Kandel, Kandel, and NEJM: flogging the gateway hypothesis to attack e-cigarettes

by Carl V Phillips

I expected to focus today’s post on all the stupid media reports on the publication that I wrote about yesterday via the CASAA press release. But a funny thing happened: almost nothing. Though this was the type of supposedly-authoritative sensationalism that the press usually eats up, there were remarkably few stories in the American press. So I decided to go old-school with the post instead, and focus on the lies and liars.

Now my readers in the UK might not realize that a miracle occurred. The British tabloids (by which I mean: most of the British national newspapers — you guys make our press corps look good) all covered it in the standard churnalistic fashion. My personal favorite:

https://twitter.com/cjsnowdon/status/507306984035864576

It seemed to be dutifully reported in India too, as expected. But in the USA, the only major news outlet that reported the sensationalistic junk was Time (no shock there). The New York Daily News reported the headlines, but did at least quote from our press release. That was basically it. Even the inevitable HuffPo piece came out of the UK.

Fingers crossed, of course, because zombies can rise at any time. But perhaps our press release did some good. (Apologies to our friends in the UK — it was just too damn expensive to distribute it outside the USA.) Once a reporter saw it, they faced the choice of ignoring our analysis and running with the sensationalism (always possible), running a story about how misleading the original claims were (a bit too bold for most), or just walking away. There was no way — if they took our critique at all seriously — they could run the standard “the study showed… but some critics argued…” piece. Because, of course, the study clearly did not show what Kandel^2 and the NEJM claimed it did.

As background, all you need to know is that they did a mice torture experiment, and reported that those mice who were dosed with nicotine may have liked cocaine better later. There are obviously more details to it, but that is basically it. If this translates to humans (a big “if”) then the implications are pretty much summed up in this handy lifehack interpretation from Alex Clark:

But the authors and journals did not spin it that way, even though you would think that would be just as good for getting headlines. No, the authors clearly intended to write a paper pushing “gateway effect” claims, which of course refers to one drug causing people to use another drug. Needless to say, liking the other drug better is a far cry from actually seeking it out and using it. Moreover, the authors chose to specifically focus on e-cigarettes, as if that were the only source of nicotine. Funny how nicotine patches did not get mentioned.

I made some random observations about the paper itself to a discussion group of interested researchers (based on the few hours we had to peer review this material before the anticipated media frenzy). So I will just repost what I wrote, for what it is worth, and move on to the blatant lies in the press release.

This paper starts out by acknowledging the problems with what it typically cited as evidence for gateway effects. It is a bad sign that they do this in the context of presenting some of it as if it were informative, but whatever. But then they go on to suggest that epidemiology “cannot” determine what causes “the progression from one drug to the next” (they seem to have already forgotten that they pointed out that we do not actually know that happens) and that only testing on nonhuman animals can.

This is wrong on both counts. Epidemiology can provide useful information about gateways (though it is quite difficult), whereas testing on other species is pretty much useless for understanding phenomena that are all about human volition and social systems. It is possible to use such tests to offer an explanation (largely unprovable, and inevitably one of many candidates) for why something happens (“why do people do this thing we know they do? perhaps it is because of this biochemical phenomenon”). But it obviously cannot show that it really happens.

The next problem is the generic problem with all “brain porn” research (a term used to rightly diminish it by biological scientists and that should be aggressively adopted by social scientists too): Something always changes in the brain when you have an experience. Thus it is possible to find the change. That does not mean what you find is causing a particular outcome. It definitely does not mean that it creates the social science phenomenon in question. The biggest problems with this come from brain scan porn, which are not used here, but what they present about molecules is basically similar.

Consider: If brains do a particular thing in response to nicotine dosing, but people react differently [from one another], that bit of biology does not offer us very good information about behavior. If it turns out that the degree of the biological response is associated with the real-world reaction, then maybe(!) you have found an explanation for behavior. But think about what you need before you can make that conclusion. You would already have to have a good measure of the real-world phenomenon in question, and the biology would just be trying to explain why it happens. It still would not be showing that it does happen.

Notice that when they circle back to people, with reference to a single study, they do not actually have any tie between their mice torture and the people. They basically just repeat the claim that they gave lip-service to denying in the introduction, that correlation is always causation, even when there is obvious massive confounding.

