The primary purpose of this project is to describe and evaluate the distribution of cancer occurrence in the United States in terms of age, sex, race, place, and time in order to identify subgroups of the population that offer possibilities for mortality reduction through intervention. Cancer incidence, mortality, and survival rates are analyzed using biostatistical techniques such as log-linear models or models derived from the multistage theory of carcinogenesis. Incidence and survival data from the SEER Program and mortality data from the National Center for Health Statistics together with other data sources are often the focus of such analyses. Trends in the incidence of mesothelioma diagnosed during 1975-84 were evaluated using age-period-cohort Poisson regression models. The average annual age-adjusted incidence rate among males increased from 5.9 to 7.6 per million population during this time. Poisson regression analysis indicated that the incidence of pleural mesothelioma peaked for the male cohort born during 1905-09. The proportional hazards model is being used to evaluate the prognostic significance of serial antibody titers to Epstein-Barr virus antigens in treated nasopharyngeal carcinoma patients. Analyses of serial antibody titers are often complicated by censorship of the response variable and by incomplete knowledge of the patient's marker state. Both of these problems are accommodated by the proportional hazards model with titer measurements treated as a time-dependent covariate. In a series of 48 nasopharyngeal carcinoma patients who provided 288 serial serum samples, an increased risk of death was found to be significantly associated with the level of the most recent titer prior to death.
