Joint Model of Multiple Longitudinal Biomarkers and an Interval-Censored Survival Outcome with Application to a Cervical Cancer Etiology

• Activity Number: 493 - Section on Risk Analysis CPapers 1
• Type: Contributed
• Date/Time: Thursday, August 6, 2020 : 10:00 AM to 2:00 PM
• Sponsor: Section on Risk Analysis
• Author(s): Siddharth Roy* and Megan Clarke and Anindya Roy and Ana Gradissimo and Robert Burk and Nicolas Wentzensen and Paul S. Albert and Danping Liu
• Companies: NIH/UMBC and National Cancer Institute and University of Marland Baltimore County and Albert Einstein College of Medicine and Albert Einstein College of Medicine and National Cancer Institute and National Cancer Institute and Eunice Kennedy Shriver National Institute of Child Health and Human Development
• Keywords: Joint Model; HPV; Cervical Cancer Risk; interval censored; Sampling weights; multiple biomarkers

Abstract

Human Papillomavirus (HPV) DNA methylation is a promising biomarker to predict the risk of precancer lesions for HPV positive women. Other disease predictors such as abnormal cytology can provide information on the timing on the transition from HPV infection to precancers and need to be accounted for. We developed a joint model to simultaneously model both a continuous longitudinal HPV methylation biomarker and a binary longitudinal covariate representing cytology and assess the relationship of these covariates to time to precancerous lesions. This model uses a common latent process to drive both longitudinal variables as well as the time to precancer. The time to precancer or last follow up is subject to interval censoring due to annual or bi-annual clinical visits. The survival time was discretized to allow for closed form integrations in the likelihood. The model also incorporates sampling weights to make cohort inference from stratified sampling data. The method is evaluated via simulations and is applied to current longitudinal HPV dataset to highlight practical use.