| Recent advances in genetic research have led to the discovery of inherited genetic mutations that increase disease risk. Immediate clinical use of these advances requires counseling people concerned about disease running in their family, to help them decide about taking a genetic test and to inform them of ways to reduce their disease risk. To help make these decisions, clinicians need to estimate the probability that a person carries a mutation given family history of disease; if this probability is high enough, the clinician may offer genetic testing to the person. Mendelian mutation prediction models are statistical models that estimate this carrier probability based on family history and knowledge of the population prevalence of the mutation and of the penetrance of the mutation (probability of disease, given mutation status). The example model that this thesis will be applied to is BRCAPRO (see (Berry et al., 1997) and (Parmigiani et al., 1998)), which estimates the probability of carrying a deleterious mutation in the BRCA1 and BRCA2 genes, based on family history of breast and ovarian cancer. First, this thesis explores the effects of misreported family history on the carrier probability estimate by deriving a general mathematical framework and presenting results for BRCAPRO in particular. Second, competing risks and censoring issues are formally treated and the effects of ignoring a competing risk are derived and applied to BRCAPRO. Finally, a methodology is presented for incorporating medical interventions into Mendelian mutation prediction models and applied to incorporating oophorectomy into BRCAPRO. |
