Journal of Child Psychology and Psychiatry **:* (2016), pp **–** doi:10.1111/jcpp.

12590

Practitioner Review: Psychological treatments for

children and adolescents with conduct disorder
problems – a systematic review and meta-analysis
M.J. Bakker,1 C.U. Greven,1,2,3 J.K. Buitelaar,1,2,* and J.C. Glennon1,*
1
Department of Cognitive Neuroscience, Donders Institute for Brain, Cognition and Behaviour, Radboudumc,
Nijmegen; 2Karakter Child and Adolescent Psychiatry University Centre Nijmegen, Nijmegen, The Netherlands;
3
Medical Research Council Social, Genetic & Developmental Psychiatry Centre, Institute of Psychiatry, Psychology &
Neuroscience, King’s College London, London, UK

Background: This meta-analysis evaluates the efficacy of nonpharmacological treatments for conduct disorder (CD)
problems in children and adolescents, based on child, parent and teacher report. Methods: PubMed, PsycINFO and
EMBASE were searched for peer-reviewed articles published between January 1970 and March 2015. Main inclusion
criteria were nonpharmacological treatment, participants younger than 18 years, clinical CD problems/diagnosis,
randomized controlled trials and inclusion of at least one CD problem-related outcome. Treatment efficacy is
expressed in effect sizes (ESs) calculated for each rater (parent, teacher, self and blinded observer). Results: Of 1,549
articles retrieved, 17 (published between June 2004 and January 2014) describing 19 interventions met the
inclusion criteria. All studies used psychological treatments; only three studies included a blinded observer to rate
CD problems. Most studies were of very poor to fair quality. ESs were significant but small for parent-reported
outcomes (0.36, 95% CI = 0.27–0.47), teacher-reported outcomes (0.26, 95% CI = 0.12–0.49) and blinded observer
outcomes (0.26, 95% CI = 0.06–0.47), and they were nonsignificant for self-reported outcomes (
0.01, 95%
CI =
0.25 to 0.23). Comorbidity, gender, age, number of sessions, duration, intervention type, setting, medication
use or dropout percentage did not influence the effect of treatment. Conclusions: Psychological treatments have a
small effect in reducing parent-, teacher- and observer-rated CD problems in children and adolescents with clinical
CD problems/diagnosis. There is not enough evidence to support one specific psychological treatment over another.
Future studies should investigate the influence of participant characteristics (e.g. age of CD onset), use more
homogeneous outcome measures and allow better evaluation of study quality. Many reports failed to provide detailed
information to allow optimization of psychological treatment strategies. Keywords: Psychological; meta-analysis;
treatment; conduct disorder; aggression.

heterogeneous, and it has been suggested that a

Introduction
distinction should be made between covert and overt
To date, studies of nonpharmacological treatments
aggression and between reactive or affective aggres-
for disruptive behaviour have not specifically
sion and instrumental or proactive aggression
focused on clinically diagnosed conduct disorder
(Buitelaar et al., 2013). The DSM-5 defines CD with
(CD) nor related problem behaviour. Furthermore,
and without callous-unemotional traits, whereby
these studies have often suffered from small sample
patients with high callous-unemotional traits are
sizes. The aims of this article are to provide an
more refractory to treatment and have a poorer long-
overview of nonpharmacological treatments for chil-
term prognosis (e.g. increased risk of criminal activ-
dren and adolescents with CD and/or CD problems
ity; American Psychiatric Association, 2013; Frick &
in the clinical range [e.g. including disruptive
White, 2008; Loeber & Stouthamer-Loeber, 1998;
behaviour disorders (DBD) and oppositional defiant
Scheepers, Buitelaar, & Matthys, 2010). Further-
disorder (ODD)], to outline clinical implications for
more, individuals presenting with early-onset CD
clinical practitioners, and make suggestions for
(i.e. before 10 years of age) may have a relatively
future research to optimize clinical outcomes.
poorer outcome than those with late-onset CD (i.e.
Conduct disorder is characterized by a repetitive
onset after 10 years of age; Fairchild, van Goozen,
and persistent pattern of antisocial and rule-break-
Calder, & Goodyer, 2013; Moffitt, 2006; Moffitt,
ing behaviour, categorized into four classes: (a)
Caspi, Dickson, Silva, & Stanton, 1996) and are at
aggression to people and animals, (b) destruction of
greater risk of developing antisocial personality
property, (c) deceitfulness or theft and (d) serious
disorder in later life. In many cases, CD is comorbid
violations of rules (American Psychiatric Association,
with attention deficit hyperactivity disorder (ADHD)
2013). Aggressive and antisocial behaviour is
and is preceded by ODD (Loeber, Burke, Lahey,
Winters, & Zera, 2000). This makes it important to
*Shared last author. develop effective treatment programmes for conduct
Conflicts of interest statement: See Acknowledgements for problems.
disclosures.

© 2016 Association for Child and Adolescent Mental Health.

Published by John Wiley & Sons Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main St, Malden, MA 02148, USA
2 M.J. Bakker et al.

Treatment for CD problems can be divided into the evaluated efficacy of psychological treatment,
pharmacological and nonpharmacological appro- complicate the interpretation of findings and limit
aches. There is no first-line medication licensed for the generalizability of study outcomes. The current
this age group, and all medication is primarily used review focuses on nonpharmacological treatments in
off-label. Medications approved for other indications children and adolescents who have a clinical CD
(particularly ADHD), such as stimulants (e.g. diagnosis and/or clinical CD problems (including
methylphenidate), alpha-2 agonists (e.g. guanfacine) DBD and ODD) and who have an IQ of minimally 80
and atypical antipsychotics (e.g. risperidone, arip- points, compared with the minimally 60 points used
iprazole and quetiapine), are currently the medical as cut-off in the studies of the NICE guideline.
treatments of choice but are secondary to psychoso- Children and adolescents with an IQ lower than 80
cial interventions (Linton, Barr, Honer, & Procyshyn, have mild intellectual problems and may benefit
2013; Smith & Coghill, 2010). Six meta-analyses of from a different treatment approach from that for
psychological treatments for children and adoles- individuals with a higher IQ. The exclusion of
cents with (but not limited to) a CD diagnosis and/or children and adolescents with a lower IQ also
CD problems have been published (Grove, Evans, increases the homogeneity of our study population.
Pastor, & Mack, 2008; Lundahl, Risser, & Lovejoy, Third, the current review includes studies of non-
2006; McCart, Priester, Davies, & Azen, 2006; pharmacological treatments involving children and
National Institute for Health and Clinical Excellence, adolescents and incorporates data of adolescents
2013; Weisz et al., 2013; Wilson & Lipsey, 2007), between 12 and 18 in contrast to the NICE guidelines
and the aim of this article is to investigate the in which the overall age range was between 7 and
effectiveness of nonpharmacological interventions 14 years. Fourth, as recommended by Rosato et al.
other than psychological ones for the treatment of (2012), psychological treatments should be age
CD and/or clinical CD problems. specific, because developmental differences (i.e. in
The meta-analysis of Wilson and Lipsey (2007) cognitive, behavioural, affective and communicative
focused on the efficacy of treatment in a population- abilities) affect outcomes. Hence, the current review
based sample (e.g. through school-based pro- investigates the onset of CD as potential moderator
grammes), and the other meta-analyses focused on of treatment effectiveness in reducing CD problems,
prevention programmes for children and adolescents an aspect that has not been addressed previously.
with disruptive and delinquent behaviour but with- Fifth, it is not always clear whether outcome mea-
out a manifest disorder (Grove et al., 2008), pro- sures are scored by blinded raters (McCart et al.,
grammes for children and adolescents with 2006). Last, but not least, in contrast to the six meta-
internalizing (e.g. anxiety) and externalizing (e.g. analyses, this review takes into account whether
misconduct, ADHD) problems (Weisz et al., 2013), study participants are on stable medication during
or specific treatment methods, such as parent the study. In addition, not all previous meta-
training (Lundahl et al., 2006), or behavioural par- analyses took the potential role of child and/or
ental training and cognitive behavioural therapy environmental factors in treatment efficacy into
(McCart et al., 2006). The National Institute for account.
Health and Clinical Excellence (NICE) guidelines, Although evidence for the efficacy of psychological
published in 2013, provide a systematic review and treatments for CD problems is limited, these remain
meta-analysis of psychological treatments and con- the treatment of choice (Conduct Problems Preven-
clude that psychosocial interventions have an overall tion Research Group, 1999, 2011; Rosato et al.,
small-to-moderate effect size (ES) compared with 2012). The NICE guidelines recommend a psychoso-
care as usual or no treatment in reducing antisocial cial approach to the treatment of CD problems, with
behaviour in youths. The other five older meta- emphasis on working with parents (if possible) or
analyses showed that psychosocial treatment was guardians. Both the NICE guideline (2013) and
modestly beneficial (overall ESs ranged from 0.17 to Eyberg, Nelson, and Boggs (2008) advocate three
0.42, including beyond the end) compared with care treatment approaches, depending on the age of the
as usual or waiting-list control in children and affected individual. For children aged 3–11 years,
adolescents with DBD/CD problems. parenting training programmes to improve the par-
The present review included 17 studies (11 of enting skills of parents are recommended. These
which were not included in the NICE review). programmes typically consist of 10–12 parents in a
Although the NICE review included 154 studies, group and involve 10–16 meetings (each lasting 90–
the current review extends the NICE guidelines in 120 min) in which modelling (e.g. imitating the
several ways. First, the NICE guidelines do not focus child), rehearsal (e.g. parents rehearse new skills in
solely on clinical cases of CD/ODD but also on session) and feedback (e.g. therapists discuss and, if
population-based samples and use broad definitions necessary, modify parents’ behaviours; Pilling,
of psychopathy (e.g. the meta-analysis included Gould, Whittington, Taylor, & Scott, 2013) are used
samples in which only a minority had CD/ODD to improve parenting skills. For children aged 9–
and/or that were not clinically at risk for CD/ODD). 14 years, cognitive behavioural approaches are the
The inclusion of a mixed study population may affect treatment of choice and consist of 10–18 weekly

