Cummings 2000.

Methods	Design: Cross‐over randomised controlled trial Unit of randomisation: Individual family Follow‐up: 4 weeks Study dates: 1999
Participants	Participants: Parents (32 mothers; 5 fathers) and other caregivers of children with negative behaviour Mean age of parents: Not reported Age of child: 2 and 3 years (mean, SD, range: NR); 17 boys and 11 girls Ethnicity: 54% Black, 46% White Number randomised: 37 (intervention 21; control 16)* Country & setting: USA; single‐site; recruited from community settings; intervention delivered in health centre or at drug treatment centre Eligibility criteria: Not specified
Interventions	Two conditions: Group‐based parent training (not specified); wait‐list control Duration of intervention: 6 (1.5 hours) sessions over 3 weeks Therapist training: Researcher
Outcomes	Secondary outcomes: Parent‐child interaction as measured by the Dyadic Parent‐Child Coding System, reported by observing clinical staff Timing of outcome: postintervention and at four‐week follow‐up Primary outcomes/adverse events: No other outcomes relevant to this review specified
Notes	*Six participants who were initially in the wait‐list control group were admitted in subsequent cohorts and were included in the analysis twice over: “21 parents were assigned to the treatment group (six also had been in the wait list control group), and 16 were assigned to the Wait List Control group” (p 41) Level of prevention: Secondary or tertiary Funding: no information provided
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	A random sequence was generated by flipping a coin. Participants who were initially assigned to the wait‐list control group were offered the opportunity to participate in the intervention group after the first training series. Those who wanted to participate were placed in the intervention group in the next training series, while the new participants were randomly placed in either the intervention or wait‐list control group (see p 41 and Table 4, p 43). Participants who were initially randomly assigned to the wait‐list control group (n = 6) were not randomly assigned to the further intervention groups; the process of randomisation was compromised
Allocation concealment (selection bias)	Unclear risk	Information reported insufficiently for a judgement to be made
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participants: It would not have been possible to fully blind participants in this type of study (there is an obvious difference between receiving group training and being placed on a waiting list). Participants were made aware of their group allocation Personnel: The design of the study meant that personnel delivering the intervention were aware which groups had been assigned to the different study conditions
Blinding of outcome assessment (detection bias) Independent observer reported outcomes	High risk	All observational assessments were coded by the researcher, who was not blinded to group allocation
Blinding of outcome assessment (detection bias) Parent reported outcomes	Low risk	There were no parent‐reported outcomes
Incomplete outcome data (attrition bias) All outcomes	High risk	The study permitted control group members to cross‐over into the intervention arm; 6 out of 16 chose to do so. However, data is not clearly broken down (i.e. it is not clear where the final numbers for intervention outcomes came from and how many of original intervention arm dropped out)
Selective reporting (reporting bias)	Low risk	The published report included all expected outcomes, including those that were prespecified
Other bias	Unclear risk	Study investigators reported "The six parents who first participated in the Wait List Control group, then in the treatment group were counted twice for demographic purposes." (p 29). There is insufficient information to assess whether these items could introduce bias