DOLL, BRADFORD HILL, AND LUNG CANCER

17 May

I want to end my account of delayed progress in medicine with a final case study. By 1948, when streptomycin was shown to be effective against tuberculosis, it looked as if all the major infectious diseases had been or would soon be conquered. Since then HIV has emerged as a major threat, and other diseases, such as tuberculosis, have developed drug resistance. In our hospitals, death rates from post-operative bacterial infections are rising because of the spread of MRSA (methicillin-resistant staphylococcus aureus), and doctors and nurses, used to relying on antiseptics and antibiotics, are finding it hard to learn new disciplines when it comes to washing hands and changing clothes. A world flu pandemic, we are told, is an imminent possibility. Nevertheless, if we look at the second half of the twentieth century in the developed world then we can think of it as the period after the defeat of the contagious diseases. And as deaths from contagious diseases fell, then of necessity more and more people died from non-contagious diseases: from cancer, heart attacks, strokes. As fast as people suffering from tuberculosis were moved out of hospital wards, people suffering from lung cancer were moved into them. As doctors turned their attention to these diseases that were far from new, but were suddenly much more important than they had been in the past, they had to abandon the safety of the germ theory of disease which had been the key factor in progress for almost a cen- tury. How to make sense of diseases that had no known cause? The first, and still perhaps the greatest, breakthrough was in the study of lung cancer. In 1950 Richard Doll and Austin Bradford Hill published the first major study demonstrating that smoking was the principal cause of lung cancer. I began to take an interest in that article because I read an interview with Richard Doll who said that nobody paid much attention to their research when it was first pub- lished; it was only in 1954, when they published a quite different study, or even 1957, when the government agreed that smoking caused lung cancer, that people began to take them seriously. Here, I thought, must be yet another example of bad medicine; as late as 1950 it was evidently still impossible to persuade doctors to take statistical arguments seriously. The true story, however, is rather different from the story Doll liked to tell in 2004. The interviews with Doll that accompanied the fiftieth anniversary of his 1954 study (and the completion of the research programme announced in that study), and the obituaries that appeared when Doll died the next year consistently misrepresented the true story of the impact of the early work of Doll and Bradford Hill. If I had to summarize that story in a single phrase it would be not ‘yet more bad medicine’, but ‘at last, good medicine’. A consideration of that study provides a fitting end to the main story of this book. Of the two authors of the 1950 study, Bradford Hill was the senior and the better known. He had published The Principles of Medical Statistics in 1937, the first textbook on the subject. It is a mark of the difficulty of teaching doctors statistics at that time that in his book he carefully avoided using the word ‘randomization’ in the belief that it would scare doctors. But it was Bradford Hill who designed the first randomized clinical trial, conducted in 1948, on streptomycin, which was in such short supply that it had been decided that it was ethical to select by lottery the patients who should be treated with it. This was an early double blind test, where neither doctors nor patients knew who was receiving the new drug. Streptomycin was shown to be highly efficacious. (‘Efficacious’ is here a term of art: drugs are effica- cious if they cure patients in trials; they are ‘effective’ if they cure patients in the real world. The two are not the same, as patients in trials are not normally very old, very young, or even very sick.) The streptomycin trial has been the model for all drug testing since that date. Bradford Hill and his new colleague Richard Doll (who had previously carried out a study of the effectiveness of different treat- ments for ulcers) then turned to an attempt to find out the causes of lung cancer. Lung cancer rates were rising sharply: in males, the rate had increased twentyfold between 1905 and 1945. In 1950, the num- ber of deaths from lung cancer, at 13,000, exceeded for the first time the number of deaths from tuberculosis –– which had been falling for some time. But the cause of the increase in lung cancer was a mystery. Doll himself suspected that the asphalting of roads might be respon- sible. Cigarettes were not suspected. Although in Britain legislation in 1908 had banned the sale of cigarettes to children under 16, this was because cigarettes were thought to stunt growth and so render young people unfit for military service. There was no generally accepted view that they were bad for the health of adults, indeed many claimed that they were good for you. In order to clear up the mystery as to why lung cancer rates were on the increase Doll and Bradford Hill (Doll was to be the lead investigator) devised a detailed questionnaire to be administered to patients suspected of having lung cancer –– in all 709 patients in London hospitals were interviewed, of whom 649 