PART ONE

THE "LALONDE DOCTRINE" AND PASSIVE SMOKING

Public health authorities often say that other people's smoking harms us. Should we accept this without question or should we wonder how they know that? This raises the need for public debate about government policy in technical areas that can be fully understood only by experts on whose advice politicians must rely. It is instructive to examine in this wider context the empirical basis for legislation that purports to protect us from passive smoking.

Some of the problems associated with decision-making based on expert scientific advice have been documented by C.P. Snow in his essay Science and Government. Such decisions, he noted, are made by people who not only have no first-hand knowledge of their bases but, accustomed to thinking for a short time about the interconnections between many things, are temperamentally quite different from their scientific advisers, who have been trained to think deeply and obsessively about one thing for a long time. Snow recorded that, in his experience:

"Even at the highest level of decision, men do not really relish the complexity of brute reality, and they will hare after a simple concept whenever one shows its head" (Snow,1962:66).

He went on to point out that, while science rarely provides simple answers to complex questions, even intelligent and highly- placed non-scientists believe so strongly that such answers do exist that they are influenced more by the force and authority with which an unambiguous opinion is expressed than by whether it is right or wrong: something they are seldom qualified to judge.

Snow was warning us against uncritical reliance on authoritative scientific opinion. Unfortunately, his message has been taken by some health promoters as showing how easy it is to influence both decision-makers and the public by the authoritative statement of unequivocal opinion. Indeed, such a strategy was openly advocated by Marc Lalonde, the former Canadian Minister of National Health and Welfare, in A New Perspective on the Health of Canadians, published in 1974. This report is acknowledged by the Australian Institute of Health, in its 1988 biennial report Australia's Health, as having had "a major impact on thinking about health, health services, health promotion and illness prevention" (p.17). Chapter 9 of the Lalonde report, entitled Science versus Health Promotion, makes the point unashamedly: "Science is full of 'ifs', 'buts' and 'maybes' while messages designed to influence the public must be loud, clear and unequivocal".

Noting that scientists are divided on issues like the bearing of exercise and diet on coronary artery disease, it goes on to say that nevertheless "action has to be taken ... even if all the scientific evidence is not in" (1974:57). It concludes:

"The scientific 'yes, but' is essential to research but for modifying human behaviour of the population it sometimes produces the 'uncertain sound' this is all the excuse needed by many to cultivate and tolerate an environmental and lifestyle that is hazardous to health" (1974:58).

The ruse is, of course, that we are not told how it is to be decided that there is enough evidence to justify action or who is to determine what that action should be. By giving primacy to behaviour modification, rather than to honest reporting of conflicting scientific opinion, the Lalonde Doctrine seems to encourage both misleading the public and, as the closing remarks suggest, dealing with dissenters by imputing to them preconceptions and prejudices rather than reasons that might be worth discussing.

Nevertheless, those who implement the Lalonde doctrine are clearly motivated by the desire to improve our health. One cannot claim that their "loud, clear and unequivocal" messages are deliberate lies: indeed, they may well turn out to be unarguably true.[1] Paradoxically, it is this very fad, interacting with the dogmatic separation of ambiguous science from unambiguous policy, that leads to disquieting consequences.

The prudent, whether scientists or not, qualify their opinions with ifs, buts, and maybes because they admit there could be new facts and different interpretations of old ones that would force a change of mind. They discover this by talking to other people; this of course is how we reach consensus. Those who deliberately avoid such qualifications commit themselves to a position that they cannot later admit was wrong without calling their credibility into question. Such people cannot,in principle, reach consensus with others. The nearest they can come to it is to form groups of like-minded people to whom, by mutual reinforcement of belief, even the idea that there could be a reasonable but unlike-minded person, gradually becomes puzzling and incomprehensible. Dissent is then seen as an attack on the aims that bind the group, and can be made comprehensible only by attributing it to the agency of an opposing group with different aims.

An example of this is an editorial comment in The Medical Journal of Australia (8 April 1978, pp. 384-7) on Sir Ronald Fisher who, it acknowledged, probably did more than anyone to establish statistics as an applied science. The editorial remarked that:

"For reasons which have puzzled many people, and which will perhaps always remain a mystery, he chose to lend his great reputation to support the tobacco interests in their attempts to defuse the health issue when it was first raised ".

