EVALUATION OF ADAPTIVE TEST STRATEGIES FOR CONTROL AND ERADICATION OF PARATUBERCULOSIS WITHIN DAIRY CATTLE HERDS

Abstract

Paratuberculosis is a chronic infection caused by Mycobacterium avium ssp. paratuberculosis (MAP). A long subclinical phase challenges the test-strategies and interpretation of diagnostic information. A bio-economic herd simulation model was used in this study to adapt the sampling interval in response to the estimated true prevalence in the herd, so the farmer could reduce the prevalence to a given tolerance level focusing on profit maximisation. When the prevalence was below the tolerance level, the sampling interval was longer, and when the prevalence was above the tolerance level, the sampling interval was shorter. The results showed that the adaptive test strategy could be used to reduce the prevalence in the simulated herd, which depended on both the sampling interval and the tolerance level used. Moreover, the simulations showed a potential for saving costs for testing when combining the adaptive strategy and a reduced risk-based test strategy, while still preserving disease control.

Full text

EVALUATION OF ADAPTIVE TEST STRATEGIES FOR CONTROL AND
ERADICATION OF PARATUBERCULOSIS WITHIN DAIRY CATTLE HERDS
C. KIRKEBY*, K. GRÆSBØLL, S.S. NIELSEN, L.E. CHRISTIANSEN, N. TOFT, T.
HALASA
SUMMARY
Paratuberculosis is a chronic infection caused by Mycobacterium
avium ssp. paratuberculosis (MAP). A long subclinical phase challenges the test-strategies
and interpretation of diagnostic information. A bio-economic herd simulation model was used
in this study to adapt the sampling interval in response to the estimated true prevalence in the
herd, so the farmer could reduce the prevalence to a given tolerance level focusing on profit
maximisation. When the prevalence was below the tolerance level, the sampling interval was
longer, and when the prevalence was above the tolerance level, the sampling interval was
shorter. The results showed that the adaptive test strategy could be used to reduce the
prevalence in the simulated herd, which depended on both the sampling interval and the
tolerance level used. Moreover, the simulations showed a potential for saving costs for testing
when combining the adaptive strategy and a reduced risk-based test strategy, while still
preserving disease control.
INTRODUCTION
Paratuberculosis is a chronic infection caused by Mycobacterium avium ssp.
paratuberculosis (MAP). It is widely spread in dairy cattle in Europe (Nielsen & Toft, 2009).
Transmission is mainly via MAP shed in the environment, and infected animals are often
infected as calves, because the susceptibility decreases with the age of the animal (Sweeney,
2011). Paratuberculosis has a long subclinical phase during which infected animals can shed
MAP (Whitlock et al. 2000). MAP specific antibody ELISA for milk and serum samples,
bacteriological culture and PCR on faecal samples exist for detection of infected and
infectious animals (Nielsen, 2014). None of these tests are perfect, particularly for
discrimination between infected and non-infected, and infected and infectious stages.
However, test-positivity is highly associated with the probability of being infectious, and
consequently the timing of testing is essential when constructing test-strategies (Nielsen &
Toft, 2006). Subclinical animals can have a lower milk yield and body weight, thus reducing
the farmer’s income from milk and slaughter values (Ott et al., 1999). In the clinical phase
the animals have severe diarrhoea and will eventually die.
One way forward in a control strategy is to use a strict test scheme to continuously screen
cows for MAP infection (Whittington & Sergeant 2001, Sergeant et al. 2008). This can help
the farmer to identify infectious or affected cows for culling before they negatively impact
* Corresponding author: Carsten Kirkeby, National Veterinary Institute, Technical University
of Denmark, Bülowsvej 27, 1870 Frederiksberg C. Email: ckir@vet.dtu.dk
1

