Two Serological Tests Show High Accuracy for Detecting Crimean-Congo Haemorrhagic Fever Virus Exposure in Tanzanian Febrile Patients

Crimean-Congo haemorrhagic fever virus (CCHFV) is a tick-borne zoonotic pathogen that can cause severe haemorrhagic disease in humans, with fatality ratios during outbreaks ranging from 5 to 40 percent. The virus circulates across Europe, Asia, and Africa, and Hyalomma species ticks are considered a reservoir. Transmission occurs through infected ticks as well as through contact with blood or tissues of infected persons or animals. Despite its virulence and the absence of effective antiviral treatments or a licensed vaccine, CCHFV is recognised as a high-priority pathogen in the World Health Organization R&D Blueprint. Clinical diagnosis is particularly challenging in low-resource settings because early symptoms such as fever, headache, myalgia, arthralgia, nausea, and vomiting overlap substantially with many other febrile illnesses, including malaria, dengue, typhoid fever, leptospirosis, and Rift Valley fever, increasing the risk of misdiagnosis and delayed detection.

A major obstacle to accurate CCHF diagnosis is the absence of a universally recognised gold standard serological test. The reference standard for antibody detection requires neutralisation assays, which must be performed in biosafety level 4 (BSL-4) laboratories. These facilities are scarce and largely unavailable in sub-Saharan Africa, including Tanzania, where CCHFV is endemic. Although several commercial molecular and serological diagnostic tests are available, including nucleic acid amplification tests such as reverse transcriptase polymerase chain reaction and loop-mediated isothermal amplification, these approaches face challenges due to substantial genetic variation among viral lineages. Antibody-based serological assays are generally less impacted by genetic variation but have limitations such as delayed antibody response, cross-reactivity, and inability to distinguish recent from past infection. Neutralisation assays are also of limited diagnostic utility for CCHF because the virus elicits low levels of neutralising antibody.

Against this backdrop, a research team led by Elichilia R. Shao of Kilimanjaro Christian Medical Center and the Kilimanjaro Clinical Research Institute in Moshi, Tanzania, together with collaborators from institutions including the University of Glasgow, Duke University, Washington State University, and Duke-National University of Singapore Medical School, conducted a study to evaluate the diagnostic performance of two serological assays for detecting prior CCHFV exposure in humans. The work, published in Frontiers in Microbiology, employed a Bayesian latent class modelling approach in the absence of a reference standard, allowing simultaneous estimation of test sensitivity, specificity, and true seroprevalence without relying on a gold standard.

The study was a secondary analysis of archived serum samples from a prospective febrile illness surveillance cohort conducted between 2012 and 2014 in Moshi, Tanzania, at Kilimanjaro Christian Medical Centre (KCMC) and Mawenzi Regional Referral Hospital (MRRH). At the time, KCMC was a 450-bed zonal referral and teaching hospital serving more than 15 million people in northern Tanzania, while MRRH, five kilometres away, was a 300-bed referral hospital for the Kilimanjaro Region. Of 870 febrile admissions enrolled across both sites, serum samples were available for 427 participants, representing 49 percent of the cohort. Participants were classified as livestock owners if their household owned one or more livestock species, including cattle, sheep, goats, pigs, poultry, camels, or horses, and as non-livestock owners otherwise. Of the 427 participants with available serum, 147 were livestock owners and 280 were non-livestock owners.

Two assays were compared. The first was the commercial ID Screen CCHF double-antigen multi-species enzyme-linked immunosorbent assay (ELISA) kit from IDvet Innovative Diagnostics, France, which detects IgG antibodies to the CCHFV nucleocapsid protein (NP). In this assay, test sera were diluted and incubated at 25 degrees Celsius for 45 minutes, followed by conjugate and substrate steps at 25 degrees Celsius for 30 and 15 minutes respectively. Absorbance was measured at 450 nanometres using an automated ELISA reader, and samples were deemed positive if their sample positivity percentage exceeded 30 percent. The second was a laboratory-developed microsphere immunoassay using Luminex technology, a qualitative multiplex magnetic bead-based assay in which virus-specific microsphere sets enabled simultaneous detection of NP-targeted IgG in a single well, relying on fluorescence and flow cytometry principles.

The analytical framework was based on the Hui-Walter Bayesian latent class model, a statistical framework for assessing diagnostic test accuracy when true disease status is uncertain. Three key assumptions underpin this model. First, the sensitivity and specificity of each test are considered constant across the subpopulations studied. The researchers considered this assumption not violated because livestock keepers and non-livestock keepers were from the same geographic area with broadly similar demographic and environmental conditions. Second, diagnostic tests have the same performance across subpopulations despite differences in seroprevalence, which was supported by standardised protocols for sample collection and assay implementation, and by laboratory personnel being blinded to subpopulation status. Third, conditional independence is presumed, where tests are independent given true disease status. However, because both assays measure the same underlying biological process, host antibodies to CCHFV antigens, their results are likely correlated even after accounting for true infection status. The model was therefore extended to incorporate conditional dependence using a covariance parameterisation.

