Preventive Veterinary Medicine 173 (2019) 104805 Contents lists available at ScienceDirect Preventive Veterinary Medicine journal homepage: www.elsevier.com/locate/prevetmed Evaluation of the effectiveness of the surveillance system for tuberculosis in T cattle in Spain Sebastian Nappa,*, Giovanna Ciaravinob, Bernat Pérez de Vala, Jordi Casala,b, José Luis Saézc, Ana Albaa a IRTA, Centre de Recerca en Sanitat Animal (CReSA, IRTA-UAB), Campus de la Universitat Autònoma de Barcelona, Bellaterra, Spain b Department of Animal Health and Anatomy, Faculty of Veterinary Medicine, Universitat Autònoma de Barcelona (UAB), Bellaterra, Barcelona, Spain c Ministry for Agriculture, Fisheries and Food (MAPA), Madrid, Spain A R T I C LE I N FO A B S T R A C T Keywords: In Spain, the eradication of tuberculosis (TB) in cattle has had some setbacks and therefore we need to re- Surveillance evaluate the current surveillance strategies. The Spanish cattle TB surveillance system consists of three com- Animal tuberculosis ponents: a) routine skin testing, b) slaughterhouse surveillance, and c) pre-movement controls. Our objectives Bovine tuberculosis were to assess the effectiveness of the whole surveillance system and evaluate the relative contribution of each of Cattle those components, both at national and at the province level. The effectiveness was estimated based on both the Stochastic model sensitivity of detection per year and the time until detection. Moreover, we evaluated the impact of different factors on that effectiveness. We used a stochastic model that simulated the spread of Mycobacterium tuberculosis- complex (MTC) infection within cattle herds that was modified to incorporate the three components of TB surveillance. Input data, at the province level, on cattle movements off-farm or the frequency of the routine screening were provided by the Spanish Ministry for Agriculture, Fisheries and Food. Under the current conditions, 96.1% of Spanish herds were detected within one year after their infection (i.e. mean sensitivity of TB surveillance), although that detection took on average 178 days. The surveillance system effectiveness was highly dependent on the routine skin testing, responsible for the detection of almost 90% of the infected herds, while slaughterhouse surveillance and pre-movement controls contributed only to the identifi- cation of a small proportion of infected herds. We observed substantial differences in the effectiveness of the surveillance components among Spanish provinces, although in general, the sensitivities were high. The most influential factor on the efficiency of TB detection in Spain was, by far, the frequency of routine controls, followed by the sensitivity of the test used. In a context of reduced funding for cattle TB eradication, the frequency of testing should be adapted based on risk-based surveillance strategies, i.e. efforts should focus on herds more likely to be infected, but also in herds more likely to infect other herds. While slaughterhouse surveillance seems a cost-effective strategy, the use of pre-movement testing at least in areas of low incidence should be further evaluated. 1. Introduction zoonotic TB) which makes the disease a public health concern (Cosivi et al., 1998; Thoen et al., 2010). Traditionally, zoonotic TB was con- Tuberculosis (TB) in cattle, formerly known as bovine TB, is defined sidered to have a much higher impact on low-income countries, as a chronic infectious disease caused by any of the three mycobacterial whereas in high-income countries with mandatory eradication pro- species within the Mycobacterium tuberculosis-complex (MTC): M. bovis, grammes in place TB cases in humans were infrequent (Müller et al., M. caprae and M. tuberculosis (EU - DG SANCO, 2013; EU - DG SANTE 2013; Olea-Popelka et al., 2017). However, recent studies have high- (European Commission, Directorate-General for Health & Food Safety), lighted that the real burden of zoonotic TB may be underestimated in 2018; OIE, 2018). In addition to cattle, other domestic species, mainly both developing and developed countries, mainly due to technical goats, may be affected. Also wildlife, of which wild boars and red deer constrains in the isolation and differentiation of the MTC members are considered the main reservoirs in Spain (De Mendoza et al., 2006). (Lombardi et al., 2017; Palacios et al., 2016; Olea-Popelka et al., 2017; M. bovis and M. caprae also pose a risk of infection to humans (known as WHO, 2019). ⁎ Corresponding author. E-mail address: sebastian.napp@irta.es (S. Napp). https://doi.org/10.1016/j.prevetmed.2019.104805 Received 3 June 2019; Received in revised form 5 September 2019; Accepted 10 October 2019 0167-5877/ © 2019 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/BY-NC-ND/4.0/). S. Napp, et al. Preventive Veterinary Medicine 173 (2019) 104805 In Spain, according to official reports, the number of confirmed herds (Ciaravino et al., 2018), was combined with another dynamic cases of tuberculosis in humans due to M. bovis/M. caprae were 33 in model that simulated detection by the different components of sur- 2014, 38 in 2015, 35 in 2016 and 55 in 2017 (European Food Safety veillance. The within-herd spread model was fed with transmission Authority and European Centre for Disease Prevention and Control parameters obtained from Spanish herds (see Ciaravino et al., 2018), (EFSA and ECDC, 2018). Given the zoonotic potential and its high while the model that simulated detection was fed with real data from economic impact, the objective in the EU countries is the eradication of Spanish cattle herds provided by the Ministry for Agriculture, Fisheries TB in cattle (Reviriego Gordejo and Vermeersch, 2006). In Spain, era- and Food (MAPA). dication programs with “test-and-slaughter” strategies have been im- Our objectives were to assess the effectiveness of the whole cattle TB plemented for decades, but despite the progressive reinforcement of the surveillance system, and evaluate the relative contribution of each of its program the disease has not yet been eliminated. Between 2005 and components, both at national and at provincial level. The effectiveness 2013, the cattle TB eradication program only achieved a slight reduc- was estimated based on both a) the sensitivity of detection per year and tion in herd prevalence (from 1.52% to 1.39%). However, between b) the time until detection of infected herds. Finally, we evaluated the 2013 and 2016, apparent herd prevalence increased from 1.39% to impact of different factors on the effectiveness of cattle TB surveillance. 2.87%. (Anonymous, 2019v). Moreover, the distribution of the disease Ultimately, the aim is to use those results for the improvement of the in Spain is highly heterogeneous, with herd prevalences close to zero in surveillance and control strategies implemented in Spain for the era- the majority of the northern regions, but above 10% in most of the dication of cattle TB. south-central areas (Anonymous, 2019v). In this context, the measures currently implemented within the eradication program need to be re- 2. Materials and methods evaluated. In Spain, the detection of infected cattle herds relies mainly on the 2.1. Within-herd spread model periodic screening of the Officially Tuberculosis Free (OTF) herds (whole-herd tests) with the single intradermal tuberculin test (SITT) The MTC infection spread within-herds was simulated using a followed by the culling of positive cattle and the establishment of compartmental stochastic SOEI (Susceptible, Occult, Exposed and movement restrictions for infected herds. Herd-testing interval varies Infectious) model (Ciaravino et al., 2018). Occult animals (O) re- between once every two years and twice a year depending on the presented animals that were infected but were not yet detectable by prevalence in the area where the herd is located (Anonymous, 2019v). SITT and were not infectious. Exposed animals (E) represented animals Regular TB testing is complemented by slaughterhouse surveillance, as that were infected and were detectable by SITT but were not yet in- all cattle intended for human consumption undergo post-mortem ex- fectious. Finally, infectious animals (I) represented animals that were amination at the slaughterhouse (Anonymous, 2019v). If lesions com- infected, were detectable by SITT and were infectious. It was assumed patible with MTC infection are detected, samples are collected and sent that the occult and exposed sojourn states followed the Erlang dis- for laboratory confirmation. Moreover, since 2006, the Spanish eradi- tribution; thus, the O and E compartments were divided into 3 se- cation program established measures to prevent MTC spread by cattle quential sub-compartments each (see Ciaravino et al., 2018 for further trade. Thus, with a few exceptions, such as some movements to calf- details). Animals susceptible to MTC infection (S) became occult (O), by fattening units, cattle are subject to SITT before their transportation to contact with infectious cattle at a rate β , the transmission coefficient. other holdings, which also contributes to the detection of infected Occult cattle became exposed (E), at a rate α1. Exposed animals became herds. Consequently, three major components: a) routine (skin) testing, infectious and detectable by SITT (I) at a rate α2 . Animals were born b) slaughterhouse surveillance, and c) pre-movement controls, can be into the susceptible class at a rate μ (Fig. 1). considered within the cattle TB surveillance system in Spain. While it is The model