –18 years of age (Watanabe et al.,
to the child’s psychological developmental 2007). Furthermore, in most published studies
level. For example, the educative component the children have not been diagnosed
about the chain of emotion-thought-action (Asarnow, Scott, & Mintz, 2002; Butler et al.,
may use various animal characters and is 1980; de Cuyper, Timbremont, Braet, De
shared in a language carefully adapted to the Backer, & Wullaert, 2004; Liddle & Spence,
age of the patient. Furthermore, the social 1990; Stark et al., 1987; Weisz, Thurber,
VOL 43, NO 4, 2014 CBT for Children with Depressive Symptoms 277

Sweeney, Proffitt, & LeGagnoux, 1997) and in analysis of the child studies did not find a
a number of studies prevention samples have significant difference between treatment and
been included (e.g., Gillham, Reivich, Jaycox, control groups (RR ¼ 1.51, p ¼ 0.08), which
& Seligman, 1995; Jaycox, Reivich, Gillham, & corresponds approximately to a Cohen’s d ES
Seligman, 1994). Including prevention of 0.3 (Chinn, 2000).
samples, however, may hamper the generaliz- It can be concluded that earlier meta-
ations to be made, as they seem to have greater analyses of different forms of psychotherapy
gains from treatment as compared to clinical for children and adolescents with depression
samples (Klein, Jacobs, & Reinecke, 2007). or subclinical depressive symptoms have
According to a meta-analysis by Durlak, obtained a large variation in mean ES:
Weissberg, Dymnicki, Taylor, and Schellinger Cohen’s d between 0.34 (Weisz et al., 2006)
(2011), preventive school-based interventions and 1.27 (Lewinsohn & Clarke, 1999). Even if
have a positive impact on youths’ mental it is unclear how Lewinsohn and Clarke (1999)
health. Also, for lasting benefits from pre- calculated their ES due to insufficient presen-
ventive programs, interventions for preadoles- tation of the method, it is important to better
cent children may be more effective than those understand the underlying reasons for this
for adolescents (i.e., grade 6 vs. grade 9; large variation. Differences in ESs have
Barrett, Farrell, Ollendick, & Dadds, 2006). previously been shown to covary with the
The current meta-analysis differs from publication year of the study: in a cumulative
previous meta-analyses in that it specifically meta-analysis of CBT treatments for adoles-
concerns the effect of CBT for children. Earlier cents, Klein et al. (2007) reported that the ES
meta-analyses have included all types of was reduced as years go by. When Klein et al.
psychological treatment (Erford et al., 2011; (2007) investigated the differences between
Watanabe et al., 2007; Weisz et al., 2006) and studies, these primarily consisted of smaller
studies of both children and adolescents ESs being associated with intent-to-treat
(Lewinsohn & Clarke, 1999; Luby et al., analyses, comparison between CBT and
2012; Watanabe et al., 2007; Weisz et al., another active treatment, treatment done in
2006). In the Michael and Crowley (2002) clinical contexts, and application of more
meta-analysis, a moderate effect size (ES) of strict methodological procedures.
0.65 was found for the controlled studies of Weisz et al. (2006) found a relatively large
treatments for children with depression. How- ES at follow-up two to three months after the
ever, of the nine included studies, three termination of therapy, but essentially no
investigated other interventions than CBT. treatment effect was seen one year posttreat-
In Weisz et al. (2006), a small ES of 0.41 was ment. A diminishing effect with time was also
found for children studies, although the found in the Michael and Crowley (2002) and
authors also included prevention samples in Watanabe et al. (2007) meta-analyses, where
their meta-analysis. The findings from the the latter study found that the effect of
Weisz et al. meta-analysis for children and psychotherapy was no longer significant at
adolescents were highly similar to those of a follow-up six months or longer. Erford et al.
recent meta-analysis of all types of psychologi- (2011) stated that depression treatments seem
cal treatments by Erford et al. (2011). Interest- to have significant effects for perhaps up to
ingly, a subgroup analysis in the Erford et al. two years, but we have far too few follow-
study suggests that CBT was not significantly up studies to ascertain any long-term benefit.
more effective than other treatments. This Major depression remits spontaneously within
finding was also reported by Spielmans, Pasek, nine months for the majority of patients,
and McFall (2007) who found no difference whereas the relapse risk is as high as 50%
between CBT- and non-CBT-based treatments within two years (Fonagy et al., 2005). Hence,
for youths with depression or anxiety. Unfor- it is important to know much more about the
tunately, none of these two reviews conducted effects of CBT in relation to spontaneous
separate analysis for children. Also in contrast remission and if the effect is maintained across
to the current analysis, Watanabe et al. (2007) time. There is, as far as we know, no study that
included all psychotherapeutic methods and has analyzed the long-term effects of CBT
used relative risk of response (RR) as the specifically in the group of patients up to 13
primary outcome measure. Their separate years of age.
278 Arnberg and Öst COGNITIVE BEHAVIOUR THERAPY

The aim of the current meta-analysis was to cINFO), and from previous meta-analyses
investigate the efficacy of CBT for depressive and relevant review papers (Fonagy et al.,
symptoms in children 8 –12 years of age. The 2005; Öst, 2010; Whittington, Kendall, &
specific research questions were the following: Pilling, 2005). The reference lists of these
publications and of studies with a potential to
1. What is the mean ES of CBT for children be included were manually searched for
with depressive symptoms? further studies (Figure 1).
2. Is the ES associated with the publication The search in PsycINFO included the
year of the study? following search terms: “((treatment or therapy
3. Is the ES maintained at follow-up? or intervention*) AND (depression OR
depressive OR mood) AND (children* OR
In addition to these primary questions, teen* OR preadolescen* OR adolescen*)).ab.
secondary analyses will be performed on AND ((randomised OR randomized).ab. OR
variables, which, recent meta-analyses in the (randomised OR randomized).ti.).” The
area indicate, show a possible relation to the searches were limited to journal articles with
outcome, such as age, gender, severity of peer-review or unknown peer-review status and
symptoms, treatment time, number of ses- were written in English. The search yielded 561
sions, quality of the study, attrition, and publications. The search in PubMed included
therapist experience (Klein et al., 2007; the following search terms: “(treatment[Title/
Michael & Crowley, 2002; Weisz et al., 2006). Abstract] OR therapy[Title/Abstract] OR
intervention*[Title/Abstract]) AND (children
[Title/Abstract] OR teen*[Title/Abstract] OR
Methods youth*[Title/Abstract]) AND (depression
Literature review [Title/Abstract] OR depressive[Title/Abstract]
Studies published between 1980 and Decem- OR mood[Title/Abstract]).” The search was
ber 2013 were retrieved from major databases limited to publications indexed as journal
in medicine (PubMed) and psychology (Psy- articles of RCTs, and they were written in

