Bovine Tuberculosis Antemortem Diagnostic Test Agreement and Disagreement in a Naturally Infected African Cattle Population

The interferon-gamma (IFN-γ) assay and single comparative cervical skin test (SCITT) are used to estimate bovine tuberculosis (bTB) prevalence globally. Prevalence estimates of bTB, caused by Mycobacterium bovis, are poorly quantified in many Sub-Saharan African (SSA) cattle populations. Furthermore, antemortem diagnostic performance can vary at different stages of bTB pathogenesis and in different cattle populations. In this study, we aim to explore the level of agreement and disagreement between the IFN-γ assay and SCITT test, along with the drivers for disagreement, in a naturally infected African cattle population. In, 2013, a pastoral cattle population was sampled using a stratified clustered cross-sectional study in Cameroon. A total of 100 pastoral cattle herds in the North West Region (NWR) and the Vina Division (VIN) were sampled totalling 1,448 cattle. Individual animal data and herd-level data were collected, and animals were screened using both the IFN-γ assay and SCITT. Serological ELISAs were used to detect exposure to immunosuppressing co-infections. Agreement analyses were used to compare the performance between the two bTB diagnostic tests, and multivariable mixed-effects logistic regression models (MLR) were developed to investigate the two forms of IFN-γ assay and SCITT binary disagreement. Best agreement using the Cohen's κ statistic, between the SCITT (>2 mm) and the IFN-γ assay implied a ‘fair-moderate' agreement for the NWR [κ = 0.42 (95%CI: 0.31–0.53)] and ‘poor-moderate' for the VIN [κ = 0.33 (95% CI: 0.18–0.47)]. The main test disagreement was the animals testing positive on the IFN-γ assay and negative by the SCITT. From MLR modeling, adults (adults OR: 7.57; older adults OR = 7.21), females (OR = 0.50), bovine leucosis (OR = 2.30), and paratuberculosis positivity (OR = 6.54) were associated with IFN-γ-positive/SCITT-negative disagreement. Subsets to investigate diagnostic test disagreement for being SCITT-positive and IFN-γ-negative also identified that adults (adults OR = 15.74; older adults OR = 9.18) were associated with IFN-γ-negative/SCITT-positive disagreement. We demonstrate that individual or combined use of the IFN-γ assay and SCITT can lead to a large variation in bTB prevalence estimates. Considering that animal level factors were associated with disagreement between the IFN-γ assay and SCITT in this study, future work should further investigate their impact on diagnostic test performance to develop the approaches to improve SSA prevalence estimates.