Mathematical modeling of cancer incidence rates: Linear models of colon cancer in connecticut

1980; Pergamon Press; Volume: 33; Issue: 11-12 Linguagem: Inglês

10.1016/0021-9681(80)90061-2

ISSN

1878-0679

Autores

Marcia A. Testa, J. Wister Meigs, John Flannery,

Tópico(s)

Statistical Distribution Estimation and Applications

Resumo

Abstract Mathematical models of cancer incidence were developed to evaluate substantively trends in age and period of diagnosis and to identify differences in these trends with regard to qualitative variables such as sex and anatomical localization of tumor. Parameter estimates were obtained through a weighted least-squares procedure, and hypothesis testing was carried out using minimum modified χ 2 statistics. These procedures were applied to colon cancer incidence rates for Connecticut males and females during the period 1940–1974. The initial model of changes in incidence over calendar time for males and females showed a different pattern and excess risk among males and a possible birth cohort effect among females. The changing pattern of risk in males was modeled as an exponential function of age and a two-stage, age-related process was identified. While proportional comparisons of time trends between the ascending and sigmoid colon showed significant differences for certain groups, patterns in absolute changes were not found to be statistically different. Using mathematical models of cancer incidence rates for substantive analysis can be useful for hypothesis generation in cancer epidemiology. Such modeling proves worthwhile because it allows the investigator to summarize the trends and patterns of cancer incidence through model parameters which then may be tested statistically for the purpose of identifying significant effects of independent variables.

Referência(s)