Again, in The Journal of Chronic Diseases, M. Kristein (1985) said of a critic who had argued that factors other than smoking might contribute substantially to lung cancer, that:

"I am led to the conclusion that he is not engaging in an exchange concerned with advancing knowledge and public policy. Rather (he) appears to see his mission as one of exonerating smoking ... "

He then attacked the approaches adopted by his critic and explained that it was necessary to reply at length because they had been used "to oppose needed public policy". Similarly, the closing paragraph of the editorial comment in The Medical Journal of Australia quoted above dismissed a particular dissenting view as "arguments about the exact interpretation of data in a piecemeal way", as if that in some way invalidated it, and concluded:

"... critics appear to argue that, because there are certain aspects of the evidence which are possibly subject to error, people should continue to smoke until the matter is cleared up. In view of the massive evidence of the dangers of smoking we believe that this proposition is clearly contrary to all reason. "

For our purposes, what is important about these examples, which are by no means atypical, is not who had the better of the argument but the way in, which the norms of scientific enquiry have been eroded. Dissent is no longer seen as part of the normal process of assessing evidence, but as an attempt to thwart the public policy of behaviour modification that has been based upon it.

As research groups become dominant, they increase their influence on the editorial and refereeing policies of professional journals and on the research funding strategies of granting bodies; consequently, it becomes increasingly difficult to an and examine all sides of a question that bears on the group's aims. Not encouraged to decide for itself, the public may be fed only what the group itself believes to be true. The dissemination of different interpretations of the facts may be discouraged and it becomes more difficult to finance research suggested by those interpretations. The end result is a common presumption that no serious scientist doubts the group dogma such as "smoking kills".

This would not perhaps be a major problem if it were only a question of the dangers of smoking, for it is clearly possible to live without smoking. But the Lalonde doctrine is part of a trend towards dogmatism by pressure groups intent on persuading politicians to adopt their beliefs and to force everyone to behave in the ways they believe to be good for us. The success of the Lalonde doctrine in health promotion has made it legitimate for such groups to replace the ifs, buts and maybes of human inquiry by the loud, unequivocal message that can brook no dissent. Consensus is now seen as the outcome not of a reasoning process but of a head count of those who support this or that action, regardless of their reasons for doing so. Unfortunately, people are not prevented from taking passionate and vociferous stands for or against dogmatic claims by lack of knowledge of their authenticity. As psychologists have noted, the content of a claim is relatively unimportant compared to the readiness to believe it, either because it fits into one's belief system or because of the context in which it is presented (Watzlawick,1976:29,139).

The context of health promotion is one in which we are all ready to hare after simple prescriptions for a longer and better life. Few of us have the time to assess the evidence for ourselves; this is as true of medical scientists in other specialities as it is of the general public. It is a matter of concern, therefore, when we are presented only with a version of the evidence that has been deliberately shorn of caveats, since this denies us any possibility of deciding for ourselves.

Another consequence of the success of the Lalonde doctrine has been the politicisation of areas of medical science that can attract a high level of public funding by maintaining a correspondingly high public profile. This is just one aspect of the way in which the social dimension of science affects its progress. At a time when many areas of health research are short of funds, it is conceivable that money used to indict tobacco, for example, could be better spent on what some might argue are more pressing and important areas of medicine.

The State of Victoria boasts of being a world leader in the crusade for a smoke-free environment. Smoking on buses, trams and trains is banned. The Victorian Tobacco Act of 1987 banned the advertising of cigarettes in cinemas, on billboards and on shop awnings; television advertising was banned earlier. Smoking is now banned in Victorian taxis and, throughout Australia, on all domestic airline flights, in all government offices and in most government business enterprises and local council offices. Many private sector organisations, universities and other bodies are rapidly adopting similar restrictions. Opposition to these aggressive developments of the anti-smoking campaign has been suppressed by promoting the idea that smoking is not only harmful to the smoker but that it is also harmful to non-smokers.