© 2016 Association for Child and Adolescent Mental Health.

 Psychological treatments for conduct disorder 3

meetings (lasting 2 hr) during which modelling (e.g. hand-searched to identify additional related publications.
imitating peers), rehearsal (e.g. the child rehearses Books and unpublished articles were not included. Authors
were contacted to gather further information about details not
new skills) and feedback (e.g. therapists discuss the
reported in the selected papers. Then, the following inclusion
child’s behaviour, with a view to modifying it; Pilling criteria were applied: (a) participants younger than 18 years
et al., 2013) are used to improve the child’s skills. with CD and/or ODD diagnosis, or scores on a dimensional
And for adolescents aged 11–17 years, multimodal construct of CD problems in the clinical range; (b) at least one
programmes (e.g. multisystemic therapy) are pre- quantitative measure of CD problem outcome reported (e.g.
rating scale or observation scale pre- and/or post-measure-
ferred, consisting of 3–4 meetings a week for 3–
ments, follow-up); (c) study published in a peer-reviewed
5 months. These programmes are based on a social journal; and (d) randomized controlled trial of nonpharmaco-
learning model with interventions provided at differ- logical intervention versus control (placebo, waiting list, no
ent levels (i.e. individual, family, school, criminal treatment or treatment as usual). Comorbidity and use of
justice and community) and have an explicit and medications were not exclusion criteria; articles not written in
English, case reports and review articles were excluded. In
supportive focus on the family. They are provided by
each article, all outcome measures relevant to CD problems
appropriately trained case managers (Pilling et al., were selected.
2013).
An up-to-date comparison of nonpharmacological
treatments targeting specifically CD or related prob- Data extraction and statistics
lems is needed. This systematic review and meta- In total, 1,549 articles were retrieved and an additional 42 were
analysis of nonpharmacological treatments for CD, identified from reference lists. Ultimately, 17 articles describ-
specifically on outcomes related to CD problems, ing 19 interventions met the inclusion and exclusion criteria
(Figure 1). These articles were published between June 2004
provides strategies for improvement.
and January 2014. The following information was extracted:
age range or mean if available, clinical diagnosis and/or
diagnostic criteria used, sample size, gender distribution,
treatment name, treatment duration, setting (e.g. outpatient,
Methods clinic), outcomes (in terms of CD-related problems), medica-
Inclusion and exclusion criteria tion use, rater (e.g. parental report, teacher report and self-
report) and rater blinding to treatment allocation. All study
A search of the PubMed, EMBASE and PsycINFO databases for characteristics and participant details of the included studies
peer-reviewed papers published up to March 2015, using the are summarized in Table 1. Study quality was assessed by two
following keywords: (((conduct disorder OR CD OR disruptive authors (MJB and JG), using the standard definition for
behavio* AND disorder) OR (disruptive behavio* AND disor- randomization; missing data were accounted for as described
ders)) NOT (disease)), identified 1,549 articles. The titles and by Jadad et al. (1996), with the following scores for random-
abstracts of the retrieved articles were read by at least two of ized controlled trial quality: 5 = excellent, 4 = good, 3 = fair,
the authors (MJB and JKB), and reference lists and relevant 2 = poor, 1 = very poor (Crowther, Lim, & Crowther, 2010;
published reviews (i.e. Von Sydow, Retzlaff, Beher, Haun, & Jadad et al., 1996). Two points were awarded for (appropriate)
Schweitzer, 2013; Woolfenden, Williams, & Peat, 2001) were
Identification

Records identified through Records identified through

database searching (n = 1544) other sourcesb (n = 5)
Screening

Articles screened (n = 1549) Records excluded at title

and abstract (n = 1507)

Full-text articles assessed Full text excluded (n = 25)

Eligibility

for eligibility (n = 42) No RCT (n = 12)

Diffused target sample (n = 11)
No control group (n = 1)
Data for calculating effect
sizes not available on conduct
Studies included in systematic problem outcome (n = 1)
Included

review and meta-analysis (n = 17)

Parent raters Teacher raters Self-reports Blinded raters

n = 15 n=6 n=2 n=2

Figure 1 Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMAa) flow diagram on psychological treatments.
a
PRISMA = Preferred Reporting Items for Systematic Reviews and Meta-Analyses (www.prisma-statement.org). bReferences from
identified articles and relevant published reviews were manually searched to identify additional related publications

© 2016 Association for Child and Adolescent Mental Health.

 Table 1 Study and participant characteristics of included randomized controlled trials. 4
Study population Intervention Averaged
Reported CD Effect size effect size
Participants Type N Treatment N Dropout % Male design problems based on regardless
Study Age (yrs) (comorbidity) Control (Control) (Control) (Control) Name Time range N sessions (duration) Setting qualitya measuresb last time range of rater

Bagner et al. 2–5 EP (n.r.) WL 14 (14) 3 (0) Overall Parent–Child 12 weeks 13 (60 min) Clinic 3 CBCL-AP 0.98 1.46
(2010) 71 Interaction Therapy CBCL-EP 1.36
ECBI-IP 2.07
ECBI-PP 1.43
M.J. Bakker et al.

Perrin et al. 2–4 DB-I (n.r.) WL IYP 89 17 63 Incredible Years 10 weeks 10 (120 min) Clinic 3 IYP NR IYP 0.30
(2014) Program – Parent FU 1 year ECBI-PP 0.59 0.39 NR 0.27
NR 123 (61) 51 (11) 62 (62) Training (IYP) and not ECBI-IP 0.43 0.56
randomized DPICSRBO 0.19 0.12
IYP (NR)

Somech and 3–5 DB-S (n.r.) AC 140 (69) 15 (12) 76 (81) Hitkashrut – Parent 4 weeks 14 (2 hr) n.r. 3 ECBI-IP 0.77 0.82
Elizur (2012) Training Control 2 (n.r.) CU traitsP 0.86

Hanisch et al. 3–6 ODD (ADHD) TAU 91 (64) 16 (18) 74 (70) Prevention Program 10 weeks 10 (90–120 min) Kinder- 2 CSP 0.36 0.31
(2010) for Externalizing FU 8 weeks garten CST 0.26
Problem Behaviour

Eyberg et al. 3–6 ODD (ADHD) TAU 31 (30) 14 (13) Overall Parent–Child PCIT n.r. PCIT n.r. Phone & home 1 CBCL-EPm 0.48 0.09
(2014)M 69 Interaction CBCL-EPf
0.37
Therapy + FU 2 years 24 (range 5–60 min)
Maintenance 1–3 home visits
Treatment
(PCIT-MT)

Jones et al. 3–8 DB-E (n.r.) AC 7 (8) n.r. Overall Helping the 8–12 weeks HNC 8–12 (n.r.) Clinic 2 ECBI-IP 1.05 0.98
(2014) 53 Noncompliant Child TE text reminders (n.r.) Phone ECBI-PP 0.91
(HNC) + 3-min video’s (number
Technology- n.r.), mid-week video-
Enhanced (TE) calls (n.r.), 8–12
videotaped home
practice (n.r.)

