COMPARATIVE ANALYSIS OF DIFFERENT STRATEGIES FOR THE CONTROL OF CLASSICAL SWINE FEVER IN THE REPUBLIC OF SERBIA USING MONTE CARLO SIMULATION

Several diff erent strategies for control and eradication of Classical Swine Fever (CSF) were compared using a Monte Carlo method-based simulation model. Th e control strategy analyzed in this paper, in addition to other CSF control measures, includes application of biosecurity measures on pig farms and rural backyard holdings. Elements of the control strategy are based on applicable regulations and include the simulation of detection of the disease, setting up the protected and surveillance zones, standstill of pig movements and restricted movement of animals, vehicles, equipment, and people with strong control measures in protection and surveillance zones, euthanasia of susceptible pigs, protective vaccination of pigs, compensation etc. During the simulation, diff erent output parameters were compared such as: duration of epizootic of a disease, number of aff ected holdings and animals, direct costs such as those for dead or culled animals, costs of surveillance, disposal of infectious materials, cleaning and disinfection. Depopulation of aff ected animals with early diagnostics and vaccination in protection and surveillance zone proved to be the most eff ective measures to stop spreading and eradication of the disease. However, during the simulation, systematic implementation of biosecurity measures in all pig production clusters demonstrated to be appropriate strategy for sustainable 1 Corresponding author: slavoljub.stanojevic@minpolj.gov.rs Arhiv veterinarske medicine, Vol. 9, No. 1, 43 59, 2016 Stanojević S. ... et al.: Comparative analysis of diff erent...


INTRODUCTION
Classical swine fever (CSF) is highly contagious disease of viral etiology affecting domestic and wild pigs. From economic aspect, CSF is the most severe threat to national pig industry in all countries. Th e disease is spread worldwide and is reported at all continents. Models and simulation of CSF epizootics enables assessment of disease dynamics as well as economic eff ects of implemented control measures. Th e objective of this article is to analyse potential control strategies for eradication of CSF in the Republic of Serbia based predominantly on improvement of biosecurity measures on a typical rural holdings and gradual upgrading of rural holdings to higher farm categories.

Description of the simulation model
Th e simulation has been conducted in the territory of the municipalities of Sremska Mitrovica and Šid. Th e area is characterized by high density of pigs as well as substantially heterogeneous pig-breeding tech nology, production habits and policies as well as production scale. Th e simulation encompassed all pig-breeding holdings including commercial farms, family farms type A, family farms type B and backyard holdings. Th e data on the number of pigs, production categories, and geographical locations were collected during the fi eld investigation or obtained from the Central Database of Ministry of the Agriculture. Data processing was performed using ARC GIS 10.0. soft ware package (Gatrell, 2004;Stanojevic, 2014). In cases where data about geographical location of rural holdings were not accessible, these were obtained from the Central Database of Ministry of Agriculture and the geographic coordinates were determined in ARC GIS 10.0, by randomly selecting (Gatrell, 2004). Th e simulation was performed applying the North American Animal Disease Spread Model based on the Monte Carlo method. NAADSM is a computer program based on the Monte Carlo method and is developed for the simulation of contagious animal diseases. Th e soft ware was developed by a team of experts of the Center for Epidemiology and Animal Health US Department of Agriculture from Fort Collins, Colorado (Jalvingh et al., 1999;Harvey et al., 2007;Reeves et al., 2012).
Th e basic idea underlying the Monte Carlo method is the approxi mation of the expected value E(X) by the arithmetic mean of the results of a large number of independent tests all with the same distribution as X. Th e stochastic simulations use random variables and are based on the law of ran dom numbers (Jalvingh et al., 1999).Th e model simulates daily disease transmission between farms and rural holdings for pig production. Th e simulation includes both direct and indirect contacts. Th e events such as "contact between the various epizootical units -ad equate contact" and "contact between the various units that caused the transmission of diseases -eff ective contact" are generated stochastically. Th e variability of the obtained results aft er 1,000 replications provide the information about the potential pattern of disease spreading (Gatrell, 2004;Engel et al., 2005;Harvey et al., 2007).At the beginning of the simulation, all backyard holdings and farms are considered "susceptible", except in cases where a number of pigs are vaccinated. Once acquiring the status "infected", a holding/farm has to pass through all other statuses predefi ned in the model. Table 1. describes the defi nition of the disease transition states. Aft er a short period of latency, all infected pigs disseminate the virus among susceptible population. However, there are certain diff erences in probability of an outbreak of the disease aft er adequate contact. Such diff erences are determined by intensity of direct and indirect contact between animals, the type of the holding and farming system itself as well as the level of implemented biosecurity measures. Some potential scenarios entail that certain number of pigs is vaccinated, thus possessing artifi cially induced immunity, which makes them non-susceptible. Defi ning parameters for disease spread sets down the modelling of control measures, laid down by relevant regulations. Upon completing the simulation, the following data are analysed: total number of infected farms and holdings, total number of diseased and culled animals, number of farms and holding where euthanasia was performed, duration of the outbreak, fi nancial data such as costs of euthanasia, disinfection and cleaning, expenses of safe disposal of carcasses, costs of laboratory examination etc. Depending on the scenario, several hypothetic situations were simulated including preventive vaccination of animals and no-vaccination scenario. Th e initial scenario describes actual status of CSF control in Serbia.
Other scenarios simulated spread of the disease in conditions with no vaccination or emergency vaccination aimed at preventing virus transmission outside of infested area ( Table 2).

