Article Text
Statistics from Altmetric.com
As previous issues of Tobacco Control have reported, BAT has long been putting out disinformation by the truck-load. For example, it sets up special seminars to reach journalists from developing countries, no doubt viewed as potentially more easily persuaded of its propaganda than those from countries where tobacco’s number is up. And it has certainly played its full part in getting the tobacco industry the sort of name for scientific integrity that, say, a pit full of vipers has for personal safety. But such activities have usually been undertaken in a situation where it has relatively firm control. Recently, however, in its home territory of the United Kingdom it has stepped out of the sidelines and put its foot straight into, well, onto the public stage. In short, it has tried some up-front, pro-active media initiatives.
First came an extraordinary attempt to set the record straight after damaging publicity in British newspapers following some of the gems scattered to the world’s news media from the Minnesota litigation, in the form of internal documents from tobacco companies. Stung by revelations about high-nicotine tobacco grown in Brazil for Brown and Williamson (B&W), BAT’s American subsidiary, and about BAT having accepted 20 years ago that tobacco was addictive, the company took out a five-column advertisement in The Observer, one of Britain’s leading serious weekend newspapers. Breezily headed “Smoking Gun?”, it was a statement by Dr Chris “Ground-breaking” Proctor (see Tobacco Control1996;5:262–3), these days billed as head of science and regulation.
Space for the 1200 or so words (probably unconvincing to virtually every reader) was presumably bought on the assumption (probably fair) that the newspaper would never publish a letter even a fraction of the size of an advertisement, where it could get its propaganda printed in full, and without the inconvenience of damning comments from health advocates. Among its defences to numerous charges was an appeal to readers as to whether it was “really so surprising that in one paragraph in an ancient six page document, the question was merely raised whether we should make a cigarette with a small amount of marijuana in it if that substance should become legal?”
Elsewhere, Proctor’s essay covered the familiar, if now somewhat superseded industry line on addiction. At least here it was consistent with company policy—B&W was the only big American company whose boss earlier this year broke ranks with his peers when they admitted their belief that smoking, after all, might be addictive. Here was vintage industry stuff: mentions of addiction being an emotional subject and a colloquial term with a much broader definition than the purely scientific—why, we even had the term “addicted to love” in a movie. No, said Proctor, this colloquial definition applied to many common substances that have “similar mild pharmacological effects to cigarettes, such as coffee, tea, chocolate and cola drinks.” No need for smoking cessation clinics, then, unless they have them for cola drinkers.
Bolder by far, however, was BAT’s outrageous hijack of an unborn report from the International Agency on Research in Cancer (IARC), an offshoot of the World Health Organisation (WHO) based in Lyon, France. A week after its Observer ad, BAT managed to convince another British Sunday newspaper, albeit one renowned for publishing the most reactionary “butter-stops-heart-attacks” type of story, to run a massive front-page article with the headline: “Passive smoking doesn’t cause cancer—it’s official”, and to tell its readers that “The world’s leading health organisation has withheld from publication a study which shows that not only might there be no link between passive smoking and cancer but that it could even have a protective effect” (Sunday Telegraph, 8 March 1998). In case its readers had missed the point, the newspaper also carried an editorial, which could not have been drafted better by BAT itself, under the headline: “A setback for nanny”. Versions of the story, which quoted Proctor, appeared all round the world, apparently facilitated by BAT’s international public relations machine. The initiative was no doubt timed to try to pre-empt the effects of a report assessing the risks of passive smoking which was due to be published the following week by the Scientific Committee on Tobacco and Health (SCOTH), an independent committee that advises the British government on scientific aspects of tobacco. A major part of the evidence reviewed by SCOTH was a study it had commissioned from British epidemiologists, and published last October in the British Medical Journal. 1
In fact, the WHO-IARC study leaked to the Sunday Telegraphwas undergoing peer review before being published in an academic journal, and thus could not be described as “withheld”. The study’s figures for relative risk (or odds ratio) of a non-smoker contracting lung cancer as a result of living with a smoking spouse or working in a smoky workplace were: spouse smokes—1.16; smoky workplace—1.17. So how did BAT and the newspaper manage to turn this upside down? The error (or deception) was to misinterpret (or misrepresent) a statistical test applied to the results, as explained by Clive Bates, the director of Action on Smoking and Health, the UK’s leading tobacco control agency, in the item reprinted in the adjoining box.
