The Scientific Scandal of Antismoking
J. R. Johnstone, PhD (Monash)
P.D.Finch, Emeritus Professor of Mathematical Statistics (Monash)
Science is not always a neutral, disinterested search for knowledge, although it may often seem that way to the outsider. Sometimes the story can be very different.
Smoking and health have been the subject of argument since tobacco was introduced to Europe in the sixteenth century. King James I was a pioneer antismoker. In 1604 he declared that smoking was “a custome lothsome to the eye, hatefull to the Nose, harmefull to the braine, dangerous to the Lungs, and in the blacke stinking fume thereof, neerest resembling the horrible Stigian smoke of the pit that is bottomelesse.” But like many a politician since, he decided that taxing tobacco was a more sensible option than banning it.
By the end of the century general opinion had changed. The Royal College of Physicians of London promoted smoking for its benefits to health and advised which brands were best. Smoking was compulsory in schools. An Eton schoolboy later recalled that “he was never whipped so much in his life as he was one morning for not smoking”. As recently as 1942 Price’s textbook of medicine recommended smoking to relieve asthma.
These strong opinions for and against smoking were not supported by much evidence either way until 1950 when Richard Doll and Bradford Hill showed that smokers seemed more likely to develop lung cancer. A campaign was begun to limit smoking. But Sir Ronald Fisher, arguably the greatest statistician of the 20th century, had noticed a bizarre anomaly in their results. Doll and Hill had asked their subjects if they inhaled. Fisher showed that men who inhaled were significantly less likely to develop lung cancer than non-inhalers. As Fisher said, “even equality would be a fair knock-out for the theory that smoke in the lung causes cancer.”
Doll and Hill decided to follow their preliminary work with a much larger and protracted study. British doctors were asked to take part as subjects. 40.000 volunteered and 20,000 refused. The relative health of smokers, nonsmokers and particularly ex-smokers would be compared over the course of future years. In this trial smokers would no longer be asked whether they inhaled, in spite of the earlier result. Fisher commented: “I suppose the subject of inhaling had become distasteful to the research workers, and they just wanted to hear as little about inhaling as possible”. And: “Should not these workers have let the world know not only that they had discovered the cause of lung cancer (cigarettes) but also that they had discovered the means of its prevention (inhaling cigarette smoke)? How had the MRC [Medical Research Council] the heart to withhold this information from the thousands who would otherwise die of lung cancer?”
Five year’s later, in 1964, Doll and Hill responded to this damning criticism. They did not explain why they had withdrawn the question about inhaling. Instead they complained that Fisher had not examined their more recent results but they agreed their results were mystifying. Fisher had died 2 years earlier and could not reply.
This refusal to consider conflicting evidence is the negation of the scientific method. It has been the hallmark of fifty years of antismoking propaganda and what with good reason may well be described as one of the greatest scandals in 500 years of modern science.
A typical example of such deception appeared in the same year from the American Surgeon General. This was “Smoking and Health”,
the first of many reports on smoking and health to be produced by his office over the next 40 years. It declared that in the Doll and Hill study “…no difference in the proportion of smokers inhaling was found among male and female cases and controls.” Fisher had shown this was not so. Fisher’s assessment and criticism of the Doll and Hill results is not mentioned, not even to be rejected. Unwelcome results are not merely considered and rejected. They cease to exist.
The work of Doll and Hill was continued and followed up over the next 50 years. They reintroduced the question about inhaling. Their results continued to show the inhaling/noninhaling paradox. In spite of this defect their work was to become the keystone of the modern anti-smoking movement: Defects count for nothing if they are never considered by those who are appointed to assess the evidence.
But their work had a far more serious and crippling disability.
From its inception the British doctors study was known to have a critical weakness. Its subjects were not selected randomly by the investigators but had decided for themselves to be smokers, nonsmokers or ex-smokers. The kind of error that can result from such non-random selection was well demonstrated during the 1948 US presidential election. Opinion polls showed that Dewey would win by a landslide from Truman. Yet Truman won. He was famously photographed holding a newspaper with a headline declaring Dewey the winner. The pollsters had got it wrong by doing a telephone poll which at that time would have targeted the wealthier voters. The majority of telephone owners may have supported Dewey but those without telephones had not. A true sample of the population had not been obtained.
The new Doll and Hill study was subject to a similar error. Smokers who became ex-smokers might have done so because they were ill and hoped quitting would improve them. Alternatively, they might quit because they were exceptionally healthy and hoped to remain so. Quitting could appear either harmful or beneficial. To avoid this source of error another project, the Whitehall study, was begun.