A recent newspaper article prominently displayed the statement:

"Studies have linked passive smoking with conditions such as lung cancer, heart disease, pneumonia and bronchitis in infants. The findings have created a phobia among non-smokers" (McGuinness,1989).

It would be more accurate to say that the phobia has been created by the reports of those findings as presented to the public by the health promoters.

The extent to which the Lalonde doctrine was invoked in that presentation can be judged from the review of passive smoking recently published under the auspices of the World Health Organisation (1987), which is well known for its anti-smoking stance. referring to the evidence of the effects of exposure to smoke from other people's cigarettes on the health of human populations, it said:

"Most of these possible effects of passive smoking, however, have not been established incontrovertibly. "

It would seem that although at that time there was no firm evidence of the harmful effects of passive smoking, health promoters had convinced both decision-makers and public that action was needed to protect the health of non-smokers. They did so by the deliberate suppression of any ifs, buts and maybes.

One tactic was to refer, first of all, to two 1981 studies of lung cancer in non-smoking spouses of smoking men, one from Hirayama in Japan and the other from Trichopoulos in Greece; then to claim that these studies had been supported by subsequent studies elsewhere; and, finally, to quote an overall figure of a roughly 30 per cent increase in the risk of lung cancer for the non-smoking spouses. But this unequivocal message failed to tell us that the later studies did not in fact confirm the possibly harmful effects discovered by the Hirayama and Trichopoulos studies. That is why the initial Hirayama finding of a threefold increase in the risk of lung cancer was substantially reduced by a factor of 10 to the 30 per cent increase of the overall figure. A review of the evidence up to 1987 considered 13 subsequent studies and concluded that none of them produced a result that could be definitely seen as harmful. Sane of them reported a small percentage increase in the risk of lung cancer among the non-smoking spouses, whereas others reported a small percentage decrease in that risk (Lee,1988). An earlier review considered the subsequent studies then available and came to substantially the same conclusions. Referring to the eight studies then in question, it considered that the association between lung cancer and passive smoking "can not at present be regarded as casual" (Pershagen,1986).

Yet, the honest dissenter cannot claim unequivocally that passive smoking is harmless. The most that can be done is to examine claims that passive smoking is harmful and indicate the ways in which they might be misleading. Against this, the Lalondists simply assert that action must be taken even if the evidence is inconclusive. But the sources of this imperative is not given and people seldom ask for it.

The three reviews mentioned above show that, at the time health promoters were claiming a linkage between passive smoking and lung cancer, only two out of 15 studies could be said to support that claim.

One case for action has been put by C Papier and S Stellman (1987). But this does little more than repeat the message of the closing sentence in its abstract:

"Despite uncertainties and differences of interpretation of various cancer studies, there is ample justification for public health measures now in place or proposed such as restriction or elimination of smoking in the workplace and in public places."

After admitting that "methodological uncertainties do not yet permit the absolutely firm conclusion that passive smoking causes cancer in those exposed" they go on to say that the quality of evidence "leaves no doubt about the propriety of public health measures intended to reduce the risks to non-smokers ".

Just how "uncertainties" lead to "no doubt" is not explained, nor is it explained why public health measures are needed to reduce risks that may not be there in the first place. Research studies have been undertaken into the effects of passive smoking on other diseases in adults, including several hundred into the relationship between parental smoking and the health of children. It is fair to say that some claim to have found evidence of possibly harmful effects. In addition, numerous studies have been undertaken into the effects on the foetus of maternal smoking during pregnancy and its association with birth weight, peri-natal mortality, spontaneous abortions and congenital malformations. These studies are not reviewed here because the principal issue under discussion is not whether passive smoking is harmful, but the way health promoters edit the facts to match their convictions; the examples already cited are enough to make that point.

Health promoters have sometimes presented facts in such a way as to support their assertions that 'action must be taken'. This has obscured the fact that the nexus between evidence and action sometimes consists of little more than strength of convictions. The principal aim of such action is to modify behaviour to reduce risks that, even if they exist, may not be as serious as is claimed. To achieve this effect, scientific dissent has sometimes been dismissed as eccentric opposition to desirable health policy, and the public has been denied knowledge of the normal caveats of scientific enquiry.

