Reefer Madness - the evidence

Peter Hitchens asks: I want to know precisely which claims he [i.e. I, T.W.] views as ‘overblown’, and whether he accepts that cannabis has - or might in future be found to have - any dangers for those who use it. And if so, what he believes those dangers are..

I certainly do accept that cannabis carries some risks and is implicated in some harmful outcomes.

Cannabis - harms and dangers

First, I've already mentioned that smoking cannabis carries a risk of cancer. It hardly need be added that I would assume it imposes the other major risks associated with smoking. [UPDATE 17 Jul 2011 18:50: But see following post.] In any case, in Europe cannabis is most commonly smoked mixed with tobacco (and unfiltered, though that may be something of a red herring). So far as the present debate goes, Hitchens though has already said As for the risk of cancer, while it undoubtedly exists, it is not my principal concern..

Second, there are the effects of intoxication, some of which may be unwelcome. While stoned, for example, users tend not to be very attentive to some things, and may fail to lay them down in their memory - this is the fabled 'short-term memory loss'.

A stoned person may become rather self-absorbed and 'self-conscious' - that is, conscious of themself as the object of others' attention. This can mean not only being oversensitive to actual attention (yes, those 12 year old girls at the back of the bus might well be giggling about you, that shopkeeper might well be looking at you funny - but normally you wouldn't even notice, let alone care), but also 'paranoid' or oversuspicious of others.

In cases of extreme intoxication - for example after ingesting relatively large quantities of hashish (cannabis resin) - it is even possible to experience hallucinations.

Stoned folk often become lazy and may spend an inordinate amount of time arguing about who is going to go to the 24-hour garage to get Twixes and chocolate-flavoured long-life milk. Some people find the whole thing distinctly unpleasant, and may even in some circumstances feel sick - in which case once the effect has largely worn off in an hour or so, they can give it a miss and do something else. But all of these effects are transient, that is they do not persist once the drug is out of the user's system.

Third, problematic patterns of usage. I have no doubt that just like a hundred other activities, from overeating to computer games to routine and joyless sexual activity, cannabis can become the focus of obsessive and excessive behaviour. Being very stoned all the time is, just to be clear, not a good way of going about things. It is not going to do one's state of mind any good. Distancing oneself from reality for a prolonged period is bound to end up distorting your view of the world, your relationships and even to cause such warped perceptions and habits of mind as to count, technically, as 'psychotic symptoms' (this covers much more than persistently hearing voices, or harbouring bizarre and grandiose delusions). But such symptoms are symptoms of cannabis use or of an ongoing cannabis 'dependency', not of anything else.

Cannabis can no doubt be used as a crutch to deal with problems which would be better faced up to. (Or in some cases, perhaps for problems which are not best faced up to - in this vale of tears, we've all got to try and muddle through somehow and if you can fend off misery for a while I'm not going to frogmarch you off to the cells for it).

It's possible to waste a lot of time sitting around getting stoned, and it's not the kind of thing you want to be doing all day every day. But these are not, in the end, the fault of the evil weed. It doesn't enslave anyone, and it doesn't force anyone to take it (and as I've pointed out, dealers rarely need to do any 'pushing' - they're more likely to be worried about keeping enough back for their regulars, and heading off such new customers as seem likely to draw heat or otherwise cause trouble.) People can take cannabis or not, and most people are quite capable of enjoying a bit of a smoke in moderation without any associated behavioural problems whatsoever.

It's worth noting that problematic cannabis use - particularly very heavy cannabis use, or 'dependence' is not something that simply infects people, as the 70-year-old tropes of invasion and contagion would suggest. Rather, it has its own antecedents and risk factors. It can be explained statistically to some degree, and even be regarded as a symptom. A subfield of psychiatric and psychological research is dedicated to investigating this issue:

Hüsler et al., 'Psychosocial predictors of cannabis use adolescents at risk'

Green et al., 'Predictors of cannabis use in men with and without psychosis'

Creemers et al., 'Predicting onset of cannabis use in early adolescence: the interrelation between high-intensity pleasure and disruptive behavior. The TRAILS Study.'

Wilens, et al., 'Alcohol, drugs, and attention-deficit/ hyperactivity disorder: a model for the study of addictions in youth'

To pre-empt for a moment the next post, none of this gets us anywhere near justifying either the kind of Draconian laws that would be required to effectively banish the stuff from the realm, nor the kind of paternalistic threats required to stop people from taking it. And even if those things could work, I happen to think they are not worth the price. If people are turning to problematic patterns of drug use, that is likely to suggest a lack of hope, direction, the realistic possibility of ambition. The plight of the poor, marginalised and 'socially excluded' is not primarily about the effects of ganja use.

I should point out that being brought up in a city is said to increase the probability of a schizophrenia diagnosis by a factor of 3-4. That is substantially more than current figures - which I dispute in any case - would suggest for the cannabis link. It is nice to hear Hitchens speak of the 'weak and vulnerable', rather than telling people to pull their socks up and their fingers out. Shame the next sentiment to follow should be 'bang 'em up'. (Or is that - infantilise them before they have a chance to do it to themselves?)


Overblown claims

So much for relatively uncontroversial harms and disadvantages of cannabis use. What of 'overblown' claims? I want - Hitchens states - to know precisely which claims he views as ‘overblown’. I don't think propositions about cannabis use can be overblown in themselves - only more or less accurate. For a claim to be overblown, there has to be someone doing the claiming, and along with it the overblowing.