How the WHO-IARC study was misinterpreted by BAT and the Sunday Telegraph
Because the estimate of risk (“relative risk” or “odds ratio”) is based on a sample of 650 lung cancer cases, the risk in the whole population might be different because the sample may not be exactly representative. So the statisticians use a “confidence interval”. This allows them to give the central estimates—spouse smokes, 1.16; smoky work place, 1.17—and then say, in effect, “We are 95% confident that the real value for the whole population lies between the limits x andy” (see below). In other words, the chance that the actual risk lies outside this range is 1 in 20, or 5%. The table shows the limits of the WHO study.
The fact that the lower limits drop below 1.0 shows that the statisticians cannot be 95% confident that the survey has detected a link between passive smoking and lung cancer. In other words there is a greater than 5% probability of obtaining the central estimates of 1.16 and 1.17 by the play of chance. This is what statisticians mean when they say the result is “not statistically significant”—they cannot be 95% certain that they have detected a link between passive smoking and lung cancer. This could happen because there is no real effect, or because a small sample size reduces the “statistical power” to detect a real effect. However, this uncertainty was inverted in the story covered by the media and was reported as evidence of “no effect”. The tobacco industry has translated this formal statistical meaning of the word “significance” into lay language, to give the meaning “the study shows the risk is insignificant”. Furthermore, because the lower limit is 0.93, it was translated to a possible “protective effect”. Of course, the study no more shows a protective effect than it shows a 44% increase in risk—the other extreme of the confidence interval.
This is an outrageous misinterpretation of the results and it is difficult to know if this was naivete on the part of the Sunday Telegraph or manipulation by BAT, who should know better, or both. Also included above for comparison are the figures from a major review of the evidence published in theBMJ last October1 The table and figure show that the ranges overlap and therefore that the results are consistent. The results of the BMJ study have a much smaller confidence interval, because the use of several studies in a “meta-analysis” increased the overall sample size, and therefore the “pooled” sample is more likely to be representative.
The WHO swiftly hit back at BAT and the newspaper, unleashing an unusually strongly worded press release headed: “Passive smoking does cause lung cancer; do not let them fool you”. It said the reporting of the findings was “false and misleading” and “From these and other previous reviews of the scientific evidence emerges a clear global scientific consensus—passive smoking does cause lung cancer and other diseases” (WHO press release, 9 March 1998). British media probably gave more coverage in total to this response, and to other denunciations of BAT’s role, than had been gained by the original story. But no doubt this will not deter BAT from similar efforts in the future, because they generate the sort of press coverage it wants. And in other countries, especially those which most matter, BAT is more likely to be the net winner. In Brazil, for example, where BAT’s Souza Cruz subsidiary holds sway, massive anti-health hype prevailed in the week following the original British newspaper article before health advocates could establish what had happened and take corrective action.
As this issue of Tobacco Control was going to press, ASH was awaiting a judgment from the Press Complaints Commission (PCC), the British newspaper industry’s self-regulation body. ASH asserted that the original Sunday Telegraph articles were false and misleading, and that far from withdrawing them, the newspaper made a number of unfounded accusations against its critics in a further article and leader a week later, inexplicably arguing that it had been right all along.
What seems most likely is that the Sunday Telegraph’shealth correspondent had little understanding of scientific methodology or statistics, and was simply led by the nose by Proctor; and that BAT’s public relations people did the rest. We await the PCC’s verdict with interest; but it is doubtful this will be the last time we report this sort of trick by BAT.