- Email: [email protected]
Control of African swine fever epidemics in industrialized swine populations
Veterinary Microbiology 197 (2016) 142–150
Contents lists available at ScienceDirect
Veterinary Microbiology journal homepage: www.elsevier.com/locate/vetmic
Control of African swine fever epidemics in industrialized swine populations Tariq Halasaa,* , Anette Bøtnera , Sten Mortensenb , Hanne Christensenb , Nils Tofta , Anette Boklunda a b
National Veterinary Institute, Technical University of Denmark, Copenhagen, Denmark Danish Veterinary and Food Administration, Ministry of Environment and Food, Glostrup, Denmark
A R T I C L E I N F O
Article history: Received 30 August 2016 Received in revised form 11 November 2016 Accepted 14 November 2016 Keywords: African swine fever Simulation model Spread Control
A B S T R A C T
African swine fever (ASF) is a notifiable infectious disease with a high impact on swine health. The disease is endemic in certain regions in the Baltic countries and has spread to Poland constituting a risk of ASF spread toward Western Europe. Therefore, as part of contingency planning, it is important to explore strategies that can effectively control an epidemic of ASF. In this study, the epidemiological and economic effects of strategies to control the spread of ASF between domestic swine herds were examined using a published model (DTU-DADS-ASF). The control strategies were the basic EU and national strategy (Basic), the basic strategy plus pre-emptive depopulation of neighboring swine herds, and intensive surveillance of herds in the control zones, including testing live or dead animals. Virus spread via wild boar was not modelled. Under the basic control strategy, the median epidemic duration was predicted to be 21 days (5th and 95th percentiles; 1-55 days), the median number of infected herds was predicted to be 3 herds (1–8), and the total costs were predicted to be s326 million (s256–s442 million). Adding pre-emptive depopulation or intensive surveillance by testing live animals resulted in marginal improvements to the control of the epidemics. However, adding testing of dead animals in the protection and surveillance zones was predicted to be the optimal control scenario for an ASF epidemic in industrialized swine populations without contact to wild boar. This optimal scenario reduced the epidemic duration to 9 days (1–38) and the total costs to s294 million (s257–s392 million). Export losses were the driving force of the total costs of the epidemics. ã 2016 Elsevier B.V. All rights reserved.
1. Introduction African swine fever (ASF) is a notifiable infectious disease in pigs. It is caused by ASF virus (ASFV), a DNA virus from the family Asfarviridae, genus Asfivirus (cited from Gallardo et al., 2009). The disease is endemic in Africa (Chenais et al., 2015), the Russian Federation, and in certain regions of the Baltic countries (Gallardo et al., 2014, 2015b; Olsevskis et al., 2016). It is considered to be a substantial threat for Western Europe (EFSA-Panel, 2014). In countries with a large production and/or export of swine and swine products, an outbreak of ASF may result in devastating economic consequences for the swine industry due to export restrictions. Therefore, it is important to explore the effectiveness
* Corresponding author at: National Veterinary Institute, Technical University of Denmark, Bülowsvej 27, 1870 Frederiksberg C, Copenhagen, Denmark. E-mail address: [email protected] (T. Halasa). http://dx.doi.org/10.1016/j.vetmic.2016.11.023 0378-1135/ã 2016 Elsevier B.V. All rights reserved.
and consequences of strategies to control outbreaks of ASF in the industrialized swine populations. The EU has established a set of strategies that should be followed in the case of an outbreak of ASF in the domestic swine populations (CEC, 2002). To our knowledge, the effectiveness of these strategies and their combination with other strategies, such as pre-emptive depopulation of neighboring swine herds or intensive surveillance in the control zones has never been investigated before. Identifying effective control strategies will assist the national veterinary authorities in the development of national guidelines for ASF control and contingency planning. Simulation models of disease spread is a widely used tool to assist the national veterinary authorities in contingency planning (e.g. Backer et al., 2009; Martínez-López et al., 2011; Boklund et al., 2013; Halasa et al., 2015). Such models are invaluable for exploring mechanisms of disease spread and control, taking into account the complexities of agricultural systems.
T. Halasa et al. / Veterinary Microbiology 197 (2016) 142–150
The objective of this study was to compare the epidemiological and economic effectiveness of different strategies to control a hypothetical epidemic of ASF in industrialized swine populations, where the role of wild boar in ASFV spread is negligible, exemplified by the Danish swine population. 2. Materials and methods This section describes the simulation of ASFV spread between swine herds using different control strategies. It provides information about the herd data used in the model, ASFV-spread mechanisms, ASFV detection and the control strategies that are simulated. Important input parameters are presented in Table S1 in the Supplementary materials. Details on model parameters and equations can be found in Halasa et al. (2016a). 2.1. Herd and movement data We used geographical data (UTM coordinates), the number of animals, and specification of herd types for the 8262 swine herds registered in the Danish Central Husbandry Register (CHR) in 2014. Descriptive data on the number of herds, herd sizes and frequencies of outgoing movements was presented in Halasa et al. (2016a). For each herd, the daily frequency of moving animals to another herd was calculated as the sum of all registered herd movements in the period from 01 January 2014 to 31 December 2014 divided by 365, based on the registrations in the Pig Movement Database. For each herd, this frequency was used as the mean (l) in a Poisson distribution, describing the number of daily outgoing movements (batches) of animals. Similarly, the probability of moving animals to an abattoir was calculated for each herd. By analyzing the distances between source herds and receiving herds, it was found that animals from nucleus herds were moved further than animals moved from other herd types. As a result, two separate distributions for movement distances were used to model the movements of animals from nucleus herds and from other herd types, respectively (Halasa et al., 2016a). The probabilities of animal movements from one herd type to another were calculated based on the Pig Movement Database and are presented in Halasa et al. (2016a). 2.2. The simulation model The DTU-DADS-ASF model (version 0.15.1) (Halasa et al., 2016a) was used. The model runs in the statistical computing language R (version 3.1.3) (R Core Team, 2015). The parameterization of the model was based on the Georgian strain of ASFV, reflecting the original strain of the epidemics currently running in Eastern Europe as explained earlier (Halasa et al., 2016a). 2.2.1. Modelling ASF spread ASFV spread was modelled in two processes: 1) spread within a herd; 2) spread between herds by different mechanisms (transmission routes). 2.2.1.1. Modelling ASFV spread within a herd. The infection model for the individual animals is a state transition model with the following states: susceptible-latent-subclinical-clinical-removed (SLSCR model; Halasa et al., 2016b). The infection model for the herd follows the same model, but includes the possibility of an infected herd to become susceptible again, should the infection fade out before all animals in the herd are infected. Infected herds will start as latent and progress to the subclinical and clinical states following infection. The infection is then either detected, and therefore the herd is removed (culled), or it becomes susceptible again.
143
2.2.1.2. Modelling ASFV spread between herds. ASFV is simulated to spread between herds via animal movements, abattoir movements, via indirect medium-risk contacts (direct contact to animals such as contacts by veterinarians or artificial inseminators) or low-risk contacts (no direct contacts to animals, such as feed trucks and visitors), or via local spread. Each type of contact was modelled as a Poisson distribution. For movements of animals to other herds or to abattoirs, the mean (l) was calculated for the individual herd, as described above, while for indirect medium and low risk contacts, a Poisson distribution was modelled for each herd type. For animal movements, each movement represented a batch of animals moved from the sending infectious herd to the receiving herd. The probability of transmitting the ASFV from the infectious herd to the receiving herd was dependent on the prevalence of the disease within the infectious herd and the number of animals moved in the batch (Halasa et al., 2016a). Local spread was modelled in a distance up to 2 km around infectious herds, and was assumed to consist of a mixture of unregistered animal movements, shared equipment and tools, and spread via rodents and insects. Detailed information including the equations and steps for modeling each of these mechanisms can be found in Halasa et al. (2016a). The risk of ASFV spread and/or maintenance through wild boar was not modelled as the number of wild boar in Denmark is limited due to intensive farming in the country, leaving few suitable habitats for wild boar (Alban et al., 2005; Jordt et al., 2016). There is also a Danish legal requirement to eliminate stray wild boar (Anonymous, 2015c). 2.2.2. Modelling ASFV detection In the model, the ASFV infection can be detected by three different mechanisms: passive surveillance before first detection; passive surveillance after first detection; and active surveillance. In passive surveillance, before first detection, detection was modelled to occur when 1) the proportion of sick or dead animals (referred to as SIED throughout the paper) reached 2.55% (Halasa et al., 2016a); and 2) the proportion of SIED animals relative to the expected cumulative mortality level within the herd (in the period from the appearance of ASF clinical signs until the current time step) had increased by 2; and 3) the number of SIED animals within the herd was minimum 5 (Halasa et al., 2016a). In passive surveillance, after first detection, the first two conditions were assumed to be the same as before first detection, while the minimum number of SIED animals was set to 1, to represent a higher awareness of the disease in the country (Halasa et al., 2016a). In active surveillance, detection occurred as a result of surveillance visits to infected herds by official veterinarians, either due to tracing or because the herd was located in a control zone (Halasa et al., 2016a). Herds to be surveyed were set in a queuing system, and visited as soon as resources were available. The daily surveillance capacity is dynamic over time (Table S1 in the Supplementary materials). The active surveillance includes either clinical surveillance alone (clinical signs and mortality), or clinical surveillance combined with serological and/or PCR testing, depending on the control strategy modelled. In case of clinical surveillance only, suspicion was assumed to occur if points 2 and 3 (in passive surveillance) were reached. Suspicions were then followed up by serological and/or PCR testing for confirmation of ASFV (Halasa et al., 2016a). When live animals were sampled for laboratory testing, the sample was dependent on the herd size (see details in Halasa et al., 2016a). For finishers and weaners, it was assumed that 30 animals were sampled per 500 animals in the herd. If sows were present in the herd, it was assumed that 30 of them were tested. If fewer than 30 animals were present in the herd, we assumed all animals were tested.
