Title: A hybrid method for the stratified mark-specific proportional hazards models with missing data with applications
Authors: Yanqing Sun - University of North Carolina at Charlotte (United States) [presenting]
Li Qi - Biostatistics and Programming at Sanofi (United States)
Peter Gilbert - University of Washington and Fred Hutchinson Cancer Research Center (United States)
Abstract: The motivation comes from the objective to understand how the dengue vaccine efficacy is modified by neutralizing antibodies and whether it depends on dengue genetics. The immune responses in the CYD14 dengue efficacy trial were measured through a two-phase sampling design and there is a high percentage of missing dengue sequences. We develop estimation and hypothesis testing procedures for the stratified mark-specific proportional hazards model with missing covariates and missing marks, where the mark is the genetic distance of an infecting dengue sequence to the dengue sequence represented inside the vaccine. We propose a hybrid method that takes advantages of both the augmented inverse probability weighted method and multiple imputation. A simulation study is conducted to examine the finite-sample performances of the estimators for the mark-specific relative risks and the conditional mark-specific cumulative incidence functions as well as the proposed testing procedures. Our simulation study shows that the proposed hybrid method performs well. The hybrid estimator is better than a direct application of Rubin's multiple imputation with the estimated variances close to the empirical variances, and thus better coverage probabilities. The developed hybrid methods are applied to the CYD14 efficacy trial to assess association of dengue infection with the immune responses.