Figure 1. Flowchart of inclusion and exclusion of studies in the meta-analysis.

VOL 43, NO 4, 2014 CBT for Children with Depressive Symptoms 279

English, yielding 608 publications. The final depression. Because of the few studies of CBT
search was conducted on 7 December 2013. with children samples, it was deemed imposs-
The abstracts were read and full text ible to include only samples where a diagnosis
versions were retrieved if it was clear or had been established. Potentially, including
possible that the publication should be both clinical and subclinical samples could
included. The inclusion criteria were set a increase the heterogeneity of the present meta-
priori and were as follows: the study has to analysis and, thus, negatively affect the
evaluate CBT, the sample’s mean age must be external validity of the findings. However,
below 13 years, the study has to be published previous meta-analyses that have performed
in a English-language peer-reviewed journal, subgroup analyses on studies of children also
the participants should have depressive symp- included prevention samples. Hence, the
toms above a threshold score on self-rating studies included in the present analysis
scales of depression or fulfill criteria for a would be more homogenous than in previous
diagnosis of depression, and a RCT design has meta-analyses.
to be used. RCT design. Reviews of RCTs across health-
Evaluating CBT. CBT was defined as psy- care disciplines suggest that the inclusion of
chotherapeutic methods based on empirical low-quality RCTs is related to higher pooled
and theoretical research on behavioral, cogni- ES (Moher et al., 1998). Thus, the quality of
tive, or social psychology, and developed to included studies was assessed and analyzed as
change emotions and behaviors by teaching a moderator variable.
the children to affect their thoughts and overt We used a number of exclusion criteria.
behaviors in an active and problem-focused Pilot studies where the participants were later
way. Also, the authors should have specified included in a full-scale trial were excluded, as
the intervention as CBT. We decided to were studies with prevention samples. For the
include all CBT methods because there is no Children’s Depression Inventory (CDI) to
clear consensus as to whether specific CBT identify children with depression with accep-
techniques or methods are more effective than table sensitivity, a threshold score of . 13 has
others in the treatment of children with been recommended (Kovacs, 1992). However,
depression (Fonagy et al., 2005). one study that was included herein used a CDI
A sample mean age , 13 years. As in previous threshold score . 8 (Asarnow et al., 2002),
meta-analyses, all individuals under 13 years which is lower than the recommended cutoff.
of age were classified as children (Michael & The study was included because 48% of the
Crowley, 2002; Weisz et al., 2006), although sample was judged to fulfill tentative criteria
studies that had included adolescents but for depression and 39% for depression NOS
where the mean age of the sample was below (Asarnow et al., 2002).
13 years were included.
Published in an English-language peer-reviewed Study characteristics
journal. Peer review may be seen as a proxy Table 1 summarizes descriptive data of the 10
measure of methodological rigor in the included studies. The studies were published
included studies (Klein et al., 2007). This between 1980 and 2009; six studies were from
inclusion criterion is further motivated in that the USA and one each from Canada,
Weisz et al. (2006) in their meta-analysis found Australia, England, and Belgium. Four
no support for differences in treatment effects studies included samples diagnosed with
between published and unpublished studies, depression. All other studies applied CDI as
which was true also for a study of publication an outcome measure and used an inclusion
bias in clinical research (Easterbrook, Berlin, threshold of 8 or 13 points. The majority of the
Gopalan, & Matthews, 1991). Other reviews studies (n ¼ 8) recruited participants from
of meta-analyses, however, have found that schools and two studies recruited from mental
published studies have 15 – 33% larger ESs health clinics. The most common methods
compared to unpublished studies (McAuley, used in the interventions were psycho-edu-
Pham, Tugwell, & Moher, 2000). cation about the association between
Participants should have depressive symptoms thoughts, emotions, and behaviors, problem-
above a threshold score in self-rating scales of solving strategies, and behavioral activation.
depression or fulfill criteria for a diagnosis of Two studies included a component specific to
 280