144
T. Halasa et al. / Veterinary Microbiology 197 (2016) 142–150
2.2.3. Modelling control strategies In total, 16 control scenarios were modelled: the basic control strategy, the basic control combined with pre-emptive depopulation in 8 different scenarios, and the basic control with intensive surveillance in 7 different scenarios. 2.2.3.1. The basic control strategy. The basic control strategy reflect the EU and Danish control measures described in the legislation and in the contingency plan for ASF (Anonymous, 2015a,b; CEC, 2002). The Danish regulations reflect the EU directive, but include a 3-day national standstill (Anonymous, 2016) as an extra control measure. The national standstill was modelled to start immediately after detection of the first infected herd in the country and involved a complete ban on all swine movements in the country for a period of 3 days. As described in the directive, all herds tested positive for ASFV infection must be culled (CEC, 2002). The maximum available resources for culling were assumed to be 4800 animals per day (Boklund et al., 2013). Furthermore, according to the directive a protection zone of minimum 3 km and a surveillance zone of minimum 10 km must be implemented surrounding the detected herds. Within these zones, movements must be restricted and herds within the zones must be surveyed for ASFV. In the model, a protection zone was created around detected herds for 50 days. Herds within the zone were simulated to be visited twice within this period. All herds included in a new zone were placed in a queuing system 2 days after the zone was created and visited as soon as resources for surveillance were available. A second visit would be performed 45 days later, after which the protection zone was lifted. It was assumed that herds were surveyed clinically during the first visit, and that samples would be collected for serological analysis during the second visit. In both zones, all herds were under movement restriction. Probabilities of effective movement restrictions are presented in Table S1 in the Supplementary materials Similarly, the surveillance zone was created around detected herds for 45 days. It was assumed that herds within the zone were visited only once before lifting the zone. Therefore, these herds were added to the queuing system to be surveyed 40 days following the creation of the zone, and only for clinical surveillance. As part of the epidemiological investigation in detected herds, contacts must be traced both backwards and forwards. Both types of tracing were modelled for a period of 3 weeks from the time of infection as described by Halasa et al. (2016a). All traced herds from animal movements and 10% of traced herds from indirect contacts were actively surveyed by clinical visits and serological and PCR testing. 2.2.3.2. The basic control plus pre-emptive depopulation. The purpose of pre-emptive depopulation was to eliminate susceptible animals that are located close to infected animals, in order to eliminate animals that potentially could have been infected and to limit disease spread. In total, 8 depopulation scenarios were modelled. Each consisted of the basic control strategies plus depopulation of neighboring swine herds in 1 or 2 km zones around detected herds. The pre-emptive depopulation was initiated either related to time from the first herd was detected (1 or 7 days after first detection) or related to the numbers of detected herds (2 or 5 detected herds). All scenarios were simulated in separate model runs. 2.2.3.3. The basic control plus intensive surveillance. The purpose of implementing intensive surveillance was faster detection of infected farms. In total, 7 scenarios were modelled with intensive surveillance. All tested scenarios consisted of the basic
control scenario, plus serological and/or PCR testing of live or dead animals in different combinations within the protection and/or surveillance zones. In a few scenarios, additional clinical surveillance was included. The scenarios were: Intensive-1) PCRtesting of all herds in the protection zone when the zone is created; Intensive-2): serological testing all herds in the surveillance zone before lifting the zone; Intensive-3) Intensive-2 plus clinical surveillance all herds in the surveillance zone when the zone is created; Intensive-4) Intensive-1 plus Intensive-2; Intensive-5) PCR testing of all herds in the surveillance and protection zone, when the zones are created, and serological testing in both zones, before the zones are lifted; Intensive-6) PCR and serological testing of 1–5 dead animals per herd per week in the protection zone; and Intensive-7) PCR and serological testing of 1–5 dead animals per herd per week from all herds in both zones. 2.3. Sensitivity analysis In Halasa et al. (2016a), a comprehensive sensitivity and robustness analysis was carried out by varying up to 3 input parameters at the same time. The influence of input parameters on model predictions was examined. It was shown that the model was robust, and only 2 parameters (transmission rate of ASF virus within a herd and the proportion of SIED animals for ASF detection) had most influence on model predictions. Therefore, we followed up on previous results by examining the impact of these parameters on the ranking of the control scenarios, in order to examine, whether our conclusions would change with changes to these parameters. Thus, all control scenarios were run with 25% increase or decrease of these parameters in separate model runs. We also examined the impact of changing the default resources for depopulation, when the pre-emptive scenarios were run, and the resources for clinical surveillance and blood sampling, when the scenarios of intensive surveillance of live animals were run. The default resources were increased or decreased by 25% for the sensitivity analysis. 2.4. Initialisation of ASFV spread All epidemics were initiated in sow herds. The selected sow herds had to have 100 animals and 3 movements per month as in previous studies (Boklund et al., 2009). There were 973 herds that fulfilled this criterion, of which 1000 sow herds (with replacement) were randomly selected as index herd. For each selected sow herd, 2 iterations were run to include potential variation from the same index herd, resulting in a total of 2000 iterations. It was shown earlier that the model would converge using this number of index herds and iterations (Halasa et al., 2016a). 2.5. Model outputs The epidemiologic outputs were: day of first detection; epidemic duration; number of infected herds; number of detected herds; number of culled animals; number of herds detected from surveillance; number of clinically examined herds; number of serologically tested herds, and number of herds tested with PCR. The economic outputs included the direct costs and the export losses and were calculated as described by Halasa et al. (2016a). All input parameters for the economic analysis are presented in Halasa et al. (2016a) together with the source. The direct costs included surveillance costs, depopulation costs, cleaning and disinfection costs, compensation costs, costs of empty stables, costs of welfare slaughter and costs of a 3 days national standstill. When preemptive depopulation was implemented, extra costs of
T. Halasa et al. / Veterinary Microbiology 197 (2016) 142–150
145
Table 1a Predicted epidemiological and economic outcomes of a simulated ASF-epidemic in Danish swine herds, when epidemics are initiated in sow herds. Values are presented as median and percentiles (5th–95th). Basic control and pre-emptive depopulation of neighboring herds. Control scenario Basic controla Depopulation Depop-1 Depop-2 Depop-3 Depop-4 Depop-5 Depop-6 Depop-7 Depop-8 a b c
Start
Zone
Day1 Day7 2 Herds 5 Herds
1 km 2 km 1 km 2 km 1 km 2 km 1 km 2 km
Epidemic durationb
Infected herds
Detected herds
Culled herds
Direct costs (s million)
Export loss (s million)
Total costs (s million)
21 (1–55)
3 (1–8)
3 (1–7)
3 (1–7)
9 (7–15)
318 (248–428)
326 (256–442)
20 (1–55) 19 (1–53)c 20 (1–55) 20 (1–55) 20 (1–55) 19 (1–53)c 21 (1–55) 20 (1–51)
3 3 3 3 3 3 3 3
3 3 3 3 3 3 3 3
4 8 4 6 4 8 3 3
9 (7–16) 10 (7–17) 9 (7–16) 10 (7–17) 9 (7–16) 10 (7–17) 9 (7–15) 9 (7–16)
316 310 317 316 316 311 318 317
326 322 326 326 326 322 326 326
(1–8) (1–8) (1–8) (1–8) (1–8) (1–8) (1–8) (1–8)
(1–7) (1–6) (1–7) (1–7) (1–7) (1–6) (1–7) (1–7)
(1–13) (1–27) (1–12) (1–23) (1–13) (1–27) (1–10) (1–18)
(248–424) (248–417) (248–424) (248–421) (248–424) (248–416) (248–424) (248–416)
(256–441) (256–433)c (256–441) (256–435) (256–441) (256–433)c (256–439) (256–428)
The basic control reflects the EU and national control. The basic control measures are included in the alternative scenarios, while additional control measures are added. The period between detection of the first infected herd and disease fade out. Statistically significantly different from the corresponding value in the basic control scenario (P-value <0.05).
depopulation and compensation for animals and empty stables were added. When diagnostic testing of dead animals (Intensive-6 or Intensive-7) was included, extra costs for submission of samples and testing were added. However, sampling costs in these two scenarios were assumed to be 0, as the collection of carcasses for testing was assumed to be carried out by the farmer. 2.6. Statistical analysis The results of the basic control strategy and the other scenarios were compared in order to test if the differences were statistically significant. Similarly, the results of the optimal control scenario were compared to the other control scenarios. The non-parametric Wilcoxon test, using the wilcox.test function in R (R Core Team, 2015), was used to test for significance of the differences between results. A p-value threshold of <5% was used to determine statistical significance.
3. Results 3.1. Epidemiological and economic outputs Tables 1a and b show epidemiological and economic results for the different control scenarios. Using the basic control strategy, which reflects the EU and national control strategy, the median epidemic duration was 21 days (5th-95th percentiles: 1–55). Furthermore, the median number of infected herds was 3 (5th– 95th percentiles: 1–8), and the median total costs was s326 million (5th-95th percentiles: s256–s442 million), mainly due to the export losses (Table 1a). Combining the basic control with pre-emptive depopulation or intensive surveillance of live animals resulted in only a marginal impact on the epidemic duration and the total costs of the epidemic (Tables 1a and b). However, there was a statistically significant reduction in epidemic duration and the total costs when scenarios of pre-emptive depopulation in 2 km was initiated after 1 day after first detection or after detection of 2 infected herds were used. Combining the basic control with pre-emptive depopulation
Table 1b Predicted epidemiological and economic outcomes of a simulated ASF-epidemic in Danish swine herds, when epidemics are initiated in sow herds. Values are presented as median and percentiles (5th–95th). Basic control and intensive surveillance in the control zones. Control scenario
Protection zoneb
Control visits Basic controla
1. Clc
Intensive-1
Epidemic durationd
Infected herds
Detected herds
Culled herds
Direct costs (s million)
Export loss (s million)
Total costs (s million)
2. Cl
21 (1–55)
3 (1–8)
3 (1–7)
3 (1–7)
9 (7–15)
+PCR +sero
Cl
20 (1–56)
3 (1–8)
3 (1–7)
3 (1–7)
9 (7–15)
Intensive-2
Cl
+sero
+sero
21 (1–55)
3 (1–8)
3 (1–7)
3 (1–7)
9 (7–15)
Intensive-3
Cl
+sero Cl
+sero
20 (1–55)
3 (1–8)
3 (1–7)
3 (1–7)
9 (7–15)
Intensive-4
+PCR +sero
+sero
20 (1–56)
3 (1–8)
3 (1–7)
3 (1–7)
10 (7–16)
Intensive-5
+PCR +sero +PCR
+sero
20 (1–55)
3 (1–8)
3 (1–7)
3 (1–7)
10 (7–18)
326 (256– 442) 326 (256– 443) 327 (256– 442) 325 (256– 438) 327 (256– 442) 326 (257–441)
3 (1–8)
3 (1–7)
3 (1–7)
9 (7–15)
318 (248– 428) 317 (249– 429) 318 (249– 428) 316 (248– 426) 317 (249– 428) 315 (249– 426) 295 (249– 401)
3 (1–8)
3 (1–7)
3 (1–7)
9 (7–15)
285 (249– 378)
294 (257– 392)e
Intensive-6
Intensive-7
a b c d e
Surveillance zoneb
2. 1. +sero
1–5 dead animals for PCR + sero 1–5 dead animals for PCR + sero
12 (1–47)
1–5 dead animals for PCR + sero
9 (1–38)e
e
304 (257– 415)e
The basic control reflects the EU and national control. The basic control measures are included in the alternative scenarios, while additional control measures are added. When samples are tested using serological (sero) or PCR, it is assumed that clinical surveillance is also performed. Cl indicates clinical surveillance. The period between detection of the first infected herd and disease fade out. Statistically significantly different from the corresponding value in the basic control scenario (P-value <0.05).
