Modeling the effect of test-and-slaughter strategies to control
bovine tuberculosis in endemic high prevalence herds
Running title: Modeling bTB control strategies in endemic herds
C. Picasso-Rissoa,b*, J. Alvarezc,d, K. VanderWaala, A. Kinsleya, A.
Gilb, Scott J. Wellsa, A.
Pereza
a Dept. of Veterinary Population Medicine,
University of Minnesota, Saint Paul, MN;
b Facultad de Veterinaria, Universidad de la
Republica, Montevideo, Uruguay;
c VISAVET Health Surveillance Centre,
Universidad Complutense, Madrid, Spain;
d Departamento de Sanidad Animal, Facultad de
Veterinaria, Universidad Complutense de Madrid, Spain.
* Corresponding author.
E-mail address:picas001@umn.edu (C.
Picasso-Risso).
Abstract
Bovine tuberculosis (bTB) prevalence substantially increased over the
past two decades with relatively high impact on large dairy herds,
raising the concern of regulatory authorities and industry stakeholders,
and threatening animal and public health. Lack of resources, together
with the economic and social consequences of whole-herd stamping-out,
makes depopulation an impractical disease control alternative in these
herds. The increase in bTB-prevalence was associated with demographic
and management changes in the dairy industry in Uruguay, reducing the
efficacy of the current control program (i.e. status quo ) based
on intradermal serial testing with caudal fold- and comparative
cervical- tuberculin test-and slaughter of reactors (CFT-CCT). Here, we
aimed to assess the epidemiological effectiveness of six alternative
control scenarios based on test-and-slaughter of positive animals, using
mathematical modeling to infer bTB-within-herd dynamics. We simulated
six alternative control strategies consisting of testing adult cattle
(>1 year) in the herd every three months using one test
(in-vivo or in-vitro) or a combination in parallel of two tests (CFT,
interferon-gamma release assay –IGRA- or Enzyme-linked immunosorbent
assay). Results showed no significant differences overall in the time
needed to reach bTB-eradication (median ranging between 61 to 82 months)
or official bovine tuberculosis-free status (two consecutive negative
herd-tests) between any of the alternative strategies and thestatus quo (median ranging between 50 and 59 months).However, we demonstrate how alternative strategies can significantly
reduce bTB-prevalence when applied for restricted periods (6, 12, or 24
months), and in the case of IGRAc (IGRA using peptide-cocktail
antigens), without incurring on higher unnecessary slaughter of animals
(false-positives) than the status quo in the first 6 months of
the program (P-value <0.05). Enhanced understanding
bTB-within-herd dynamics with the application of different control
strategies help to identify optimal strategies to ultimately improve
bTB-control and -eradication from dairies in Uruguay and similar endemic
settings.
Keywords: simulation modeling, test-and-slaughter, disease
control, Uruguay, Mycobacterium bovis , dairy cattle, disease
modeling
Introduction
Mycobacterium bovis (M. bovis ) is the primary cause of
bovine tuberculosis, one of the most widespread zoonotic bacterial
infections affecting cattle and other mammals (OIE 2016). Limited
success in disease control has been achieved worldwide (Bezos et al.,
2014; Good et al., 2018; More, Radunz, & Glanville, 2015; Morris, 2015)
due to the chronic nature of the disease, the presence of wildlife
reservoirs and the imperfect accuracy of currently available diagnostic
strategies (Schiller et al. 2010).
The use of the tuberculin intradermal test has proven useful in
detecting infected herds (Bezos et al., 2014; de la Rua-Domenech et al.,
2006); however, complex host-pathogen interactions can lead to low
accuracy when considering the infection status of an individual (Gormley
et al., 2006; Gormley et al., 2004), challenging bTB-eradication after
the establishment of infection. In herds with high prevalence
(>10%), bTB-eradication with the use of the intradermal
test and individual slaughter is difficult yet important, as can help to
reduce pathogen circulation and consequently decrease potential zoonotic
risk, and negatively impact animal health and welfare.
In bTB-high prevalence settings, two approaches to overcome the limited
sensitivity of the skin tests have been used for eradication of the
disease at the herd level globally. One is the use of ancillary in-vitro
tests to maximize the ability to detect infected cattle in the herd
(Council Directive 64/432/EEC, Casal et al. 2014), and the other is the
depopulation of the herd from which individual reactors were detected
(More, Radunz, & Glanville, 2015; Verteramo Chiu et al., 2019).