Also, it turns out that mice are not people, so the behavioral observations are not terribly informative. And that is actually an understatement, because lab mice that are forced to live in unnatural captive situations are very much unlike not just people, but mice. It was famously demonstrated that rodent propensity for “addiction” decreased dramatically when they were allowed to live in happy semi-natural colonies.

Another factor to keep in mind is that mouse scientists are every bit as bad about model shopping as bad epidemiologists. We have to wonder how many different versions of this experiment they ran before the mice cooperated to give the results they could publish in NEJM.

Finally, regarding their grossly inappropriate and irresponsible implicit policy conclusions:  Imagine that their results are not cooked, and that they really apply to humans, and that what they observe really translates into real-world preferences. Presumably if you ran the analogous experiment (on humans) enough times with enough different foods/pharmaceuticals/lifestyles/experiences, you would find quite a few of them that increased people’s appreciation of cocaine (caffeine, thrill sports, wild sex, and attending Harvard come to mind — either individually or in conjunction). This would not mean that they cause cocaine use to any measurable degree, and even if they did, it would most certainly not be sufficient to recommend restricting them.

The e-cigarettes emphasis looks like a ploy to turn what was basically boring research (for everyone other than the poor mice and those trying to get better value from their cocaine) into headlines. It was not out-of-the-blue, however. Rodu points out that Denise Kandel has been flogging disreputed Drug War gateway claims for 40 years, and so one could argue that this was a natural extension of her political goals. Still, given that e-cigarettes deliver far less nicotine to people every day than does smokeless tobacco and probably NRT (and obviously cigarettes), it is difficult not to settle on the publicity ploy explanation.

From the paper:

The gateway hypothesis was developed by Denise Kandel, who observed that young people become involved in drugs in stages and sequences.[Reference to her 1975 paper.]

Now some might think that a hypothesis that had been part of the Drug Wars since the 1950s was, at best, labeled in 1975, but whatever. What follows is some of the bait-and-switch that is common among gateway proponents, where they define gateway to mean merely using one drug before the other (which is easy to observe and support), and then quietly drift into claiming the former caused the latter (which is very difficult to try to show, and is seldom supported). It is important to note that after a bit of that game, the authors clearly settle on the normal causal interpretation of the term, and thus their use of the term clearly is a causal claim.

In the Columbia University press release the authors did not even pretend to recognize the difficulties with drawing gateway conclusions, and certainly did not let their emphasis on e-cigarettes get lost amidst any of that pesky sciencey stuff.  The headline was:

E-Cigarettes May Promote Illicit Drug Use and Addiction

When discussing D.Kandel’s claims about the gateway, even the veneer of scientific skepticism in the article, let alone a recognition of the widespread disagreement with her conclusions, was absent. Instead:

Dr. Denise Kandel’s further analysis of 2004 epidemiologic data from a large, longitudinal sample suggested that nicotine also primes human brains to respond to cocaine. She found that the rate of cocaine dependence was highest among users who started using cocaine after having smoked cigarettes. “Our findings provided a biologic basis for the sequence of drug use observed in people,” said Dr. Eric Kandel. “One drug alters the brain’s circuitry in a way that enhances the effects of a subsequent drug.”

What was the basis for saying the data showed a priming of human brains? Basically just the fact that lots of people used cigarettes before cocaine, and not many the other way around. I trust I do not have to explain why this does not show anything interesting, given that cigarettes are much easier to obtain. It is also interesting to contrast this claim (based on epidemiology) with the statement in the paper I refer to above, in which the authors state that epidemiology cannot show such a thing.

The observation that “dependence was highest” (even setting aside the tenuousness of measuring dependence) is easily explained by confounding. People who never smoked but gave cocaine a try are different in many obvious ways from those who acquire the easy smoking habit then add others on top of it. Still, if the tenuous claim really is true, the mouse experiment might indeed provide an explanation for why it happens, as I noted above. That is potentially a real scientific discovery, but one so fraught with uncertainty that even the statement in this paragraph is already going overboard.

So did they leave it at that? What do you think?

in light of the skyrocketing popularity of e-cigarettes, particularly among adolescents and young adults, the researchers say that more effective prevention programs need to be developed for all products that contain nicotine.

…and…

“E-cigarettes may be a gateway to both combustible cigarettes and illicit drugs. Therefore, we should do all we can to protect young people from the harmful effects of nicotine and the risks of progressing to illicit drugs.”

