ABSTRACT
This chapter discusses a few models for estimating absolute risk, pure cumulative risk, and relative risk from family-based studies, such as kin-cohort studies and studies of families ascertained because they have several diseased members. Key considerations include whether the model allows for residual familial correlations apart from those attributable to the genes that are explicitly included in the model and how the family was ascertained, which determines the likelihood equations. The chapter compares several genetically based models and an empirical model, the Breast Cancer Risk Assessment Tool (BCRAT) for projecting breast cancer risk. Some of these models, such as BCRAT project absolute breast cancer risk, whereas others project pure breast cancer risk that treats competing risks as random censoring. The emphasis is on breast cancer models because various family-based designs have been used to study breast cancer risk, and because both strongly associated mutations and weakly associated genetic markers (SNPs) have been identified for breast cancer.
