This Article Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Doll, R. Search for Related Content PubMed PubMed Citation Articles by Doll, R.

Interpretations of Epidemiologic Data^*

( Medical Research Council's Statistical Research Unit, University College Hospital Medical School, Gower Street, London, England)

Epidemiological observations have a special part to play in cancer research because of the limited possibilities for human experiment. Such observations cannot prove that an agent is the cause of disease in man—not even when coupled with the results of animal experiments—and decisions with regard to etiology usually have to be made on the basis of evidence that is logically incomplete. In these circumstances it may be desirable to attempt to prevent the disease in the manner suggested by the evidence and to see whether the results of the action support the conclusions on which it was based. This policy has paid well in regard to occupational cancers, and examples are given illustrating how similar epidemiological observations can be used to test the validity of current opinions concerning the etiology of other cancers.

For most cancers epidemiological evidence is limited to showing that the incidence varies from place to place or from time to time, and it has so far failed to link these differences with differences in the prevalence of any specific agent. The discovery of such links is the greatest contribution that epidemiology can hope to make, but it may also be able to contribute in another field that has been regarded as the province of the laboratory worker—that is, in the study of the mechanism of carcinogenesis.

Examples include the observation that, after a single exposure to radiation, the incidence of leukemia rapidly rises to a maximum and then slowly falls; it is possible therefore, that those cancers that normally show a similar type of distribution with age are also due to brief periods of exposure to the agent concerned. Prolonged exposure, as occurs with cigarette smoke, gives rise to a continuous increase in incidence which is similar to that seen with most of the common epithelial cancers. In these cancers the incidence increases approximately in proportion to the sixth power of the age; but other mathematical relationships can be adduced, and it is dangerous to build theories on this relationship alone. Data have been obtained for the relationship between the intensity of exposure and the resulting incidence, and for the effect of ceasing exposure and of starting exposure at different ages. All these results must be taken into account in constructing an adequate model. If they sometimes appear to conflict with laboratory experiments it may be because the conditions of life for laboratory animals do not reflect the complexity of the human environment.

^* The Gordon Richards Memorial Lecture given at the annual meeting of the American Association for Cancer Research, Toronto, Canada, on May 23, 1963.