As described in paragraph 2.9.3, the primary outcome measure in the study was based on self-report data in the areas of physical and mental health, social functioning, and substance use. In accordance with the study protocol, the self-report data pertaining to the patient's substance use and criminality were verified by analysis of urine samples and by investigation of registered data in the police records, respectively. Urine samples were collected both at the regular two-monthly assessment-visits, and on a randomly selected day during the interval between these two-monthly assessments, and were analyzed by a central laboratory. The registered data in the police records were investigated by an independent research-assistant.

Substance use
From the data collected in the areas of substance use, the use of cocaine and/or amphetamines was part of the primary outcome measure. Verification of self-reported substance use focussed, therefore, on these two substances. In the self-report questionnaire, related to the urinalysis, the subjects were asked about their cocaine and amphetamine use in the 48 hours prior to the assessment. Due to the low prevalence of amphetamine use, the analyses were restricted to the use of cocaine.
A first investigation pertained to the degree of underreporting at the time of the month 12 outcome assessment in treatment conditions A and B combined. To this end, the level of agreement was determined between the use of substances (i.e. cocaine) according to self-report and according to urinalysis, in the intention-to-treat population in conditions A and B combined, using Kappa as measure of concordance, and regarding the results of the urinalyses as "golden standard". Additionally, a possible difference in degree of underreporting between the month 12 and month 10 assessment was analyzed by means of the McNemar test. In this case, the analysis set consisted of all patients with a positive urine sample on month 12 and month 10, in conditions A and B combined.
The second type of investigation pertained to the occurrence of differential underreporting in conditions A and B. Differential underreporting was investigated for the month 12 and month 10 assessment separately, among patients with a positive urine sample, using a logistic regression model, with underreporting (yes/no) as dependent variable, treatment condition (A or B) as predictor, and treatment site as covariate.
The third type of investigation focussed on possible differences in level of underreporting between the regular month 12 assessment and the random - hence unpredictable - assessment between month 10 and 12 ("month 11 assessment"). As in the first investigation, differences in level of underreporting were analyzed by means of the McNemar test, among patients with a positive urine sample on month 12 and month 11, in conditions A and B combined.

At month 12, 76.8% of the urine samples were positive for cocaine, whereas 64.4% of the patients reported the use of cocaine in the previous 48 hours. At month 10, 76.5% of the subjects had positive urines for cocaine, whereas 62.6% reported cocaine use. Results for the random assessment at month 11 were similar: 78.7% positive urine samples; 69.4% self-reported cocaine use.
Agreement between the two assessment procedures was generally good. At the month 12 assessment, overall agreement between self-report and the results of urinalysis amounted to 85.8%, with a Kappa of 0.66 (95%-CI: 0.58-0.75). At month 10, overall agreement amounted to 84.6%, with a Kappa of 0.64 (95%-CI: 0.55-0.73). Similar results were obtained at the random assessment at month 11 (overall agreement: 88.5%; Kappa=0.70; 95%-CI: 0.59-0.82). Between the two trials, no differences in level of agreement were detected (injectable heroin trial: overall agreement at month 12: 84%; Kappa=0.64; 95%-CI: 0.51-0.78; inhalable heroin trial: overall agreement at month 12: 87%; Kappa=0.68; 95%-CI: 0.57-0.78).
Of all patients with a positive urine sample on cocaine at month 12, 82.6% also reported cocaine use, indicating 17.4% underreporting. A similar result was obtained at month 11: 13.2% underreporting. In the trial on injectable heroin, underreporting at month 12 amounted to 20.2%, and in the trial on inhalable heroin to 15.5%.
In the trial on injectable heroin, no significant differences between the treatment groups A and B in underreporting were observed at month 12: group A: 17.2% underreporting; group B: 23.9% underreporting. However, in the trial on inhalable heroin, underreporting in group B (24.3%) was significantly higher than in group A (8.8%). A similar difference in underreporting occurred at the month 10 assessment (group B: 25.7%; group A: 12.1%). In the logistic regression model with underreporting as the dependent variable, and treatment site as covariate, a group effect remained significant for both the month 12 (adjusted OR=3.31; 95%-CI: 1.32-8.26; p=0.01) and month 10 assessment (adjusted OR=2.57; 95%-CI: 1.07-6.17; p=0.03). In summary, the data indicate acceptable agreement between the two assessment procedures, with some differential underreporting in the inhalable heroin trial.

Criminality
Whereas the primary outcome data could be verified directly with regard to the self-reported use of substances, the self-reported data in the criminality domain of the primary outcome measure (i.e. number of days in the previous month of illegal activities) did not allow for such direct verification, because no external verification-source was available. Therefore, verification took place with regard to the variable whether the person had been charged by the police for possession and dealing of drugs, crimes against property, crimes of violence, and other crimes, a variable which was available in the CRF and could be retrieved from the police records. Analogous to the types of investigations described for self-reported substance use, the self-reported charges by the police were verified by analyzing (1) the degree of underreporting at the time of the month 12 outcome assessment in treatment conditions A and B combined, (2) the difference in degree of underreporting between the month 12 and month 10 assessment, and (3) the difference in degree of underreporting between conditions A and B at the month 12 assessment, and separately at the month 10 assessment.
Given that the data in the police records are routinely registered with some delay, the analyses were limited for the present report to the patients in conditions A and B in the cities of Amsterdam and Rotterdam (n=137), who entered the study during the first study stage (see paragraph 2.4). In the four months prior to the month 12 assessment, these patients reported 20 charges, whereas in the police records, 18 charges were registered. Overall agreement amounted to 89.6%, with a Kappa of 0.62 (95%-CI: 0.43-0.82). These preliminary data suggest acceptable agreement and no systematic underreporting by the patients. A more comprehensive analysis will be provided in future reports.