Comparison of Trends in Cancer Incidence and Mortality in Osaka, Japan, Using an Age-Period-Cohort Model

Background. We aimed to estimate the effects of age, period and birth cohort on trends in cancer incidence and death for all sites and selected sites of cancer in Osaka using an age-period-cohort model. Methods. Cancer incidence data during 1968-2003 were obtained from the Osaka Cancer Registry, and cancer mortality with population data in Osaka during 1968-2007 were obtained from vital statistics departments. We estimated age, period and birth cohort effects for incidence and mortality using Nakamura's Bayesian Poisson age-period-cohort model. Results. For most sites of cancer, linear ageing effects were observed, the exceptions being breast and cervix which levelled-off at around 40 years old, while period effects were small. Decreasing cohort effects were observed in stomach and liver cancer. Cohort effects peaked at the generation born in the early 1950s for colorectal, lung, breast cancers. For most sites of cancer, incidence and mortality showed similar trends, but in the late cohorts for cervical cancer, cohort effects decreased in mortality, while increasing in incidence. Conclusion. Period effects reflecting immediate effects to cancer incidence and mortality, such as development of the effective treatment and screening programme were stable in most sites of cancer. Cohort effects influenced by long-term risk factors were prominently observed for every site, decrease in stomach and liver cancer cases being related to reduction in risk factor prevalence. Cancer control activities could be evaluated through the results, indicating utility for future cancer control planning.