146
T. Halasa et al. / Veterinary Microbiology 197 (2016) 142–150
Fig. 1. Predicted number of depopulated animals using the different control scenarios in a simulated ASF-epidemic in Denmark. The basic control reflects the EU and national control. All other scenarios include the basic scenario plus extra control measures.
resulted in larger number of depopulated animals compared to the other control scenarios (Fig. 1). When the basic control was combined with diagnostic testing of dead animals in either the protection zone (Intensive-6) or the protection and surveillance zones (Intensive-7) a significant reduction in the epidemic duration (p-value <0.001) and the total costs of the epidemics was predicted (p-value <0.001) (Table 1b). For instance, in Intensive-7 a median epidemic duration of 9 days (5th–95th percentiles: 1–38) and a total costs of s294 million (5th–95th percentiles: s257–s392 million) was predicted. These values were significantly different from the corresponding values in the other control scenarios (p-value < 0.001 in all cases). While pre-emptive depopulation did not impact the number of herds detected from active surveillance (Table 2a), intensified surveillance of live animals potentially increase the chance to detect infected herds (Table 2b). Using the basic control strategy with added diagnostic testing of dead animals in the protection and surveillance zones (Intensive-7) resulted in a median number of herds detected from surveillance of 1 (5th and 95th percentiles 0–5) (Table 2b). Nonetheless, the intensive surveillance scenarios resulted in larger number of herd samplings for laboratory testing (serology and/or PCR) than the other control scenarios (Tables 2a and b). Diagnostic testing of dead animals in the control zones (Intensive-6 and Intensive-7) was predicted to reduce the time between infection and detection (Fig. 2). This explains the shorter
Table 2a Predicted epidemiological outcomes of a simulated ASF-epidemic in Danish swine herds, when epidemics are initiated in sow herds. Values are presented as median and percentiles (5th–95th). Basic control and pre-emptive depopulation of neighboring herds. Control scenario Basic controla Depopulation Depop-1 Depop-2 Depop-3 Depop-4 Depop-5 Depop-6 Depop-7
Start
Day1 Day7 2 Herds 5 Herds
Depop-8 a
Zone Detected herds from active surveillance
1 km 2 km 1 km 2 km 1 km
Serology tests (herd- PCR tests (herdlevel) level)
Survey dead animals (herd-level)
Clinical surveillance only (herdlevel)
0 (0–2)
19 (5–51)
3 (0–8)
0 (0–0)
110 (35–279)
0 0 0 0 0
18 18 19 18 18
3 3 3 3 3
0 0 0 0 0
(0–0) (0–0) (0–0) (0–0) (0–0)
110 (35–277) 107 (35–266) 110 (35–278) 109 (35–279) 110 (35–277)
(0–2) (0–2) (0–2) (0–2) (0–2)
(5–52) (5–51) (5–52) (5–52) (5–52)
(0–8) (0–7) (0–8) (0–7) (0–8)
2 km 0 (0–2) 1 km 0 (0–2)
18 (5–50) 19 (5–52)
3 (0–7) 3 (0–8)
0 (0–0) 0 (0–0)
107 (35–266) 110 (35–275)
2 km 0 (0–2)
19 (5–52)
3 (0–8)
0 (0–0)
110 (35–271)
The basic control reflects the EU and national control. The basic control measures are included in the alternative scenarios, while additional control measures are added.
Table 2b Predicted epidemiological outcomes of a simulated ASF-epidemic in Danish swine herds, when epidemics are initiated in sow herds. Values are presented as median and percentiles (5th–95th). Basic control and intensive surveillance in the control zones. Control scenario
Protection zoneb
Control visits Basic controla Intensive-1 Intensive-2 Intensive-3 Intensive-4 Intensive-5 Intensive-6
1. Clc
Intensive-7
a b c
Surveillance zoneb
2. 1. +sero
+PCR +sero Cl +sero Cl +sero +PCR +sero +PCR +sero 1–5 dead animals for PCR + sero 1–5 dead animals for PCR + sero
Cl +PCR
2. Cl Cl +sero +sero +sero +sero
1–5 dead animals for PCR + sero
Detected herds from active surveillance
Serology tests (herd-level)
PCR tests (herd-level)
Survey dead animals (herd-level)
Clinical surveillance only (herd-level)
0 (0–2)
19 (5–51)
3 (0–8)
0 (0–0)
110 (35–279)
0 0 1 0 1 1
18 (5–51) 112 (38–282) 112 (38–288) 111 (38–282) 111 (37–293) 19 (5–52)
19 (5–55) 3 (0–8) 3 (0–9) 19 (5–55) 119 (38–290) 3 (0–8)
0 (0–0) 0 (0–0) 0 (0–0) 0 (0–0) 0 (0–0) 11 (0–108)
93 (29–237) 17 (4–49) 118 (37–285) 0 (0–0) 0 (0–0) 110 (35–274)
19 (5–53)
3 (0–8)
72 (0–498)
109 (36–260)
(0–3) (0–2) (0–3) (03) (0–3) (0–4)
1 (0–5)
The basic control reflects the EU and national control. The basic control measures are included in the alternative scenarios, while additional control measures are added. When samples are tested using serological (sero) or PCR, it is assumed that clinical surveillance is also performed. Cl indicates clinical surveillance.
T. Halasa et al. / Veterinary Microbiology 197 (2016) 142–150
147
consistent effect on model predictions regardless of which control scenario that was simulated (results not shown). Changing resources for depopulation or resources for surveillance had marginal impact on epidemic duration and the total costs. 4. Discussion
Fig. 2. Predicted number of days between infection and detection for detected herds using the different control scenarios in a simulated ASF-epidemic in Denmark. The basic control reflects the EU and national control. All other scenarios include the basic scenario plus extra control measures.
epidemic duration and total costs in these scenarios compared to the other control scenarios (Table 1b). 3.2. Sensitivity analysis For all control scenarios, reducing the transmission rate (white boxes) is predicted to result in longer epidemic duration, while increasing it (light gray boxes) is predicted to result in shorter epidemic duration compared to the default transmission rate (Fig. 3). The total costs of the epidemic (Fig. 4) are also predicted to follow the same trend as observed for the epidemic duration. Changing the proportion of SIED animals in the herd also had
Adding diagnostic testing of dead animals in protection and/or surveillance zones to the basic control strategy (Intensive-6 and Intensive-7) reduced the epidemic duration and the total costs of the epidemics substantially compared to using the basic control alone (Table 1b). This is due to the faster detection of infected herds using these combinations compared to the other control scenarios (Fig. 2). As mentioned above, this Georgian ASFV strain is characterized by high mortality, and hence constant diagnostic testing of the dead animals in the control zones would increase the chance of early virus detection, which would shorten the epidemic duration. Because of the large export of swine and swine products, the shorter epidemic duration would reduce the export losses reducing the total costs (Table 1b). Testing of dead animals in the protection and surveillance zones (Intensive-7) is predicted to give a shorter epidemic duration and lower total losses than testing only in the protection zone (Intensive-6) (Table 1b), and therefore, control scenario Intensive-7 is predicted to be the optimal control strategy for an ASF epidemic in Denmark. Despite of the statistical difference of median costs between this optimal scenario and the other control scenarios, the relevance of this difference may be challenged by the large variation in the total costs using these scenarios. The model predictions show that combining the basic control with pre-emptive depopulation does not provide much additional benefit compared to using the basic control alone (Table 1a). In addition, this combination is predicted to result in larger numbers of depopulated animals (Fig. 1), which may raise societal concern regarding the mass culling of healthy animals. The reasons behind the failure of this combination are due to the relatively small
Fig. 3. Boxplot of the predicted epidemic duration of an ASF epidemic in Denmark by varying transmission rates of the virus within the herds and with different control scenarios. The dark gray, white and light gray boxes represent the default value, a 25% reduction and a 25% increase of the transmission rate, respectively. The black line in the boxes represents the median value, the box represents the interquartile range and the whiskers represent the range. The basic control reflects the EU and national control. All other scenarios include the basic scenario plus extra control measures.
148
T. Halasa et al. / Veterinary Microbiology 197 (2016) 142–150
Fig. 4. Boxplot of the predicted total costs of an ASF epidemic in Denmark by varying transmission rates of the virus within the herds and with different control scenarios. The dark gray, white and light gray boxes represent the default value, a 25% reduction and a 25% increase of the transmission rate, respectively. The black line in the boxes represents the median value, the box represents the interquartile range and the whiskers represent the range. The basic control reflects the EU and national control. All other scenarios include the basic scenario plus extra control measures.
epidemics predicted by the model and that the majority of the infected herds were predicted to be infected before the first detection as shown by Halasa et al. (2016a). As the purpose of preemptive depopulation is to prevent new infections by emptying the neighbored area from susceptible herds, the addition of preemptive depopulation to the basic control would not provide an extra advantage. Combining the basic control with intensive surveillance scenarios Intensive-1 to Intensive-5 (Table 1) did not show any statistical significant improvement compared to the basic control. These intensive surveillance scenarios focus on clinical surveillance and/or testing animals from herds within the protection and/ or surveillance zones. By infection of pigs with the Georgian ASFV strain, as used for this modeling study, a short period of clinical signs is normally observed, followed by the death of the vast majority of the infected animals and hence resulting in a short infectious period (Gallardo et al., 2015a,b). In addition, ASFV has been described to be moderate in contagiousness in the sense that it does not spread fast within a herd (Olsevskis et al., 2016). Thus the fast death of infected animals and the slow spread of the virus within the herd make it unlikely to detect the disease quickly by clinical surveillance and/or testing live animals within the control zones. Similarly, surveillance programs of ASFV in wild boar have shown a higher probability of detecting ASFV in dead wild boar than in live (hunted) wild boar (EFSA-Panel, 2015). Furthermore, the late surveillance visit is, with its occurrence on day 45–50, intended to assure freedom before lifting the zones, rather than to achieve early detection. Results of the sensitivity analyses have shown that consistent effect of changing the two most influential parameters on model predictions regardless the simulated control scenario of ASF (Figs. 3 and 4). Hence, the conclusion regarding the optimal control scenario (Intensive-7) is not expected to change, when the influential input values of the model are changed. This confirms the robustness of the model as also shown by Halasa et al. (2016a).