Scott and 4–6 CP and/or AC ED 36 (32) n.r. ED 75 Incredible Years SP 12 weeks SP 28 (2.5 hr) School & 1 PACS-CPP ED 0.87 H 0.14 ED 0.87
O’Connor (2012) ODDED or H School Program (SP) Home H 0.14
(n.r.) H 25 (19) H 64 (73) + Child Literacy CL n.r. CL 16 (n.r)
Program (CL) FU 1 year

Drugli and 4–8 ODD/CD WL PT 47 23 80 (n.r.) Incredible Years-PT PT 12–14 weeks PT: 12–14 (2 hr) Clinic 2 TRFT + PBQT PT 0.06 PT + CT PT 0.06
Larsson (2006) (ADHD) PT+CT 52 (28) Incredible Years-CT CT 18 weeks CT: 18 (2 hr) (Combined 0.68 PT + CT 0.68
FU 1 year score only)

Jouriles et al. 4–9 ODD/CD (n.r.) TAU 32 (34) 5 (5) 58 (41) Project Support 32 weeks 20 (range 2–40; SD 9); Home 3 CBCL-EP 0.97 0.76
(2009)1 FU 80 weeks control: 4 (range 0–9, ECBI-PP 0.89
SD 2.7) OCBSBO 0.42

McDonald et al. 4–9 CD/ODD (n.r.) TAU 32 (34) n.r. n.r. Project Support 32 weeks 20 (range 2–40; n.r.), in Home 2 PSDP 0.51 0.51
(2011)1 addition 5.4 contacts
(n.r.); control: 4 (range
0–9; n.r.)

Owens et al. 6–10 CD/ODD WL 91 (26) n.r. 78 (73) Daily Report Card 1 school-year Parent: 8 (range School & 2 DBD-CDP 0.51 0.41
(2008)M (ADHD) (DRC) – procedure (fall-spring) 3–63, n.r.) Home DBD-ODDP 0.30
for children, teacher Children: 20 (n.r.) DBD-CDT 0.35
consultation, Teacher: 26 with clinician DBD-ODDT 0.46
and behaviourally (n.r.)
based parent
sessions

(continued)

© 2016 Association for Child and Adolescent Mental Health.

 Table 1 (continued)

Study population Intervention Averaged

Reported CD Effect size effect size
Participants Type N Treatment N Dropout % Male design problems based on regardless
Study Age (yrs) (comorbidity) Control (Control) (Control) (Control) Name Time range N sessions (duration) Setting qualitya measuresb last time range of rater

Kolko et al. 6–11 CP and/or ODD AC 83 (80) 9 (14) 65 (65) Protocol for Onsite 12–24 weeks 6 (1.5 hr) Clinic 1 PSC-17-EP 0.00 0.08
(2010)2,M (ADHD) Nurse-administered FU 1 year 2–4 booster sessions SDQ-TP 0.05
Intervention (PONI) (n.r.) SDQ-TT
0.20

Boylan et al. 8–11 ODD/CD (mood WL + TAU 78 (87) 17 (26) 76 (71) Multifamily Psycho- n.r. 8 (90 min) University 2 ChIPS-
(2013) disorders) educational FU 48 weeks CDP
0.12
Psychotherapy (MF- ODDP 0.29 0.16
PEP) +TAU DBDP 0.30

Van Manen et al. 9–13 ODD/CD (no WL SC 42 n.r. Overall Social Cognitive (SC), 11 weeks SC 11 (70 min) Clinic 2 SC SS SC 0.30
(2004) formal ADHD) SS 40 100 and Social Skills SS 11 (70 min) CBCL-EP
0.13 0.37 SS 0.18
(15) (SS) TRF-ET 0.16 0.23
TRA-RT 0.57 0.05
TRA-PT 0.41 0.55
TRA-CT 0.26 0.06
SCRST 0.16
0.18

Kolko et al. 10–14 CD/ODD AC 59 (59) 4 (37) 86 (79) Booster Treatment 24 weeks Maximum of 15 hr Clinic 1 APSD-CUT 0.14 0.02
(2013)2,M (ADHD) for PONI FU 2 years TRF-ET
0.17
CBCL-EP 0.16

Sundell et al. 12–17 CD (n.r.) TAU 79 (77) 24 (29) 61 (n.r.) Multi-systematic 28 weeks Weekly by phone (n.r.); Home 3 CBCL-OP 0.27 0.06

© 2016 Association for Child and Adolescent Mental Health.

(2008) Therapy quarterly on-site booster YSR-EC 0.01
sessions (n.r.); TAU: SRDC
0.03
biweekly individual PYSC
0.12
sessions (1–2 hr) or
family therapy (n.r.)

Hendriks et al. 13–18 CD/ODD AC 55 (54) 11 (38) 80 (80) Multi-dimensional 20–24 weeks 48 (1 hr) Clinic 2 VPCC 0.01 0.01
(2012) (Cannabis Family Therapy FU 1 year
Depen-dence)

Other: n.r., not reported; 1-1; 2-2Same dataset, but different outcome measures; mincluding participants on medication use regardless of group. Abbreviations participants: ADHD,
attention deficit hyperactivity disorder; CD, conduct disorder; CP, conduct problems; DB-I, S, or EDisruptive Behaviour: Ibased on Infant-Toddler Social-Emotional Assessment Scale >80th
percentile; SSDQ score >80th percentile or based on EECBI-severity T-score >60; EP, externalizing problems-based on CBCL-externalizing T-score >60; ODD, Oppositional Defiant
Disorder; ODD-ED or HChildren with CP and/or ODD subgroup Emotionally Dysregulated or Headstrong. Abbreviations comparison group: AC, active control; TAU, treatment as usual; WL,
wait list. Abbreviations N treatment (control): CT, child training; NR, not randomized to treatment group; PT, parent training. Time-range: FU, follow-up; hr, hour; min, minutes.
Abbreviations outcome measures: APSD-CU, antisocial process screening device-callous/unemotional; CBCL-O/A/E, child behaviour checklist-overall score/aggressive behaviour/
externalizing behaviour; ChIPS-CD/ODD/DBD, Children’s Interview for Psychiatric Syndromes – Conduct Disorder/Oppositional Defiant Disorder/Overall Disruptive Behaviour
Disorder symptoms; CP, conduct problems (based on Dyadic Parent–Child Interaction Coding System and Multi-Option Observation System for Experimental Studies); CS, composite
score: CBCL-total/TRF-total + PCL ADHD and ODD + questionnaire on Judging Parental Strains; CU traits, Callous/Unemotional (based on APSD and ICU); DBD-CD/ODD, disruptive
behaviour disorder-conduct disorder and oppositional deviant disorder symptoms; DPICSR_CII, Dyadic Parent-Child Interactive Coding System – Coder Impresssion Inventory; ECBI-I/P,
Eyberg Child Behaviour Inventory-Intensity/Problem Behaviour Scale; OCBS, Oppositional Child Behaviour Score; PACS-CP, Parent Account of Child Symptoms-Conduct Problem score;
PBQ, Preschool Behaviour Questionnaire; PSC
17-E, Pediatric Symptom Checklist-Externalizing; PSD, Psychopathy Screening Device-Callous/Unemotional and Narcissism scales; PYS,
Pittsburgh Youth Study, including subscale bad friends; SCRS, Self-Control Rating Scale; SDQ-T, Strengths and Difficulties Questionnaire-Total; SRD, Self-Report Delinquency Scale;
TRA-R,P,C, Teacher Rating of Aggression-Reactive, Proactive Aggression, Covert Antisocial; TRF-E, Teacher Report Form-Externalizing; VPC, Violent/Property Crimes past 90 days; YSR-
E, youth self report-externalizing behaviour problems.
a
Reported design quality of the included studies, based on the Jadad ratings, 5 = excellent, 4 = good, 3 = fair, 2 = poor, 1 = very poor.
b
Source of information: BO, blinded observer; C, child; P, parent-not specified; Pf, parent-father; Pm, parent-mother; T, teacher.
Psychological treatments for conduct disorder
5
6 M.J. Bakker et al.

