Descriptive analysis and spatial epidemiology of porcine reproductive and respiratory syndrome (PRRS) for swine sites participating in area regional control and elimination programs from 3 regions of Ontario

Abstract

The objectives of this study were to describe demographics, basic biosecurity practices, ownership structure, and prevalence of porcine reproductive and respiratory syndrome (PRRS) in swine sites located in 3 regions in Ontario, and investigate the presence of spatial clustering and clusters of PRRS positive sites in the 3 regions. A total of 370 swine sites were enrolled in Area Regional Control and Elimination projects in Niagara, Watford, and Perth from 2010 to 2013. Demographics, biosecurity, and site ownership data were collected using a standardized questionnaire and site locations were obtained from an industry organization. Status was assigned on the basis of available diagnostic tests and/or assessment by site veterinarians. Spatial dependence was investigated using the D-function, the spatial scan statistic test and the spatial relative risk method. Results showed that the use of strict all-in all-out (AIAO) pig flow and shower before entry are uncommon biosecurity practices in swine sites, but a larger proportion of sites reported having a Danish entry. The prevalence of PRRS in the 3 regions ranged from 17% to 48% and localized high and low risk clusters were detected. Sites enrolled in the PRRS control projects were characterized by membership in multiple and overlapping ownership structures and networks, which complicates the way the results of monitoring and disease management measures are communicated to the target population.

Introduction

Porcine reproductive and respiratory syndrome (PRRS) was first recognized in the late 1980s in North America and Europe (1). In 1991 and 1992, the causative agent was identified as small single-stranded enveloped RNA viruses in Europe (2) and North America (3), and it was discovered that these so-called PRRS viruses (PRRSV) had properties similar to those of the family Arteriviridae, genus Arterivirus. A few characteristics of the virus are essential for its ability to maintain itself in pig populations, which is a constant challenge for disease control. The first one is persistence of the virus in lymphoid tissues of individual pigs for long periods of time (4). The virus is also able to mutate and recombine (5), and is stable in cold and moist conditions (6). Additional challenges in PRRS control include the dynamics of modern swine production, which is characterized by a high degree of connectedness between sectors of the industry and the concentration of swine herds in production systems, further connected through specialized service providers. Routine pig movements among herds not only facilitate disease spread but also regularly introduce pools of susceptible animals (nursery pigs, gilts, boars) into the herds (7). In addition, it is not uncommon for sites to be located close to feed resources, creating areas of high pig density that may facilitate disease transmission. Area spread is a common term that corresponds to site-to-site transmission for which precise modes of transmission are unknown, but it is hypothesized that factors such as aerosol transmission, spread via avian or insect vectors, and contaminated fomites are involved (8). Some strains of the virus are able to remain viable after travelling in the air for up to 9 km (9).

Disease control programs that are focused on individual herds and production systems have limited effectiveness in the long-term control of PRRS infection (10). Consideration of PRRS dynamics at the regional level has been the focus of current PRRS Area Regional Control and Elimination (ARC&E) projects. Even though PRRS ARC&E programs have recently become popular in North America, reports on past or current programs are very limited in the peer-reviewed literature (11), and thorough analysis of data from such programs is almost entirely lacking. Descriptions of such programs (including their implementation, infrastructure, and infection control practices) would be valuable since they represent rare examples of approaches initiated by producers and the industry rather than imposed by the government towards control of infectious disease in animal populations with complex demographics (12). Furthermore, the investigation on the role of spatial proximity to other sites on PRRS status for particular regions might aid in future PRRS outbreak investigations. The objectives of the present study were first to describe demographics, biosecurity practices, ownership structure, and PRRS prevalence in swine sites participating in PRRS ARC&E projects located in the regions of Niagara, Watford, and Perth in Ontario and second to investigate the presence of spatial clustering and clusters of PRRS-positive sites located in those regions.

