Eradication of bTB at a herd level, which is prerequisite for disease eradication at a region and country level, has been impaired by the chronic nature of the disease that can lead to relatively long periods of “silent infection” in infected animals, especially at low infectious doses [8
]. This feature of the disease, together with limitations in accuracy of the diagnostic techniques, can substantially extend the time needed for declaring a herd bTB free. Several studies have been published dealing with the identification of risk factors at an individual or herd level associated with bTB [6
] or with the detection of predictors of future herd breakdowns in bTB free herds [16
], but longitudinal studies aiming at elucidating the disease dynamics in infected herds under a strong eradication pressure such as the one presented here have not been published in the peer reviewed literature. Here, we analyzed data from all cattle herds in ACM included in the Spanish bTB eradication program, therefore providing information on its situation without sampling-biases. ACM was selected for this study due to its high herd-prevalence, its relative small size (which allows for including the entire herd population in the analysis) and the presence of most epidemiological situations described in the rest of the country; these include different herd types, presence of wildlife reservoirs, and extensive and intensively managed herds. Thus, results here are likely to be representative of other Spanish regions. Our data suggest that herd type influences the probability of success of the bTB control program in ACM, and perhaps, in other bTB-infected regions of Spain and Europe.
The median value of βt
, estimated here was similar to the values reported in Argentina [10
], New Zealand [9
] and the Netherlands [17
], suggesting that, on average, every infectious animal led to 2–5 infections per year. However, the value of βt
was significantly higher for dairies compared with other herd types. Dairy farms were at higher risk for bTB than beef and bullfighting herds, as indicated by the larger values of βt
, proportion of outbreaks, and proportion of recurrently infected farms observed in dairy herds compared to other herd types. This finding may be related to a combination of factors that favor disease spread in these herds including, for example, high contact rates, high density, and presence of stressors associated with the intensive management of animals, as previously suggested [6
]. Arguably, some animals may have been culled before becoming positive, which would result in an underestimation of beta (especially in beef herds with a higher replacement rate); however, animals culled prior to detection were, consequently, not considered as sources of infection at the subsequent estimation of beta, which would result in an overestimation of the parameter. In addition, animals from bTB-infected cannot be moved to other herds, therefore decreasing the likeness of removal of infected animals from the herd before detection. Imperfect sensitivity was also incorporated to estimate the number of true infected animals. Therefore, estimates here would be, on average, accurate. The number of herd tests required to obtain OTF status was not influenced by the herd type; however, the time to achieve OTF status was shorter for dairy herds compared to beef and bullfighting herds, probably as a consequence of a short duration of the between-test period in dairy farms. In addition, the occurrence of occasional individual infections due to contact with infected wildlife could also contribute to extend duration of outbreaks in extensively managed herds (typically beef and bullfighting herds). These results suggest that a possible herd-type effect should be taken into account when measures for bTB eradication are being implemented.
Introduction of the IFN-γ assay for the identification of infected animals in positive herds resulted in an increase on the values of βt
(Table ), likely as a consequence of an increase in the sensitivity of the detection, as expected when this technique is used as an ancillary test [18
]. Consequently, use of the IFN-γ assay also decreased significantly the median time to recover OTF status after an outbreak (Table ). Misclassification of still-infected farms as OTF due to the change in the regulations implemented in 2006 does not seem likely, as only 3 (6%) of the 50 farms that achieved OTF status after 2006 were reinfected by 2010. These findings suggest that incorporation of the IFN-γ was beneficial for the control program in ACM, accelerating bTB eradication at a herd-level.
Interestingly, results of the multivariable model suggest that the larger the number of bTB-positive animals detected at the first herd test, the shorter the time to eradicate the disease from the farm. A large number of positive animals at the first herd test may be interpreted as an indication of low probability of false negative results, thus, decreasing the time of eradication. This finding is consistent with the results obtained by the use of IFN-γ, which most likely increases the sensitivity of the detection.
Outliers were removed from the dataset before the survival analysis (10% outbreaks) and before identifying the theoretical distribution that best fitted the within-herd transmission data (6% herds) because of the uncertainty related to the accuracy of such estimations. There are a number of factors that could lead to unusual lengths of outbreaks and high values of βf (referred to as outliers) including, for example, recording bias, introduction of infected animals, or participation of wildlife reservoirs in the maintenance and transmission of the disease (whose effect on bTB transmission would be included in the same βf estimate). Still, from the decision making perspective, identification and further investigation of such outliers is of interest, because they may represent farms with particular epidemiological conditions that truly increase the risk for the disease, or, alternatively, misreporting of data related with disease identification. Nevertheless, if outliers were not excluded from the analyses the results obtained did not change significantly as shown by the overlapping 95% CI in all estimates (data not shown).
Arguably, the formulation used here to compute the transmission rate is relatively simple, because certain factors that may affect transmission, such as absence of random contact among susceptible and infected animals, management practices, or the potential role that certain wildlife species may play in disease transmission, were not considered in the formulation. Increasing model complexity would require quantitative knowledge on the value that certain epidemiological factors took in the study population that was not available to the authors including, for example, number and distribution of wildlife species or rate of contact between cattle per herd. Most important, it is uncertain whether inclusion of such information in the model formulation would result in significantly different estimates for the transmission rate compared to those presented here. In the absence of such information, estimates presented here may be useful approximations of the true value values of the transmission rate of bovine tuberculosis in Spain that allow to compare the estimates between different herd types and other epidemiological factors. This could aid in the evaluation of the impact of control strategies for different production systems. In any case, the methodological approach presented here may help the administration of control programs to identify and further investigate those farms in which unusually high transmission rates suggest the occurrence of an anomaly in the pattern of disease transmission.