randomization, two points for (appropriate) blinding and one intervention arm than to the control arm (1,225
point for reporting the fate of all patients (including dropouts). and 774, respectively); overall, the median dropout
Treatment efficacy is expressed in terms of the ES. Per
was 17.6%. Interventions were given in a clinical
study, for each instrument and rater, the ES was calculated by
using the unbiased estimate of ES (dppc2) developed by Morris setting (52.6%), home, school or a combination of
(2008). locations. Control conditions were treatment as
usual (n = 12) or a waiting-list control group
ðMpost; T
Mpre; T Þ
ðMpost; C
Mpre; C Þ
dppc2 ¼ Cp (n = 9). Ten studies focused on group interventions
SDpre and seven on individual interventions; four stud-
ies used a combination. For nine interventions,
Within brackets, the Mpost/pre, T and Mpost/pre, C refer to the
pre- and post-treatment outcomes were recorded
mean CD problem score, for the post- and pre-test for
treatment and control group, respectively. SDpre is referred to x = 16 weeks), and for 10 interventions, pretreat-
(

the standard deviation of the pretest and Cp is the sample size ment and follow-up outcomes were recorded
bias correction, based on the n in the treatment group (Nt) and x = 1.2 years). Details on intervention duration are
(

the n in the control group (Nc; Morris, 2008). provided in Table 1. Two interventions did not have a
3 set duration – Eyberg, Boggs, and Jaccard (2014)
Cp ¼ 1
: continued treatment until the parent’s skills reached
4ðNt þ Nc
2Þ
1
a preset level, and Sundell et al. (2008) did not
Cp is defined as the standardized difference between pre- regulate contact with therapists, who were available
and post-treatment and was subtracted from the standardized 24/7. Two interventions made use of audiovisual
mean difference (SMD) of the control condition. Because we material (Jones et al., 2014; Scott & O’Connor,
were interested in a sustained effect of the interventions, we
2012). Most interventions (n = 17) made use of
included (where possible) the follow-up outcomes when calcu-
lating ESs. By using the SMD, we could take into account the parent-reported information; seven interventions
use of different instruments to monitor CD-related problem made use of teacher-reported information, and 10
behaviours. The SMD and 95% confidence interval were interventions made use of self-reported information
calculated using Review Manager 5.3 (‘Review Manager (Rev- [Child Behaviour Check List (CBCL) – Externalizing
Man)’, 2014). A SMD higher than zero indicates that the active
subscale, n = 5, Achenbach, 1991a; Eyberg Child
treatment is better than the control condition in reducing
problem behaviour in children and adolescents diagnosed Behaviour Inventory (ECBI) – Intensity and Problem
with CD. Behaviour subscales, n = 8, Robinson, Eyberg, &
An overall ES (indicated as ES-all) was calculated per rater Ross, 1980]. Two studies collected parent-reported
if multiple outcome measures were used in a study. In information on specific aggression-related outcomes
addition, as an exploratory step, the data obtained with the
(e.g. callous-unemotional traits; McDonald et al.,
lowest and the highest scoring instruments were calculated
per rater (indicated as ES-low and ES-high). One study 2011; Somech & Elizur, 2012) and one study
(McDonald, Dodson, Rosenfield, & Jouriles, 2011) did not collected similar teacher-reported information
provide data to enable us to calculate the ES as described (Kolko, Lindhiem, Hart, & Bukstein, 2013). None of
above, and so we used an online calculator to calculate an the parent-scored instruments assessed subtypes of
ES (Thalheimer & Cook, 2002; see Table 1). In accordance
aggression, such as reactive and proactive aggres-
with the literature, an ES of 0.2 was considered to be small,
0.5 to be moderate and ≥0.8 to be large (Thalheimer & Cook, sion; only one teacher-scored instrument did (van
2002). Heterogeneity was calculated using chi-square (v2) Manen, Prins, & Emmelkamp, 2004). The most
and I-squared (I2) tests. If heterogeneity was present, meta- widely used instrument to score teacher-rated
regression analyses were performed with possible modera- aggression was the Teacher Report Form – External-
tors, such as participant or treatment characteristics, using
izing subscale (n = 2; Achenbach, 1991b). A wide
Bonferroni correction for multiple testing; the adjusted p-
value was .003. range of instruments was used for child-rated CD
problem outcomes (n = 5). Only two studies included
blinded observer-rated outcomes (Jouriles et al.,
Results 2009; Perrin, Sheldrick, McMenamy, Henson, &
The 17 studies recruited 1,999 participants (of Carter, 2014). Four of the 19 interventions allowed
which 73.4% boys), with a mean age of 7.5 years participants to use medication: in three, ADHD
(range 2.8–16.8 years). Nine studies included par- medication (not further specified) was used, and in
ticipants (n = 782) diagnosed according to DSM-IV one, no details were provided (see also Table 1).
or DSM-III with CD and/or ODD; the remaining eight Information about whether medication use was
studies included participants at risk of conduct stable was not provided in these studies. On the
problems or with externalizing problems in the basis of the Jadad scale score, most studies were of
clinical range. ADHD was the most common comor- ‘very poor’ or ‘poor’ quality (n = 12); five studies were
bidity (n = 6); eight studies did not report comorbid- of ‘fair’ quality (Table 1).
ity data. The 17 studies examined the effect of 19
different psychological interventions with a median
Effectiveness of psychosocial treatment – parent
duration of 14 hr for children/adolescents and
report
21.8 hr for parents. Only half of the studies provided
information about the duration of teacher involve- Seventeen interventions made use of parent-reported
ment. More participants were allocated to the information (Figure 2). Thirteen interventions were

© 2016 Association for Child and Adolescent Mental Health.

 Psychological treatments for conduct disorder 7

focused on parent management skills (e.g. on proac- the study by van Manen et al. (2004) included
tive and nurturing parenting, effective limit setting teacher-reported information on the effect of a cog-
and handling misbehaviour) and psycho-education nitive behavioural intervention. One study involved a
(Bagner, Sheinkopf, Vohr, & Lester, 2010; Boylan, multimodal programme (including the child, parents
Macpherson, & Fristad, 2013; Drugli & Larsson, and teachers; Owens, Murphy, Richerson, Girio, &
2006; Eyberg et al., 2014; Hanisch et al., 2010; Himawan, 2008), and another involved two types of
Jones et al., 2014; Jouriles et al., 2009; Kolko, interventions, namely, a multimodal programme and
Campo, Kelleher, & Cheng, 2010; Kolko et al., a separate parent-focused programme, and included
2013; McDonald et al., 2011; Perrin et al., 2014; teacher-reported ratings (Drugli & Larsson, 2006).
Somech & Elizur, 2012; Sundell et al., 2008). Some On the basis of available teacher-reported posttreat-
interventions focused on additional aspects such as ment/follow-up outcomes, the weighted mean ES-all
mandated participation of the father (Somech & was 0.26 (95% CI = 0.12–0.49; random effect mod-
Elizur, 2012), stress of the parent (Boylan et al., els), indicating a small but significant effect of the
2013; Eyberg et al., 2014), and instrumental and intervention in reducing CD problems in children/
emotional support to mothers (Jouriles et al., 2009; adolescents with clinical CD problems and/or a CD
McDonald et al., 2011). On the basis of posttreat- diagnosis. In addition, the weighted mean ES-low
ment/follow-up data, the weighted mean ES-all was was 0.18 (95% CI = 0.00–0.36; random effect mod-
0.37 (95% CI = 0.27–0.47; random effect models), els) and ES-high was 0.27 (95% CI = 0.08–0.46;
indicating that psychosocial treatment had a small random effect models). Heterogeneity ranged from
but significant effect in reducing CD problems in I2 = 32% (v2 = 10.3, p = .05) to I2 = 39% (v2 = 11.5,
children/adolescents with clinical CD problems p < .006), which shows inconsistency of study
and/or a CD diagnosis. Because the 95% confidence results. Therefore, the random effect model was
interval did not contain zero, the null hypothesis that used.
dppc2 = 0 was rejected at the 0.05 level.
In addition, the weighted mean ES based on the
Effectiveness of psychosocial treatment – self-report
lowest parent-reported score in each study was 0.30
(95% CI = 0.20–0.40; random effect models) and that Two interventions made use of self-reported infor-
based on the highest parent-reported score was 0.42 mation (Figure 4). These were multimodal pro-
(95% CI = 0.33–0.52; random effect models). Hetero- grammes involving children and parents (Sundell
geneity was calculated in RevMan 5.3 (Review et al., 2008) and/or family, school and courts (Hen-
Manager (RevMan), 2014), using the I-squared and driks, van der Schee, & Blanken, 2012). In these
chi-square test for the lowest and highest ES parent- cases, the focus was on the training of specific skills
reported scores in each study. If the mean scores of by means of cognitive behavioural therapy (Boylan
different samples differ, then the samples may et al., 2013; Drugli & Larsson, 2006; Kolko et al.,
originate from different populations (heterogeneity). 2010, 2013) or on improving motivation (Hendriks
Heterogeneity ranged from I2 = 62% (v2 = 40, et al., 2012; van Manen et al., 2004). On the basis of
p = .0005) to I2 = 63% (v2 = 40.13, p = .0004), which self-reported posttreatment/follow-up outcomes, the
shows inconsistency of study results (Higgins, weighted mean ES-all was
0.01 (95% CI =
0.25 to
Thompson, Deeks, & Altman, 2003). This necessi- 0.23; random effect models), indicating that neither
tated the use of a random effect model, which intervention reduced CD problems. Only Sundell
corrects for heterogeneity, to test for differences in et al. (2008) included multiple self-report informa-
outcomes in the meta-analysis (Borenstein, Hedges, tion, which yielded ESs ranging from d =
0.12 (95%
Higgins, & Rothstein, 2010; Ried, 2006). As seen in CI =
0.44 to 0.19; random effect models) to
Table 1, the ES reported in the study by Bagner et al. d = 0.01 (95% CI =
0.22 to 0.41; random effect
(2010) deviated substantially from that reported in models). The inclusion of multiple reports resulted in
other studies. Therefore, this study was excluded a weighted mean ES-low of 0.00 (95% CI =
0.02 to
from the sensitivity analysis, which showed that 0.21; random effect models) and ES-high of 0.10
specific participant and intervention characteristics (95% CI =
0.11 to 0.31; random effect models).
did not affect treatment efficacy. Heterogeneity was I2 = 0% (v2 = 1.72, p = .63), which
shows consistency of study results. Nevertheless,
the random effect model was used to correct for the
Effectiveness of psychosocial treatment – teacher
different outcome measures used in the meta-
report
analysis.
Seven interventions made use of teacher-reported
information (Figure 3). Teachers were not always
Effectiveness of psychosocial treatment – observer
directly involved in an intervention other than filling
report
out some questionnaires. Two parent-focused pro-
grammes also included teacher report alongside Three interventions also made use of blinded obser-
parent report (Hanisch et al., 2010; Kolko et al., vers (not further specified) to score the children/
2010). The follow-up study by Kolko et al. (2013) and adolescents’ CD problems (Jouriles et al., 2009;