Materials and methods

Selection of sites, questionnaire administration, and data management

The source population for this cross-sectional study was the PRRS ARC&E database for the regions of Niagara, Watford, and Perth. Regions were not necessarily defined according to county or city boundaries, but according to the extent of producer’s interest and geographical plausibility (kept to a reasonable distance). The only county-level project was Perth, and the other 2 were considered regional-level projects. All swine sites enrolled in the control projects up to August of 2013 for Niagara sites, October of 2013 for Watford, and January of 2014 for Perth were included in the current study. Sites participating in such ARC&E projects were enrolled on a voluntary basis and, to be eligible, the sites had to be located within the determined area and the producer had to agree to share site location, attributes, and PRRS status among participants in the same PRRS ARC&E. The producer also had to answer a 20-minute questionnaire over the phone.

The questionnaire applied for data collection contained up to 40 questions regarding basic site demographics (e.g., number of animals and type of system), ownership structure (e.g., contact information for the premises owner, pig owner, and contract producer), sources (e.g., semen, boars, gilts, and feed), networks (e.g., truck and feed companies), biosecurity [e.g., use of all-in all-out (AIAO), shower-in, and Danish entry], presumed PRRS status and producer’s opinion on PRRS control. Site location was obtained from an industry partner (Ontario Pork, Guelph), and was based on the center point of the premises or parcel. A production system was defined as ≥ 2 sites linked by a common owner or management structure. It was possible for a single-site farrow-to-finish operation to be part of a production system if the owner had > 1 facility.

Diagnostic testing

After enrollment, sites were assigned an unknown PRRS status until results of diagnostic testing became available. The sample size recommended for PRRS status definition was 9 for oral fluids (composite sample) and 11 serum for individual animals, based on sample size calculations to declare freedom of infection (detection of at least one positive sample). For oral fluids, the assumptions for sample size calculations were that an individual test had low sensitivity (56%) and perfect specificity (13), expected within-herd prevalence of 10%, and that 5 pigs would contribute to one rope sample, with the confidence level set at 90%. For individual animal serum sample size, perfect test sensitivity and specificity were assumed (13), with expected within-herd prevalence of 20% and 90% confidence. Even though a high within-herd prevalence is expected for PRRS (14), prevalence levels were set to relatively low levels for sample size calculations to increase the ability to detect the virus in low prevalence situations. Serological testing was done with the aim of assessing evidence of exposure to PRRSV. Additional sampling and/or diagnostic testing using polymerase chain reaction (PCR) was done if the site veterinarian had reason to believe, on the basis of epidemiological or clinical assessment, that PRRSV was circulating on the site. All samples were submitted for testing to the Animal Health Laboratory at the University of Guelph. The immunofluorescent antibody assay (IFA) was used to rule out potential false positives on serum ELISA results.

Porcine reproductive and respiratory syndrome site status based on actual diagnostic testing was considered to be “confirmed” on a case-by-case basis, declared by the veterinary coordinator according to the use of recommended sample size being achieved and a combination of tests. A site was considered positive if at least one animal tested positive by ELISA (previous PRRSV exposure) or PCR (current PRRSV infection). As such, sites that were vaccinating animals (and therefore had positive ELISA results) were also considered positive. For cases in which diagnostic tests were not available, status was defined based on the site veterinarian’s knowledge on pig flow (e.g., confirmed positive feeder pigs being moved to a downstream finisher site), and such sites were considered “presumed” (15). For the purposes of this study, confirmed and presumed statuses were combined to define the binary outcome of interest for spatial analysis (PRRS positive or negative site).

Statistical analysis

Descriptive analysis

All captured data including geographical location, enrollment information, test submissions, and test results were entered in an online-based software (Fluid Surveys, Ottawa, Ontario) by project coordinators, investigators, and supervised technicians and were linked to a central database. In the later phase of the project, a Microsoft SQL 2008 database had been constructed in collaboration with industry groups to maintain the data in a standardized form.

