Original articlesThe Zelen Design May Be the Best Choice for a Heroin-Provision Experiment
Introduction
In agreement with the advice of the Health Council, the Dutch Parliament agreed upon a medical experiment regarding the effects of heroin provision for long-term heroin addicts 1, 2. The main hypothesis is that provision of heroin, in addition to available standard treatment, decreases the harm caused by drug dependence. The outcome measures, in hierarchical order, will be general health and psychosocial and criminal behavior of long-term heroin addicts. Previous studies suffered particularly from high drop-out and noncompliance rates in the control group 3, 4, 5 or did not use a control group [6]. Therefore, these studies have been insufficiently informative with regard to the effectiveness of heroin provision. We set out to design an alternative study that would address this pragmatic question and would be less susceptible to drop-out and noncompliance while retaining the usual experimental rigor. In The Netherlands, standard treatment for long-term drug addicts is delivered by health services and consists of methadone maintenance and social aid.
Section snippets
The traditional design versus the zelen design
Previous trials 3, 4, 5 were conducted as traditional randomized trials in which addicts’ consent was obtained before randomization. Addicts who were randomized to a no-heroin provision control group had a clear reason to be disappointed, complied badly, dropped out in large numbers, and, in this way, adversely affected the trials’ validity. Thus, the challenge is to design a study in which disappointment leading to disruptive behavior is avoided. In a design that is known as the Zelen design,
Dropout and noncompliance
Potentially, the most serious threat to the comparability of treatment groups is dropout leading to incomplete data collection and missing data on effect measurements. A study’s internal validity is reduced if the addicts who drop out are prognostically different from the addicts who complete the study and the drop-out rates differ between the treatment groups.
Noncompliance is another factor that may reduce a trial’s validity. Addicts may either reject their allocated treatment or may take
Validity and precision
In the preceding section, we argued that the chances of obtaining valid results in a heroin-provision trial are probably higher with the use of the Zelen design than with the traditional design. However, the use of the Zelen design precludes taking nonstandard measurements in the control group at baseline. In this section, we qualitatively discuss the implications this has for the trial’s precision. Note also that the points 1 and 2 in the preceding list are likely to reduce the trial’s
A heroin provision experiment in badly integrated addicts
The attractiveness of the intervention implies that many heroin addicts will volunteer for heroin provision. So, in the traditional design, the addicts in the control group will be frustrated after being refused the desired intervention. Theoretically, in the Zelen design there will be no frustration; however, in practice some addicts in the control group, being aware of the experiment, may demand participation in the provision group, which will then be denied. This may cause disappointment and
Ethical considerations
In clinical medicine, both the Zelen design and the traditional design have been the subject of ethical scrutiny and controversy [9]. In both designs, the physician–patient relationship can be adversely affected because of the physician’s preference for one of the treatments [13]. However, in the traditional design, this relationship may be under even greater pressure because patients know that treatment allocation is purely a matter of chance [14]. In addition, sometimes, complex information
Conclusion
We have focused on the use of the Zelen design in a heroin-provision experiment in badly integrated long-term addicts. However, we think that our design considerations have a wider applicability. In this final section, we describe these considerations more generally.
The Zelen design as an alternative for the traditional randomized design should be considered if: (a) the experimental intervention is highly attractive for all potential participants, and (b) the control group will receive standard
References (18)
- et al.
Patients’ preference and randomised trials
Lancet
(1996) Randomised consent trials
Lancet
(1992)- The Drugs Policy of the Netherlands: Continuity and Change. Rijswijk: Ministry of Foreign Affairs, Ministry of Health,...
- The prescription of heroin to heroin addicts: Sub-report from a committee of the Health Council of the Netherlands. The...
- et al.
Evaluation of heroin maintenance in controlled trial
Arch Gen Psychiatry
(1980) Diversifizierte Verschreibung von Betäubungsmitteln an HeroinabhängigeGrundlagen, Versuchsplan, Begleitforschung
Schweiz Rundsch Med Prax
(1994)- et al.
Versuche für eine ärztliche Verschreibung von BetäubungsmittelnZwischenbericht der Forschungsbeauftragten
(1995) Das Englische System in Widnes, Merseyside. Heroinvergabe
A new design for randomized clinical trials
N Engl J Med
(1979)
Cited by (16)
The adapted Zelen was a feasible design to trial exercise in myeloma survivors
2020, Journal of Clinical EpidemiologyCitation Excerpt :Five researchers of the MASCOT team attended the ethics meeting, including a myeloma consultant (KY), behavioral scientists (AF and RB), a senior physiotherapist (BP), and the trial statistician (AH). This design can evoke strong emotions and rejection from ethicists and researchers [20], and we believe that the attendance of our diverse team providing the rationale behind the design helped mitigate a lengthy review. Overall, 313 patients were invited by post to take part in the lifestyle cohort study.
Details about informed consent procedures of randomized controlled trials should be reported transparently
2019, Journal of Clinical EpidemiologyIndications and requirements for the use of prerandomization
2009, Journal of Clinical EpidemiologyCitation Excerpt :However, more and more investigators are becoming aware that the conventional design, in which informed consent is obtained in all participants before randomization, is not the panacea for all research questions, because in certain situations, information in the reference group about the experimental intervention could lead to serious protocol violations. Therefore, in the last two decades, several investigators applied prerandomization designs to counter this contamination problem, questioning the absolutism of the conventional design [1–4]. The term prerandomization implies that the randomization takes place before seeking informed consent.
Randomized consent designs in randomized controlled trials: Systematic literature search
2006, Contemporary Clinical TrialsHeroin maintenance treatment for chronic heroin-dependent individuals: A Cochrane systematic review of effectiveness
2006, Journal of Substance Abuse TreatmentMembers of research ethics committees accepted a modification of the randomized consent design
2005, Journal of Clinical EpidemiologyCitation Excerpt :In The Netherlands, we mailed the questionnaires to the 77 registered medical ethics committees (local research ethics committees), and to the central medical ethics committee (CCMO). The questionnaire described a fictitious study on the provision of heroin to long-term heroin addicts, which met three preconditions for the proper use of a randomized consent design [21], namely, (1) the experimental intervention cannot be blinded and is very attractive (in which situation full prior consent would evoke contamination, and differential noncompliance and dropout in the reference group); (2) no personal measurement(s) are necessary until informed consent is obtained; and finally, (3) the reference group receives standard treatment (“usual care”). A randomized consent design with informed consent for effect measurements in the reference group, indicated as randomized consent design (Fig. 1a) and the modified version, indicated as modified randomized consent design (Fig. 1b), were explained as feasible design options for this study.