It is in the public interest to monitor these consequences of the Lalonde doctrine and to make it widely known that the dicta of health promoters are sometimes designed to influence both public and governments by the deliberate suppression of anything that might suggest that the bases of their proposals are less than certain.

PART TWO

MISLEADING CLAIMS ON SMOKING AND HEALTH

I have argued that the campaign against passive smoking was an example of how health promotion policy was ignoring the uncertainties and ambiguities of scientific evidence in order to make its messages "loud, clear and unequivocal", as recommended by Marc Lalonde, formerly Canada's Minister of National Health and Welfare.

I now propose to demonstrate further how the anti-smoking lobby has distorted evidence about the health effects of smoking so as to persuade people to support its cause. I also dispute claims that reductions in smoking bring substantial savings in overall health costs. My point throughout is not that smoking is safe but that health promotion policy is not in the public interest unless it is based on sound interpretation of the available evidence.

There are two ways in which relevant evidence has been distorted. First, studies that do not support the anti-smoking lobby's claims have been ignored. Second, evidence has been interpreted in such a way as to suppress other but equally valid interpretations of it.

An example of the first kind of distortion is contained in the latest report on the health effects of smoking by the US Surgeon General. On the subject of coronary heart disease (CHD) - a term embracing all forms of cardiac disorder resulting from impairment of the coronary arterial circulation by atherosclerosis (thickening or obstruction of the arteries) - the Report claims that:

"The findings of several prospective studies involving more than 20 million person-years of observation ... have been remarkably similar: cigarette smokers are at an increased risk for fatal and non-fatal myocardial infarction and for sudden death. Overall smokers have a 70 percent greater CHD death rate, a two-to-fourfold greater incidence of CHD, and a two-to-four fold greater risk for sudden death than non-smokers." (US Department of Health and Human Services,1989: 58).

But even someone who accepts these findings could well wonder whether there are studies that conflict with them. We could have been told that, as reported by team-members of the respected Framingham Heart Study:

"... men who never smoked had higher mortality-rates than men who quit and there are other inversions in the morality trends by level of smoking. We think this is explained by sampling variability but clearly some other explanation would serve as well."

With commendable objectivity they also remark that:

"We do not pretend the Framingham data are definitive: they simply provide another set of facts to consider. " (Gordon et al., 1977).

Similarly, we could have been told that that study did not find any statistically significant effect of smoking on CHD in women.

It is also worth noting that in the so-called seven-country study there was no significant association between CHD incidence and smoking in the different countries (Keys,1970) and that in the prospective necropsy series of the Oslo study no significant correlation of coronary-raised atherosclerotic lesions and smoking could be shown (Holme et al.,1981). Again, various Heart Disease Intervention Studies involving changes in lifestyle have produced mixed results, ranging from the possibly harmful to the possibly beneficial. A recent review of these studies argues that they have provided no evidence that CHD is preventable in this way. In reply to the claim that to await definitive evidence is to deny people possible benefits and become an enemy of public health, the authors remark that: "This might be a tenable stance if the interventions were themselves harmless, but they are not". They conclude, "The financial and human resources used in this possibly ineffective crusade could be better employed elsewhere" (McCormick & Skrabanek, 1988). (For the other side of the argument, see Gunning-Schepers et al. [1989]).

To see how easy it is to highlight one's message by stressing one interpretation of the facts at the expense of other, equally valid interpretations, consider the following statistical evidence, taken from Breslow & Day, (1980).

It has been estimated that males below age 45 who smoke 25 or more cigarettes a day suffer almost 15 times the risk of similarly-aged male non-smokers of dying from ischaemic heart disease (IHD - a term synonymous with coronary heart disease). This 15-fold increase is calculated thus: the annual death rate from IHD for the non-smokers is seven per 100,000 men whereas that for the smokers is 104 per 100,000 men: 104 divided by seven is a little under 14.9, i.e. almost 15. The calculation is usually seen as telling us that, in this age-group, the effect of smoking is to multiply the risk of dying from IHD by 15. This way of telling it is a very effective way of highlighting the message that 'smoking kills'.