All descriptive analyses were conducted at the site level, stratified by region. Initial descriptive statistics were determined and plots were generated for exploratory data analysis (SAS version 9.3; SAS Institute, Cary, North Carolina, USA). Computer software was used to visualize point data, create digital maps, and transform geographic coordinates from longitude/latitude to Cartesian coordinates (ArcMap version 10.1; ESRI, Environmental Systems Resource Institute, 2012, Redlands, California, USA). Computer software (R package “spatstat v.1.33-0,” R version 3.0.2; [R Development Core Team, 2013; (16)] was used to create risk maps for each of the regions. Two raster surfaces were created: the first representing the intensity of PRRS-positive sites and the second representing the intensity of all swine sites. The ratio of the case population over the population at risk was then calculated for each grid in the map and smoothed using Gaussian kernel smoothing. Finally, Stata-IC was used to describe PRRS prevalence over time for the region of Niagara (Stata-IC version 10; StataCorp, 2007, College Station, Texas, USA).

Spatial analysis

Spatial analysis was conducted through clustering and cluster analysis, the first being a global measure of spatial dependence and the second being an investigation of localized aggregation of positive sites.

Clustering analyses was conducted using computer software [R version 3.0.2; R (17)]. Sites that clearly did not belong geographically to the study area, although they were part of a specific project, were excluded from the spatial analysis. The study area for the spatial analysis was based on the convex hull of the remaining sites in each region. For clustering analysis, first the K-function was calculated using the “splancs v. 2.01-34” package (18), which is a measure of the number of events of the same type occurring within a certain distance. As this information is not very informative on its own, the K-function calculated for positive sites was compared with the one calculated for negative sites, and the difference between the 2, known as the D-function, is a measure of the extra aggregation of positive sites over and above that observed for the negative sites (19,20). Monte Carlo randomization was used to randomly permute locations of positive and negative sites, and values of the difference between the two were computed for each permutation. The 95% simulation envelope of these permutations, as well as 95% confidence limits calculated using a normal approximation, were plotted together with the observed D-functions (19) for each of the regions analyzed. Significant evidence of clustering was declared when the observed D-function deviated from the envelope formed by the upper and lower bounds calculated from the simulations (P < 0.05). The D-function was estimated over a distance of 20, 10, and 25 km for Niagara, Watford, and Perth, respectively.

The presence of localized clusters in the 3 regions was also investigated. The SaTScan™ program uses the spatial scan statistic method (21), and in this case the test was based on a purely spatial Bernoulli model and a circular shaped window that scans the region with gradually increasing sizes to include an increasing population of sites up to a limit of 50%. The risk of disease within each circle was compared to the risk outside using a maximum likelihood test and the window(s) with the maximum likelihood ratio function indicated the location of the most likely cluster(s), with significance being declared when P < 0.05. Both high and low risk clusters were investigated.

A different method for cluster detection was also applied for each area using the packages “splancs v. 2.01-34,” “spatstat v.1.33-0,” and “sparr v.0.3-4” on R (22). In this case, a spatial relative risk surface was constructed to represent the ratio of positive and negative densities estimated using quartic kernels with fixed or adaptive bandwidth, depending on which fitted better to the particular region. The areas were divided into grids and the observed surface relative risk was calculated for each grid. Case locations were randomly assigned to the grids, and a surface relative risk was simulated on the basis of 999 iterations. Finally, the function compared the observed with the expected spatial relative risk (simulations) for each grid and joined the ones with similar probabilities. Statistically significant high risk clusters were declared when P < 0.05.