In 1968 fourteen hundred British civil servants, all smokers, were divided into two similar groups. Half were encouraged and counselled to quit smoking. These formed the test group. The others, the control group, were left to their own devices. For ten years both groups were monitored with respect to their health and smoking status.
Such a study is known as a randomised controlled intervention trial. It has become increasingly the benchmark, or as it is often referred to, the “gold standard” of medical investigation. Any week you can open The Lancet or British Medical Journal and you will likely find an example of such a trial to determine the benefits or harm of some new therapy. Such trials are fundamentally different to that of Doll and Hill. This is ironic because Hill had published the influential and much-reprinted textbook “Principles of Medical Statistics” where he considers the relative merits of controlled and uncontrolled trials. His praise is reserved for the former. Of the latter he is particularly critical: Such work uses “second-best” or “inferior” methods. “The same objections must be made to the contrasting in a trial of volunteers for a treatment with those who do not volunteer, or in everyday life between those who accept and those who refuse. There can be no knowledge that such groups are comparable; and the onus lies wholly, it may justly be maintained, upon the experimenter to prove that they are comparable, before his results can be accepted.” This criticism by Hill can accurately be applied to the Doll and Hill study. According to Hill’s own criteria, his work with Doll can only be described as second-rate, inferior work. It would be for others to conduct properly controlled trials.
So what were the results of the Whitehall study? They were contrary to all expectation. The quit group showed no improvement in life expectancy. Nor was there any change in the death rates due to heart disease, lung cancer, or any other cause with one exception: certain other cancers were more than twice as common in the quit group. Later, after twenty years there was still no benefit in life expectancy for the quit group.
Over the next decade the results of other similar trials appeared. It had been argued that if an improvement in one life-style factor, smoking, were of benefit, then an improvement in several – eg smoking, diet and exercise – should produce even clearer benefits. And so appeared the results of the whimsically acronymed Multiple Risk Factor Intervention Trial or MRFIT, with its 12,886 American subjects. Similarly, in Europe 60,881 subjects in four countries took part in the WHO Collaborative Trial. In Sweden the Goteborg study had 30,022 subjects. These were enormously expensive, wide-spread and time-consuming experiments. In all, there were 6 such trials with a total of over a hundred thousand subjects each engaged for an average of 7.4 years, a grand total of nearly 800,000 subject-years. The results of all were uniform, forthright and unequivocal: giving up smoking, even when fortified by improved diet and exercise, produced no increase in life expectancy. Nor was there any change in the death rate for heart disease or for cancer. A decade of expensive and protracted research had produced a quite unexpected result.
During this same period, in America, the Surgeon General had been issuing a number of publications about smoking and health. In 1982, before the final results of the Whitehall study had been published, the then Surgeon General C. Everett Koop had praised the study for “pointing up the positive consequences of smoking in a positive manner”. But now for nearly ten years he fell silent on the subject and there was no further mention of the Whitehall study nor of the other six studies, though thousands of pages on the dangers of smoking issued from his office. For example in 1989 there appeared “Reducing the Health Consequences of Smoking: 25 Years of Progress”. This weighty work is long on advice about the benefits of giving up smoking but short on discussion of the very studies which should allow the evaluation of that advice: you will look in vain through the thousand references to scientific papers for any mention of the Whitehall study or most of the other six quit studies. Only the MRFIT study is mentioned, and then falsely:
“The MRFIT study shows that smoking status and number of cigarettes smoked per day have remained powerful predictors for total mortality and the development of CHD [coronary heart disease], stroke, cancer, and COPD [chronic obstructive pulmonary disease]. In the study population, there were an estimated 2,249 (29 percent) excess deaths due to smoking, of which 35 percent were from CHD and 21 percent from lung cancer. The nonsmoker-former smoker group had 30 percent fewer total cancers than the smoking group over the 6-year follow up.”
This was untrue, as the Surgeon General was later to admit.