My initial mention of the self-defeating nature of overblown warnings was tangential to the issue of criminalisation. If cannabis were legal, I would still, in a helpful spirit, warn against inflated claims. People are not stupid. when you're telling them what they want to hear, or something uncheckable about what is happening in a foreign country they don't have any stake in, you can get away with rampant lies. When you're giving them information that directly affects their lives and about which they have other sources of information, any hint of the 'noble lie' is going to scupper any chance of being taken seriously. Consider the damage done to the global warming campaign as a result of the behaviour of the climate scientists at the University of East Anglia and the subsequent fudged inquiries, just as one example.

Criminalisation, though, tends itself to produce palpably exaggerated warnings. After all, the current criminalisation of cannabis doesn't tend to be defended by reference to any moral arguments about the charred corpses of toddy-drinkers, or by alluding to something Alan Bloom once wrote (more on those topics in part three). The justification that is offered - and which would be offered for any remotely plausible future 'crackdown' - is primarily danger. Accordingly, the warnings given to provide that justification have to be suitably, well, overblown.

This is perhaps especially ironic in the case of those who advocate criminal sanctions without believing, as Hitchens does, that an acceptably civilised prohibition regime could come close to eradicating cannabis use. For some such people, criminalisation is seen as a means of providing public information messages, on an 'only language they understand' kind of basis. In the recent hoo-hah over the finer points of drug classification (which neither I nor Hitchens consider to be of more than marginal relevance to altering behaviour via deterrence), this was just about the only relevant consideration. Having downgraded cannabis from class B to C, politicians responded to pressure from those they judged as having the strongest feelings - and loudest voices - and rapidly performed an about-face.

This memorable advertisement managed to be appear rather overblown without providing much actual content:

Just to complicate things, this was introduced as part of a a campaign which followed the short-lived downgrading of cannabis - and was intended to offset a reduction of penalties rather than to justify an increase.

As this video suggests, and my first post indicated, though only in outline, the main strand of widely exaggerated - and very probably misconceived - rhetoric I'm concerned about relates to the link between cannabis and schizophrenia.


Cannabis and Schizophrenia

A number of clinicians and academics are pursuing this avenue of investigation with some optimism, notable among them being Robin Murray, whom Hitchens suggests it would be 'arrogant' to contradict. Of course it would also be arrogant to contradict those scientists like Nutt (the Drugs Kaiser fired by the last government for expressing the opinion he was employed to give) or many others who have less invested than Murray in the cannabis/schizophrenia issue. So going by Hitchens's logic, arrogance is unavoidable if one is to make any useful judgement about the matter. So I'll embrace my arrogance with regard to Dr Murray, and see if I can explain to others my reasons for doing so.

Hitchens's question concerned my acceptance that cannabis has - or might in future be found to have - any dangers for those who use it. As well as accepting that cannabis is associated with some harms and dangers, I certainly accept that all kinds of things might yet be discovered. Such general possibilities are however of little interest, since they cut both ways. One should certainly observe Cromwell's precept think it possible that ye may be mistaken - not to do so is rank dogmatism. But in the context of adversarial debate, unilaterally admitting such a possibility in general terms is apt to be taken as a concession to the other side.

So arrogance aside, of course I may be mistaken. I may be mistaken about anything. Am I especially likely to be mistaken about this matter? No. In fact given that I've taken considerable interest in the matter, done a fair bit of research, and have some general training in methodology, pretty good analytical ability and even a personal connection with the world of psychiatric research, my judgement is that I'm less likely to be mistaken about this than many other things. The key is to address the issue of how likely one is to be mistaken before finalising an opinion, and thus moderating the certainty with which you hold that opinion. If you want to be really fancy about it, you can even consider some alternatives and assess how much credence you might give those, then formulate a complex opinion, with a probability distribution and everything. There is always a generalised risk (or uncertainty) that one might be mistaken, but that is a common factor which can be ignored so far as choosing among opinions goes.

So I am happy to 'fess up to the generalised and unquantified possibility of unknown error. But I don't consider such a possibility to undermine my opinions. And - importantly - I apply the same principle to my criticism of the research into this topic - I have a good and coherent reasons to think that there are specific problems with the research. I am not just taking pot shots at a view I'm hostile to, or waving my hands in the general direction of error and claiming victory for the 'druggies'. It is one thing to point out, for example, the common hazards of using complex statistical methods to derive conclusions from data. It is another to indicate specific ways in which such hazards appear to have affected research in this area. Doing that properly is not a straightforward business, though, and as a result these comments are lengthy and go into some detail.

Murray: marijuana warner

The research programme has been under way sporadically for some decades, but has not yet produced sufficiently impressive results for any scientist, even a true believer (and advocate) like Murray, to be willing unequivocally to assert, in an academic paper, a causal link leading from cannabis to schizophrenia. Murray, like anyone who thinks he is onto something, has a tendency to bury his caveats and forget them as quickly as possible, but he still has to include them, and their relative lack of prominence ought if anything to draw more attention to them.

Murray has been busy prematurely pseudo-replicating his findings in a variety of reports, survey articles, and meta-analyses, and has been involved in rehashing and revisiting existing studies, to squeeze every drop of statistical goodness from the data. I will consider at least one of these later, but for a foretaste, consider that a widely-cited survey article in which Murray was involved states:

Consistent with the previous findings, a follow-up study of the same Swedish conscript cohort showed that ‘heavy cannabis users’ by the age of 18 years were 6.7 times more likely than non-users to be diagnosed with schizophrenia 27 years later (Zammit et al, 2002). This risk held when the analysis was repeated on a subsample of men who used cannabis only, as opposed to using other drugs as well.

