Reconstructing the silent circulation of West Nile Virus in a Caribbean island during 15 years using sentinel serological data

The dynamics of zoonotic infectious diseases with silent circulation may be imperfectly understood and monitored using passive (or reactive) epidemiological surveillance data only, highlighting the interest of quantitative methods like modelling. West Nile virus (WNV) is a widespread mosquito-borne virus transmitted from birds to “dead-end” hosts including humans and horses, in whom it can be fatal. It was first detected in Guadeloupe archipelago, Caribbean, in 2002, although no WNV clinical case in humans nor horses had been reported before 2024. Undetected infections represent a risk as WNV can be transmitted via blood and organ donations. In Guadeloupe, epidemiological surveillance started in 2002 in sentinel chickens and horses and in 2015 in mosquitoes, to detect WNV and to improve knowledge on its epidemiology and dynamics. In order to reconstruct the WNV force of infection (FOI), we built a model assessing different hypotheses regarding its dynamics using serological results in respectively 1,022 and 3,649 blood samples collected from 256 horses and 317 chickens between 2002 and 2018. We fitted the model to the serological data using Markov Chains Monte Carlo. We found that WNV FOI in Guadeloupe Island presented both within-year (seasonal) and between-years fluctuations. We identified three main episodes of WNV circulation on the island between 2002 and 2017. During years with circulation, the FOI was predicted to be highest around the months of October-November, although transmission could occur all year long. We estimated a very low weekly seroreversion rate, which is consistent with a lifelong persistence of WNV IgG antibodies in many infected individuals. To conclude, combining longitudinal serological data to a mathematical model allowed reconstructing the recurrent and silent circulation of WNV in this Caribbean island, which could improve surveillance design for better virus detection.