Table 1. Study characteristics

Symptom
severity Follow-up months
Arnberg and Öst

Study Recruitment Inclusion criteria T-scorea Treatment (n) Control (n) Outcome (% attrition)b
Butler et al. School Score . 1.5 SD for $2 of 4 69.9 Role play (14) AP (14) CDI –
(1980) measuresc Cognitive restructuring (14) WL (14)
Stark et al. School CDI $ 13 67.2 Self-control (9) WL (9) CDI 2 (– )
(1987) Behavior problem-solving (10) 2 (– )
Kahn and School CDI $ 15 þ diagnosis 74.5 CBT (17) WL (16) CDI 1 (0)
Kehle (1990) from clinical interview Relaxation training (17) 1 (0)
Self-modeling interventions (17) 1 (0)
Liddle and School CDI $ 19 66.0 Social competence training (11) AP (10) CDI 2 (36)
Spence (1990) WL (10)
Vostanis et al. Clinic Diagnosis from K-SADS 67.8 CBT (29) AP (28) MFQ 24 (5)
(1996)
Weisz et al. School CDI $ 11d 61.8 Primary and secondary control WL (32) CDI 9 (– )
(1997) enhancement training (16)
Asarnow et al. School CDI $ 8 63.3 CBT þ family education (12) WL (11) CDI –
(2002)
Nelson et al. School Diagnosis from K-SADS 56.0 Video conference (14) Face-to- CDI –
(2003) face (14)
de Cuyper School CDI $ 11d 55.9 Taking action (9) WL (11) CDI 4 (18)
et al. (2004)
Fristad et al. Clinic, Diagnosis from ChIPS and 77.7 MF-PEP þ TAU (78) WL þ TAU CDRS-R 4 (37) þ 18 (– )
(2009) adverts P-ChIPS (87)
Note. AP ¼ attention placebo; CDI ¼ Children’s Depression Inventory; CDRS-R ¼ Children’s Depression Rating Scale-Revised; ChIPS ¼ children’s interview
for psychiatric syndromes; P-ChIPS ¼ parent version of ChIPS; K-SADS ¼ schedule of affective disorders and schizophrenia for school-age children;
MF-PEP ¼ Multifamily Psychoeducational Psychotherapy; NTC ¼ no treatment control; TAU ¼ treatment as usual; WL ¼ wait-list.
a
The group mean scores were transformed to T-scores according to normative population samples.
b
Follow-up interval with respect to the treatment group.
c
Self-esteem scale, CDI, Moyal-Miezitis Stimulus Appraisal Questionnaire and Nowicki-Strickland Locus of Control Scale for Children.
d
CDI without the suicide item.
COGNITIVE BEHAVIOUR THERAPY
VOL 43, NO 4, 2014 CBT for Children with Depressive Symptoms 281

parents. Of seven studies with a follow- attrition from post to follow-up, treatment
up assessment, only four reported data on format (individual, group), treatment dur-
the number of treatment group participants ation, treatment context, study quality, and
who completed the assessment and the time to therapist experience. Age was coded as mean
follow-up varied from one to nine months. and range; if the mean age was not reported, it
was assumed to be the midpoint of the range.
Study quality assessment To assess symptom severity, all scores were
To assess the potential effect of an interven- transformed to a T-distribution because two
tion, the studies that evaluate this method studies did not report CDI ratings. First, the
should be well designed and of high methodo- study by Vostanis, Feehan, Grattan, and
logical quality (Chambless & Ollendick, 2001). Bickerton (1996) reported only the Mood
The quality of the included studies was and Feelings Questionnaire (MFQ), for which
therefore assessed using the psychotherapy norms were not reported in the original
outcome study methodology rating form by publication (Angold, Costello, Messer, &
Öst (2008). The rating form has previously Pickles, 1995), and so a T-score was arrived
been used in a meta-analysis by Öst (2008) to at by computing the mean value of all studies
assess the study quality and was used as a that included the CDI. Second, the study by
moderator variable. The rating form includes Fristad, Verducci, Walters, and Young (2009)
22 items: (1) clarity of description of partici- included the Children’s Depression Rating
pants, (2) severity/chronicity of the disorder, Scale-Revised (CDRS-R), and a T-score was
(3) representativeness of participants, (4) calculated from a normative sample (Poz-
reliability of diagnosis, (5) specificity of out- nanski et al., 1984). Treatment duration was
come measures, (6) validity and reliability of coded as the number of planned sessions times
outcome measures, (7) use of assessors blind to the session duration in minutes. One study did
assignment, (8) method-specific training of not report the session duration (Vostanis
assessors, (9) group allocation sequence, (10) et al., 1996), which then was assumed to be
study design, (11) power analysis, (12) timing 50 minutes. The number of actual sessions that
of assessments, (13) manualized, replicable, were delivered was not coded because of a lack
specific intervention programs, (14) number of of reported data in the included studies.
therapists, (15) therapist method-specific Therapist experience was assessed by item
training and experience, (16) assessment of 15 in the quality rating form (Öst, 2008).
adherence, (17) therapist competence assess- To achieve the highest rating (2), the therapists
ment, (18) assessment of concurrent treat- needed to have extensive clinical experience
ments, (19) methods to adjust for attrition, with both depression and the relevant treat-
(20) statistical analysis and presentation of ment method. The coded data were assessed
results, (21) clinical significance, and (22) and reassessed twice by both authors to
equality of therapy duration (only for designs increase reliability and arrive at consensus.
with treatment arms $ 2). The sum score was
used to assess study quality. Primary outcome
The quality of all the included studies was The primary outcome measure was self-report
assessed by the first author who is qualified to of depressive symptoms since this was used in
use the rating form after receiving training all studies. The CDI (Kovacs, 1992) was used
from the author of the rating form. The form in eight of the included studies. It is a self-
has good inter-rater reliability for both the rating scale for depression in the 7– 17 year age
total rating (ICC ¼ 0.92) and for each item range and includes 27 items. The CDI has
(kappa mean ¼ 0.75, range ¼ 0.50 –1.00) and satisfactory psychometric properties in the
good internal consistency (Cronbach’s case of internal consistency and excellent
a ¼ 0.86) (Öst, 2008). validity generalization (Dougherty, Klein,
Olino, & Laptook, 2008). And many studies
Coding procedure have reported good short-term test –retest
The following variables were selected a priori reliability (Brooks & Kutcher, 2001; Kovacs,
to be retrieved in order to be used in 1992). CDI has been widely used both in
moderator analyses: age, sex, symptom sever- clinical practice and in research, and shows
ity at baseline, attrition from baseline to post, high sensitiveness to change. One study
282 Arnberg and Öst COGNITIVE BEHAVIOUR THERAPY