This is flatly incorrect: the first figure was in fact 3.1 rather than 6.7 - though because of an accident of the numbers, it is, thankfully, possible to attribute this to carelessness.

Murray is though without doubt committed - to a degree that I would consider prejudicial - to establishing a causal link between cannabis use and subsequent onset of schizophrenia. His writings in the press make this quite clear: for example, he claims that Nutt and the rest of the 2002 ACMD 'boobed' by advising that cannabis has no serious mental health consequences. He then states that the evidence that cannabis contributes causally to the onset of schizophrenia "although not yet conclusive, has been mounting steadily over the past six years." Since this was written in 2009, elementary arithmetic tells us that the inconclusive statistics Murray refers to had not even begun the process of 'mounting steadily' at the time the ACMD 'boobed' so badly by ignoring them! I should point out that I don't accept Murray's statement about the mounting data, nor much of anything else he says, at least not in newspaper commentary. My purpose is to expose one of the reasons for my lack of confidence in his capacity to maintain scientific detachment - or indeed basic coherence - in discussing these matters.

Murray states:

Professor Nutt contrasts a 2.6 fold increase in risk of psychosis carried by using cannabis with a twentyfold increase in risk of lung cancer if one smokes cigarettes. Unfortunately, he is not comparing like with like. The twentyfold increased risk is not carried by just being a cigarette smoker but rather by being a long-term heavy smoker. For cannabis, the risk of psychosis goes up to about six times if one is a long-term heavy cannabis smoker.

One of the more thoughtful of the proliferating survey articles, Semple et al, suggests that the best estimate of statistical correlation between cannabis use and subsequent schizophrenia diagnosis, adjusting for some but by no means all sources of error, is an odds ratio of around 1.41.

This ought to mean, if it were a directly usable statistic, that on finding out someone has smoked cannabis you should increase your assessment of the probability that they will subsequently receives a schizophrenia diagnosis by a factor of 0.4 (for an average person - somewhat less for a higher-risk individual) while for the very heaviest cannabis users, the best guess - though admittedly still probably too high, and by no means indicating causation - is given as an odds ratio of 3 - not 6, which is the figure Murray has cherry-picked from somewhere. Even if unlike me you think this research is on the right track, this figure can't be justified.

In 2007, too, Murray was at it, in the Independent, making strange claims about skunk:

Skunk is to old-fashioned hash as is whisky to lager. You can become alcoholic by just drinking lager; but you have to drink a lot more lager than whisky.

Hash is resin, and the more refined forms of it are roughly as strong, in terms of THC content, as skunk. Typical skunk is stronger than most other herbal cannabis - but is accordingly more expensive. And the difference in strength, i.e. bulk, is negligible - like the difference between whisky and whisky with a dash of water. There is nothing stopping people from getting just as stoned on relatively weak cannabis as they do on the strongest stuff.

Murray cites a UN report which, like his own article, reads like a polemical document. So much for no interest from government - the UN and various government bodies have sponsored a great deal of the work that has been done in pursuit of cannabis-induced schizophrenia). (I happened to notice p117 has a graphic showing a 5% increase in cannabis use over 12 years, but is described in the text as showing a 10% increase in a different 5 or 6 year period (between the 'late 90s' and publication in 2006. The graphic itself is a fantastically misleading, giving the visual impression of cannabis use shooting up from near zero to some huge figure.

In the preface to that UN report, Antonio Maria Costa, Executive Director of the UN Office on Drugs and Crime, feels able to say

With supply virtually unlimited and demand subject to the vagaries of government policy, traffickers have invested heavily in increasing the potency - and therefore the market attractiveness - of cannabis. The result has been devastating: today, the characteristics of cannabis are no longer that different from those of other plant-based drugs such as cocaine and heroin. (emph. mine).

It is hard to say what this is supposed to mean, but it's even harder to find any possible meaning at all one which it comes out true.

(In an even less convincing variation on Murray's non-sequitur, the report also states on p.174 that "Most of the cannabis smoked in the 1960s would be considered to be of low quality today. In addition to seeds, it comprised a great deal of the large leaf, twigs, and other material that would be currently regarded as waste." But it would also have been regarded as waste then, and thrown away, with the possible exception of large leaves.)

The UN report's distinctly partial approach to cannabis might suggest that Hitchens is not on particularly strong ground when he suggests that funding for studies seeking a causal path from cannabis to schizophrenia is unlikely to be available from governmental sources. While funding is not a central issue in the absence of specific allegations of prejudicial motivation, I would suggest that neither drug companies, from whom much funding for medical research comes, nor - contrary to Hitchens's quixotic protestations - governments seem likely to be very interested in research which would debunk the findings of this recently-hewn niche in schizophrenia literature. And as I pointed out in my first post, alcohol interests have been known to give financial assistance to anti-cannabis causes. Trying to paint cannabis as a populist issue, with powerful pro-legalisation forces ranged against the little man who favours prohibition is not a viable position so far as I can see.

Having said that, as far as it is of any relevance at all, Murray himself makes it clear that, wedded though he is to the RM-causation hypothesis, he does not subscribe to Hitchens's prohibitionist agenda:

I care little whether cannabis is classified as a class B or class C drug. Fourteen year olds starting daily cannabis use do not agonise over its exact classification; many do not even think it is a drug and few have any knowledge of its hazards. By comparison, most adults in the UK drink alcohol in moderation, but do so in the knowledge that drinking a bottle of vodka a day is likely to be injurious to health, and few are in favour of daily drinking from age 14 years.


