To be or not to be (confirmed through bacteriology): spatial epidemiology of bovine tuberculosis in a low prevalence region in Spain

Comunicación oral en GeoVet 2019. Novel spatio-temporal approaches in the era of Big Data

8 de octubre de 2019

Pozo P., Minguez O., Grau A., Bezos J., Nacar J., Romero B., de Juan L. y Alvarez J.

Bovine tuberculosis (bTB) is an infectious chronic disease caused by members of the Mycobacterium tuberculosis complex, mainly M. bovis and M. caprae. It is heterogeneously distributed worldwide, and despite the efforts invested on its eradication, it is still prevalent in several parts of the world (OIE, World Organisation for Animal Health, 2019). In Europe, bTB is still endemic in areas of England, Wales, Ireland and Spain (EFSA, 2018). Bovine tuberculosis causes a high economic impact due to movement restrictions, diagnostic costs and insurances payments. Wildlife species (i.e., wild boar, badger, and red deer) may act as reservoirs of bTB for livestock, adding up additional complexity to the epidemiology of the disease (Nugent et al., 2015; Phillips et al., 2003; de la Rua-Domenech et al., 2006). Eradication programs are based on skin test and passive surveillance at the abattoir. All reactors are subjected to post-mortem analysis to confirm the presence of bTB through bacteriological culture.
An extensively assessed feature of bTB is its persistence in certain cattle herds, in terms of either herd recurrence or prolonged periods of restriction (Doyle et al., 2015; White et al., 2013). Recurrence and infection persistence in a herd may indicate failure of testing, thereby allowing infected animals not only to remain in the population, but potentially to act as an ongoing source of infection to other herds and wildlife (Broughan et. al., 2016). Accurate diagnosis of bTB in live animals is often difficult. Several factors have been identified that influence test performance and hence can lead to failures in detecting infection, such as limited sensitivity and specificity, the prevalence of the disease or disease stage in the animal, among others (de la Rua-Domenech et al., 2006).
In Spain, challenges to the final eradication of bTB include limitations in the accuracy of currently available diagnostic tests, the presence of wildlife reservoirs (mostly in wild boar and red deer), and undetected bTB in imported cattle (Ciaravino et al., 2018). These factors are associated to persistence of the disease at the farm level and in the environment, respectively. For these reasons, the key to controlling TB in Spain is an integrated approach at all levels in the ecosystem. Herd prevalence in Spain has remained relatively constant during the last 15 years (2.2% in 2002, 2.3% in 2017) (MAPA, 2018). However, the high diversity in the epidemiological situations in different areas of Spain adds even more complexity to the problem. The Castilla y Leon autonomous community, which holds 20% of the Spain’s total cattle population, is classified as a high prevalence region (>1%, 1.43% in 2018), but includes provinces with both high and low bTB herd prevalence (MAPA, 2018). Soria, a low bTB prevalence province (Figure 1) with a median number of 285 herds tested every year for bTB, has however a higher apparent herd prevalence compared with other neighboring provinces (10-15% positive herds detected every year) that are however not confirmed through bacteriology (with only 20-40% of skin positive herds being confirmed every year). These results may be due to a lack of specificity of in-vivo diagnostic tests (i.e., too many false positive animal/herds in the skin test) or a lack of sensitivity of the post-mortem tests (i.e., too many false negative animal/herds in bacteriology). For this reason, we performed a study to describe the spatial patterns in the distribution of positive (confirmed and non-confirmed) bTB cattle/herds in Soria, in an attempt to identify factors that could help establish if there is a lack of sensitivity or specificity in the diagnostic tests used in the bTB national eradication program in Soria.
For this reason, we collected information on individual (age, years in the herd) and herd (location, history of bTB, herd size, prevalence at a county level) risk factors from animals and herds with a positive reaction in the skin test (SIT)/IFN-Ƴ assay in Soria during 2015-2018. In addition, we analyzed data of reactors detected in Burgos, a neighboring province in Castilla y Leon, to assess whether patterns identified in Soria were similar in other regions. In each of the regions, positive animals to the in-vivo tests were classified as bTB-confirmed or true positive (SIT or IFN- ƴ+/Culture+) and bTB-unconfirmed or false positive (SIT or IFN-ƴ+/Culture-). Parametric and non-parametric tests were used to detect significant differences between individual/herd variables and bacteriology results.
