Protocol amendments

In the course of the study, several changes were made in the original study protocol (CCBH, 1997). These changes have been integrated in the second, revised edition of the protocol (CCBH, 1999a), and are described in detail in a separate addendum of this second protocol edition (CCBH, 1999c). All amendments were submitted to and approved by the Central Committee on Medical Ethics, prior to implementation of the proposed adaptations. In addition, all adaptations were made prior to the availability - and without knowledge - of outcome results. Most protocol amendments represented more explicit formulations, further operationalizations, or minor adaptations of certain specific study procedures. Two major changes were made in the original protocol - one concerning dropping a treatment condition in the injectable heroin trial, and one with regard to a change in the definition of the study population for the primary outcome analysis. The protocol amendments are summarized in this section.

Omission of treatment condition C in the injectable heroin trial
Against the background of declining prevalence rates of injecting heroin use among heroin addicts in the Netherlands, it became apparent during the recruitment and selection phase of the first study stage in the cities of Amsterdam and Rotterdam, that the number of eligible subjects for the injectable heroin trial would be insufficient to achieve at the - according to the power analysis -necessary total of 375 subjects. To cope with this problem, the CCBH decided to omit the treatment condition in the injectable heroin trial in which patients were treated with methadone alone during the first six months of the experimental study phase, and subsequently treated with co-prescribed heroin and methadone during the following six months (treatment condition C). This omission - which reduced the required number of participants in the injectable heroin trial from n=375 to n=250, and hence the total number of required participants in both trials from n=750 to n=625 - did not affect the primary analysis of effectiveness of 12 months treatment with oral methadone and co-prescribed heroin, compared with that of standard methadone maintenance treatment, since this analysis involved only the treatment conditions A and B. However, in the absence of condition C, the secondary study question into the treatment effects after six months, which involved a comparison between the treatment conditions B and C at the time of the month 12 assessment, could consequently be answered only for the trial on inhalable heroin (where treatment condition C was maintained).

Definition of the intention-to-treat population in the primary analysis
In the original study protocol, the intention-to-treat population in the primary analysis of 12 months effectiveness was defined as all patients who were randomized and who had received at least one post-baseline assessment. Since this definition would implicate exclusion from the primary outcome analysis of all patients with only a baseline assessment - an approach that would be inconsistent with the basic meaning of the intention-to-treat principle - the definition of the intention-to-treat population was changed to all patients who had been randomized in the study and had been notified about the result of their randomization. In addition, a final choice was made in the second version of the study protocol to deal with missing endpoint-assessments by means of the Last Observation Carried Forward approach.

Response definition
In the original protocol, response was defined in global terms, as percentage of change from the baseline assessment in medical condition, social functioning and substance use of the patients. In the second edition of the protocol, the response definition was further operationalized by referring to specific instruments, scales and items, and by including explicit inclusion thresholds and required degree of change on each of the primary outcome domains. As part of this operationalization, the original requirement for response of at least 20% change on any of the outcome domains was adapted to at least 40% change.

Randomization
According to the first edition of the protocol, patients would be randomized in blocks of 12 subjects, with gender and ethnicity as stratification variables. Given the total maximum of 80-90 participants in each study site for both trials combined, the use of two stratification variables, and the allocation of subjects to five treatment conditions, the randomization block length was adapted to three subjects in the inhalable heroin trial, and to two subjects in the injectable heroin trial.

Partners
During the execution of the first stage of the trials, it became apparent that some study subjects had partners who were also eligible for the study. In those cases where the partners were randomized to different treatment conditions (i.e. one partner receiving co-prescribed heroin and the other partner receiving methadone alone), this tended to cause considerable stress in their relationship. In these cases, and provided that both partners met the inclusion criteria of the trial, the CCBH decided to randomize both partners to the same treatment condition, and to designate one partner (with the lowest patient identifier number) as the index partner to be included in the primary outcome analysis. Hence, the other partner had the possibility to receive the same type of treatment, but was excluded from the primary analysis.