was run in continuous time (with days as units) and clear that the three contribute to the detection of infected herds, their transitions between compartments occurred according to the corre- relative impact has never been evaluated. Once cattle TB has been sponding differential equations using the Gillespie’s direct algorithm detected in a herd by any of these components and the MTC infection (see Ciaravino et al., 2018 for further details) (Keeling and Rohani, has been confirmed, control measures are established. They include the 2007). The values of the transmission parameters (α1, α2 and β ) used in culling of all positive cows and the implementation of movement re- the simulations were randomly drawn from the probability distribu- strictions until the herd has been cleared (for which two consecutive tions estimated by Ciaravino et al. (2018) for Spanish cattle herds. negative whole-herd tests are required) (Anonymous, 2019v). In in- Regardless of the mechanism of introduction of MTC within the herd, fected herds, SITT is used in parallel with the the interferon-γ assay infection was assumed to start with a single infected (occult) animal (IFN-γ) to increase the probability of detection of the infected animals. (time 0) and the model was run for 5 years. The success of TB eradication relies not only on the identification of infected herds but also on the clearance of the herds identified as in- 2.2. Modifications to include the surveillance components fected. In Spain, the persistence of TB in cattle herds, in particular in southern and central regions, poses an important challenge to the era- 2.2.1. Modelling detection at slaughterhouses dication program (Guta et al., 2014). Movement restrictions impose a Province-level data on cattle movements to slaughterhouses in 2017 significant burden for farmers, which adds to the direct cost of the was provided by the MAPA. Data included the average number of eradication program, estimated in 30.7 million euros (19.6 for testing batches sent to the slaughterhouse per herd per year, and the average and 11.1 for compensation), of which 50% is co-financed by the EU number of animals in those batches (sizeSi ). Therefore, for province i , (Anonymous, 2017). given the average frequency of movement of cattle to the slaughter- For surveillance and control of TB in cattle herds, understanding the house, measured in batches transported per year (rateSi ), the time of the dynamics of MTC spread within Spanish cattle herds is essential first movement from a herd to the slaughterhouse (t S1i ) was simulated (Ciaravino et al., 2018). However, the study of cattle TB dynamics is as: hampered by factors such as the long incubation periods, the lack of t S1i ∼ uniform (0, 365/rateSi ) clinical symptoms in infected animals, or the uncertainty about the mechanisms of transmission. That is why mathematical modelling of- At time t S1, the spread of the infection within the herd determined fers a practical option for the study of MTC spread in cattle herds the number of animals in the different compartments (S , O , E and I ). (Brooks-Pollock et al., 2014). To evaluate the effectiveness of the sur- The composition of the herd at time t S1i and the average size of the veillance system for tuberculosis in cattle in Spain, a model previously batches (sizeSi ) defined the composition of the batch (SS1i , O S1i , ES1i and developed to simulate the spread of MTC infection within Spanish cattle IS1i ). 2 S. Napp, et al. Preventive Veterinary Medicine 173 (2019) 104805 Fig. 1. Modifications of the within-herd transmission model developed by Ciaravino et al., 2018 to include births (with birth rate μ ), plus models for the different components of the TB surveillance system: a) whole-herd tests (i.e. routine testing), b) Batches for trade (i.e. pre-movement testing), and c) Batches for slaughter (i.e. slaughterhouse surveillance). The black main box represent the herd composition and simulate the spread of the disease in the herd. Cattle are distributed in the following compartments: S for susceptible animals, O for occult, E for exposed and I for infectious cattle. The probability of detection (by slaughterhouse surveillance) for O H1i , E H1i and IH1i ) was calculated taking into account the average size of that batch at time t S1i (Se S1i ) was given by the probability that at least the batch (sizeHi ) and the composition of the herd at time t H1i . one of the infected animals was detected at the post-mortem inspection: The probability of detection (by pre-movement tests) for that batch at time t H1i (Se H1i ) was be given by: Se S1i ∼ binomial ((O S1i + ES1i + IS1i ), Seslaug ) Se H1i ∼ binomial ((E H1i + IH1i ), SeSITT ) where Seslaug was the sensitivity of post-mortem detection at the slaughterhouse for a single infected cow (Fig. 1). The value of Seslaug where E H1i and IH1i were the number of exposed and infectious cattle on used in the model was 31.4%, which was derived from a study carried the batch subject to pre-movement tests at time t H1i . And where SeSITT out in North-Eastern Spain (Catalonia) (Garcia-Sáenz et al., 2015). The was the sensitivity of the SITT for a single animal. As in (Ciaravino estimated sensitivity of slaughterhouse for a MCT-infected cow was the et al., 2018), the value of SeSITT was set at 94%. After the movement, the product of three probabilities: first, the probability that a MCT-infected composition of the herd was also re-adjusted, and if any infected animal cow from an OTF herd arrived at the slaughterhouse with TB-Macro- (O , E or I ) remained, within-herd spread was resumed. If further scopically Detectable Lesions (MDL); second, the probability that those movements to other herds fell within period considered, the whole MDL were detected by the routine meat inspection; and third, the process was repeated (Fig. 2). probability that the Veterinary Officer suspected of TB and sent a sample to the laboratory for confirmation or directly notified the au- 2.2.3. Routine testing by SITT thorities (Garcia-Sáenz et al., 2015). In Spain, the frequency of routine controls on OTF herds varies After the movement to the slaughterhouse, the composition of the between once every two years and twice a year depending on the herd was re-adjusted by subtracting the number of cattle in the different prevalence in the area (province and county) where the herd is located compartments in the batch (SS1i , O S1i , ES1i and IS1i ) from the number of (Anonymous, 2019v). animals in the different compartments in the herd (S , O , E and I ). If an Therefore, given the frequency of routine controls per year in a herd infected animal (O , E or I ) remained in the herd, within-herd spread in province i (rateCi ), the time of the first routine control (tC1i ) was si- was resumed. If further movements to slaughterhouses fell within the mulated. At time tC1i , the spread of MTC within the herd determined the period considered for the simulation (5 years), the whole process was number of animals in the different compartments (SC1i , OC1i , EC1i and repeated (Fig. 2). IC1i ), and therefore the probability of detection for that first control (SeC1i ): 2.2.2. Modelling of TB detection by pre-movement testing SeC1i ∼ binomial ((EC1i + IC1i ), SeSITT ) Province-level data on cattle movements to other holdings in 2017 was also provided by the MAPA. Data included, for province i , the where EC1i and IC1i were the number of exposed and infectious cattle in average number of batches transported to other holdings per herd per the herd at the time of the first routine control (tC1i ), and SeSITT was the year and the average number of animals in those batches (sizeHi ). Given sensitivity of the SITT for a single animal, which was set at 94%, as in the frequency of movements to other herds (rateHi ), the time of the first Ciaravino et al. (2018). movement (t H1i ) was simulated, and the composition of the batch (S H1i , The times for the subsequent routine controls followed a regular 3 S. Napp, et al. Preventive Veterinary Medicine 173 (2019) 104805 Fig. 2. Timeline summarizing the disease spread and control activities. Since the introduction of TB, infection progresses within the herd. That progress determines the composition of the herd at each point in time, which will influence on the probability of detection by the different components of the TB surveillance system. pattern, where: Then, we also compared how well the overall sensitivity of TB surveillance in a province correlated with the TB herd prevalences in tC2i = tC1i + (365/ rateCi ) that province. In order to do that, we standardized the overall sensi- tC3i = tC2i + (365/ rateCi ) tivities of TB surveillance and TB prevalences in the different provinces (i.e. we set both values in a scale from 0 to 100): For any further routine controls that fell within the period con- Previ sidered for the simulation of within-herd spread, the process was re- Prev. STi = × 100 max (Previ ) peated (Fig. 2). Where, Prev. STi was the standardized prevalence for province i , and 2.2.4. Further modifications of the TB within-herd model Previ the prevalence for province i . The inclusion of animal movements to both slaughterhouses and SeALLi other holdings implied the progressive decrease of the number of ani- Se . STi = × 100 max (SeALLi ) mals in the herd. To avoid that, a crude birth rate ( μ ) term (Keeling and Rohani, 2007) was included in the equation that determines the var- Where, Se . STi was the standardized sensitivity for province i , and SeALLi iation in the number of