3.1 Baseline demographics and drug use
The two groups were very similar on most baseline characteristics (). There was a significant difference only for composite legal score between the varenicline and placebo groups (F=4.44, P=0.04).
3.2 Missing Data
There were a relatively small number of missed visits resulting in missing data (see , top panel). Each group provided 77% of their scheduled treatment urines. Time to last visit was not significantly different between the groups (F=2.77, p=0.10).
Attendance rates for the varenicline and placebo groups. Both groups attended approximately 77% of visits, with no significant difference in time to last visit (F=2.77, p = 0.10).
3.3 Cocaine Use (UDS)
For the cocaine use results, treating missing urines as cocaine-positive, and controlling for baseline use, the placebo group was about twice as likely to use cocaine as the varenicline group (OR = 2.02, 95% CI = (0.91, 4.48), p = 0.08) (, second panel). If missing urines are treated as ignored, the direction of the effect is the same, although the estimated effect is slightly smaller (OR = 1.86, 95% CI = (0.74, 4.70), p = 0.19).
3.4 Secondary Outcomes
Changes from baseline to mid study, baseline to end of the study, and mid study to end of the study were examined for the Addiction Severity Index (ASI). Varenicline-treated participants were more likely to have less days of alcohol use (F=5.98, p=.015) from baseline to mid study, but there were no time by medication group interaction effects. Varenicline-treated participants were more likely to have lower ASI composite medical scores compared with placebo-treated participants from mid study to the end of the study (F=3.86, p=.05), but there were no time by medication group interaction effects. Most measures showed significant improvement across the three timepoints, but no difference between groups. These included days of cocaine use, dollars spent on drugs, as well as composite drug, medical and psychiatric scores (all ps ≤ .05).
As measured by the Multiple Choice Procedure (MCP), varenicline treated participants had significantly decreased rates of cocaine reward compared to placebo-treated participants (p = .02) (, third panel).
The Brief Substance Craving Scale (BSCS) composite measure combines three cocaine craving domains: intensity, duration and frequency. Cocaine craving showed a significant decline over the trial in both groups (F=20.34, P<.001).
The Cocaine Selective Severity Assessment (CSSA) composite, a measure of cocaine withdrawal symptoms, showed a significant decline over the trial in both groups (F=8.42, P=.004) (, bottom panel).
The Hamilton Depression Scale (HAM-D) examines depression intensity. Depression Scores showed a significant decline over the trial in both groups (F=8.62, P=.003). Varenicline-treated participants were more likely to have lower HAM-D scores compared with placebo group, but there was no significant difference between these two groups (F=0.01, P=0.90).
The severity and improvement scores for the Clinical Global Impression Scale (CGI) were treated as either continuous, dichotomized, or categorical (3 categories). If treated as continuous variables, CGI severity scores (F=46.07, P<.001) and CGI improvement scores (F=37.11, P<.001) changed significantly in both groups over time, with CGI severity decreasing and CGI improvement increasing, but there was no significant time by treatment interaction for CGI severity scores and CGI improvement scores. If CGI severity and improvement were dichotomized into low and high categories, there was a significant time main effect for CGI severity (F=6.90, P=.009) associated with a decreasing trend over time with no significant time by treatment group interaction. CGI improvement showed a significant increasing trend (F=25.19, P<.001), but no significant time by treatment group interaction. If there were three categories (low, mild and high scores) for CGI severity and improvement, there was a decreasing trend for CGI severity (F=12.00, P<.001). For CGI improvement, there was a significant increasing trend for CGI improvement (F=12.63, P<.001); there was no significant interaction shown for time by treatment interaction.