Abstract PR06: Using frailty models to improve familial cancer risk prediction

Abstract

There are numerous statistical models used to identify individuals at high risk of cancer due to inherited mutations. We focus on models using Mendelian laws of inheritance to calculate the probability that a counselee is a mutation carrier and their future risk of cancer based on family history and known mutation prevalence and penetrance (the probability of having a disease at a certain age given the person9s genotype). Mendelian risk prediction models for various cancers have previously been developed. These models include BRCAPRO, which identifies individuals at high risk for breast or ovarian cancer by calculating the probabilities of germline deleterious mutations in BRCA1 and BRCA2. These models do not account for the heterogeneity of risk across families due to sources such as environmental or unobserved genetic risk factors. We aim to improve breast cancer risk prediction in the BRCAPRO model by incorporating a frailty model that contains a family-specific variate to account for this heterogeneity. We apply our proposed model to data from the Cancer Genetics Network, and preliminary results show that model calibration (measured by the ratio of observed to expected number of events) improves, while discrimination (measured by the area under the receiver operating characteristic (ROC) curve) remains the same. This abstract is also being presented as Poster A18. Citation Format: Theodore Huang, Danielle Braun, Malka Gorfine, Giovanni Parmigiani. Using frailty models to improve familial cancer risk prediction. [abstract]. In: Proceedings of the AACR Special Conference: Improving Cancer Risk Prediction for Prevention and Early Detection; Nov 16-19, 2016; Orlando, FL. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2017;26(5 Suppl):Abstract nr PR06.