Multistate Modeling Approaches for Estimating Disease Risk in Gene Mutation Carriers from Family Data

• Activity Number: 642
• Type: Invited
• Date/Time: Thursday, August 5, 2010 : 10:30 AM to 12:20 PM
• Sponsor: SSC
• Author(s): Karen Kopciuk*+ and Yun-Hee Choi and Laurent Briollais and Patrick Parfrey and Jane Green
• Companies: Alberta Health Services and The University of Western Ontario and Samuel Lunenfeld Research Institute and Memorial University and Memorial University
• Keywords: multi-state models ; disease risk (penetrance) ; Expectation-Maximization (EM) algorithm ; missing genotype data ; multiple phenotypes ; disease screening

Abstract

Hereditary Nonpolyposis Colorectal Cancer (HNPCC) results in multiple cancers within the same family so understanding the patterns of various cancers among individuals can help dissect the genetic origin of the disease and predict future events in susceptible individuals. We employ progressive and illness-death multi-state models to estimate age-specific penetrance estimates of developing successive phenotypes and to assess the impact of interventions in pedigree data. An EM algorithm is developed to estimate all regression model parameters by inferring missing genotypes; our methods were applied to a retrospective cohort of 12 HNPCC families from Newfoundland who harbour a founder mutation. We found the modulating time scale term unnecessary, lifetime penetrance for a second cancer high (70% for males, 50% for females) and hazard ratios of 2nd to 1st cancer in carriers fell with age.