Correction by the author Carl Phillips, TobaccoHarmReduciton.org 3 July 2010 It has been called to my attention that one of my conclusions does not match the calculation that is presented. In particular, the statement in the abstract, “for the average smoker, smoking for just one more month before quitting causes greater health risk than switching to a low-risk nicotine source” should read either “for a smoker who is doomed to die from smoking if he does not quit, smoking for just one more month….” or “for the average smoker, smoking for just a few more months….”. The latter is probably better as the take-away message from the analysis. The explanation is that I ran two different calculations when carrying out this analysis, one for the average smoker and one for those smokers who are doomed to die from smoking if they do not quit. The advantage of looking at the former is that it describes an identifiable characteristic; the advantage of the latter is that it distills the analysis to those smokers we are most concerned about. Obviously the average smoker is at less risk from smoking another month than the doomed-but-for-quitting smoker, since the population of non-doomed smokers dilutes the average risk (by a factor of 2, 2.5, 3, or whatever the reciprocal of one’s estimate of the proportion of lifetime smokers who die from their habit). In the published version of the paper I ended up including only the calculation for the doomed smokers, but erred by describing it as applying to the average smoker. Had I observed such a switch in an analysis that was intended to attack tobacco harm reduction (THR), or something similar in an empirical study to spin it as anti-THR, I would have characterized it as certainly misleading and probably dishonest. I hope I can avoid the latter accusation for the following reasons: (a) I am highlighting the error and voluntarily running this correction with contrition (I do not recall anti-THR activists ever doing such a thing despite the enormous flaws that have been identified in some of their articles); and (b) When I have previously cited my results, including in a press release about the article, I believe I always described the break-even point as “a few months” or something similar, the accurate result I had in my head but did not correctly put on paper (which contrasts with anti-THR activists who often make claims to the public that are far stronger than what their studies support). Frankly, I doubt the quantitative error really matters to anyone. It should not. Is anyone really going to think “you had me sold on promoting THR when you said that smoking for only one more month is just as bad as a lifetime of a low-risk alternative, but if it is two or three months then forget it – let them keep smoking”? This makes it no less an error, but it means that it changes nothing about the practical implications of the paper. Always the teacher and social scientist, I cannot resist drawing two interesting conclusions from this error. I wish I could say that I planted it as an Easter Egg to teach these lessons, but I might as well take advantage of it anyway. First, this is a great reminder that the status “peer-reviewed publication” is far from sufficient to conclude that something is correct. A version of this paper that included the error was read and checked many times over many months before being published (of course, I do not blame referees or anyone else who reviewed this for my error). We can only guess how often there are comparable errors in the health science papers where it is impossible for peer reviewers, editors, or readers to check the authors’ work (the 99% of papers where, unlike this one, the authors do not report enough information to assess the accuracy of the conclusions). Moreover, when confidence intervals or other error statistics are reported (which would have been completely inappropriate for something like this, but is disturbingly common in other cases where it is equally inappropriate), they are based on countless assumptions, many of which are false (e.g., the assumption the author did not switch what he was calculating in the middle of reporting his results). This illustrates how such statistics typically serve to obscure the most important sources of uncertainty, giving a false sense of accuracy to non-expert readers by pretending to quantify the potential inaccuracy. Second, the paper was in a journal (free online) for half a year before anyone pointed it out the problem. While it may seem nice to be able to “get away with it” for that long, the broader implications are disturbing. This correction to an overly-strong pro-THR claim did not come from one of the tens-of-thousands of people whose job descriptions (self-determined or institutionally defined) include advocating against THR. (It was discovered by Peter Lee, who, while not an activist, conducts analyses that have strongly supported the case for THR and was in the process of contributing to our THR yearbook.) No one took advantage of the opportunity to catch the error and use it to undermine the analysis. Since the article is quite damning to both the ethics and quantitative health claims of anti-THR activism, why might the anti-THR activists not care to read it carefully enough to find the error? It would be too glib to suggest that none of them can do the math – it is actually pretty easy and was all laid out for the reader. The only apparent explanation that is also consistent with other evidence is that the anti-tobacco extremists have decided that getting what they want is purely a matter of exercising their immense wealth and political power (a not unreasonable expectation). It is not merely that they adopted a worldview that is something like a religion, and thus no mere scientific argument could ever change their views. (Why bother to read scientific or ethical analysis when you do not care whether you are wrong by either of these standards?) It is that they are completely uninterested in ethical or scientific analyses – whether it be criticizing mine or presenting their own – because they expect to get what they want through the exercise of pure power, and so figure it really makes no difference that the legitimate arguments are arrayed against them. Competing interests I trust that people will get it in this case: Some things matter a lot more, in terms of being real competing interests, than funding. In this case, being the author of the paper that is being corrected clearly matters rather more.