But if we ask not only by what factor smoking increases the chance of dying from IHD, but also by what factor smoking reduced the chance of escaping death from IHD, the same facts appear in a different light. To answer this second question, note that 99,993 per 100,000 non-smokers escape death from IHD whereas only 99,896 smokers do so. Since 99,896 divided by 99,993 is 0.999, the factor in question is 0.999.

In other words, the data tell us not only that smoking is associated with a 15-fold increase in risk of death from IHD but also that the smoker has 99.9 percent of the chance of a non-smoker of escaping death from IHD.

The point here is not that the second calculation exonerates smoking but that presenting only the result of the first calculation suppresses an aspect of the data that could lessen the force of the message that 'smoking kills'. For telling a 40- year old male heavy smoker that his chance of not dying from IHD is about 99.9 percent of that of a comparable non-smoker is much less likely to persuade him to abandon smoking than telling him that his chance of dying from IHD is almost 15 times that of a comparable non-smoker.

To illustrate that the preceding result is not just a vagary of one particular data set, consider the death rates from lung cancer as given in one of the benchmark papers of the anti-smoking movement (Doll & Bradford Hill,1986). These were 166 per 100,000 for smokers of 25 or more grams of tobacco a day and sever per 100,000 for non-smokers. Smoking was thus associated with an almost 24-fold increase in the risk of death from lung cancer. But it would be equally true to say that a smoker has about 99.8 per cent of the chance of a non-smoker of escaping death from lung cancer. Similarly, a 30 per cent increase in the risk of lung cancer, which is the magnitude of the alleged overall effect of passive smoking on spouses of smoking men, would correspond to about 99.998 per cent of a non-smoker's chance of escaping lung cancer (1). While this figure is not based on the passive smoking data itself, it does give a rough idea of how one-sided it can be to report only a per centage increase in risk.

These relatively high per centage chances of escaping the disease in question do not by themselves establish that smoking is harmless. For one thing, they ignore death from other causes and refer only to one disease at a time rather than to the spectrum of diseases with which smoking has been associated. Again, the calculation for IHD refers to one specific age-group and does not consider a possible cumulative effect as it becomes older (2). But that same is true, of course, of the corresponding per cent increases in risk.

Nevertheless, even if one accepts the edited versions of the facts presented by health promoters, then their message should perhaps more accurately be: smoking kills relatively infrequently. Moreover, in the case of the possible effect of passive smoking on lung cancer, it is arguable that the effect, if any, is too small to be measured accurately (3).

The fact that a smoker has almost as much chance as a comparable non-smoker of escaping a disease does not itself make that a chance worth taking. One can best see this in an unemotional context that has nothing to do with smoking and the diseases that have been associated with it. Early work on the vaccination of children against polio showed that polio was contracted by 57 in 100,000 unvaccinated children, but only 16 in every 100,000 vaccinated children. This meant that an unvaccinated child was about 3.6 times more likely to contract polio that a vaccinated child. On the other hand, an unvaccinated child had 99.96 per cent of the chance of a vaccinated child of escaping polio. Nevertheless most people did not see this as chance worth taking.

This does not mean that vaccination is pointless. It is mentioned here only because it has been argued that the logic that sees polio vaccination as worthwhile should also suggest that smoking be abandoned. However, this argument ignores what the smoker sees as the benefits of smoking. Thus, in the case of smoking, as opposed to refusing to be vaccinated, some people do see a benefit in it and may value that benefit so highly that they see 99.9 per cent of the chance that a non-smoker has of escaping death from IHD as worth taking. A person's decisions about what risk factors to avoid involve balancing what he or she perceives as their benefits against the chance of falling to what some say might be their consequences.

Facts about the amount of disease attributable to smoking can also be presented in a one-sided way. Consider the lung cancer data cited above. Of the 166 in every 100,000 smokers who die of lung cancer, it could be said that seven of them would have died from that disease even if they had not smoked, since that is the rate at which non-smokers die from it. In other words, we could attribute 159 of the 166 lung cancer deaths among smokers to their smoking. We could paraphrase this by saying that almost 96 percent of all lung cancer deaths among smokers are attributable to smoking, particularly if we want to suggest that smoking is very harmful. But, once again, the same facts can be presented in another way.

