From tests to truth: A misclassification-aware machine learning framework for estimating brucellosis seroprevalence in wild canids

Brucellosis is an important zoonotic disease affecting humans, livestock, and wildlife, yet prevalence estimates in wild species are often underestimated due to limited attention to surveillance, as well as insufficient and biased sampling. To clarify exposure patterns in wild canids, we searched PubMed, Scopus, Web of Science, SciELO, and Google Scholar (1962-2025) for primary prevalence studies of Brucella species (spp.) in free-ranging canids. Serological data were analyzed using a misclassification-aware, multi-assay model that corrects for imperfect test sensitivity and specificity. Confirmatory polymerase chain reaction (PCR) and culture results were considered separately from serological data, as each offers a different perspective on disease status. Across 48 wild serology populations (n = 3,925 animals), the global misclassification-adjusted true seroprevalence was 8.2% (95% confidence interval CI: 5.1-11.3%). Confirmed active infection, based on PCR or culture, was uncommon (3.9%; 95% CI: 3.0-5.1%). Exposure levels varied across continents, with higher estimates in South America (approximately 18%) and lower levels in Europe (approximately 0.8%) and North America (approximately 4.1%). Data from Africa were limited, and Asian estimates were based on sparse wild samples, leading to wide uncertainty. Seroprevalence was consistently influenced by assay type, host species, and region. Overall, wild canids exhibit modest but widespread serological exposure to Brucella spp., whereas confirmed active infection remains rare. Because evidence quality and diagnostic rigor varied considerably across regions, especially in Africa and parts of Asia, results from these areas should be interpreted with caution. By correcting for imperfect tests and explicitly accounting for study heterogeneity, this framework provides more reliable and transparent prevalence estimates and highlights geographic gaps where improved, targeted One Health surveillance is most urgently needed.