© 2016 Association for Child and Adolescent Mental Health.

8 M.J. Bakker et al.

Figure 2 Forest plot conduct disorder problems, comparison of effect sizes (ESs) per study based on parent report. ■ [Name study]-All
Weighted ES based on multiple aggression outcomes. ♢ [Name study]-High Weighted ES based on single and highest aggression outcome.
[Name study]-Low
Weighted ES based on single and lowest aggression outcome. Abbreviations studies: ED or H Children with CP and/or
oppositional defiant disorder (ODD) subgroup emotionally dysregulated or headstrong; NR, not randomized to treatment group; SC,
social cognitive; SS, social skills

Perrin et al., 2014) (Figure 5). The weighted mean child-focused or multimodal programmes), group
ES-all was 0.26 (95% CI = 0.06–0.47; random effect size (i.e. group, individual or combination), setting
models), indicating a moderate but significant effect (i.e. school, clinic, home or combination) or dropout
in reducing CD problems. As both studies reported percentage (all p-values were at least >.07). Individ-
only one observer-rated outcome, it was not appro- ual studies, in particular of children aged 10 years
priate to calculate the weighted mean ES-low and or older, did not report whether their participants
ES-high. In these two studies, the mean weighted had early- or late-onset CD, with the exception of
ES-all based on parent-reported information was Kolko et al. (2010, 2013). However, we found a trend
higher than that based on observer-reported infor- towards smaller ESs in studies involving children
mation (d = 0.41, 95% CI = 0.21–0.62). aged 10 years or older compared with those of
studies involving children younger than 10 years
(i.e. average treatment efficacy based on all parent-
Moderators of treatment effect
reported information was 0.21 and 0.52, respec-
In order to explain the heterogeneity in outcomes, we tively). A similar trend was seen with teacher-
investigated whether ESs (per ES level and per rater) reported information (average ES = 0.32 in children
were moderated by specific participant and inter- aged <10 years and average ES = 0.11 in children
vention characteristics. However, analyses revealed aged >10 years). It was not possible to compare
no effect of comorbidity, gender, age, type of control treatment efficacy in the two age groups based on
(i.e. waiting-list control, treatment as usual or active self-reported and observer-reported information
control group), number of sessions, duration because of the limited data available. Treatment
(hours), intervention type (i.e. parent-focused, efficacy appeared not to be influenced by whether CD

© 2016 Association for Child and Adolescent Mental Health.

 Psychological treatments for conduct disorder 9

Figure 3 Forest plot conduct disorder problems, comparison of effect sizes (ESs) per study based on teacher report. ■ [Name study]-All
Weighted ES based on multiple aggression outcomes. ♢ [Name study]-High Weighted ES based on single and highest aggression outcome.
[Name study]-Low
Weighted ES based on single and lowest aggression outcome. Abbreviations studies: T or PT + CTParent Training, Parent
Training and Child Training; SC or SSSocial Cognitive subgroup, Social Skills subgroup

Figure 4 Forest plot conduct disorder problems, comparison of effect sizes (ESs) per study based on self-report. ■ [Name study]-All
Weighted ES based on multiple aggression outcomes. ♢ [Name study]-High Weighted ES based on single and highest aggression outcome.
[Name study]-Low
Weighted ES based on single and lowest aggression outcome

had been formally diagnosed or by early- versus late- (n = 10). This was the case for parent-, teacher-, self-
onset CD. Moreover, timing of assessment (post- and observer-rated outcomes. As most studies
treatment or follow-up) did not influence results: ES (n = 15) did not provide details about the medication
ranged from 0.06 to 1.46 for interventions with used, it was not possible to perform sensitivity
posttreatment assessment (n = 9) and from 0.01 to analyses for medication use. Lastly, trials with high
0.76 for interventions with a follow-up assessment Jadad ratings did not necessarily yield large ESs,

© 2016 Association for Child and Adolescent Mental Health.

10 M.J. Bakker et al.

An alternative explanation is that different interven-

tions are used at different ages. For example, sys-
tematic treatment of children older than 10 years
might be more effective if accompanied by parent
training.
Our findings regarding the efficacy of psychologi-
cal treatment for CD-related problems are partly in
line with the current literature, in which small-to-
moderate effects have been reported (Excellence,
2013; Grove et al., 2008; Lundahl et al., 2006;
McCart et al., 2006; Weisz et al., 2013; Wilson &
Lipsey, 2007). The studies included in this analysis
made use of different raters (teacher, parent, chil-
dren and blinded observers) and different assess-
ment instruments. We noted that parents tended to
consider treatment more effective than did teachers
Figure 5 Forest plot conduct disorder problems, comparison or blinded observers. This might be because, unlike
of effect sizes (ESs) per study based on blinded rater report. blinded observers, parents expect treatment to be
■ [Name study]-All Weighted ES based on multiple aggression effective (Sonuga-Barke et al., 2013). That teachers
outcomes. Abbreviations studies: NR, not randomized to considered treatment to be less effective than par-
treatment group
ents did, might be because children are slower to
implement acquired skills in the classroom than at
although the statistical power to identify such effects home or because peer-pressure at school prevents
was relatively low, as reported in another study them from doing so. Therefore, psychological treat-
(Field & Gillett, 2010). ment should not only impart skills but also take into
account the school environment, so that children are
encouraged to use these skills (Berryhill & Prinz,
Discussion 2003). Interventions may be more effective if they
We carried out this systematic review and meta- also target teachers (e.g. helping teachers adapt their
analysis to investigate the effect of nonpharmacolog- behaviour so that negative interactions with children
ical interventions for CD problems and to provide are reduced) or peers.
recommendations for future research, with a view to In contrast, on the basis of self-report information
optimizing clinical outcomes in children/adolescents provided by the children/adolescents, we did not
with CD. Although we specifically searched data- find psychological interventions to have a significant
bases for nonpharmacological treatments other than effect. This might be because children/adolescents
psychological treatments, all identified studies used are usually less inclined to report on their external-
psychological treatments. Overall, we found that CD izing behaviour (Smith, Pelham, Gnagy, Molina, &
is more common among males than females and that Evans, 2000). The self-report measures of CD prob-
based on information provided by teachers, parents lems used in the two studies that included self-
and blinded observers, psychosocial treatments have reported data (Hendriks et al., 2012; Sundell et al.,
a small but significant effect in reducing CD problems 2008) have been shown to have a high internal
in children/adolescents. However, based on the chil- consistency and re-test validity. The review by Grove
dren/adolescents’ ratings, these interventions were et al. (2008) suggested that children who have com-
not effective. Within the same rater and between pleted psychological treatment are better at avoiding
raters, different instruments evaluating the same being caught by parents, teachers and schools than
construct of CD problems yielded widely different before participation. Therefore, children might con-
ESs. Treatment effects remained significant when we tinue to exhibit antisocial behaviour at the same rate
focused on the instrument with the lowest ES, to and intensity as before treatment but they are not
account for this variability. While our findings sup- caught doing so; however, they continue to report
port the use of psychological treatments for CD, there exhibiting this behaviour.
is a lack of evidence about what the best treatment is. Conduct disorder behaviour may be exhibited in
While gender, age, number of sessions, duration, different ways (e.g. covert vs. overt aggression) and in
intervention type, setting, medication use or dropout different contexts (e.g. home vs. school). In the
percentage did not influence treatment efficacy, our studies included in this review, CD problem out-
findings suggested that treatment was less effective in comes were mainly scored with the CBCL and the
older children (>10 years) than in younger children ECBI questionnaires. In the CBCL, information
(<10 years), based on parent-rated outcomes. It is about misconduct behaviour (both aggression and
possible that the diminished efficacy of treatment in delinquency) is represented in the Externalizing
older children is because treatment is often delayed in scale, and in the ECBI, the Intensity scale represents
children with early-onset CD and/or CD problems. misconduct behaviour. Unfortunately, the two scales

© 2016 Association for Child and Adolescent Mental Health.

