Comparisons between treatment groups on baseline characteristics were performed by means of either 2-sample t tests or Wilcoxon rank sum tests for continuous measures and χ2 tests for categorical variables. The t tests were also used to compare treatment groups on mean number of documented cocaine-negative urine toxicology test results, mean percentage of positive breath alcohol tests, and Treatment Services Review composite scores for services received. Treatment Services Review composite scores represent the sum of services in each area per month during the period an individual was in treatment (ie, not influenced by retention). A log-rank test was used to compare treatment groups on retention through the 24-week treatment period. The χ2 tests were used to compare treatment groups on percentages of subjects who completed 12 and 24 weeks, percentage abstinent from cocaine and other drugs on the basis of urinalysis assessments, and percentage of subjects who self-reported abstinence from cocaine and alcohol for the 30-day periods before each assessment. Repeated-measures analyses of covariance were used for treatment comparisons corresponding to ASI composite scores, individual ASI items, and BDI scores. The statistical model used for the analyses of covariance consisted of fixed factors representing treatment group; time, which corresponded to the 3 periodic assessments during the scheduled treatment period (6, 12, and 24 weeks), 2 assessments during the initial 6 months of follow-up when aftercare was still available (9 and 12 months), and 4 assessments during the year after the end of the recommended course of clinical care (15, 18, 21, and 24 months); and their interaction. Subjects' intake values for each outcome measure were used as covariates to adjust for potential treatment group differences at intake and to reduce error variance. To differentiate between treatment differences during the scheduled treatment period and during the 2 follow-up periods, a priori contrasts were constructed to test the significance of the main effect of treatment separately for these 3 assessment periods. Additional analyses of covariance were used to perform treatment comparisons within strata defined by whether subjects were appropriate for and willing to use disulfiram therapy. Analyses were performed with SAS, PROC MIXED (SAS Institute Inc, Cary, NC), which allows for incomplete subject data. Treatment means presented for periodic assessments are least-squares means ± SEMs adjusted for the covariate (ie, intake values).