Age-period-cohort analysis of cervical cancer mortality in Taiwan, 1974-1992.

OBJECTIVES To develop a hypothesis about the carcinogenesis of cervical cancer from a descriptive analysis. METHODS The mortality data of cervical cancer were analyzed over the period from 1974 to 1992 among Taiwanese women using a log-linear Poisson model modified from the method of Osmond and Gardner to examine the effects of age, calendar period of death, and birth cohort on cervical cancer mortality. RESULTS This age-period-cohort model provides a summary guide for interpretation of cancer mortality trends. According to this model, age was found to be the strongest factor predicting cervical cancer mortality. Women in 50-54 age group have 89.3-fold risk of cervical cancer compared to those in the 30-34 age group. The cohort effect is also of particular interest because the generation at greatest risk for cervical cancer is the one born between 1893 and 1938, and a dramatically declining trend is observed thereafter for 1938-1963 birth cohort. Interest has emerged about the increasing trend in recent cohorts (after 1963 birth cohort). However calendar time only has a slight effect in the APC analysis. The model also identified a possible role of female sex hormones as the age effect, promiscuous sexual activity as the period effect (promoter) and the change in reproductive behavior as the cohort effect (initiator). CONCLUSIONS These results may help to develop a hypothesis of carcinogenesis of cervical cancer in Taiwan.