included the MFQ (Vostanis et al., 1996) and 50% of the maximum score on the quality
one study used the CDRS-R (Fristad et al., rating form (range 18– 69%) and only two
2009): both measures have provided psycho- studies were judged to be of good quality.
metric properties that are similar to those of There was a uniform distribution of studies
the CDI. with experienced therapists, somewhat experi-
enced therapists, and not experienced thera-
Statistical analyses pists. For seven studies, a CDI score was
The computer software Comprehensive Meta- reported at follow-up assessment with an
Analysis, version 2.2 (Biosoft, 2006), was used average time to follow-up of 17.7 weeks after
to calculate ESs and heterogeneity and to the treatment ended (range 4 –36 weeks).
perform meta-regression analyses for the
moderator variables. Corrections were made Controlled effect sizes
for small samples by calculating Hedges’ g as Based on 16 comparisons, CBT had a
the ES. To evaluate the homogeneity of ESs, moderate ES (g ¼ 0.66) at postassessment
we used the Q-statistic; if statistically signifi- and was associated with a better outcome than
cant heterogeneity was found, this was both attention placebo (AP) and wait-list
examined by moderator analyses. A random- (WL); however, there were substantial differ-
effects model was used if the Q-statistic was ences in ES among the studies (Figure 2). For
significant. Comparison of the ESs at post- CBT versus AP (n ¼ 5), there was a moderate
treatment and follow-up was done by a ES (g ¼ 0.54) favoring CBT. For CBT versus
dependent t-test performed in SPSS v. 19 for WL (n ¼ 11), there was also a moderate ES
Windows (IBM, Chicago, IL). (g ¼ 0.70) in favor of CBT (Table 2).
Sensitivity analysis. Two of the included
studies differed from the others in specific
Results ways. Vostanis et al. (1996) included youth
In total, the 10 included studies included 267 up to 17 years of age, and Asarnow et al.
participants in active treatment groups and (2002) used a CDI score of 8 as inclusion
256 participants in control groups. The mean criterion. When these two studies were
age of the participants was 10.5 years (range excluded, the overall ES increased somewhat
9.2 –12.7 years). In six studies, all participants to g ¼ 0.72 [0.41; 1.02], z ¼ 4.62, p , 0.0001.
were , 13 years of age (Asarnow et al., 2002; The CBT versus AP ES also increased to some
Butler et al., 1980; de Cuyper et al., 2004; extent, g ¼ 0.71 [0.01; 1.41], z ¼ 1.99,
Liddle & Spence, 1990; Stark et al., 1987; p ¼ 0.046, and the CBT versus WL ES to
Weisz et al., 1997). In two studies, there were g ¼ 0.73 [0.41; 1.05], z ¼ 4.49, p , 0.0001.
participants who were up to 14 years of age The removal of the Vostanis et al. and
(Kahn & Kehle, 1990; Nelson, Barnard, & Asarnow et al. studies yielded nonsignificantly
Cain, 2003). In one study, the participants higher ESs than when they were included, and
were up to 17 years old (Vostanis et al., 1996). thus we decided to keep them in the meta-
One study did not report the age range analysis.
(Fristad et al., 2009). The proportion of boys
was on average 54.2% (range 25– 76%). There Publication bias
was no difference in symptom severity Publication bias might pose a potential threat
( p ¼ 0.46) between studies with diagnosed to the validity of the findings. Therefore, we
samples (n ¼ 4; T-score M ¼ 69) and not estimated the number of unpublished studies
diagnosed samples (n ¼ 6; T-score M ¼ 64) with null results needed to render the observed
(see Table 1). The attrition ranged from 0% to ES nonsignificant, by using the methods
37%. Eight studies provided treatment in proposed by Rosenthal (1991) and Orwin
group format. The number of sessions ranged (1983). The number of unpublished studies
from 8 to 18 and the length of the sessions with null findings needed is 173, whereas
from 50 to 90 minutes. The total treatment 24 studies with a g-value of 0.00 are needed to
duration was on average 656 minutes (range reduce the ES to a clinically insignificant level
450– 1080 minutes). The treatment context (g ¼ 0.20).
was predominantly in a school environment However, the trim-and-fill method by
(n ¼ 7). The study quality was on average Duval and Tweedie (2000) indicated that in
VOL 43, NO 4, 2014 CBT for Children with Depressive Symptoms 283

Figure 2. Forest plot for comparisons of CBT versus attention placebo (AP) and wait-list (WL). Note.
CI ¼ confidence interval; CDI ¼ Children’s Depression Inventory; CDRS-R ¼ Children’s Depression
Rating Scale-Revised; MF-PEP ¼ Multifamily Psychoeducational Psychotherapy; MFQ ¼ Mood and
Feelings Questionnaire; TAU ¼ treatment as usual.

Table 2. Between-subjects effect sizes posttreatment

Comparison k g 95% CI z P Q p
Total 16 0.655 0.40 – 0.91 4.96 , 0.001 35.5 0.002
CBT vs AP 5 0.536 0.02 – 1.05 2.05 0.04 13.8 0.008
CBT vs WL 11 0.695 0.40 – 1.00 4.54 , 0.001 16.2 0.09
Note. AP ¼ attention placebo; CI ¼ confidence interval; g ¼ Hedges’ g effect size; k ¼ number of comparisons;
Q ¼ measure of heterogeneity; WL ¼ wait-list.

order to obtain a symmetric funnel plot, four 0.78]. There was a large uncontrolled ES from
studies should be imputed to the left of the pretreatment to follow-up for the 10 treatment
mean, which would reduce the Hedges’ g to groups, g ¼ 1.34, 95% CI [0.97; 1.71]. From
0.44 (95% CI [0.15 –0.73]), which is still posttreatment to follow-up, there was a
significantly different from zero. The rank subsequent improvement in six treatment
correlation test of publication bias using conditions. However, the increase in uncon-
Kendall’s tau (Begg & Mazumdar, 1994) was trolled ES for those studies having follow-up,
not significant (0.21, z ¼ 1.13, p ¼ 0.26). from 1.20 at post to 1.44 at follow-up, was not
Thus, it can be concluded that publication significant [t(9) ¼ 1.73, p ¼ 0.12]. There was
bias is probably not a significant problem for heterogeneity in ESs among studies at follow-
the current meta-analysis. up, Q ¼ 22.0, p ¼ 0.009.