were men. They were matched with a control group of 709 hospital patients who were as like them in every respect as possible, except for the fact that they did not have lung cancer. In the long list of factors surveyed, one quickly stood out. Of the male lung cancer patients only 2 were non- smokers; in the control group 27 were non-smokers. Of the female lung cancer patients 19 were non-smokers, while in the control group 32 were non-smokers. A smoker, it should be said, was defined very broadly, as someone who had smoked at least one cigarette a day for at least a year in the course of their life. Since we now know smoking causes a number of diseases, smokers will have been disproportion- ately present in the control group. A fair guess would be that 80 per cent of adult men smoked, and 40 per cent of adult women; the men averaging 15 cigarettes a day, and the women half as many. Doll and Bradford Hill calculated that if there was no statistical connection between smoking and lung cancer, and thus if the difference between the two groups was only a matter of chance, then one would have to conduct the trial more than a million times for a difference on this scale to occur once. These rather small numbers –– 21 non-smokers amongst the lung cancer patients, 59 in the control group –– amounted to proof of a causal connection. Further evidence also showed that the more you smoked the greater the risk: their initial estimate was that heavy smokers were fifty times more likely to die of lung cancer than non-smokers. One criticism made of this first study was that the results might have been skewed in some way because all the patients in the study came from London. Doll and Bradford Hill quickly conducted a larger survey including patients in Bristol, Cambridge, Newcastle and Leeds, publishing the results in December 1952. These two studies mark the completion of the first phase of their research. Doll’s and Bradford Hill’s early work had some critics, including R. A. Fisher, the first statistician to advocate random trials (Fisher’s area of expertise was agriculture rather than medicine, so his were trials of seeds rather than drugs). Fisher pointed out that the fact that there was a statistical correlation did not mean there was a causal link: people with grey hair tend to have short life expectancies, but this is not because grey hair causes death; it is because old age is one of the causes of grey hair, and old age is a major cause of death. There is a real correlation between grey hair and death, but not a causal link. So there might be a genetic trait, for example, which made one both disinclined to smoke and relatively immune to lung cancer. But the crucial fact about Doll’s and Bradford Hill’s first two publications is that they met with widespread acceptance amongst medical experts. As early as the middle of 1951 the Secretary of the Medical Research Council (which had funded their work) was prepared to state that ‘the case against smoking as such is proven’ and that there was no need for further statistical work. Within the new National Health Service, however, there existed a body called the Standing Advisory Committee on Cancer whose job it was to advise on government policy. This committee was reluctant to accept Doll’s and Bradford Hill’s conclusions, and even more reluctant to see any action based on them. They therefore called in an independent committee of experts, chaired by the government actu- ary, to reassess the evidence. In November 1953, this committee unequivocally backed Doll and Bradford Hill. As a result, on 12 February 1954, the minister of health, Iain Macleod, announced to the House of Commons that there was a ‘real’ link (the word ‘real’ was slightly equivocal –– he did not say there was a causal link) between smoking and lung cancer. He went over the same ground at a press conference the same day, smoking as he did so. At that confer- ence the minister tried to walk a tightrope. He had been advised that ‘It is desirable that young people should be warned of the risks appar- ently attendant on excessive smoking’, but he insisted that ‘the time has not yet come when the Ministry should offer public warnings against smoking’. In other words, the Ministry had accepted that the case against smoking was proven, but they planned to do nothing about it. In this they were following the general policy of the SAC on Cancer, which held that public education about cancer would only provoke anxiety without saving lives. It is important to stress that the argument that smoking causes cancer had been won by February 1954, because later in life Doll himself used to claim that their research was not taken seriously until later in 1954, with the publication of the first results of an entirely new project, and as a result the 1954 announcement is often presumed to have occurred later in the year, or is confused with an announce- ment made in 1957. In both 1954 and 1957 the minister announced that smoking was linked to lung cancer, and on both occasions he is supposed to have smoked through the press conference –– this story may be true of both occasions, or the two events may have become hopelessly confused. According to Doll’s later account, he and Bradford Hill devised a new statistical study because nobody had taken their early work ser- iously. The very design of their study shows that this account is