by Carl V Phillips
A recent study by Arefalk et al., published in Circulation, was claimed by the authors, the AHA, and the corporate media to show that continuing snus use after an MI (heart attack) is harmful. In reality – if we are to just blindly interpret the study results, without context or thinking, as the authors imply we should – it shows that among people who have MIs, snus users and even smokers are better off than non-users.
To explain this, start with the letter to the editor Brad Rodu and I submitted to the journal (which, given the integrity typically displayed by health sector journals when it comes to admitting errors and allowing dialog, has a 4% chance of being published). It explains the basic problems and our methodology, but due to word limits does not present the full picture, which I continue below. (Note: Brad expects to post about this too, and I will update with a link when he does. Update: It is now live, here.)
Letter by Rodu and Phillips Regarding Article, “Discontinuation of Smokeless Tobacco and Mortality Risk after Myocardial Infarction”
Arefalk et al. report mortality incidence rates (deaths per 1,000 person-years at risk, PYAR) following a myocardial infarction among Swedish tobacco users who, following the MI, continued to use snus (18.7), quit snus (9.7), continued to smoke (28.4) or quit smoking (13.5) (1). We point out several significant problems with this study.
From a methodological perspective, the authors’ inclusion of women in the estimates is a mistake. The number of women among snus users is so small that the gender control variable is extremely unstable. Any gain in power is more than offset by the resulting potential bias. Moreover, combining men and women smokers assumes that the effects of product use are the same for both, which is not justified. In this and similar studies, stratified analysis by gender is always the more informative approach.
The authors report that the mortality rate for the entire sample was 18.9, which appears to be incorrect. It is based on 812 deaths and 40,370 PYAR, which actually yields 20.1. Alternatively, one of the count numbers may be wrong; for example, the rate is 18.9 if the death count is 765. The error requires correction because it affects the article’s major conclusion about snus and smoking; the reported mortality rate for the entire sample (18.9) is similar to that for continuing snus users (18.7).
Arefalk et al. did not report the mortality rate among nonusers of tobacco, which comprise more than half of the sample. We estimated the nonusers’ rate after subtracting the deaths and PYAR for snus, cigarette, and dual users from the sample totals (our estimate of deaths and PYAR among 934 dual users assumed that they were proportional to the mean of those statistics for snus users and smokers).
The estimated mortality rate among nonusers is 21.4, based on 494 deaths and 20,031 PYAR. This is higher than the rate for continuing snus users (18.7), and far higher than that for snus quitters (9.7). In addition, it is higher than the rate for smoke-quitters (13.5). The only group that fared worse than non-users was continuing smokers (28.4). (If the death count was actually 765, the non-users disadvantage would be reduced but not eliminated.)
Thus, those who quit using tobacco products reduced their mortality risk by half compared to those who continued, but also compared to non-users. By the logic of the article’s conclusions, MI victims are better off if they use tobacco so that they can experience the benefits of quitting. The authors have the data to calculate the exact mortality rate for nonusers and adjust for age (which probably does explain part of the disadvantage among nonusers) and other factors. These additional results might explain how the reported harms from snus use can be reconciled with the apparent advantages of quitting.
1. Arefalk G, Hambraeus K, Lind L, Michaelsson K, Lindahl B, Sundström J. Discontinuation of smokeless tobacco and mortality risk after myocardial infarction. Circulation, published online June 23, 2014. DOI:10.1161/CIRCULATIONAHA.113.007252
Brad Rodu, DDS, Department of Medicine, School of Medicine, University of Louisville
Carl V. Phillips, PhD, Consumer Advocates for Smoke-free Alternatives Association
The bottom line is that the authors tried to spin the results as suggesting that continuing snus use is bad for you after an MI. But those who did not quit snus after the MI had a lower death rate than people who did not use either snus or cigarettes at the time of their MI, and those who used snus but quit were far better off still. A simplistic interpretation of the data, then, is: “You had an MI. But the good news is that you use snus! If you now quit snus, your mortality risk compared to people who do not use snus – and thus do not have the opportunity to quit – is magically lowered by more than half. Moreover, even if you do not quit, your risk is still lower than that of non-users.” The authors really buried the lead on this one – or more accurately, did not report the lead at all.