Uncontrolled effect sizes Moderator analyses

The estimation of within-group (uncontrolled) Meta-regression analyses were carried out
ES was based on 15 treatment groups, 4 AP using a fixed effect model (Borenstein, 2009).
groups, and 7 WL groups. The ES was large Five of the moderator variables were signifi-
for CBT, g ¼ 1.07, 95% CI [0.70; 1.43], small cantly associated with the ES (Table 3).
for AP, g ¼ 0.38, 95% CI [0.13; 0.62], and Publication year was a negative predictor in
moderate for WL, g ¼ 0.54, 95% CI [0.31; that later year (newer studies) was associated
284 Arnberg and Öst COGNITIVE BEHAVIOUR THERAPY

Table 3. Meta-regression analyses of moderator variables

Moderator B SE z p
Publication year 2 0.0284 0.0081 2 3.506 0.0005
Mean age 0.1571 0.0798 1.967 0.0492
Sex, % girls 0.0102 0.0061 1.667 0.0955
Symptom severity 2 0.0065 0.0144 2 0.453 0.6509
Number of sessions 0.2751 0.0577 4.766 0.0000
Study quality 2 0.0101 0.0099 2 1.018 0.3089
Attrition pre– post 2 0.0230 0.0071 2 3.251 0.0012
Therapist experience 2 0.3126 0.1084 2 2.884 0.0039

with a lower ES. Age was a positive predictor, younger children separately (d ¼ 0.65).
indicating that the higher the mean age of the A number of the studies in the present meta-
sample, the higher was the ES. The removal of analysis were also included in that meta-
the Vostanis et al. (1996) study led to an even analysis. Michael and Crowley had nine
higher B-coefficient (0.38) than when it was studies, six of which are included in the
included (0.16), so we kept the study in the present meta-analysis (the remaining three
analysis. Another positive predictor was the studies did not use CBT). Weisz et al. (2006)
number of therapy sessions; a higher number found a lower ES (d ¼ 0.41), which may be
of sessions were associated with a larger ES. due to the fact that they had a somewhat
Attrition and therapist experience were nega- younger sample or that they used a procedure
tively associated with ES; higher dropout rate for calculating ES that has been criticized
and higher therapist experience were both (Klein et al., 2007).
associated with lower ES. Finally, proportion The present meta-analysis did not find a
of girls and mean symptom severity of the significant correlation between pretreatment
sample, as well as methodological quality of symptom severity and ES. However, there
the study, did not emerge as significant were indications that recruitment from and
moderators of ES. treatments in a clinical context were associated
with a lower ES, which has been reported
earlier (Klein et al., 2007). A possible expla-
Discussion nation may be that comorbidity is higher in a
The present meta-analysis investigated the clinical population, which can contribute to
effect of CBT on depressive symptoms in maintenance of symptoms. It can be con-
young children, the long-term effects of the cluded that we so far know little about the
treatment, and potential moderators of the effect of CBT for clinical samples with a
ES. The controlled ES was moderate (0.66) diagnosed depressive disorder.
when CBT was compared to a control For children and adolescents, Weisz et al.
condition; CBT was significantly better than (2006) did not find a difference in ES between
AP (0.54) and WL control (0.70). Regarding CBT and other psychosocial treatment
pre- to postassessment, CBT had a large methods. However, the database for calcu-
within-group ES (1.07). The effect was large lation of ES almost exclusively contained CBT
(1.44) also at follow-up, although the scarcity studies, which makes the conclusion of no
of follow-up data limits the reliability of the differences tentative at best.
finding. Older studies, older age of the The analysis of moderators showed that the
participants, and more treatment sessions ES was lower in later compared to earlier
were associated with a larger effect. None- published studies. This finding corroborates
theless, differences in efficacy and methodo- that of Klein et al. (2007) in their cumulative
logical shortcomings in the studies indicate meta-analysis of CBT for adolescents. There
areas in need of improvement. are a number of methodological differences
The controlled ES is in agreement with that between the studies in the present meta-
found in the meta-analysis by Michael and analysis, which may be a possible reason for
Crowley (2002) that analyzed studies with the lowering of the ES for CBT. In contrast to
VOL 43, NO 4, 2014 CBT for Children with Depressive Symptoms 285

