July 28th, 2007 by Ben Goldacre in drurrrgs |

The Guardian
Saturday July 28 2007

You know when cannabis hits the news you’re in for a bit of fun, and this week’s story about cannabis causing psychosis was no exception. The paper was a systematic review and then a “meta-analysis” of the data which has already been collected, looking at whether people who smoke cannabis are subsequently more likely to have symptoms of “psychosis” or diagnoses of schizophrenia. Meta-analysis is, simply, where you gather together all of the numbers from all the studies you can find into one big spreadsheet, and do one big calculation on all of them at once, to get the most statistically powerful result possible.

Now I don’t like to carp, but it’s interesting that the Daily Mail got even these basics wrong, under their headline “Smoking just one cannabis joint raises danger of mental illness by 40%”. Firstly “the researchers, from four British universities, analysed the results of 35 studies into cannabis use from around the world. This suggested that trying cannabis only once was enough to raise the risk of schizophrenia by 41%.”

In fact they identified 175 studies which might have been relevant, but on reading them, it turned out that there were just 11 relevant papers, describing seven actual datasets. The Mail made this figure up to “35 studies” by including 24 separate papers which the authors also found on cannabis and depression, although the Mail didn’t mention depression at all.

They also said that “previous studies have shown a clear link between cannabis use in the teenage years and mental illness in later life”. They then described some of these previous studies. These were the very studies that are summarised in the new Lancet paper.

But what was left out is as interesting as what was added in. The authors were clear – as they always are – that there were problems with a black-and-white interpretation of their data, and that cause and effect could not be stated simply. For ongoing daily users, as an example, it’s difficult to be clear that cannabis is causing people to have a mental illness, because their symptoms may simply be due to being high on cannabis all the time. Perhaps they’d be fine if they were clean.

It was also interesting to see how the risk was numerically reported. The most dramatic figure is always the “relative risk increase”, or rather: “cannabis doubles the risk of psychosis”, “cannabis increases the risk by 40%”. Because schizophrenia is comparatively rare, translated this into real numbers this works out – if the figures in the paper are correct, and causality is accepted – that about 800 yearly cases of schizophrenia are attributable to cannabis. This is not belittling the risk, merely expressing it clearly.

But what’s really important, of course, is what you do with this data. Firstly, you can mispresent it, and scare people. Obviously it feels great to be so self-righteous, but people will stop taking you seriously. After all, you’re talking to a population of young people who have worked out that you routinely exaggerate the dangers of drugs, not least of all with the ridiculous “modern cannabis is 25 times stronger” fabrication so beloved by the media and politicians.

And craziest of all is the fantasy that reclassifying cannabis will stop six million people smoking it, and so eradicate those 800 extra cases of psychosis. If anything, for all drugs, increased prohibition may create market conditions where more concentrated and dangerous forms are more commercially viable. We’re talking about communities, and markets, with people in them, after all: not molecules and neuroreceptors.
(http://www.badscience.net/?p=476#more-476)