A meta-analysis was conducted on 78 studies of drug treatment conducted between 1965 and 1996. Each study compared outcomes among clients who received drug treatment with outcomes among clients who received either minimal treatment or no treatment. Five methodological variables were significant predictors of effect size. Larger effect sizes were associated with studies with the following characteristics: smaller numbers of dependent variables, significant differences between groups at admission, low levels of attrition in the treatment group, a passive comparison group (no treatment, minimal treatment) as opposed to an active comparison group (standard treatment), and drug use determined by a drug test. Controlling for these methodological variables, further analyses indicated that drug abuse treatment has both a statistically significant and a clinically meaningful effect in reducing drug use and crime, and that these effects are unlikely to be due to publication bias. For substance abuse outcomes, larger effect sizes tended to be found in studies in which treatment implementation was rated high, the degree of theoretical development of the treatment was rated low, or researcher allegiance to the treatment was rated as favorable. For crime outcomes, only the average age of study participants was a significant predictor of effect size, with treatment reducing crime to a greater degree among studies with samples consisting of younger adults as opposed to older adults. Treatment modality and other variables were not related to effect sizes for either drug use or crime outcomes