what Klein et al. (2007) reported, the present than in earlier meta-analyses due to exclusion
meta-analysis did not find a relation between of prevention studies and the use of a cutoff
study quality and ES. It should be mentioned score on a depression rating scale or clinical
that the validity of the research quality rating diagnosis as inclusion criteria. It should be
scale used in the present study has not noted that there was not a significant
formally been evaluated. One difference difference in degree of depressive symptoms
compared to other quality scales is that the between diagnosed and undiagnosed samples.
one used here does not weigh different quality The decision to include children without a
variables when the total score is calculated. depression diagnosis was motivated by the few
Thus, the result concerning quality of the studies in the area. In order to increase
studies should be interpreted cautiously. homogeneity among studies, it was decided to
The within-group ES increased from post- use CDI as the primary outcome measure.
assessment (g ¼ 1.20) to follow-up assessment In the analysis, the size of the study was also
(g ¼ 1.44), but the variations were large taken into consideration and a correction for
concerning both ES and time to follow- small sample size was done by transforming
up. Furthermore, the description of attrition Cohen’s d to Hedges’ g. Despite taking these
left a lot to be desired. Only three studies had a measures, the ESs of the included studies were
follow-up period of nine months or longer. heterogeneous.
Taking these factors into account, the results The calculation of fail-safe N, the trim-and-
are promising but equivocal concerning long- fill method by Duval and Tweedie (2000), and
term effects. The findings regarding follow- the rank correlation test of publication bias
up in the present study partly concur with using Kendall’s tau indicated that publication
previous meta-analyses of depression that bias was improbable but it cannot be ruled out
have aggregated children and adolescent completely. However, it is worth noticing that
samples, which show a relatively large ES at Weisz et al. (2006) did not find a difference in
follow-up two to three months after the ES between published and unpublished studies
termination of therapy. However, these in their meta-analysis.
meta-analyses reported that no effect could In conclusion, CBT for children with
be seen after one to two years posttreatment depressive symptoms is a neglected research
(Erford et al., 2011; Weisz et al., 2006), area with a few RCTs. There is a great need for
something the present meta-analysis could not more studies of high quality. Large variations
evaluate since only two follow-ups were longer in ES between studies point to the importance
than one year. The relative lack of long-term of more methodologically stringent studies
follow-up assessment highlights an important with properly diagnosed children, which
area for improvement since we know that would strengthen the translation of research
childhood depression entails an increased risk into clinical practice. So far, the findings are
of psychological and physical problems later based on group treatments, almost exclusively
on. It is particularly important, therefore, to carried out in school settings in the USA,
develop and evaluate treatments for children which hampers the generalizability to clinical
that can prevent depressive symptoms to practice and other countries. In addition,
progress into chronic depression or more more data are needed to understand the effect
severe conditions. of individual treatment and parental involve-
The present meta-analysis has its limi- ment. Future research should adjust the
tations. In order not to risk excluding studies treatments to the developmental level of the
that could contribute to increased knowledge patients to an even larger extent than has been
in the area, few exclusion and generous the case hitherto and should include long-term
inclusion criteria were used. This contributed follow-ups. The research area would be
to the fact that the studies varied concerning developed through the reporting of follow-
treatment format and treatment components, up data to enable reliable assessments of how
which complicate the interpretation of the the treatment effects are maintained in the
results. There were large differences in degree long run. Despite current limitations in our
of symptoms between studies, and we included knowledge of CBT for children with depress-
studies of both diagnosed and undiagnosed ive symptoms, CBT is a well-documented
children. However, the variation was lower treatment, compared to other forms of
286 Arnberg and Öst COGNITIVE BEHAVIOUR THERAPY

therapy, with a moderate ES for children with preadolescents. American Educational Research
depressive symptoms. Journal, 17, 111– 119. doi:10.2307/1162512
Chambless, D.L., & Ollendick, T.H. (2001).
Empirically supported psychological interven-
tions: Controversies and evidence. Annual
Acknowledgements Review of Psychology, 52, 685–716. doi:10.
We thank Filip Arnberg for generously 1146/annurev.psych.52.1.685
Chinn, S. (2000). A simple method for converting
sharing his statistical and language an odds ratio to effect size for use in meta-
knowledge. analysis. Statistics in Medicine, 19, 3127– 3131.
Disclosure statement: The authors have doi:10.1002/1097-0258(20001130)19:22,3127:
declared that no conflict of interest exists. AID-SIM784.3.0.CO;2-M
Compton, S.N., March, J.S., Brent, D., Albano, A.
M., Weersing, V., & Curry, J. (2004). Cognitive-
behavioral psychotherapy for anxiety and
depressive disorders in children and adoles-
References cents: An evidence-based medicine review.
Abela, J.R.Z., & Hankin, B.L. (2008). Handbook of Journal of the American Academy of Child &
depression in children and adolescents. New Adolescent Psychiatry, 43, 930– 959. doi:10.
York, NY: Guilford Press. 1097/01.chi.0000127589.57468.bf
Angold, A., & Costello, E.J. (2001). The epidemiol- Costello, E., Erkanli, A., & Angold, A. (2006). Is
ogy of depression in children and adolescents. there an epidemic of child or adolescent
In I.M. Goodyer (Ed.), The depressed child and depression? Journal of Child Psychology and
adolescent (2nd ed., pp. 143–178). New York, Psychiatry, 47, 1263– 1271.
NY: Cambridge University Press. Costello, E., Pine, D.S., Hammen, C., March, J.S.,
Angold, A., Costello, E.J., Messer, S.C., & Pickles, Plotsky, P.M., Weissman, M.M., . . . Leckman,
A. (1995). Development of a short question- J.F. (2002). Development and natural history of
naire for use in epidemiological studies of mood disorders. Biological Psychiatry, 52,
depression in children and adolescents. Inter- 529 – 542. doi:10.1016/S0006-3223%2802%
national Journal of Methods in Psychiatric 2901372-0
Research, 5, 237– 249. Cummings, C.M., Caporino, N.E., & Kendall, P.C.
Asarnow, J.R., Scott, C.V., & Mintz, J. (2002). (2014). Comorbidity of anxiety and depression
A combined cognitive-behavioral family edu- in children and adolescents: 20 years after.
cation intervention for depression in children: Psychological Bulletin, 140, 816– 845. doi:10.
A treatment development study. Cognitive 1037/a0034733
Therapy and Research, 26, 221– 229. doi:10. David-Ferdon, C., & Kaslow, N.J. (2008). Evi-
1023/A:1014573803928 dence-based psychosocial treatments for child
Barrett, P.M., Farrell, L.J., Ollendick, T.H., & and adolescent depression. Journal of Clinical
Dadds, M. (2006). Long-term outcomes of an Child and Adolescent Psychology, 37, 62 – 104.
Australian universal prevention trial of anxiety doi:10.1080/15374410701817865
and depression symptoms in children and youth: de Cuyper, S., Timbremont, B., Braet, C., De
An evaluation of the friends program. Journal of Backer, V., & Wullaert, T. (2004). Treating
Clinical Child & Adolescent Psychology, 35, depressive symptoms in schoolchildren: A pilot
403–411. doi:10.1207/s15374424jccp3503_5 study. European Child & Adolescent Psychiatry,
Begg, C.B., & Mazumdar, M. (1994). Operating 13, 105– 114. doi:10.1007/s00787-004-0366-2
characteristics of a rank correlation test for Dougherty, L.R., Klein, D.N., Olino, T.M., &
publication bias. Biometrics, 50, 1088– 1101. Laptook, R.S. (2008). Depression in children
Birmaher, B., Ryan, N.D., Williamson, D.E.,
Brent, D.A., Kaufman, J., Dahl, R.E., . . . and adolescents. In J. Hunsely & E. J. Mash
Nelson, B. (1996). Childhood and adolescent (Eds.), A guide to assessments that work
depression: A review of the past 10 years, Part I. (pp. 69 – 95). New York, NY: Oxford University
Journal of the American Academy of Child & Press.
Adolescent Psychiatry, 35, 1427–1439. doi:10. DuBois, D.L., Felner, R.D., Bartels, C.L., &
1097/00004583-199611000-00011 Silverman, M.M. (1995). Stability of self-
Borenstein, M. (2009). Introduction to meta- reported depressive symptoms in a community
analysis. Chichester: Wiley. sample of children and adolescents. Journal of
Brooks, S.J., & Kutcher, S. (2001). Diagnosis and Clinical Child Psychology, 24, 386– 396. doi:10.
measurement of adolescent depression: 1207/s15374424jccp2404_3
A review of commonly utilized instruments. Durlak, J.A., Weissberg, R.P., Dymnicki, A.B.,
Journal of Child and Adolescent Psychopharma- Taylor, R.D., & Schellinger, K.B. (2011). The
cology, 11, 341 – 376. doi:10.1089/ impact of enhancing students’ social and
104454601317261546 emotional learning: A meta-analysis of school-
Butler, L., Mietzitis, S., Friedman, R., & Cole, E. based universal interventions. Child Develop-
(1980). The effect of two school-based interven- ment, 82, 405 – 432. doi:10.1111/j.1467-8624.
tion programs on depressive symptoms in 2010.01564.x
VOL 43, NO 4, 2014 CBT for Children with Depressive Symptoms 287

