Richard Doll and the death of smoking

James A Dickinson 狄堅信

In July Sir Richard Doll died at the age of 92. Many doctors have never heard of him, yet he had a profound influence on the way we behave and practice today. His work with Bradford Hill laid the foundations of our understanding of the effects of smoking. Perhaps his most profound influence on doctors is illustrated by the papers in the British Medical Journal in June 2004.^1,2 One can read them at http://www.bmj.com. There are two papers which are about the same study, and have the same principal author, Richard Doll; but one is new, while the other is reprinted from 1954, 50 years earlier. Both are written in masterful English, easy to read and understand. The earlier one was delightfully sub-titled "a preliminary report". The authors then could hardly have known how true that was.

The British doctors' study was the first great modern epidemiological cohort study, and its features are worth considering. A simple questionnaire went to every doctor on the British medical register, and this was followed up with further medical inquiries every few years. The reasons for choosing doctors were simple and wonderful. These people are all effectively of the same social class, well educated, and aware of the issues, and so able and willing to complete questionnaires. Since doctors must be registered to work, we can be located easily. Because we are all likely to get high quality medical care including investigations, whenever we fall ill and die, our death certificates are likely to be accurate. Consequently the results are not misled by errors in measurement, or confounded by these other social variables.

When the study started, the majority of British doctors were smokers: 87% in the first questionnaire. Initially this frequency reduced the power to show differences between smokers and non-smokers, but over time, as more doctors quit, a new and large group of ex-smokers developed, who could also be followed up. Richard Doll himself was in this group: his own work convinced him that smoking was so dangerous that he should quit.

The first report was interesting, in that it showed the higher death rate of smokers, and relative risk for lung cancer of 40 times that in non smokers. This in itself, might not have been fully convincing, since after all lung cancer was still uncommon, especially early in life. Though report also showed high rates for coronary thrombosis, the authors dismissed smoking as the cause: perhaps just a subgroup "in which tobacco has a significant adjuvant effect". Fifty years later the assessment is different. However, lung cancer accounts for only 5% of deaths of smokers, and one seventh of excess deaths among smokers. Because cardiovascular disease is so common, smokers are still more likely to die of this. Most smokers die of heart disease and strokes, at rates about twice that of non-smokers.

A large study like this is very powerful for demonstrating relationships between subgroups. Cohort analyses show that smoking had less effect in the oldest age groups, those who grew up when early heavy smoking was uncommon. Those in younger cohorts who started when teenage smoking was usual, and heavier smoking was more common, have death rates of three times those of non-smokers. The total increase of risk is such that smokers die at twice the rate of non-smokers in middle age, up to the age of 70. Indeed for men born in the 1920's, who started smoking early and heavily, the death rate in their middle age between 35 and 69 years is 43%, almost three times the 15% rate for the never-smoked. For those over 70 years, being a smoker reduced the chance of surviving to age 90 from 33% to 7%. It may be that for those who started smoking early, life-long smoking trebles their death rate, cutting off at least ten years from their life.

The power of the cohort study also shows the long-term effect for those who quit at different ages. The graphs in the recent paper are wonderful evidence to show our patients. They show that those who quit before age 35 recover from many of the effects of smoking: their death rate is barely distinguishable from non-smokers. At progressively older ages the effect becomes more pronounced: by 50, the risk of death after quitting improves about half. A 60 year old can only gain an average of 3 years of life expectancy by stopping. Quitting is worthwhile at any age: some excess mortality is reduced, but a small excess risk of lung cancer is still present even for those who quit at young ages.

It is reasonable, then, to say that after quitting the risk of cancer remains, but stops increasing; while the risks for other disease, such as heart disease or obstructive pulmonary disease, decrease.

The external validity of this study is a concern. Does this apply to other groups of people besides doctors? There is no biological reason for thinking this is not so, that we react to cigarettes any differently from the rest of society. The factors that select us into medical school are unlikely to cause a bias in our bodies' reactions to smoking. Our diets are the same as our other fellow-citizens, doctors are scattered all over the country in the same way as the rest of the population, and other data from other groups shows similar trends, even if not in such beautiful detail. The number of female doctors when the study started was small, so their results were not analysed, but other studies of women point in the same direction, so we can legitimately apply these findings to anyone.

As evidence accumulated, these and other results provided evidence against smoking, leading to the Royal College of Physicians report of 1962,³ and the US surgeon-general's report of 1964.⁴ I wonder whether doctors took these results to heart because the research applied to them and their colleagues. When I was a child, most doctors smoked. Later when I was a medical student, medical meetings included smokers: indeed no allowances were made for non-smokers. Possibly 30% of my classmates smoked. How things have changed! Now it would be inconceivable for medical meetings to allow smoking, at least in the English-speaking world. On the continent of Europe, and other countries many doctors still smoke, and do so in medical meetings. But there appears to be a downwards trend.

As smoking prevalence has dropped, rates for cardiovascular disease, chronic airways disease and lung cancer also have dropped to equivalent extents; while in countries where smoking rates have increased, so have these diseases increased.

The message is now clear. The effect of smoking is so strong, that if a person is a smoker, nothing else can make as much difference to multiple aspects of their health as stopping smoking. We must encourage our patients to stop as soon as they can. Smoking is a very addictive habit, so many will not appreciate our efforts, but we must try to help them using all the tools at our disposal. These include individual level work such as advice, education, providing nicotine substitutes as appropriate, and supporting policy-level work such as prohibiting advertising, reducing sales, taxing cigarettes. The Hong Kong Council on Smoking and Health has done sterling work, and we must encourage our colleagues in China to do the same too.

We salute Sir Richard Doll, who played such a central role in helping us to understand such a major health problem. I hope this tribute does him justice. In our own small way, each of us must continue to apply his insights.