And that, for all my disagreement with Murray on science and rhetoric, is largely my approach. If there are significant dangers, then a properly independent agency staffed by professionals should be tasked with publicising them honestly and clearly. Peter Hitchens talks of the vulnerable and weak - I think he means teenagers - but there is a limit to how far we should go in trying to protect teenagers from danger, and attempting to eradicate cannabis from the country by instituting an enormous crackdown is too far. My personal opinion - in itself largely irrelevant since no-one need share it to reject increased criminalisation, nor to endorse legalisation - is that to justify a ban, the harms would have to exceptional - near-certain lethality, say - or the substance would have to be capable of taking people over and sapping all ability to resist taking it - something that neither I nor Hitchens think is remotely the case. I don't particularly approve of youngsters smoking dope in any case and under a legalisation regime I'd envisage similar - or should the evidence merit it, heavier and better enforced - penalties on selling to minors.

In fact, I have a question for PH - what kind of penalties, what expansion of prisons and of policing, do you suppose would be required to achieve your preferred cannabis-free UK? How much more stop-and-search would there have to be, how much drug testing, how many more homes smashed up looking for the stuff? Would we bring the import trade to a halt by searching every container that comes into the country?

So, to be clear: I consider that even accepting Murray's most egregious misrepresentations, the herb should still be legal. Why then, am I about to embark on extended criticism of the cannabis-induced-schizophrenia hypothesis? Well, in part simply because I consider it an interesting issue in itself, and a case study in how science works, and have accordingly done a good deal of research into the matter. That is a bad reason.

A slightly better reason is that I consider Murray in particular to have overstepped the bounds of scientific objectivity by some degree with his public statements, and suspect him of having done so - unconsciously, no doubt - in some of his academic work too. After all, here at 'Surely Some Mistake?', the policy is to aim the Cointreau of criticism squarely at the peach of privilege.

And then there is of course the urgency with which Hitchens has repeated his request - in all four tranches of his response - for me to state my position on the harms of cannabis, which of course most saliently involves the schizophrenia issue.

But the most pressing reason, as intimated in round one of the debate, is that overblown claims of cannabis-induced-schizophrenia have gained such traction with the public, and have been illustrated with highly emotive anecdotes calculated to appeal, in particular, to parents' deepest fears. Any attempt to actually change minds in favour of legalisation requires that this issue be put firmly in perspective, and the tenuous basis of the fearmongering be established.

The rhetoric of prohibition

One of the less decorative rhetorical devices in the canon is 'correctio' - making a bold statement, then moderating it - the aim of course being to gain an (unfair) advantage in debate. The stronger statement may be unjustified, but is more persuasive than a moderate one, so assent is first gained by use of the stronger proposition, and even after this is pared back to pre-empt criticism, the assent remains. In psychology, a similar tactic is known as the foot-in-the-door technique - a person is first asked to agree to some modest proposal, and once they have done so, the demand is ramped up. This can be done in stages, so as to avoid escalating demands too abruptly.

The engineering of public support for, among others, the US/NATO invasions of Kuwait, Iraq, and Libya all showed observable elements of this technique. A foot in the door was gained by telling extreme tales of the dreadful events that were going on or were imminent and which required military intervention (tales which to many were rather obviously fabricated; consent need not be universal). As it became clear that these tales were indeed false and fabricated (in the case of Libya, this has not yet really happened - that's progress for you), they were abandoned or replaced with slightly more realistic reasons for war. Classically, the psychological foot-in-the-door is supposed to depend on a 'preference for consistency'. War rhetoric is rather different - once you get people marching to that drum beat, the momentum gained owes more to the lower reaches of the brain than to any preference for consistency.

The same is true of an opening gambit that involves the highly emotive elements of parental protectiveness and the spectre of rampant insanity. Once the image of murderous lunatics is out there, and parents are provided with appropriately lurid scare stories about their own offspring's chances of developing schizophrenia, subsequent gradual moderation of the message has little effect.

Neither I nor, as far as I know, any other prominent advocate of cannabis legalisation denies that cannabis use by schizophrenics interferes with standard treatments and can worsen their prognosis. I don't challenge the idea that ongoing very heavy cannabis use and its associated lifestyle can be a factor in bringing forward the onset of frank psychosis in those who are on the pathway to schizophrenia. And as I suggest above, ongoing heavy cannabis use is itself likely to be associated with phenomena which will show up on a checklist of 'psychotic symptoms'. What I do deny is that there is good reason to believe in what I - in honour of the Mail story linked above - will call RM-causation (for 'Reefer Madness' - a famously lurid propaganda film of the 30s, when the failure and reversal of alcohol prohibition was leaving FBN agents short of work, and (some say) timber and Nylon interests were anxious to see off the threat of new hemp processing technology, among them a newspaper tycoon who was responsible for much 'yellow journalism' demonising hemp... but I digress). Furthermore, considerable effort has been made to establish that cannabis causes schizophrenia, and at least one scientist at the centre of these efforts shows signs of excessive zeal. Failure to discover a link in those circumstances, goes beyond mere absence of evidence to constitute evidence of absence.