Duval, S., & Tweedie, R. (2000). Trim and fill: Psychiatry, 46, 1403 – 1413. doi:10.1097/chi.
A simple funnel-plot-based method of testing 0b013e3180592aaa
and adjusting for publication bias in meta- Klein, D.N., Lewinsohn, P.M., & Seeley, J.R.
analysis. Biometrics, 56, 455– 463. (1997). Psychosocial characteristics of adoles-
Easterbrook, P.J., Berlin, J.A., Gopalan, R., & cents with a past history of dysthymic disorder:
Matthews, D.R. (1991). Publication bias in Comparison with adolescents with past his-
clinical research. Lancet, 337, 867– 872. tories of major depressive and non-affective
Erford, B.T., Erford, B.M., Lattanzi, G., Weller, J., disorders, and never mentally ill controls.
Schein, H., Wolf, E., . . . Peacock, E. (2011). Journal of Affective Disorders, 42, 127–135.
Counseling outcomes from 1990 to 2008 for doi:10.1016/S0165-0327%2896%2901403-6
school-age youth with depression: A meta- Kovacs, M. (1992). The Children’s Depression
analysis. Journal of Counseling & Development, Inventory (CDI) manual. Toronto: Multi-
89, 439 – 457. doi:10.1002/j.1556-6676.2011. Health Systems.
tb02841.x Kovacs, M., & Lopez-Duran, N.L. (2012). Con-
Fonagy, P., Target, M., Cottrell, D., Phillips, J., & textual emotion regulation therapy:
Kurtz, Z. (2005). What works for whom? A A developmentally based intervention for
critical review of treatments for children and pediatric depression. Child and Adolescent
adolescents. London: Guilford Press. Psychiatric Clinics of North America, 21,
Fristad, M.A., Verducci, J.S., Walters, K., & 327– 343. doi:10.1016/j.chc.2012.01.002
Young, M.E. (2009). Impact of multifamily Lewinsohn, P.M., & Clarke, G.N. (1999). Psycho-
psychoeducational psychotherapy in treating social treatments for adolescent depression.
children aged 8 to 12 years with mood disorders. Clinical Psychology Review, 19, 329 – 342.
Archives of General Psychiatry, 66, 1013– 1020. doi:10.1016/S0272-7358%2898%2900055-5
doi:10.1001/archgenpsychiatry.2009.112 Liddle, B., & Spence, S.H. (1990). Cognitive-
Gillham, J.E., Reivich, K.J., Jaycox, L.H., & behaviour therapy with depressed primary
Seligman, M.E. (1995). Prevention of depressive school children: A cautionary note. Behavioural
symptoms in schoolchildren: Two-year follow- Psychotherapy, 18, 85 – 102. doi:10.1017/
up. Psychological Science, 6, 343– 351. doi:10. S0141347300018218
1111/j.1467-9280.1995.tb00524.x Luby, J., Lenze, S., & Tillman, R. (2012). A novel
Harter, S. (1990). Adolescent self and identity early intervention for preschool depression:
development. In S. Feldman & G. Eliot (Eds.), Findings from a pilot randomized controlled
At the threshold: The developing adolescent trial. Journal of Child Psychology and Psychia-
(pp. 352 – 387). Cambridge, MA: Harvard try, 53, 313– 322. doi:10.1111/j.1469-7610.2011.
University Press.
Jaycox, L.H., Reivich, K.J., Gillham, J., & Selig- 02483.x
man, M.E. (1994). Prevention of depressive McAuley, L., Pham, B., Tugwell, P., & Moher, D.
symptoms in school children. Behaviour (2000). Does the inclusion of grey literature
Research & Therapy, 32, 801– 816. influence estimates of intervention effectiveness
Jonsson, U., Bohman, H., von Knorring, L., reported in meta-analyses? Lancet, 356,
Olsson, G., Paaren, A., & von Knorring, A. 1228– 1231.
(2011). Mental health outcome of long-term Michael, K.D., & Crowley, S.L. (2002). How
and episodic adolescent depression: 15-year effective are treatments for child and adolescent
follow-up of a community sample. Journal of depression? A meta-analytic review. Clinical
Affective Disorders, 130, 395– 404. doi:10.1016/ Psychology Review, 22, 247– 269. doi:10.1016/
j.jad.2010.10.046 S0272-7358%2801%2900089-7
Kahn, J., & Kehle, T. (1990). Comparison of Moher, D., Pham, B., Jones, A., Cook, D.J., Jadad,
cognitive-behavioral, relaxation, and self-mod- A.R., Moher, M., . . . Klassen, T.P. (1998).
eling interventions for depression. School Does quality of reports of randomised trials
Psychology Review, 19, 196– 211. affect estimates of intervention efficacy reported
Kaslow, N.J., & Thompson, M.P. (1998). Applying in meta-analyses? Lancet, 352, 609– 613. doi:10.
the criteria for empirically supported treatments 1016/S0140-6736(98)01085-X
to studies of psychosocial interventions for child Nelson, E.-L., Barnard, M., & Cain, S. (2003).
and adolescent depression. Journal of Clinical Treating childhood depression over videocon-
Child Psychology, 27, 146– 155. doi:10.1207/ ferencing. Telemedicine Journal and E-Health,
s15374424jccp2702_2 9, 49– 55.
King, N.J., Ollendick, T.H., & Gullone, E. (1991). Orwin, R.G. (1983). A fail-safe N for effect size in
Negative affectivity in children and adolescents: meta-analysis. Journal of Educational Statistics,
Relations between anxiety and depression. 8, 157–159. doi:10.2307/1164923
Clinical Psychology Review, 11, 441 – 459. Öst, L.-G. (2008). Efficacy of the third wave of
doi:10.1016/0272-7358%2891%2990117-D behavioral therapies: A systematic review and
Klein, J.B., Jacobs, R.H., & Reinecke, M.A. (2007). meta-analysis. Behaviour Research and Therapy,
Cognitive-behavioral therapy for adolescent 46, 296–321. doi:10.1016/j.brat.2007.12.005
depression: A meta-analytic investigation of Öst, L.-G. (Ed.). (2010). KBT inom barn- och
changes in effect-size estimates. Journal of the ungdomspsykiatrin [CBT in child and adolescent
American Academy of Child & Adolescent psychiatry]. Stockholm: Natur & Kultur.
288 Arnberg and Öst COGNITIVE BEHAVIOUR THERAPY