Of course, I am not saying I believe that MI victims are genuinely better off if they use snus. But it is what these statistics suggest (and it is certainly a possibility). The authors carefully avoided reporting an obviously useful and important statistic, the mortality rate among non-users, because they presumably realized that it would call their anti-snus conclusions into question. And it does. Whatever is happening in this population, it clearly does not support the “snus is bad” message that the authors and other naive and politically-motivated commentators have presented.
There is a glaringly obvious explanation for why people who quit snus or smoking after an MI fare better than those who do not: Those who are generally healthy (except for the recent MI, of course) and expect to recover and live a long time are much more likely to take every possible precaution to minimize their risks. After being advised to give up snus, get proper physical therapy, and eat better, many of them will do it. Meanwhile, those who are sickly, have serious cancer, are depressed, etc. will have little incentive to make lifestyle sacrifices in an attempt to regain long-term health. There was no attempt to control for this obvious confounding in the study.
Of course, this would only explain the better outcomes of the quitters compared to those who continued tobacco product use, and does nothing to explain why all of them (except those who continued to smoke) fared better than the non-users. There are possible explanations for this in the form of statistical artifacts or real effects. But the key observation is that the data clearly do not support the main interpretation that the authors touted, that it is terribly bad for you to keep using snus after an MI.
Another important conclusion that should be drawn from this article is just how bad the peer review process is for medical journals. The out-and-out error we discovered in the reported numbers was not some obscure statistic buried in the depths of the paper, but was the key number reported in the first short paragraph of the results section. Brad immediately noticed that the calculation appeared to be wrong on his first read-through, and it took me fifteen seconds to provide the correct arithmetic while creating the spreadsheet that is the basis for our letter. Yet the vaunted “peer review process” let the error through. Those of us who are familiar with journal peer review understand that – contrary what most people seem to believe – the reviewers and editors cannot check anything that readers cannot check for themselves, because they have access to only the same paper the reader has, not the data or calculations. Thus it is, at best, a fairly weak check. But the failures here are a reminder that reviewers seldom even check what is in the paper.
Even without correcting that error or calculating the mortality rate for non-users, the (incorrect) reported number for the population as a whole can still be compared to rates for people who used snus or cigarettes at the time of their MI. This alone is enough to raise red flags about the analysis and conclusions, since it is still higher than the rates for smokers (averaging together those who quit and those who kept smoking) and for those who kept using snus, and far higher than the rates for those who stopped using either product.
Moreover, the obvious confounding described above is not something that only leading scientific experts understand, but is so well known that it even has a name in medical research: “confounding by indication”. It is a constant source of bias when an action (e.g., a particular medical treatment, or in this case, obeying the advice to quit tobacco) is associated with how healthy someone is. A new experimental surgical technique might have a higher death rate, not because it is worse than the conventional practice, but because it is only tried on people whose condition is particularly dire. Or to take an everyday example, someone slamming on their car brakes is much more likely to crash than someone not doing it; this is because the action indicates that there is a risk of imminent crash, not because it is harmful. But the naive logic from blindly looking at the data the way these authors did would be “never slam on your brakes – it will cause you to crash.” Anyone involved in medical research should understand this common problem, and yet the authors do not so much as acknowledge the possibility of such confounding, and the editors and reviewers let them get away with that.
In sum, the way these results were presented and interpreted is a simple classic piece of anti-tobacco propaganda. Researchers who were both skilled and honest would have tried to address the reasons that MI victims who quit using snus or cigarettes fare far better than those who did not use them in the first place, and even those who did not quit snus still did a bit better (after reporting this was the case instead of hiding it, of course). Researchers who were honest but not so skilled would have just reported that result. But the authors, and the journal editors and reviewers who enabled them, instead endeavored to hide that rather remarkable result and to lie that their results showed that snus was harmful.
[Update: Brad recalled that Arefalk is a serial producer of anti-THR junk. He posted about a previous claim here.]