Since 159 lung cancer deaths among every 100,000 smokers are attributed to smoking, the proportion of smokers who die from lung cancer attributed to smoking is 159 divided by 100,000, or 0.00159. We could paraphrase this by saying that only about 0.16 per cent of smokers die from cancer attributable to their smoking.

Both paraphrases are correct. Lung cancer is a somewhat rare disease that it is more common among smokers. By reporting only the high proportion of lung cancer deaths that are attributed to smoking, we highlight a possible harmful effect of smoking but suppress the fact that, even so, relatively few smokers get lung cancer attributable to their smoking; slightly fewer than one in every 620 smokers (which follows from the fact that 0.00159 is about 1/628.9). Some would see this as lessening the force of the message that smoking is harmful to health. Similarly, reporting only the one-in-620 figure would mislead by highlighting the rarity among smokers of lung cancer attributable to smoking at the expense of the fact that their lung cancer attributable to other causes is much rarer.

Another feature of the anti-smoking campaign is the suggestion that reducing smoking will correspondingly reduce the costs of treating the illnesses allegedly caused by it. This is likely to be seen as a persuasive argument by a government trying to provide health care at a manageable cost.

At first sight, the argument is appealing. For example, since 96 per cent of all smoker's lung cancer deaths are attributable to cigarette smoking, then there would be a corresponding reduction in the burden of treating lung cancer. But a moment's reflection is enough to uncover the other costs that could also be take into account.

On one side of the ledger, there are the dead smoker's net loss of future earnings and, perhaps, the provision of pensions to surviving dependants. On the other side there are possible savings from the age that pension might be drawn by a surviving ex-smoker and the costs of providing the medical and social services that might be required as he or she ages. There is also the loss to government of tax revenue from tobacco. The direction of the balance of all these costs is not obvious (4). One early suggestion was that smoking has premature ageing effects by increasing the rate of living, so that smokers die from the same diseases as non-smokers but tend to do so at younger ages. For if quitting smoking is the way of postponing illness and death, then the medical and other costs associated with them are not saved but are themselves simply postponed, with the possibility of additional costs being incurred along the way. What this shows is that the issue of thc cost of smoking is something of a red herring.

Suppose, for the sake of argument, first, that smoking is as harmful as has been claimed and, second, that persuading people to quit smoking increases public spending. In those circumstances it would be callous to suggest that people should be left to harm themselves by smoking because it is cheaper than getting them to quit. In other words, if one is convinced that smoking is harmful, then the direction of the balance of costs is irrelevant to the issue of promoting public health by dissuading people from smoking. One would take that action regardless of the net costs.

It cannot be emphasised too strongly that this review of the evidence on the health effects and costs of smoking has not sought to refute the anti-smoking lobby's claim that "smoking kills". Even if one sees the balance of evidence as showing that smoking is harmful to health, it is perhaps not so harmful as is now generally assumed. But when health Promotion becomes a matter of public policy, with politicians trying to modify our behaviour and committing taxpayer's money to that end, the public has a right to have the relevant evidence presented in an undistorted manner. In my judgement, the anti-smoking lobby has failed to ensure this!

( 1 ) This does not mean they are necessarily correct. For example, T D Sterling (1977) , referring to a study of the effects of smoking on morbidity by the US National Centre for Health Statistics, argues that various British, Canadian, and US health reports mistakenly claim that it demonstrated that illness increases with the amount smoked when the study in fact showed just the opposite.

Kristein, M. ( 1985), "Author's Response", Journal of Chronic Diseases, 38: 471-3.

Lalonde, M. (1974), A new Perspective on the Health of Canadians. Information Canada, Ottawa.

Lee, P. (1988)Misclassification of Smoking Habits and Passive Smoking, Springer-Verlag, Berlin.

McGuinness, K. (1989), "No Smoke Without Ire", The Herald , 1Oth November, pp 11-12.

Papier, C. & S. Stellman (1987), "Health Risks of Passive Smoking", Women Health 11 (1-4): 267-77.

Pershagen, G. (1986), "Review of Epidemiology in Relation to Passive Smoking", Archives of Toxicology, Supplement 9, pp. 63-73.

Snow, C. (1962), Science and Government, Mentor/Harvard University Press, Cambridge, Mass.

