CONTENTS
FOREWORD
At a time when, in most things, most of the world is becoming more liberal, Westem governments have taken to treating their own citizens like children who are to be protected from risks to their health and finances. This paper concerns itself only with a major aspect of the former, health intervention programs, but other publications, such as even the Campbell Report on financial markets, have discussed the latter. Strangely, governments have taken it upon themselves to protect our bodies from our own imprudence at the time when Western governments, at least, have abandoned much of the interest they once evinced in individual moral well-being. It is as though, despairing of offering us a place in heaven, or perhaps doubting heaven's existence, Nanny government has felt it necessary to give us earthly immortality!
Aware that variants of Puritanism have provided the justification -- in the eyes of the governments themselves -- for much mischief done by them to their peoples, and having an almost innate distrust of unlimited high-handed exercise of authority, I encouraged others to look at the Nanny-State phenomenon. At the time, I did not doubt that ingestion of cholesterol, alcohol, tobacco smoke, salt etc. and lack of exercise etc. would, on average, shorten life span. Nor did I doubt that governments could reduce the incidence of morbidity and prolong average life expectancy by intervening. I did, however, ask whether old-age was an unmitigated blessing, and whether a few more years of it was worth the price of many of life's pleasures. I, further, asked myself idly whether prolonging life expectancy beyond the age of reproduction -- including the proper care of children -- was in the interests of the species. I kept this latter speculation idle because I did not want to be accused of wishing to shorten anyone's life, but now that I am 34 years old, I am becoming more courageous. I, of course, marvel that a government that takes so much interest in how much alcohol, nicotine and fat I consume should care so little about whether I continue adequately to support the children I sired -- surely, if there is a place for paternalism it is with children. And, finally, I asked myself what damn business my life is of the government.
I have not changed my view about the lethal properties of alcohol, cholesterol et al. Why would I, when we all know that most substances and practices are lethal when carried to a sufficient excess? I have, however, been surprised to learn of the failure of the lifestyle intervention programs. But I can't make up my mind whether to be glad or sorry for that fact. I have given substance to my instincts about the proper place of government in my life and why it has no business Nannying me. My other questions will have to wait upon another publication.
In the meantime I advise the reader to stop worrying about his or her health for long enough to enjoy the argument and not to get apoplexy thinking about the taxes spent upon Australia's Life-Be-In-It program. I support my case with the results of truly huge trials conducted by arms of the United States, British and other governments. It is a pity that Life-Be-In-It was not similarly monitored.
Richard J. Wood
INTRODUCTION
"Corporations have neither bodies to be punished, nor souls to be condemned, they therefore do as they like."
Lord Thurlow, 1731-1806
This is the story of a remarkable episode in the long history of man's attempt to understand and control nature, a story which in a number of cases reveals bad science, misinformation and abuse of trust. In itself, there is nothing new about such behaviour. What distinguishes this example is its sheer scale. Just as the Great War was unique because it involved virtually all the nations of the world and so later came to be known as World War One, so today we have for the first time a public health war which involves most nations of the world regardless of their politics or religion. It is a war whose aim is to rid us of all those elements of modern life which are held by those in authority to be injurious to our health. Its warriors are to be found at every level of society, from the national leader who fiscally assaults the smoker to the citizen who abuses his neighbour for being overweight. Governments around the world, supported by the public, are spending enormous amounts of money in an attempt to persuade us to live properly and adopt a "healthy life-style".
Perhaps the nearest historical parallel to this international campaign is to be found in the great witchhunts which occurred at various times throughout Europe and North and South America reaching their peak in the seventeenth century. About a million witches were tortured and executed over a period of 300 years. Today those witchhunts are looked on as a terrible aberration in the evolution of a progressively more humane civilisation, as something never to be repeated. Yet there are still elements of the witchhunt in today's lifestyle campaigns. Lifestyle heretics are harassed in public and even pursued by junk mail, television and newspaper to their very hearths in an unending attempt to make them recant, confess their unhealthy life-style and turn over a new behavioural and nutritional leaf.
