Wednesday, 2 September 2009

1+1=11: 10,000 deaths (part 4)

As part of the campaign to ban smoking in pubs in England, the British Medical Journal published a number of studies which purported to show that secondhand smoke was a clear and present danger to the health of nonsmokers. None of them stand up to scrutiny and any one of them could have been the subject of this article. In terms of the publicity it generated, however, one sticks out above the rest. It was produced by Konrad Jamrozik, a fanatical anti-smoker from Australia who compared those who disputed the passive smoking theory with medieval anti-Semites who preached that Jews ate Christian babies and stated that smoking should be undertaken "only by consenting adults in private". Jamrozik's study deserves closer attention because his conclusions have been parrotted many times since and were a major  factor in manipulating the eventual 'yes' vote for the smoking ban. As with the other studies covered in this series of articles, his study was flawed by a basic mistakes of mathematics and logic.

Drawing on conclusions made by one epidemiologist - Ichiro Kawachi - Jamrozik assumed relative risks of 1.24 for lung cancer and 1.20 for heart disease for nonsmoking women married to smokers (ie. 24% and 20% increases in risk). He then tripled these figures on the basis that nonsmoking bar staff supposedly had three times as much cotinine in their saliva as nonsmokers married to smokers, thereby turning negligible relative risks into more substantial ones; 1.73, 1.61 and 2.52 for lung cancer, heart disease and stroke respectively. (Cotinine is a bio-marker for nicotine. It is not unique to tobacco and it is not harmful, but it is found in relatively high quantities in the saliva of smokers)

Herein lay the problem. Once trebled, the heart disease risk for nonsmoking barworkers came in line with the known heart disease risk for regular smokers despite it being biologically implausible that secondhand smoke exposure was as damaging to the body as actually smoking. The risk ratio Jamrozik arrived at for stroke was still more unlikely. His figure of 2.52 was substantially higher than that seen in smokers and was completely out of kilter with two studies the BMJ had recently published which found no association at all between secondhand smoke and stroke. Only a few weeks earlier the BMJ had stressed that "cotinine is not associated with lung cancer or other diseases" and that "previous studies have stressed the limitations of cotinine as a biomarker of exposure."

Jamrozik's hypothesis rested on a previous study which had found that nonsmokers married to smokers had a cotinine level of 1.2 ng/ml (nanograms per millitre) while barworkers had a level of 3.65 ng/ml. He assumed this threefold increase in cotinine must result in a threefold increase in risk and adjusted his numbers accordingly.

What he did not mention was that cotinine levels in nonsmokers, whether they worked in a bar or not, were always a tiny fraction of those found in smokers. Pack-a-day smokers consistently showed cotinine levels of at least 300 ng/ml, and often much more than that (1).

Jamrozik multiplied the nonsmoking wives' supposed relative risk of 1.20 for heart disease to arrive at a figure of 1.61 risk for bar-workers ie. a raised risk of 61%. Since numerous studies had shown that smokers themselves had around a 70% increased risk of heart disease, this figure stood out as being unrealistically high and the folly of Jamrozik's approach was writ large if one applied the same kind of mathematics to smokers. Smokers have cotinine levels 250 times higher than those of the nonsmoking wives and therefore should, by Jamrozik's logic, be 250 times more likely to develop heart disease. This would mean they had a heart disease risk of 50.0, making them 5,000% - or fifty times - more likely to die of a heart attack than nonsmokers ie. walking timebombs liable to drop dead at any minute. But we know that smoking raises the risk of heart disease by 70%, 4,930% away from what would be calculated using Jamrozik's model.

Jamrozik's premise was that cotinine provided an accurate measure of risk and that the risk rose or fell at a steady gradien (like James Repace, he never entertained the notion that there is a safe level of exposure to a substance, even when that substance can only be measured in nanograms or pictograms). But that premise could be used to form a very different conclusion. Since the nonsmoking wives have 250 times less cotinine that smokers, and since it is widely assumed that smokers have a 70% increased risk of heart disease, it must follow that the nonsmokers' raised risk is 0.28% (70 divided by 250). Such a figure is less than negligible in both statistical and practical terms.

It goes without saying that, as an ardent anti-smoker with an obvious agenda, Jamrozik did not chose not to go down this path. Still, it is intriguing to imagine a parallel universe in which someone used exactly the same flawed premise to prove a completely different point. It would still be nonsensical, of course, but who would seriously believe that the BMJ would publish such nonsense?

Jamrozik was demonstrably wrong. By extrapolating rates of disease from very low doses of harmless cotinine he had overlooked the fact that the doses involved for both nonsmoking wives and bar-workers were, for all practical purposes, zero. But should anyone be tempted to question his wild mathematical assumptions he had a stubborn response: "Given that authorities on three continents have concluded that passive smoking causes disease in adults, my calculations have a firm foundation"; a reference to the EPA's discredited 1992 report and copycat efforts in Britain and Australia. As a result of his (mis)calculations, he announced that secondhand smoke was killing exactly 7 bar-workers a year from lung cancer, plus 12 from heart disease and 6 from stroke. These figures were feeble, even after being arbitrarily trebled using the method described above, but after adding in estimates for hotel and restaurant workers, Jamrozik came to a combined figure of 54 deaths per year; still feeble, but a convenient estimate for the media since it worked out at about one a week. He then added in the deaths he believed secondhand smoke caused in the home and other workplaces and concluded that: "Adoption of smoke free policies in all workplaces in the United Kingdom might prevent several hundred premature deaths each year."

Jamrozik's study was a gift to the British pro-ban activists and it was a gift that kept on giving. The BBC was sent an abstract of the paper a year before publication, in May 2004, and reported that "Smoking at work kills hundreds". The following March, they used a preliminary copy to report: "Passive smoke killing thousands", and when, a few weeks later, the paper was finally published, a spokeswoman for the British Heart Foundation declared that the evidence against passive smoking was so strong that there was "no room left for scientfic debate." Based on Jamrozik's work, the British Medical Association now gave a figure of 11,000 as their official estimate of passive smoking deaths in the UK, an eleven-fold increase on the figure they had previously touted. Two years earlier, a few UK newspapers had tentatively given a figure of 300 UK deaths from passive smoking. The calculations of Konrad Jamrozik (who, ironically, was credited as being a "professor of evidence based healthcare") led to a figure of 10,000 being quoted for the remainder of the campaign. The rest, as they say, is history.

(1) Robbins et al. 'Cigarette Smoking and Stroke in a Cohort of U.S. Male Physicians', Annals of Internal Medicine, 15/3/94, Volume 120 Issue 6, Pages 458-462

No comments: