Sloboda 2009.

Methods	Country: US Site: 83 school clusters from 6 metropolitan areas including Detroit, Houston, Los Angeles, Newark, New Orleans and St Louis ( 41 intervention, 42 control) ‘Adolescent Substance Abuse Prevention Study’ (ASAPS) implementing ‘Take Charge of Your Life’ program (TCYL) Focus: prevention and reduction of tobacco, alcohol, marijuana / substance use at grade 11 Design: cluster RCT collecting longitudinal data over 5 years (Group 3: point prevalence)
Participants	Baseline: 19529 consented, 17320 (10028 treatment, 7292 control) completed baseline survey Age: intervention: 12.4 years (mean); control: 12.5 (mean) Gender: intervention: 44.5% M; control: 43.7% M Ethnicity: intervention: White 32.8%, Black 12.6%, Latino/Hispanic 27.8%, Asian 4.2%, American Indian 8.3%, Other 11.8%; Control: White 39.4%, Black 15.4%, Latino/Hispanic 17.9%, Asian 4.7%, American Indian 8.3%, Other 12.9% Baseline smoking data: 30 day: 672/10,028 (6.7%)
Interventions	Category: social influences & social competence vs. control Programme deliverer: trained D.A.R.E. police officers (six 3 days training for each of 7th and three 3‐day for the 9th grade lessons, role plays) Intervention: Take Charge of Your Life (TCYL) ‐ personal, social, legal risks and consequences of tobacco, alcohol, illicit drugs, normative beliefs, communication, decision‐making, assertiveness, refusal skills, interactive tasks for students to "make sense of their experiences...", role playing; 10 lessons in 7th grade (9 on tobacco), 7 in 9th grade (3 on tobacco). Control: no statement
Outcomes	Past 30 days, past 12 months: scored 0 to 6 (2 packs/day), "As students' responses across surveys were heavily skewed, with most subjects reporting no use, substance use variables were converted into dichotomous variables coded 0 for no use and 1 for any use." Follow‐up: tests at baseline, annually until the 11th grade and 2 years post intervention; total 5 years from baseline.
Notes	Quality of intervention delivery: "Fidelity of implementation of TCYL was examined using independent observations, student assessments, and officer‐instructor assessments... the officers taught every lesson and implemented the curricula as designed with an average content coverage (i.e. activities within each lesson) score of 74% and used the appropriate instructional strategy on average, 55.5% of the time." Statistical quality: Was a power computation performed? Yes. To detect 8% difference in marijuana use in 11th grade, with 300 students/cluster/3 loss to follow up, ICC 0.05, required 40 school clusters for power = 0.80 and P = 0.05. "Although the actual loss to follow‐up through 11th grade was higher than the initial power calculations (45.7% compared to 33 1/3%), intra‐cluster correlations were much lower than expected,… this study is adequately powered." Was an intention‐to‐treat analysis performed? No, not explicitly stated, note that analysis has been done on imputed data. Both original and imputed data shown and the result for 30 day smoking is significant in both. Was a correction for clustering made? Yes Were appropriate statistical methods used? Yes. Multilevel logistic model.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"School clusters... were randomly assigned to either the treatment or control conditions." "To achieve diversity of study participants, the study consisted of school clusters within school districts in and around Detroit, Houston, Los Angeles, Newark (NJ), New Orleans, and St. Louis. A stress index representing poverty based on the percentage of students eligible for free lunch programs and the percentage of minority students attending schools  within the districts was calculated... Two strata of high and low stress districts  were created… Cities were randomly assigned to represent either low or high stress and then one inner city school district in the appropriate stress condition was randomly selected… school clusters within a 50‐mile radius of the inner city school cluster were randomly selected and recruited." Email from author 12 Jan, 2012 confirmed randomisation by computer generated selection. Clusters: "School clusters consisting of a high school and its feeder middle schools..." Cluster constraint: stratification Baseline comparability: equivalent on demographics and substance use "... the only significant difference noted was for region, with Detroit having a greater number of control students."
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	High risk	10434 completed 11th grade survey (of baseline 17320) Intervention: 5756/10028 (57%) Control: 4678/7292 (64%) 80/83 school districts retained in 11th grade (2 schools destroyed in Hurricane Katrina); "... because of the No Child Left behind Act of 2001, many inner city students transferred from the study high school to schools outside the study." "The treatment and control samples at the 11th grade were more likely to be white with a loss of Blacks and Latinos. Also there were fewer alcohol users in the control group by the 11th grade and fewer treatment students in Los Angeles and New Orleans. Because of the nested nature of our data, in order to describe attrition at the time of the 11th grade survey, we utilized a design adjusted logistic regression… attritors were more likely to be older at baseline, female, non‐white, users of alcohol, marijuana and tobacco, and from Los Angeles and Detroit."; (no significance levels stated). "To address the problem of missing data, we used a multiple imputation approach under a model that assumes values are missing at random... Multiple imputations were carried out using the NORM program with separate imputations for the treatment and control conditions…" "We employed multiple imputation techniques to estimate missing data… readers should interpret the results presented in this paper with some caution."
Selective reporting (reporting bias)	Low risk	No selective reporting