What the MRFIT authors themselves had to say about their work was quite different:
“In conclusion we have shown that it is possible to apply an intensive long-term intervention program against three coronary risk factors with considerable success in terms of risk factor changes. The overall results do not show a beneficial effect on CHD or total mortality from this multifactor intervention.” (Multiple Risk Factor Intervention Trial Research Group, 1982)
But in 1990 the Surgeon General published “The Health Benefits of Smoking Cessation” and at last the subject was addressed. The Whitehall study was rejected because of its “small size”. A once praiseworthy study had become blameworthy. The MRFIT results were described, this time truthfully: “there was no difference in total mortality between the special intervention [quit] and usual care groups.” This and the other studies were rejected because the combined change in other factors – eg diet and exercise – made it impossible to apportion benefit due to smoking alone. This is absurd and illogical reasoning. If, say, a 10% improvement in life expectancy had been found then it might indeed be difficult if not impossible to say how much was due to smoking alone. But there was no improvement. There was nothing to apportion. Nevertheless, with such deceptive words the Surgeon General turned to an unpublished, unreviewed, un-controlled, non-intervention, non-randomised survey conducted for the American Cancer Society (“American Cancer Society: Unpublished tabulations”). The gold standard of modern science was rejected and replaced by the debased currency of what is by comparison little better than opinion and gossip.
This rejection of consistent results from controlled trials and the acceptance of far inferior data would not be countenanced in any other area of medical science. Anyone who suggested doing so would be met with howls of derision and questions as to their intelligence if not their sanity. But where smoking and health are being considered this debasement of science is commonplace and passes without comment.
In Australia in the same year there appeared a similar publication “The Quantification of Drug Caused (sic) Mortality and Morbidity in Australia” from the Federal Department of Community Services and Health. Its authors waste no time in discussing intervention trials. These receive not a mention, not even to be rejected. Instead the authors turned to several surveys of the kind ultimately used by the Surgeon General. In particular they used yet another study conducted for the American Cancer Society by E.C.Hammond, a gigantic study of a million subjects, another uncontrolled, non-intervention, non-randomised survey. This was a particularly bad choice. The dangers of very large surveys are well known to statisticians: because of their size it is difficult to do them accurately. The flaws in Hammond’s work were revealed when the initial results were published in 1954. Hammond himself was later to admit that his study had not been conducted as he had intended and as a consequence his results are to an unknown extent erroneous. But it was worse than that. His work became literally a textbook example of how not to do research. It can be found as example 287 in “Statistics A New Approach” by W.A.Wallis and H.V.Roberts. This was the ignominious and undignified fate of work which should only be quoted as a salutary example of the pitfalls which can await the researcher.
Two problems bedevil both Hammond’s work and other similar studies.
First, some of the volunteers who enrolled their subjects told Hammond that contrary to his instructions they had selectively targeted ill smokers. These results he was able to scrap but necessarily an unknown proportion of his final results must be suspect. Second, as was demonstrated at the time, his subjects were quite unrepresentative of the general public in a number of respects. In particular, there were relatively few smokers. It seems quite plausible that many healthy if indignant smokers would refuse to take part in his trial and this would produce such an aberration. These two vitiating defects are of the kind which have led to the widespread preference for gold standard trials.
But the continuation of Hammond’s work, with its demonstrated
faulty methodology, was used by the Australian authors to deduce that smoking causes premature death to the extent of 17,800 per year in Australia. Their conclusions should be compared with the results of a survey by the Australian Statistician in 1991 of 22,200 households, chosen at random. This showed “long term conditions”, including cancer and heart disease, to be more common in non-smokers than smokers.
Even if they had used sound data to calculate deaths caused by smoking, this still would not have shown that smoking is overall harmful or causes an excess of deaths. Antibiotics kill some susceptible, allergic individuals but this fact does not show that antibiotics reduce life expectancy. If the data used by these authors is examined more closely it can in fact be shown that the mean age at death from smoking-related causes (eg lung cancer) is about 1 year greater than from nonsmoking-related causes (eg tetanus). See: http://members.iinet.net.au/~ray/finch2.pdf for details. This result does not necessarily show that smokers live longer than nonsmokers: smokers as well as nonsmokers die from both nonsmoking-related causes and smoking-related causes. But it is certainly not evidence for the belief that smoking reduces life expectancy.
During all this time health authorities have repeatedly and persistently lied to the public. Consider just one of innumerable examples. In June 1988, in Western Australia the Health Department in full page advertisements in local papers declared: “The statistics are frightening. Smoking will kill almost 700 women in Western Australia this year. If present trends continue, lung cancer will soon overtake breast cancer as the most common malignant cancer in women”. What was frightening was not the statistics but the fact that a Health Department should lie about them. In 1987 the same Health Department in its own publications had said: “Suggestions by some commentators that lung cancer deaths in women will overtake breast cancer deaths in the next few years look increasingly unlikely…female lung cancer death rates have fallen for the last 2 years.” It was predicted that breast cancer would far outweigh lung cancer for the next 14 years. What the public were told was not just an untruth but the reverse of the truth. This is classic Orwellian Newspeak. The public are given what George Orwell in “1984″ named “prolefeed” – lies. Orwell must have smiled wryly in his grave.
