An agent-based model of dengue virus transmission shows how multiple uncertainties about vaccine efficacy influence public health impact projections

Abstract

Given the limited effectiveness of strategies based solely on vector control to reduce dengue virus (DENV) transmission, it is expected that an effective vaccine could play a pivotal role in reducing the global disease burden of dengue. Of several dengue vaccines under development, Dengvaxia® from Sanofi Pasteur recently became the first to become licensed in select countries and to achieve WHO recommendation for use in certain settings, despite the fact that a number of uncertainties about its profile complicate projections of its public health impact. We used a stochastic, agent-based model for DENV transmission to perform simulations of the public health impact of dengue vaccines in light of two key uncertainties: (1) "statistical uncertainty" about the numerical value of the vaccine's efficacy against disease, and (2) "biological uncertainty" about the extent to which its efficacy against disease derives from the amelioration of symptoms, blocking of DENV infection, or some combination thereof. Simulations of a generic dengue vaccine showed that the proportion of disease episodes averted following 20 years of routine vaccination of nine-year olds at 80% coverage was sensitive to both the numerical value of vaccine efficacy and to the extent to which efficacy derives from blocking of DENV infection. Simulations of a vaccine resembling Dengvaxia® took into account that vaccine trial results substantially reduced statistical uncertainty but did not address biological uncertainty, resulting in the proportion of disease episodes averted being more sensitive to biological uncertainty than to statistical uncertainty. Taken together, our results indicate limitations associated with the use of symptomatic disease as the primary endpoint of dengue vaccine trials and highlight the importance of considering multiple forms of uncertainty in projections of a vaccine's public health impact.

Biased efficacy estimates in phase-III dengue vaccine trials due to heterogeneous exposure and differential detectability of primary infections across trial arms

“…We simulated vaccine trials using an agent-based model of DENV transmission that simulates the transmission dynamics of DENV in a population of 200,000 in a core population area of Iquitos, Peru [42,43]. Our model includes simulated daily human movement patterns calibrated to data from retrospective, semi-structured interviews of residents in Iquitos [44,45].…”

“…Our model includes simulated daily human movement patterns calibrated to data from retrospective, semi-structured interviews of residents in Iquitos [44,45]. This model has been demonstrated to reproduce the dynamics of all four DENV serotypes [42] and was used to investigate the epidemiological impacts of hypothetical vaccination campaigns to project the population-level impact of Dengvaxia [46,47]. Because a formal description of the version of the transmission model that was used here has been published elsewhere [42], we provide an overview of its key features and assumptions in S2 Text.…”

“…Three sets of parameters were calibrated through this process: (1) serotype-specific population immunity on 1 January 2000, when estimates of time-varying, serotype-specific incidence of infection began; (2) time-varying, serotype-specific patterns of unexplained infections from visitors to the study area or residents that were infected outside the study area; and (3) constants for scaling mosquito population density at two different time points during which mosquitoes were collected using different methods. We calibrated these parameters using a particle filtering algorithm described in detail elsewhere (Table 3) [42].…”

Modeling Diseases: Prevention, Cure and Management

“…Dengue is a mosquito-borne disease and occurs in more than 100 countries in tropical and subtropical regions. Recent medical advances in vaccine development has prompted some of the recent modeling work to focus on vaccination strategies (e.g., Knerer et al (2015), Perkins et al (2018)). ABM is a natural tool for modeling these vector-borne diseases and can enable geographical factors to be incorporated into the model (de Lima et al 2016;Isidoro et al 2011).…”

Model-based assessment of public health impact and cost-effectiveness of dengue vaccination following screening for prior exposure

“…Generalized additive model of vaccination impact projections.Despite efforts to minimize stochastic differences between paired simulations under a given parameter set, the proportions of cases averted resulting from our simulations were relatively noisy. This is due to the dynamics of the model, which are characterized by large epidemics in some years separated by low levels of incidence during inter-epidemic periods[23]. Thus, even small differences in the sequence of random number draws (due to differences in infection outcomes associated with protective effects of vaccination) can lead otherwise similar pairs of simulations to diverge in their behavior over time.…”