Bayesian Estimation of a Joint Semiparametric Recurrent Event Model of Multiple Cancer Types with Applications to the Li-Fraumeni Syndrome

• Activity Number: 338 - Novel Bayesian Methods in Genetic and Genomic Studies
• Type: Contributed
• Date/Time: Tuesday, August 9, 2022 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Genomics and Genetics
• Author(s): Nam Hoai Nguyen* and Seung Jun Shin and Jing Ning and Elissa Dodd-Eaton and Wenyi Wang
• Companies: Rice University and Korea University and UT MD Anderson Cancer Center and The University of Texas MD Anderson Cancer Center and The University of Texas MD Anderson Cancer Center
• Keywords: survival analysis; age-at-onset penetrance; individual likelihood; family-wise likelihood; Markov chain Monte Carlo

Abstract

Multiple primary cancers (MPC) are increasingly more frequent, hence personalized risk evaluation on cancer survivors is critical to effective clinical management. Studies have shown that the risk landscapes of cancer survivors are different from those who have never developed cancer, and that the risk of subsequent primary depends on the first primary. We propose a Bayesian semiparametric framework, where the intensities of non-homogeneous Poisson processes are cancer-type-specific and depend on the type and timing of the first primary. To overcome the scarcity of MPC, we apply our model to a dataset on the Li-Fraumeni syndrome (LFS), a rare hereditary disorder with germline mutations in the TP53 tumor suppressor gene. We use the peeling algorithm to account for the pedigree structure, and the ascertainment-corrected joint approach to correct for ascertainment bias. The training cohort consists of 431 families collected at MD Anderson Cancer Center. The validation on an independent cohort of 189 families showed a good performance of our model in risk prediction, which could not be remotely reached by a naïve model that assumes independence between the first and second primary.