Doll and Hill smoking and lung cancer case-control study

Deaths from lung cancer in Britain rose steeply in the first half of the twentieth century, and the rise demanded explanation. Several hypotheses circulated: atmospheric pollution from coal and motor exhaust, the tarring of roads, general urban air quality, and the increasing use of tobacco. Richard Doll, then working for the Medical Research Council, and Austin Bradford Hill designed a case-control study to test these explanations against each other. They recruited patients admitted to London hospitals with confirmed lung cancer and matched them with controls admitted for other conditions during the same period.

Interviewers questioned 709 lung cancer patients, 649 men and 60 women, and 709 general medical and surgical patients of the same sex and similar age, gathering detailed smoking histories. The association with smoking was stronger than any of the other exposures examined. Only 0.3 percent of the men with lung cancer were non-smokers, against 4.2 percent of the male controls, and risk rose with the amount smoked: those who smoked 25 or more cigarettes a day had about 25 times the risk of non-smokers. The study appeared in the BMJ on September 30, 1950. Four months earlier, Ernst Wynder and Evarts Graham had reported a study of 684 proven cases in JAMA that pointed the same way.

The response from the scientific community was mixed, and the tobacco industry's response was hostile. Critics raised methodologically legitimate concerns: case-control studies were susceptible to recall bias, and an unmeasured confounding variable could in principle explain the results. The tobacco companies funded alternative hypotheses and cast doubt on the epidemiological evidence through the 1950s and into the 1960s, a campaign later documented through internal industry documents obtained in litigation. Doll himself was a smoker at the time of the 1950 study and stopped after reviewing the data.

To address the objections, Doll and Hill launched a prospective cohort study in 1951, sending a questionnaire about smoking habits to every doctor on the British Medical Register; the analysis followed 34,440 male doctors. That cohort was followed for 50 years. The doctors study provided dose-response data, showed that stopping smoking reduced risk over time, and extended the evidence to heart disease, chronic obstructive pulmonary disease, and other conditions. It supplied the kind of prospective evidence that the critics of the 1950 case-control design had demanded.

The U.S. Surgeon General's report of January 11, 1964, whose committee reviewed more than 7,000 scientific articles, concluded that cigarette smoking is causally related to lung cancer in men. The case-control method Doll and Hill used in 1950, efficiently matching cases to controls to study outcomes too rare for a cohort of practical size, is still a standard design in chronic disease epidemiology.