Role of Human Population Study Centers in Studies of Cancer Etiology

Abstract

RECENTLY interest has quickened in the use of centers as sources of data for medical research. The idea is not new. An early example of studies was the series of investigations in Hagerstown, Mid., initiated by the Public Health Service in 1921 (1). Other such studies have been supported in part by National Institutes of Health research grants. The Framingham, Mass., survey of heart disease (2) and the Veterans Administration-National Cancer Institute of smoking and health (3) were National Institutes of Health intramural activities. To assess their role in the support of centers the NIH Division of Research Grants Study Sections concerned with grant applications sponsored the Conference on Health Studies of Human Populations at the University of Pittsburgh School of Public Health in November 1960 (4). The question now is what steps the National Cancer Institute should take in the further development of centers. Any review and appraisal of centers is conditioned not only by events past but by the expectation of things to come and so must be regarded as speculative judgments not unlike those involved in predicting the future course of business and market activities. Humane and ethical considerations deprive investigators of the distribution of disease in man of many advantages conferred by experimental work with pedigreed animals, such as precise control of environmental and related factors, detailed and sophisticated comparative measurements in experimental and control animals, and the relatively large numbers of subjects available for experiments. Studies of disease in man must rely heavily on observations and on the substitution of statistical for experimental control. Situations, then, must be sought which improve the power of the observational method, usually through resort to more detailed cross-classifications of data and more elaborate determinations of the sequence of events, in order to extend the range of inferences permitted by the observational associations. Many investigators look upon studies as a specialized activity which has its own techniques and discipline. In a statistical context, this distinction arises from recognition that statistical methods suitable for analysis of experimental data are often not appropriate for handling observational data. The human center label has been applied to a number of different concepts so that it is difficult to obtain general agreement on all features of these centers. At the outset it would be well to distinguish between populations for continuing study and community-based population facilities. Mr. Haenszel is chief of the Biometry Branch and Dr. Miller is chief of the Epidemiology Branch, National Can7cer Institute, National Institutes of Health, Public Health Service.' This paper was prepared for presentation at a meeting of the Consultant Panel on Biometry and Epidemiology of the National Cancer Institute on November 5-6, 1961. While the present text benefits from the comments of panel members and other reviewers, the authors assume responsibility for the views expressed.