The most useful lesson I ever had in my life was at my cooking school. A splendid member of staff who had cooked single-handed for a hospital with 170 staff and patients for three years during the war, lectured us on dietetics. The lecture went thus: “We used to be told that spinach was good for you because it had a lot of iron in it. Now they say it isn’t good for you because the iron is in a form you can’t digest. No doubt in ten years’ time they’ll be saying something else.”
End of lecture on dietetics, and we went on to talk about food, which we much preferred. We have seen many such reversals of opinion: many of us can remember being told to go to work on an egg, and we all remember the great egg hysteria. In the thirties it was axiomatic that a healthy diet included plenty of red meat, butter, milk, and cheese. Bread and potatoes were plain bad. Fibre was something that the unfortunate Africans had to eat.
The point about all this was that it was supported by all kinds of “experts”. (The definition of an expert given by a cynical Central European was: “a boy from the next village”). These experts, as is their custom, had all sorts of figures to support their claims. Now the trouble with statisticians, and their close relatives the epidemiologists (epidemiology is the branch of medicine concerned with the control of epidemics, and since the great epidemics of tuberculosis, small-pox, cholera and the rest have been largely checked, epidemiologists haven’t got so much to do), is that they have, like other experts, their own language.
What is the layman to make of statements like: “the rate for men in this category in 1951 (10.78) is based on the experience of only 127 men and may therefore not be very accurate (95% confidence limits 6.57 and 14.99)”. He will probably seize on “not very accurate”, and decide the whole thing isn’t worth his attention. [1]
Nevertheless he continues to be deluged with figures of this kind, terms like “multifactorial” and “quantitative estimates”, tables of figures that all too often seem to refer to nothing at all, and graphs that would mean just as much to him any other way up.
However, if he wants to have an idea of what is going on it is necessary that he gets to grips with some figures, and it is probably easiest and most reliable to look at figures on deaths, particularly expectation of life. It is not really difficult to decide whether someone is dead or not, and in these days of passports, certificates, and forms of all kinds, ages are generally known.
So let us consider the expectation of life in various nations: Greece, with the world’s highest recorded consumption of cigarettes, has the highest expectation of life in Europe. The highest in the world is in Japan, fifth in cigarette consumption. (These are figures from the World Health Organization and the Economist Book of Vital World Statistics. Other figures given by other bodies may differ slightly). [2 & 3]
In fact five of the world’s heaviest smoking nations are in the list of the ten longest lived (Japan, Greece, Spain, Canada, and Australia). Cuba, with the second highest smoking rate, in spite of having a Communist government, which usually seems to shorten lives, has the third highest life expectancy in the Americas.
It is difficult to tie in these figures with what we are so often told about the dangers of smoking. For instance, that every cigarette smoked shortens the smoker’s life by five and a half minutes. How long would the Greeks or the Japanese, then, live if they didn’t smoke? Would it still be longer than the very low-smoking Norwegians? Or the Indians, with the world’s lowest recorded cigarette smoking rate, and a life-expectancy of fifty- eight?
46% of the nation smoked in 1976. This dropped to 35% in 1986. Yet we are constantly told that smoking-related deaths have gone up: from the figure of 50,000 given by the Royal Society of Medicine in 1984 to 115,000 in 1991, and an abrupt jump to 150,000 in 1992.
Ralph W. Moss, formerly assistant director of public affairs at the Memorial Sloan-Kettering in New York, one of the nation’s most prestigious cancer clinics, tells us that in 1962 about 278,000 Americans died of cancer; in 1988, an estimated 494,000. [4] Reducing the Health Consequences of Smoking, a report to Congress in 1989 [5], says that in 1965 50.2% of men and 31.9% of women in the USA smoked. In 1985 this went down to 33.2% of men and 28% of women. Yet the cancer death-rate rose..We have no reason to disbelieve these figures, since they are not at odds with the figures given in this US Surgeon General’s report of 1989, which gives “smoking-related” deaths in 1965 at 183,000, (150,000 men and 30,000 women), and in 1985 at 337,000. These are very embarrassing figures for those who believe that not smoking is good for you, so they are “adjusted” to allow for population growth and an aging population; thus “smoking-related”, “population-corrected” deaths are estimated at 232,000, 167,000 for men and 67,000 for women.
These may well be trustworthy (on their own terms, of course), since they contradict the message that the report is trying to put across: that the large increase in mortality was because of women’s adoption of smoking, which does not seem to agree with its own figures showing that 4% of women gave up.