There is no suggestion that health professionals at every level have anything but our well-being at heart. Indeed, a parallel is often drawn between the worthy and very successful attempts to eliminate the most serious infectious diseases and modem man's a quest for the elimination of the degenerative diseases, particularly cancer and cardio-vascular disease. However, good intentions are not enough. It is my contention that, well-meaning though they may be, our advisers have too often misread the evidence.
As a libertarian, I believe that people should be free to lead their private lives as they choose. What is private? That question is really one for the political scientists to wrangle over. My personal view is that since there can be no personal activity without some consequence to others, toleration is required by all. I find strong deodorants offensive, and the zest emitted from a freshly peeled orange offensive to the point of injury (there has been at least one fatality resulting from inhaled orange fumes triggering a fatal vaso-vagal attack). And I tolerate them. Reciprocally, I expect the deodorised and the orange-eaters to tolerate my cigar smoke.
My cholesterol-rich boiled breakfast egg and my fatty lamb chop dinner are strictly my affair whether or not they bring about my premature death. No-one castigates the skier for breaking his leg and thereby increasing the national health bill and no more should I be for supposedly taking risks in the much murkier waters of cardiovascular disease causation.
The principal argument which will be made in this paper is that the elements of the supposedly healthy life-style -- exercise, fresh air, fat-free diet etc. -- are not necessarily conducive to long life, good health or general well-being and may have quite the reverse effect. That is quite apart from the concomitant infringement of our rights as free citizens.
CHAPTER 1
"Passive smoking" is the inhalation of tobacco smoke produced by other people. Because smoking itself is widely regarded as a cause of ill-health and in particular lung cancer, concern has been expressed that passive smoking also might be harmful. Although both smokers and non-smokers may be passive smokers, most concern has naturally been felt for the non-smoker and that is the question considered here: specifically, does passive smoking cause lung cancer in non-smokers? (1) The answer necessarily must come from epidemiology, the study of the frequency, distribution and cause of illness and death. It is quite insufficient to quote the evidence that cigarette smoke contains poisons -- it certainly does. The Surgeon General's report marshalls a battery of such evidence, but as such it is circumstantial and incomplete. Rhubarb and spinach contain about one per cent oxalic acid but we need not quail on being offered Eggs Florentine or Rhubarb and Apple Crumble, even though a lethal dose of oxalic acid is only 10 grams (Diem, 1962; Dreisbach, 1977). A properly controlled study would be needed to determine the incidence of rhubarb and spinach poisoning in the community. So too with passive smoking and lung cancer. If the purpose is to frame public policies which may limit individual freedoms then it is of particular importance that a correct and reliable answer be obtained.
There are two important kinds of study. In a case-control study, people with lung cancer are examined and compared to similar individuals without lung cancer. How do the individuals in the 2 groups differ? For example, if 100 non-smokers with lung cancer had all been exposed to tobacco smoke while 100 disease-free neighbours matched for age, sex, race etc had never been so exposed then this would constitute a strong link between passive smoking and lung cancer. In practice no one expects such clear cut results. If 60 of the lung cancer cases had been exposed compared with 40 of the healthy group then the result would be less clear. The methods of statistics are used to determine whether such a result is "significant". Here "significant" means "statistically significant". It has nothing to do with the magnitude of the effect observed. A study which showed that passive smokers were 1% more likely to contract lung cancer than non-exposed individuals might still have found a significant effect.
The second kind of study is the prospective study. For example, select a group of healthy people married to non-smokers and follow than for 10 or 20 years. Compare them with a similar group of people married to smokers. Which group has the highest incidence of lung cancer and, once again, is the difference significant? A major difficulty with such studies is the time and money required to obtain a result. According to the orthodox view, it takes on average 40 to 50 years of cigarette smoking to induce lung cancer (Smoking and Health, 1962: 20). Only a small proportion will succumb much sooner. Therefore a large population must be studied over a long period if any significant effect is to be detected.
Table 1.1 summarises the results of prospective studies in men and women, with the duration of the trial and the number of subjects who were followed.