Since the introduction of ASFV to Georgia in 2007, it has spread to the Russian Federation and several Eastern EU countries, including Latvia, Estonia, Lithuania, Moldova and Poland (Gallardo et al., 2015b; Guinat et al., 2016). The control actions in the Russian Federation have been hampered by the lack of a centralized program to combat the disease, the illegal transport of animal and animal products, swill feeding and socio-economic factors (Sánchez-Vizcaíno et al., 2013). In the EU infected member states, the EU guidelines were followed from start, focusing on strict surveillance (Gallardo et al., 2015b). Following the continuous spread of the virus, stricter control strategies were added, such as disinfection of vehicles, suspension of markets, stricter biosecurity, regionalization, awareness-raising campaigns and reducing wild boar densities and limiting their capabilities to cross frontiers (Gallardo et al., 2015b). Still (until summer 2016), ASFV is being detected in both wild boar and domestic pigs in the Baltic countries and in Poland, which poses a constant risk for other countries in EU with large swine industries, such as Denmark, Germany and the Netherlands. Thus, ASF-free countries are encouraged to update their contingency plans, in order to mitigate a potential ASF epidemic in the country. The EU control regulations demand that in both protection and surveillance zones: “all dead or diseased pigs on a holding shall be immediately notified to the competent authority, which shall carry out appropriate investigations in accordance with the procedures laid down in the diagnostic manual” (CEC, 2002; art. 10e). The veterinary and food administration will consider which of these reports on diseased and dead animals will lead to suspicion of ASF and result in clinical surveillance and testing. In the model, only ASF was included, and therefore it was not possible to simulate differential diagnostic diseases. Furthermore, knowledge on how many reports on dead or sick animals would lead to investigation for ASF was not available. Therefore, suspicion based on dead animals was not modelled separately in the basic scenario, but was considered as part of the passive surveillance by the farmer. Nevertheless, it is important to show the potential effects of
T. Halasa et al. / Veterinary Microbiology 197 (2016) 142–150
sampling dead animals for disease detection. We proposed an effective and well-defined approach to sampling dead animals in the control zones: sampling up to 5 dead animals from all herds every week (Intensive-7). Mortality was modelled not only as result of ASF, but also the baseline mortality in Danish herds was modelled and used in the simulation of surveillance of dead animals. The practicalities on how the sampling of dead animals could be performed were not considered in this study, and must be addressed by the authorities and other stakeholders. Results of simulation modeling of disease spread are increasingly used for contingency planning (e.g. Garner et al., 2014). The descriptive results from the model can inform the veterinary authorities about the number of herd visits that have to be performed and samples that have to be collected during an outbreak. The veterinary authorities can use this information to estimate the necessary resources to be able to meet these demands. Still, more details on resource estimations are needed in terms of needs in different stages of the epidemic and needed capacity for different work tasks and equipment. In the current study, we do not consider the effect of wild boar on ASFV spread. It has been shown that wild boar can be involved in the spread of ASFV in certain areas (Vergne et al., 2015). Thus the results of this study should be interpreted with caution for regions were established wild boar populations exist, as the success of the control scenarios might be affected by potential spread of ASFV between wild boar and domestic pigs. Still, scenarios designed to monitor mortality within domestic pigs might be effective, as early detection is important for limiting ASFV spread (Gallardo et al., 2015b). It could be interesting to investigate the effect of monitoring mortality in domestic herds as well as in wild boar, and to examine the effect in control of ASFV spread in regions where established wild boar populations exist. In the current simulation, the Danish population of swine herds was used as an example. We highly recommend other countries to repeat this exercise using their own data, as the optimal control strategy might be different. This difference could be due to differences in herd structure, contact and movement patterns between herds, density of the population, possible presence of wild boar and export patterns of swine and swine products. For instance, our results showed that under Danish conditions export losses are the major driver of the total cost and as a consequence the epidemic duration is of particular importance. In other countries without large exports of swine and swine products, other considerations might apply, e.g. minimizing the number of pre-emptive culling. Still, the approaches and modelled processes from this work can be followed. Conflict of interest None. Acknowledgements This project was financially supported as part of an agreement of commissioned work between the Danish Ministry of Food, Agriculture and Fisheries and the Technical University of Denmark. The work was commissioned by the Danish Veterinary and Food Administration. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j. vetmic.2016.11.023.
149
References Alban, L., Andersen, M.M., Asferg, T., Boklund, A., Fernandez, N., Goldbach, S., Ydesen, B., 2005. Wildrisk: classical swine fever and wild boar in Denmark: a risk analysis. Danish Inst. for Food Vet. Res.. http://orbit.dtu.dk/files/3390946/ WILDRISK_2005.pdf. Anonymous, 2015a. Bekendtgørelse Om Bekæmpelse Af Afrikansk Svinepest, BEK Nr 1469 Af 08/12/2015 (In Danish). . https://www.retsinformation.dk/Forms/ R0710 aspx?id=176036. Anonymous, 2015b. Udbrudsmanual for Afrikansk Svinepest. , pp. 26 oktober 2015 (In Danish). http://www.foedevarestyrelsen.dk/SiteCollectionDocuments/ Dyresundhed/Sygdoms-gruppen/Udbrudsmanual%20afrikansk%20svinepest% 2026-10-2015.pdf. Anonymous, 2015c. Bekendtgørelse Af Lov Om Jagt Og Vildtforvaltning. BEK nr. 1617 af 08/12/2015 x 38 (In Danish). https://www.retsinformation.dk/forms/r0710. aspx?id=175262. Anonymous, 2016. Veterinær Beredskabsplan Ved Udbrud Af Alvorlige Husdyrsygdomme. marts 2016 (In Danish). https://www.foedevarestyrelsen. dk/SiteCollectionDocuments/Dyresundhed/Sygdoms-gruppen/Veterin%C3% A6r%20beredskabsplan%20ved%20udbrud%20af%20husdyrsygdomme%20% 20marts%202016%20-%20uden%20anneks.pdf. Backer, J.A., Hagenaars, T.J., Van Roermund, H.J.W., 2009. Modelling the effectiveness and risk of vaccination strategies to control classical swine fever epidemics. J. R. Soc. 6, 849–861. Boklund, A., Alban, L., Toft, N., Uttenthal, Å., 2009. Comparing the epidemiological and economic effects of control strategies against classical swine fever in Denmark. Prev. Vet. Med. 90, 180–193. Boklund, A., Halasa, T., Christiansen, L.E., Enøe, C., 2013. Comparing control strategies against foot-and-mouth disease: will vaccination be cost-effective in Denmark? Prev. Vet. Med. 111, 206–219. CEC, 2002. Council Directive 2002/60/EC. Council Directive 2002/60/EC of 27 June 2002 laying down specific provisions for the control of African swine fever and amending Directive 92/119/EEC as regards Teschen disease and African swine fever. Off. J. Eur. Commun. L192 (June), 27–46. Chenais, E., Sternberg-Lewerin, S., Boqvist, S., Emanuelson, U., Aliro, T., Tejler, E., Cocca, G., Masembe, C., Ståhl, K., 2015. African swine fever in Uganda: qualitative evaluation of three surveillance methods with implications for other resource-poor settings. Front. Vet. Sci. 2, 51. doi:http://dx.doi.org/10.3389/ fvets.2015.0005. EFSA Panel on Animal Health and Welfare (AHAW), 2014. Scientific opinion on African swine fever. EFSA J. 12, 3628. doi:http://dx.doi.org/10.2903/j. efsa.2014.3628. EFSA Panel on Animal Health and Welfare (AHAW), 2015. Scientific opinion on African swine fever. EFSA J. 13, 4163. doi:http://dx.doi.org/10.2903/j. efsa.2015.4163. Gallardo, M.C., Ries, A.L., Kalema-Zikusoka, G., Malta, J., Soler, A., Blanco, E., Parkhouse, R.M.E., Laitao, A., 2009. Recombinent antigen targets for serodiagnosis of African swine fever. Clin. Vac. Immunol. 16, 1012–1020. Gallardo, C., Fernandez-Pinero, J., Pelayo, V., Gazaev, I., Markowska-Daniel, I., Pridotkas, G., Nieto, R., Fernandez-Pacheco, P., Bokhan, S., Nevolko, O., Drozhzhe, Z., C, Perez, C., Soler, A., Kolvasov, D., Arias, M., 2014. Genetic variation among African swine fever genotype II viruses, eastern and central Europe. Emerg. Infect. Dis. 20, 1544–1547. Gallardo, M.C., Soler, A., Nieto, R., Cano, C., Pelayo, V., Sanchaz, M.A., Pridotkas, G., Fernandez-Pinero, J., Briones, V., Arias, M., 2015a. Experimental infection of domestic pigs with African swine fever virus Lithuania 2014 Genotype II field isolate. Transbound. Emerg. Dis. doi:http://dx.doi.org/10.1111/tbed.12346. Gallardo, M.C., de la Torre Reoyo, A., Fernández-Pinero, J., Iglasias, I., Jesús Munoz, M., Luisa Arias, M., 2015b. African swine fever: a global view of the current challenge. PHM 1, 21. doi:http://dx.doi.org/10.1186/s40813-015-0013-y. Garner, M.G., Bombarderi, N., Conzens, M., Conway, M.L., Wright, T., Paskin, R., East, I. J., 2014. Estimating resource requirement to staff a response to a medium to large outbreak of foot and mouth disease in Australia. Transbound. Emerg. Dis. doi:http://dx.doi.org/10.1111/tbed.12239. Guinat, C., Gogin, A., Blome, S., Keil, G., Pollin, R., Pfeiffer, D.U., Dixon, L., 2016. Transmission routes of African swine fever virus to domestic pigs: current knowledge and future research directions. Vet. Rec. 178, 262–267. doi:http://dx. doi.org/10.1136/vr.103593. Halasa, T., Toft, N., Boklund, A., 2015. Improving the effect and efficiency of FMD control by enlarging protection or surveillance zones. Front. Vet. Sci. 2, 70. doi: http://dx.doi.org/10.3389/fvets.2015.00070. Halasa, T., Bøtner, A., Mortensen, S., Christensen, H., Toft, N., Boklund, A., 2016a. Simulating the epidemiological and economic effects of an African swine fever epidemic in industrialized swine populations. Vet. Microbiol. 193, 7–16. Halasa, T., Boklund, A., Bøtner, A., Toft, N., Thulke, H.-H., 2016b. Simulation of spread of African swine fever including the effects of residues from dead animals. Front. Vet. Sci. 3, 6. doi:http://dx.doi.org/10.3389/fvets.2016.00006. Jordt, A.M., Lange, M., Kramer-Schadt, S., Nielsen, L.H., Nielsen, S.S., Thulke, H.H., Vejre, H., Alban, L., 2016. Spatio-temporal modeling of the invasive potential of wild-boar: a conflict-prone species-using multi-source citizen science data. Prev. Vet. Med. 124, 34–44. Martínez-López, B., Ivorra, B., Ramos, A.M., Sanchez-Vizcaino, J.M., 2011. A novel spatial and stochastic model to evaluate the within- and between-farm transmission of classical swine fever virus. I. General concepts and description of the model. Vet. Microbiol. 147, 300–309.
150
T. Halasa et al. / Veterinary Microbiology 197 (2016) 142–150
Olsevskis, E., Guberti, V., Serzants, M., Westergaard, J., Gallardo, C., Rodze, I., Depner, K., 2016. African swine fever virus introduction into the EU in 2014: experience of Latvia. Res. Vet. Sci. 105, 28–30. R Core Team, 2015. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project. org/.
Sánchez-Vizcaíno, J.M., Mur, L., Martínez-López, B., 2013. African swine fever (ASF): five years around europe. Vet. Microbiol. 165, 45–50. Vergne, T., Gogin, A., Pfeiffer, D., 2015. Statistical exploration of local transmission routes for african swine fever in pigs in the Russian Federation. Transbound. Emerg. Dis. 2007–2014. doi:http://dx.doi.org/10.1111/tbed.12391.