the herd. In Denmark, the ID Screen ELISA test (IDvet, Grabels, France) is used on milk
samples in the voluntary national control programme for paratuberculosis. Lactating cows are
sampled quarterly and cows are classified into three risk categories: high, medium and low
risk cows based on the repeated ELISA results (Nielsen et al., 2007). The test costs are
covered by the farmers, and some farmers may be reluctant to enter the paratuberculosis
programme or stay in the programme when the prevalence has been reduced to a level where
the infection no longer appears to affect the production performance. Furthermore, farmers
that have always had a low prevalence may not want to spend money on quarterly testing
when positive cows are rarely found. The farmer may think that the tests (and associated
costs) are to no avail, even if the continuous testing helps to keep the prevalence low.
This study evaluated a new test strategy that can adapt to the purpose of retaining control
of paratuberculosis in a herd without necessarily eradicating the infection (Kirkeby et al.
2016b). The strategy includes automatic adjustment of the sampling interval using a long
sampling interval when the within-herd prevalence is low, and a shorter sampling interval
when the prevalence is above a certain threshold. For the first time, a reduced testing
approach was also simulated, where only the test-positive cows in the herd follow the full test
scheme and cows with so far only negative tests are tested once per lactation. This reduced
testing approach is currently used in the Danish control program for paratuberculosis. The
first objective of the present study was to evaluate the epidemiological and economic
consequences of the adaptive test strategy. All scenarios were run in a herd with standard
hygiene (and mean prevalence of paratuberculosis, following Kirkeby et al. 2016a). The
second objective was to evaluate the efficiency of the reduced test scheme for eradication of
MAP in the herd.
MATERIALS AND METHODS
Simulations were performed using the iCull model, a mechanistic, stochastic, dynamic
bio-economic model that simulates individual cows in a standard Danish dairy cattle herd
with daily time steps (Kirkeby et al. 2016a). The model simulates a realistic dairy farm with
lactation curves, somatic cell counts and more details based on data (Græsbøll et al. 2016).
All simulations in this study were of a herd with 200 dairy cows over 10 years. A 3 year
burn-in period and 500 repetitions were used in all simulations. As the infection progresses in
the individual animal, the milk ELISA value increases. The milk yield of the individual cow
is therefore adjusted according to the ELISA value (Græsbøll et al. 2014, Kirkeby et al.
2016a). All simulated scenarios were done in a herd with an initial true prevalence of 5.6%
(corresponding to a herd with a median prevalence and thus standard level of hygiene,
Kirkeby et al. 2016a). All scenarios in this study were with closed herds (i.e. without
introduction of livestock), which is common in 50% of all dairy herds in Denmark. This
means that the probability of eradication is higher than when an open herd is simulated.
All simulated scenarios used the test-and-cull strategy, which may be sufficient for
reducing the prevalence within a herd (Lu et al. 2008, Kirkeby et al. 2016a). For comparison,
we simulated baseline scenarios (for each of the two herd types), reflecting the current
strategy for Danish herds with a sampling interval fixed to three months. In the adaptive
strategy, a prevalence cutoff (PC) is set to reflect the prevalence that the farmer will accept.
Lowering the prevalence below the PC can be costly for example due to increased test costs
or costs associated with culling of false-positive reactors. Therefore, the farmer may wish to
keep the prevalence low, without necessarily aiming for eradication. In the adaptive strategy,
the sampling interval is varied according to the estimated true prevalence in the herd. The
2