Why? The implication is because they are causing people to later use other drugs (recall the headline). But that is not what their research suggested even if you take their interpretation at face value. It merely suggested that you will like cocaine (not “illicit drugs” generally; not cigarettes) better if you use nicotine (not e-cigarettes in particular) first. Does this call for “more effective prevention”? Certainly not if the supposed increased liking for cocaine does not cause anyone to act differently. Do the authors make any attempt to assess whether anyone is acting differently, or even argue that it is happening? Nope.

And even if there were evidence of that, their conclusions would still not follow. Is it necessarily bad that someone likes cocaine more and thus chooses to use it whereas he would not have done so otherwise? That can certainly be argued, but it does not go without saying. Liking something more is normally considered a good thing, and so at least some case must be offered as to why that is not true here. If this leads to more life-ruining cocaine use, there is a case to be made. If it just means that some people enjoy an occasional indulgence that they might have otherwise not found worth the effort, not so much.

Moreover, what if the prevention efforts needed to keep people from using cocaine interfere with THR, or cost millions of dollars per cocaine-user-prevented, or cause most of the population to think that public health people are tyrannical nannying loons and thus ignore useful advice? The real world of public policy making is not nearly as simple as mouse torturers seem to think it is.

Apparently they do not even understand that actions by individual humans are more complicated than mouse biology (“of mice and men” joke omitted — those have already been done to death):

We don’t yet know whether e-cigarettes will prove to be a gateway to the use of conventional cigarettes and illicit drugs, but that’s certainly a possibility.

We don’t yet know if they will destroy the galaxy either, but the mere fact that it is a theoretical possibility is certainly not a basis for policy recommendations. Or a basis for headlines that are clearly intended to be interpreted at “it does that!”

Nicotine clearly acts as a gateway drug on the brain, and this effect is likely to occur whether the exposure comes from smoking cigarettes, passive tobacco smoke, or e-cigarettes.

For something to act “as a gateway drug on the brain”, it needs to cause volition — actual action to use the other drug — not just change preferences. Even if their claims about preferences were supported, they would not tell us that nicotine is acting as a gateway. In fact, the previous sentence in the paragraph (the previous quote) says as much. Notice the bait-and-switch. The previous sentence talks speculatively about the alarming actual gateway effect, while the second conveniently redefines it to a much weaker claim that they then assert to be true.

Oh, and that massive dose of nicotine from “passive tobacco smoke”? Seriously? Do they even understand orders of magnitude? Perhaps not:

the researchers point out that e-cigarette use is increasing exponentially among adolescents and young adults.

This was from the paper and cited to the first CDC scare report, which was based on figures for each of two years (h/t Bill Godshall for noticing that). Two points do not show whether a trend is exponential, linear, geometric, or any other basic shape. You can fit any curve shape to two points. Except we know it is obviously not following the purely exponential shape that would hit those two points, because if it were, the prevalence would exceed 100% by 2016. I tend to think it is a bad idea to trust any claim made by a scientist who does not understand basic math.

And finally,

“The effects we saw in adult mice are probably even stronger in adolescent animals,” said Dr. Eric Kandel.

But are they stronger in people animals or just mice? You don’t really know, do you? I am willing to bet that people of any age are much less inclined than mice to spend their day in a little chamber with cocaine in the water supply. Indeed, if offered the choice, I suspect the mice would have left too.

So who is to blame for this fiasco? The authors, obviously. They were clearly on a mission to trump up support for personal pet theories. Columbia University, no doubt. They are the one who sent out the shark-jumping press release.

What about the New England Journal of Medicine? That one is trickier, since this was not just a research paper, but the presentation (with lots of extra material) of the content of an honorary lecture. In some sense, this magazine was just acting as a typical magazine, merely publishing the thoughts of a couple of authors as-is. But they do have the word “Journal” in their name and normally claim to be a respected peer-reviewed publication. (Unjustly respected, it turns out. They were always my go-to source when I wanted to quickly find a teaching example of bad epidemiology.) There were no caveats accompanying this article to suggest it was not reviewed according to their normal methods and standards. Thus the journal and those who reviewed this paper also bear responsibility for allowing a thinly veiled political statement to be published as if the claims in it were supported by the reported study.