Associate Professor, Canada Research Chair in Marine Epidemiology University of Toronto, Canada
Abstract: While compartmental models have long been used to describe among-host disease dynamics, within-host dynamics have been comparatively understudied. It has been theorized that Lotka-Volterra models may be well-suited to describing within-host dynamics, given that a host’s immune system can seek out and destroy infecting parasites as a predator does its prey. However, the practical difficulties associated with collecting within-host infection data have precluded empirical model validation. Aiming to determine the kind and quality of data required to reliably recover disease parameters and identify dynamical behaviors from within-host infection data, we used two models of within-host population dynamics (Antia et al. (1994); Nowak and May (2000)) to simulate five sets of stochastic time series data. We used the Antia et al. model to simulate an acute infection, wherein the parasite elicits a Type II functional response from the host’s immune system. We used the Nowak and May model to simulate four dynamically distinct chronic infections. Within three of these parameterizations, the parasite elicits a Type II functional response from the host’s immune system; within the fourth parameterization, the parasite elicits a Type III functional response from the host’s immune system. We then used Bayesian methods to fit the models to the simulated data in autoregressive and global frameworks, and re-estimate parameter values and population trajectories.
We’ve found that the statistical model is able to reliably re-estimate the parameter values and population trajectories of the Antia et al. model, but must be provided with the true values of parasite recognition rate and immune system handling time to reliably re-estimate the parameter values and population trajectories of the various parameterizations of the Nowak and May model. The ability of the statistical model to reproduce the population trajectories of different parameterizations of the Nowak and May model is of particular importance, given that very small changes in parameter values produce radically different dynamical behaviors and infection outcomes. Taken together, these findings imply that to empirically validate within-host population dynamics models and improve our understanding of within-host disease dynamics, we’ll not only need to collect within-host infection data—we’ll also need to quantify key disease parameters. Improving our understanding of within-host disease dynamics will bolster our understanding of the factors that underlie and drive among-host dynamics, and allow us to make robust and timely predictions regarding disease spread, and implement actionable mitigation strategies in support of public health.