true prevalence is continuously estimated with the Rogan-Gladen approach (Rogan & Gladen
1978). If the estimated true prevalence drops below the chosen PC, the farmer uses a long
sampling interval (LSI). Then, if the prevalence exceeds the PC, the farmer switches to a
short sampling interval (SSI). In this study, two different PC levels of 1% and 5% were
simulated (Table 1). Summary information of the number of simulated days before switching
to LSI (when the estimated prevalence was lower than the chosen PC) was saved during the
simulations. Likewise, it was recorded if the farmer switched back to SSI (meaning that the
prevalence went above the chosen PC again). The probability of switching to LSI, as well as
the probability of switching back to SSI, was then calculated from the 500 repetitions of each
scenario. Lastly, the probability of eradication of MAP was also calculated from these data.
Table 1. Parameters used in this study.
Parameters
Setting
Unit
Long sampling
interval (LSI)
365/730
a
Days
Short sampling
interval (SSI)
31/91/182 Days
Initial
prevalence
5.6
%
Prevalence
cutoff (PC)
1/5 %
Reduced testing
On/Off
a) Reduced testing was not combined with LSI = 730 because
the reduced testing approach uses a one-year cycle.
Besides the adaptive test scheme, it was also simulated that the farmer could choose to use
a reduced testing approach. This reflects a current alternative test strategy in Denmark, where
farmers can choose to test cows just once per lactation, but cows that receive a positive test
result (which may be true positive or false positive), still have to be tested every third month.
In this way the farmer saves costs for testing because only positive cows will follow the full
test scheme while the rest will follow the reduced test scheme. Cows in the reduced test
scheme are sampled at least 180 days prior to calving. The reduced testing approach was
combined with the adaptive test scheme to evaluate the efficiency of both (Table 1).
The iCull model is a bio-economic model, and thus is able to retrieve information on the
economy in each scenario, of particular interest the costs of testing. The economic
calculations in the model are further described in Kirkeby et al. (2016a).
RESULTS
Generally, the higher the PC, the higher the prevalence was in the end of the simulations
(Fig. 1). Likewise, the longer LSI and SSI, the longer time it took to reduce the prevalence.
When the PC was set to 1% in the scenario with SSI/LSI = 31/365, the median number of
simulated days before switching to LSI was 849. This was the fastest among all the scenarios,
excluding the reduced testing scenarios. The probability of reaching LSI within the ten
simulated years was generally high, between 64% and 100% in all simulations (Table 2). The
probability of switching back to SSI (after LSI was reached) was between 48% and 94% in
the scenario with PC = 1% and between 46% and 81% when PC = 5% (Table 2).
3

Fig. 1. Boxplot of the true prevalence after ten simulated years for each scenario. In these
scenarios the farmer chose a prevalence cutoff (PC) as the maximum MAP prevalence he will
tolerate. When the prevalence exceeds this value, a short sampling interval is used. When the
prevalence is lower than this value, a long sampling interval is used. X-axis labels show the
short sampling interval and long sampling interval and if the reduced testing strategy is used.
SSI = Short sampling interval, LSI = Long sampling interval. So in the second scenario the
SSI is 31 days, the LSI is 365 days and reduced testing was not used. Solid bar is the median
value, box represents 25th–75th percentiles and whiskers show range of values.
Table 2. Results of the simulations. Parameters used in this study. SSI = Short sampling
interval, LSI = Long sampling interval, Reduced = Reduced testing strategy. The
4

