This research extends the previous study of Connecticut population-based cancer incidence trends for purposes of monitoring, etiology and data quality. Where sample size permits, each type of analysis will be applied to 20 cancer sites: a) Population-based rates will be modeled by age, period and cohort in order to (i) compare Connecticut mortality with Connecticut incidence regarding the relative strengths of period and cohort effects. (ii) compare Connecticut mortality trends with those for the entire United States to address the question of the generalizability of long-term Connecticut incidence trends; (iii) examine Connecticut rates for carcinoma in situ, with emphasis on data since 1965; and (iv, with a common set of assumptions, compare long-term cancer incidence trends among 16 regions of the world. Trends among racialethnic groups will also be examined. b) Statistical methods for survival analyses will be applied to study (i) trends in survival, (ii) trends in the occurrence of second primary invasive cancers in those with a first primary, and (iii) the occurrence of invasive cancer in those with carcinoma in situ. Differences in trends between whites and non-whites will also be examined. c) Two types of analyses pertain to the concept of a linear relationship of log (age) to the log (cancer incidence rate) and mathematical models of carcinogenesis: (i) The age distribution of Connecticut cancer incidence rates will be examined by adjusting for time trends. Age distributions so obtained will be fit to a linear relationship of log(incidence) with log(age) and curvature estimates computed, as was previously accomplished on cross-sectional data by Cook et al (1969). (ii) In order to deal with the estimability problems in age-period-cohort models, the slopes of the birth cohort and period trends will be estimated by assuming a linear increase of log(incidence) with log(age). For each cancer, the assumption for the age distribution will be examined and may be modified in light of current epidemiologic data and knowledge of carcinogenic mechanisms. The slope so obtained will be compared with the slope from other """"""""reasonable"""""""" assumptions, such as a zero linear effect for time period. d) Approaches toward the synthesis of incidence, mortality and survival in Connecticut population-based cancer data will be explored.
