Data analysis

3.1 Calculation of sample size

As described in paragraph 2.9.3, treatment response was defined in the present study in terms of a dichotomous, multi-domain outcome index (responders versus non-responders), based on the difference in patient status after 12 months compared to the baseline measurement. To determine the sample size needed to demonstrate effect, a comparison of two groups, two-tailed testing, alpha=0.05, a power of (1-beta) 0.80, and - based on the most often required minimum difference in effectiveness in trials on depression and schizophrenia - a difference in percentage responders between the treatment conditions of at least 20% were used as starting points. On the basis of these parameters, and the procedures and formulas described in Armitage and Berry (1987, 1990) and Fleiss (1981), the number of subjects needed was estimated to be 108 subjects per treatment group (CCBH, 1997, 1999a).

Furthermore, based on a recent study in Amsterdam into the stability of attending methadone maintenance treatment of patients with similar characteristics as those projected for the target population of the current clinical study (Buster and Van Brussel, 1996), it was estimated that the drop-out rate during the experimental study-phase could amount to 20% in the control condition (group A), 6% in the experimental condition (group B), and 14% in group C (inhalable heroin trial only). Taking these anticipated drop-out rates into account, the number of subjects required at the start of the study amounted to n=135 in group A, n=115 in group B, and n=125 in group C.