Anticipation effects
Since the participants in the study were well aware of the fact, that the treating physician had the possibility to reinstate the experimental treatment with heroin on individual, medical indication for responders who deteriorated considerably after the termination of the experimental treatment at month 12, it could not be excluded that the data of the month 12 outcome assessment would be biased to some extent by anticipation effects, in particular because the outcome measure in the study consisted of self-report data, which could only partly be verified with objective data. To gain insight into such possible bias due to anticipation effects, the findings were investigated in relation to the course of the changes in the patient during the last months of the experimental study-phase. A sudden and considerable increase in self-reported improvements in the months preceding the outcome assessment would be indicative of anticipation bias. Therefore, the effectiveness of the treatments, if significantly different between the experimental and the control condition (incorporating LOCF) at the month 12 assessment, was similarly investigated on the basis of the data collected at the month 10 assessment. If the observed difference in effectiveness between the conditions at month 12 would also occur at month 10, it would be sufficiently demonstrated that anticipation effects could not be held responsible for the observed effect after twelve months of treatment.

Treatment completers analysis
An additional treatment completers analysis was conducted into the efficacy (as opposed to effectiveness) of the treatments among those patients who had completed the planned treatment with methadone alone and/or co-prescribed heroin. To this end, "treatment completers " were defined as patients who still participated in treatment (i.e. received methadone in the control condition, and co-prescribed heroin in the experimental condition) in month 12. The percentage responders and the difference in response between the treatment conditions among the treatment completers were subsequently similarly investigated as in the intention-to-treat analyses.
In the Statistical Analysis Plan (CCBH, 1999b), an additional per-protocol analysis of treatment compliant patients was described, but since the definitions for treatment compliance were based on different criteria for both treatment groups, this type of per-protocol analysis was omitted.