Why do the study?

Heroin addiction is a serious problem in the Netherlands and all over the world with considerable consequences for the addicts and their communities, but we still do not know the best way to treat the addiction. The most common medical approach is to prescribe methadone to heroin users, but in the Netherlands up to a third of people addicted to heroin who take prescribed methadone also still regularly use illegal heroin. Recently, a study in Switzerland suggested that prescribing heroin may be a more effective way to treat the addiction.

The reasoning behind this approach is that when heroin users are able to obtain the drug from a controlled environment they are less likely to associate with other drug users and will have better access to the medical and social support services they need. Although the Swiss findings are promising, they are not completely reliable because the researchers did not use a control group in their study against which they could compare the effect of heroin treatment.

How is this study different?

The current study builds on the Swiss findings by testing the effect of heroin treatment in a randomised controlled trial. This means that two groups of people are used: the experimental group receives the intervention that the researchers want to find out about and the control group receives no new treatment but the researchers still monitor it. When the people in both groups are the same in various important ways the researchers can be more certain that any differences in outcome measures between the two groups are caused only by the intervention that they are interested in. In reality it is impossible to ensure that all the people in a study are the same, so to overcome this problem participants are randomly allocated to the control and experimental groups. When this happens any important differences between the people in the two groups will also be randomised and are unlikely to influence the outcome more in one group than the other.

The current study is a little more complicated because there were actually five study groups in all: two control groups and three experimental groups which varied in the route in which heroin was taken (injecting or inhaling) and also whether methadone was prescribed at the same time. Essentially, however, the design follows the same principles as a randomised controlled trial with two groups. Applying this logic to the current study we can say that the researchers' objective was to find out how heroin treatment compares to methadone treatment in people addicted to heroin who are not currently responding to treatment with methadone. The people in the control groups continued to take their usual methadone treatment (no intervention) while the people in the experimental group took methadone and heroin.

How did the researchers test how well the treatments worked?

Another important aspect to think about when reading a research paper is what the outcome measure is. These are measurements that the researchers take to find out if their intervention is having any effect. Deciding on an outcome measure may sound fairly straightforward and in some cases it is, but in other situations researchers must make deliberate choices about the criteria they will use to know whether and how well the treatment is working. In this study the researchers used a diagnostic interview and a measure of the severity of addiction. These measures are essentially interviews which require the people in the study to describe how they match up to a set of predefined criteria which measure various aspects of addiction. Since these measures have probably been developed and tested by many researchers over many years we can be fairly sure they test what they are supposed to, as long as the participants giving the answers can be relied upon to be truthful.

The next problem is deciding how much of a change in the outcome measure is a meaningful result. In this study, the researchers decided that a 40% improvement in one of three aspects of the participants' physical, mental, and social well being would indicate a successful treatment. But why 40%? The researchers do not state in their paper why they chose this number, which leads the discerning reader to conclude it is essentially arbitrary. This is more important than it may sound: if the cut-off point for improvement had been higher, the study could have been deemed a failure if none of the heroin treated participants improved sufficiently, when in reality the treatment could be beneficial. This kind of conclusion, a false negative, is known in research as a type II error. The converse problem, when outcome measures are too easy to pass and a treatment that seems to work well is in fact not that useful, is a type I error.

<
CHRIS YOUNG/PA

Doing research necessitates making many decisions-what to study, how to study it, and what all the numbers you get at the end really mean. Researchers must justify these decisions because they are fundamental to the findings of the study.

One of the advantages of randomised controlled trials is that they allow researchers to control for, and so effectively rule out, certain other factors which may have an unwanted effect on the outcome measure. Normally, randomised controlled trials are double blind, which means that neither the researcher nor the participants know which study group they are in until after the end of the study. If participants know in which group they are then they may behave less naturally or in ways that might be expected of them, according to their group (the Hawthorne effect). Similarly, researchers might find it difficult to remain objective if they know which participants are in which group.

In many studies, however, blinding researchers and participants is impossible. In this study, because methadone and injectable and inhaled heroin are taken in different ways the researchers were unable to disguise which treatment each participant received. This could be a particular problem for this study because the participants were also patients being treated for drug addiction in the same clinics used in the study and might therefore have had a motive for conforming to the expectations of the study group. To some extent prescription of heroin both during and on conclusion of the study might depend on how participants responded to the heroin maintenance treatment during the trial.¹

Because the study's main outcome measures depended on self reported data, knowing how much the "open label" study design might have influenced the findings of the study is difficult.

The study has also been criticised on grounds of validity because the dose of methadone used in the control groups may not have been adequate for the heroin users to cope with their addiction. In contrast, the people in the heroin treatment groups were allowed to choose, up to a limit, how much heroin they used each day.² This means that the experimental and control groups may not be directly comparable and any differences in the effectiveness of either treatment could be as much due to the different doses as the treatment itself. The problem is that there is no way of telling how much different doses might have influenced the findings compared with different treatments.

What can we learn from this study?

The results of the study show that addicts who were treated with heroin plus methadone were a fifth more likely to show an improvement on the study's outcome measures than the group receiving treatment with only methadone. The authors interpret these findings as showing that coprescribing heroin with methadone is "more effective and probably just as safe as methadone alone." Is this interpretation accurate? Given the differences in dose between the treatment groups, the authors' interpretation should be qualified: what the study shows is that coprescription of heroin with methadone is more effective than methadone at the doses used in this study. But to generalise these findings to other doses and even equivalent doses of heroin and methadone treatment would be to go beyond the current evidence. It was this latter point though that the researchers intended to make.

So what does the study tell us?

The findings do at least suggest that treatment with heroin can be effective in improving the quality of life of heroin addicts. If you look closely at the findings though, the groups treated with only methadone also showed some improvement (as much as 12% in the injecting group). Given the issues surrounding differences in doses between the treatment groups, additional research is needed to show convincingly that treatment with heroin is better.

1. Dehue T. An examination rather than an experiment [rapid response to van den Brink W et al. Medical prescription of heroin to treatment resistant heroin addicts: two randomised controlled trials]. BMJ 2004. http://bmj.com/cgi/eletters/327/7410/310#35634.
2. Reed LJ. Not a level playing field [rapid response to van den Brink W et al. Medical prescription of heroin to treatment resistant heroin addicts: two randomised controlled trials]. BMJ 2004.http://bmj.com/cgi/eletters/327/7410/310#36714.