Results

Descriptive analysis

Demographics of regions and sites

A description of region and site characteristics is provided in Table I. A total of 370 sites were enrolled in the 3 PRRS ARC&E projects in Southern Ontario during the years 2010 to 2013. For Niagara and Watford regions, participation rate was high (95% of all swine sites in the regions were enrolled for both projects). For the Perth region, participation rate was estimated at approximately 50%. Perth was the region with the largest study area (2589 km²), followed by Niagara (1666 km²), and Watford (510 km²). Figure 1 shows a graphical representation of the boundaries for each region. Assuming 50% of swine sites in the Perth region participate in the control program and represent a random selection of all sites in the region in terms of number of animals, the pig density is estimated to be twice as high as the one calculated with the current enrolled sites, or approximately 207 pigs/km². Therefore the region of Perth had the highest pig density (207 pigs/km²), followed by Watford (153 pigs/km²) and Niagara (61 pigs/km²). The distribution of production types for the 3 regions is shown in Table II.

Table I.

Region and site descriptors. The numbers in brackets correspond to number of sites with missing information

Descriptors	Niagara	Watford	Perth
Number of sites enrolled as part of the regiona	75	72	223
Mean number of buildings per siteb	—	1.25	1.23
Number of veterinarians	9 (n = 1)	11	14
Number of production systems	12	22	49
Number of premises owners	63 (n = 1)	54 (n = 1)	163
Number of sites part of a production system	56	56	191
Number of pig owners	—	32	82
Number of sites included in the polygonc	73	66	218
Polygon area (km2)d	1666.30	509.83	2589.10
Total number of animals in the polygon	100 050 (n = 1)	77 800 (n = 11)	267 725 (n = 2)
Number of animals per site in the polygon
Mean	1370.55	1178.79	1222.49
Minimum	75	154	52
Maximum	4500	5750	6000
Number of nursery/finishing animals in the polygon	87 165 (n = 1)	66 240 (n = 9)	237 995 (n = 2)
Number of sows/boars in the polygon	12 885	11 560 (n = 2)	29 730
Pig density (pigs/km2)	60.04	152.60	103.41e
Site density (sites/km2)	0.04	0.13	0.08e
Number of sites included in the spatial analysisf	65	56	197
Distance to closest neighbor
All sites (mean km)	2.32	1.66	1.58
Positive sites (mean km)	2.38	1.63	1.63
Negative sites (mean km)	2.31	1.70	1.55
All sites (max km)	8.20	8.15	13.52
All sites (min km)	0.22	0.18	0.15

^aIncludes all sites participating in the control program for the regions (those sites were included in all descriptive analysis regarding demographics and biosecurity practices).

^bInformation collected for nurseries, wean-finish, and finishers from the Watford and Perth regions only.

^cIncludes all sites that fit inside the created polygon (status positive, negative, and unknown).

^dArea corresponds to the area of the polygon that was created for spatial analysis purposes and does not represent geographic or political boundaries.

^eAs we estimate that about 50% of the sites in the region are enrolled on the project, the reader should be aware that pig and site density are underestimated on this table.

^fIncludes only sites with known status and within the areas of the polygon.

Figure 1.

Map of southern Ontario showing the 3 regions participating on porcine reproductive and respiratory syndrome (PRRS) Area Regional Control and Elimination (ARC&E) project that are described in this study.

Table II.

Description of production types of swine sites participating in the porcine reproductive and respiratory syndrome (PRRS) Area Regional Control and Elimination (ARC&E) control programs by region, in percent (n = number of sites)

Production type	Niagara	Watford	Perth
Farrow-wean	5.33 (n = 4)	6.94 (n = 5)	16.14 (n = 36)
Farrow-finish	10.68 (n = 8)	12.50 (n = 9)	7.62 (n = 17)
Farrow-feeder	9.33 (n = 7)	6.94 (n = 5)	4.93 (n = 11)
Nursery	9.33 (n = 7)	4.17 (n = 3)	15.25 (n = 34)
Wean-finish	5.33 (n = 4)	9.72 (n = 7)	6.73 (n = 15)
Finish	60.00 (n = 45)	56.94 (n = 41)	46.64 (n = 104)
Isolation/acclimatization	0.00 (n = 0)	2.79 (n = 2)	2.24 (n = 5)
Dry sows	0.00 (n = 0)	0.00 (n = 0)	0.45 (n = 1)
Total	100.00 (n = 75)	100.00 (n = 72)	100.00 (n = 223)