Contents lists available at ScienceDirect
Veterinary Microbiology journal homepage: www.elsevier.com/locate/vetmic
Control of African swine fever epidemics in industrialized swine populations Tariq Halasaa,* , Anette Bøtnera , Sten Mortensenb , Hanne Christensenb , Nils Tofta , Anette Boklunda a b
National Veterinary Institute, Technical University of Denmark, Copenhagen, Denmark Danish Veterinary and Food Administration, Ministry of Environment and Food, Glostrup, Denmark
A R T I C L E I N F O
Article history: Received 30 August 2016 Received in revised form 11 November 2016 Accepted 14 November 2016 Keywords: African swine fever Simulation model Spread Control
A B S T R A C T
African swine fever (ASF) is a notifiable infectious disease with a high impact on swine health. The disease is endemic in certain regions in the Baltic countries and has spread to Poland constituting a risk of ASF spread toward Western Europe. Therefore, as part of contingency planning, it is important to explore strategies that can effectively control an epidemic of ASF. In this study, the epidemiological and economic effects of strategies to control the spread of ASF between domestic swine herds were examined using a published model (DTU-DADS-ASF). The control strategies were the basic EU and national strategy (Basic), the basic strategy plus pre-emptive depopulation of neighboring swine herds, and intensive surveillance of herds in the control zones, including testing live or dead animals. Virus spread via wild boar was not modelled. Under the basic control strategy, the median epidemic duration was predicted to be 21 days (5th and 95th percentiles; 1-55 days), the median number of infected herds was predicted to be 3 herds (1–8), and the total costs were predicted to be s326 million (s256–s442 million). Adding pre-emptive depopulation or intensive surveillance by testing live animals resulted in marginal improvements to the control of the epidemics. However, adding testing of dead animals in the protection and surveillance zones was predicted to be the optimal control scenario for an ASF epidemic in industrialized swine populations without contact to wild boar. This optimal scenario reduced the epidemic duration to 9 days (1–38) and the total costs to s294 million (s257–s392 million). Export losses were the driving force of the total costs of the epidemics. ã 2016 Elsevier B.V. All rights reserved.
1. Introduction African swine fever (ASF) is a notifiable infectious disease in pigs. It is caused by ASF virus (ASFV), a DNA virus from the family Asfarviridae, genus Asfivirus (cited from Gallardo et al., 2009). The disease is endemic in Africa (Chenais et al., 2015), the Russian Federation, and in certain regions of the Baltic countries (Gallardo et al., 2014, 2015b; Olsevskis et al., 2016). It is considered to be a substantial threat for Western Europe (EFSA-Panel, 2014). In countries with a large production and/or export of swine and swine products, an outbreak of ASF may result in devastating economic consequences for the swine industry due to export restrictions. Therefore, it is important to explore the effectiveness
* Corresponding author at: National Veterinary Institute, Technical University of Denmark, Bülowsvej 27, 1870 Frederiksberg C, Copenhagen, Denmark. E-mail address: [email protected] (T. Halasa). http://dx.doi.org/10.1016/j.vetmic.2016.11.023 0378-1135/ã 2016 Elsevier B.V. All rights reserved.
and consequences of strategies to control outbreaks of ASF in the industrialized swine populations. The EU has established a set of strategies that should be followed in the case of an outbreak of ASF in the domestic swine populations (CEC, 2002). To our knowledge, the effectiveness of these strategies and their combination with other strategies, such as pre-emptive depopulation of neighboring swine herds or intensive surveillance in the control zones has never been investigated before. Identifying effective control strategies will assist the national veterinary authorities in the development of national guidelines for ASF control and contingency planning. Simulation models of disease spread is a widely used tool to assist the national veterinary authorities in contingency planning (e.g. Backer et al., 2009; Martínez-López et al., 2011; Boklund et al., 2013; Halasa et al., 2015). Such models are invaluable for exploring mechanisms of disease spread and control, taking into account the complexities of agricultural systems.
T. Halasa et al. / Veterinary Microbiology 197 (2016) 142–150
The objective of this study was to compare the epidemiological and economic effectiveness of different strategies to control a hypothetical epidemic of ASF in industrialized swine populations, where the role of wild boar in ASFV spread is negligible, exemplified by the Danish swine population. 2. Materials and methods This section describes the simulation of ASFV spread between swine herds using different control strategies. It provides information about the herd data used in the model, ASFV-spread mechanisms, ASFV detection and the control strategies that are simulated. Important input parameters are presented in Table S1 in the Supplementary materials. Details on model parameters and equations can be found in Halasa et al. (2016a). 2.1. Herd and movement data We used geographical data (UTM coordinates), the number of animals, and specification of herd types for the 8262 swine herds registered in the Danish Central Husbandry Register (CHR) in 2014. Descriptive data on the number of herds, herd sizes and frequencies of outgoing movements was presented in Halasa et al. (2016a). For each herd, the daily frequency of moving animals to another herd was calculated as the sum of all registered herd movements in the period from 01 January 2014 to 31 December 2014 divided by 365, based on the registrations in the Pig Movement Database. For each herd, this frequency was used as the mean (l) in a Poisson distribution, describing the number of daily outgoing movements (batches) of animals. Similarly, the probability of moving animals to an abattoir was calculated for each herd. By analyzing the distances between source herds and receiving herds, it was found that animals from nucleus herds were moved further than animals moved from other herd types. As a result, two separate distributions for movement distances were used to model the movements of animals from nucleus herds and from other herd types, respectively (Halasa et al., 2016a). The probabilities of animal movements from one herd type to another were calculated based on the Pig Movement Database and are presented in Halasa et al. (2016a). 2.2. The simulation model The DTU-DADS-ASF model (version 0.15.1) (Halasa et al., 2016a) was used. The model runs in the statistical computing language R (version 3.1.3) (R Core Team, 2015). The parameterization of the model was based on the Georgian strain of ASFV, reflecting the original strain of the epidemics currently running in Eastern Europe as explained earlier (Halasa et al., 2016a). 2.2.1. Modelling ASF spread ASFV spread was modelled in two processes: 1) spread within a herd; 2) spread between herds by different mechanisms (transmission routes). 2.2.1.1. Modelling ASFV spread within a herd. The infection model for the individual animals is a state transition model with the following states: susceptible-latent-subclinical-clinical-removed (SLSCR model; Halasa et al., 2016b). The infection model for the herd follows the same model, but includes the possibility of an infected herd to become susceptible again, should the infection fade out before all animals in the herd are infected. Infected herds will start as latent and progress to the subclinical and clinical states following infection. The infection is then either detected, and therefore the herd is removed (culled), or it becomes susceptible again.
143
2.2.1.2. Modelling ASFV spread between herds. ASFV is simulated to spread between herds via animal movements, abattoir movements, via indirect medium-risk contacts (direct contact to animals such as contacts by veterinarians or artificial inseminators) or low-risk contacts (no direct contacts to animals, such as feed trucks and visitors), or via local spread. Each type of contact was modelled as a Poisson distribution. For movements of animals to other herds or to abattoirs, the mean (l) was calculated for the individual herd, as described above, while for indirect medium and low risk contacts, a Poisson distribution was modelled for each herd type. For animal movements, each movement represented a batch of animals moved from the sending infectious herd to the receiving herd. The probability of transmitting the ASFV from the infectious herd to the receiving herd was dependent on the prevalence of the disease within the infectious herd and the number of animals moved in the batch (Halasa et al., 2016a). Local spread was modelled in a distance up to 2 km around infectious herds, and was assumed to consist of a mixture of unregistered animal movements, shared equipment and tools, and spread via rodents and insects. Detailed information including the equations and steps for modeling each of these mechanisms can be found in Halasa et al. (2016a). The risk of ASFV spread and/or maintenance through wild boar was not modelled as the number of wild boar in Denmark is limited due to intensive farming in the country, leaving few suitable habitats for wild boar (Alban et al., 2005; Jordt et al., 2016). There is also a Danish legal requirement to eliminate stray wild boar (Anonymous, 2015c). 2.2.2. Modelling ASFV detection In the model, the ASFV infection can be detected by three different mechanisms: passive surveillance before first detection; passive surveillance after first detection; and active surveillance. In passive surveillance, before first detection, detection was modelled to occur when 1) the proportion of sick or dead animals (referred to as SIED throughout the paper) reached 2.55% (Halasa et al., 2016a); and 2) the proportion of SIED animals relative to the expected cumulative mortality level within the herd (in the period from the appearance of ASF clinical signs until the current time step) had increased by 2; and 3) the number of SIED animals within the herd was minimum 5 (Halasa et al., 2016a). In passive surveillance, after first detection, the first two conditions were assumed to be the same as before first detection, while the minimum number of SIED animals was set to 1, to represent a higher awareness of the disease in the country (Halasa et al., 2016a). In active surveillance, detection occurred as a result of surveillance visits to infected herds by official veterinarians, either due to tracing or because the herd was located in a control zone (Halasa et al., 2016a). Herds to be surveyed were set in a queuing system, and visited as soon as resources were available. The daily surveillance capacity is dynamic over time (Table S1 in the Supplementary materials). The active surveillance includes either clinical surveillance alone (clinical signs and mortality), or clinical surveillance combined with serological and/or PCR testing, depending on the control strategy modelled. In case of clinical surveillance only, suspicion was assumed to occur if points 2 and 3 (in passive surveillance) were reached. Suspicions were then followed up by serological and/or PCR testing for confirmation of ASFV (Halasa et al., 2016a). When live animals were sampled for laboratory testing, the sample was dependent on the herd size (see details in Halasa et al., 2016a). For finishers and weaners, it was assumed that 30 animals were sampled per 500 animals in the herd. If sows were present in the herd, it was assumed that 30 of them were tested. If fewer than 30 animals were present in the herd, we assumed all animals were tested.
