Results across studies need to be interpreted in the context of possible limitations to generalizability. All studies relied on self-report to assess substance use outcomes, which might be susceptible to bias. However, all studies carefully describe a procedure for assuring confidentiality to participants and no consequences for reporting substance use. Some studies controlled for cluster-level effects (e.g. school), but not all. Only two studies included an active control intervention as a comparison group [22, 78]. None of these studies were able to mask intervention condition to participants or therapists but most described a procedure in which assessment of outcomes was conducted by experimenters blind to treatment condition.
