Given the response rates in the course of the trial, displayed in Figure 12, and particularly considering the substantial response rate in the heroin group present at the month 2 assessment-point, subsequent exploratory analyses were conducted to investigate the relative contribution of the outcome domains to the percentage responders at subsequent assessments during the trial. First, the type(s) of outcome domains on which the patients responded in the course of the trial ("response profiles") were investigated. Second, the number of response domains were investigated.

Response profiles in the course of the trial
Analogous to the inclusion profiles described earlier in paragraph 6A.1.2, the three outcome domains of physical health, psychiatric status, and social functioning were combined into seven types of response, each representing a unique and mutually exclusive response profile. In Figure 13, the resulting bar charts are displayed for the heroin group and methadone group separately. Notably, the total values of each column for the heroin and methadone group in Figure 13 are identical to the response rates of the treatment groups at each corresponding assessment-point, displayed earlier in Figure 12.

As indicated by the upper bar chart in Figure 13, response in the heroin group - both at the start, during the course, and at the end of the trial - was based on contributions from all three outcome domains. For example, with response rates at month 12 of 35.5%, 27.6%, and 23.7% respectively, the social, physical health and psychiatric domain each contributed considerably to the occurrence of overall (56.6%) response after twelve months of treatment. In addition, as suggested by the 18.4% response due to the social domain alone at month 12, response in the heroin group could not be attributed to a dominating influence of decreased illegal activities by the participants.

This limited influence on response of the social domain alone is also illustrated when subjects, who were included in the trial on the basis of their poor functioning in the social domain alone (22.4% of the intention-to-treat population), are left out of the analysis (not shown in Figure 13). In this case, overall response at month 12 still amounted to 56.5% in group B (n=62), and increased from 31.6% to 39.7% in group A (n=73). Using the same logistic regression model as

Figure 13. Contribution of the outcome domains to response at subsequent assessments

described earlier, the difference in percentage responders remained significant (adjusted OR=2.18; 95%-CI: 1.06-4.45; p=0.0333).

In the methadone group (lower chart in Figure 13), the three outcome domains contributed to the occurrence of response as well. Here, however, response at month 12 (31.6%) often occurred on the social domain alone (9.2%), physical health domain alone (9.2%), and psychiatric domain alone (8.2%). Hence, response in the methadone group was often based on improvements in one single domain.

Single-domain versus multi-domain response
The difference in comprehensiveness of response between the methadone alone and the co-prescribed heroin group is further illustrated in Figure 14. As before, it should be noted that the total values of each column for the heroin and methadone group in Figure 14 are identical to the response rates of the treatment groups at each corresponding assessment-point, displayed in the earlier figures in this paragraph.

Figure 14. Comprehensiveness of response at subsequent assessments

In the heroin group, at the primary assessment-point at month 12 (left-hand chart in Figure 14), 30.3% of the participants had responded on one domain only, 22.4% on two domains, and 4.0% on three domains simultaneously. Expressed as a percentage of the total response rate at month 12, 46.5% of the response in the heroin group occurred on more than one domain. In addition, Figure 14 shows that treatment comprehensiveness in the heroin group increased with time-in-treatment.

In contrast, in the methadone group (right-hand chart in Figure 14), 26.5% of the subjects had responded on one domain only, 5.1% on two domains combined, and none of the subjects had responded on three domains simultaneously. Hence, given the total response rate of 31.6% at month 12 in this group, 83.9% of the response in the methadone group was limited to improvements on one single domain, and - conversely - only 16.1% of the total response occurred on more than one domain. The difference in percentage multi-domain response between the two treatment groups (30.4%) was tested in a logistic regression model, with treatment site as the only covariate, and with non-significant relationships between multi-domain response and both treatment site and treatment-by-site interaction. In the model, the difference in percentage multi-domain responders was significant (adjusted OR=4.80; 95%-CI: 1.48-15.59; p=0.009).