O'Neill 1996.

Methods	RCT.
Participants	92 pregnant women enrolled in MMT who injected drugs. Mean age 26.2; mean years education 10.2; 53% ever sex worker; mean gestational age 22 weeks; DSM‐III‐R: 85% opiate dependent, 15% cocaine, 59% marijuana, 32% alcohol, 98% nicotine.
Interventions	All participants received MMT (mean methadone dose 49 mg) and counselling about HIV risk. 6 sessions of manual based CBT lasting 60 to 90 minutes, the first being more of a MI (n = 47). No intervention (n = 45). Duration 9 months.
Outcomes	Retention was measured as a proportion. Attendance was measured as the average number of missed appointments.
Notes	Researchers attempted to contact patients lost to follow‐up through the Department of Social Security and Department of Health. We attempted to contact the trial authors but received no answer.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Study states that subjects were randomly allocated to the intervention or control group, but does not outline specific randomization methods.
Allocation concealment (selection bias)	Unclear risk	No specific references to allocation concealment methods were made.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	"Five subjects dropped out of the intervention condition; three before any sessions and four after two or more sessions. There were no differences in any of the variables (demographic, drug use or HIV‐risk taking behaviour) between those who remained in the study and those who dropped out. There were 40 subjects in each group at the post‐intervention assessment. Of these 80 subjects, 74 (92.5%) were contactable 9 months later.  One refused to participate further, giving a 91% follow‐up rate of those who complete pre‐ and post‐intervention assessments." Attrition information provided, but differences between the intervention and control group not detailed.  In addition, specific participant numbers for different outcomes not stated.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"Assessments occurred at pre‐intervention, post‐intervention and at 9‐month follow‐up.  Follow‐up assessments were conducted by an interviewer blind to the subject’s group membership." Specific references to the blinding of the interviewer conducting follow‐up assessments makes the risk of detection bias low.