We made eight amendments.

Why it is important to do this review: we amended the second paragraph in order to provide a more thorough description and analysis of previous reviews conducted in the area. Thus, the differences between the current and previous reviews are more clearly delineated.

Types of studies: in the protocol, we inadvertently omitted to mention that studies involving children with serious comorbid physical and intellectual impairments (for example, autism spectrum disorders, Down Syndrome, tic disorders, significant language delay and learning problems) were also excluded from the review.

Types of outcomes: in the protocol, we stated that the follow‐up should be at least three months following treatment. This proved to be too restrictive and it was amended, therefore, to include all short‐term outcomes, whether conducted immediately post‐treatment or up to three months following treatment.

Types of outcomes: in the protocol, we stated that we would investigate the impact of the intervention on the outcome of parenting skills. However, during the review process, it was necessary to differentiate between positive (for example, praise, positive affect, play, proactive discipline) and negative (for example, physical and verbal criticism, negative commands) parenting practices as the intervention could possibly effect change in either one or the other, in both or in none of them. Moreover, it was necessary to distinguish between positive and negative parenting practices in order to explore which aspects of parenting practices act as causal mechanisms within behavioural and cognitive‐behavioural group‐based parenting interventions.

Measures of treatment effect: dichotomous data for child conduct problems and child emotional problems are presented as risk ratios (RR) rather than as odds ratios (OR) as specified in the protocol. After seeking advice at Cochrane workshops, we understand that risk ratios are more commonly reported and are much easier for the reader to interpret and use.

Subgroup analyses: in the protocol, we stated that we would conduct a subgroup analysis on implementation fidelity, as measured by assessing the training and supervision of facilitators delivering the programme. However, in conducting the review, we found that this measure of fidelity was rather basic and incomplete and so we included additional measures of implementation fidelity, namely, evidence of adherence to protocols, exposure to the programme, quality of delivery and programme differentiation.

Sensitivity analyses: in the protocol, we stated that we would exclude studies with attrition rates larger than 20%. In order to provide a more rigorous examination of the impact of missing data, we added that we would also exclude studies without an intention‐to‐treat analysis. Furthermore, within the review, we provided a more operational definition of quasi‐randomisation, i.e. the removal of studies with inadequate sequence generation or inadequate allocation concealment. We also included one additional sensitivity analysis that was not specified in the protocol; in order to provide an overall picture of the effect of risk of bias on the meta‐analyses, we excluded any studies with evidence of risk of bias in any key domain of inadequate randomisation, blinding, or attrition greater than 20% in either the control or intervention groups. Lastly, the sensitivity analyses on fidelity and short‐term versus long‐term outcomes were not conducted within the current review as all of the studies reported on fidelity and all included studies reported only short‐term outcomes.