Table 5.

Methodological quality of included studies.

Study	Treatment	Control	NHMRC evidence level	Randomization	Blinding	Methodological quality
Abikoff et al. [53]	Child and parent training with medication	Attention control with medication	II	Block randomization scheme with blocks of 4 children. The groups were balanced for age, sex, oppositional defiant disorder, and ethnicity	Trained observers, blind to treatment and diagnosis, observed the study children and classmates as a primary outcome	Strong quality (score 23/28). Reliable use of peers. Sampling strategy appropriate with subject characteristics sufficiently described in companion methodological article presented by Klein et al. (2004). Insufficient data to assess sample size. Blinding for one of the primary outcomes was reported. Randomization not reported in detail. No estimates of variance reported.
Choi and Lee [54]	Child training	Waitlist control	II	Block randomization scheme with blocks of 5 children. No evidence of stratification	No blinding of participants or personnel reported. All instruments were either parent- or child-report measures	Good quality score (18/28). Reliable use of peers. Sampling strategy appropriate with but only subject age, gender and school grade described. Blinding not reported for outcome measurement. Insufficient data to assess sample size. Estimates of variance provided for time points of interest, but not around the difference.
Frankel et al. [55]	Child and parent training with medication	Waitlist control with medication	III	No randomization evident	No blinding of participants or personnel reported	Good quality (score 18/28). Reliable use of peers. Sampling strategy appropriate with subject characteristics sufficiently described. Insufficient data to assess sample size. Blinding and randomization not reported. Authors reported the treatment and waitlist groups differed significantly in mean socioeconomic status however did not correlate with any outcome variable. Authors reported ADHD children were prescribed medication by their own private physicians and dosage was not verified by the present authors thus possibly affecting the results.
Guli et al. [56]	Child training	Clinical control	III	No randomization evident	No blinding of participants or personnel reported	Good quality (score 20/28). Reliable use of peers. Sampling strategy appropriate with subject characteristics sufficiently described. Insufficient data to assess sample size. Blinding and randomization not reported. Estimates of variance provided. Results reported in sufficient detail with supporting conclusions. Authors reported some children in the treatment and control group were taking prescription medication thus possibly affecting the results.
Haas et al. [57]	Child training	Child training for no diagnosis control	III	No randomization evident	Counsellors who rated children's behaviour were naïve to the conduct problems and callous/unemotional traits status of each child	Good quality (score 20/28). Reliable use of peers. Sampling strategy not described. Subject characteristics sufficiently described. Blinding reported for one of the primary outcomes. Randomization not reported. Authors reported a relatively small sample with limited power to detect trends. As the sample only included children with high levels of CP and ADHD the data can only be safely generalised to children with high levels of CP and ADHD. Treatment outcomes may have been affected by medicated and unmedicated behaviour as some children were on medication for some (but not all) days therefore this may have affected results.
Hannesdottir et al. [58]	Child training	Waitlist control	III	Randomization procedure not described	No blinding of participants or personnel reported	Good quality score (18/28). Reliable use of peers. Sampling strategy appropriate with subject characteristics sufficiently described. Randomization was reported, but procedure not described. Blinding not reported. Insufficient information to calculate sample size. No estimates of variance around differences reported.
Hantson et al. [59]	Child and parent training	Treatment as usual control	III	No randomization evident	No blinding of participants or personnel reported	Good quality (score 19/28). Reliable use of peers. Sampling strategy appropriate with subject characteristics sufficiently described, however comorbidities were not reported. Blinding and randomization not reported. Relatively small sample size. No estimates of variance reported.
Huang et al. [60]	Child and parent training	Community care control	II	No randomization evident	No blinding of participants or personnel reported	Good quality score (17/28). Reliable use of peers. Sampling strategy appropriate with subject characteristics sufficiently described. Control group consisted of families interested in the program but unable to attend at specific appointment times, thus not truly randomized. No blinding reported. Insufficient information to calculate sample size. No estimates of variance around differences reported. Authors recognise that participants with good medication compliance had better outcomes on some measures, but did not control for medication compliance in all analyses.
Jensen et al. [61]	Child, parent, and school-based training with medication	Community care control	II	Randomization was done centrally and stratified by site in blocks of 16 (4 to each group). Sealed, ordered envelopes were sent to sites for successive entries. Treatment assignment was concealed until the family confirmed agreement to accept randomization	No blinding of personnel reported	Good quality (score 21/28). Reliable use of peers. Sampling strategy appropriate with subject characteristics sufficiently described. Blinding was reported however it is not clear who was blinded. Randomization not described in detail. Appropriate sample size. Variance estimates reported inappropriately as study provided variance around the parameters of interest however not around the difference.
Kolko et al. [62]	Child training	Child socialisation group	III	No randomization evident	Second trained research assistant unaware of child's group assignment recorded in vivo behavioural observations. Interrater agreement of behavioural role-play test were assessed by comparing ratings assigned by a trained research assistant unaware of group assignment	Good quality (score 18/28). Reliable use of peers. Sampling strategy may have introduced bias. Subject characteristics sufficiently described. Blinding was adequately reported for some of the raters however not all. Randomization was not reported. No estimates of variance reported. Appropriate sample size.