Table 1.1: Lung Cancer and Passive Smoking Caused by Spouse: Prospective Studies
| Study | Duration (Years) | Men | No. | Women | No. |
| Garfinkel (1981) | 12 | • | 94,000 | ○ | 375,000 |
| Gillis et al. (1984) | 6 | ○ | 827 | ○ | 1917 |
| Hirayama (1981a,b,c, 1984a,b) | 17 | ↑ | 20,289 | ↑ | 91,540 |
Notes:
○ = a non-significant result
• = this condition was not examined in the paper
↑ = a positive association
Tables 1.2 and 1.3 summarise the results of case control studies in men and women respectively. Table 1.4 summarises the results of case control studies of men and women together. The authors of each study are listed first, then the number of (nonsmoking) lung cancer cases. In Tables 1.2, 1.3 and 1.4 the source of the cigarette smoke -- spouse, other family member, workmates etc, is then given. Different authors have classified the source of the cigarette smoke according to their own criteria; I have used the classification which here seems most accurate. The source of the cigarette smoke for the prospective studies was always effectively a spouse.
Table 1.2: Lung Cancer in Men and Passive Smoking: Case Control Studies
| Study | No. of cases | Spouse | Children Family | Parents | Work | Leisure | Travel | |
| F | M | |||||||
| Akiba et al. (1986) | 19 | ○ | • | • | • | • | • | • |
| Buffler et al. (1984) | 5 | • | ○ | • | • | • | • | • |
| Correa et al. (1983) | 8 | ○ | • | ○ | ○ | • | • | • |
| Dalager et al. (1986) | 29 | ○ | • | ○ | • | • | • | |
| Kabat and Wynder (1984) | 37 | ○ | ○ | • | • | ↑ | • | • |
| Lee et al. (1986) | 32 | ○ | ○ | • | • | ○ | ○ | ○ |
Notes: As for Table 1.1
Table 1.3: Lung Cancer in Women and Passive Smoking: Case Control Studies
| Study | No. of cases | Spouse | Children Family | Parents | Work | Leisure | Travel | |
| F | M | |||||||
| Akiba et al. (1986) | 94 | ○ | • | • | • | • | • | • |
| Buffler et al. (1984) | 33 | • | ○ | • | • | • | • | • |
| Chan and Fung (1982) | 84 | ○ | • | • | • | • | • | • |
| Correa et al. (1983) | 22 | ↑ | • | ○ | ○ | • | • | • |
| Dalager et al. (1986) | 70 | ○ | • | ○ | • | • | • | |
| Garfinkel et al. (1985) | 134 | ○ | • | • | • | • | • | • |
| Humble et al. (1987) | 20 | ○ | • | • | • | • | • | • |
| Kabat and Wynder (1984) | 97 | ○ | ○ | • | • | ○ | • | • |
| Koo et al. (1984, 1985, 1987) | 88 | ○ | ○ | ○ | • | • | • | |
| Lee et al. (1986) | 12 | ○ | ○ | • | • | ○ | ○ | ○ |
| Pershagen et al. (1987) | 77 | ○ | • | • | • | • | • | • |
| Trichopolous et al. (1981, 1983) | 77 | ↑ | • | • | • | • | • | • |
| Wu et al. (1985) | 31 | ○ | • | ○ | ○ | ○ | • | • |
Notes: As for Table 1.1
Table 1.4: Lung Cancer in Men and Women and Passive Smoking: Case Control Studies
| Study | No. of cases | Spouse | Children Family | Parents | Work | Leisure | Travel | |
| F | M | |||||||
| Humble et al. (1987) | 28 | ↑ | • | • | • | • | • | • |
| Sandler et al. (1985a,b) | 2 | • | ○ | ○ | ○ | • | • | • |
Notes: As for Table 1.1
For each category an upward arrow (↑) indicates a positive association i.e. a passive smoker was significantly more likely to contract lung cancer than someone who was not a passive smoker. "o" indicates no significant association was found and a decimal point (•) indicates that the condition was not examined in the particular study.