probabilities of eradication, switch to LSI and back to SSI again were calculated from 500
simulations.
Scenario Days before
LSI: Median
(5%:95%)
Prob. of
eradication Prob. of
LSI Prob. of
SSI again
Baseline
0.87
PC = 1%:
SSI31-LSI365
849 (212:1824)
0.92
1
0.94
SSI91-LSI365
1640 (361:3259)
0.78
0.89
0.82
SSI182-LSI365
1911 (452:3364)
0.43
0.64
0.8
SSI31-LSI730
863 (209:1886)
0.79
1
0.72
SSI91-LSI730
1726 (361:3182)
0.69
0.9
0.5
SSI182-LSI730
1927 (452:3364)
0.39
0.71
0.48
Reduced-SSI31-LSI365
246 (119:398)
0.65
1
1
Reduced-SSI91-LSI365
799 (179:1731)
0.61
1
0.98
Reduced-SSI182-LSI365
1585 (270:3364)
0.41
0.81
0.81
PC = 5%:
SSI31-LSI365
257 (88:553)
0.4
1
0.79
SSI91-LSI365
520 (88:1362)
0.41
1
0.75
SSI182-LSI365
756 (88:2090)
0.34
1
0.7
SSI31-LSI730
248 (88:491)
0.24
1
0.55
SSI91-LSI730
516 (88:1271)
0.24
1
0.46
SSI182-LSI730
717 (88:1908)
0.17
0.99
0.5
Reduced-SSI31-LSI365
153 (88:243)
0.31
1
0.79
Reduced-SSI91-LSI365
270 (88:634)
0.32
1
0.81
Reduced-SSI182-LSI365
475 (88:1726)
0.29
0.99
0.79
Use of the reduced testing strategy decreased the probability of eradication when
comparing scenarios with the same PC, SSI and LSI. The reduced testing approach did not
have any noticeable impact on the probability of switching to LSI. Furthermore, the reduced
testing approach generally increased the probability of switching back to SSI after LSI was
reached.
In the baseline scenario, approximately 2750 EUR were spent yearly on ELISA tests. This
cost was reduced in all scenarios with PC = 1%, except those with SSI = 31. When PC = 5%,
all scenarios reduced the costs for testing compared to the baseline (Fig. 1). All the reduced
testing scenarios had lower costs than the equivalent scenarios without reduced testing.
Increasing the LSI to 730 days reduced the costs for testing in the normal hygiene herd. This
was also true for the level of PC, so that higher PC reduced the costs for testing.
5

Fig. 2. Boxplots of the mean annual expenses (EUR) for ELISA testing for each scenario.
X-axis labels show the short sampling interval, long sampling interval, prevalence cutoff and
when the reduced test scheme is used. SSI = Short sampling interval, LSI = Long sampling
interval. So in the second scenario the SSI is 31 days, the LSI is 365 days and reduced testing
was not used. Solid bar is the median value, box represents 25th–75th percentiles and
whiskers show range of values.
DISCUSSION
This study showed that the adaptive test strategy is a feasible way to save costs for testing
while still controlling paratuberculosis within a dairy farm, supporting previous results
(Kirkeby et al. 2016b). In the baseline scenario the costs for testing were about 2750 EUR per
year. In the scenario where PC was set to 1%, SSI = 3 months and LSI = 1 year, the normal
hygiene herd could generally save 500 EUR yearly for testing (Fig. 2). If this strategy was
combined with the reduced testing approach, the costs for testing were around 1200 EUR
yearly, saving about 1550 EUR yearly (Fig. 2).
6

When combining the reduced test approach with the adaptive strategy, the costs for ELISA
testing were generally reduced (Fig. 2). However, the reduced test strategy was not as
efficient in decreasing the prevalence (Fig. 1). How to balance between reducing test costs
and prevalence in the herd must be decided by the individual farmer. This consideration
should also include the speed of decrease in prevalence and the likely economic effect of the
chosen strategy. This study described some new tools to perform this task, namely PC, SSI,
LSI and reduced testing. The likely outcomes of using the different strategies in a typical
Danish herd are provided in this paper, and should be used as a tool for farmers to prioritize.
Increasing the PC resulted in an increase in the prevalence at the end of the simulations.
But it also dramatically reduced the costs for testing, especially in the scenarios with long test
intervals (Fig. 2). If a farmer believes that reducing the prevalence to 5% is achievable and
realistic, he may use the results from this study to decide on a higher PC. This can be an
important first step towards control of paratuberculosis within a herd. Farmers have to
prioritise their efforts against different diseases, and a more relaxed control scheme could
potentially be an affordable method to get more farmers to control paratuberculosis. Another
argument for using the adaptive test approach is to keep farmers within a control program.
Once the prevalence is close to zero, the farmer may want to leave the control programme
because the prevalence of MAP is now low and he has to prioritize other diseases etc.
However, if the adaptive test scheme is used, the costs for testing will also be low, increasing
the incentive to stay in the programme, but now with a purpose that more resemblance
surveillance than control.
Long SSI intervals caused the prevalence to increase, because no tests were performed for
long periods. This prevents advantages of repeated testing, which has generally been an
advantage in the Danish paratuberculosis control programme. When the SSI was decreased
from three months to one month, no noticeable change could be seen (Fig. 1). This is
probably because the SSI of three months is really good for identification of infected animals.
In a herd with lower hygiene and thus higher prevalence of MAP, the results will likely be
very different because it is often not possible to prioritise culling of MAP cows only (Kirkeby
et al. 2016a).
ACKNOWLEDGEMENTS
This project was funded by the Green Development and Demonstration Program (GUDP)
under the Danish Directorate for Food, Fisheries and Agriculture, grant no. 34009-13-0596.
REFERENCES
Græsbøll, K., Nielsen, S.S., Halasa, T.H., Christiansen, L.E., Kirkeby, C. and Toft, N. (2015).
Association of map specific ELISA-responses and productive parameters in 367 Danish
dairy farms. In Proceedings of the 12th International Colloquium on Paratuberculosis,
Parma, Italy, June 22-26. 139–141
7

