Duke University - Institute of Statistics & Decision Sciences
Recent advances in the understanding of genetic susceptibility to breast cancer, notably identification of the BRCA1 and BRCA2 susceptibility genes, raise important questions for clinicians, patients and policy makers. Answers to many of these questions hinge on accurate assessment of the probability of carrying a genetic susceptibility mutation. In particular, it is important to predict genetic susceptibility based on easy-to-collect data about family history of breast and related cancers.
In this talk I will describe a simple prediction model of genetic susceptibility. Our model is based on combining published data about the genes' prevalence, penetrance and inheritance mechanism, together with a straightforward application of Bayes' rule. I will briefly outline the basic elements of our model and then describe four areas of current research: (i) directly using the model in genetic counseling; assessing and communicating uncertainty about the model's estimated probability; (ii) validating the model using pedigree data on tested individuals, accounting for errors in genetic testing; (iii) using the model for exploring differences in prognosis between carriers and noncarriers of the genes, by incorporating detailed pedigree information in a survival analysis; (iv) developing a comprehensive model of breast cancer risk, including pedigree information as well as other risk factors. The latter raises interesting issues about jointly modeling case-control and prospective studies.