144
T. Halasa et al. / Veterinary Microbiology 197 (2016) 142–150
2.2.3. Modelling control strategies In total, 16 control scenarios were modelled: the basic control strategy, the basic control combined with pre-emptive depopulation in 8 different scenarios, and the basic control with intensive surveillance in 7 different scenarios. 2.2.3.1. The basic control strategy. The basic control strategy reflect the EU and Danish control measures described in the legislation and in the contingency plan for ASF (Anonymous, 2015a,b; CEC, 2002). The Danish regulations reflect the EU directive, but include a 3-day national standstill (Anonymous, 2016) as an extra control measure. The national standstill was modelled to start immediately after detection of the first infected herd in the country and involved a complete ban on all swine movements in the country for a period of 3 days. As described in the directive, all herds tested positive for ASFV infection must be culled (CEC, 2002). The maximum available resources for culling were assumed to be 4800 animals per day (Boklund et al., 2013). Furthermore, according to the directive a protection zone of minimum 3 km and a surveillance zone of minimum 10 km must be implemented surrounding the detected herds. Within these zones, movements must be restricted and herds within the zones must be surveyed for ASFV. In the model, a protection zone was created around detected herds for 50 days. Herds within the zone were simulated to be visited twice within this period. All herds included in a new zone were placed in a queuing system 2 days after the zone was created and visited as soon as resources for surveillance were available. A second visit would be performed 45 days later, after which the protection zone was lifted. It was assumed that herds were surveyed clinically during the first visit, and that samples would be collected for serological analysis during the second visit. In both zones, all herds were under movement restriction. Probabilities of effective movement restrictions are presented in Table S1 in the Supplementary materials Similarly, the surveillance zone was created around detected herds for 45 days. It was assumed that herds within the zone were visited only once before lifting the zone. Therefore, these herds were added to the queuing system to be surveyed 40 days following the creation of the zone, and only for clinical surveillance. As part of the epidemiological investigation in detected herds, contacts must be traced both backwards and forwards. Both types of tracing were modelled for a period of 3 weeks from the time of infection as described by Halasa et al. (2016a). All traced herds from animal movements and 10% of traced herds from indirect contacts were actively surveyed by clinical visits and serological and PCR testing. 2.2.3.2. The basic control plus pre-emptive depopulation. The purpose of pre-emptive depopulation was to eliminate susceptible animals that are located close to infected animals, in order to eliminate animals that potentially could have been infected and to limit disease spread. In total, 8 depopulation scenarios were modelled. Each consisted of the basic control strategies plus depopulation of neighboring swine herds in 1 or 2 km zones around detected herds. The pre-emptive depopulation was initiated either related to time from the first herd was detected (1 or 7 days after first detection) or related to the numbers of detected herds (2 or 5 detected herds). All scenarios were simulated in separate model runs. 2.2.3.3. The basic control plus intensive surveillance. The purpose of implementing intensive surveillance was faster detection of infected farms. In total, 7 scenarios were modelled with intensive surveillance. All tested scenarios consisted of the basic
control scenario, plus serological and/or PCR testing of live or dead animals in different combinations within the protection and/or surveillance zones. In a few scenarios, additional clinical surveillance was included. The scenarios were: Intensive-1) PCRtesting of all herds in the protection zone when the zone is created; Intensive-2): serological testing all herds in the surveillance zone before lifting the zone; Intensive-3) Intensive-2 plus clinical surveillance all herds in the surveillance zone when the zone is created; Intensive-4) Intensive-1 plus Intensive-2; Intensive-5) PCR testing of all herds in the surveillance and protection zone, when the zones are created, and serological testing in both zones, before the zones are lifted; Intensive-6) PCR and serological testing of 1–5 dead animals per herd per week in the protection zone; and Intensive-7) PCR and serological testing of 1–5 dead animals per herd per week from all herds in both zones. 2.3. Sensitivity analysis In Halasa et al. (2016a), a comprehensive sensitivity and robustness analysis was carried out by varying up to 3 input parameters at the same time. The influence of input parameters on model predictions was examined. It was shown that the model was robust, and only 2 parameters (transmission rate of ASF virus within a herd and the proportion of SIED animals for ASF detection) had most influence on model predictions. Therefore, we followed up on previous results by examining the impact of these parameters on the ranking of the control scenarios, in order to examine, whether our conclusions would change with changes to these parameters. Thus, all control scenarios were run with 25% increase or decrease of these parameters in separate model runs. We also examined the impact of changing the default resources for depopulation, when the pre-emptive scenarios were run, and the resources for clinical surveillance and blood sampling, when the scenarios of intensive surveillance of live animals were run. The default resources were increased or decreased by 25% for the sensitivity analysis. 2.4. Initialisation of ASFV spread All epidemics were initiated in sow herds. The selected sow herds had to have 100 animals and 3 movements per month as in previous studies (Boklund et al., 2009). There were 973 herds that fulfilled this criterion, of which 1000 sow herds (with replacement) were randomly selected as index herd. For each selected sow herd, 2 iterations were run to include potential variation from the same index herd, resulting in a total of 2000 iterations. It was shown earlier that the model would converge using this number of index herds and iterations (Halasa et al., 2016a). 2.5. Model outputs The epidemiologic outputs were: day of first detection; epidemic duration; number of infected herds; number of detected herds; number of culled animals; number of herds detected from surveillance; number of clinically examined herds; number of serologically tested herds, and number of herds tested with PCR. The economic outputs included the direct costs and the export losses and were calculated as described by Halasa et al. (2016a). All input parameters for the economic analysis are presented in Halasa et al. (2016a) together with the source. The direct costs included surveillance costs, depopulation costs, cleaning and disinfection costs, compensation costs, costs of empty stables, costs of welfare slaughter and costs of a 3 days national standstill. When preemptive depopulation was implemented, extra costs of
T. Halasa et al. / Veterinary Microbiology 197 (2016) 142–150
145
Table 1a Predicted epidemiological and economic outcomes of a simulated ASF-epidemic in Danish swine herds, when epidemics are initiated in sow herds. Values are presented as median and percentiles (5th–95th). Basic control and pre-emptive depopulation of neighboring herds. Control scenario Basic controla Depopulation Depop-1 Depop-2 Depop-3 Depop-4 Depop-5 Depop-6 Depop-7 Depop-8 a b c
Start
Zone
Day1 Day7 2 Herds 5 Herds
1 km 2 km 1 km 2 km 1 km 2 km 1 km 2 km
Epidemic durationb
Infected herds
Detected herds
Culled herds
Direct costs (s million)
Export loss (s million)
Total costs (s million)
21 (1–55)
3 (1–8)
3 (1–7)
3 (1–7)
9 (7–15)
318 (248–428)
326 (256–442)
20 (1–55) 19 (1–53)c 20 (1–55) 20 (1–55) 20 (1–55) 19 (1–53)c 21 (1–55) 20 (1–51)
3 3 3 3 3 3 3 3
3 3 3 3 3 3 3 3
4 8 4 6 4 8 3 3
9 (7–16) 10 (7–17) 9 (7–16) 10 (7–17) 9 (7–16) 10 (7–17) 9 (7–15) 9 (7–16)
316 310 317 316 316 311 318 317
326 322 326 326 326 322 326 326
(1–8) (1–8) (1–8) (1–8) (1–8) (1–8) (1–8) (1–8)
(1–7) (1–6) (1–7) (1–7) (1–7) (1–6) (1–7) (1–7)
(1–13) (1–27) (1–12) (1–23) (1–13) (1–27) (1–10) (1–18)
(248–424) (248–417) (248–424) (248–421) (248–424) (248–416) (248–424) (248–416)
(256–441) (256–433)c (256–441) (256–435) (256–441) (256–433)c (256–439) (256–428)
The basic control reflects the EU and national control. The basic control measures are included in the alternative scenarios, while additional control measures are added. The period between detection of the first infected herd and disease fade out. Statistically significantly different from the corresponding value in the basic control scenario (P-value <0.05).
depopulation and compensation for animals and empty stables were added. When diagnostic testing of dead animals (Intensive-6 or Intensive-7) was included, extra costs for submission of samples and testing were added. However, sampling costs in these two scenarios were assumed to be 0, as the collection of carcasses for testing was assumed to be carried out by the farmer. 2.6. Statistical analysis The results of the basic control strategy and the other scenarios were compared in order to test if the differences were statistically significant. Similarly, the results of the optimal control scenario were compared to the other control scenarios. The non-parametric Wilcoxon test, using the wilcox.test function in R (R Core Team, 2015), was used to test for significance of the differences between results. A p-value threshold of <5% was used to determine statistical significance.
3. Results 3.1. Epidemiological and economic outputs Tables 1a and b show epidemiological and economic results for the different control scenarios. Using the basic control strategy, which reflects the EU and national control strategy, the median epidemic duration was 21 days (5th-95th percentiles: 1–55). Furthermore, the median number of infected herds was 3 (5th– 95th percentiles: 1–8), and the median total costs was s326 million (5th-95th percentiles: s256–s442 million), mainly due to the export losses (Table 1a). Combining the basic control with pre-emptive depopulation or intensive surveillance of live animals resulted in only a marginal impact on the epidemic duration and the total costs of the epidemic (Tables 1a and b). However, there was a statistically significant reduction in epidemic duration and the total costs when scenarios of pre-emptive depopulation in 2 km was initiated after 1 day after first detection or after detection of 2 infected herds were used. Combining the basic control with pre-emptive depopulation
Table 1b Predicted epidemiological and economic outcomes of a simulated ASF-epidemic in Danish swine herds, when epidemics are initiated in sow herds. Values are presented as median and percentiles (5th–95th). Basic control and intensive surveillance in the control zones. Control scenario
Protection zoneb
Control visits Basic controla
1. Clc
Intensive-1
Epidemic durationd
Infected herds
Detected herds
Culled herds
Direct costs (s million)
Export loss (s million)
Total costs (s million)
2. Cl
21 (1–55)
3 (1–8)
3 (1–7)
3 (1–7)
9 (7–15)
+PCR +sero
Cl
20 (1–56)
3 (1–8)
3 (1–7)
3 (1–7)
9 (7–15)
Intensive-2
Cl
+sero
+sero
21 (1–55)
3 (1–8)
3 (1–7)
3 (1–7)
9 (7–15)
Intensive-3
Cl
+sero Cl
+sero
20 (1–55)
3 (1–8)
3 (1–7)
3 (1–7)
9 (7–15)
Intensive-4
+PCR +sero
+sero
20 (1–56)
3 (1–8)
3 (1–7)
3 (1–7)
10 (7–16)
Intensive-5
+PCR +sero +PCR
+sero
20 (1–55)
3 (1–8)
3 (1–7)
3 (1–7)
10 (7–18)
326 (256– 442) 326 (256– 443) 327 (256– 442) 325 (256– 438) 327 (256– 442) 326 (257–441)
3 (1–8)
3 (1–7)
3 (1–7)
9 (7–15)
318 (248– 428) 317 (249– 429) 318 (249– 428) 316 (248– 426) 317 (249– 428) 315 (249– 426) 295 (249– 401)
3 (1–8)
3 (1–7)
3 (1–7)
9 (7–15)
285 (249– 378)
294 (257– 392)e
Intensive-6
Intensive-7
a b c d e
Surveillance zoneb
2. 1. +sero
1–5 dead animals for PCR + sero 1–5 dead animals for PCR + sero
12 (1–47)
1–5 dead animals for PCR + sero
9 (1–38)e
e
304 (257– 415)e
The basic control reflects the EU and national control. The basic control measures are included in the alternative scenarios, while additional control measures are added. When samples are tested using serological (sero) or PCR, it is assumed that clinical surveillance is also performed. Cl indicates clinical surveillance. The period between detection of the first infected herd and disease fade out. Statistically significantly different from the corresponding value in the basic control scenario (P-value <0.05).