There is no doubt that many nervous people are very disturbed about passive smoking, and it is unkind to light up in front of them even on Hampstead Heath in a high wind.
There have been a number of studies on the subject: the best known, or most often quoted, being that of Hirayama in Japan. He found a relationship between smoking husbands and lung cancer in their wives. [6] Unfortunately, he made a mistake in his figures, mistaking an X and an X2. When this was pointed out to him, he dismissed it without replying.
“In any other area of science – or indeed in intellectual discourse generally – this would be enough to negate Hirayama’s contribution”, says J.R. Johnstone. [7]
However, Hirayama added some new information, which showed that lung cancer was commoner for non-smoking unmarried women than for the non-smoking wives of smokers. [8] Also that non-smoking men with smoking wives were as likely to get lung cancer as occasional smokers, if not more so. In fact, a cigarette killed you whether you or someone else smoked it. But active smokers must breathe the same air as passive smokers, so they should get a double dose of the cancer-causing stuff.
It appeared also that Hirayama had made another mistake: in factors of up to 1000 percent. [9] His reply was to the effect that it didn’t make any difference...
Trichopolous gave the results of a study of 40 non-smoking women with lung cancer, comparing them with 149 who did not have it. [10] It seemed from this that non-smoking wives of heavy smokers were more likely to get lung cancer than wives who actively smoked. It seemed, therefore, that passive smoking was more harmful than active smoking...
Trichopolous published further results, with some odd figures that he blamed on a “typing error”, when they were pointed out to him. [11] This was politely described by Johnstone as: “casual”.
The Surgeon-General of the USA and the National Health and Medical Research Council of Australia, nevertheless, accepted these results, quoting only those parts of the studies that confirmed their opinions, and leaving out what did not suit them. [12] This must have been very encouraging for other statisticians who weren’t very good at elementary maths.
Garfinkel et al. found that women who had been exposed to passive smoking for 1-2 hours a day for 25 years were significantly less likely to develop lung cancer. [13]
In general, in the 26 studies discussed in Health, Lifestyle and Environment, 20 have shown no results one way or the other, and in the remainder the results, when not invalidated by the originator not being able to count, were not very striking. It should also be pointed out that except in the quite impossible Hirayama study, numbers of cases were all very small: Kabat & Wynder (1984) 37; Correa et al (1983) 22; Trichopolous (1981) 40, (1983) 77; Humble et al (1987) 28. [14]
It is also worth noting that the National Health and Medical Research Council of Australia misquoted two of the studies, implying that certain results in one case were statistically significant when the authors of it stressed that they were not, and actually claiming that Correa et al. [15] had found a positive trend when they had not and nowhere said so.
On the subject of surveys, there have been a number carried out world-wide over the years. Probably the best known, or the most quoted, is the Multiple Risk Factor Intervention Trial (MRFIT). 12,866 men in the USA took part, over a period of some seven years. They were all thought to be at risk (to show "risk factors") from Coronary Heart Disease because of their life-style and their general health. Half were assigned to a special intervention group, treated with drugs for their high blood pressure, and encouraged to stop smoking and eat more healthily. The others were allowed to carry on as usual.
At the end of the follow-up period, 41.2 per thousand of the first group (the intervention group, doing all the healthy things) were dead, as against 40.4 per thousand of the others.
The investigators didn’t like these results, and investigated further. They found that the drugs to reduce high blood pressure had in fact increased the death rate among the men given them.
They were forced to conclude that the “risk factors” had no influence on the actual risks. [16 & 17] A study in Oslo that took in all men aged between 40-49 in the city in 1972-73, was similarly inconclusive. [18]
In 1970 there was a trial in Goteburg in Sweden in which middle-aged men in the town were assigned either to an intervention group or to one of two control groups. As usual the intervention group was encouraged to change diet, give up smoking and the rest, and the result was as usual. [19]
The Finnish businessmen’s study (1985), took 612 48-year-old businessmen and encouraged them to do much of what had been done in MRFIT: change their diet, give up smoking, take more exercise, and take various drugs for high blood pressure. There was a control group of 610 who went on as usual. [20] After five years 10 of the 612 were dead, and 5 of the 610.