Above all has been the repeated and world-wide directive that smokers should quit and live longer when every controlled trial without exception has demonstrated this claim to be false.
Is there anything that can be said with certainty about the health and life expectancy of smokers and non-smokers? The evidence indicates little difference. One important fact often causes confusion: an agent can be a certain cause of death and yet have the effect of extending life. Smoking could be a major cause of lung cancer or even the only cause yet also be associated with long life. The Japanese are amongst the heaviest smokers in the world. They also live the longest. The Frenchwoman Jeanne Calment smoked for a hundred years before dying at 122 as the world’s oldest ever person.
The resolution of this paradox lies in the simple fact that most agents have both good and bad effects on health and life expectancy and it is the net result which is of primary importance. This simple but crucial fact is often ignored or forgotten by medical researchers. Coffee causes pancreatic cancer says the newspaper article. Perhaps it does, but if it has a bigger and beneficial effect on heart disease then those who drink coffee may well live longer than those who don’t. Hormone replacement therapy may increase the incidence of certain cancers yet still have overall a beneficial effect. (See “The Contrapuntists”).
It may now be apparent why there is such a general belief that smoking is dangerously harmful. There are 3 reasons. First, studies which in any other area of science would be rejected as second-rate and inferior but which support antismoking are accepted as first-rate. Second, studies which are conducted according to orthodox and rigorous design but which do not support the idea that smoking is harmful are not merely ignored but suppressed. Third, authorities who are duty-bound to represent the truth have failed to do so and have presented not just untruths but the reverse of the truth.
It may be argued that this is news about an old and settled subject. And who cares about smoking anyway. But smoking is really a secondary issue. The primary issue is the integrity of science. This has no use-by date. When the processes of science are misused, even if for what seems a good reason, science and its practitioners are alike degraded.
In a few years time an accidental by-product of genetic engineering leads to the discovery that certain living vibrating crystals can be manufactured very cheaply. When encased in a suitable holder and inserted in the ear one can hear, just for a few minutes, until body heat kills the crystal, beautiful melodies, rhythms and fascinating counterpoint. They are marketed as aural contrapuntive devices. Since they are cheap and become very popular, the Government taxes them. Users of the device become known as contrapuntists.
Some years later a new disease is identified when an increasing number of people drop dead, suddenly, for no apparent reason. Autopsies reveal a strange deterioration in the brain cells of those affected. An observant pathologist notes that in most of the associated post-mortem examinations an aural contrapuntive device was found in an ear of the deceased and the disease becomes known as SADS, an acronym for Sudden Aural Death Syndrome. Epidemiologists find that people who are not contrapuntists seldom fall victim to SADS and that, in fact, about 98 per cent of all such deaths are either current or former contrapuntists. The strength of association between aural contrapuntism and SADS is undeniable, the relative risk is as high as 50, i.e. a contrapuntist has about 50 times the chance of falling to SADS as does a non-contrapuntist.
An anti-contrapuntist health campaign is initiated and aural contrapuntive devices are taxed more and more heavily in an attempt to dissuade people from using them. The campaign is very successful and is vigorously supported by an unexpected alliance between animal liberationists, the music industry and the tone-deaf. Attention then shifts to passive aural contrapuntism, viz. the dangers posed by the sidestream melodic overflow from the devices in the ears of contrapuntists, in particular on the occurrence of SADS in non-contrapuntal spouses of contrapuntal men, the harm contrapuntal parents may do their children and the possible ill-effects suffered by the foetus of a contrapuntal pregnant woman.
After great initial success, however, the campaign falters when it becomes widely known that even though aural contrapuntism is so strongly associated with SADS, relatively few contrapuntists die from it each year and those that do have lived, on average, about one year longer than do non-contrapuntists and, moreover, at each age, are much more likely to die of other causes than of SADS itself. Politicians realise very quickly that they can now, with a clear conscience and with profit, tax aural contrapuntal devices even more heavily.
The above is reproduced from http://members.iinet.com.au/~ray/TSSOASb.html