Advantages and limitations of the use of in-vitro ancillary diagnostic
tests have been reviewed elsewhere (Bezos, Casal, et al. 2014; de la
Rua-Domenech et al. 2006), but essentially these tests target infected
animals that may be missed by the intradermal test. Whole herd
depopulation, though costly, is an effective strategy to control and
eradicate bTB since it ensures disease elimination (More et al. 2015).
However, the complexity of its implementation increases with the size of
the bTB-infected herd, making it implausible at times, given the limited
resources for the compensation to farmers for culled animals in many
countries. The lack of resources, together with economic and social
implications makes depopulation difficult to justify to stakeholders,
especially for herds in which a low proportion of the animals test
positive, as often seen in endemic settings (Ciaravino et al., 2017).
In Uruguay, bTB-control programs rely on the serial application of the
caudal fold tuberculin test –CFT- followed by the comparative cervical
tuberculin test –CCT- in the reactors and subsequent slaughter of CCT
positive animals for confirmation of M. bovis infection by a
bacteriological culture of selected tissues sampled at the
slaughterhouse. Herds in which bTB-infection is confirmed are subjected
to retesting using intradermal tests (CFT+CCT) every 60 to 180 days,
until two consecutive negative results are achieved, leading to
regaining the officially tuberculosis-free status (OTF) (MGAP 1989).
Government indemnity is provided to farmers for those animals
slaughtered under the regulations of the bTB-control program (Law 19300,
26/12/2014.DGSG/MGAP).
In the Uruguay cattle population, with the above described bTB-control
program in place, a low bTB-prevalence was traditionally reported
(<0.001 at the herd level) (WAHIS_OIE 2014). Within the past
two decades, however, the number of bTB-infected dairy herds, the
within-herd bTB-prevalence, and the time required from detection to
regaining the officially tuberculosis-free status (OTF) increased (
Picasso-Risso et al. 2019), leading to unprecedented challenges in the
control of bTB in Uruguay. The difficulty in controlling bTB has been
associated with changes in dairy demographic structure and management
(Picasso et al. 2017; Picasso-Risso et al. 2019), including larger herds
(>360 animals), higher animal density (DIEA 2018),
increased animal movements, and more intensive animal rearing than
traditional dairy farming prior to the 1990s (Picasso et al. 2017). In
this context, the question is whether the current Uruguayan bTB-program
is sufficient to control bTB in herds once the infection is confirmed
and some level of within-herd transmission is suspected.
Mathematical models have been broadly used to understand within-herd
bTB-transmission patterns and to evaluate control and surveillance
strategies (Alvarez et al., 2014; Brooks-Pollock, Roberts, & Keeling,
2014; Ciaravino et al., 2018; Perez, Ward, & Ritacco, 2002; Rossi, et
al., 2019). These models account for the chronic nature of bTB,
considering the long and variable incubation periods, biological
variabilities, and the influence of different production systems
(Alvarez et al. 2014) while avoiding the risks and the costs of in-vivo
implementation (Halasa and Dürr 2017). An integrated within-and
between-herd model has been parameterized and validated to evaluate the
performance of risk-targeted bTB-surveillance using the current
test-and-slaughter bTB- strategies in Uruguay (VanderWaal et al. 2017).
However, previous studies have suggested that the sensitivity of the
test-and-slaughter program is impaired in high prevalence dairy herds in
Uruguay (Picasso-Risso et al. 2019), and given that depopulation of
these large herds is not economically, logistically, or socially
feasible, the use of alternative diagnostic in-vitro assays is a
reasonable alternative strategy to evaluate for control in these herds.
In this study, we aimed to assess the effectiveness of different
alternative bTB-control strategies, to ultimately elucidate the optimal
option for control in bTB-high prevalence dairy herds in Uruguay when
depopulation is not an alternative.
Methods:
Model description
Our objective was to model disease dynamics at the herd level for in
dairies in Uruguay, considering herd demographics, bTB transmission, and
bTB-control measures. We evaluated the relative effectiveness of thestatus quo and six alternative control scenarios in dairy herds
with bTB-high prevalence (>10%) (Table 1) using a modified
within-herd bTB transmission model developed and parameterized for
Uruguayan cattle herds (VanderWaal et al. 2017). The outputs of the
model were a) time-to-bTB-eradication, b) bTB-prevalence at six months
and ten years post-infection, c) time-to-regain the OTF status, and d)
proportion of animals slaughtered in the herd.