The only study that tested smoking specifically was the Whitehall study in 1968. 1445 British civil servants were recruited. Half were encouraged to give up smoking; the others were not. After a year smoking in the intervention group was down by 75%. After ten years, 17.2% of the test group was dead, against 17.5% of the control group (the difference is “statistically insignificant”).
There was no change in deaths from heart disease or lung cancer, and the only other unexpected result was that the intervention group had 28 deaths from cancer other than lung cancer, compared with the control group in which the number of deaths from such cancers was 12. [21 & 22]
The letter from Philip J. Burch in the British Medical Journal already quoted says, on the MRFIT and Whitehall studies: [23]
“because of the similarity in the designs of the two trials their results can be combined to maximize numbers.”
In the low smoking intervention groups [the nagged] 56 cases of lung cancer were recorded in a total starting population of 7,142 men (0.78%); the corresponding numbers for the more heavily smoking normal care groups being 53 in 7,169 (0.74%). Findings for cancer other than those of the lung were even more surprising. Some 88 cases (1.23%) were recorded in the low smoking intervention groups, but only 60 cases (0.84%) in the normal care groups. Thus in the category ’all cancers’ there were 144 cases (2.02%) in the intervention groups but 113 cases (1.58%) in the more heavily smoking normal care groups. Reduced levels of smoking were associated with increases in cancer incidence.
“It is fair to ask” he concludes “...experts...to explain why these remarkable findings from methodologically reputable trials conflict so drastically with their claims.”
“Strenuous efforts have been made to rescue something from the wreckage, though Stallones risked the creation of many personal enemies when he wrote: ’No amount of squirming on the hook alters the fact that for every 1,000 test subjects 41.2 died and for every 1,000 control subjects 40.4 died,’” he adds, in Can Epidemiology Become a Rigorous Science? [24]
The authors believed that these results were due to chance...and the Surgeon General, C.E. Koop, omitted them from his 1986 report on passive smoking, though four years earlier, in 1982, he had praised the study for “pointing up the positive consequences of cessation in an authoritative manner”. [25] It can be understood why Fisher, Brownlee and Yerushalmy, statisticians of the highest eminence, have spoken of “catastrophic and conspicuous howlers”. [26]
Some other claims about smoking include the statement that the number of fires, and consequent deaths and injuries, has gone up to its greatest number in ten years. [27] But why, since the number of smokers has dropped so much? And can they be sure that all these fires are caused by cigarettes?
Whenever fires are caused by certain fuels, rumours grow about efforts made by certain authorities to fudge the facts, so that these fuels are not blamed. There is of course no proof, but how can there be proof that a cigarette causes a particular fire?
Any fire worthy of the name is not likely to leave any evidence in the form of cigarette packets or dog-ends. Can it be that many of these fires are like that at King’s Cross, where a cigarette end could have ignited the accumulated filth, and so got blamed rather than the people who allowed the filth to accumulate (though it seems that the cigarette was exonerated at the inquiry)?
Incidentally, Michael Wharton of the Daily Telegraph states that “reported incidents of fire on the Underground have nearly tripled since the smoking ban was imposed.” [28] We are also told that ex-smokers have less days of illness, less health complaints, and better reported health status than smokers. [29] This does not seem to tie in with the figures on life expectancy and the results of the studies described above.
The most authoritative study on this is unquestionably the Framingham Heart Study. In this town in Massachusetts, 5,127 men and women have been studied for a period of more than three decades. They have had the fullest details taken on their health and life-style, and have been checked every two years. [30]
The results show that there is no relationship at all between coronary heart disease in women, except possibly a very slightly favourable one (women who smoke have a marginally lower rate of the disease).
There is what is known as the “risk ratio”; that is, the ratio of cigarette smokers to nonsmokers in the heart disease incidence. A ratio of 1.0 shows no link at all between smoking and coronary heart disease. For men, the risk ratio begins at slightly over 1.0, but this goes down to l.0 exactly in the 30 year follow-up. Smokers of 40 or more cigarettes a day were found to have a risk ratio of only 1.3. (The risk ratio of 2 has been designated as the boundary of a weak association).
Thus the best and most often quoted evidence that cigarette smoking causes heart disease completely fails to live up to this description: at best there is a very weak relationship for men and none at all for women.
Since Hopkins and Williams in 1981 listed 246 suggested coronary risk factors in their survey, it was not difficult to find other factors that could explain this (very slight) relationship. When data about “type A” personality, (which is thought to be prone to heart disease), systolic blood pressure and serum cholesterol were included in the analysis, [31] cigarette smoking was not found to be a significant factor in coronary heart disease or myocardial infarction in men, any more than in coronary heart disease and angina pectoris in women.