Prior distributions were informed by a literature review covering peer-reviewed articles published from January 2018 to December 2025, with emphasis on East African studies. Sensitivity and specificity priors assumed beta distributions. For the Luminex sensitivity, no literature was identified, so the final model specified the Luminex sensitivity prior the same as the IDvet prior, given the comparable serological detection principles of the two assays. Seroprevalence priors were guided by East African CCHFV seroprevalence estimates ranging from 1.8 to 19 percent, using an informative beta distribution with a prior mean of 9.1 percent. The model was implemented in Just Another Gibbs Sampler (JAGS) using the rjags package in R, with three chains, a burn-in phase of 10,000 iterations, and 6,000 sampling iterations thinned by 10, resulting in 1,800 posterior samples. Convergence was assessed with the Gelman-Rubin diagnostic, the Heidelberger-Welch stationarity test, trace plots, and effective sample size checks.

The results demonstrated high diagnostic performance for both assays. The median posterior estimate for IDvet sensitivity was 0.893 with a 95 percent Bayesian credibility interval (BCI) of 0.555 to 0.996, and specificity was 0.972 with a 95 percent BCI of 0.936 to 0.996. For the Luminex assay, the median posterior sensitivity estimate was 0.884 with a 95 percent BCI of 0.541 to 0.995, and specificity was 0.987 with a 95 percent BCI of 0.947 to 1.00. The model generated seroprevalence estimates for the two subpopulations. The livestock owner seroprevalence estimate was 0.027 with a 95 percent BCI of 0.002 to 0.069, while for non-livestock owners it was 0.017 with a 95 percent BCI of 0.001 to 0.044. Crude seroprevalence was 4.0 percent by IDvet and 2.3 percent by Luminex across all participants. Both assays showed very high negative predictive values exceeding 0.99 across all groups, while positive predictive values were lower, particularly for IDvet, with Luminex showing higher PPV among non-livestock owners consistent with its higher specificity.

The slightly higher seroprevalence among livestock owners might be explained by the age difference among livestock owners, who were slightly older, meaning more years of potential cumulative occupational and environmental exposure. The presence of seropositive individuals among non-livestock owners suggests that CCHFV exposure extends beyond groups with close livestock links typically considered at high risk. Possible explanations include environmental exposure to infected ticks, handling of contaminated animal products outside the home, or other unrecognised community pathways. Alternatively, reports of livestock ownership may not correlate with livestock contact-related activities such as milking or assisting deliveries. The covariance parameters for exposed and non-exposed individuals were centred near zero, with credibility intervals encompassing zero, indicating limited residual dependence between the two assays after adjusting for infection status.

The findings carry practical implications for CCHFV surveillance in endemic regions lacking reference-standard testing. Both platforms are suitable for CCHFV serosurveillance and epidemiological investigations, with the IDvet ELISA offering operational advantages because of its simplicity and suitability for decentralised laboratories, while the Luminex platform may offer additional value in specialised reference laboratories and research applications. At the country level, the IDvet can be used by the Tanzania Veterinary Laboratory Agency and human health facilities in tandem to provide a unified data stream for national CCHF preparedness and outbreak response. Several limitations should be acknowledged, including the relatively small number of seropositive samples leading to wide credibility intervals particularly for sensitivity and seroprevalence estimates, the measurement of IgG antibodies reflecting prior exposure rather than acute infection, prior distributions informed by a limited number of published diagnostic studies, and the use of archived sera collected in 2012 to 2014. Future studies with larger sample sizes, molecular confirmation, and evaluation across diverse endemic populations will be important for refining diagnostic performance estimates and supporting harmonisation of CCHFV serological testing strategies.

Subject of Research: Virology

Article Title: Comparison of two Crimean-Congo haemorrhagic fever virus serological tests in Tanzanian febrile patients using a Bayesian latent class approach

Article References: Shao, E. R., Bodenham, R. F., Chilongola, J., Cleaveland, S., Maro, V. P., Mujaga, B., Mlay, E. A., Shirima, L. J., Pyuza, J. J., Njau, J., Lankester, F., Kayaga, R., Rubach, M. P., Crump, J. A., Tan, C. W., Wang, L.-F., Rostal, M. K., & Mmbaga, B. T. (2026). Comparison of two Crimean-Congo haemorrhagic fever virus serological tests in Tanzanian febrile patients using a Bayesian latent class approach. Frontiers in Microbiology, 17, Article 1894712. https://doi.org/10.3389/fmicb.2026.1894712

Image Credits: AI Generated

DOI: 10.3389/fmicb.2026.1894712

Keywords: Crimean-Congo haemorrhagic fever virus, CCHFV, serology, Bayesian latent class model, diagnostic accuracy, ELISA, Luminex, Tanzania, febrile illness, zoonoses, Comparison, Crimean-Congo