Cannabis use in teens, suicide and school dropout: the jury is still out

A recent study found associations between cannabis use and later poor educational achievement and suicide attempts. But was the cannabis really to blame?

The potential harms of adolescent cannabis use is a hotly discussed topic, particularly in recent months as various locations across the globe are decriminalising cannabis use or even legalising it. A recent study published in Lancet Psychiatry found evidence of associations between adolescent cannabis use and a variety of negative outcomes in early adulthood.

The study combined data from three cohort studies based in Australia and New Zealand to assess cannabis use by age 17 and its possible association with low educational achievement, later substance use and dependence, welfare dependence and poor mental health outcomes, including depression and suicide attempts. Across the cohorts, cannabis use was associated with all of these outcomes, apart from depression and welfare dependence.

Perhaps unsurprisingly, the strongest association was between daily cannabis use by 17 and later cannabis dependence. Daily use was also associated with a 60% decrease in the likelihood of finishing high school or getting a degree, compared with those who had not ever used cannabis.

The study has been picked up by media outlets as evidence that adolescent cannabis use might cause these negative outcomes, and therefore that legalisation or decriminalisation of cannabis could be a bad idea as they might make adolescent cannabis use more common. However, causation is quite a leap from these findings.

There are a number of strengths of this study. First, that it has combined findings from three studies means the sample size is larger and therefore the results are more meaningful than any of the studies would be by themselves. Also, it has combined them at the level of individual data, rather than the more usual meta-analysis method of running the analysis in each study individually, and then pooling the results. Combining individual data allows for the authors to more carefully control for other factors that could be affecting the associations.

However, the sample size is still not large where outcomes are very rare. In the case of suicide attempts, despite a sample size of 2,537 participants, only 78 people in the analysis attempted suicide, and of them, only two were daily cannabis users. This means the strength of evidence is low (as can be seen by the uncertainty around the estimated risk, which ranges from a doubling of risk for daily users to a 22-fold increase), and any bias in the data will have a larger effect when the numbers are very small.

The authors of this study note that they control for over 50 other variables which might affect the relationship between cannabis and the outcomes of interest, including factors from early in the child's life, the behaviours of the child's parents, and socio-economic factors. This is because you can never be fully sure that the differences seen between cannabis users and non-users are not due to some other difference between those who choose to smoke cannabis or not. Even though these authors take into account a lot of these potential differences, not all of them were measured in all three studies, so some confounding factors may remain.

Not only that, but if a meta-analysis is going to be undertaken, it is more usual to do a systematic review so that all the available studies are included. Selecting only some studies could bias the results. That said, there are some advantages to their selective meta-analysis. First, all the studies have relatively similar measures so can be merged. Second, they are all from similar geographical locations, and were conducted at roughly the same time, so the statistical noise that results from combining studies would be relatively low.

Perhaps more importantly, the associations seen could be operating in either causal direction. Although cannabis use was measured before the outcomes, it is still possible that people could have already been performing poorly in school, or have had suicidal tendencies, before cannabis use was measured. And which might have then led them to take up cannabis smoking. If people with pre-existing symptoms had been excluded from the analysis, this would provide stronger evidence that cannabis was the cause rather than the effect (if indeed it is either: associations might be due to bias, confounding or chance).

Finally, the studies did not ask the participants about the type of cannabis that they smoked. As more research is occurring into the various compounds that make up cannabis, there is more evidence for differing effects of them. THC is the main active compound in cannabis, but it seems to interact with cannabidiol, the levels of which vary dramatically across different strains and preparations of cannabis.

Daily cannabis use before age 17 is not very common, and the teenagers who smoke every day are very likely to be different from those who either use less frequently or not at all, in a variety of other ways. Heavy cannabis use as a teenager being associated with later heavy cannabis use is a relatively unsurprising finding.

The associations between cannabis use and poor educational outcomes are interesting, but it's hard to be sure whether these heavy users were already struggling at school (an issue seen in another much-covered paper suggesting cannabis use was associated with lower IQ), which led to their cannabis use. So while the author's (and the media's) interpretation of these findings might be correct, and heavy cannabis use could lead to negative outcomes in later years, the evidence from this study is by no means conclusive that cannabis causes a reduction in educational ability and mental health.

http://www.theguardian.com/science/sifting-the-evidence/2014/sep/18/cannabis-teens-suicide-school-dropout-education 17/09/2014