by Carl V Phillips
[This is a continuation of yesterday's post on this topic and will probably not make much sense if you have not read that one.]
I will pick up the analysis with a thoughts that came up in the comments (h/t to the comment by “mav” in particular): On the topic of whether Glantz is a liar, it is possible that he genuinely believes his absurd claims about how mere imagery of smoking is what causes people to benefit from nicotine/tobacco, and that he might have genuinely humanistic concern about the harm such images would cause. (I.e., on that count, he is not the calculated corporate liar that dominates the tobacco control industry leadership.) But then, it might also be that he is so worried about e-cigarettes creating such imagery (despite most of the decent e-cigarettes not looking much like cigarettes) that he is motivated to lie about e-cigarettes specifically.
There was also the observation that a single reported example of an exploding e-cigarette battery or contamination of the e-cigarette liquid is considered by the anti-harm-reduction activists to be true and worth repeatedly reporting. So obviously they do understand the concept that a single observation is often adequate scientific evidence of a particular claim. This argues in favor of the explanation that they are consciously lying when they deny the usefulness of personal success stories of how e-cigarettes caused someone to quit smoking. If a personal testimonial of an exploding battery constitutes evidence, then countless testimonials about smoking cessation certainly do.
One of the reasons that both of these are so compelling is the plausibility of the claim. Batteries (in all devices, whether they cost $5 or $200 million) do occasionally fail catastrophically, especially if someone uses the wrong kind. Thus, it is not at all surprising that a few batteries have burned or exploded. The e-cigarette is purpose-built for smoking cessation. Thus, it is not at all surprising that e-cigarettes have caused lots of people to quit smoking. There are many other examples of how to use prior knowledge in our scientific reasoning: Since nicotine is derived from tobacco, we would expect to find a miniscule (inconsequential) trace of other molecules that occur in plants — e.g., TSNAs — in the nicotine that is used in NRTs or e-cigarettes, and indeed it has been found. This contrasts with, say, a study of e-cigarette vapor that finds tobacco smoke combustion products that we do not expect to find; in that case, the explanation is probably sloppy lab procedure and equipment that was contaminated by previous analyses of smoke.
So for something that is easily predicted to be true, apart from the empirical evidence, we only need a little bit of empirical evidence to convince us that it is indeed true. That evidence is valuable, though, no matter how strong the theory is. So, recent efforts by smokeless tobacco manufacturers to bring new products to the American market have almost certainly caused some smokers to quit, but absent a bit of evidence to confirm success, it is not unreasonable to consider that conclusion less than definitive. This relates back to the light switch point from the previous post: We have strong reason to believe that the switch on the wall and not, say, the fact that it is 8:00 is what causes the lights to be on, so when our switch-flipping experiment seems to confirm that, we can be confident of the claim.
This does bring up the one bit of possible doubt about personal experiments: What if someone is quite sure about a particular cause and effect, but they are wrong? You only need to wander through the internet for a few minutes to find examples where this is pretty clearly the case. But how could someone possibly be wrong about their smoking cessation? Glantz tries to make up such a story by claiming “placebo effect”.
Glantz is in good company in clearly not understanding what this concept even means. Most people who talk about placebo effects or placebo controls clearly do not understand what they are talking about. So to explain…
The concept of a placebo, as part of a scientific inquiry, only makes sense with specific reference to context, and needs to be separate from similar but quite distinct effects. When a placebo is referred to without a research context, it generally refers to an actual treatment method, in which someone is cured of a disease by intentionally tricking them into believing they are receiving a treatment with known benefits. This is a good thing when it works, obviously. To the extent that the word is interpreted this way, then for someone to quit smoking due to the placebo effects of e-cigarettes means that they quit smoking due to e-cigarettes. Success! (Michael Siegel wrote more about this observation yesterday.)
But that brings up the issue of what the word means in a research context, and that, rather than the intentional medical use of a placebo, is presumably what Glantz was referring to. What would it mean for an e-cigarette to work as a placebo, rather than because… well, rather than what? This points out that the concept of an experimental placebo has to refer to a specific characteristic of an intervention, and not the intervention in general. Someone might wonder if the nicotine delivery from an e-cigarette really matters much, or whether the device and action itself is causing most of the effect — i.e., whether the nicotine content was merely a placebo. In that case, the placebo control would consist of a nicotine-free e-cigarette that the subject did not know was nicotine-free. The placebo would still be something real, as most placebos are, but not real in one specific sense. But this would obviously not constitute a placebo if the question were “is the act of mimicking smoking useful for cessation?” The concept of a placebo in research only makes sense when we consider the question that is being asked.
To further complicate it: In clinical studies where some subjects are just given a sugar pill (Glantz’s example of what a placebo is), there is perhaps some placebo effect. However, this is actually probably dwarfed by the “Hawthorne effect”, the tendency of people to behave differently just because they know they are being studied, regardless of whether anything is being done to them. (Ironically, the Hawthorne experiment that it is named after probably actually demonstrated as much of a placebo effect as anything, but that is another story. The concept is clearly right even though the name is wrong.) When smoking cessation studies are performed, and it is found (as is pretty much always the case) that the NRT or other intervention had the same effect as a placebo, most of the effect is probably Hawthorne and not placebo. That is, the cessation success rate is elevated for both groups not so much because of a placebo effect, but because people who are being studied are more likely to behave in different ways.
Indeed, I have long believed that the following intervention would have approximately the same success rate as NRT in clinical trials (or the placebos that have the same effect): Have a reasonably attractive member of the opposite sex approach a smoker and say “I am from <organization> and we are very interested in finding out how many people are quitting smoking this year and how they are doing it. Would you mind if I get contact information for you and check back in six months to see if you have quit?” In both that scenario and in the real clinical studies, extra cessation (in excess of the population average rate) would mostly result from people who had been seriously thinking about quitting one of these days, and who — because they know that someone is watching them to see if it happens right now — go ahead and do it. (We were actually considering doing that experiment when I was in Alberta, but did not manage to get it started.)
It should be obvious that both placebo and Hawthorne effects are much more likely when the outcome of interest is decision-based rather than biological. That is, occasionally someone actually overcomes biological disease due to the power of belief, but it is much more likely for someone to choose an action because of their belief. Indeed, people only take volitional action because of belief. So talking generically of a placebo (rather than specifically, as with the nicotine-free e-cigarettes) when the outcome is behavioral actually makes no sense. It seems likely that Glantz was confusing the Hawthorne effect, which occurs in study settings, with the placebo effect. He was probably trying to suggest that since the placebo group in clinical trials quit at the same rate of those who received “FDA approved”-type interventions, that maybe that was also occurring with e-cigarettes. But since he was talking about people making personal decisions in their normal lives and not in a study, there was probably almost no Hawthorne effect (there might have been a bit, if a friend who turned someone on to e-cigarettes was monitoring their success, but not much).
As I said, Glantz is in good company in not understanding this. Well, I should say he is in a lot of company — the widespread failure to understand this among people who claim to be experts is not really forgivable. When someone refers to a placebo effect, without actually identifying what specific aspect of an intervention is being replaced with a placebo, it is yet another example of someone understanding just enough about how to do science to create confusion rather than knowledge.