wrong. It was funded by a government-appointed body, the Medical Research Council. Its basis was a questionnaire sent in October 1951 to every doctor in the country, all 60,000 of them, by the British Medical Association –– it was thus endorsed by the organization that represented all doctors. And finally, it required the active involvement of the Registrar General of Births, Marriages and Deaths, represent- ing another government agency. The second phase of their research was thus only possible because their work had been taken extremely seriously. Some two-thirds of the doctors approached by Doll and Bradford Hill responded to the questionnaire. The Registrar General then sent Doll and Bradford Hill the death certificate for every doctor who died. By March 1954, 36 doctors who smoked had died of lung cancer, and no non-smoking doctor had done so –– since 12.7 per cent of the doctors in their study were non-smokers, nearly 4 should have done so. Their second study thus followed a population over time and showed that smokers died younger than non-smokers. By 1954, Doll and Bradford Hill were already able to point to eleven studies (starting with their own study of 1950) linking smoking and lung cancer. Fifty years later, when the study was finally concluded (the number of doctors alive in 1954 was rapidly diminishing by this point), Doll and Bradford Hill had shown that smoking reduced life expectancy by approximately ten years. They also were able to show by 1956 that stopping smoking, particularly stopping smoking when young, significantly extended life expectancy. Had smoking been banned in 1950, therefore, the whole population (of whom about two-thirds were smokers) would have gained about six years in life expectancy, a gain probably greater than that achieved by the whole of medical science prior to that date. As it is, although lung cancer rates in the UK amongst males peaked in the 1960s as many smokers began to give up, the overall incidence of lung cancer now is still higher than it was in 1950. In 2001 there were 37,500 new cases of lung cancer in the UK. The accumulation of this new research meant that in 1957 the UK government, although now facing sustained lobbying from the tobacco industry, was finally prepared to accept a ‘causal’ relationship between smoking and lung cancer –– the US Surgeon-General reached the same conclusion at the same time. The UK government’s view was based on a report from the Medical Research Council, which had supported Doll’s and Bradford Hill’s work from the beginning. In fact, it would have been easier to get a straightforward statement on smoking and lung cancer out of the MRC in the autumn of 1952 than it was in 1957: from December 1952 on (partly as a result of the great London smog of 5–12 December 1952 which was thought to have killed 12,000 people) attention increasingly focused on atmospheric pollution as a likely cause of lung cancer, the incidence of which was greater in cities than in the countryside. The BMJ editorial which accompanied Doll’s and Bradford Hill’s publica- tion of 13 December 1952, and which was probably written during the great smog, calculated that air pollution was responsible for roughly 17 per cent of lung cancer deaths, with the rest being attrib- utable to smoking –– but it concluded with a call for government action against air pollution, not smoking. The conviction that smog caused disease and death soon led to the Clean Air Act of 1956. In 1957 the MRC initially wanted to issue a statement saying that smok- ing was the main cause of lung cancer, but that as much as 30 per cent of lung cancer might be caused by atmospheric pollution –– a view the government wanted to prevent being articulated, since the gov- ernment, it was thought, would be held responsible for pollution but not for smoking. At the time no one seems to have made the obvious argument that smokers pollute the air that other people breathe: the key studies showing that passive smoking increases the risk of lung cancer were not made until 1981. When the MRC’s report finally emerged, the British Medical Journal at last called for ‘the dangers of smoking’ to be ‘brought home to the public by all the modern devices of publicity’. The government now began, extremely hesitantly, to act: anti-smoking literature was made available, although no real campaign began until the Royal College of Physicians reported in 1962 that smoking caused cancer. By this time, indeed, it was becoming apparent that a range of diseases were associated with smoking. Health warnings were placed on cigar- ette packets in the US in 1965, but in the UK they were not required until 1971 when experiments on beagle dogs finally showed that you could induce lung cancer by making them breathe cigarette smoke. In the mid-1970s it finally became UK government policy to discourage smoking by raising the tax on cigarettes. By 2000, less than 30 per cent of British men were smokers; the decline in the US had been com- parable. Half that decline had taken place by 1970, and thus before the full campaign of health warnings and tax increases had begun. There are various responses one may reasonably have to this story. One can lament the vast number of lives –– worldwide, hundreds of millions of lives –– lost because of the failure to ban tobacco, or at least to ban smoking in public places. One can regret that when