Speaker
Description
African animal trypanosomosis (AAT) imposes a major constraint on livestock productivity throughout sub-Saharan Africa, but its true prevalence is underestimated due to the limited sensitivity of widely used diagnostic methods, namely the haematocrit centrifugation and buffy coat techniques (HCT/BCT). Although polymerase chain reaction (PCR) testing provides improved sensitivity, it remains expensive and is seldom accessible in routine field conditions. We developed a Bayesian hierarchical latent class framework using bovine trypanosomosis data compiled from more than 40 studies across Africa to infer the sensitivity and specificity of HCT/BCT and PCR. These estimates were then used to correct apparent prevalence, followed by geostatistical modelling in two high-burden, data-rich settings: Ethiopia and Nigeria. Two key forms of uncertainty were explicitly characterised: diagnostic uncertainty arising from imperfect tests, and spatial coverage uncertainty due to uneven sampling. Estimated sensitivities were 29.8% for HCT/BCT and 68.9% for PCR, while specificities were above 99% for both. After accounting for diagnostic performance, mean prevalence increased from 0.086 to 0.299 in Nigeria and from 0.092 to 0.382 in Ethiopia, corresponding to approximately 1.93 million and 3.39 million infected cattle within the tsetse belt. Across endemic regions, diagnostic uncertainty was the predominant contributor. Incorporating test uncertainty into spatial modelling provides a more accurate representation of AAT burden, supporting improved surveillance and more effective targeting of control strategies, with broader relevance to other neglected tropical diseases.
This work was developed in the framework of the FAO project ‘Disease intelligence and modelling for progressive control of animal trypanosomosis in Africa’ (DIMCAT), supported by the Gates Foundation (INV-067581).
