|
Abstract:
|
Repeated low-dose (RLD) challenge studies provide valuable information when evaluating candidate vaccines for founder viruses since they resemble the typical exposure of natural transmission and inform on the number of exposures prior to infection. This work provides methods to characterize candidate vaccine’s protective effect, by determining the vaccine's action model (VAM). The VAMs we consider are a null model (no protection), a Leaky model in which the probability of infection is reduced by some factor in vaccinated subjects, the All-or-none model in which the vaccine either offers complete protection or no protection in vaccinated subjects, and a Combination model with both Leaky and All-or-none mechanisms. We consider two competing models involving maximum likelihood methods for a discrete survival model. These models assume that the founder virus population follows a Poisson distribution with either a fixed mean parameter (Poisson model), or a random mean parameter that follows a Gamma distribution (Negative Binomial Model). We illustrate the performance of these methodologies with a data example of SIV on non-human primates and a simulation study.
|
