Printed in U.S.A. AGE DEPENDENCE OF COHORT PHENOMENA IN BREAST CANCER MORTALITY IN THE UNITED STATES

Breast cancer mortality has increased in most parts of the world, and many explanations have been postulated. In this paper, the authors examined the evolution of mortality rates for white and nonwhite females in the United States from 1950-1979. Using both graphic techniques and Poisson regression models, they found that there has been strong modification of apparent cohort effects by age. For both white and nonwhite females, they observed an increase in mortality rates limited to the postmenopausal ages. breast neoplasms; mortality; regression analysis Using rates for white females from the Connecticut Cancer Registry, MacMahon (1, 2) reported that time trends in breast cancer incidence were occurring in a pat-tern that indicated an important contribu-tion of attributes related to the year of birth; he noted that incidence rates were slowly increasing, although younger age groups experienced a leveling off. Stevens et al. (3), examining data from the United States, Canada, England and Wales, and Japan, concluded that the temporal evolu-tion of breast cancer mortality has been similar among these countries. In all of the populations that these investigators re-viewed, the cohort effect parameters fell until the birth cohort born around 1900, followed by a steady increase thereafter. Received for publication May 5, 1986, and in final form December 24, 1986. Abbreviation: GLIM, generalized linear interactive modeling