 Psychological treatments for conduct disorder 11

are not comparable at the level of individual items. versus unblinded raters, because only two primary
Furthermore, the ECBI lacks information about how studies used blinded observer report, (d) the influ-
children or adolescents act at home and at school, ence of callous-unemotional traits and/or subtypes
whereas the CBCL lacks information on the fre- of CD problems (e.g. covert), and (e) onset of CD (i.e.
quency of misconduct behaviour. Thus, while the early or late onset), which was reported in only one
two questionnaires are complementary in observing individual study. Surprisingly, nearly half of the
and rating CD problem behaviour, they may miss included studies did not report data on comorbidity,
important information about function (e.g. covert vs. not even on ADHD, which is the most common
overt; Olson et al., 2013) and subtypes of aggression comorbidity in cases with CD.
(e.g. proactive vs. reactive; Raine et al., 2006). The Another limitation is that our standardized quality
predominant subtype of aggression may influence assessment (Jadad scale) reflects the information
the focus of treatment focus, which in turn may lead provided in individual study reports and may not
to better outcomes. For example, in the case of fairly represent the trials themselves (Jadad et al.,
predominant reactive aggression, treatment could be 1996). For example, there is some debate about the
focused on improving cognitive control and improv- use of the Jadad scale, which includes double
ing strategies for controlling negative emotions of blinding, because the design of many psychological
distress, frustration and anger. In the case of pre- trials makes it difficult to blind patients or use
dominant proactive aggression, the focus of treat- blinded raters.
ment could be on improving sensitivity to moral
issues and moral reasoning and on increasing emo-
Recommendations for future research
tional empathy (Blair, 2013).
Although we did not find the dropout rate to The small ES of psychological interventions may lead
influence treatment efficacy, the dropout varied to two conclusions that are not necessarily opposing,
considerably in the different studies. Unfortunately, namely that CD problems are persistent and rather
the studies did not provide information about the refractory to treatment and that psychological inter-
characteristics of the dropouts, so we could not ventions for CD problems could be improved, to
analyse whether dropout is related to the subtype of make them more effective. Future studies should
aggression. Interventions for young children address a number of shortcomings identified in our
(aged < 10 years) have the benefit that the involve- review. First, we found evidence for rater effects so
ment of parents and/or guardians means that they that future studies should integrate information
can make sure the child attends treatment sessions. from multiple informants (e.g. parents, teachers
This is in contrast to interventions for older children and blinded observers) and assess CD problems in
(>10 years), where treatment compliance is less more than one environment (e.g. home and school
likely to be under parent and/or guardian control. situations). Second, we need to understand whether
For this reason, the dropout rate might be higher treatments are more effective in certain subgroups,
among older children. classified by the time of onset, the presence and
We did not find treatment duration to influence severity of callous-unemotional symptoms and the
treatment efficacy. Thus, while one could infer that subtype of aggression. Third, future randomized
shorter treatments are as effective as longer treat- controlled trials should be more precise in reporting
ments, this would need to be confirmed in a cost– their methods of randomization and blinding, the
benefit analysis that includes the actual cost of fate of all patients (including dropouts) within the
treatment over time and the individual’s involvement trial and the medication used by participants, all of
in different systems. which might influence the effect of treatment on CD
outcomes. Fourth, more randomized controlled trials
with large samples are needed because there have
Limitations
been relatively few such studies, which limits the
This review overcame some of the shortcomings of generalizability of findings and makes it difficult to
previous studies by including randomized controlled evaluate possible moderators and mechanisms of
trials with a specific target group and incorporating change. Larger, possibly multisite, studies are
sensitivity analyses to take possible moderators (e.g. needed to optimize psychological treatment efficacy
study quality) into account. However, we could not and maintenance. Fifth, as the nonpharmacological
control for other potential modifying factors because interventions in the current meta-analysis were
of a lack of information or a small number of studies, behavioural/psychosocial treatments, there is a
such as those involving an active control interven- need to investigate the effectiveness of other psycho-
tion (e.g. treatment as usual). For instance, we were logical interventions, such as dietary interventions
unable to control for the following aspects: (a) and cognitive training. One study suggested that
potential effects of psychiatric medication used by food intolerance (based on primarily ADHD and/or
participants, (b) potential gender effects primarily CD) and deficient intake of either micronutrients
due to the small sample sizes of the CD cohorts in (e.g. vitamins) or fatty acids (e.g. omega-3) may
general and of females in particular, (c) blinded predispose imprisoned delinquents to aggressive

© 2016 Association for Child and Adolescent Mental Health.

12 M.J. Bakker et al.

behaviour (Benton, 2007). Finally, while we found

Conclusion
that intervention duration (in hours) did not moder-
The current review demonstrates that psychological
ate study outcomes, few studies explicitly investi-
treatments for children/adolescents with CD and
gated this and future studies should establish
CD-related problems are modestly effective, as rated
whether there is an optimal treatment duration. To
by parents, teachers and blinded observers, but not
the best of our knowledge, guidelines are not con-
by children; however, the ES varied according to the
sistent about the duration of treatment, and so we
instrument used. There is not enough evidence to
would suggest a standardized treatment duration of
support one type of psychological treatment over
3 months, in order to allow interstudy comparison of
another. We could not compare treatment efficacy
treatment efficacy.
between children/adolescents based on self-report
and blinded observer report because of the limited
Implications for healthcare policy and clinical data available. In addition, too few studies examined
practice the possible moderating role of participant and study
characteristics, and so more research is needed in
This review suggests that psychological treatments
this area. The overall quality of the studies was poor,
have modest efficacy in reducing problems in chil-
and many studies failed to provide important details,
dren/adolescents with diagnosed CD or with clinical
such as on subtypes of aggression and presence of
CD problems and that treatments appear to be more
callous-unemotional traits, information that is
effective in younger children (<10 years) than in
needed to further optimize psychological treatment
older children (>10 years). The limited evidence
strategies. Future studies of CD should pay close
based on parent-reported information suggests that,
attention to these details, be of high quality and be
regardless of group- or individual-focused sessions,
adequately powered.
the following treatment programmes may be espe-
cially effective in reducing CD problems: (a) the
parent–child-based intervention: ‘Parent–Child
Acknowledgements
Interaction Therapy’, and the multicomponent inter-
The study is supported by the European Union 7th
vention ‘Incredible Years Program plus Child Liter- Framework Programs AGGRESSOTYPE (602805)
acy Program’ [in an emotionally dysregulated (e.g. and MATRICS (603016) (to J.C.G. and J.K.B.) and
angry and resentful) subgroup as opposed to a head- by a NWO Brain & Cognition grant (056-24-011) (to
strong (e.g. arguing with adults) subgroup]; and (b) J.K.B).
two specific parent-focused interventions: ‘Parent- J.K.B. has been a consultant to/member of advisory
Training Hitkashrut’ and ‘Project Support’. In addi- board of and/or speaker for Janssen Cilag BV, Eli Lilly,
tion, teacher-reported information suggested that a UCB, Shire, Novartis, Roche, Lundbeck, and Servier.
multicomponent group-focused treatment pro- J.C.G. has been a consultant to Boehringer Ingelheim
gramme such as ‘Parent Training plus Child Train- GmbH. Neither are employees of these companies nor
ing’ is also effective in reducing CD problems. are they shareholders in these companies. They have no
other financial or material support, including expert
However, as to date only single trials of specific
testimony, patents and royalties. The authors have
types of treatment for children/adolescents with declared that they have no potential or competing
particular clinical CD diagnoses and/or CD prob- conflicts of interest in relation to this report.
lems in the clinical range have been performed, there
is a need for further research. The finding that
psychological interventions had a modest effect not Correspondence
only on CD symptoms but also on CD-related Mireille J. Bakker, Karakter Child- and Adolescent
problems, such as academic performance, in both Psychiatry University Centre, Reinier Postlaan 12,
the home and school environments is promising and 6525 GC Nijmegen, The Netherlands; Email: mireille.
shows that such interventions are broadly beneficial. bakker@radboudumc.nl

© 2016 Association for Child and Adolescent Mental Health.

