Taylor 2005.

Methods	Randomised controlled trial (computer‐generated randomisation further balanced to ensure that participants matched on key variables) Setting: study conduct in the UK; inpatient forensic service (National Health Service) No study dates are provided
Participants	Inclusion criteria: male between 18 ‐ 60 years of age; full scale IQ between 55 to 80 (mild‐to‐borderline intellectual disability); detained on a section under the Mental health Act; self‐report score of 90 of more on the Novaco Anger Scale; self‐report score of 55 or more on the Provocation Index Exclusion criteria: presence of an active, uncontrolled axis 1 mental disorder (DSM IV); presence of epilepsy if thought to be related to episodes of aggression; and if patient was likely to be discharged or transferred within 6 months of starting treatment Demographic profile: 40 men. Mean age in intervention group 29.4; mean age in control group 29.9; mean IQ in treatment group 67.1; mean IQ in control group 70.7 25 of the participants had comorbid mental disorders Most of the participants had reported physical aggression, including sexual assault
Interventions	Participants were allocated to treatment and control groups, but also received treatment‐as‐usual. 20 participants in each group Treatment 1: Manual‐based individual cognitive‐behavioural treatment for anger management in people with mild intellectual disabilities Same therapist delivered 18 weekly sessions. Sessions split into a 6‐week preparatory phase and then 12 weeks of treatment Weekly peer supervision sessions for the therapists and random checks of treatment files to check treatment fidelity and therapist confidence Treatment 2: Routine care (RC) wait‐list control
Outcomes	Assessments were at baseline, immediately pre‐treatment, at completion of treatment, and 4 months post‐treatment Outcomes measures included self reports of Novacos Anger Scale and the Anger Expression Scale. Ward staff completed the Ward Anger Rating Scale over same time
Notes	No information in source of funding No declaration of conflicts of interest statement
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation was carried out by independent statistician using computer‐generated randomisation codes. The groups were subsequently "balanced" on the following variables: age, IQ, and length of hospital stay
Allocation concealment (selection bias)	Low risk	In an email (Taylor 2007 [pers comm], 4 September 2007) author reported "This was done by an assistant psychologist who was not involved in delivering treatment, administering study assessments, or in the routine care of the patients involved."
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participants and study personnel were not blinded to the allocation group
Blinding of outcome assessment (detection bias) All outcomes	High risk	Outcome assessors were not blinded to the treatment group
Incomplete outcome data (attrition bias) All outcomes	High risk	Data from 36 out of 40 participants were reported (10% loss). All dropouts were in the treatment group. 2 dropped out for reasons "unrelated to the treatment itself", and the other 2 were discharged from the hospital and lost to follow‐up. The researchers did not carry out an intention‐to‐treat analysis as the data of those four non‐completers were excluded from the final analysis
Selective reporting (reporting bias)	Low risk	All prespecified outcomes were reported
Other bias	Unclear risk	The effects of comorbidity and psychotropic medication on treatment response is unclear