RM-causation is the big headline item, the one that, thanks to the likes of Murray, gives every parent nightmares (and my interest in the issue was certainly influenced by my own parenthood). This is the claim that there is a more than negligible chance that a person who would otherwise have been unaffected by the horrors of schizophrenia might develop the condition through taking cannabis. I repeat, the key claim I am referring to as RM-causation and which has lodged - quite understandably - in the mind of the public, in particular of parents, is that cannabis is capable of causing schizophrenia in those who would otherwise not have developed the condition. Of course, at the margins it will no doubt be possible to point the finger to aspects of cannabis use, especially very heavy and prolonged use, as factors in tipping borderline cases into full schizophrenia - the same is true of almost anything you care to mention. Remember, social isolation is regarded as a risk factor for schizophrenia as is, for another example, being brought up in a city - the relative risk of that factor supposedly being higher than that currently claimed with any credibility at all - that is, assuming I am completely wrong - by the Murray camp.

Given what they are being told, I don't blame parents for being worried. While we didn't refuse the combined MMR vaccine on our son's behalf, we certainly gave it some thought before opting to volunteer our own flesh and blood in the cause of herd (and his own) immunity. And elsewhere, I've defended the decision other parents took to refuse the jab as not only understandable but rational, given the widely available information. The present case is of course not one of an Andrew Wakefield, though it is worth remembering that his research was published in the Lancet. the point is that his warnings, while not backed with strong evidence, concerned a sufficiently severe outcome to be of great concern even at low probability, especially given that the risk could be very easily avoided without much wider impact.

So a head-on approach seems better. I'll explain why I have been suspicious of this strand of research (other than the selfish desire to keep open the option of revisiting youthful habits with impunity), and why I believe it is on the wrong track. I reiterate that to do this I'll need to go into some detail, since otherwise my criticisms will amount to little more than denial, or else may appear to be general nitpicking objections of the kind which could be made to any study. It is important that the root problems I detect are all of the same two kinds, and that more ad hoc objections to individual studies are geared towards showing how these general problems are manifested.

I'll also point out how the claims currently being made bear little resemblance to those touted so loudly when the issue first found its way into our esteemed press. A quiet and gradual correctio has been under way, with the result that the original headline-grabbing claims of RM-causation are just about extinct, and some much more modest positions taken up in their stead - positions which would never on their own have been considered remarkable, and which I have very good reason to expect will as the research project continues, become more modest still.

The decline effect and bias

Once this vocal but heterodox group has made more headway in influencing scientific opinion, results which produce the finding 'nothing to see here' will become more interesting and might stand a chance of being published. The fact that negative results are commonly rejected for publication, especially when a new 'discovery' is in its infancy has, in conjunction with related forms of bias, been dubbed the decline effect. The linked article observes that initially promising findings tend to fall off, and even to disappear altogether, as further studies are conducted. There are a number of reasons for this: the usual standard of 95% confidence implies that one in 20 such statistical results should be expected to be wrong without containing any errors or design flaws, or to put it another way, the probability that a given statistic falls outside the 95% confidence range is higher than that of rolling snake eyes on a pair of dice. If a one-in twenty chance of this kind appears to make a new and exciting finding, a wild goose chase ensues. (And if multiple re-analyses are run on the same data, any bias in that data will be manifested on each occasion.)

Various forms of bias can of course creep in - including the all-too-human tendency for scientists to want to believe, and to interpret results accordingly, even to re-run a statistical analysis using a different test, then discard the 'anomalous' result (this, though generally impossible to detect, is entirely unethical and invalidates any findings). Not that I'm saying that Murray himself has ever done anything like that; this is a general point. Another reason why results 'disappear' from subsequent studies is that the subsequent studies refine and improve experimental design, and eliminate flaws that had affected previous work. That seems to be one of the things that has happened in the case of the RM-causation hypothesis.

(The popular article linked has been extensively criticised by scientists. The key criticism has been that it poses as a bold debunking of science, but is in fact just a perfectly ordinary account of how science works - to the scientists themselves, it is too obvious! I would agree with these criticisms as far as they apply, indeed I'd consider them to reflect the same point I have already made - that general possibilities of error are ubiquitous and care should be taken not to exaggerate their importance. In this case, the criticisms amount to pointing out that the proposition 'for every claim, it is possible that it is mistaken' is different from and far weaker than 'it is possible that every claim is mistaken'.)

The decline effect has certainly occurred in this case - in the academic literature, multiples like 6, and even 12, times increased expectation of schizophrenia in cannabis smokers are now nowhere to be seen. Instead, figures from recent studies have fallen below 2. 2 of course here represents a doubled expectation (or 'risk'), but that is double a very small starting probability, so even if it were accounted for entirely by RM-causation, that would still be a very small risk.

Publication bias is a real and widely-acknowledged problem. There are methods for detecting it, such as funnel plots though they are of limited usefulness, especially as a means of ruling it out. On the one occasion I have seen such methods used in this context, in a Semple et al. survey article, the funnel plot does seem to show skew:

(and the Egger bias plot, another test, fails to show a line passing close to the origin), yet the funnel plot is described as follows:

Where studies with results in a certain direction are not published or identified, the spread of studies about the horizontal line will tend to be asymmetrical. The study effect sizes shown above are approximately symmetrical about the line of overall effect and the presence of publication bias is not supported

But the study effect sizes are not symmetrical about the line - 5 of 7 lie above the line. Furthermore, it is the smaller studies which are supposed to reveal publication bias if it is present - the larger studies effectively decide the position of the line. All of the four smaller studies are above the line. If the plot is to be taken seriously, the conclusion from this one ought to be that there is evidence of bias.