The arrows reflect the risk associated with passive smoking, the risk-ratio. In these studies the risk-ratio ranged from approximately three times more likely for the largest positive association to three times less likely for the largest negative association.
The tables summarise the results of 26 studies of lung cancer in nonsmoking men, women, and men and women. Of these, 20 revealed no association. Of 51 separate exposure situations (spouse, leisure, etc.), 45 revealed no association. The failure to find a positive association in most examples suggests a tenuous connection at most, unless the minority studies are of convincing persuasiveness.
The most widely quoted results are perhaps those of Hirayama. Hirayama (1981a) analysed his data with a test devised by N. Mantel. Mantel (1981) replied in a letter to the British Medical Journal in which he expressed concern about ambiguities and omissions in Hirayama's paper. In particular it appeared that in Mantel's test Hirayama had mistaken χ for χ2. Mantel provided an arithmetically detailed criticism including Mantel's own calculated values. He concluded "The question then is whether he has conducted a more refined analysis, about which he is giving us no clues, or he has mistakenly interpreted his χ2 value as a χ value". In other words, had Hirayama simply blundered, as even the best of us can do. Hirayama (1981b) replied to Mantel in the British Medical Journal, 3 October 1981 with the sentence: "The validity of the significance test used in my paper was kindly confirmed by prominent statisticians in many institutes, including the US National Cancer Institute and the Massachusetts Institute of Technology." With an appeal to unnamed authorities Hirayama therefore dismissed Mantel's queries without answering. In any other area of science -- or indeed intellectual discourse generally -- this would be enough to negate Hirayama's contribution. However he went further and added new information in the form of a histogram (Figure 1.1). This revealed some extraordinary results. Rutsch (1981) showed that from Hirayama's data it could be deduced that lung cancer was commoner for non-smoking unmarried women than for the non-smoking wives of smokers. Hirayama (1981d) agreed with Rutsch but noted that most of the "unmarried" women were widows, adding "Although their late or former husbands' smoking habits were not asked about at the time of enrolment there is little doubt that the majority of them were smokers". Sloppy data collection, it appears, leaves little room for doubt.
Other strange features were described. Non-smoking men with smoking wives were as likely to get lung cancer as occasional smokers, if not more so. Lee (1981) pointed out that in effect this meant that a cigarette apparently had the same carcinogenic effects on a person whether it was smoked by him or by someone else! Because "active" smokers must necessarily breathe the same air as "passive" smokers this would be a remarkable result indeed. (2) Surely "active" smokers should get a double dose -- smoke they inhale while their cigarette is in their mouth, plus the polluted air they breathe between puffs. Lee went further and showed that Hirayama's 11 printed confidence intervals (3) were all in error by factors of up to 1000%.
In reply to this new criticism Hirayama (1981c) admitted that his evidence appeared to show that it mattered not whether a cigarette was actively or passively smoked. His response to the errors pointed out by Lee was "I regret that errors have been found in the 95% confidence intervals shown in the Figure in my last letter. The correct values are given in the accompanying Figure [Figure 1.2]. The errors do not, however, influence the substance of my letter". After such a comedy of errors it is difficult to take seriously either his work of 1981 or his later paper (Hirayama, 1984). There can be few authors in the history of science who have distinguished themselves by replying to a published criticism with a reply which itself contains 11 errors of between 100% and 1000%.
Trichopolous et al. (1981) originally presented the results of his study for 40 non-smoking women with lung cancer. He compared them with 149 controls who did not have lung cancer. The results as risk-ratios (4) for women whose husbands were non-smokers, ex smokers, 1-20 cigarettes/day and 21+/day are shown in Table 1.5.