RESULTS
Simulation of the scenario No.1 entailed analysis of the eff ects of implementation of currently relevant measures for CSF control in case of disease outbreak, taking into consideration factors such as actual production conditions and capacities of fi eld veterinary service. Scenarios 2-4 dealt with potential modifi cations of current control strategies and the assessment of their eff ects when applied in conditions of CSF control using preventive vaccination policy. In scenarios No. 5-14, hypothetical situations of no-vaccination CSF control were tested. Th e testing included also the hypothesis on CSF control using protective vaccination as well as the improvement of biosecurity meas-ures on rural holdings and family farms type B such as restriction of natural mating and intensity of indirect contacts. Fourteen diff erent hypothetical scenarios have been analysed (Table 2).
Th e analysis of obtained results revealed that CSF control using preventive vaccination strategy results in less direct economic losses, less number of diseased and culled animals as well as signifi cantly shorter duration of the epizootic. Th e simulation indicated that the area for preventive euthanasia of pigs should be set to a radius of max 50m around extensive rural holdings and family farms type A and B, and 500m around commercial pig farms. Th e simulation also revealed no statistically signifi cant diff erences regarding duration of epidemics, number of infected holding or animals even if the depopulation radius was limited to 100m around the infected commercial farm.
Th e most severe losses were observed in a scenario that was identical with the current fi eld conditions, yet presuming cessation of vaccination program and absence of protective vaccination in case of CSF outbreak. In scenarios No. 12, 13 and 14 characterized by absence of preventive vaccination but with restricted natural mating and improved biosecurity measures on rural holdings and family farms type B, the results revealed statistically signifi cant decrease in number of diseased animals as well as lower economic damage.
In conditions of termination of vaccination, the scenario No. 9 proved most appropriate, that is, the following measures are most eff ective: depopulation radiuses set to 500m and 50 m around commercial farms and rural holdings/family farms type A and B, respectively; implementation of protective vaccination policy and other measures laid down in relevant legislation. In all simulation models, there were no statistically signifi cant diff erences between the eff ects of protective vaccination applied in the radius of 10km or 3 km around the commercial farms. Table 4 depicts the results obtained in simulation models for 14 diff erent scenarios. In scenarios No.1-4, potential modifi cation of current CSF control strategy relying on preventive vaccination are analysed. Th e results obtained in simulation scenario No. 5 indicated that veterinary service is unable to control CSF without protective vaccination. As obvious from Table 4, implementation of controlled natural mating in case of CSF outbreak results in statistically signifi cant decrease in number of diseased animals as compared with scenarios lacking this measure (scenario No. 12 in Table 4).
Furthermore, if controlled natural mating in case of an outbreak of CSF would be associated with a decrease in intensity of indirect contacts for 25% and 50%, the number of diseased animals would be even more decreased (scenarios No. 13-14, Table 4).