(The Egger plot is also explained away by a very odd argument: although the line does not pass through the origin exactly, the confidence interval for the intercept (indicated by two small circles on the vertical axis) includes the origin. This may be interpreted as a lack of statistically significant publication bias.)

The pachyderm in the parlour - no rise in schizophrenia

Perhaps the most obvious stumbling block in the way of establishing RM-causation is the rather obvious fact that if cannabis were actually causing people to become schizophrenic who wouldn't otherwise have done, we should expect, in the absence of countervailing factors, to see a proportionate increase in the rate of schizophrenia.

This has of course been pointed out by a number of eminent academics in the field, for example:

there is no evidence that the incidence of schizophrenia has risen dramatically over the past 50 years, in parallel with the huge increase in cannabis use.
- Dr. Colin Blakemore, Chair of the Department of Physiology at the University of Oxford; Dec. 27, 2002 email to ProCon.org.

Given that the incidence of schizophrenia declined substantially in Western societies in the 1970s, at the same time cannabis use was rising, it seems highly unlikely that marijuana causes schizophrenia in otherwise healthy people....
Cannabis psychosis is self-limiting, disappearing in a few days with or without medical treatment. Toxic psychosis probably occurs more commonly in individuals with preexisting psychiatric disorders...
- Lynn Zimmer, Former Professor Emeritus at the Queens College, City University, New York (CUNY), in Marijuana Myths, Marijuana Facts, 1997)

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.
- Professor Leslie Iverson, Oxford University. BBC News

To quote from the relatively thoughtful meta-analysis of Semple et al:

The question of whether cannabis is a precipitating or a causative factor in the development of schizophrenia remains. A recent study that used mathematical modelling to explore the possible effects of cannabis use and schizophrenia (Degenhardt et al., 2003) supported the possibility that cannabis precipitated psychosis in vulnerable individuals and that cannabis use is more likely among individuals with schizophrenia, but did not support a direct causal hypothesis. The main reason for this finding was the absence of any increase in the incidence of schizophrenia, despite clear increases in the use of cannabis in the Australian population studied. Any hypothesis that suggests that cannabis causes schizophrenia must explain this discrepancy in the epidemiological data.

Again, Degenhardt et al., Comorbidity between cannabis use and psychosis: modelling some possible relationships, Conclusions:

This study has examined a hotly debated issue from a different point of view - through the use of modelling, using data-based parameter estimates to predict what changes we would expect to see in the incidence and prevalence of schizophrenia if each of four hypotheses about the relationships were true. If cannabis caused schizophrenia among persons who would not otherwise have
developed the disorder, significant increases would have occurred in the number of persons with the illness. Given that the incidence of schizophrenia is either unchanged or decreasing, such a causal relationship is unlikely.

All other hypotheses provided a better fit to the available data. However, it is difficult to judge which of these is the best fit, or whether a combination of them is most appropriate. If cannabis use acts as a precipitant of psychosis, we would have seen small increases in the number of early onset cases. If cannabis use made relapse to psychotic symptoms, we would have seen small increases in the number of chronic cases. Finally, if persons with psychosis were simply more likely to become regular cannabis users, we would expect to see no differences in the number incident or chronic cases, but a higher prevalence of regular use in this population. Future research needs to examine these possibilities in prospective studies.

This approach has suggested that cannabis use is probably not causally related to psychosis in the strong sense of causing cases that would no otherwise have occurred. Nonetheless, cannabis use may affect persons who are vulnerable to developing psychosis, or who have already developed the disorder. This suggests that such persons may be advised of this possible relationship and counselled against using cannabis.

This is a big problem for Murray and, to the extent they share his premature zeal, others pursuing this line of investigation. It has led Murray to suggest that it is too early for the supposed effects to be seen in the general population, because it mostly affects young people, and is associated with the new 'super skunk' (something of a myth in itself, though I believe that newer strains and hybrids used in producing intensively cultivated sinsemilla have little or none of the chemical cannabidiol which, it has been suggested, moderates or otherwise interacts with the action of THC. Of course no large studies into that topic can be conducted, nor most probably would any recommendations be worth making, because without regulation, no-one knows what they are smoking).

The idea that RM-causation is restricted to those taking cannabis at a relatively young age in any case seems to have been forced on Murray since the results of recent, more rigorous studies are, not for want of trying, inconsistent with much of an effect in any other group.

In fact, this research programme, though not (yet) fairly describable as degenerating, exhibits a certain degree of retrenchment, not limited to the moderation of claimed effect sizes typical of the 'decline effect. In some ways, this retrenchment resembles the psychological 'foot-in-the-door' and rhetorical 'correctio' techniques, though of course it is not intended as a technique. Murray's readiness to proselytise for an unproven proposition and his insistence on sticking to what are by now clearly inflated figures have certainly had much the same effect as a deliberate foot-in-the-door though.

These retrenchments include, as mentioned, a reduction in the claimed effect size from initial estimates of increased risk, which I believe were at one time around a factor of 10 or even more, to a factor of about 2 (i.e., if this statistic were to apply to the average person at average risk, an increased risk of 1 percentage point).

Another slippage has been in the range of the effect - while no-one ever really suggested that those not already genetically vulnerable to developing schizophrenia (it appears that about 50% of schizophrenia incidence is down to genetics) is going to be 'infected' by cannabis use, it seems that recently it has been suggested that those who are 'vulnerable' in this sense form a very small proportion of the population. Some even seem to be veering dangerously close to agreeing that RM-causation itself may not occur, and instead the only causal influence of heavy cannabis use is only to speed up the course of an already existing condition.