Poznanski, E.O., Grossman, J.A., Buchsbaum, Y., 9-month follow-up. Journal of Affective
Banegas, M., Freeman, L., & Gibbons, R. Disorders, 40, 105 – 116. doi:10.1016/0165-
(1984). Preliminary studies of the reliability and 0327%2896%2900054-7
validity of the children’s depression rating scale. Watanabe, N., Hunot, V., Omori, I., Churchill, R.,
Journal of the American Academy of Child & Furukawa, T. (2007). Psychotherapy for
Psychiatry, 23, 191– 197. doi:10.1097/00004583- depression among children and adolescents:
198403000-00011 A systematic review. Acta Psychiatrica Scandi-
Rosenthal, R. (1991). Meta-analysis: A review. navica, 116, 84 – 95. doi:10.1111/j.1600-0447.
Psychosomatic Medicine, 53, 247– 271. 2007.01018.x
Rowlison, R.T., & Felner, R.D. (1988). Major life Weissman, M.M., Warner, V., Wickramaratne, P.,
events, hassles, and adaptation in adolescence: Moreau, D., & Olfson, M. (1997). Offspring of
Confounding in the conceptualization and depressed parents: 10 years later. Archives of
measurement of life stress and adjustment General Psychiatry, 54, 932– 940. doi:10.1001/
revisited. Journal of Personality and Social archpsyc.1997.01830220054009
Psychology, 55, 432 – 444. doi:10.1037/0022- Weissman, M.M., Wolk, S., Goldstein, R.B.,
3514.55.3.432 Moreau, D., Adams, P., Greenwald, S., . . .
Spielmans, G.I., Pasek, L.F., & McFall, J.P. (2007). Wickramaratne, P. (1999). Depressed adoles-
What are the active ingredients in cognitive and cents grown up. JAMA: Journal of the American
behavioral psychotherapy for anxious and Medical Association, 281, 1707– 1713. doi:10.
depressed children? A meta-analytic review.
Clinical Psychology Review, 27, 642 – 654. 1001/jama.281.18.1707
doi:10.1016/j.cpr.2006.06.001 Weisz, J.R., McCarty, C.A., & Valeri, S.M. (2006).
Stark, K.D., Reynolds, W.M., & Kaslow, N.J. Effects of psychotherapy for depression in
(1987). A comparison of the relative efficacy of children and adolescents: A meta-analysis.
self-control therapy and a behavioral problem- Psychological Bulletin, 132, 132– 149. doi:10.
solving therapy for depression in children. 1037/0033-2909.132.1.132
Journal of Abnormal Child Psychology, 15, Weisz, J.R., Thurber, C.A., Sweeney, L., Proffitt,
91– 113. doi:10.1007/BF00916468 V.D., & LeGagnoux, G.L. (1997). Brief treat-
SBU (Swedish Council on Technology Assessment ment of mild-to-moderate child depression
in Health Care) (2004). Behandling av depres- using primary and secondary control enhance-
sionssjukdomar. (Report no. 166/3). Stockholm: ment training. Journal of Consulting and Clinical
Author. Retrieved from http://www.sbu.se/en/ Psychology, 65, 703 – 707. doi:10.1037/0022-
Published/Yellow/Treatment-of-depression 006X.65.4.703
Vostanis, P., Feehan, C., Grattan, E., & Bickerton, Whittington, C.J., Kendall, T., & Pilling, S. (2005).
W.-L. (1996). A randomised controlled out- Are the SSRIs and atypical antidepressants safe
patient trial of cognitive-behavioural treatment and effective for children and adolescents?
for children and adolescents with depression: Current Opinion in Psychiatry, 18, 21 – 25.
Copyright of Cognitive Behaviour Therapy is the property of Routledge and its content may
not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's
express written permission. However, users may print, download, or email articles for
individual use.