One hundred and forty-three studies were included. Of these, 28 studies (approximately 8,025 patients) were of a treatment-comparison design, and 115 (approximately 28,854 patients) were of a single-group design. Of the treatment-comparison studies, 12 studies were randomised or quasi-randomised, 2 studies were non-randomised and the patients were matched, and 14 studies were non-randomised and the patients were not matched.
Most studies received funding from federal sources (52%). The number of patients per study ranged from 8 to 3,440 for single-group studies, and from 16 to 2,544 for treatment-comparison studies. All but 2 studies were conducted in the United States.
There were 38 studies of MMT, 32 studies of TC, 29 studies of ODF programmes, 29 studies of DETOX, and 15 studies of other modalities. Seventeen studies were conducted in the 1960s, 42 in the 1970s, 30 in the 1980s, and 37 studies in the 1990s. Seventeen studies did not report the study decade.
Drug use.
Treatment-comparison studies (28 studies): the random-effects weighted mean effect size was 0.29 (95% CI: 0.14, 0.45); the homogeneity (Q) was 96.64 (p<0.0001).
Single group pre-test post-test studies (103 studies): the random-effects weighted mean effect size was 1.20 (95% CI: 1.15, 1.38); the homogeneity (Q) was 11495.43 (p<0.0001).
Most correlations with potentially explanatory variables were not statistically significant, even when the level of significance was set at a p-value of less than 0.10. Statistically-significant correlations were found among treatment-comparison studies for implementation of treatment (negative correlation, p<0.10) in MMT programmes. Statistically-significant correlations were found among single-group studies for the following: decade of treatment (negative correlation, p<0.10), average methadone dose (positive, p<0.05), implementation of treatment (negative, p<0.10), and mean number of weeks in treatment (positive, p<0.10) for MMT programmes; implementation of treatment (positive, p<0.50) for TC programmes; mean number of hours of contact time (positive, p<0.10) for ODF programmes; and average methadone dose (positive, p<0.05) for DETOX programmes.
Crime. Treatment-comparison studies (17 studies): the random-effects weighted mean effect size was 0.17 (95% CI: 0.00, 0.33); the homogeneity (Q) was 30.62 (p<0.01).
Single group pre-test post test studies (42 studies): the random-effects weighted mean effect size was 0.75 (95% CI: 0.61, 0.90); the homogeneity (Q) was 3,176.18 (p<0.0001).
Most correlations with potentially explanatory variables were not statistically significant, even when the level of significance was set at a p-value of less than 0.10. Statistically-significant correlations were found among treatment-comparison studies for: maturity of programme (negative correlation, p<0.10), mean number of weeks in treatment (positive, p<0.10) and mean number of hours of contact time (positive, p<0.10) for TC programmes; and maturity of programme (negative, p<0.10) for ODF programmes. Statistically-significant correlations were found among single-group studies for: role of researcher (positive correlation, p<0.10) for MMT programmes; decade of treatment (negative, p<0.10) counsellor-to-client ratio (positive, p<0.05), and intensity of treatment (positive, p<0.10) for TC programmes; and implementation of treatment (positive, p<0.05) for ODF programmes.
Publication bias (Rosenthal's fail-safe N)..
Treatment-comparison group: for substance use (28 studies), N was 344 studies; for criminal activity (17 studies), N was 30 studies. Single-group: for substance abuse (reported as 102 studies in Table 7, but 103 studies in Table 4), N was 1,030,601 studies; for criminal activity (42 studies), N was 74,755 studies.