To gain further insight into the development of response over time, the percentage of "stable" or "sustained" responders in the treatment groups was determined at each assessment-point, as before using LOCF for the missing assessments during the trial. Sustained responders were defined as patients who (1) became responder prior to the month 12 assessment, and (2) remained a responder during the course of the trial, once they had become responder for the first time. For the month 12 assessment, this definition implied that patients who became responder for the first time at month 12, were considered as non- (sustained) responders. Similarly, patients who were responder at some earlier assessment-point, but non-responder at a later assessment, were considered as non- (sustained) responders at all assessment-points. Hence, the analysis represents a very conservative approach of response. Figure 15 shows the results of the analysis.

After 12 months of treatment, the percentage sustained responders amounted to 39.5% in the co-prescribed heroin group, and to 18.4% in the methadone alone group. In a logistic regression model, with percentage sustained responders as the dependent variable, and treatment site as covariate, the observed difference of 21.1% between the treatment groups corresponded with an adjusted Odds-Ratio of 2.94 (95%-CI: 1.47-5.87; p=0.002).
At the month 2 assessment, the percentage sustained responders amounted to 13.2% in the co-prescribed heroin group, and to 3.1% in the methadone alone group. These early response percentages for the sustained responders are considerably lower than those found for the responders at this assessment-point (43.4% and 18.4%, respectively; see paragraph 7A.3.2).

Figure 15. Percentage sustained responders at subsequent assessments