146
T. Halasa et al. / Veterinary Microbiology 197 (2016) 142–150
Fig. 1. Predicted number of depopulated animals using the different control scenarios in a simulated ASF-epidemic in Denmark. The basic control reflects the EU and national control. All other scenarios include the basic scenario plus extra control measures.
resulted in larger number of depopulated animals compared to the other control scenarios (Fig. 1). When the basic control was combined with diagnostic testing of dead animals in either the protection zone (Intensive-6) or the protection and surveillance zones (Intensive-7) a significant reduction in the epidemic duration (p-value <0.001) and the total costs of the epidemics was predicted (p-value <0.001) (Table 1b). For instance, in Intensive-7 a median epidemic duration of 9 days (5th–95th percentiles: 1–38) and a total costs of s294 million (5th–95th percentiles: s257–s392 million) was predicted. These values were significantly different from the corresponding values in the other control scenarios (p-value < 0.001 in all cases). While pre-emptive depopulation did not impact the number of herds detected from active surveillance (Table 2a), intensified surveillance of live animals potentially increase the chance to detect infected herds (Table 2b). Using the basic control strategy with added diagnostic testing of dead animals in the protection and surveillance zones (Intensive-7) resulted in a median number of herds detected from surveillance of 1 (5th and 95th percentiles 0–5) (Table 2b). Nonetheless, the intensive surveillance scenarios resulted in larger number of herd samplings for laboratory testing (serology and/or PCR) than the other control scenarios (Tables 2a and b). Diagnostic testing of dead animals in the control zones (Intensive-6 and Intensive-7) was predicted to reduce the time between infection and detection (Fig. 2). This explains the shorter
Table 2a Predicted epidemiological outcomes of a simulated ASF-epidemic in Danish swine herds, when epidemics are initiated in sow herds. Values are presented as median and percentiles (5th–95th). Basic control and pre-emptive depopulation of neighboring herds. Control scenario Basic controla Depopulation Depop-1 Depop-2 Depop-3 Depop-4 Depop-5 Depop-6 Depop-7
Start
Day1 Day7 2 Herds 5 Herds
Depop-8 a
Zone Detected herds from active surveillance
1 km 2 km 1 km 2 km 1 km
Serology tests (herd- PCR tests (herdlevel) level)
Survey dead animals (herd-level)
Clinical surveillance only (herdlevel)
0 (0–2)
19 (5–51)
3 (0–8)
0 (0–0)
110 (35–279)
0 0 0 0 0
18 18 19 18 18
3 3 3 3 3
0 0 0 0 0
(0–0) (0–0) (0–0) (0–0) (0–0)
110 (35–277) 107 (35–266) 110 (35–278) 109 (35–279) 110 (35–277)
(0–2) (0–2) (0–2) (0–2) (0–2)
(5–52) (5–51) (5–52) (5–52) (5–52)
(0–8) (0–7) (0–8) (0–7) (0–8)
2 km 0 (0–2) 1 km 0 (0–2)
18 (5–50) 19 (5–52)
3 (0–7) 3 (0–8)
0 (0–0) 0 (0–0)
107 (35–266) 110 (35–275)
2 km 0 (0–2)
19 (5–52)
3 (0–8)
0 (0–0)
110 (35–271)
The basic control reflects the EU and national control. The basic control measures are included in the alternative scenarios, while additional control measures are added.
Table 2b Predicted epidemiological outcomes of a simulated ASF-epidemic in Danish swine herds, when epidemics are initiated in sow herds. Values are presented as median and percentiles (5th–95th). Basic control and intensive surveillance in the control zones. Control scenario
Protection zoneb
Control visits Basic controla Intensive-1 Intensive-2 Intensive-3 Intensive-4 Intensive-5 Intensive-6
1. Clc
Intensive-7
a b c
Surveillance zoneb
2. 1. +sero
+PCR +sero Cl +sero Cl +sero +PCR +sero +PCR +sero 1–5 dead animals for PCR + sero 1–5 dead animals for PCR + sero
Cl +PCR
2. Cl Cl +sero +sero +sero +sero
1–5 dead animals for PCR + sero
Detected herds from active surveillance
Serology tests (herd-level)
PCR tests (herd-level)
Survey dead animals (herd-level)
Clinical surveillance only (herd-level)
0 (0–2)
19 (5–51)
3 (0–8)
0 (0–0)
110 (35–279)
0 0 1 0 1 1
18 (5–51) 112 (38–282) 112 (38–288) 111 (38–282) 111 (37–293) 19 (5–52)
19 (5–55) 3 (0–8) 3 (0–9) 19 (5–55) 119 (38–290) 3 (0–8)
0 (0–0) 0 (0–0) 0 (0–0) 0 (0–0) 0 (0–0) 11 (0–108)
93 (29–237) 17 (4–49) 118 (37–285) 0 (0–0) 0 (0–0) 110 (35–274)
19 (5–53)
3 (0–8)
72 (0–498)
109 (36–260)
(0–3) (0–2) (0–3) (03) (0–3) (0–4)
1 (0–5)
The basic control reflects the EU and national control. The basic control measures are included in the alternative scenarios, while additional control measures are added. When samples are tested using serological (sero) or PCR, it is assumed that clinical surveillance is also performed. Cl indicates clinical surveillance.
T. Halasa et al. / Veterinary Microbiology 197 (2016) 142–150
147
consistent effect on model predictions regardless of which control scenario that was simulated (results not shown). Changing resources for depopulation or resources for surveillance had marginal impact on epidemic duration and the total costs. 4. Discussion
Fig. 2. Predicted number of days between infection and detection for detected herds using the different control scenarios in a simulated ASF-epidemic in Denmark. The basic control reflects the EU and national control. All other scenarios include the basic scenario plus extra control measures.
epidemic duration and total costs in these scenarios compared to the other control scenarios (Table 1b). 3.2. Sensitivity analysis For all control scenarios, reducing the transmission rate (white boxes) is predicted to result in longer epidemic duration, while increasing it (light gray boxes) is predicted to result in shorter epidemic duration compared to the default transmission rate (Fig. 3). The total costs of the epidemic (Fig. 4) are also predicted to follow the same trend as observed for the epidemic duration. Changing the proportion of SIED animals in the herd also had
Adding diagnostic testing of dead animals in protection and/or surveillance zones to the basic control strategy (Intensive-6 and Intensive-7) reduced the epidemic duration and the total costs of the epidemics substantially compared to using the basic control alone (Table 1b). This is due to the faster detection of infected herds using these combinations compared to the other control scenarios (Fig. 2). As mentioned above, this Georgian ASFV strain is characterized by high mortality, and hence constant diagnostic testing of the dead animals in the control zones would increase the chance of early virus detection, which would shorten the epidemic duration. Because of the large export of swine and swine products, the shorter epidemic duration would reduce the export losses reducing the total costs (Table 1b). Testing of dead animals in the protection and surveillance zones (Intensive-7) is predicted to give a shorter epidemic duration and lower total losses than testing only in the protection zone (Intensive-6) (Table 1b), and therefore, control scenario Intensive-7 is predicted to be the optimal control strategy for an ASF epidemic in Denmark. Despite of the statistical difference of median costs between this optimal scenario and the other control scenarios, the relevance of this difference may be challenged by the large variation in the total costs using these scenarios. The model predictions show that combining the basic control with pre-emptive depopulation does not provide much additional benefit compared to using the basic control alone (Table 1a). In addition, this combination is predicted to result in larger numbers of depopulated animals (Fig. 1), which may raise societal concern regarding the mass culling of healthy animals. The reasons behind the failure of this combination are due to the relatively small
Fig. 3. Boxplot of the predicted epidemic duration of an ASF epidemic in Denmark by varying transmission rates of the virus within the herds and with different control scenarios. The dark gray, white and light gray boxes represent the default value, a 25% reduction and a 25% increase of the transmission rate, respectively. The black line in the boxes represents the median value, the box represents the interquartile range and the whiskers represent the range. The basic control reflects the EU and national control. All other scenarios include the basic scenario plus extra control measures.
148
T. Halasa et al. / Veterinary Microbiology 197 (2016) 142–150
Fig. 4. Boxplot of the predicted total costs of an ASF epidemic in Denmark by varying transmission rates of the virus within the herds and with different control scenarios. The dark gray, white and light gray boxes represent the default value, a 25% reduction and a 25% increase of the transmission rate, respectively. The black line in the boxes represents the median value, the box represents the interquartile range and the whiskers represent the range. The basic control reflects the EU and national control. All other scenarios include the basic scenario plus extra control measures.
epidemics predicted by the model and that the majority of the infected herds were predicted to be infected before the first detection as shown by Halasa et al. (2016a). As the purpose of preemptive depopulation is to prevent new infections by emptying the neighbored area from susceptible herds, the addition of preemptive depopulation to the basic control would not provide an extra advantage. Combining the basic control with intensive surveillance scenarios Intensive-1 to Intensive-5 (Table 1) did not show any statistical significant improvement compared to the basic control. These intensive surveillance scenarios focus on clinical surveillance and/or testing animals from herds within the protection and/ or surveillance zones. By infection of pigs with the Georgian ASFV strain, as used for this modeling study, a short period of clinical signs is normally observed, followed by the death of the vast majority of the infected animals and hence resulting in a short infectious period (Gallardo et al., 2015a,b). In addition, ASFV has been described to be moderate in contagiousness in the sense that it does not spread fast within a herd (Olsevskis et al., 2016). Thus the fast death of infected animals and the slow spread of the virus within the herd make it unlikely to detect the disease quickly by clinical surveillance and/or testing live animals within the control zones. Similarly, surveillance programs of ASFV in wild boar have shown a higher probability of detecting ASFV in dead wild boar than in live (hunted) wild boar (EFSA-Panel, 2015). Furthermore, the late surveillance visit is, with its occurrence on day 45–50, intended to assure freedom before lifting the zones, rather than to achieve early detection. Results of the sensitivity analyses have shown that consistent effect of changing the two most influential parameters on model predictions regardless the simulated control scenario of ASF (Figs. 3 and 4). Hence, the conclusion regarding the optimal control scenario (Intensive-7) is not expected to change, when the influential input values of the model are changed. This confirms the robustness of the model as also shown by Halasa et al. (2016a).