Mikami et al. [63]	Child training for ADHD children	Child training for TD children	II	Randomly assigned via a computer-generated sequence either to a classroom in the MOSAIC treatment condition in Session 1 and a different classroom in the COMET treatment condition in Session 2 or vice versa. Assignment was stratified by child age and sex	Trained research assistants unaware of treatment group administered all primary outcome measures	Strong quality (score 25/28). Reliable use of peers. Sampling strategy appropriate with subject characteristics sufficiently described. Blinding was adequately reported for all outcome measures. Randomization was adequately reported. Estimates of variance provided. Authors reported a small sample size which limits the confidence for moderation results.
The MTA Cooperative Group [64]	Child, parent, and school-based training with medication	Community care control	II	Randomization was done centrally and stratified by site in blocks of 16 (4 to each group). Sealed, ordered envelopes were sent to sites for successive entries. Treatment assignment was concealed until the family confirmed agreement to accept randomization	The open parent, teacher, and child ratings for 5 out of 6 outcomes were augmented by blinded ratings of school-based ADHD and oppositional/aggressive symptoms. Raters blind to treatment condition performed standardized laboratory tasks to assess parent-child interactions	Strong quality (score 24/28). Reliable use of peers. Sampling strategy appropriate with subject characteristics sufficiently described. Blinding was adequately reported. Randomization not reported in detail. Appropriate sample size based on power analyses. Explicit use of manualised, evidence-based treatments and comprehensive range of outcome assessments. Variance estimates reported inappropriately as study provided variance around the parameters of interest however not around the difference.
MTA Cooperative Group [65]	Child, parent, and school-based training with medication	Community care control	II	Randomization was done centrally and stratified by site in blocks of 16 (4 to each group). Sealed, ordered envelopes were sent to sites for successive entries. Treatment assignment was concealed until the family confirmed agreement to accept randomization	No blinding of personnel reported	Good quality (score 21/28). Reliable use of peers. Sampling strategy appropriate with subject characteristics sufficiently described. Blinding not reported. Randomization not described in detail. Appropriate sample size. Variance estimates reported inappropriately as study provided variance around the parameters of interest however not around the difference.
Pfiffner et al. [66]	Child and parent training with teacher consultation	Waitlist control or treatment as usual	II	Randomization was stratified by sex (when two children of the same sex were identified, one was randomly assigned to CLAS program and one to the control group). Investigators required at least two participants of the same sex in the treatment group	Interviewers and raters who were blind to child's group assignment administered the Test of Life Skill Knowledge outcome measure	Good quality (score 22/28). Reliable use of peers. Sampling strategy appropriate with subject characteristics sufficiently described. Blinding was adequately reported. Randomization was not reported in detail. No estimates of variance reported. Appropriate sample size. Participant use of medication was a confounding variable however unlikely to have seriously distorted results.
Shechtman and Katz [67]	Child training for ADHD children	Waitlist control	II	The group was randomly divided (by alphabetical order) into an experimental and a waitlist control group (to be treated the next year)	No blinding of participants or personnel reported	Strong quality (score 24/28). Reliable use of peers. Sampling strategy appropriate with subject characteristics sufficiently described. Blinding not reported. Randomization was adequately reported. Estimates of variance provided. Appropriate sample size. Participant use of medication was not reported therefore may be confounding.
Storebø et al. [68]	Child and parent training with standard treatment	Standard treatment alone	II	The Copenhagen Trial Unit conducted central randomization with computer-generated, permuted randomization sequences in blocks of four with an allocation ratio of 1 : 1 stratified for sex and comorbidity	The interventions given were not “blind” to participants, parents, treating physicians, or personnel. However, the outcome assessors (teachers) were kept blinded of the allocated intervention. Blinded data were then handed over for data entry and statistical analyses	Strong quality (score 26/28). Reliable use of peers. Sampling strategy appropriate with subject characteristics sufficiently described. Blinding was adequately reported for all outcome measures. Blinding was not apparent for participants. Randomization was reported in detail. Estimates of variance provided. Sample size calculation completed with an appropriate sample size collected.
Waxmonsky et al. [69]	Child and parent training with school-based daily report card and medication	Medication alone	II	One-half of the subjects were randomly assigned to receive atomoxetine + BT and the remaining subjects randomly assigned to receive atomoxetine alone	The second (reliability) observer for the behaviour therapy arm was blinded to group assignment	Good quality (score 22/28). Reliable use of peers. Sampling strategy appropriate with subject characteristics sufficiently described. Primary investigators were not blinded. Randomization was not reported in detail. Variance estimates reported inappropriately as study provided variance around the parameters of interest however not around the difference. Sample size was calculated for the primary outcome measure however not the secondary measures therefore sample may not have had sufficient subjects to detect group differences for secondary measures.

Notes. NHMRC level II = RCTs; level III = quasi-experimental designs without random allocation; ADHD = Attention-Deficit Hyperactivity Disorder; CP = conduct problems; MOSAIC = Making Socially Accepting Inclusive Classrooms; COMET = contingency management training; and CLAS = Child Life and Attention Skills program.