Graph 1. Comparative graph of total economic damage expressed in EUR
Legend: -CFS control applying preventive vaccination (scenarios 1-4); -No-vaccination CFS control (scenario5); -CFS control applying protective vaccination as the alternative to mass pig (scenarios 6-11); -CFS control applying protective vaccination and improvement of biosecurity measures at rural holding and family farms type B (scenarios 12-14).

DISCUSSION
Simulation and mathematic modelling enable the pre-estimation of optimal control strategies, quantifi cation of potential epizootic outcomes, adjustment of relevant control plans, assessment of veterinary service and necessary resources (Jalvingh et al., 1999;Karsten et al., 2005a). Th is research off ered a review of potential outcomes of CSF outbreak in a limited area characterized by high pig density and highly heterogeneous pig production system. Potential dynamics of CSF epizootic as well as the level of consequent damage were described relating to diff erent approaches to disease control and eradication. Th e simulation revealed that rural holdings are highly susceptible to CSF; however, in such holdings, the potential for virus spread over large distance is lower, which corresponds with simulation results reported in Bulgaria (Backer et al., 2011). Spread of the disease over large distances is mainly associated with family farms type A and B and commercial farms. Rural holdings producing pigs for their own needs are not considered to be of high potential risk for disease transmission. Th e simulation model also suggested that, when speaking of rural holdings, local transmission is the most common route of infection with CSF virus. As regards the farms type B, most common infection routes include both local spreading and indirect contacts. Th e obtained results correspond with the results of the study analysing the potential of local spread of CSF virus conducted in Holland during a CSF epidemics in 1997-1998 (Karsten et al., 2005b; Klinkenberg et al., 2005). Th e simulation emphasized the role of uncontrolled natural mating in disease spreading at settlement level. Occurrence of CSF on family farms type A is mostly associated with the purchase of animals for fattening from producers from family farms type B and rural holdings. When speaking of commercial farms, two potential risk factors are most commonly associated with CSF outbreaks -introduction of infected animals into the herd and contacts with rural holdings through the personnel employed at the farm, who have their own pigs at home. Safe elimination of animals and protective vaccination proved most eff ective. On the other hand, the least economic losses are observed in conditions of continuous maintenance of appropriate immunity status of animals. Improvement of biosecurity measures proved highly important for substantial reduction of both disease transmission and economic damage in case of epidemic outbreak. Combined with other biosecurity measures such as good on-farm production practices, controlled access of visitors and vehicles, elimination of unnecessary contacts with other pig owners and preventing contacts with wild boars contributes to substantial reduction of the risk for CSF outbreaks.

CONCLUSIONS
Based on the results obtained in this research, we may conclude as following: 1. In endemic areas with predominate extensive pig production, relying on CSF control, strategy based on preventive vaccination proved most cost eff ective. 2. Preventive slaughtering of pigs should be carried out in the radius of max 50 m around the infected rural holdings, whereas destruction radius for commercial farms is set to 100 m. 3. If selecting the no-vaccination CSF control policy, modifi ed EU strategy based on protective vaccination and limited pig depopulation in a radius of 500m around infected commercial farms, i.e., 50 m around infected family farms type A and B and rural holdings, has proved most cost eff ective. 4. For all simulation models, where protective vaccination was used as control measures in combination with limited depopulation, there were no statistically signifi cant diff erences between the eff ects of protective vaccination if applied within the radius of 10 km around infected farm as compared to the radius of 3 km. 5. Restriction of natural mating and its limitation to one's own herd significantly reduces the risk of virus transmission and CSF outbreak. 6. Under the present conditions in the Republic of Serbia, it is not reasonable to implement a SCF control strategy without vaccination, particularly on rural holdings and family farms with lower levels of implementation of biosecurity measures.