Herd demographics
We used two animal categories, adults (≥12 months), and calves
(<12 months), with calves becoming adults after reaching one
year of age (rate 1/12 months) (VanderWaal et al. 2017). Slaughter,
birth, and replacement of death/culled rates were assumed to follow a
Poisson distribution based on demographic information available. The
average proportion of routine slaughter for adults
(λsl.a) and calves (λsl.c ) was 0.268
and 0.007 respectively (VanderWaal et al. 2017). Then, in order to
maintain a stable herd size, births and adult replacements were
happening every four months (s = 4), with births at the same rate
of calves slaughters (λsl.c), and adult replacements at
similar rates as adult slaughter (λsl.a). The model was
initialized with a population of 500 animals and an adult to calf ratio
of 75/25 following the typical demographic characteristics of large
dairy herds in the past decades in Uruguay (DIEA 2018).
Individual-based bTB-transmission
dynamics
bTB-transmission was simulated using a stochastic, discrete, compartment
model, in which animals transitioned through four mutually exclusive
stages; susceptible (S), Occult (O), Reactive (to diagnostic tests) (R),
and Infectious (I; SORI model) (Alvarez et al. 2014; Ciaravino et al.
2018; Conlan et al. 2012; Perez et al. 2002). When healthy animals from
the susceptible compartment (S) are infected with M. bovis , they
transition to the Occult (O) state for a latent period
(λ1) in which even though infected, they are not
detectable by antemortem bTB tests or infectious.
As the disease progresses, occult animals become detectable by
diagnostic tests and move to the sub-compartments Ra and
Rb based on two different times for detection
(λ2a and λ2b) respectively:
λ2a represents the period until IGRA is able to identify
the infection, and λ2b is the time until all the other
diagnostic tests implemented can identify bTB-infected animals.
The final compartment represents animals that become infectious (I)
while remain also detectable by the antemortem diagnostic tests (Figure
1a). A similar SORI-model was applied for the two age categories
(adults/calves). At each time step, the number of animals that
transitioned between compartments was selected from a Poisson
distribution with the purpose of incorporating stochasticity to the
model as previously described (Gillespie 2001; Keeling and Rohani 2008).
Transition rates were based on the following deterministic backbone
differential equations:
\(\frac{\text{dS}_{\text{cal}}}{\text{dt}}=-\left(\text{β\ }\frac{S_{\text{calv}}(I_{\text{calv}}+I_{\text{ad}})}{N}\right)\);\(\frac{\text{dS}_{\text{ad}}}{\text{dt}}=-\left(\text{β\ }\frac{S_{\text{ad}}(I_{\text{calv}}+I_{\text{ad}})}{N}\right)\)
\(\frac{\text{dO}_{\text{calv}}}{\text{dt}}=\left(\text{β\ }\frac{S_{\text{calv}}(I_{\text{calv}}+I_{\text{ad}})}{N}\right)-\left(O_{\text{calv}}\frac{1}{\lambda_{1}}\right)\);\(\frac{\text{dO}_{\text{ad}}}{\text{dt}}=\left(\text{β\ }\frac{S_{\text{ad}}(I_{\text{calv}}+I_{\text{ad}})}{N}\right)-\left(O_{\text{ad}}\frac{1}{\lambda_{1}}\right)\)
\(\frac{\text{dRa}_{\text{calv}}}{\text{dt}}=\left(O_{\text{calv}}\frac{1}{\lambda_{1}}\right)-\left(\text{Ra}_{\text{calv}}\frac{1}{\lambda_{2a}}\right)\);\(\frac{\text{dRa}_{\text{ad}}}{\text{dt}}=\left(O_{\text{ad}}\frac{1}{\lambda_{1}}\right)-\left(\text{Ra}_{\text{ad}}\frac{1}{\lambda_{2a}}\right)\)
\(\frac{\text{dRb}_{\text{calv}}}{\text{dt}}=\left(\text{Ra}_{\text{calv}}\frac{1}{\lambda_{2a}}\right)-\left(\text{Rb}_{\text{calv}}\frac{1}{\lambda_{2b}}\right)\);\(\frac{\text{dRb}_{\text{ad}}}{\text{dt}}=\left(\text{Ra}_{\text{ad}}\frac{1}{\lambda_{2a}}\right)-\left(\text{Rb}_{\text{ad}}\frac{1}{\lambda_{2b}}\right)\)
\(\frac{\text{dI}_{\text{calv}}}{\text{dt}}=\left(\text{Rb}_{\text{calv}}\frac{1}{\lambda_{2b}}\right)\);\(\frac{\text{dI}_{\text{ad}}}{\text{dt}}=\left(\text{Rb}_{\text{ad}}\frac{1}{\lambda_{2b}}\right)\)
We assumed that animals within the adult and calf compartments interact
with equal probability (homogeneous mixing) and that transmission is
frequency-dependent (Smith et al., 2013; VanderWaal et al., 2017).