We then explored the spatial variation in the risk of being confirmed through bacteriology for herds in Soria and Burgos: clustering of confirmed/unconfirmed animals in positive herds was assessed using the Poisson model of the spatial scan statistic, implemented using the SaTScan software (Kulldorff, 2009). We then assessed the association between being confirmed through culture (outcome) and the available predictor variables (age, years in the herd, confirmed bTB infection in the previous year to the bacteriology result, herd size and county level prevalence in the previous year) using univariable models with a liberal p-value (p <0.3) for Soria and Burgos. Subsequently, all variables potentially associated were evaluated using a mixed multivariate logistic regression model. To account for the lack of independence between the observations, all models (uni- and multivariable) included herd as a random effect. A backward stepwise selection to obtain the adequate model and comparison of the fit of different models based on likelihood-ratio tests were performed. Residuals of the final model were analyzed for spatial autocorrelation using Moran’s test to determine whether any spatial autocorrelation detected had been explained by the final model (Bivand et al., 2013).
Our source population consisted of 30,242 animals subjected to bTB testing and located in 137 and 70 bTB positive farms in Burgos (n=17,845, 59% animals) and Soria (n=12,397, 41% animals), respectively during 2015-2018. Recorded information included herd type (beef, dairy, mixed), herd size, herd location (latitude and longitude), date of testing, bTB diagnostic test results at every round of testing for each animal, and age. All herds in Soria were beef herds, while in Burgos there were 108 (78.8%) beef, 19 (13.9%) dairy, and 10 (7.3%) mixed herds. Post-mortem bacteriology (Culture, C) results were available for 2,124 animals, of which 1,538 (66 C+ animals) and 586 (22 C+ animals) of them were located in farms in Burgos and Soria, respectively. Annual number of animals subjected to bacteriological culture ranged between 165-679 (yearly median C+ animals=15) and 73-238 (yearly median C+ animals=6) in Burgos and Soria, respectively.
bTB-confirmed animals located in Burgos were significantly (Mann Whitney test, p-value <0.001) older (median= 7.5 years, IQR [3.9-10.1]) than unconfirmed animals (median= 4.7 years, IQR [2.5-8.3]), while no significant differences were found between bTB-confirmed and unconfirmed animals in Soria (median= 4.3 years for both categories).
Poisson spatial models identified the presence of clusters of bTB-confirmed animals in the north-east part of Soria both when analyzing data of the cumulative 2015-2018 time-span or separately for years 2015 and 2016 (Figure 2). In addition, non-confirmed reactors were also clustered in the cumulative 2015-2018 dataset and in each of the years of study in different areas in the north and center part of the province. Same pattern was observed for both bTB-confirmed and unconfirmed herds in Burgos (Figure 3).
Among the variables analyzed in the univariable models (Table 1), years in the herd and bTB-confirmation in the herd in the previous year were selected for inclusion in the multivariable analysis in Soria (Table 2). County level prevalence in the previous year was also considered in Burgos (Supplementary material 1, Supplementary material 2). The model with bTB-confirmation in the herd in the previous year and the model with county level prevalence in the previous year had the best fit in the final multivariable models in Soria and Burgos, respectively (Supplementary material 2).
Moran’s I test performed on the standardized residuals of the final models for Soria and Burgos revealed no spatial autocorrelation once the effect of the selected variables had been considered in the model. According to the final models, animals in herds in which bTB infection had been confirmed the previous year had a 90% reduction in the odds of being a confirmed in the following year in Soria, while in Burgos this variable was not associated with the outcome. The risk of finding positive culture results in reactors animals increased in animals located in counties in Burgos with a bTB-prevalence higher than 0.5%, while in Soria this variable produced no effect. High (in Soria and Burgos) and low (in Soria) risk bTB-confirmation clusters were found, but when accounting for the effect of significantly associated variables to bTB-confirmation, the odds of being a confirmed animal was similar regardless of the location of the herd. In general, we conclude that the previous history of bTB and prevalence in the area seem to be associated with the success in bTB-confirmation in Soria and Burgos, respectively. Notwithstanding, other variables (i.e., animals in herds located within/outside a cluster of bTB-confirmation, reactors to either SIT or IFN-ƴ) should be considered to further assess whether the risk of being a bTB-confirmed animal is due to the lack of performance of diagnostic tests in Soria

Participantes:

	Servicio de Micobacterias (MYC). Centro de Vigilancia Sanitaria Veterinaria (VISAVET). Universidad Complutense (UCM).

	MAEVA SERVET, S.L..

	Ministerio de Agricultura, Pesca y Alimentación (MAPA).

	Junta de Castilla y León.

	Departamento de Sanidad Animal. Facultad de Veterinaria. Universidad Complutense (UCM).