What use does the US Surgeon General’s 1983 report make of these facts?
It declares that “cigarette smoking is a major cause of coronary heart disease in the US for both men and women”.
“1. In men, the incidence of CHD is two fold greater in cigarette smokers than in nonsmokers and fourfold greater in heavy smokers.
2. In women, the rates of CHD are lower than in men but are commensurately higher when the smoking patterns are similar to those in men.
3. The risk of developing CHD increases with the duration (in years) of cigarette smoking.
4. The cessation of smoking leads to CHD death rates that are substantially lower in the stopped smokers than are in the continuing smokers, and after 10 years of nonsmoking, the CHD incidence of former light smokers approximates those of nonsmokers.
To which Seltzer comments “the Framingham data substantially disagree” and “the anomaly remains unexplained”. “The explanation of the discrepancy”, he continues, with great restraint, “is an intriguing challenge for future research”. [32]
This might be the place to speak of various other claims made by the anti-smoking industry.
“Independent scientists [unnamed] have estimated that the life-long risk from passive smoking is some 100 times greater than the estimated effect of 20 years’ exposure to chrysotile asbestos, of the type normally found in asbestos-containing buildings.” [33] But chrysotile asbestos is a naturally occurring mineral. All of us take in something like one to two million fibres a year. A WHO study in 1986 [34] describes the risks as “undetectably low”.
If the risk is “undetectable”, how do you estimate what a hundred times of it is? There is also the claim that smoking mothers have smaller babies. [35] Apart from the fact that women always used to be told that big fat babies were more at risk from breathing problems, and that therefore they should keep their own weight down in pregnancy, (and of course a smaller baby was easier to deliver, and therefore it would benefit them both – and the doctors and midwives too, though this was seldom pointed out), this conflicts with an odd fact that was described in the report of the advisory committee to the Surgeon General of the US public health service, Smoking and Health. [36]
This is that a study of 922 former Harvard students showed that: “Smokers were consistently greater than non-smokers in height, weight, and in the dimensions of the head, face, shoulders, chest, hip, leg, and hand. In addition, in eight out of ten bodily indices or proportions the smoker types showed mean deviations from the non-smoker that were all in the same direction and indicative of the same trend”.
How can the indirect effect of a smoking mother stunt a child when actual smoking plainly does not stop you from being bigger?
There remains the endlessly repeated claim that 90% of cases of lung caner are caused by smoking. This seems to have originated with a study made by Doll and Hill starting in 1951, results of which were published in the British Medical Journal [37]
This is what is known as a “self-selected” study: a simple questionnaire was sent to all those doctors named in the current British Medical Register who were then living in the UK. 69% replied: that is, slightly more than two in three.
Eysenck states roundly that “no relevant conclusions regarding causality can be drawn from studies of this kind”. Only the “randomized” studies like MRFIT and the work in Framingham, where two groups selected by the researchers are compared, can be taken seriously. [38]
The “nine out of ten cases of lung cancer caused by smoking”, then, comes from this self-selected study. The first thing to observe is that the results are spectacularly different from those of the randomized studies, which found no difference between smokers and non-smokers (except for the very curious affair of the increased number of cancers other than those in the lung in the "intervention groups", who were doing all the healthy things).
However, it is only fair to add that Eysenck himself gives a report on two studies, made in Yugoslavia and Germany, where there is shown to be a connection between lung cancer and smoking in one particular group of people, who combined smoking with a certain type of personality, and also, in many cases, other factors such as a close relative who had, or had had, the disease.
(There was also some very interesting information on victims who did not smoke, but did have the “cancer-prone” personality). [39] The question of a close relationship with other cancer victims is of course of considerable importance, since Doll himself, in the Stevens lecture in 1986 [40] admitted that there were fourteen different types of cancer that were definitely or probably caused by a virus, and the virus theory is now being in bits and pieces and fits and starts agreed among scientists, even those who like Doll stick to the smoking theory (which suggests a charming picture of a virus smoking a tiny cigarette).
The study started in 1951, as stated, and was reported in the British Medical Journal in 1964.
To Part Two Of This Pamphlet
To Notes and References
Back To Judith Hatton Pamphlet Index
Back To Site Index