Individual-based bTB-control
dynamics
We compared the current control strategy (i.e. status quo ) to six
alternative scenarios applied to adult animals (>12 months)
in the herd. Testing was performed every three months for all adult
animals, consistent with the status quo control strategy in
Uruguay (MGAP 1989). The alternative strategies aimed to improve the
sensitivity of the control program (status quo testing) by using
a maximum of two diagnostic tests every testing period. The six
alternatives were the use of CFT only, the use of IGRA (assessing two
different antigens commercially available for the region), and the
parallel combination of CFT+IGRA, CFT+ELISA, or IGRA+ELISA (Table 1).
For the parallel combination, the IGRA selected used peptide cocktail
antigens given it demonstrated to have an improved specificity with
similar sensitivity under Uruguayan field conditions (Picasso-Risso et
al. 2019).
The sensitivity (Se) and specificity (Sp) of each testing strategy were
modeled using different beta distributions for the different stages
(Table 2) based on previous estimates (Picasso-Risso et al. 2019).
Test-positive animals were immediately removed from the herd before the
following testing period. To track the number of slaughtered animals we
created four mutually exclusive compartments, one for the
false-positives (susceptible false positives -SS-) and
three for the true-positives (reactors or infectious positives
-RaS, RbS, IS-) to each
control strategy (Figure 1b). The number of reactor animals in each of
these four compartments were assumed to be drawn from a Poisson
distribution centered on the expected number of false-positive animals
given the susceptible population (Ss), and on the
expected number of true-positives in the infected compartments
(Ras, Rbs, Is) using
equations 1 and 2 for strategies involving one-test or parallel testing
respectively. Dependency between the test results when applying two
tests was introduced through positive and negative correlation
coefficients (ρDc and ρD ) following the distributions
described previously for Uruguay (Picasso-Risso et al. 2019).
Eq 1. False-Positives:\(S_{s}=[1-Sp)]*S_{\text{ad}}\) (single
testing)
\(S_{s}=[1-{(Sp}_{1}*\ \text{Sp}_{2}\ +\ rhoD)]*S_{\text{ad}}\)(parallel testing)
Eq 2. True-Positives:\(\text{Ra}_{s}=Se*\text{Ra}_{\text{ad}}\) ;\(\text{Rb}_{c}=Se*\text{Rb}_{\text{ad}}\) ;\(I_{s}=Se*I_{\text{ad}}\) (single
testing)
\(\text{Ra}_{s}={(Se}_{1}*\ \text{Se}_{2}\ +\ rhoDc)*\text{Ra}_{\text{ad}}\)(parallel testing)
\(\text{Rb}_{s}={(Se}_{1}*\ \text{Se}_{2}\ +\ rhoDc)*\text{Rb}_{\text{ad}}\)
\(I_{s}={(Se}_{1}*\ \text{Se}_{2}\ +\ rhoDc)*I_{\text{ad}}\)
Assessment of alternative strategies
The model was run without application of any control strategy until the
median apparent prevalence (the sum of the animals in compartments R and
I) of 500 iterations reached 10% (high prevalence herd). Then, the
median number of animals in each of the infected compartments (O-R-I)
was used to seed each of the six models evaluating bTB-control
strategies (Supplementary Figure S1).
