Patients' characteristics and health outcomes
Experimental (n = 193) and control (n = 237) patients were similar at baseline. The mean age was 39.3 (SD 5.7) years. Most patients were men (80%, 345), of Dutch nationality (92%, 397), of low education (74%, 316), living independently or with relatives and friends (84%, 359), and unemployed or disabled (82%, 351). Major sources of income were social welfare (48%, 203), disability benefit or pension (12%, 50), and illegal activities or prostitution (28%, 121). Sixty per cent (256) inhaled heroin. Sixty one per cent (261) had previously attempted at least one treatment based on abstinence. On average, they spent €900 (£610; $1133) a month on drugs. About once every three days, patients engaged in illegal activities to acquire money or drugs.
After randomisation, 13 participants in the experimental group and one in the control group rejected treatment. During the study period 45 and 32, respectively, discontinued treatment for various reasons. Thus 135 (70%) in the experimental group and 204 (86%) in the control group completed the full year of treatment. One hundred (51.8%) participants in the experimental group responded to treatment compared with 68 (28.7%) in the control group. Two deaths occurred, one in each group, but both were unrelated to treatment. Over one year, mean QALYs per patient () were significantly higher for experimental than control patients with a mean difference of 0.058 (95% confidence interval 0.016 to 0.099).
Responses to EQ-5D during treatment after imputation of missing data and the related QALYs, by treatment group. Figures are means; medians (interquartile ranges)
Use of resources and illegal activity
Methadone intake was similar in both groups, at about 19 g a year. Experimental patients received 143 g heroin on average. The use of healthcare resources outside the programme was low with 8.6 and 7.4 consultations on average a year for experimental and control patients, respectively, and with each patient spending less than half a week as an inpatient ().
Mean volume of healthcare resources used by treatment group. Figures are means per 100 patients (SD) unless stated otherwise
shows that participants in the experimental group engaged in criminal activities less often than those in the control group. They reported fewer days with crime against property (10.3 v 37.5), were arrested less often (2.1 v 2.8 times a year), were less frequently convicted (0.25 v 0.54 times), and stayed in prison almost one day less (11.7 v 12.5 days).
Mean number of days with crime against property, arrests, convictions, days of imprisonment, and contacts with the probation officer, by treatment group. Figures are means per 100 participants (SD) unless stated otherwise
After we excluded patients with missing data or with more than 200 illegal activities a month (n = 47 remaining), the mean number of crimes against property per day of illegal activity with possible damage to third parties amounted to 3.4, with 2.5 (74%) crimes against companies and 0.9 (26%) against civilians.
Costs and cost utility
Tables and show the mean and median costs per patient. Mean cost differences between the groups resulted from the maintenance programme, law enforcement, victim damage, and travel. The mean total net savings amounted to €12 793. With these net savings and the higher mean QALYs (), the experimental treatment was superior to the control treatment ().
Comparison of per patient costs (€) between treatment groups. Figures are means; medians (interquartile ranges)
Table 6 Differential costs per patient (€) in treatment groups (methadone plus heroin v methadone alone). Figures are differential mean costs (95% confidence limit or interval based on bias corrected and accelerated non-parametric bootstrapping; one or (more ...)
Figure 1 Cost effectiveness plane, showing differences between experimental treatment (methadone plus heroin) and control treatment (methadone alone) after 25 000 bootstrap replications. Vertical axis shows differences in mean total costs, horizontal axis shows (more ...)
Sensitivity and subgroup analyses
The dominance of the experimental treatment persisted (95% upper limit - €12 911) when we used Dutch time trade-off based health utility values rather than UK values, with a difference in QALYs between the groups of 0.053 instead of 0.058.
The programme costs of experimental treatment included the costs of rebuilding existing treatment centres. Exclusion of these initial implementation costs reduced the unit costs per heroin dispensation from €30.32 to €28.64 and strengthened the superiority of experimental over control treatment (95% upper limit - €30 953).
Having participated in any abstinence-oriented treatment in the past did not substantially influence the effectiveness of the experimental treatment. For willingness-to-pay values between €5000 and €50 000, the cost acceptability of experimental against control treatment ranged from 89% to 96% for patients with at least one previous abstinence attempt and from 96% to 97% for patients without such attempt.
Completion of treatment strongly influenced the programme's efficiency. The cost acceptability of experimental against control treatment for willingness-to-pay values up to €50 000 did not exceed 32% in case of non-completers. For treatment completers, however, the cost acceptability was higher than 99% for willingness-to-pay values of 3