It is easy to agree with Temple, Brown & Hine [1] that a number of methodological problems in cannabis research limit our ability to draw firm conclusions about prevalence, prevention/policy and treatment. Clearly, compared to tobacco or alcohol research, drug research, including cannabis, is less developed and refined. One reason is the relative immaturity of this research field. Another concerns the obvious problems related to the illegal nature of the phenomena under study.

One important problem raised by the authors is the poor quality of exposure data. Few studies go further than giving frequency figures and often very crude estimates at that, e.g. ever use, last year use, last month use. One should be particularly wary of life-time use data. In a recent Swedish study [2], reported life-time use was found to decrease with increasing age. It was lowest in the oldest age group, 55–64 years, which is curious, considering that when this group was in its 20s, around 1970, cannabis prevalence was considerably higher than currently. The only credible explanation is that old drug use experiences are forgotten, repressed or denied.

The illicit nature of drug use leads to a reluctance in many users to bring attention to their drug habits, e.g. by seeking treatment or participating in research. This creates large hidden populations, distorting the picture. The authors rightly warn against generalizing findings from clinical samples to the whole population of cannabis users. We have considerable difficulties in determining how large is the proportion of users seeking treatment to the total population of users. This problem is not entirely unique to the drug field; similarly, a large majority of alcohol-dependent people also do not seek treatment.

In an attempt to get closer to the truth, various methods to estimate the prevalence of illicit drug use were triangulated in a Swedish project [2]. The main study was a large general population survey, which was combined with surveys of high-risk groups, a respondent-driven sampling study and register-based methods. Survey data, with a 52% response rate, indicated that 127 000 Swedes were regular drug users. Register data from the patient register and from the Prison and Probation Services estimated 29 500 as problematic drug users. This means that there is a large hidden population of people with regular drug use, but where this is not identified in the health or corrections systems.

One part of this hidden population consists of drug users who do not experience difficulties and therefore do not seek treatment. It is also clear, however, that we have another group of users who experience severe problems but nevertheless refuse to seek treatment. A number of reasons may explain this, but one with special significance in the addictions field is the stigma attached to drug use. This stigma also permeates most treatment facilities, including staff attitudes, labelling, etc., making treatment unappealing to many of those in need of it. Drug users who do seek treatment for various complications to drug use, e.g. depression and infections, tend not to volunteer information about their drug use and, in the majority of cases, are not asked about drug use. In most of these treatment episodes, drug use or diagnoses of harmful use or dependence are not recorded.

This gives rise to a number of methodological difficulties. Based on the available data it is, for instance, difficult to determine whether there is an epidemiological paradox within the drug field of the same nature as within the alcohol field. In the alcohol field it is clear that most problems and costs arise within the larger group of heavy drinkers that do not belong to the most severe group [3–5]. It seems reasonable to assume that the same principles would apply within the drug field, but the lack of data does not allow any firm conclusions. This has important policy implications: are we only seeing the tip of the iceberg in the drug field—or are we seeing most of the problem?

Declaration of interests