There seems also to have been slippage regarding the extent of use required to find any effect. While the Mail story linked above claims a single exposure increases risk by a factor of 1.4, recent studies almost exclusively limit their conclusions to the supposed effects of relatively heavy use and 'dependence'.

Perhaps the most significant and slippage, however, is in the outcomes examined. Schizophrenia is a recognised and very serious chronic condition with a fairly well-understood course. Initially, studies were primarily concerned with diagnoses of that condition. But as later studies sought to extract more information from data, they began instead to measure schizophreniform disorder, symptoms of schizophrenia, schizophrenia-like symptoms, and psychotic symptoms, none of which are the same thing as, and none of which are necessarily very closely associated with, schizophrenia.

This leads me to the first of two problems which I see as manifested in this line of research, which I don't believe have yet been adequately addressed and which in combination provide a specific alternative explanation for the observed results, without relying on unspecified residual confounding factors nor on general problems with statistical methods such as multivariate regression analysis.

The two problems

1. Transient cannabis related psychotic symptoms.

In those studies which measure psychotic, or schizophrenia-like symptoms rather than schizophrenia proper, there is a danger that transient psychosis-like symptoms of cannabis use and of ongoing heavy cannabis use will be interpreted as manifestations of some independently subsisting psychosis. Such effects of cannabis as sensory or cognitive disturbances and so-called 'paranoia' would surely show up in test for psychotic symptoms. This possibility is explicitly acknowledged in some papers, and ought in any case to be taken account of by diagnostic criteria and methods, but it doesn't appear to have been addressed properly in many cases. As more recent studies have attempted to address the issue of confounding more thoroughly, they have also tended to expand the range of outcomes measured so as to include mere 'psychotic symptoms', thus introducing this new source of error. Little more can be said at this point. I'll return to the issue later in discussion of some individual papers.

2. 'Self-medication' (common predisposition)

In my first post, I used this term somewhat inaccurately to describe a wider phenomenon: a common predisposition both to take cannabis and to develop schizophrenia. Self-medication describes only those cases in which the subject is aware of taking cannabis to deal with symptoms of their illness. This does not accurately describe the case in which a person has some mild precursor or 'prodromal' mental disturbance which which motivates them, consciously or not, to seek cannabis or mind-altering substances in general.

Neither does it describe other explanations for taking cannabis which might be correlated with a predisposition to schizophrenia: tendency to rebellious behaviour (cannabis being illegal and generally frowned upon means taking it has this character - and smoking it all the time certainly does), propensity to 'addictive' behaviour, a general desire to alter consciousness and so on.

A substantial literature exists on the prodrome and precursor; to offer some haphazardly selected examples:

Cannon et al., Early and late neurodevelopmental influences in the prodrome to schizophrenia: contributions of genes, environment, and their interactions

Yung and McGorry, 'The Prodromal Phase of First-episode Psychosis: Past and Current Conceptualizations'

White et al., 'The Schizophrenia Prodrome'

Owens et al., 'Precursors and prodromata of schizophrenia: findings from the Edinburgh High Risk Study and their literature context.'

Amminger et al., 'Relationship Between Childhood Behavioral Disturbance and Later Schizophrenia in the New York High-Risk Project'

Kwapil, 'Social anhedonia as a predictor of the development of schizophrenia-spectrum disorders'

Hans et al., 'Interpersonal behavior of children at risk for schizophrenia'

Roff et al., 'Childhood antecedents of schizophrenia: developmental sequencing and specificity of problem behavior'

Self-medication properly so-called would be a case of reverse causation - schizophrenia causally contributing to cannabis use, rather than the other way round. Reverse causation is not a special kind of causation, just one that runs in in the opposite direction from that expected or hypothesised. Consider the fictitious tabloid headline: 'BLACK INFLUX LOWERS HOUSE PRICES'. The headline suggests that the arrival of black people in an area causes the house prices to drop. Reverse causation would be said to occur if in fact the reason for the co-incidence of lower prices and more black people arriving in the area is that black people have on average a lower income than the general population, and more were attracted to the area as a consequence of greater affordability.

In other cases of a common disposition, cannabis use might not be causally mediated by schizophrenia, but there may instead be a common causal influence - perhaps some genetic factor and/or an abnormal brain chemistry - which gives rise both to cannabis use and to schizophrenia. This would then be a confounding factor. Both reverse causation and confounding factors have to be addressed when a causal influence is to be established between correlated phenomena - most of the studies try to do this to some extent, though few of them, especially the earliest ones, even attempt to do so very thoroughly (by the way, one likely confounder, city living, is apparently believed to be associated risk of schizophrenia increasing by a factor of 3-4, substantially higher than that indicated by studies of RM-causation, even taken at face value).

In any case, I think it is sensible for my purposes to treat this family of explanations, variants of which have generally been considered the obvious explanation for high cannabis use, as one hypothesis, despite the fact that it cuts across the distinction between reverse causation and confounding factors. I'll refer to it by the catchy name 'common predisposition'.

The common predisposition hypothesis has been around in one form or another for a long time, and is the obvious competitor to the RM-causation hypothesis, as all the authors of RM papers acknowledge in one way or another. A range of scientific work supports the general idea that the schizophrenia-prone are also likely to be cannabis-prone.