Since the introduction of ASFV to Georgia in 2007, it has spread to the Russian Federation and several Eastern EU countries, including Latvia, Estonia, Lithuania, Moldova and Poland (Gallardo et al., 2015b; Guinat et al., 2016). The control actions in the Russian Federation have been hampered by the lack of a centralized program to combat the disease, the illegal transport of animal and animal products, swill feeding and socio-economic factors (Sánchez-Vizcaíno et al., 2013). In the EU infected member states, the EU guidelines were followed from start, focusing on strict surveillance (Gallardo et al., 2015b). Following the continuous spread of the virus, stricter control strategies were added, such as disinfection of vehicles, suspension of markets, stricter biosecurity, regionalization, awareness-raising campaigns and reducing wild boar densities and limiting their capabilities to cross frontiers (Gallardo et al., 2015b). Still (until summer 2016), ASFV is being detected in both wild boar and domestic pigs in the Baltic countries and in Poland, which poses a constant risk for other countries in EU with large swine industries, such as Denmark, Germany and the Netherlands. Thus, ASF-free countries are encouraged to update their contingency plans, in order to mitigate a potential ASF epidemic in the country. The EU control regulations demand that in both protection and surveillance zones: “all dead or diseased pigs on a holding shall be immediately notified to the competent authority, which shall carry out appropriate investigations in accordance with the procedures laid down in the diagnostic manual” (CEC, 2002; art. 10e). The veterinary and food administration will consider which of these reports on diseased and dead animals will lead to suspicion of ASF and result in clinical surveillance and testing. In the model, only ASF was included, and therefore it was not possible to simulate differential diagnostic diseases. Furthermore, knowledge on how many reports on dead or sick animals would lead to investigation for ASF was not available. Therefore, suspicion based on dead animals was not modelled separately in the basic scenario, but was considered as part of the passive surveillance by the farmer. Nevertheless, it is important to show the potential effects of
T. Halasa et al. / Veterinary Microbiology 197 (2016) 142–150
sampling dead animals for disease detection. We proposed an effective and well-defined approach to sampling dead animals in the control zones: sampling up to 5 dead animals from all herds every week (Intensive-7). Mortality was modelled not only as result of ASF, but also the baseline mortality in Danish herds was modelled and used in the simulation of surveillance of dead animals. The practicalities on how the sampling of dead animals could be performed were not considered in this study, and must be addressed by the authorities and other stakeholders. Results of simulation modeling of disease spread are increasingly used for contingency planning (e.g. Garner et al., 2014). The descriptive results from the model can inform the veterinary authorities about the number of herd visits that have to be performed and samples that have to be collected during an outbreak. The veterinary authorities can use this information to estimate the necessary resources to be able to meet these demands. Still, more details on resource estimations are needed in terms of needs in different stages of the epidemic and needed capacity for different work tasks and equipment. In the current study, we do not consider the effect of wild boar on ASFV spread. It has been shown that wild boar can be involved in the spread of ASFV in certain areas (Vergne et al., 2015). Thus the results of this study should be interpreted with caution for regions were established wild boar populations exist, as the success of the control scenarios might be affected by potential spread of ASFV between wild boar and domestic pigs. Still, scenarios designed to monitor mortality within domestic pigs might be effective, as early detection is important for limiting ASFV spread (Gallardo et al., 2015b). It could be interesting to investigate the effect of monitoring mortality in domestic herds as well as in wild boar, and to examine the effect in control of ASFV spread in regions where established wild boar populations exist. In the current simulation, the Danish population of swine herds was used as an example. We highly recommend other countries to repeat this exercise using their own data, as the optimal control strategy might be different. This difference could be due to differences in herd structure, contact and movement patterns between herds, density of the population, possible presence of wild boar and export patterns of swine and swine products. For instance, our results showed that under Danish conditions export losses are the major driver of the total cost and as a consequence the epidemic duration is of particular importance. In other countries without large exports of swine and swine products, other considerations might apply, e.g. minimizing the number of pre-emptive culling. Still, the approaches and modelled processes from this work can be followed. Conflict of interest None. Acknowledgements This project was financially supported as part of an agreement of commissioned work between the Danish Ministry of Food, Agriculture and Fisheries and the Technical University of Denmark. The work was commissioned by the Danish Veterinary and Food Administration. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j. vetmic.2016.11.023.
149
References Alban, L., Andersen, M.M., Asferg, T., Boklund, A., Fernandez, N., Goldbach, S., Ydesen, B., 2005. Wildrisk: classical swine fever and wild boar in Denmark: a risk analysis. Danish Inst. for Food Vet. Res.. http://orbit.dtu.dk/files/3390946/ WILDRISK_2005.pdf. Anonymous, 2015a. Bekendtgørelse Om Bekæmpelse Af Afrikansk Svinepest, BEK Nr 1469 Af 08/12/2015 (In Danish). . https://www.retsinformation.dk/Forms/ R0710 aspx?id=176036. Anonymous, 2015b. Udbrudsmanual for Afrikansk Svinepest. , pp. 26 oktober 2015 (In Danish). http://www.foedevarestyrelsen.dk/SiteCollectionDocuments/ Dyresundhed/Sygdoms-gruppen/Udbrudsmanual%20afrikansk%20svinepest% 2026-10-2015.pdf. Anonymous, 2015c. Bekendtgørelse Af Lov Om Jagt Og Vildtforvaltning. BEK nr. 1617 af 08/12/2015 x 38 (In Danish). https://www.retsinformation.dk/forms/r0710. aspx?id=175262. Anonymous, 2016. Veterinær Beredskabsplan Ved Udbrud Af Alvorlige Husdyrsygdomme. marts 2016 (In Danish). https://www.foedevarestyrelsen. dk/SiteCollectionDocuments/Dyresundhed/Sygdoms-gruppen/Veterin%C3% A6r%20beredskabsplan%20ved%20udbrud%20af%20husdyrsygdomme%20% 20marts%202016%20-%20uden%20anneks.pdf. Backer, J.A., Hagenaars, T.J., Van Roermund, H.J.W., 2009. Modelling the effectiveness and risk of vaccination strategies to control classical swine fever epidemics. J. R. Soc. 6, 849–861. Boklund, A., Alban, L., Toft, N., Uttenthal, Å., 2009. Comparing the epidemiological and economic effects of control strategies against classical swine fever in Denmark. Prev. Vet. Med. 90, 180–193. Boklund, A., Halasa, T., Christiansen, L.E., Enøe, C., 2013. Comparing control strategies against foot-and-mouth disease: will vaccination be cost-effective in Denmark? Prev. Vet. Med. 111, 206–219. CEC, 2002. Council Directive 2002/60/EC. Council Directive 2002/60/EC of 27 June 2002 laying down specific provisions for the control of African swine fever and amending Directive 92/119/EEC as regards Teschen disease and African swine fever. Off. J. Eur. Commun. L192 (June), 27–46. Chenais, E., Sternberg-Lewerin, S., Boqvist, S., Emanuelson, U., Aliro, T., Tejler, E., Cocca, G., Masembe, C., Ståhl, K., 2015. African swine fever in Uganda: qualitative evaluation of three surveillance methods with implications for other resource-poor settings. Front. Vet. Sci. 2, 51. doi:http://dx.doi.org/10.3389/ fvets.2015.0005. EFSA Panel on Animal Health and Welfare (AHAW), 2014. Scientific opinion on African swine fever. EFSA J. 12, 3628. doi:http://dx.doi.org/10.2903/j. efsa.2014.3628. EFSA Panel on Animal Health and Welfare (AHAW), 2015. Scientific opinion on African swine fever. EFSA J. 13, 4163. doi:http://dx.doi.org/10.2903/j. efsa.2015.4163. Gallardo, M.C., Ries, A.L., Kalema-Zikusoka, G., Malta, J., Soler, A., Blanco, E., Parkhouse, R.M.E., Laitao, A., 2009. Recombinent antigen targets for serodiagnosis of African swine fever. Clin. Vac. Immunol. 16, 1012–1020. Gallardo, C., Fernandez-Pinero, J., Pelayo, V., Gazaev, I., Markowska-Daniel, I., Pridotkas, G., Nieto, R., Fernandez-Pacheco, P., Bokhan, S., Nevolko, O., Drozhzhe, Z., C, Perez, C., Soler, A., Kolvasov, D., Arias, M., 2014. Genetic variation among African swine fever genotype II viruses, eastern and central Europe. Emerg. Infect. Dis. 20, 1544–1547. Gallardo, M.C., Soler, A., Nieto, R., Cano, C., Pelayo, V., Sanchaz, M.A., Pridotkas, G., Fernandez-Pinero, J., Briones, V., Arias, M., 2015a. Experimental infection of domestic pigs with African swine fever virus Lithuania 2014 Genotype II field isolate. Transbound. Emerg. Dis. doi:http://dx.doi.org/10.1111/tbed.12346. Gallardo, M.C., de la Torre Reoyo, A., Fernández-Pinero, J., Iglasias, I., Jesús Munoz, M., Luisa Arias, M., 2015b. African swine fever: a global view of the current challenge. PHM 1, 21. doi:http://dx.doi.org/10.1186/s40813-015-0013-y. Garner, M.G., Bombarderi, N., Conzens, M., Conway, M.L., Wright, T., Paskin, R., East, I. J., 2014. Estimating resource requirement to staff a response to a medium to large outbreak of foot and mouth disease in Australia. Transbound. Emerg. Dis. doi:http://dx.doi.org/10.1111/tbed.12239. Guinat, C., Gogin, A., Blome, S., Keil, G., Pollin, R., Pfeiffer, D.U., Dixon, L., 2016. Transmission routes of African swine fever virus to domestic pigs: current knowledge and future research directions. Vet. Rec. 178, 262–267. doi:http://dx. doi.org/10.1136/vr.103593. Halasa, T., Toft, N., Boklund, A., 2015. Improving the effect and efficiency of FMD control by enlarging protection or surveillance zones. Front. Vet. Sci. 2, 70. doi: http://dx.doi.org/10.3389/fvets.2015.00070. Halasa, T., Bøtner, A., Mortensen, S., Christensen, H., Toft, N., Boklund, A., 2016a. Simulating the epidemiological and economic effects of an African swine fever epidemic in industrialized swine populations. Vet. Microbiol. 193, 7–16. Halasa, T., Boklund, A., Bøtner, A., Toft, N., Thulke, H.-H., 2016b. Simulation of spread of African swine fever including the effects of residues from dead animals. Front. Vet. Sci. 3, 6. doi:http://dx.doi.org/10.3389/fvets.2016.00006. Jordt, A.M., Lange, M., Kramer-Schadt, S., Nielsen, L.H., Nielsen, S.S., Thulke, H.H., Vejre, H., Alban, L., 2016. Spatio-temporal modeling of the invasive potential of wild-boar: a conflict-prone species-using multi-source citizen science data. Prev. Vet. Med. 124, 34–44. Martínez-López, B., Ivorra, B., Ramos, A.M., Sanchez-Vizcaino, J.M., 2011. A novel spatial and stochastic model to evaluate the within- and between-farm transmission of classical swine fever virus. I. General concepts and description of the model. Vet. Microbiol. 147, 300–309.
150
T. Halasa et al. / Veterinary Microbiology 197 (2016) 142–150
Olsevskis, E., Guberti, V., Serzants, M., Westergaard, J., Gallardo, C., Rodze, I., Depner, K., 2016. African swine fever virus introduction into the EU in 2014: experience of Latvia. Res. Vet. Sci. 105, 28–30. R Core Team, 2015. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project. org/.
Sánchez-Vizcaíno, J.M., Mur, L., Martínez-López, B., 2013. African swine fever (ASF): five years around europe. Vet. Microbiol. 165, 45–50. Vergne, T., Gogin, A., Pfeiffer, D., 2015. Statistical exploration of local transmission routes for african swine fever in pigs in the Russian Federation. Transbound. Emerg. Dis. 2007–2014. doi:http://dx.doi.org/10.1111/tbed.12391.