The paper by Perkonigg and colleagues in this issue of Addiction is the latest piece of evidence suggesting that a person's relationship with cannabis is not inevitably transient [1]. In the mid-1990s cross-sectional surveys, particularly in disadvantaged communities, suggested that older individuals who had used cannabis heavily and regularly for decades were not particularly unusual [2]. Longitudinal studies from Australia and New Zealand then showed that, albeit generally among users who had not yet been followed beyond their 20s, heavy regular use (including that fulfilling criteria for dependence) had a population prevalence of between 5 and 10% [3–5]. Risk factors for heavier use appeared to be male sex and troubled early lives in which behavioural disorders, family adversity, delinquency, other drug use and drug-using peers often featured. Earlier publications from the Munich-based Early Developmental Stages of Psychopathology study (EDSP) also suggested that for many cannabis users use was less intermittent and more persistent than perhaps assumed previously [6,7]. These earlier papers were based on 4-year follow-up; the current paper extends this to 10 years. Given over-sampling of younger participants (who were 14–15 years old at recruitment), this still means that the phenomenon under investigation is predominantly the natural course of cannabis use among people now in their mid-20s, although some data relate to older participants. By the end of follow-up approximately half the sample had used cannabis—a fairly typical proportion. Of the original sample of approximately 3000, more than 500 were using cannabis at each assessment point. Among people who had used cannabis more than trivially at baseline (five times or more), more than half also reported use 4 years later and the majority of these also reported use at 10-year follow-up. Other than younger age and initial level of use, the strongest baseline predictors of persistent use at 10-year follow-up were male sex, peer drug use and alcohol dependence.

Some might question the extent to which persistent cannabis use between the ages of 14 and 24 (or 24 and 34) years represents a health problem—adolescents, after all, do many things that perplex and concern their parents and may attract official disapproval. These data suggest part of the answer: 8% of users at baseline, rising to 12% at 4-year follow-up and falling back to 8% at 10 years met criteria for cannabis abuse. That is, they had experienced adverse social, occupational, legal or interpersonal consequences as a result of their cannabis use. That these consequences are predominantly socially constructed does not make them less real or less distressing for the people enduring their impact. A further 4% of users at baseline rising to 6% at 10-year follow-up were cannabis-dependent, implying impairment of their ability to control their use even when they wanted to. Add these adverse outcomes to the health risks associated with regular unfiltered tobacco smoking for 10 years or more and the possible additional risks for physical and psychological health of cannabis itself [8] and it seems reasonable to suggest that Perkonigg and colleagues are describing a significant public health problem [1].

Unfortunately, their data do not suggest a clear solution. Certainly, preventing more than trivial cannabis use in the first place seems sensible, but how to achieve this aim is not obvious. Limiting adolescent alcohol dependence, association with drug-using or delinquent peers and exposure to life stresses are similarly laudable aims whose desirability is increased by the possibility that they may also decrease the risk of problematic cannabis use; however, again the means to reach these ends are not obvious. There are probably many causes of frequent and persistent cannabis use, and only some of them will be amenable to elucidation through risk factor epidemiology. It is also likely that trajectories to harmful or hazardous patterns of cannabis use are established early in the life-course and that covariance of some problem behaviours in adolescence and early adulthood with cannabis use may reflect independent associations with earlier fundamental causes [8,9]. Studies such as the EDSP, that recruit participants in mid-adolescence, can generally access only retrospective, hence potentially biased, measures of the factors that may be causally relevant. Further, more than a third of EDSP participants were already cannabis users at the time of recruitment, introducing issues related to reverse causation.

It is also important to acknowledge that some people use drugs such as cannabis because they think that they enhance the quality of monotonous, unrewarding and often unpleasant lives [10]. Prevention strategies that ignore these ‘risk factors’ may have limited success; this notwithstanding, prevention of cannabis use makes public health sense. In many places, such as the United Kingdom, the relevant policy discourse is dominated currently by arguments focusing on criminal justice, rather than public health, strategies. Any policy in this area, within whatever systems or strategies it is delivered, should be based on sound evidence of net public benefit [11]. Without this evidential basis, policy, however well-intentioned, can be part of the problem rather than part of the solution.