Table 1.5: Risk-ratios for Lung Cancer in Non-smoking Women:
Data of Trichopolous et al. (1981, 1983)
| Smoking Habit of Husband | ||||
| Non- smoker | Ex- smoker | 1-20 per day | 21+ per day | |
| Trichopolous et al. (1981) (40 cases, 149 controls) | 1.0 | 1.8 | 2.4 | 3.4 |
| Trichopolous et al. (1983) (77 cases, 225 controls) | 1.0 | 1.9 | 2.4 | 3.4 |
| Trichopolous et al. (1983) corrected by Heller (1983) (77 cases, 225 controls) | 1.0 | 1.9 | 1.9 | 2.5 |
Burch (1981) drew attention to a puzzling feature of the original Trichopolous data, one similar to that found by Hirayama: the non-smoking wives of heavy smokers were more likely to contract lung cancer than wives who actively smoked. Burch made the pointed comment: "It is difficult to imagine that passive smokers -- even with husbands smoking more than 20 cigarettes per day -- could breathe in more carcinogens than active smokers." There comes a point at which the case against passive smoking becomes too strong: it has been reached when evidence shows -- or appears to show -- that passive smoking is more dangerous than active smoking.
Trichopolous continued to add to his observations. By 1983 he had almost doubled his number of cases and controls. He published the results for the combination of new and old subjects and controls (Trichopolous et al., 1983) (Table 1.5). Remarkably, in spite of the increase, the risk-ratios were virtually unchanged: with ex-smoking husbands, 1.8 became 1.9, with 1-20 cigarettes/day husbands the risk-ratio remained at 2.4, and for 21+ cigarettes/ day husbands the risk-ratio remained at 3.4. Such resistance to change in the risk-ratios, though possible, is most unlikely and suggests there might be another explanation for the results. There is indeed, as Heller (1983) showed. From the published data (Trichopolous, 1983) Heller recalculated the ratios (Table 1.5). Trichopolous (1984) replied to Heller in the Lancet March 24, 1984: "He is right and we apologise for our typing error (which was pointed out by others including Dr Gosta Axelson of Goteborg, Sweden)". Once again, a most casual response to a most serious criticism of the principal conclusion to his 1983 statement. Although published scientific results are normally taken on faith, such casualness in this case makes them at best dubious.
In my opinion the results of Hirayama and Trichopolous have been vitiated both by the devastating criticisms which have been made of them and by the casual and indeed inadequate responses by their two authors.
How then have such august authorities as the Surgeon-General of the United States of America (Koop, 1986) and the National Health and Medical Research Council of Australia (NHMRC, 1986) arrived at the conclusion that the epidemiological evidence is "strongly suggestive" (NNMRC, 1986: 38) or "compelling" (Koop, 1986: 97)? The answer may lie in a quite uncritical evaluation of the evidence. Consider the results of Trichopolous. The NHMRC in its Report grasped the nettle firmly (NHMRC, 1986: 33). Only the original and incorrect calculations of risk ratios given by Trichopolous et al. (1983) were quoted: no mention was made of the total recalculation of risk-ratios produced by Heller (1983) and accepted by Trichopolous (1984). The Surgeon-General of the USA (Koop, 1986: 71) gave the corrected values but attributed them to "Trichopolous et al., 1983", where they do not appear, rather than to Heller (1983) who had corrected the faulty homework of Trichopolous et al. For his trouble, Heller (1983) does not even get mentioned by the Surgeon-General. This avoidance of published corrections leads to a quite natural sense of security in acceptance of erroneous published conclusions.
The results of Hirayama receive a similarly curious treatment. The Surgeon-General in a discussion of Hirayama's critics made no mention of the enormous errors in Hirayama's 1981b paper. On the contrary, he said (page 76) that in response to Lee's (1981) criticism "the calculations were later confirmed", but did not mention the fact that Hirayama himself admitted making errors of a magnitude which put them more in the realm of cosmology than epidemiology. The NHMRC Report on the other hand simplified the issue by not even mentioning the question of error.