susceptible animals, therefore implying that all was the overall sensitivity of the surveillance (i.e. including all com- animals were born as susceptible. The differential equation for sus- ponents) for province i . ceptibles was then modified to: Then, we estimated, for each province, a parameter that we called discrepancy as: dS βS (t ) I (t ) =μ− dt N (t ) Discrepancyi = Se . STi − Prev. STi To allow the maintenance of a more or less constant number of That allowed provinces to be identified where sensitivity of TB animals, μ was estimated as a function of the number and size of the surveillance was lower than expected according to their prevalence of batches sent to both slaughterhouses and to other herds. As those herds infected with MTC. parameters vary per province, the values of μ also vary: μi = (rateSi × sizeSi ) + (rateHi × sizeHi ) 2.4. Evaluation of the factors influencing the effectiveness of TB surveillance Finally, by allowing different model parameters to vary within their 2.3. Evaluation of the effectiveness of TB surveillance at province level and range of values, the influence of those factors on the effectiveness of TB nationally detection (objective 3) were evaluated. The parameters assessed in- cluded: At each iteration, the model simulated the spread of MTC infection Sensitivity of the routine test: A value of 94% was the average es- within a cattle herd, as well as movements to other holdings or timate given in the review carried out by the European Food Safety slaughterhouses and routine SITT controls (Fig. 2). The results of Authority (EFSA) (EFSA, 2012). However, in field conditions, the dif- whether detection by any of the components occurred, were recorded, ficulties of carrying out the skin tests in extensively-reared cattle and the effectiveness of surveillance, either of the whole system or the (Álvarez et al., 2012a), or the pressure faced by the veterinarians when different components individually, was evaluated based on the sensi- performing the tests in the presence of farmers (Ciaravino et al., 2017), tivity of surveillance and time between infection and detection. may result in much lower sensitivities. Therefore, in addition to the As the input values fed to the model (e.g. frequency of movements value of 94%, we included another two scenarios in which the sensi- to slaughterhouses and to other herds, or frequency of routine SITT tivity of the SITT was reduced to 70% and 50%. Furthermore, in Spain, controls) represented the average values for the different provinces in the use of the single intradermal comparative cervical test (SICCT) with Spain in 2017, we were able to estimate the effectiveness of each of the a severe interpretation can be authorized for TB detection in areas with surveillance components and of the whole system at province level. herd prevalences below 1% (Anonymous, 2019v). The SICCT requires That allowed the assessment of the spatial variation in the effectiveness the simultaneous injection of bovine and avian tuberculin, and its in- of TB surveillance among the different provinces in Spain (objective 2). terpretation is based on the observation that MTC-infected cattle tend to Then, by combining the province-level results, weighted by the number show a greater response to the bovine tuberculin than to the avian of cattle herds in each province, the average values of effectiveness of tuberculin (De la Rua-Domenech et al., 2006). It allows minimizing the surveillance for Spain, for each component and overall (objective 1) risk of false-positive results caused by other mycobacteria, but increases were obtained. the risk of false-negative results, giving an average sensitivity of 61% 4 S. Napp, et al. Preventive Veterinary Medicine 173 (2019) 104805 Table 1 Data inputs for models. For objective 1 and 2 we used the range of values for the different provinces (minimum and maximum values represented in the table). For objective 3 we also used the median values for the parameters related to movements to slaughterhouses and other herds. Average herds Routine SITT Movements to Average size batches to Movements to other Average size batches size controls (per year) slaughterhouses (per year) slaughterhouses holdings (per year) to other holdings Minimum (provinces) 6 0.5 0,1 1 0.01 1 Maximum (provinces) 217 2 61,5 4 14.2 8 Median (provinces) 72 1.1 0.68 2 2.5 4 (with a severe interpretation) (EFSA, 2012). This value of 61% was also components and the overall surveillance system in Spain are shown in incorporated in the evaluation. Table 2. Routine testing was by far the most effective component, Frequency of routine controls: In most Spanish provinces OTF herds whereas slaughterhouse surveillance and pre-movement testing com- are tested once a year. In the provinces of Tenerife, Las Palmas, Islas ponents had a much lower contribution to the effectiveness of the Baleares, Pontevedra and Guipúzcoa the majority of OTF herds are system. Given the current conditions in Spain, we estimated that more tested once every two years, while in Extremadura and Andalucía re- than ninety-six percent of herds would be detected within the first year gions most OTF herds are tested twice a year. Therefore, for the as- after their infection (i.e. > 96% sensitivity), and detection would take, sessment of the influence of the frequency of routine controls, we in- on average, 178 days. cluded three scenarios: once every two years, once a year and twice a year. Frequency of movements and size of the batches: For the frequency 3.2. The effectiveness of the TB surveillance system among Spanish of cattle movements to slaughterhouses (rateS ), the frequency of cattle provinces (Objective 2) movements to other herds (rateH ), the average size batches to slaugh- terhouses (sizeS ) and the average size batches to other herds (sizeH ), 3.2.1. Effectiveness of the TB surveillance by province three scenarios corresponding to low, medium and high frequency, The values of sensitivity for the different components of the TB were considered. They correspond to the minimum, maximum and 50th surveillance system varied significantly among provinces (Table 3; percentile of the values provided by the MAPA for the different pro- Fig. 3). vinces (Table 1). There was a significant variation in the effectiveness of the slaughterhouse surveillance by province (Table 3; Fig. 3A). The pro- vinces with the highest sensitivities were León, Salamanca, Ciudad Real 2.5. Specifications of the models and Cáceres (Fig. 3A), which reached values above 25%. The sensitivity of pre-movement detection also varied significantly among provinces Iterations were carried out in the following way: (Table 3; Fig. 3B). The provinces with the highest sensitivities (above • Input selection: First, the values of the transmission parameters (α , 20%) were located mainly in south-western Spain (e.g. Huelva, Cáceres, 1 Ciudad Real and Albacete) (Fig. 3B). The sensitivities of routine testing α2 and β ) were randomly selected from their probability distribu- were considerably higher than the other two components, with values tions. Second, a Spanish province was randomly chosen and its in- greater than or equal to 95% in 25 out of 50 Spanish provinces puts (herd size, rates of movements to slaughterhouses or other (Fig. 3C); only in five provinces (Guipúzcoa, Pontevedra, the Balearic herds, or batch sizes) selected. • Islands, Tenerife and Las Palmas) sensitivities were in the range of 50% Then, MTC spread as well as movements to slaughterhouses or other (Fig. 3C) but those provinces were actually free of TB. The overall holdings were simulated. • sensitivity of the TB surveillance system in Spain was above 95% for the Finally, detection by routine SITT, slaughterhouse detection and majority of provinces (Table 3; Fig. 3D). pre-movement tests were also simulated. Accordingly, the times to detection of infected herds also varied significantly among provinces, for both the whole system and for each To achieve the convergence of the output parameters, a large of the different components (Table 3; Fig. 4) and, as expected, the times number of iterations were run (100,000 for objectives 1 and 2, and were inversely correlated to the sensitivities of the surveillance com- 240,000 for objective 3). Moreover, convergence was later assessed by ponents. The times to detection of infected herds by slaughterhouse splitting iterations in two and comparing the variation in the average surveillance were more than 600 days for all Spanish provinces, and for output values (Gelman and Rubin, 1992). All models were implemented 38 out of 50 provinces times were more than 800 days (Fig. 4A). The within the R environment version 3.2.1 (R Core Team et al., 2015). times to detection by pre-movement testing were more than 900 days in most of the Spanish provinces, and only a few provinces could detect 3. Results the infection of their herds by this component in less than 700 days on average (Fig. 4B). On the other hand, times to detection by the routine 3.1. The effectiveness of the cattle TB surveillance system in Spain (skin) testing were much lower, and the 80% of provinces were able to (Objective 1) detect infected herds by this component in less than 250 days (Fig. 4C). Considering the whole surveillance system for TB, the mean times to The results of the evaluation of the effectiveness of the different detection were even lower (Fig. 4D), and only in the five provinces that Table 2 Average values of effectiveness of TB surveillance in Spain. Results for the different components, as