 Psychological treatments for conduct disorder 13

Key Points
Key practitioner message

• Psychological treatments appear to be effective in reducing conduct disorder (CD) problems in children and
adolescents with clinically elevated CD problems and/or CD diagnosis.
• Effect sizes (ESs) are small, but significant, based on parent report (effect size, ES = 0.36), teacher report
(ES = 0.21) and blinded observer report (ES = 0.26). This suggests that these treatments are effective in
reducing CD problems across different raters and situations (e.g. home and school environments). However,
this is not the case when children/adolescents rated their CD problems; then psychological treatments appear
not to be effective.
• Effects are not limited to CD symptoms, but include to a range of CD problems including frequency of
misconduct behaviour, academic problems, and how children/adolescents behave at home and at school.
Nevertheless, ESs varied within the same rater and between raters, when different instruments measuring the
same construct of CD problems were used.
• Comorbidity, gender, age, number of sessions, duration, intervention type, setting, medication use or dropout
percentage do not appear to influence the effect of treatment.
• There is tentative evidence that treatment may be more effective in children younger than 10 years.
• In the light of current evidence, psychological treatment is recommended for children/adolescents with CD
problems. Future studies should address a number of key shortcomings to further bolster this recommenda-
tion.

Areas for future research

• There is not enough evidence to support one specific psychological treatment over another. More research is
needed comparing specific interventions.
• Studies included in this meta-analysis used a range of different outcome measures for CD problems. While this
allows generalization of treatment effects across a range of CD problems, future studies should use a more
homogeneous set of outcome measures to improve comparability across studies.
• To allow better evaluation of the quality of studies, future randomized controlled trials should provide
detailed information on their methods of randomization and blinding and on the fate of all trial participants
(including dropouts).
• In order to improve treatment efficacy, future studies should pay greater attention to the role of participant
characteristics, such as CD onset, presence of and severity of callous-unemotional traits, subtype of aggression
(e.g. proactive vs. reactive aggression) and severity of aggression.
• The nonpharmacological interventions included in the meta-analysis were psychosocial treatments. More
research is needed into the effectiveness of other intervention types, such as diet and cognitive training.

Borenstein, M., Hedges, L.V., Higgins, J.P., & Rothstein, H.R.

References (2010). A basic introduction to fixed-effect and random-
Achenbach, T.M. (1991a). Manual for the child behavior effects models for meta-analysis. Research Synthesis Meth-
checklist/4–18 and 1991 profile. Burlington: Department of ods, 1, 97–111.
Psychiatry, University of Vermont. Boylan, K., Macpherson, H.A., & Fristad, M.A. (2013). Exam-
Achenbach, T.M. (1991b). Manual for the teacher’s report form ination of disruptive behavior outcomes and moderation in a
and 1991 profile. Burlington: Department of Psychiatry, randomized psychotherapy trial for mood disorders. Journal
University of Vermont. of the American Academy of Child and Adolescent Psychia-
American Psychiatric Association. (2013). Available from: try, 52, 699–708.
www.dsm5.org [last accessed 3 December 2013]. Buitelaar, J.K., Smeets, K.C., Herpers, P., Scheepers, F.,
Bagner, D.M., Sheinkopf, S.J., Vohr, B.R., & Lester, B.M. Glennon, J., & Rommelse, N.N. (2013). Conduct disorders.
(2010). Parenting intervention for externalizing behavior European Child and Adolescent Psychiatry, 22(Suppl. 1),
problems in children born premature: An initial examina- S49–S54.
tion. Journal of Developmental and Behavioral Pediatrics, 31, Conduct Problems Prevention Research Group (1999). Initial
209–216. impact of the fast track prevention trial for conduct prob-
Benton, D. (2007). The impact of diet on anti-social, violent lems: II. Classroom effects. Journal of Consulting and
and criminal behaviour. Neuroscience and Biobehavioral Clinical Psychology, 67, 648–657.
Reviews, 31, 752–774. Conduct Problems Prevention Research Group (2011). The
Berryhill, J.C., & Prinz, R.J. (2003). Environmental interven- effects of the fast track preventive intervention on the
tions to enhance student adjustment: Implications for development of conduct disorder across childhood. Child
prevention. Prevention Science, 4, 65–87. Development, 82, 331–345.
Blair, R.J. (2013). The neurobiology of psychopathic traits in Crowther, M., Lim, W., & Crowther, M.A. (2010). Systematic review
youths. Nature Reviews Neuroscience, 14, 786–799. and meta-analysis methodology. Blood, 116, 3140–3146.

© 2016 Association for Child and Adolescent Mental Health.

