This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution license (http://creativecommons.org/licenses/by/3.0/).
The objective was to study how surveillance for bovine tuberculosis (bTB) could be made more resourceeffective in a bTB free country. A stochastic scenario tree model was developed to: (1) evaluate the sensitivity (CSe) of four surveillance system components (SSC) (
Documenting freedom from disease is an essential requirement for international trade of live animals and animal products. In accordance with the Agreement on the Application of Sanitary and Phytosanitary Measures (the SPS Agreement) of the World Trade Organization (WTO), sciencebased evidence is needed to determine the appropriate country sanitary level for international trade [
When meat inspection was founded more than 100 years ago, bovine tuberculosis (bTB) was endemic in most countries. Since then, many countries have managed to eradicate this infection. One example is Denmark, which has been declared free from bTB since 1980 by the European Union (EU) (Decision 2004/320/EC). The requirements for a country to be given the bTB free status according to the OIE and EU are that bTB has been found in less than 0.2% and 0.1% of the herds, respectively [
The Danish surveillance programme for bTB consists of four independent components (
The overall objective was therefore to evaluate the surveillance system for a country that has been declared free from bTB for many years to elucidate how and where surveillance could be made more effective, using Denmark as a study case. The specific aims of this study were:
to evaluate the sensitivity of each of the surveillance system components given that bTB would enter one of these components, but not necessarily spread between these due to limited contact and risk of transmission between the four components;
to estimate the probability of freedom from bTB in Denmark over time, and
to evaluate how future alternative programmes based on visual meat inspection and riskbased meat inspection would affect the probabilities of freedom from bTB compared to the current meat inspection system, at the very low animallevel design prevalence 0.0002% and a low probability of introduction of 1% and 0.1%.
The results are relevant for livestock industries and veterinary authorities in countries aiming at improving and modernising the meat inspection procedures, while lowering the resources spent on surveillance programme for bTB. Furthermore, the results can support international organisations such as EFSA and OIE in the decisionmaking processes concerning modernisation of meat inspection, and hereby ensure free trade on safe conditions.
The mean surveillance system component (SSC) sensitivity (CSe) of the individual SSCs in the current Danish surveillance programme for bTB and the alternative scenarios, together with their 95% credibility interval, are shown in
With a low probability of introduction of 1%, the pig SCC again provided the highest posterior mean probability of freedom at 98.5% (CI: 98.0%–98.8%), compared to 96.9% (CI: 96.4%–97.2%), 97.6% (CI: 96.7%–98%) and 87.4% (CI: 85.3%–89.5%) for the domestic cattle, deer and export cattle SSCs, respectively, by year 2012. The posterior mean probability of freedom never reached values above 98.5%.
Results of simulations with 10,000 iterations for the sensitivities (CSe), with design prevalence and units processed for each of four surveillance system components (SSC) (
Surveillance programme/SSC  Design prevalence  Units processed  CSe (95% CI) 

^{a} Average number of cattle slaughtered between January 2004 to June 2012 in Denmark; 