A similar lack of critical zeal has been effected with other papers on passive smoking. The study of Correa et al., (1983) is described in the Report as showing "Positive association in both males and females statistically significant". As shown by Correa et al., in their Table 1, it was in fact significant only for females. The Report also states that Correa et al., found a positive trend in lung cancer in non-smokers with passive smoking. They did not and nowhere say so. Garfinkel et al., (1985) are described in the Report's Table 1 as having reported a "Positive association in non-smoking females; statistically significant". Garfinkel et al. in fact summarise their results by saying "In conclusion, we found an elevated risk of lung cancer, ranging from 13 to 31%, in women exposed to the smoke of others, although the increase was not statistically significant" (emphasis added); the justification for the Report's claim seem to be that the small subset of women whose husbands smoked more than 20 or more cigarettes per day were significantly more likely to develop lung cancer (NHMRC, 1986: 35). The danger of such selective presentation of data is readily demonstrated: another subset of women, those who had been exposed to passive smoking for 1-2 hours per day over the last 25 years were significantly less likely to develop lung cancer (Garfinkel et al., 1985: Table 4). It is possible to go through the multiple analysis of Garfinkel et al. and find as many significant negative associations as positive associations but the overall result remains as stated by the authors and not by the NHMRC Report. Perhaps this discrepancy due to selective presentation is not surprising. The term of reference of the NHMRC Committee required "a thorough review of the literature". Yet their Section 7, "Passive smoking and cancer", in its 34-item bibliography includes no reference to Burch (1981), Heller (1983), Hirayama (1981b), Hirayama (1981c), Lee (1981), Mantel (1981), Rutsch (1981), or Trichopolous (1984).
CONCLUSION
It is evident from what has been presented that drawing a conclusion from the mass of published data on passive smoking and lung cancer is bound to be exceedingly difficult. One way of doing so is to consider the data at three successive levels of analysis. First, examine the results taken at face value. Second, examine the criticism of the work and the responses of the authors. Third, examine the reviews of experts.
- At face value the published studies do little to inspire confidence in anything stronger than a "not proven" verdict, if not one of "not guilty". The great majority have produced no association between passive smoking and lung cancer. Any true correlation must be so small as to be scarcely detectable by even the most expensive and time consuming of trials.
- The criticisms levelled at some of the most important studies have been both justified and convincing, and the responses of the authors have often been inadequate and inappropriate. It is in the nature of science that errors are made and published but few scientists make errors of the kind and order of those described here. There comes a point at which the most trusting reader, scientist or not, is justified in treating some published data with profound reserve.
- It is the duty of experts whether in committee or not, to critically assess the evidence on which they will base their conclusions. That has not in general happened. Rather there has been what can only be described as a cover-up. Even when the critical assessment and recalculation of the data have been done by others, the corrected amendments have been either ignored or incorporated without mention of the criticisms. Results have been selectively presented with unwelcome results left unpresented.
Ultimately, with the results presently available, any judgement must be subjective. But there is a way of obtaining a new perspective on the available material. Suppose the results described in this chapter referred not to cigarette smoke but to barbecue smoke wafted into a neighbour's backyard, or the vapour from deodorised family members and workmates. Would such results lead to the banning of barbecues or the exclusion of the deodorised from public society? The notion would, I think, be dismissed as absurd and resting on quite inadequate evidence, even though both meat smoke and deodorants may contain carcinogens (McGee, 1984; Selinger, 1989). But barbecues and the deodorised are not scapegoats. Not yet.
ENDNOTES
1. Other diseases have been examined in this context but lung cancer has naturally received most attention and is the only one to be considered here.
2. Burch (1985) analysed a similar situation which arose from the work of Sander et al. (1985a) on all cancers and showed that there are three possible explanations for such a result: (1) Active and passive smoking are both carcinogenic; (2) Active smoking causes cancer and passive smoking prevents cancer; or (3) Active smoking prevents cancer and passive smoking causes cancer.
3. A confidence interval is here an expression of the possible error in a measured risk ratio. For example a risk ratio of 3.0 may be the best available estimate, but if ten times as many cases had been examined a better estimate might have been obtained. Statistical techniques permit a range to be associated with the risk ratio, say 3 ± 1 (95% C.I.): in other words, we can be 95 per cent certain that the risk ratio lies somewhere between 2 and 4.
4. For example, a risk ratio of 3 means that a passive smoker is 3 times more likely to get lung cancer than someone not thus exposed.
No comments:
Post a Comment