Sterling, T. (1977), "New Evidence Concerning Smoking and Health", Medical Journal of Australia, 15th October, 99, 538-42.

Watzlawick, P. (1976), How Real is Real, Random House, New York.

World Health Organisation International Agency for Research on Cancer (1987), Environmental Carcinogens: Methods of Analysis and Exposure measurement. Vol. 9: Passive Smoking, Lyon, pp. 69-84,

(1) This is only a rough calculation for illustrative Purposes. The figure of 99.998 per cent can be obtained in two ways: first, by supposing that seven per 100,000 is the death rate among spouses of non-smokers and noting that a 30 per cent increase comes to a death rate of 9.1 per 100,000; second, by supposing that the baseline death rate of seven per 100,000 already represents a 30 per cent increase on a death rate of 5.38 per 100,000. The first method gives the factor of 99 990.1/00 993, or 0.999979, whereas the second gives 99 993/99 994.62 or 0.999984.

(2) I chose the youngest age group of the Table in Breslow & Day ( 1980) because it is the one with the highest per cent increase in risk. These percentage increases themselves decrease with age, from 14.9 for those aged less than 45 years to 1.3 for those aged 75 years or more, The corresponding percentage chances of escaping death from CHD also decrease with age from 99.9 per cent in the youngest group to 99.25 percent in the oldest group, indicating a more harmful effect with age.

(3) Note 1 shows that one is involved with measuring differences in death rates of the order of 1 in 50,000 deaths. Practical limitations in obtaining unambiguous data make this as best a very difficult exercise and errors in death certification and diagnosis may make it well nigh impossible.

(4) It is not obvious how to do the calculations. A useful references is Robinson (1986). A good starting point is Rice, 1986. These authors concluded that smoking had severe economic consequences. On the other hand, Jang et al. (1987) did not show large differences in the rates of medical care or medical costs between smokers and non-smokers. Indeed, it is suggested that the medical cost to smokers were less than those to non-smokers.

Breslow, N. & N. Day (1980), Statistical Methods in Cancer Research, IARC Scientific Publications no. 32, Lyon (Table 2.1. p. 68).

Doll, R. & A. Bradford Hill (1956), "Lung Cancer and Other Causes of Death in Relation to Smoking", British Medical Journal, 1Oth November, pp.1072-81.

Gordon, T. et al. (1977), "Stopping Smoking and CHD", The Lancet, l9th February, p. 421.

Gunning-Schepers, L. et al. (1989), "Population Interventions Reassessed", The Lancet, 4 March, pp. 479-81.

Holme, I. et al. (1987), "Risk Factors and Raised Atherosclerotic Lesions in Coronary and Cerebral Arteries: A Statistical Analysis", Arteriosclerosis 1: 250-6.

Jang, D. et al. (1987), "The Relationship Between Smoking Habit and Medical Cost Among the National Health Insurance Population", Japanese Journal of Public Health 34: 89-94 (in Japanese).

Keys, A. (1970), Coronary Heart Disease in Seven Countries, The American Heart Association Inc., New York (monograph no. 29).

McConnick, J. & P. Skrabanek (1988), "Coronary Heart Disease Not Preventable by Population Interventions", The Lancet, 8th October, PP· 839-41.

Rice, D. et al. (1986), "The Economic Costs of the Health Effects of Smoking", Millbank Quarterly 64(4): 489-547.

Robinson, J. (1986), "Philosophical Origins of the Economic Valuation of Life", Millbank Quarterly 64(1):133-55.

US Department of Health and Human Services (1989), Reducing the Health Consequences of Smoking. 25 Years of Progress: A Report of the Surgeon General, Washington.

Peter Finch has been Foundation Professor of Mathematical Statistics at Monash University, Australia, since 1964. He has contributed to countless scholarly journals including The British Journal for the Philosophy of Science, Information Sciences, The Australian Journal of Statistics, The Journal of the Royal Statistical Society, Acta Mathematica Scientia, Biometrics, The European Journal of Cancer and Clinical Oncology, as well as to such books as The Philosophical Transactions of the Royal Society of London, The Encyclopedia of Statistical Science and The Foundations of Statistical Theories in the Physical Sciences.