Models with each control strategy were run for 500 simulations for 20
years, and results were summarized as median, and 2.5, 25, 75, and
97.5% intervals, meaning the interval containing 2.5, 25, 50, 75, and
97.5% of the outcomes. Differences between the outcomes were compared
using the Kruskal-Wallis test (Kruskal and Wallis 1952), Dunn’s test for
pairwise comparison, and log-rank test to compare time to eradication
and OTF.
Results:
Epidemiological
indicators:
Estimates of the time to achieve bTB-eradication, both in the whole herd
and when considering each of the age categories separately, showed
slightly different medians depending on the scenario considered (Figures
2-3). The median time to eradication ranged from 61 to 82 months when
considering the whole herd, and if only adults were considered, from 41
to 52 months (Table 3). Towards the end of the outbreak (when
bTB-prevalence in adults reached zero), calves (which were not tested)
carried most of the residual infections, maintaining the circulation of
the disease for significant (Kruskal-Wallis P-value <0.05)
longer periods (Figure 3).
There was a significant difference (P-value<0.05) in
bTB-prevalence at the early stages of the outbreak (6, 12, and 24
months) when alternative strategies were applied compared to thestatus quo (Table 4). At the end of the second year of
simulations, bTB-prevalence estimates under the status quo were
only significantly (P-value =0.01) different from those generated when
considering the CFT+IGRA strategy. Differences with estimates from the
IGRAb and IGRAc scenarios were not considered significant though were
close to the p-value threshold (P-value=0.22) (Table 4).
Time to regain OTF status did not vary between the strategies simulated
(P-value >0.05) (Table 3), with median estimates ranging
between 50 and 59 months (4.1 to 4.9 years) under all scenarios.
Performance
effectiveness:
The simulated scenarios using ELISA as an ancillary test (IGRA+ELISA and
CFT+ELISA) led to a higher proportion of animals testing positive, with
higher rates of false-positives (0.155, 95thpercentile:0.120-0.191 and 0.177, 95th percentile:
0.141-0.212 (Figure 4) than the status quo or any of the other
alternative scenarios. The status quo scenario yielded the lowest
proportion of positive diagnostic results, and the lowest estimates for
the proportion of false-positive results (median: 0.08,
95th percentile: 0.05-0.13), which were significantly
different from all the other simulated strategies according to the
Kruskal-Wallis test (P-value <0.05).
Discussion
In this study, we simulated bTB-transmission under very specific
conditions (large herds and a high within-herd apparent prevalence) in a
dairy cattle herd in Uruguay to try to assess the performance of current
and potentially available control strategies, given that this is an
emerging problem faced by animal health authorities in the country for
which the effectiveness of the tools at hand has not been evaluated
extensively. The fixed size of the herd (500-cattle) was considered
large for the country given that it is
>75th percentile for dairies in Uruguay
(DIEA 2018; Picasso et al. 2017). Furthermore, the starting apparent
prevalence of 10% was high most challenging bTB-infected herds in
Uruguay (Picasso Risso 2016).
In this context, we assumed the presence of two independent populations
in the herd (adults and calves), of which only the former would be
subjected to every three months to seven control strategies
(status quo and six alternatives) considered alternatively.
Alternative scenarios considered tools currently available (and applied)
for bTB control elsewhere. Among these, the application of IGRAs can
help in earlier detection (~2 weeks) of the bTB-
cell-mediated immune response in comparison to the intradermal test (or
the ELISA) (Bezos, Casal, et al. 2014; de la Rua-Domenech et al. 2006).
By including two subcompartments (Ra and Rb) in the R compartment, our
analysis accounted for variations in the duration of the detection
period described for the IGRAs (de la Rua-Domenech et al. 2006a). Two
different protocols for the application of the IGRA where considered,
based on the use of PPD or a more specific set of antigens that under
the Uruguayan conditions has been proved to provide a similar Se with no
loss in Sp (Picasso-Risso et al. 2019). Furthermore, high bTB-prevalence
herds tend to have animals in various stages of the disease, including
advanced stages, which can result in an improved Se for antibody-based
diagnostics such as ELISA (de la Rua-Domenech et al. 2006; Waters et al.
2011). Hence, ELISA was assessed as an alternative strategy to the
cell-mediated diagnostics for its potential in these endemic herds.