Current surveillance system (Invasive in all slaughtered animals)  
Domestic cattle  0.000002  498,364^{a}  0.32 (0.28–0.36) 
Pig  0.000002  20,022,790  0.68 (0.52–0.80) 
Deer^{b}  0.000125  8,000  0.43 (0.31–0.53) 
Export cattle^{b}  0.00038  2,634  0.07 (0.05–0.09) 
Alternative scenario 1 (100% visual inspection cattle)  
Domestic cattle  0.000002  498,364  0.18 (0.15–0.20) 
Alternative scenario 2 (100% visual inspection pigs)  
Pig  0.000002  20,022,790  0.43 (0.30–0.54) 
The results of the scenario analyses showed that if the proportion of granulomas detected in the alternative surveillance system, based on visual inspection only, was set to 40% or 60% rather than the 50% used in the original model, no noteworthy changes were observed in the results.
If in the pig SCC the population of adults (2% of all slaughtered pigs) was inspected using the current meat inspection procedures and the rest of the pig population was only visually inspected, no noteworthy changes were observed in the results. Nor did it change the results much if the submission of detected lesions for bacteriological culture was set to 100% instead of 10%, as in the current scenario.
Temporal discounting, estimating the probability of freedom from bovine tuberculosis in Denmark, for the first study period (1995–2012) under the current surveillance system for the four surveillance system components and for the second study period (2013–2037) under the alternative surveillance programmes: Scenario 1 where all the domestic cattle would be subjected to visual meat inspection and Scenario 2 where all pigs would be subjected to visual meat inspection. Export cattle were tested using intradermal tuberculin test (ID). The top graph shows the results from the scenario with 0.1% and the bottom graph the scenario with 1% probability of introduction.
If the design prevalence was changed to 0.01% corresponding to the EU requirement of a herdlevel design prevalence of 0.1% multiplied with an assumed animallevel prevalence within infected herds of 10%, the posterior mean probability of freedom would reach a maximum level of 99.9% at the end of the first year of surveillance and would be maintained that high throughout the two study periods.
If start probabilities of freedom in 1995 were set to 0.5 (corresponding to an uninformed prior) instead of 0.9, the posterior mean probability of freedom from bTB would reach values above 99% after the end of the fifth, tenth and thirteenth year of surveillance for the pig, deer and domestic cattle SSC. If changed to 0.7 instead of 0.9, the posterior mean probability of freedom from bTB would reach values above 99% after the end of the fourth, eighth and tenth year of surveillance for the pig, deer and domestic cattle SSC.
The most influential input parameters identified in the tornado plots for the current surveillance system and posterior probability of freedom outputs were the proportion of granulomas submitted for laboratory investigation in the pig SSC and the probability of granulomas present in young animals from the domestic cattle and pig SSC. In the case of the alternative surveillance system CSes and posterior probabilities of freedom outputs, the most influential input parameters were the proportion of granulomas present in the adult animals of the domestic cattle and pig SSC, and the proportion of granulomas submitted from the domestic cattle SSC.
The present study has shown that a country officially recognized as being free from bTB for many years can retain a high confidence in being free, even if meat inspection procedures are changed to visually based procedures, but only if the probability of introduction is kept negligible. This underpins the importance of effective import testing and quarantine procedures when import of cattle into bTB free countries cannot be avoided. The prediction of freedom from bTB in Denmark was made using the SSCs independently, because of the lack of contact and hence, low risk of transmission of bTB between cattle, pigs and deer representing the different surveillance components. We believe that the introduction of bTB into one of the components is unlikely to lead to a spread to the other components; therefore the results are presented for each component separately. The benefit of not palpating and making incisions is that the inspector can have more time to focus on other food safety hazards (
The presented results indicate that the current surveillance programme  requiring palpation and/or incisions of relevant lymph nodes of cattle, pigs and deer at meat inspection  has the highest CSe of detecting at least one slaughter animal with bTB at the design prevalence used in this study (
Assuming a probability of 0.9 that Denmark was free of bTB at the beginning of 1995 and an annual probability of introduction of disease into the country of 0.1% and 1%, the posterior mean probability of freedom increased rapidly, except in export cattle under the 1% scenario. The sensitivity analyses showed that the posterior mean probability of freedom would be affected, if start probabilities of freedom in 1995 of 0.5 (corresponding to an uninformed prior) or 0.7 were assumed instead of 0.9. The posterior mean probability of freedom from bTB would reach values above 99% much later in time, which underpins the importance of striving to keep the bTBfree status continuously. The results indicated that it took around five years to obtain the highest obtainable confidence in freedom after bTB had been eradicated with the level of surveillance used in Denmark.
The probability of introduction of bTB into Denmark was set to 1% due to the fact that Denmark mainly imports a low number of animals and mostly from bTB free countries. Testing is performed before cattle are moved in from countries that are not officially free of bTB according to OIE criteria. The scenario with negligible probability of introduction (0.1%) was evaluated to investigate the effect of improving import biosecurity procedures such as import testing and quarantine procedures on the confidence of freedom over time. The number of import cattle was reduced from 1235 in 2009 to 170 in 2011, and a large proportion of the imported animals came from the bTB free countries Sweden and Germany. Therefore, the assumption about the probability of introduction of 1% is believed to be a valid reasonable estimate for current Danish practices, but lowering this probability through improved testing and quarantine practices would be beneficial for the level of confidence in freedom.
For the second study period, assuming a change in the current meat inspection procedures to visual inspection with a negligible probability of introduction of 0.1%, the posterior mean probability of freedom from bTB decreased slightly for the domestic cattle alternative scenario (Scenario 1) by the end of 2013 from 99.7% to 99.6%. Furthermore, the pig alternative scenario (Scenario 2) had the highest posterior mean probability of freedom by the end of year 2037 at 99.8% (
Overall, the level of confidence in freedom from bTB remained close to the same regardless of how large a proportion of the cattle or pigs would be subjected to visual inspection,
A posterior mean probability of freedom reaching levels above 96%, under the current and alternative surveillance programme scenarios, suggests that there is a low probability of undetected bTB cases in the country. However, in the Danish case the probability of freedom is calculated for a country where there have been no bTB cases in cattle since 1995 (more than 18 years), and therefore, the design prevalence was set much lower than the international standards. This provides a higher confidence in the level of freedom, even with a change in the meat inspection procedures, compared to countries that still have a few bTB cases in the country.
Bovine tuberculosis is a notifiable disease [
The main routes for transmission of bTB to humans are the consumption of unpasteurized milk and milk products, and direct contact through aerosols or skin cuts [
The stochastic scenario tree methodology described by Martin
The four components of the Danish surveillance programme for bTB were assessed separately to evaluate each their sensitivities, because of limited contact and hence low risk of transmission of bTB between the different surveillance components. The sensitivity of a SSC specifies the probability that at least one bTB test positive animal will be detected given that bTB has been introduced into the SSC and spread to the level of a predefined design prevalence (see below). The four SSCs of the current Danish surveillance programme for bTB are slaughtered domestic cattle, farmed deer and pigs as well as testing of adult cattle for export. Hence, there is no overlap between these components. All slaughtered domestic cattle, farmed deer and pigs are examined at meat inspection for clinical signs or lesions indicative of bTB, and suspect findings during meat inspection are tested by bacterial culture for confirmation. Adult cattle for export are tested with an intradermal tuberculin test (ID). If there is a suspect case of bTB, the herd is put under veterinary official supervision and all or a large proportion of the animals in the herd are tested with ID [
Due to very few imports of livestock into Denmark and negligible risk of infection from wildlife (
Four scenario trees were constructed, one for each of the SSCs in the Danish bTB surveillance programme. The trees consisted of nodes dividing the population into risk groups. Within each of the risk groups the units had the same probability of detection, given the presence of disease (
Conceptual scenario tree developed to evaluate the performance of the four surveillance system components (SSCs) for detection of bovine tuberculosis (bTB) at meat inspection in Denmark.
Register data were obtained from the Danish Cattle Database (DCD) for the period from January 2004 to June 2012 of all recorded cattle slaughtered in Denmark. The average number of cattle slaughtered per year during this period was calculated (
The first node in the tree was a risk category node at animal level (
Input variables used to estimate the probability of freedom from bovine tuberculosis in Denmark represented either as fixed values or distributions, and sources of information.
Input variable  Input value and distribution  Source  

