OnCo
key papersKey paper

Doll and Hill 1950: the case-control study that tied smoking to lung cancer

Comparing 649 men with lung cancer to matched hospital controls in London, Doll and Hill found that almost none of the cancer patients were non-smokers and that risk rose steeply with the amount smoked.

Richard Doll and Austin Bradford Hill interviewed patients with lung cancer and matched controls with other diseases in 20 London hospitals between 1948 and 1949. The preliminary report analysed 649 men and 60 women with lung cancer.

Only 0.3% of male lung cancer patients were non-smokers, compared with 4.2% of controls, and heavy smokers (25 or more cigarettes a day) were far over-represented among cases. The authors concluded that smoking was a factor, and an important factor, in the production of carcinoma of the lung, and set out the case for a prospective study, which became the British Doctors Study in 1951.

Along with Wynder and Graham's US study the same year, it launched the modern epidemiology of tobacco and eventually the tobacco control policies that have prevented more cancer deaths than any treatment.

Observational studyChanged practice1,465 participants
Authors
Doll R, Hill AB
Published
BMJ, 1950
What it found
  • Non-smokers: 2 of 649 male lung cancer cases (0.3%) vs 27 of 649 controls (4.2%)
  • Proportion of heavy smokers (25 or more cigarettes a day) was about twice as high among cases as controls
  • Association was specific to lung cancer, not to other cancers or diseases in the same hospitals
  • Estimated risk in heavy smokers about 50 times that of non-smokers
What it means

Doll and Hill's 1950 study is where the evidence that smoking causes cancer begins. Everything from cigarette warnings and tax to smoke-free laws and lung screening eligibility descends from this study and the cohort that followed it.

Be careful
  • Case-control design with hospital controls; recall and selection bias were the main criticisms at the time
  • Causation was not accepted by many, including Fisher, until the prospective cohort and animal data accumulated
  • Women were too few for separate analysis
  • Exposure was self-reported and unverified

Key papers

1top

Connected

9top

Pages like this

not linked directly; found by shared links