well as for the overall surveillance system. Sensitivity (%) Mean time to detection (days) First detected by (%) Not detected within 5 years (%) Slaughterhouse detection 9.5% 1124 4.5% 24.6% Pre-movement testing 10.8% 1173 5.7% 30.7% Routine testing 94.3% 208 89.0% 2.8% Overall TB surveillance 96.1% 178 NA 1.2% 5 S. Napp, et al. Preventive Veterinary Medicine 173 (2019) 104805 Table 3 combination of high overall sensitivities of the surveillance system Average values of effectiveness of TB surveillance at herd level in the different (above 95%) and very low herd prevalences (below 0.5% in 2017). Spanish provinces (objective 2). Results for the different components, as well as Those provinces were located mainly in northern Spain (Fig. 5). for the overall surveillance system. Sensitivity (%) Mean time to detection (days) 3.3. Factors influencing the effectiveness of TB surveillance in Spain (Objective 3) Slaughterhouse Minimum 1.1% 678 detection Maximum 32.1% 1684 Pre-movement testing Minimum 1.4% 454 The higher the sensitivity of the diagnostic test used, the higher the Maximum 44.6% 1725 efficiencies of both the routine herd testing and pre-movement detec- Routine testing by SITT Minimum 46% 85 tion components. For a test-sensitivity of 50%, the overall sensitivity of Maximum 99.7% 489 TB surveillance was 74.6% and the mean time for detection was 227 Overall TB surveillance Minimum 50.7% 82 Maximum 99.8% 367 days, while for a test-sensitivity of 94%, the overall sensitivity of TB surveillance was 88.1% and the mean time for detection was 153 days (Table 4). were free of cattle TB in 2017 (Guipúzcoa, Pontevedra, the Balearic The higher the frequency of routine controls, the higher the sensi- islands, Tenerife and Las Palmas) detection of infected herds took tivity of herd testing and the overall sensitivity of TB surveillance. For longer than 300 days. one control every two years, the overall sensitivity of TB surveillance was 64.3%, while under a testing interval of six months, the overall sensitivity of TB surveillance was 93.7% (Table 4). That allowed re- 3.2.2. Discrepancy between sensitivity of detection and cattle TB herd ducing the time to detection from 280 days with one control every two prevalence years, to only 115 days with two controls per year. We also evaluated whether the intensity of surveillance efforts in The increase in the frequency of cattle movements to the slaugh- each province was adequate to its TB herd prevalence (i.e. the pro- terhouse had a huge effect on the sensitivity of slaughterhouse detec- portion of infected herds), by calculating the parameter called dis- tion, but the effect on the overall sensitivity was quite limited (Table 4). crepancy (Fig. 5). On the other hand, an increase in the average size of the batches sent to In the majority of Spanish provinces, the high overall sensitivity of slaughterhouses resulted in a slight increase in the sensitivity of the TB surveillance system prevented negative discrepancy values, slaughterhouse detection, while it did not affect the overall system which would be indicative of a high TB herd prevalence together with a sensitivity. The increase of the frequency of cattle movements to other relatively low sensitivity for the detection of infected herds. We iden- holdings had a very large effect on the sensitivity of pre-movement tified only one province with negative discrepancy (Guadalajara, with a testing, but also a significant effect on the overall sensitivity of TB value of -5) (Fig. 5). This was the result of a relatively high sensitivity of detection (11.0% difference) (Table 4). The increase in the average size TB detection (95.1%) combined with an extremely high TB herd pre- of the batches sent to other herds also resulted in a large increase in the valence (28.6%), the highest in Spain in 2017. Eleven provinces were sensitivity of pre-movement testing and some effect on the overall associated to very high values of discrepancy (above 95) because of the sensitivity of TB detection (Table 4). Fig. 3. Maps of sensitivities of the different TB surveillance components and overall sensitivity by province. A: Slaughterhouse detection; in the map, the following provinces are indicated: León (S1), Salamanca (S2), Cáceres (S3) and Ciudad Real (S4). B: Pre-movement detection; in the map, the following provinces are indicated: Huelva (H1), Cáceres (H2), Ciudad Real (H3) and Albacete (H4). C: Routine SITT test detection; in the map, the following provinces are indicated: Guipúzcoa (R1), Pontevedra (R2), the Balearic Islands (R3), Tenerife (R4) and Las Palmas (R5). D: Whole TB surveillance system (overall sensitivity). ** Be aware that the scales for the maps are different **. 6 S. Napp, et al. Preventive Veterinary Medicine 173 (2019) 104805 Fig. 4. Maps of times to detection for the different components of the TB surveillance system. A: Slaughterhouse detection. B: Pre-movement detection. C: Routine SITT test detection. D: Whole TB surveillance system. ** Be aware that the scales for the maps are different **. Convergence of output values was demonstrated by splitting itera- the field were largely caused by lack of knowledge and training among tions in two and comparing the variation in the average sensitivities. veterinarians. For objective 1 (whole of Spain) the mean sensitivities differed in Another factor that has influenced the lack of progress in TB era- 0.013%, for objective 2 (provinces) 0.86%, and for objective 3 (para- dication is the time it takes to clear the infected herds once they have meters) 0.17%. been detected. In Spain, the persistence of MTC in cattle herds, in particular in southern and central regions, poses an important challenge to the eradication of cattle TB (Guta et al., 2014). Residual infection 4. Discussion could be the result of the presence of false negatives to the skin test, reviewed by De Mendoza et al. (2006), or be the consequence of the In recent years, the progress towards the eradication of cattle TB in incorrect application of the test (Humblet et al., 2009). Indirect trans- Spain has had some setbacks, particularly in some regions where the mission caused by the persistence of the microorganism in the en- prevalence has increased significantly. Given such difficulties, there is a vironment could also result in residual infections (Courtenay et al., need to re-evaluate the strategies currently implemented. 2006). Moreover, the presence of infected goats on the farm could Our results indicate that, under the present conditions in Spain, the contribute to the recirculation of MTC within cattle herds (Crawshaw sensitivity of the TB surveillance system was quite high (96.1%). et al., 2008). However, it should be noted that the evaluation of the sensitivity of a The evaluation of the relative contribution of the three components surveillance system requires the definition of the period over which evidenced that routine testing was by far the most sensitive surveillance data are analysed (Cameron et al., 2014), and for cattle TB we chose a component (94.3% sensitivity). Almost 90% of the infected herds were relatively long period (one year). Choosing a long time period increases first detected by this component, and that took on average 208 days the apparent sensitivity of the surveillance system (Cameron et al., after the infection of the herd. Therefore, the detection of MTC-infected 2014). In fact, the detection of infection in the herds took on average herds in Spain seems to be highly dependent on routine testing. This is 178 days, and during that period transmission to other herds may consistent with the fact that in the countries where TB in cattle has been occur. eradicated, that was achieved mainly through the regular skin testing of Despite that 96.1% sensitivity, between 2013 and 2016, the cattle cattle and the elimination of infected animals. In all Spanish provinces, TB herd prevalence in Spain went from 1.39% to 2.87%. That apparent the routine testing was the most sensitive surveillance component and it increase is likely to have been associated with the improvement of the strongly influenced the overall sensitivity of the system in the province. sensitivity of the surveillance system after the implementation of a In general, routine testing sensitivities were relatively high, with values series of reinforcement measures (EU, 2015, 2019a). Measures in- above 90% in more than 80% of provinces. Those sensitivity values cluded: compulsory training courses for the veterinarians who perform were clearly associated with the frequencies of testing. In fact, the only the SITT controls, official control of field veterinarians, training courses five provinces with sensitivities of routine testing below 50% were for official veterinarians, audits of the implementation of the eradica- those that had a prevalence of zero in 2017, and where most cattle tion program by the different regions carried out by the MAPA, random herds were tested only once every two years. This was further con- post-movement tests or additional control measures to herds sharing firmed by the results of objective 3, which indicated that the frequency pastures. In fact, a recent survey of farmers and veterinarians indicated of testing was the most influential factor on the effectiveness of the that they perceived an improvement in the application of the Spanish routine testing component, and therefore of the whole TB surveillance cattle TB eradication programme in recent years, which they mainly system. As a matter of fact, changing from two