14 M.J. Bakker et al.

Drugli, M.B., & Larsson, B. (2006). Children aged 4–8 years behavior disorders: A systematic review of efficacy and
treated with parent training and child therapy because of tolerability. Current Psychiatry Reports, 15, 355.
conduct problems: Generalisation effects to day-care and Loeber, R., Burke, J.D., Lahey, B.B., Winters, A., & Zera, M.
school settings. European Child and Adolescent Psychiatry, (2000). Oppositional defiant and conduct disorder: A review
15, 392–399. of the past 10 years, part I. Journal of the American Academy
Eyberg, S.M., Boggs, S.R., & Jaccard, J. (2014). Does main- of Child and Adolescent Psychiatry, 39, 1468–1484.
tenance treatment matter? Journal of Abnormal Child Psy- Loeber, R., & Stouthamer-Loeber, M. (1998). Development of
chology, 42, 355–366. juvenile aggression and violence. Some common miscon-
Eyberg, S.M., Nelson, M.M., & Boggs, S.R. (2008). Evidence- ceptions and controversies. American Psychologist, 53,
based psychosocial treatments for children and adolescents 242–259.
with disruptive behavior. Journal of Clinical Child and Lundahl, B., Risser, H.J., & Lovejoy, M.C. (2006). A meta-
Adolescent Psychology, 37, 215–237. analysis of parent training: Moderators and follow-up
Fairchild, G., van Goozen, S.H., Calder, A.J., & Goodyer, I.M. effects. Clinical Psychology Review, 26, 86–104.
(2013). Research review: Evaluating and reformulating the McCart, M.R., Priester, P.E., Davies, W.H., & Azen, R. (2006).
developmental taxonomic theory of antisocial behaviour. Differential effectiveness of behavioral parent-training and
Journal of Child Psychology and Psychiatry, 54, 924–940. cognitive-behavioral therapy for antisocial youth: A meta-
Field, A.P., & Gillett, R. (2010). How to do a meta-analysis. analysis. Journal of Abnormal Child Psychology, 34, 527–
British Journal of Mathematical and Statistical Psychology, 543.
63(Pt 3), 665–694. McDonald, R., Dodson, M.C., Rosenfield, D., & Jouriles, E.N.
Frick, P.J., & White, S.F. (2008). Research review: The impor- (2011). Effects of a parenting intervention on features of
tance of callous-unemotional traits for developmental mod- psychopathy in children. Journal of Abnormal Child Psychol-
els of aggressive and antisocial behavior. Journal of Child ogy, 39, 1013–1023.
Psychology and Psychiatry, 49, 359–375. Moffitt, T.E. (2006). Life-course persistent versus adolescence-
Grove, A.B., Evans, S.W., Pastor, D.A., & Mack, S.D. (2008). A limited antisocial behavior (Vol. 3). New York: Wiley.
meta-analytic examination of follow-up studies of programs Moffitt, T.E., Caspi, A., Dickson, N., Silva, P.A., & Stanton,
designed to prevent the primary symptoms of oppositional W. (1996). Childhood-onset versus adolescent-onset anti-
defiant and conduct disorders. Aggression and Violent social conduct problems in males: Natural history from
Behavior, 13, 169–184. ages 3 to 18 years. Development and Psychopathology, 8,
Hanisch, C., Freund-Braier, I., Hautmann, C., Janen, N., 399–424.
Pluck, J., Brix, G., . . . & Dopfner, M. (2010). Detecting Morris, S.B. (2008). Estimating effect sizes from pretest-
effects of the indicated prevention Programme for External- posttest-control group designs. Organizational Research
izing Problem behaviour (PEP) on child symptoms, parent- Methods, 11, 364–384.
ing, and parental quality of life in a randomized controlled National Institute for Health and Clinical Excellence. (2013).
trial. Behavioural and Cognitive Psychotherapy, 38, 95– Antisocial behaviour and conduct disorders in children and
112. young people: Recognition, intervention and management
Hendriks, V., van der Schee, E., & Blanken, P. (2012). (National clinical guideline 158). British Psychological Soci-
Matching adolescents with a cannabis use disorder to ety, Royal College of Psychiatrists. Available from: http://
multidimensional family therapy or cognitive behavioral www.nice.org.uk/guidance/CG158 [last accessed 4 Febru-
therapy: Treatment effect moderators in a randomized ary 2014].
controlled trial. Drug and Alcohol Dependence, 125, 119– Olson, S.L., Sameroff, A.J., Lansford, J.E., Sexton, H., Davis-
126. Kean, P., Bates, J.E., . . . & Dodge, K.A. (2013). Deconstruct-
Higgins, J.P., Thompson, S.G., Deeks, J.J., & Altman, D.G. ing the externalizing spectrum: Growth patterns of overt
(2003). Measuring inconsistency in meta-analyses. BMJ, aggression, covert aggression, oppositional behavior, impul-
327, 557–560. sivity/inattention, and emotion dysregulation between
Jadad, A.R., Moore, R.A., Carroll, D., Jenkinson, C., Reynolds, school entry and early adolescence. Development and Psy-
D.J., Gavaghan, D.J., & McQuay, H.J. (1996). Assessing the chopathology, 25, 817–842.
quality of reports of randomized clinical trials: Is blinding Owens, J.S., Murphy, C.E., Richerson, L., Girio, E.L., &
necessary? Controlled Clinical Trials, 17, 1–12. Himawan, L.K. (2008). Science to practice in underserved
Jones, D.J., Forehand, R., Cuellar, J., Parent, J., Honeycutt, communities: The effectiveness of school mental health
A., Khavjou, O., . . . & Newey, G.A. (2014). Technology- programming. Journal of Clinical Child and Adolescent
enhanced program for child disruptive behavior disorders: Psychology, 37, 434–447.
Development and pilot randomized control trial. Journal of Perrin, E.C., Sheldrick, R.C., McMenamy, J.M., Henson, B.S.,
Clinical Child and Adolescent Psychology, 43, 88–101. & Carter, A.S. (2014). Improving parenting skills for families
Jouriles, E.N., McDonald, R., Rosenfield, D., Stephens, N., of young children in pediatric settings: A randomized clinical
Corbitt-Shindler, D., & Miller, P.C. (2009). Reducing con- trial. JAMA Pediatrics, 168, 16–24.
duct problems among children exposed to intimate partner Pilling, S., Gould, N., Whittington, C., Taylor, C., Scott, S., &
violence: A randomized clinical trial examining effects of Guideline Development, G. (2013). Recognition, interven-
Project Support. Journal of Consulting and Clinical Psychol- tion, and management of antisocial behaviour and conduct
ogy, 77, 705–717. disorders in children and young people: Summary of NICE-
Kolko, D.J., Campo, J.V., Kelleher, K., & Cheng, Y. (2010). SCIE guidance. BMJ, 346, f1298.
Improving access to care and clinical outcome for pediatric Raine, A., Dodge, K., Loeber, R., Gatzke-Kopp, L., Lynam, D.,
behavioral problems: A randomized trial of a nurse-admi- Reynolds, C., . . . & Liu, J. (2006). The reactive-proactive
nistered intervention in primary care. Journal of Develop- aggression questionnaire: Differential correlates of reactive
mental and Behavioral Pediatrics, 31, 393–404. and proactive aggression in adolescent boys. Aggressive
Kolko, D.J., Lindhiem, O., Hart, J., & Bukstein, O.G. (2013). Behavior, 32, 159–171.
Evaluation of a booster intervention three years after acute Review Manager (RevMan) [Computer program]. Version 5.3.
treatment for early-onset disruptive behavior disorders. (2014). Copenhagen: The Nordic Cochrane Centre, The
Journal of Abnormal Child Psychology, 42, 383–398. Cochrane Collaboration.
Linton, D., Barr, A.M., Honer, W.G., & Procyshyn, R.M. (2013). Ried, K. (2006). Interpreting and understanding meta-analysis
Antipsychotic and psychostimulant drug combination ther- graphs–a practical guide. Australian Family Physician, 35,
apy in attention deficit/hyperactivity and disruptive 635–638.

© 2016 Association for Child and Adolescent Mental Health.

 Psychological treatments for conduct disorder 15

Robinson, E.A., Eyberg, S.M., & Ross, A.W. (1980). The and psychological treatments. American Journal of Psychia-
standardization of an inventory of child conduct prob- try, 170, 275–289.
lem behaviors. Journal of Clinical Child Psychology, 9, Sundell, K., Hansson, K., Lofholm, C.A., Olsson, T., Gustle,
22–28. L.H., & Kadesjo, C. (2008). The transportability of multisys-
Rosato, N.S., Correll, C.U., Pappadopulos, E., Chait, A., temic therapy to Sweden: Short-term results from a ran-
Crystal, S., & Jensen, P.S. (2012). Treatment of maladaptive domized trial of conduct-disordered youths. Journal of
aggression in youth: CERT guidelines II. Treatments and Family Psychology, 22, 550–560.
ongoing management. Pediatrics, 129, e1577–e1586. Thalheimer, W., & Cook, S. (2002). How to calculate effect sizes
Scheepers, F.E., Buitelaar, J.K., & Matthys, W. (2010). Con- from published research articles: A simplified methodology.
duct Disorder and the specifier callous and unemotional Available from: http://work-learning.com/effect_sizes.htm.
traits in the DSM-5. European Child and Adolescent Psychi- [last accessed 23 August 2012].
atry, 20, 89–93. van Manen, T.G., Prins, P.J., & Emmelkamp, P.M. (2004).
Scott, S., & O’Connor, T.G. (2012). An experimental test of Reducing aggressive behavior in boys with a social cognitive
differential susceptibility to parenting among emotionally- group treatment: Results of a randomized, controlled trial.
dysregulated children in a randomized controlled trial for Journal of the American Academy of Child and Adolescent
oppositional behavior. Journal of Child Psychology and Psychiatry, 43, 1478–1487.
Psychiatry, 53, 1184–1193. von Sydow, K., Retzlaff, R., Beher, S., Haun, M.W., &
Smith, H., & Coghill, D. (2010). Pharmacotherapy for children Schweitzer, J. (2013). The efficacy of systemic therapy for
and adolescents with conduct problems. In R.C. Murrihy, childhood and adolescent externalizing disorders: A system-
A.D. Kidman & T.H. Ollendick (Eds.), Clinical handbook of atic review of 47 RCT. Family Process, 52, 576–618.
assessing and treating conduct problems in youth (Vol. 2, pp. Weisz, J.R., Kuppens, S., Eckshtain, D., Ugueto, A.M., Hawley,
383–404). New York: Springer New York. K.M., & Jensen-Doss, A. (2013). Performance of evidence-
Smith, B.H., Pelham, W.E., Jr, Gnagy, E., Molina, B., & Evans, based youth psychotherapies compared with usual clinical
S. (2000). The reliability, validity, and unique contributions care: A multilevel meta-analysis. JAMA Psychiatry, 70, 750–
of self-report by adolescents receiving treatment for atten- 761.
tion-deficit/hyperactivity disorder. Journal of Consulting Wilson, S.J., & Lipsey, M.W. (2007). School-based interven-
and Clinical Psychology, 68, 489–499. tions for aggressive and disruptive behavior: Update of a
Somech, L.Y., & Elizur, Y. (2012). Promoting self-regulation meta-analysis. American Journal of Preventive Medicine, 33
and cooperation in pre-kindergarten children with conduct (2 Suppl.), S130–S143.
problems: A randomized controlled trial. Journal of the Woolfenden, S., Williams, K., & Peat, J. (2001). Family and
American Academy of Child and Adolescent Psychiatry, 51, parenting interventions in children and adolescents with
412–422. conduct disorder and delinquency aged 10–17. Cochrane
Sonuga-Barke, E.J., Brandeis, D., Cortese, S., Daley, D., Database Systematic Review, 2, CD003015.
Ferrin, M., Holtmann, M., . . . & Sergeant, J. (2013). Non-
pharmacological interventions for ADHD: Systematic review Accepted for publication: 4 April 2016
and meta-analyses of randomized controlled trials of dietary

© 2016 Association for Child and Adolescent Mental Health.