Incorporating immunological dynamics in transmission models of varicella zoster virus (VZV) and cytomegalovirus (CMV)

Varicella zoster virus (VZV) and cytomegalovirus (CMV) are herpes viruses that lead to persistent (latent) infection of the human host. Upon primary infection, VZV causes a disease called varicella or chickenpox. Later in life, the virus may reactive causing a disease called zoster or shingles. Primary infection with CMV usually causes mild disease (and sometimes CMV mononucleosis), and may later reactivate in immunocompromised individuals. The mechanisms determining whether and at what age reactivation takes place are still poorly understood. Here we analyse data of VZV and CMV antibody concentrations in the Dutch population, in conjunction with zoster incidence data. Our aim is to obtain an improved understanding of the transmission dynamics of these viruses in the population and, hopefully, also of the immunological dynamics in the human host. Inspired by Hope-Simpson’s exogenous boosting hypothesis, we analyse the combined VZV seroprevalence and zoster incidence data in an age-structured transmission model in which the reactivation rate is dependent on age, the number of boosting events, and the time since the last boosting event.  The model is based on earlier work by Guzzetta and colleagues (Guzzetta G et al (2013) Am J Epidemiol 177, 1134-1142). The analyses show that the boosting hypothesis is fully compatible with the data. However, we will argue that the data can be explained equally well with a variety of within-host models. For CMV, the  seroprofile data are complex, and no well-founded hypotheses have been put forward to explain the observed patterns, although there are suggestions that endogenous boosting of the immune system by reactivation attempts of the virus plays a role. We analyse the serological data in a statistical mixture model in which individuals are classified as seronegative or seropositive, and these classes are characterised by distributions of the antibody concentrations. The age-dependent seroprevalence (the mixing parameter) is estimated using a penalised spline. We will argue that the observed patterns of seropositivity which increase slowly with age (ranging from 50% in older adults) are difficult to capture using standard epidemiological transmission models with reactivation.