Biosecurity practices and herd immunity management

Biosecurity practices were available from the Watford and Perth regions only and are shown in Table III. For both regions, less than 50% of the swine sites reported having a shower-in facility at the entrance of the site. The sites that did not have a shower-in facility were asked whether there was a Danish entry, and most of those reported having at least this physical barrier at the entrance of their facilities. The majority of Watford and Perth swine sites had continuous-flow management (78% and 66%, respectively). The “continuous” category included both sites that had continuous flow and AIAO by room, while the AIAO category included sites reporting having AIAO by site and/or by building (Table III). The question concerning type of pig flow was not asked for farrow-to-wean operations which were considered to be strictly continuous due to the inherently open nature of such sites.

Table III.

Description of biosecurity practices for sites located in the Watford and Perth regions. Numbers are expressed as percent (n = number of sites that had the practice in place/total at risk)

	Region
	Watford	Perth
Biosecurity practice
CSHBa	78.26 (n = 54/69)	80.45 (n = 177/220)
Air filter	0.00 (n = 0/69)	0.00 (n = 0/217)
Shower in	47.69 (n = 31/65)	43.84 (n = 96/219)
Breeding sites	76.19 (n = 16/21)	44.93 (n = 31/69)
Growing pig sites	34.09 (n = 15/44)	43.33 (n = 65/150)
Danish entryb	80.65 [n = 25/(62–31)]	69.17 [n = 83/(216–96)]
Use of an external truck — allc	75.00 (n = 54/72)	74.44 (n = 166/223)
Incoming pigs	31.94 (n = 23/72)	36.77 (n = 82/223)
Outgoing pigs	63.89 (n = 46/72)	64.57 (n = 144/223)
Dead stock disposalc
Compost/burial/incineration within CAZ	26.39 (n = 19/72)	15.69 (n = 35/223)
Compost/burial/incineration outside CAZ	15.28 (n = 11/72)	13.90 (n = 31/223)
Third party pick up within CAZ	12.50 (n = 9/72)	8.07 (n = 18/223)
Third part pick up outside CAZ	2.78 (n = 2/72)	62.33 (n = 139/223)
Deliver to rendering	0.00 (n = 0/0)	2.69 (n = 6/223)
Pig flow — summary
Continuousd	77.94 (n = 53/68)	65.74 (n = 142/216)
AIAOe	22.06 (n = 15/68)	34.26 (n = 74/216)
Pig flow — detailed
Continuous	30.90 (n = 21/68)	8.80 (n = 19/216)
Continuous and AIAO by roomf	4.41 (n = 3/68)	3.24 (n = 7/216)
AIAO by room	35.29 (n = 24/68)	34.72 (n = 75/216)
AIAO by room and AIAO by buildingf	0.00 (n = 0/68)	1.40 (n = 3/216)
AIAO by building	13.23 (n = 9/68)	20.37 (n = 44/216)
AIAO by site	8.82 (n = 6/68)	13.89 (n = 30/216)
F-Wg/dry sows operations (not asked)	7.35 (n = 5/68)	17.59 (n = 38/216)
Missing information	4	7

^aSites that completed the Canadian Swine Health Board National Biosecurity Training Program.

^bOnly sites that did not have a shower facility were at risk for this outcome.

^cMore than one of these answers might have been chosen by one site.

^dCombines continuous flow (including F-W and dry sow operations), AIAO by room and any site that any of those combined or with other flows.

^eAIAO corresponds to all-in all-out practice; this category includes AIAO by site or by building. Missing information was excluded from the denominator.

^fSites had a combination of those 2 pig flows.

^gFarrow-to-wean operations.

CAZ — Controlled access zone.

