Determining Effects of Genes, Environment, and Gene X Environment Interaction That Are Common to Breast and Ovarian Cancers Via Bivariate Logistic Regression

摘要

A new method for the simultaneous genetic analysis of two or more discrete traits such as the presence of breast and ovarian cancers in twins was developed. A generalized estimation equations (GEE) logistic regression model was used for the modeling. A shared trait is defined for two discrete traits based upon explicit patterns of trait concordance and discordance within twin pairs; this shared trait is assessed for the influence of additive genetic and/ or common environmental effects. Data are summarized in the form of 2 x 2 tables (for monozygotic and dizygotic twins) by combining appropriate cells form the 16-cell multinomial distribution to define the individual and shared trait. Hypothesis tests for additive genetic and common environmental influence are performed using repeated measures logistic regression via the GEE approach. The model specification is highly flexible accounts for the correlated structure of the parameter estimates and does not require multivariate normality assumption for the underlying liability distribution. The approach was applied to data sets form the Vietnam Era Twin Registry and the Mid Atlantic twin Registry. Currently, efforts are being taken to collect adequate data on cancer outcomes that will provide enough power to apply this methodology to twins with cancer.