We found there was no difference in the time to reach bTB-eradication or
OTF-status between the six alternatives and the status quo(median 73, 36-133 months, and 59, 26-122 months respectively), although
the large variability in the estimates obtained could have limited our
ability to detect such differences. The relative costs associated with
the slaughter of uninfected (but test-positive) animals was
significantly increased with the use of most of the alternative
strategies (except for IGRAc), with the highest cost performance
observed with strategies including the ELISA (Supplementary Figure S2).
The lack of significant improvement was associated with the maintenance
of the disease in the calf category (Figure 3). In most of the simulated
scenarios, eradication was reached earlier in the adult category than
the calf category (Table 3), since calves remain undetected until
reaching the age to be tested. Undetected calves were responsible for
sustaining bTB in the herd for longer periods in most simulations (Table
3, Figure 3). We, therefore, need to consider this conclusion might not
hold when simulating control strategies that include calfhood testing.
When exploring the effect of the control strategies in shorter periods
(after 6, 12 and 24 months), we observed a significant reduction in
bTB-prevalence after the first 6 and 12 months with the use of any of
the six alternative strategies of control, and after 24 months with the
use of CFT+IGRA parallel testing. The model outputs suggest that
alternative strategies can be selected as an initial strategy, and then
could be followed by the use of the current status quo strategy
for eradication. In addition, when the cost of implementation of
alternative strategies was assessed (including for unnecessary
slaughter), the unique strategy that matched the performance of thestatus quo in the first 6 months of testing (P-value
>0.05) was the use of IGRAc (Supplementary Figure S2). This
finding suggests that IGRAc might be an effective tool to quickly reduce
bTB-prevalence at the initial stages of the control program (six months
or two consecutive tests).
The Se and Sp of the diagnostic tests considered in the different
control scenarios were based on estimates of accuracy based on
information from high bTB-prevalence dairy herds in Uruguay
(Picasso-Risso et al. 2019), which helped reduce the uncertainty on test
performance derived from the sometimes contradictory estimates described
in previous studies (Alvarez et al. 2012; Bezos, Casal, et al. 2014) ).
Testing intervals (3 months) represent a high pressure for detection ofM. bovis using intradermal testing to elude the anergy period
described as a result of multiple intradermal inoculations (Radunz and
Lepper 1985; de la Rua-Domenech et al. 2006b; Vordemeier et al. 2006),
and logistically allow slaughter before the next testing period.
Although in-vitro testing allows for more frequent testing and can
benefit from the booster effect after tuberculin inoculation when
applied in parallel (CFT+IGRA) (Casal et al. 2014; Palmer et al. 2006;
Schiller et al. 2010), we chose to assess the strategies in reference to
the status quo , and we avoided inclusion of shorter
testing-intervals. We recognize that a deeper understanding of the
effect of different testing periods is needed.
In order to identify optimal testing strategies, we balanced the
epidemiologic effectiveness of disease control while minimizing the
unnecessary culling of false reactors relative to the status quo .
While an initial useful approximation of the additional efforts imposed
by each strategy, a next step is to provide an estimation of the
economic cost, including costs of testing as well as unnecessary culling
of false-positive cattle (Kao, Roberts, & Ryan, 1997; Kao et al., 2018;
Smith et al., 2013) and social acceptance (Ciaravino et al., 2017); both
are essential before implementation.
Conclusions
In this report, we conclude that the assessed alternative strategies
were not able to improve the time to bTB-eradication, time to regain the
OTF-status, or were able to reduce the number of false-positive cattle.
Results from this study contribute to the understanding of the
implications of applying different testing pressures in highly infected
dairy herds in Uruguay. Additionally, we showed the importance of
targeting surveillance and control strategies to infected calves, the
benefit of using the IGRAc as an ancillary test in initial stages of
control. Determination of the best testing strategy will be a result of
epidemiologic, performance, and economic balance while acknowledging
logistics and socio-cultural perceptions of individual countries and
regions. Our results enhance understanding of bTB-within-herd dynamics
and identify optimal bTB-control strategies for dairies in Uruguay and
similar endemic settings.
Acknowledgments
CPR is a recipient of the Uruguayan Agency of Research and Innovation
(ANII in the Spanish language) fellowship which partially supported this
work. JA is the recipient of a Ramón y Cajal postdoctoral contract from
the Spanish Ministry of Economy, Industry, and Competitiveness (MINECO)
(RYC-2016-20422).