controls per year to one attributed to the organization of the mandatory training courses control every two years reduced the overall sensitivity of TB (Ciaravino et al., 2017). They acknowledged that some bad practices in 7 S. Napp, et al. Preventive Veterinary Medicine 173 (2019) 104805 Fig. 5. Map showing the results obtained for the parameter discrepancy by provinces. A: Map of prevalence of herds infected with MTC in 2017 in Spain. B: Map of the overall sensitivity for TB in 2017. C: Map of discrepancy; the province of Guadalajara (D1) is indicated in the map. surveillance by almost 30%, and increased the time to detection from Humblet et al., 2009; Meskell et al., 2013). Furthermore, it has been 115 to 280 days. That was in agreement with previous studies (Fischer pointed out that in Spain the correct execution and interpretation of the et al., 2005; Rossi et al., 2015) that found that the variations in the SITT results might be influenced by the pressure linked with the “pa- sensitivity of TB surveillance were primarily related to the frequency of tronage relationship” between farmers and private veterinarians, be- testing. Routine testing represents a significant part of the cost of the cause the latter carry out other duties besides TB testing, and are paid eradication program in Spain (23.7 out of the 38.8 million euros) by farmers (Ciaravino et al., 2017). Also, factors related to the profes- (Anonymous, 2019v), and that figure is clearly dependent on the fre- sional skills and awareness of veterinarians may compromise the ap- quency of testing of herds. However, any increase in the time between plication of the SITT in the field (Ciaravino et al., 2017; Humblet et al., SITT controls will come at the cost of a significant reduction in the 2011a, 2011b; Meskell et al., 2013). Also, exposure to the parasite effectiveness of cattle TB detection. Fasciola hepatica is known to interfere with the diagnosis of cattle TB The sensitivity of the test used for the detection of MTC-infected (Claridge et al., 2012). F. hepatica is considered widespread in Europe herds had a significant influence on the sensitivity of the routine but with a heterogeneous spatial distribution and significant regional testing, but also on the pre-movement component, and as a result, af- variations in prevalence (Beesley et al., 2018). In Spain, the spatial fected the overall sensitivity of surveillance. However, in comparison distribution and prevalence of F. hepatica are not known, and therefore with the frequency of controls, the sensitivity of the test used had a its influence on TB detection is difficult to predict. Spatial models in- much lower impact. The sensitivity of the SITT is generally considered dicate relatively high risks in areas of northern Spain (Ducheyne et al., as high, in fact, a value of 94% (based on the review by EFSA) was used 2015) where the prevalence of cattle TB is quite low. for the evaluation of the cattle TB surveillance (at national and pro- The estimates of the effectiveness of cattle TB surveillance in Spain vincial levels). However, much lower values have been reported when were based on the assumption of a sensitivity of 94% for the SITT, so the SITT is applied in field conditions (Álvarez et al., 2012b; Humblet any reduction of that value, for whatever reason, would result in a et al., 2011b). Among the reasons for the reduction of the sensitivity is lower sensitivity of the cattle TB surveillance and longer times for the the management and on-farm testing conditions, which can make the detection of infected herds. SITT difficult to perform, in particular in some production types such as In countries such as the UK and Ireland, the single intradermal extensively managed beef herds and fighting bulls (Álvarez et al., 2014; comparative cervical test (SICCT) is the primary screening test for TB in 8 S. Napp, et al. Preventive Veterinary Medicine 173 (2019) 104805 Table 4 out by the European Commission on the progress of the Spanish pro- Results of the evaluation of the influence of different factors on the effectiveness gram to eradicate bovine tuberculosis indicated that the optimization of of TB surveillance (objective 3). Where SeTEST was the sensitivity of routine the sensitivity of this component was still far for being accomplished testing, SeHERD was the sensitivity of pre-movement testing and SeSLAUG was the (EU - DG SANTE, 2016). Reasons that may have contributed to the low sensitivity of slaughterhouse detection. rate of detection include the lack of competence/awareness of meat Categories Value Overall Se specific inspectors and veterinarians or inadequate facilities/conditions (e.g. Se component lighting or line speed) (EU - DG SANCO, 2013; Garcia-Sáenz et al., (SeALL) 2015; Hadorn and Stärk, 2008). Moreover, in Spain, some deficiencies Sensitivity of the Minimum 50% 74.6% SeTEST = 56.0% & in the coordination between the authorities responsible for the TB routine test SeHERD = 22.9% eradication programme (animal health authorities) and those re- Other (SICCT) 61% 79.0% SeTEST = 61.6% & sponsible for slaughterhouse inspection (food safety authorities) have SeHERD = 25.8% also been identified (EU - DG SANTE, 2016). Medium 70% 82.0% SeTEST = 65.3% & SeHERD = 28.0% At the provincial level, the sensitivities of slaughterhouse detection Maximum 94% 88.1% SeTEST = 73.9% & were quite heterogeneous, although no clear spatial pattern was ob- SeHERD = 32.7% served. While the number of movements to the slaughterhouse had a Frequency of the Minimum 0.5 64.3% SeTEST = 35.4% large effect on the sensitivity of the slaughterhouse detection and lim- routine controls Medium 1 84.8% SeTEST = 70.5% ited on the sensitivity of the whole system, the size of the slaughter- (per year) Maximum 2 93.7% SeTEST = 86.8% Frequency of cattle Minimum 0.1 80.3% SeSLAUG = 0.4% house batches only had a moderate influence on the slaughterhouse movements to Medium 2.5 80.1% SeSLAUG = 2.4% component. In any case, even in those provinces where the sensitivities slaughterhouses Maximum 61.5 82.4% SeSLAUG = 58.4% of slaughterhouse detection were highest (around 30% in several pro- (per year) vinces), it took around two years for this component to detect infected Average size batches to Minimum 1 81.4% SeSLAUG = 16.3% slaughterhouses Medium 2 80.9% SeSLAUG = 21.0% herds. However, despite its apparent limited contribution to the overall Maximum 4 80.4% SeSLAUG = 23.6% sensitivity of the TB surveillance system, the slaughterhouse detection Frequency of cattle Minimum 0.01 75.9% SeHERD = 0.2% may still play an import role in the detection of “anergic” animals movements to other Medium 0.7 80.0% SeHERD = 23.3% (Domingo et al., 2014). These are chronically infected animals in which herds (per year) Maximum 14.2 86.9% SeHERD = 59.1% cell-mediated immune response may be depressed, and therefore may Average size batches to Minimum 1 78.3% SeHERD = 13.8% other herds Medium 4 81.6% SeHERD = 30.0% not be detectable by skin test or gamma-interferon, but which are likely Maximum 8 82.9% SeHERD = 38.3% to have developed macroscopic lesions and therefore be detected by post-mortem inspection at the slaughterhouses (De la Rua-Domenech et al., 2006). Moreover, according to the EU report, slaughterhouse cattle (De la Rua-Domenech et al., 2006; Frankena et al., 2007). In surveillance is in place for the detection of diseases other than TB, and Spain, the SICCT can be authorized in areas of low herd prevalence it has proven its capacity to find MCT-infected animals that otherwise (Anonymous, 2019v). While the use of the SICCT increases the speci- would have been missed (EU - DG SANTE (European Commission, ficity as compared with the SITT, it comes at the price of a reduced Directorate-General for Health & Food Safety), 2016). sensitivity (EFSA, 2012). In fact, in our scenarios, the average time to Contrary to our findings, studies carried out in Belgium (Welby detection of TB infected herds was estimated to be 153 days with the et al., 2012), Denmark (Calvo-Artavia et al., 2013) or Canada (El Allaki SITT assuming a 94% sensitivity, while it increased to 204 days with a et al., 2016) estimated that the slaughterhouse surveillance was highly test sensitivity of 61% (SICCT with the severe interpretation) and to 227 effective. However, those countries were OTF and therefore their sur- days when the test sensitivity decreased to 50% (SICCT with the stan- veillance systems did not include the same components as in Spain. In dard interpretation). In Spain, given the difficulties in the later stages of fact, some of them (Calvo-Artavia et al., 2013; El Allaki et al., 2016) did cattle TB eradication, the objective should be the maximization of not consider periodic tuberculin screenings but only the testing of im- sensitivity, and therefore the use of SITT rather than SICCT is re- ported animals, while other (Welby et al., 2012) considered reduced commended. herds screenings. Furthermore, in the Belgian study (Welby et al., 2012) One alternative for increasing the sensitivity of routine testing the sensitivity of detection at the slaughterhouse for an individual an- would be the use of the IFN-γ in combination with the SITT. However, imal (Seslaug ) was assumed to range between 50% and 99%, with a most there are several factors that limit the systematic use of IFN-γ for rou- likely value of 70%, while in our study we used a value of 31.4%, de- tine testing, among them, its highest cost, the logistic difficulties for rived from the study carried out in North-Eastern Spain (Garcia-Sáenz sending blood samples to the laboratory on time from remote areas, as et al., 2015). In agreement with our results, other studies (Fischer et al., well as the increased risk of false positives (Ciaravino et al., 2017). That 2005; Rossi et al., 2015; Schöning et al., 2013; Van Asseldonk et al., is why in Spain the routine testing of herds is carried out using the 2005) have highlighted the limitations of the slaughterhouse surveil- tuberculin test (usually the SITT), while the IFN-γ is used as a com- lance for TB detection. In any case, the comparison with studies carried plementary test to the tuberculin, but only in TB-positive herds out in other countries is difficult due to the variability in the meth- (Anonymous, 2019v). Given its importance, any measure to increase odologies applied and therefore interpretations should take with cau- the sensitivity of the test in the field would improve the effectiveness of tion. cattle TB detection. Therefore, initiatives such as compulsory training Our estimates indicated that in Spain the sensitivity of pre-move- courses for the veterinarians who perform the SITT controls or official ment testing was 10.8%, and it took over 3 years for this component to control of field veterinarians seem a cost-effective alternative to im- detect the infection of a herd. That meant that 5.7% of infected herds prove cattle TB detection. were first detected by tests prior to the movement of cattle to other The slaughterhouse surveillance component also contributed to the herds. That result is in line with the proportion of herds estimated to be detection of MTC infected herds, although its average sensitivity was detected by this component in the report of the EU on the progress of 9.5% and the mean time to detection was longer than three years. We the Spanish eradication program for the period 2012–2014, 5.2% in estimated that the slaughterhouse was responsible for the first detection 2012, 3.9% in 2013, and 3.7% in 2014 (EU - DG SANTE (European of only 4.5% of TB infected herds. In 2017, according to the MAPA, Commission, Directorate-General for Health, 2016). The sensitivities of lesions compatible with cattle TB were observed in animals from 282 pre-movement detection varied significantly among provinces (between OTF-herds, of which 119 herds were confirmed as infected by MTC by 1.4 and 44.6%), with larger values generally observed in south-western the laboratory (Anonymous, 2017).The conclusions of the audit carried Spanish provinces. In the case of pre-movement detection, both the 9 S. Napp, et al. Preventive Veterinary Medicine 173 (2019) 104805 number of movements and the size of the batches had an effect on the regularly (Anonymous, 2019v). While some herd-level factors related to sensitivity of the whole surveillance system. In Spain, the low sensi- the risk of infection and further transmission are known, further studies tivity of pre-movement testing is the result of a) the small size of the are needed for their proper quantification. That would allow a better batches (median value of 4) as compared to the size of the herds estimation of the most cost-effective frequency of testing depending on (median value of 72); and b) the low prevalence in infected herds as a the characteristics of the herds. result of the combination of the slow spread of MTC infection and the For the estimation of discrepancy, we used provinces as units be- implementation of regular controls for many years. That results in a cause they were the smallest geographical unit for which we had small probability that the batch of animals selected to be transported to complete data. However, in reality, there may be significant differences other herd contains an infected animal. Furthermore, if we also take in the prevalence of infected herds, as well as in the frequency of testing into account that in many areas of Spain the incidence (i.e. probability among the smaller geographical units (called comarcas) within pro- of appearance of newly infected herds) (Anonymous, 2019v) is very low vinces, and that may have some influence on the results. (way below 1% for all the northern regions), the contribution of pre- The understanding of the efficacy of the different components of TB movement testing to the overall system is limited. In fact, according to surveillance and their drivers allows adjusting the level of surveillance the MAPA, in 2017, pre-movement tests were carried out in 16,202 to the risk of infection in the area, maximizing the cost-effectiveness of herds and 203,989 animals, resulting in 72 positive herds (8 confirmed) cattle TB detection and contributing to the eradication of the disease in and 103 positive animals (Anonymous, 2017). Therefore, the cost-ef- the long term. An improved characterization of the risk of cattle TB at fectiveness of pre-movement testing, at least in areas of low incidence, herd level, based not only in the prevalence in the area, but in other should be evaluated taking into account not only its contribution to the characteristics of the herd (e.g. type of production or size) would allow detection, but also its cost, and that should be compared to the cost- an even better allocation of resources for the surveillance of TB. effectiveness of alternative options (e.g. increasing the frequency of A stochastic model was used to simulate the spread of MTC infection routine testing). within herds, which allowed the estimation of the effectiveness of cattle The distribution of TB in Spain is highly heterogeneous (Allepuz TB surveillance. Therefore, the model assumptions (e.g. homogeneous- et al., 2011; Garcia-Sáenz et al., 2014), and in fact, the Spanish eradi- mixing model with frequency-dependent transmission) may have had cation program is designed to account for that heterogeneity by al- an influence on the results. lowing changes in the frequency of routine testing according to the prevalence in the area (Anonymous, 2019v). Our results evidence that, 5. Conclusions in general, the sensitivities of cattle TB detection in the provinces with the highest TB herd prevalences were also quite high. The exception • Under the current conditions in Spain, and considering a sensitivity was the province of Guadalajara, which had the highest TB herd pre- of 94% for the SITT, 96.1% of herds were detected within one year valence in Spain in 2017 (28.6%), but the sensitivity of TB detection after their infection (i.e. mean sensitivity of TB surveillance), al- (95.1%) was not among the highest. However, in several provinces with though that detection took on average 178 days after the infection of very low TB prevalence, the frequencies of SITT, and therefore the the herd. overall sensitivities, were quite high resulting in large positive dis- • Variations in the efficiency of the components and overall surveil- crepancy values. The budget provided by the EU for cofinancing the lance were observed at the provincial-level, although in general, the Spanish eradication programme has progressively decreased (14.4 sensitivities were high, with the exception of OTF provinces where million in 2016, 13.2 in 2017, 13.4 in 2018 and 11.9 in 2019) (EU, the frequency of routine testing was reduced to one control every 2016, 2017; EU, 2018, 2019b). To cope with this reduction, a two years. straightforward solution could be to decrease the frequency of testing in • Routine testing was by far the most sensitive surveillance compo- areas of low prevalence. However, that would increase the time to the nent (94.3% sensitivity). Almost 90% of infected herds were first detection of the infected herds facilitating the spread of the infection, detected with routine testing; on average it took 208 days until in- which may compromise cattle TB eradication in those areas where the fected herds were detected using this method. Therefore, the de- disease was close to elimination. Surveillance of low-prevalence dis- tection of MTC-infected herds in Spain was highly dependent on eases is problematic because large sample sizes are required to achieve routine testing by SITT. an acceptable level of sensitivity, but at the same time, surveillance • The frequency of testing was the most influential factor in the effi- needs to be affordable (Cameron et al., 2014). In this context, the only ciency of routine testing. However, in the context of reduced option may be to use risk-based surveillance strategies, i.e. reducing the funding for cattle TB eradication, rather than a straightforward surveillance efforts in those animal populations that are less likely to be decrease in the frequency of testing, risk-based surveillance strate- affected, while increasing the efforts in those populations that are more gies should be promoted. Surveillance efforts should focus not only likely to be infected. That requires good knowledge of the risk factors in those herds more likely to be infected, but also in those herds for cattle TB. Risk-based surveillance strategies may be applied based more likely to infect other herds. on area-level or herd-level risk factors. In Spain, area-level risk factors • The sensitivity of the test used for TB detection had an important are better known. For example, a large number of movements from influence on the sensitivity of the routine testing component but also counties with high incidence (> 1%) or the presence of bullfighting on the pre-movement detection, and as a result, affected the overall cattle herds are known to increase the risk of cattle TB at the county sensitivity of the system. Therefore, improving