Relationships of age, period, and birth cohort for stomach cancer mortality in Japan.

The mortality data on stomach cancer in Japan during 1955 to 1980 were analyzed by the use of an age-period-cohort model. Although the model could not give a unique solution of age, period, and cohort parameters, several sets of estimates of the parameters clarified particular features of each effect. Within a plausible range, the features were as follows. The period parameter underwent linear change from 1965 to 1980 for males and from 1970 to 1980 for females. The cohort parameters decreased birth cohort by birth cohort, at least among those who were born after 1891. The curve of the age parameter in females showed a shoulder around age 40 and a subsequent upward slope to age 80, while that in males was unimodal. The curve in females was compatible with the idea that there might be two different main entities in stomach cancer. The model also gave estimates of future trends in stomach cancer mortality under the assumption that there would be no additional change in period effect other than that expected from the recent quinquennia. The number of deaths from stomach cancer in males would be rather stable until 2000, and that in females would decline slightly. The age-adjusted and crude mortality rates would continue to decrease at least until 2000.