This paper attempts a partial explanation of substance abuse by reference to underlying aspects of schizophrenia which need not be contemporaneous with manifestation of psychotic symptoms:

Chambers et al., A Neurobiological Basis for Substance Abuse Comorbidity in Schizophrenia

Other papers of of relevance examine various behaviours which can be precursor or prodromal predictors of schizophrenia, and link those directly to cannabis use:

Korhonen et al., 'Externalizing behavior problems and cigarette smoking as predictors of cannabis use: the TRAILS Study'

Buckner et al., 'Specificity of Social Anxiety Disorder as a Risk Factor for Alcohol and Cannabis Dependence'


This type of consideration might be of special relevance in the context of a finding that only heavy use is of relevance to subsequent development of schizophrenia, since it may be the case that such heavy use, as distinct from typical recreational and social use, tends to arise from other abnormal behaviours or dispositions.

As is the way of these things, those investigating (hoping to find) RM-causation have a natural tendency to regard the common predisposition hypothesis as an obstacle. It is to be controlled for, eliminated, ruled out. That is quite reasonable. Less reasonable but equally understandable is the inadequacy of the measures taken to rule it out. If you are, like Murray, a true believer in the RM-causation hypothesis, you will approach the matter as a necessary but rather tiresome technicality, much as a sleuth who's convinced he has his man may regard the task of observing proper procedures in gathering evidence. As a result, few of the attempts to rule out the common predisposition explanation in studies not directly aimed at that end are convincing, even among those which do attempt to account for a wide variety of other confounding factors.

Two brief examples:

Henquet et al., 'Prospective cohort study of cannabis use, predisposition for psychosis, and psychotic symptoms in young people', which appears to do a relatively thorough job in addressing standard confounding factors, seems vulnerable to challenge on both counts. The outcomes tested are merely 'psychotic symptoms - a far cry from schizophrenia, on which early studies concentrated, and which remains the headline topic of the kind of discussion in which papers such as this are commonly cited (or alluded to). So far as common predisposition is concerned, while the paper attempts to evaluate predisposition for psychosis, it states: we used self reported psychotic experiences on the symptom checklist to determine predisposition for psychosis. This is not a test of predisposition for psychosis, but of levels of psychosis itself.

Fergusson et al., 'Tests of causal linkages between cannabis use and psychotic symptoms' takes some trouble to attempt to adjust both for confounding factors and for common predisposition, but its results are solely in terms of 'psychotic symptoms'. It states:

participants were questioned on current (over the past month) psychotic symptomatology using items from the Symptom Checklist 90 (SCL-90) [20]. A series of 10 items were selected as representative of psychotic symptoms.These items spanned the following
symptoms: hearing voices that other people do not hear; the idea that someone else can control your thoughts; other people being aware of your private thoughts; having thoughts that are not your own; having ideas and beliefs that others do not share; the idea that something is seriously wrong with your body; never feeling close to another person; the idea that something is wrong with your mind; feeling other people cannot be trusted; feeling that you are watched or talked about by others.

Confirmatory factor analysis of the item set has shown previously that the items formed a unidimensional scale reflecting the extent of psychotic symptomatology. (p. 356)

It is far from clear that the claim that the item set forms a unidimensional scale is robust to the presence or absence of prolonged heavy cannabis use. The last three items in particular might seem especially likely to be found in heavy habitual users of cannabis as compared to the general population. It would be interesting to know whether these were especially heavily represented among heavy cannabis users. I'm not inclined to rest a great deal of weight on the presence of items like having ideas and beliefs that others do not share in questionnaires of this kind - even though both I and Hitchens, neither of whom are remotely psychotic, might well give a positive response. Questions and questionnaires of this kind have generally been carefully calibrated and weighted so as to provide results which are overall statistically accurate.

I should add that these two relatively refined studies reported low increases in risk rate compared to the kinds of figures quoted in newspapers. In Fergusson et al., the highest estimate - for the heaviest cannabis users - was a 1.77 times greater risk of 'psychotic symptoms'. Henquet et al. found an odds ratio (a comparable measure in this instance) of 2.44 for increased risk of any 'psychotic symptoms'; this occurred among the second most frequent users at 3-4 times/week.

And finally

In the second part of my response, to follow shortly, I'll attempt to nail down and substantiate the position outlined here by examining in more detail some prominent studies in this area. The third and concluding part of my response will return to moral and policy arguments relating to legalisation.

Apologies to Peter Hitchens for the delay in responding - other obligations have taken up a good deal of my free time over the past week, and though my opinions were clear enough, as intimated in my first post, I'd not anticipated focusing on this empirical issue. Revisiting and documenting references, and attempting to impose some semblance of organisation on the whole thing, has been quite time-consuming.

Further apologies to readers for the length and relatively technical nature of the discussion. There seemed little point in addressing the issue of harm without addressing this central rhetorical issue, and there seemed in turn little point in addressing that without going into some detail, especially since RM-causation has a considerable head start so far as public discourse is concerned.

Finally, apologies to any scientists whom I may have offended - while I do think Robin Murray has gone further than he ought in his newspaper articles, the remaining limited ascriptions of bias and my criticism of specific papers, are not intended to suggest that any of those engaged in this research were acting in bad faith, nor that their conduct was in any way atypical of the normal progress of scientific inquiry.



[Edited 15 Jul 2011 11:30 - minor corrections, added missing references for prodrome and precursors]
[Edited 15 Jul 2011 16:01: moved references from one list to another]
[Edited 17 Jul 2011 17:55: belated minor corrections to spelling, syntax, sentence construction etc.]
[Edited 20 Jul 2011 10:44: 'correctio', not 'retractio']