A Bayesian analysis of birth pulse effects on the probability of detecting Ebola virus in fruit bats

Abstract

Since 1976 various species of Ebolavirus have caused a series of zoonotic outbreaks and public health crises in Africa. Bats have long been hypothesised to function as important hosts for ebolavirus maintenance, however the transmission ecology for these viruses remains poorly understood. Several studies have demonstrated rapid seroconversion for ebolavirus antibodies in young bats, yet paradoxically few PCR studies have confirmed the identity of the circulating viral species causing these seroconversions. The current study presents an age-structured epidemiological model that characterises the effects of seasonal birth pulses on ebolavirus transmission within a colony of African straw-coloured fruit bats (Eidolon helvum). Bayesian calibration is performed using previously published serological data collected from Cameroon, and age-structure data from Ghana. The model predicts that annual birth pulses most likely give rise to annual outbreaks, although more complex dynamic patterns – including multi-annual cycles and skip years – may be possible. Weeks 30 to 31 of each year were estimated to be the most likely period for isolating the circulating virus in Cameroon. The probability that a previous PCR campaign failed to detect Ebola virus, assuming that it was circulating, was estimated to be one in two thousand. This raises questions such as (1) what can we actually learn from ebolavirus serology tests performed without positive controls? (2) are current PCR tests sufficiently sensitive? (3) are swab samples really appropriate for ebolavirus extraction? The current results provide important insights for the design of future field studies aiming to extract Ebola viruses from sylvatic hosts, and can contribute to risk assessments concerning the timing of zoonotic outbreaks.

Using serosurveys to optimize surveillance for zoonotic pathogens

“…We used publicly available data from 28 to predict the peak period of active infection, , using our methodology. This data set includes seroprevalence and the proportion of animals lactating for each species.…”

“…They also tested for active infections in E. helvum using a semi-nested PCR assay specific to Ebola virus (EBOV; Zaire ebolavirus ) targeting a 184 bp fragment on the VP35 gene see 38 . For analyses in this study, we used the results from the Res1GP.ZEBVkiss antigenic test, a test for on the glycoprotein of EBOV, following 28 . Specific details of all methods and data are publicly available from 38 and 28 .…”

“…For analyses in this study, we used the results from the Res1GP.ZEBVkiss antigenic test, a test for on the glycoprotein of EBOV, following 28 . Specific details of all methods and data are publicly available from 38 and 28 .…”

“…To parameterize our models for E. helvum and H. monstrosus , we used values previously estimated values for the recovery rate and rate of waning antibodies. Pleydell et al 28 estimated the recovery rate ( γ = 1 / 1.5 weeks) and rate of waning antibodies ( ω A = 1 / 75 weeks) for the E. helvum population in Cameroon, but did not estimate these value for H. monstrosus . For H. monstrosus , we used measurements from experimental studies in Egyptian fruit bats ( Rousettus aegyptiacus ) with Marburg virus (MARV; Marburg marburgvirus ) to approximate parameter values for the recovery rate ( γ = 1 / 1.43 weeks) 43 and the rate of waning antibodies ( ω A = 1 / 12.9 weeks) 44 .…”

“…For example, Borremans et al 27 used information about multiple antibodies over time, pathogen presence, and demographic information to back-calculate the time since infection for individuals to estimate incidence of Morogoro virus infection in multimammate mice ( M. natalensis ). Using a different approach, Pleydell et al 28 fit an age-structured epidemiological model specific to Ebola virus in straw-colored fruit bats ( Eidolon helvum ) to estimate the timing of peak prevalence in the adult population. Although these methods are robust, adapting them quickly to other systems would be laborious and not always feasible depending on data availability.…”