the sensitivity level (Garcia-Sáenz et al., 2014). Some herd-level risk factors were through initiatives such as compulsory training courses for the ve- identified in previous studies, e.g. a previous infection of the herd or terinarians in charge of the testing seems a cost-effective alternative sharing of pastures (Guta et al., 2014). Those factors have already been to improve cattle TB surveillance. incorporated into the Spanish eradication program, so that in areas of • Slaughterhouse surveillance also contributed to detection of MTC- low prevalence, herds receiving animals from high incidence areas, infected herds with an average sensitivity of 9.5% and a mean time bullfighting herds, herds previously infected or those sharing pastures to detection longer than three years. Despite this relatively low ef- are tested more frequently than other herds (Anonymous, 2019v). In fectiveness, this component is important as it allows the detection of addition to the risk of infection of the herds, the consequences also need diseases other than TB, and it has proven its capacity to find MCT- to be taken into account for the prioritization of the surveillance effort infected animals that otherwise would have been missed (i.e. “an- (Cameron, 2012). Some factors that make herds more likely to infect ergic” animals) although those cases are less and less frequent in other herds (e.g. large number of movements to other herds) are also Spain. considered in the program, and therefore those herds are tested more • In Spain, 10.8% of infected herds were detected by pre-movement 10 S. Napp, et al. Preventive Veterinary Medicine 173 (2019) 104805 testing within the first year following infection, and on average Ciaravino, G., Ibarra, P., Casal, E., Lopez, S., Espluga, J., Casal, J., Napp, S., Allepuz, A., detection took over 3 years. The cost-effectiveness of pre-movement 2017. Farmer and veterinarian attitudes towards the Bovine tuberculosis eradication programme in Spain: what is going on in the Field? Front. Vet. Sci. 4, 1–28. https:// testing, at least in areas of low incidence, should be evaluated taking doi.org/10.3389/fvets.2017.00202. into account not only its contribution to the detection, but also its Ciaravino, G., García-Saenz, A., Cabras, S., Allepuz, A., Casal, J., García-Bocanegra, I., De cost, and should be compared to the cost-effectiveness of alternative Koeijer, A., Gubbins, S., Sáez, J.L., Cano-Terriza, D., Napp, S., 2018. Assessing the variability in transmission of bovine tuberculosis within Spanish cattle herds. options. Epidemics 23, 110–120. https://doi.org/10.1016/j.epidem.2018.01.003. Claridge, J., Diggle, P., McCann, C.M., Mulcahy, G., Flynn, R., McNair, J., Strain, S., Funding Welsh, M., Baylis, M., Williams, D.J., 2012. Fasciola hepatica is associated with the failure to detect bovine tuberculosis in dairy cattle. Nat. Commun. 3, 853. Cosivi, O., Grange, J.M., Daborn, C.J., Raviglione, M.C., Fujikura, T., Cousins, D., This research was funded by the Ministerio de Economía y Robinson, R.A., Huchzermeyer, H.F.A.K., de Kantor, I., Meslin, F.X., 1998. Zoonotic Competitividad (MINECO) of Spain (Ministry of Economy and tuberculosis due to Mycobacterium bovis in developing countries. Emerg. Infect. Dis. 4, 59–70. https://doi.org/10.3201/eid0401.980108. Competitiveness, EPITUBER Project, number AGL2013-49159-C2-1-R) Courtenay, O., Reilly, L.A., Sweeney, F.P., Hibberd, V., Bryan, S., Ul-Hassan, A., Newman, and was supported by “Integrated epidemiological models for risk- C., Macdonald, D.W., Delahay, R.J., Wilson, G.J., Wellington, E.M.H., 2006. Is based surveillance approaches (Epi-risk)” Era-Net ANIHWA Project. Mycobacterium bovis in the environment important for the persistence of bovine tu- The PhD of Giovanna Ciaravino was funded by the Universitat berculosis? Biol. Lett. 2, 460–462. Crawshaw, T., Daniel, R., Clifton-Hadley, R., Clark, J., Evans, H., Rolfe, S., De La Rua- Autònoma de Barcelona (UAB) of Spain (Autonomous University of Domenech, R., 2008. TB in goats caused by Mycobacterium bovis. Vet. Rec. 163 127- Barcelona, grant number D045702/B14P0024). 127. De la Rua-Domenech, R., Goodchild, A.T., Vordermeier, H.M., Hewinson, R.G., Christiansen, K.H., Clifton-Hadley, R.S., 2006. Ante mortem diagnosis of tuberculosis Acknowledgments in cattle: a review of the tuberculin tests, γ-interferon assay and other ancillary di- agnostic techniques. Res. Vet. Sci. 81, 190–210. https://doi.org/10.1016/j.rvsc.2005. The authors express their thanks to all the field veterinarians and 11.005. Domingo, M., Vidal, E., Marco, A., 2014. Pathology of bovine tuberculosis. Res. Vet. Sci. Veterinary officers involved in the Spanish Bovine Tuberculosis 97, S20–S29. https://doi.org/10.1016/j.rvsc.2014.03.017. Eradication Programme for their dedication and effort in implementing Ducheyne, E., Charlier, J., Vercruysse, J., Rinaldi, L., Biggeri, A., Demeler, J., Brandt, C., the campaign. We would also like to acknowledge the animal health de Waal, T., Selemetas, N., Höglund, J., Kaba, J., 2015. Modelling the spatial dis- tribution of Fasciola hepatica in dairy cattle in Europe. Geospat. Health 9, 261–270. and administrative staff of the Spanish Ministry for Agriculture, El Allaki, F., Harrington, N., Howden, K., 2016. Assessing the sensitivity of bovine tu- Fisheries and Food (MAPA) for their constructive assistance in the berculosis surveillance in Canada’s cattle population, 2009–2013. Prev Vet Med. 134, management of the database and for facilitating the data. Databases on 145–152. https://doi.org/10.1016/j.prevetmed.2016.10.012. EFSA AHAW Panel (EFSA Panel on Animal Health and Welfare), 2012. Scientific opinion cattle movements and routine controls are propriety of the Ministry for on the use of a gamma interferon test for the diagnosis of bovine tuberculosis. EFSA J. Agriculture, Fisheries and Food (MAPA) and subjected to con- 10 (12), 2975. https://doi.org/10.2903/j.efsa.2012.2975. fidentiality issues. EU, 2015. Animal Health - Regulatory Committee–Presentations. Evolution of the National Eradication Programme on Bovine tb 2014 in Spain. 2015. Available at:. https://ec.europa.eu/food/sites/food/files/animals/docs/reg-com_ahw_20150506_ References pres_bovine-tuberculosis_spain.pdf. EU, 2019a. Animal Health - Regulatory Committee–Presentations. Evolution of the Allepuz, A., Casal, J., Napp, S., Saez, M., Alba, A., Vilar, M., Domingo, M., González, M.A., National Eradication Programme on Bovine tb 2018 in Spain. 2019. Available at:. Duran-Ferrer, M., Vicente, J., Álvarez, J., Muñoz, M., Saez, J.L., 2011. Analysis of the https://ec.europa.eu/food/sites/food/files/animals/docs/reg-com_ahw_20190612_ spatial variation of Bovine tuberculosis disease risk in Spain (2006–2009). Prev. Vet. pres_bov-tub_esp.pdf. Med. 100, 44–52. https://doi.org/10.1016/j.prevetmed.2011.02.012. EU, 2016. Animal Health - Regulatory Committee–Presentations. Final Outcome of the Álvarez, J., Bezos, J., de la Cruz, M.L., Casal, C., Romero, B., Domínguez, L., de Juan, L., Evaluation Procedure of Eradication, Control and Surveillance Programmes Pérez, A., 2014. Bovine tuberculosis: within-herd transmission models to support and Submitted by the MSs for Union Financial Contribution for 2016 and Following direct the decision-making process. Res. Vet. Sci. 97, S61–S68. https://doi.org/10. Years. Available at:. https://ec.europa.eu/food/sites/food/files/animals/docs/reg- 1016/j.rvsc.2014.04.009. com_ahw_20160112_pres_eval-procedure_ecsp.pdf. Álvarez, J., Perez, A.M., Bezos, J., Marqués, S., Grau, A., Saez, J.L., et al., 2012a. EU, 2017. Animal Health - Regulatory Committee–Presentations. Outcome of the Evaluation of the sensitivity and specificity of bovine tuberculosis diagnostic tests in Evaluation Procedure of the 2017 EU Co-financed Veterinary Programmes. naturally infected cattle herds using a Bayesian approach. Vet. Microbiol. 155, 38–43. Available at:. https://ec.europa.eu/food/sites/food/files/animals/docs/reg-com_ https://doi.org/10.1016/j.vetmic.2011.07.034. ahw_20170117_pres_erad-2017_eur.pdf. Álvarez, J., Perez, A.M., Bezos, J., Casal, C., Romero, B., Rodriguez-Campos, S., Saez- EU, 2018. Animal Health - Regulatory Committee–Presentations. Outcome of the Llorente, J.L., Diaz, R., Carpintero, J., de Juan, L., Domínguez, L., 2012b. Eradication Evaluation Procedure of the 2019 Veterinary Programmes for Which MSs Request EU of bovine tuberculosis at a herd-level in Madrid, Spain: study of within-herd trans- Co-financing. Available at:. https://ec.europa.eu/food/sites/food/files/animals/ mission dynamics over a 12 year period. BMC Vet. Res. 8, 100. https://doi.org/10. docs/reg-com_ahw_20181123_progs-2019.pdf. 1186/1746-6148-8-100. EU, 2019b. Animal Health - Regulatory Committee–Presentations. Outcome of the Anonymous, 2017. Informe Final Técnico-Financiero Programa Nacional de la Evaluation Procedure of the 2019 Veterinary Programmes for Which MSs Request EU Tuberculosis Bovina año 2017. Available at:. Ministerio de Agricultura Pesca y Co-financing. Available at:. https://ec.europa.eu/food/sites/food/files/animals/ Alimentación (MAPA) - Dirección General de Sanidad de la Producción, Madrid, docs/reg-com_ahw_20190116_2019-vet-finance-progs_eur.pdf. Spain. https://www.mapa.gob.es/es/ganaderia/temas/sanidad-animal-higiene- EU - DG SANCO (European Comission, Directorate-Generale for Health & Consumers ganadera/informefinaltecnicotb2017_tcm30-496047.pdf. Protection), 2013. Working Document on Eradication of Bovine Tuberculosis in the Anonymous, 2019v. Programa Nacional de