Græsbøll, K,, Kirkeby C, Nielsen SS, Halasa TH, Toft N and Christiansen LE. (2016).
Models to estimate lactation curves of milk yield and somatic cell count in dairy cows at the
herd level for the use in simulations and predictive models. Front. Vet. Sci. 3.
DOI:10.3389/fvets.2016.00115
Kirkeby, C., Græsbøll, K., Nielsen, S.S., Christiansen, L.E., Toft, N., Rattenborg, E. and
Halasa, T.H. (2016a). Simulating the epidemiological and economic impact of
paratuberculosis control actions in dairy cattle. Front. Vet. Sci. 3,
DOI:10.3389/fvets.2016.00090
Kirkeby, C., Græsbøll, K., Nielsen, S.S., Christiansen, L.E., Toft, N., Rattenborg, E. and
Halasa, T.H. (2016b). Adaptive Test Schemes for Control of Paratuberculosis in Dairy
Cows. PLOS ONE 11, e0167219
Lu, Z., Mitchell, R.M., Smith, R.L., Van Kessel, J.S., Chapagain, P.P., Schukken, Y.H. and
Gröhn, Y.T (2008). The importance of culling in Johne's disease control. J. Theor.
Biol. 254, 135-146
Nielsen, S.S. (2014). Developments in diagnosis and control of bovine paratuberculosis. CAB
Reviews. 9, 1-12
Nielsen, S.S., Jepsen, O.R. and Aagaard, K. (2007). Control programme for paratuberculosis
in Denmark. Bull. Int. Dairy Fed. . 410, 23
Nielsen, S.S. and Toft, N. (2009). A review of prevalences of paratuberculosis in farmed
animals in Europe. Prev. Vet. Med. 88, 1-14
Ott, S.L., Wells, S.J. and Wagner, B.A. (1999). Herd-level economic losses associated with
Johne's disease on US dairy operations. Prev. Vet. Med. 40, 179-192
Rogan, W.J. and Gladen, B. (1978). Estimating prevalence from the results of a screening
test. Am. J. of Epidemiol. 107, 71-76
8

Sergeant, E.S.G., Nielsen, S.S. and Toft, N. (2008). Evaluation of test-strategies for
estimating probability of low prevalence of paratuberculosis in Danish dairy herds. Prev.
Vet. Med. 85, 92-106
Sweeney, R.W. (2011). Pathogenesis of paratuberculosis. Veterinary Clinics of North
America: Food Animal Practice. 27, 537-46
Whitlock, R.H., Wells, S.J., Sweeney, R.W. and Van Tiem J. (2000). ELISA and fecal
culture for paratuberculosis (Johne’s disease): sensitivity and specificity of each
method. Vet Microbiol. 77, 387-398
Whittington, R.J. and Sergeant, E.S.G. (2001). Progress towards understanding the spread,
detection and control of Mycobacterium avium subsp. paratuberculosis in animal
populations. Aust. Vet. J. 79, 267-278
9