Thursday, 24 October 2013

Meth: I choose you! Does study show addicts' choices are ultimately rational?

This post is a slightly modified version of one that was written by Jay Duckworth and posted on his own blog

Last month (16/09/13) The New York Times published an article in their science section entitled The Rational Choices of Crack Addicts, which was promoted online by numerous sources and proved to be quite popular. Incidentally, this popularity is evidenced by the numerous iterations of the original piece in various media outlets (some of which can be seen here, here, here, here, here, here and here). The version published in The Atlantic further enjoyed a promotional tweet by the famous Harvard psychologist Steven Pinker:

The Science of Choice in Addiction - Sally Satel - The Atlantic
— Steven Pinker October 2, 2013
The articles possess the same basic implication - that the predominant view of addicts as zombie-like creatures bound by their compulsion is wrong, and now we have evidence to prove it. The problem with this, however, is two-fold: first, the view outlined above is not the predominant view of most addiction researchers, but a caricatured version of a more measured and nuanced conception (Marc Lewis's description of addiction as "biased choice" succinctly parses this quite well; he also has a piece tackling a similar subject to this post). The second part of the problem is that the conclusions of many of the media outlets vastly exaggerate the reach of the study they cite.

The study in question is that of Kirkpatrick et al. (2012): Comparison of intranasal methamphetamine and d-amphetamine self-administration by humans - a title rendered sufficiently bland by its competitors in the media (Gawker's headline was Why Everything You Know about Crack Addiction is Wrong). The study's primary aims were to test the effects of different molecular amphetamines on cardiovascular health and psychomotor performance, as well as users' subjective experiences (i.e. ratings of mood change and behaviour towards each drug).

The general procedure (of interest here) was as follows; participants were subjected to two different sessions on two different days - the first was a "sample" session in which participants were provided with samples of both drug and monetary rewards; the second a "choice" session wherein these rewards were actively chosen by participants during a computerised task. Depending on their choices, participants received either some ratio of drugs and money, or maximal amounts of either reward. Participants were also assigned to either a high value condition or low value condition; those in the former received $20 on sample days and had the option of obtaining $20 on choice days (given they chose only monetary options during the task), while those unfortunate enough to be assigned to the latter condition received only $5. (Drug doses were the same for each condition.)

It was found that those in the $5 condition possessed no strong preference for a single reward, with only a slight choice tendency towards the money (59%). Those in the $20 condition, however, were far more likely to choose money over drugs, with 83% of options chosen being monetary. The authors concluded that when given a valuable-enough alternative, participants could be dissuaded from drug self-administration: a reasonable conclusion. The media then took this modest hatchback of a conclusion and inflated it to Ferrari-like proportions. The labelling of the media's conclusions as exaggerated is justified for two reasons: first, such significant conclusions are unlikely to ever be validated by a single study (this, as a rule, is not how science generally works). Second, such conclusions based on this particular study would be unwise given its numerous limitations.

We can first consider flaws with the study's cohort. With just 13 participants, the study's sample size was small; but further, since the design was between subjects (meaning conditions are effectively analysed separately) this reduces the numbers further to just 6 and 7 participants in each condition. These numbers are minuscule. Moreover, all participants were male and self-selected - none had ever sought help for their drug use nor were they interested in receiving any - meaning that what we might have here is a particular subgroup of drug users with different goals, intentions, personalities and experiences with drugs, which do not apply to many others not in this group. Furthermore, the authors note that only three of the participants met the criteria for drug dependence - this is only 23% of an already tiny sample! Overall this means that any conclusions derived from this study cannot simply be extrapolated to clinical populations.

There also exist problems with the study's design (given the conclusions of the media). Since participants were automatically supplied with drugs on "sample" days, we cannot make any solid deductions about whether the same results would be garnered in periods of abstinence. Also, given that many of the participants may have been occasional users, a hit of meth every few days may have been ample, meaning the only real option to go for would have been money anyway. This also means that the employed strategy of many may have been to sequester funds for purchase of drugs after the experiment was over. However, this is unlikely to have been the case - at least if the plan was to spend the monetary rewards on meth. The values of the drug and monetary rewards over the course of the study were well matched ($60-$80 and $100, respectively), and given the average dose of meth (250mg) will set you back around $80 on the streets of New York (the study's location), the participants would receive the same amounts of meth whether they obtained it from the researchers during the study or from Walter White afterwards (Breaking Bad references are still going strong, right?).

However, another small problem appears when we consider that these participants were also users of cocaine and marijuana, both of which are cheaper than meth (marijuana substantially so; see Table 1).
Illicit drugs prices in New York City: 1lb
~$2,000 - $3,000
Table 1: The links for these estimates can be found in the reference section.

This means that the above strategy would still be valid when applied to these other drugs, and so again any inference that these results alone show the way forward in how we treat addicts is mistaken. Finally, the doses used in this study were 5-20 times lower than average street use, and simply scaling these results upward and assuming they would remain the same is bad practice.

So as we can see, though this is by no means the worst study you're likely to find, the conclusions made by the media extended far beyond the study's grasp. This is a shame, not just because scientific research has been misrepresented - though this would surely be enough - but also because there are other studies out there which actually test whether alternative monetary reinforcers dissuade addicts' drug inclinations, and they support the results found here. Some might ask "Then why write this article?!" But to ask this question would be to miss the point. The point here is not to say that offering alternative reinforcers to addicts could never work, but to highlight the fact that this lone study could never provide the evidence necessary to show that it does. In the progress of science, evidence is more often than not carefully accumulated, not acquired in its entirety. The media would do well to remember this.


Hser, Y. I., Huang, D., Brecht, M. L., Li, L., & Evans, E. (2008). Contrasting trajectories in Heroin, Cocaine and Methamphetamine use. Journal of Addictive Diseases, 27, 13-21.

Kirkpatrick, M. G., Gunderson, E. W., Johanson, C., Levin, F. R., Foltin, R. W., & Hart, C. L. (2012). Comparison of intranasal methamphetamine and d-amphetamine self-administration by humans. Addiction, 107(4), 783-791.

No comments:

Post a Comment