Erradicación de Tuberculosis Bovina pre- EU Accepted by the Bovine Tuberculosis Subgroup of the Task Force on Monitoring sentado por España para el año 2019. Available at:. Ministerio de Agricultura Pesca Animal Disease Eradication. Brussels, SANCO/10067/2013. Available at:. https:// y Alimentación (MAPA) - Dirección General de Sanidad de la Producción, Madrid, ec.europa.eu/food/sites/food/files/animals/docs/diseases_erad_tb_workingdoc2006_ Spain. https://www.mapa.gob.es/es/ganaderia/temas/sanidad-animal-higiene- en.pdf. ganadera/programatb2019verdefinitiva_tcm30-500265.pdf. EU - DG SANTE (European Commission, Directorate-General for Health & Food Safety), Beesley, N.J., Caminade, C., Charlier, J., Flynn, R.J., Hodgkinson, J.E., Martinez‐Moreno, 2018. Commission Delegated Regulation 2018/1629/ EU of 25 July 2018 Amending A., Martinez‐Valladares, M., Perez, J., Rinaldi, L., Williams, D.J.L., 2018. Fasciola and the List of Diseases Set Out in Annex II to Regulation (EU) 2016/429 of the European fasciolosis in ruminants in Europe: identifying research needs. Transbound. Emerg. Parliament and of the Council on Transmissible Animal Diseases and Amending and Dis 65, 199–216. Repealing Certain Acts in the Area of Animal Health (‘Animal Health Law’). Brussels. Brooks-Pollock, E., Roberts, G.O., Keeling, M.J., 2014. A dynamic model of bovine tu- Available at:. http://data.europa.eu/eli/reg_del/2018/1629/oj. berculosis spread and control in Great Britain. Nature 511, 228–231. https://doi.org/ EU - DG SANTE (European Commission, Directorate-General for Health & Food Safety), 10.1038/nature13529. 2016. Final Report of an Audit Carried Out in Spain from 01 February 2016 to 12 Calvo-Artavia, F., Alban, L., Nielsen, L., 2013. Evaluation of surveillance for doc- February 2016 in Order to Evaluate the Effectiveness of, and the Progress Made by umentation of freedom from bovine tuberculosis. Agriculture 3, 310–326. https:// the Programme Co-financed by the European Union to Eradicate Bovine Tuberculosis. doi.org/10.3390/agriculture3030310. Available at:. . http://ec.europa.eu/food/audits-analysis/audit_reports/details.cfm? Cameron, A.R., 2012. The consequences of risk-based surveillance: developing output- rep_id=3711. based standards for surveillance to demonstrate freedom from disease. Prev. Vet. European Food Safety Authority and European Centre for Disease Prevention and Control Med. 105, 280–286. (EFSA and ECDC), 2018. The European Union summary report on trends and sources Cameron, A., Njeumi, F., Chibeu, D., Martin, T., 2014. Risk-Based Disease Surveillance. of zoonoses, zoonotic agents and food‐borne outbreaks in 2017. EFSA J. 16 (12). Available at:. FAO. http://www.fao.org/3/a-i4205e.pdf. https://doi.org/10.2903/j.efsa.2018.5500. p.e05500. Available at:. Fischer, E.A., van Roermund, H.J.W., Hemerik, L., van Asseldonk, M.A.P.M., de Jong, 11 S. Napp, et al. Preventive Veterinary Medicine 173 (2019) 104805 M.C.M., 2005. Evaluation of surveillance strategies for bovine tuberculosis Meskell, P., Devitt, C., More, S.J., 2013. Challenges to quality testing for bovine tu- (Mycobacterium bovis) using an individual based epidemiological model. Prev. Vet. berculosis in Ireland; Perspectives from major stakeholders. Vet. Rec. 173, 94. Med. 67, 283–301. https://doi.org/10.1016/j.prevetmed.2004.12.002. https://doi.org/10.1136/vr.101676. Frankena, K., White, P.W., O’Keeffe, J., Costello, E., Martin, S.W., Van Grevenhof, I., Müller, B., Dürr, S., Alonso, S., Hattendorf, J., Laisse, C.J.M., Parsons, S.D.C., van Helden, More, S.J., 2007. Quantification of the relative efficiency of factory surveillance in P.D., Zinsstag, J., 2013. Zoonotic Mycobacterium bovis –induced Tuberculosis in hu- the disclosure of tuberculosis lesions in attested Irish cattle. Vet. Rec. 161, 679–684. mans. Emerg. Infect. Dis. 19, 899–908. https://doi.org/10.3201/eid1906.120543. https://doi.org/10.1136/vr.161.20.679. OIE (World Organisation for Animal Health), 2018. Infection with Mycobacterium tu- Garcia-Sáenz, A., Napp, S., Lopez, S., Casal, J., Allepuz, A., 2015. Estimation of the in- berculosis complex, 27th ed. OIE, Terrestrial Animal Health Code, vol. 2 session 8, dividual slaughterhouse surveillance sensitivity for bovine tuberculosis in Catalonia chapter 8.11. Paris. Available at: http://www.oie.int/en/standard-setting/ (North-Eastern Spain). Prev. Vet. Med. 121, 332–337. https://doi.org/10.1016/j. terrestrial-code/access-online/. prevetmed.2015.08.008. Olea-Popelka, F., Muwonge, A., Perera, A., Dean, A.S., Mumford, E., Erlacher-Vindel, E., Garcia-Sáenz, A., Saez, M., Napp, S., Casal, J., Saez, J.L., Acevedo, P., Guta, S., Allepuz, Forcella, S., Silk, B.J., Ditiu, L., El Idrissi, A., Raviglione, M., Cosivi, O., LoBue, P., A., 2014. Spatio-temporal variability of bovine tuberculosis eradication in Spain Fujiwara, P.I., 2017. Zoonotic tuberculosis in human beings caused by Mycobacterium (2006–2011). Spat. Spatiotemporal. Epidemiol. 10, 1–10. https://doi.org/10.1016/j. bovis—a call for action. Lancet Infect. Dis. 17, e21–e25. https://doi.org/10.1016/ sste.2014.06.002. S1473-3099(16)30139-6. Gelman, A., Rubin, D.B., 1992. Inference from iterative simulation using multiple se- Palacios, J.J., Navarro, Y., Romero, B., Penedo, A., González, Á.M., Hernández, M.D.P., quences. Stat. Sci. 7 (4), 457–472. https://doi.org/10.1214/ss/1177011136. Fernández-Verdugo, A., Copano, F., Torreblanca, A., Bouza, E., Domínguez, L., 2016. Guta, S., Casal, J., Garcia-Saenz, A., Saez, J.L., Pacios, A., Garcia, P., Napp, S., Allepuz, A., Molecular and epidemiological population-based integrative analysis of human and 2014. Risk factors for bovine tuberculosis persistence in beef herds of Southern and animal Mycobacterium bovis infections in a low-prevalence setting. Vet. Microbiol. Central Spain. Prev. Vet. Med. 115, 173–180. 195, 30–36. https://doi.org/10.1016/j.vetmic.2016.08.019. Hadorn, D.C., Stärk, K.D.C., 2008. Evaluation and optimization of surveillance systems for R Core Team, core Team, R, Team, R.D.C, 2015. R: a language and environment for rare and emerging infectious diseases. Vet. Res. 39, 57. https://doi.org/10.1051/ statistical computing. R A Lang. Environ. Stat. Comput Available from: https://www. vetres:2008033. r-project.org/. De Mendoza, J.H., Parra, A., Tato, A., Alonso, J.M., Rey, J.M., Pena, J., García-Sánchez, Reviriego Gordejo, F.J., Vermeersch, J.P., 2006. Towards eradication of bovine tu- A., Larrasa, J., Teixidó, J., Manzano, G., Cerrato, R., 2006. Bovine tuberculosis in berculosis in the European Union. Vet. Microbiol. 112, 101–109. https://doi.org/10. wild boar (Sus scrofa), red deer (Cervus elaphus) and cattle (Bos taurus) in a 1016/j.vetmic.2005.11.034. Mediterranean ecosystem (1992–2004). Prev. Vet. Med. 17 (74), 239–247. https:// Rossi, G., De Leo, G.A., Pongolini, S., Natalini, S., Vincenzi, S., Bolzoni, L., 2015. doi.org/10.1016/j.prevetmed.2005.10.005. Epidemiological modelling for the assessment of bovine tuberculosis surveillance in Humblet, M.-F., Boschiroli, M.L., Saegerman, C., 2009. Classification of worldwide bovine the dairy farm network in Emilia-Romagna (Italy). Epidemics 11, 62–70. https://doi. tuberculosis risk factors in cattle: a stratified approach. Vet. Res. 40, 50. https://doi. org/10.1016/j.epidem.2015.02.007. org/10.1051/vetres/2009033. Schöning, J.M., Cerny, N., Prohaska, S., Wittenbrink, M.M., Smith, N.H., Bloemberg, G., Humblet, M.-F., Moyen, J.-L., Bardoux, P., Boschiroli, M.L., Saegerman, C., 2011a. The Pewsner, M., Schiller, I., Origgi, F.C., Ryser-Degiorgis, M.P., 2013. Surveillance of importance of awareness for veterinarians involved in cattle tuberculosis skin testing. bovine tuberculosis and risk estimation of a future reservoir formation in wildlife in Transbound. Emerg. Dis. 58, 531–536. https://doi.org/10.1111/j.1865-1682.2011. Switzerland and Liechtenstein. PLoS One 8, e54253. https://doi.org/10.1371/ 01228.x. journal.pone.0054253. Humblet, M.-F., Walravens, K., Salandre, O., Boschiroli, M.L., Gilbert, M., Berkvens, D., Thoen, C.O., Lobue, P.A., de Kantor, I.N., 2010. Why has zoonotic tuberculosis not re- Fauville-Dufaux, M., Godfroid, J., Dufey, J., Raskin, A., Vanholme, L., Saegerman, C., ceived much attention? Int. J. Tuberc. Lung Dis. 14, 1073–1074. 2011b. Monitoring of the intra-dermal tuberculosis skin test performed by Belgian Van Asseldonk, M.A.P.M., van Roermund, H.J.W., Fischer, E.A.J., de Jong, M.C.M., field practitioners. Res. Vet. Sci. 91, 199–207. https://doi.org/10.1016/j.rvsc.2010. Huirne, R.B.M., 2005. Stochastic efficiency analysis of bovine tuberculosis-surveil- 12.004. lance programs in the Netherlands. Prev. Vet. Med. 69, 39–52. https://doi.org/10. Keeling, M.J., Rohani, P., 2007. Modeling Infectious Diseases in Humans and Animals. 1016/j.prevetmed.2005.01.012. Princeton university presshttps://doi.org/10.1097/01.ede.0000254692.80550.60. Welby, S., Govaerts, M., Vanholme, L., Hooyberghs, J., Mennens, K., Maes, L., Van Der 408 pp. Stede, Y., 2012. Bovine tuberculosis surveillance alternatives in Belgium. Prev. Vet. Lombardi, G., Botti, I., Pacciarini, M.L., Boniotti, M.B., Roncarati, G., Dal Monte, P., 2017. Med. 106, 152–161. https://doi.org/10.1016/j.prevetmed.2012.02.010. Five-year surveillance of human tuberculosis caused by Mycobacterium bovis in WHO (World Health Organization), 2019. Zoonotic TB. In: Tuberculosis (TB). Available Bologna, Italy: an underestimated problem. Epidemiol. Infect. 145, 3035–3039. from:. WHO. https://www.who.int/tb/areas-of-work/zoonotic-tb/en/. https://doi.org/10.1017/S0950268817001996. 12