Headline writers are having a field day with the results of a meta-analysis to be published in The Lancet which found that cannabis users are 40% more likely than non-users to develop psychotic symptoms. Radio, print, and internet publications urge that more must be done to warn the public of the risks of cannabis use. But hold on, did the headline writers miss the lesson on correlation vs. causation in their Science for Journalists class?
It’s official: a meta-analysis to be published in The Lancet looking back over studies from the 70s, 80s, and 90s (you can listen to a Lancet podcast on the subject here) has found that cannabis users are around 40% more likely than non-users to suffer from a mental illness such as schizophrenia. The popular media this morning are full of commentary about how the public must be warned about these ‘risks’ from cannabis use, and how the evidence of this ‘link’ should inform policy-making.
For example, Marjorie Wallace, chief executive of the mental health charity SANE, is quoted on the BBC website as saying, “This analysis should act as a serious warning of the dangers of regular or heavy cannabis use.”
Even study author Professor Glyn Lewis seems pretty keen on the sensational interpretation and is quoted in bold print in that same story: “All the studies have found an association and it seems appropriate to warn members of the public about the possible risk.” Less prominence is given to what Lewis said just before that sentence, in the main text of the BBC article: “It is possible that the people who use cannabis might have other characteristics that themselves increase risk of psychotic illness”.
And therein lies the distinction between basic science and sensational journalism: correlation does not imply causation. In other words, the mere fact that cannabis users are more likely to develop psychotic symptoms does not, in and of itself, tell us anything whatsoever about the ‘risks’ of using cannabis. Nor does it tell us anything whatsoever about whether cannabis can actually cause a user to develop psychotic symptoms. What it tells us is that there is significant overlap between the set of people who use cannabis and the set of people who develop psychotic symptoms.
It might be the case that the sensational interpretation will ultimately prove to be correct: perhaps cannabis use really does cause psychotic symptoms directly; it’s just that the meta-analysis does not appear to provide any evidence to support that conclusion.
To be fair, the BBC article on the study does go on to highlight some of the basic science that other publications (and, apparently, the chief executive of SANE) are missing completely. Near the end of the article, we do find a brief nod in the direction of actual science:
But Professor Leslie Iverson, from the University of Oxford, said there was still no conclusive evidence that cannabis use causes psychotic illness.
“Their prediction that 14% of psychotic outcomes in young adults in the UK may be due to cannabis use is not supported by the fact that the incidence of schizophrenia has not shown any significant change in the past 30 years.”
So…let’s see…cannabis use has increased and decreased significantly over the last 30 years…yet the incidence of schizophrenia has not…and the headline writers would still have us believe that one is causing the other?
Does all this about distinguishing correlation from causation sound familiar? We’ve been here before. (For example, see our article from last summer, “Popular Media Get it Wrong About Music Lyrics With Sexual Content”, about evidence supposedly demonstrating that music with sexual lyrics prompts teens to have sex.) Ignoring basic science is a great way of crafting an attention-grabbing headline, if you’re a journalist, and in fact it’s a great way to get journalists’ attention if you’re a scientist.
All clinical material on this site is peer reviewed by one or more clinical psychologists or other qualified mental health professionals. This specific article was originally published by Dr Greg Mulhauser, Managing Editor on .on and was last reviewed or updated by