govern- ments finally acknowledged the nature of the problem, in 1957, the action they took was too little, too late. But you can scarcely argue that the original work undertaken by Doll and Bradford Hill was unreasonably opposed, ignored, or misunderstood. What is impressive is not the small amount of obstruction they faced from the SAC on Cancer (which was opposed to public education on cancers in general), but the speed with which the medical establishment –– the BMA, the BMJ, the MRC, and those with close links to the medical establishment, such as the government actuary and the Registrar General –– acknowledged the importance of their research and sought to foster it. It was this support that made it possible for Doll and Bradford Hill to strengthen their arguments in 1952, 1954, and 1956. The real puzzles are elsewhere. Late in life, Doll said of their 1950 research, ‘When we showed the results to Sir Harold Himsworth, the [Medical Research] Council’s Secretary, he said it would have a huge impact. We really thought people would give up immediately.’ He himself had given up smoking a few months before –– he found it quite easy to do so. Instead the results were completely ignored by the press. On 14 October, a fortnight after Doll’s and Bradford Hill’s pioneering publication, a letter from D. J. Parr was published in the British Medical Journal. It is worth quoting in full: Sir, Readers of the popular press are regularly treated to sensational news items based on gleanings from the Journal or one of its con- temporaries, often before the publication reaches its medical sub- scribers. In view of the wide range of subjects covered in this way, we must wonder why the usual publicity has not so far been given to the conclusions of Dr. R. Doll and Professor A. Bradford Hill on ‘Smoking and Carcinoma of the Lung’ (September 30, p. 739) –– a topic of much more general interest than the design of perambulators, the dangers of staying in bed, or many others recently in the headlines. Do newspaper editors fear that their public may resent being disillusioned about the ‘harmlessness’ of the tobacco habit, or are they a little unwilling to risk the displeasure of some of their major advertisers? –– I am, etc. In the very issue in which Doll and Bradford Hill’s first research was published, an editorial discussing their work included a wry joke: ‘It is said that the reader of an American magazine was so disturbed by an article on the subject of smoking and cancer that he decided to give up reading.’ Most journalists were smokers, and far from keen to think about smoking objectively; most of their readers were in the same position. Anti-smoking stories were not likely to sell news- papers to smokers. Another correspondent in 1950, Lennox Johnston, pointed out that smokers could not be trusted to make rational decisions about smoking for the simple reason that they were addicts. He objected to the very language of Doll’s and Bradford Hill’s paper: I take the investigators to task for their 36 references to the habit factor in smoking whilst completely ignoring the much more important crav- ing factor . . . It is . . . a violent euphemism to refer to smoking as a habit. Tobacco smoking is a drug addiction (to be quite precise, a means of administering a drug of addiction, nicotine), and the drug addictions are specific diseases –– specific intermittent intoxications . . . Since tobacco smoking is a disease and a preventable one, it is our plain duty to prevent it. Two years later, when Doll and Bradford Hill published their second paper, another correspondent wrote in to make a similar point. Since a significant proportion of smokers were addicts, there was no point in waiting for people to voluntarily give up smoking. What was needed was a ban on smoking. Had the full significance of these letters been understood, more than twenty years of delay could have been avoided. It was not until the 1980s that the medical profession’s attention finally began to focus on the addictive element in smoking. If there was a major intellectual failure by doctors in 1950, it was not in any failure to understand Doll’s and Bradford Hill’s research; it was in their failure to recognize that smoking was not a habit but an addiction, and that there was therefore nothing straightforward about persuading people not to smoke. Doll, who suffered from this intel- lectual failure just as much as everyone else did, never faced it: instead he invented a myth where nobody understood the importance of his and Bradford Hill’s work until 1957. But the truth is that by 1950 medicine was becoming powerfully progressive, and new knowledge was eagerly seized upon. Lung cancer proved the first of the non- contagious diseases to be largely preventable, and progress in other fields was much slower. In a hundred years time, historians looking back may identify missed opportunities and unnecessary delays. It’s too soon to tell. But the story of this one discovery suggests doctors were quick to appreciate its significance, even if their failure to under- stand addiction meant that they allowed the government to waste twenty years before taking effective action. Looking back more than fifty years later, Doll thought no one had understood his early work. The truth is they understood it rather well. They just did not know how to act on it.

Random Posts

Comments are closed.