Design prevalence ( 
0.000002  Adjusted from [ 

Design prevalence ( 
0.000125  Calculated: 1 animal out of 8000  
Design prevalence ( 
0.00038  Calculated: 1 animal out of 2634  
Proportion of young and adult animals


Domestic cattle  Young  0.3  Danish Cattle Database (DCD) 
Adult  0.7  
Pig  Young  0.98  Danish Agriculture & Food Council 
Adult  0.02  
Deer  Young  0.9  Danish Agriculture & Food Council 
Adult  0.1  
Export cattle  Young  0.77  Danish Knowledge Centre of Agriculture, Cattle 
Adult  0.23  
Relative risk of infection between the young and adult groups

RiskPert (1.5, 2, 3)  [ 

Proportion of lesions present ( 

Adult animals  RiskBeta (84, 58)  [ 

Young animals  50% of adults  
Proportion of lesions detected ( 

Current programme  RiskPert (0.9, 0.95, 0.99)  [ 

Alternative programme  50% of the current  
Proportion of lesions submitted ( 

Pig SSC  RiskPert (0.005, 0.1, 0.15)  Expert opinion  
Other three SSC  RiskPert (0.7, 0.8, 0.9)  [ 

Meat inspection Se ( 
RiskPert (0.29, 0.6, 0.86)  [ 

Bacteriological culture test Se ( 
RiskPert (0.92, 0.95, 0.98)  [ 

Intradermal tuberculin test Se ( 
RiskPert (0.53, 0.7, 0.96)  [ 

Probability of introduction ( 
0.001 and 0.01  Evaluation of import practices 
The next node in the scenario tree was the infection node, where the design prevalence (
Where
The next nodes correspond to the probability of detecting bTB in each group. For the domestic cattle and pig SSC, the meat inspection was divided into three nodes:
The proportion of lesions present (
The proportion of granulomas detected at meat inspection, modeled using a pert distribution based on [
The proportion of granulomas submitted for confirmation by bacterial culture. Based on expert opinion, the proportions were set higher for domestic cattle than for pigs (
For the export cattle SSC the detection node was the probability of testing positive to the ID test, which was also modeled using a pert distribution, based on a field trial literature review by [
To estimate the overall CSe of each SSC, the component unit sensitivity (CSeU) was first calculated by the sum of the estimated probabilities with a positive outcome at the end of each limb (Equation 3). These probabilities were the result of the product of all branch probabilities and proportions. For the tree of the domestic cattle SSC (
Where
Once the CSeU was calculated, then the probability that one or more positive units would be detected given that the SSC was infected at the design prevalence (CSe) was estimated for each of the four SSC (Equation 4).
Where
To calculate the posterior probability of freedom (
The first posterior probability of infection (
The models were set up in @Risk 6 (Palisade Corporation^{®}, Middlesex, UK) and run with 10,000 iterations. The input parameters were described by use of distributions to account for the uncertainty in the parameter estimates (
A new period was simulated (
To estimate the probability of freedom for the alternative surveillance programme, the
In conclusion, for a country considered free from bTB, lowering the intensity of surveillance e.g., by introducing visual meat inspection of cattle and pigs, will not reduce the confidence in freedom from bTB markedly providing the probability of introduction of the infection can be kept low. However, with such a slow spreading infection with a long incubation time, new introduction might take a long time to be detected. Hence, both continued surveillance and rigid import testing and quarantine practices are recommended to acquire a negligible risk of introduction of bTB.
The authors would like to thank the following people for providing data and good discussions about the methods and results: Camilla Brasch Andersen, Danish Veterinary and Food Administration. Flemming ThuneStephensen, Danish Agriculture and Food Council. Preben Willeberg, UCDavis, California, USA/Former CVO from Denmark.