Hansen 1991.

Methods	Country: USA Site: 12 Junior high schools in LA and Orange County, CA. Adolescent Alcohol Prevention trial (AAPT). Focus: Preventing onset of alcohol abuse, marijuana and tobacco use; primary outcome was alcohol use. Design: Cluster RCT (excluded from analysis).
Participants	Baseline: (1987) 3011 Age: 7th graders Gender: 48 ‐ 55% F. Ethnicity (range by intervention group): Asian 9% ‐ 26% (significant differences); B 1 ‐ 3%; H 11 ‐ 43% (significant differences); W 33 ‐ 52%. Baseline smoking data: smoking public schools = 4%, private schools = 4%.
Interventions	Category: information vs social influences vs information/perceptions vs social influences. Programme deliverer: project staff (received 2 wks intensive training) Intervention: Information (32 classrooms): 4 x 45‐min lessons about the social and health consequences of alcohol, tobacco and drugs. Resistance training [RT] (33 classrooms): 4 lessons on consequences of using substances, 5 on resisting peer and media pressures to use alcohol, tobacco and other drugs (ATOD) Normative Education [NE] (27 classrooms): 4 information lessons, 5 lessons on perceptions on prevalence and acceptability of using ATOD Combined programme of NE and RT (26 classrooms): 3 information, 3.5 resistance skills, 3.5 conservative norms Control: No control (author considers intervention 1 a placebo comparison)
Outcomes	Smoking index, and never/ever smoking/ 30 day smoking. Follow up: 8th grade, 1 yr from baseline.
Notes	Part of Adolescent Alcohol Prevention Trial (AAPT); Rohrbach 1993 discusses techniques of implementing the AAPT in Los Angeles, but without any data on student smoking. Quality of intervention delivery: Process analysis showed high fidelity in the delivery (average 6 on a 7‐point scale for 8 aspects of programme implementation were achieved) of the interventions; but 3 of the independent variables (skill, resistance knowledge and acceptability) were judged by programme specialists to have been affected by programme integrity. Statistical quality: Was a power computation performed? No. Was an intention‐to‐treat analysis performed? Not stated. Was a correction for clustering made? Yes. Were appropriate statistical methods used? The unit of allocation was the school, and the unit of analysis in the 1991 paper was class. General linear model analysis of covariance approach was used with classroom means for each composite index and for each dichotomous item. In the 1998 re‐analysis, a combination of multilevel analysis (ML3 programme) and ordinary least‐squares analysis for the post‐test at 2 yrs were used for: (i) the 2370 individuals, (ii) the 120 classes, and (iii) the 12 schools.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No statement on how schools were selected. "Schools were stratified by size, test scores and ethnic composition and then randomly assigned to receive one of four intervention programs". Method of randomisation not stated. Clusters: Schools Cluster constraint: Stratification by size, test scores and ethnic composition. Baseline comparability: Not stated.
Allocation concealment (selection bias)	Unclear risk	No statement
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No statement
Incomplete outcome data (attrition bias) All outcomes	Low risk	Pre‐test (1987) = 3011; Follow‐up at 1 year: 20% attrition with differential attrition in the resistance training group (P < 0.01), but the authors comment: "Since main effects of Resistance Training did not even approach significance, the interpretation of findings is not threatened".
Selective reporting (reporting bias)	Low risk	No selective reporting