Herd immunity management differed between regions. In the Watford area, < 9% of the sites were exposing animals to PRRSV (in the last 6 mo of questionnaire administration): 6 sites were vaccinating using either the Ingelvac ATP (Boehringer Ingelheim Vetmedica, Burlington, Ontario, n = 4) or the Ingelvac MLV vaccine (Boehringer Ingelheim Vetmedica, n = 2). In the Perth area, less than 20% of the sites were exposing their animals, either by using the FOSTERA™ (Zoetis Animal Health, Kirkland, Quebec, n = 1), the Ingelvac ATP (Boehringer Ingelheim Vetmedica, n = 3), the Ingelvac MLV vaccine (Boehringer Ingelheim Vetmedica, n = 28), or doing live virus inoculation (n = 11).

Ownership structure

The mean number of sites owned per premises owner was similar for the 3 regions (1.2 site/premises owner for Niagara, 1.3 site/premises owner for Watford, and 1.4 site/premises owner for Perth). However, the mean number of sites per pig owner was somewhat higher for Perth (2.7 sites/pig owner) than for Watford (2.2 sites/pig owner). The premises owner was not the pig owner for 43.06% of the premises in Watford and 46.67% in Perth. The information regarding pig owner was not available for Niagara at the time of analysis. A relatively small number of veterinarians (n = 18) were responsible for the farms located in the 3 regions, and 56% of the veterinarians were responsible for sites in more than one region. For the region of Niagara, there were approximately 8 sites per veterinarian [range: 1 to 38, inter-quartile range (IQR): 6.0], for Watford 7 sites per veterinarian (range: 1 to 26, IQR: 6.5), and for Perth, 16 sites per veterinarian (range: 2 to 69, IQR: 10.0).

Diagnostic testing

A total of 335 laboratory submissions were available during the period examined herein for the 3 regions: 183 submissions for Perth, 71 for Niagara, and 81 for Watford. Of those submissions, 194 submissions were based on serology only (either serum or oral fluids ELISA), 58 were based on virology (either serum, oral fluids or tissue PCR) and 77 were based on both types of diagnostic tests. For 6 submissions, neither serology nor virology was done (3 submissions were for sequencing only and three had missing information). The 335 submissions were from 252 premises (including all premises enrolled in the projects): 206 premises with only one submission, 25 premises with 2 submissions, 12 premises with 3 submissions, and 9 premises with ≥ 4 submissions.

Out of the 271 submissions tested using ELISA, all of them were tested individually (i.e., none were pooled for testing), and out of the 135 samples tested using PCR, 78% were pooled. For a total of 209 submissions, serum samples were submitted; for 108 submissions, oral fluids were submitted; for 11 submissions, both serum and oral fluid samples were submitted; for one submission, fresh tissue was submitted; and for 2 submissions both serum and tissue samples were submitted. For 4 submissions, information regarding type of sample was missing (data entry error).

Porcine reproductive and respiratory syndrome regional prevalence

The mean within-site prevalence of PRRS calculated using the serum ELISA data (n = 37 positive sites) was 71.41% (median = 90.91% and SD = 0.35). The mean overall prevalence of PRRS was lowest for the Niagara region, followed by Perth, and finally Watford (Table IV). For the region of Niagara, prospective data allowed for a temporal description of PRRS prevalence during the previous 3 y. Overall prevalence in the region of Niagara decreased over time (Figure 3). In February of 2012, PRRS was diagnosed on some sites in a few production systems, and control and elimination strategies, such as site closure (herd closure) with or without homogenization (exposure of all animals in the herd to PRRSV), were put in place then. The prevalence of PRRS decreased to approximately 16% by August of 2013.

Table IV.

Mean prevalence of the porcine reproductive and respiratory syndrome (PRRS) for the 3 different control regions and stratification on presumed versus confirmed status

	Region
	Niagara	Watford	Perth
Prevalence % (na/Nb)	16.92 (11/65)	48.21 (27/56)	40.61 (80/197)
95% CI (%)	7.81, 26.04	35.13, 61.30	33.75, 47.47
Exact 95% CI (%)	8.76, 28.27	34.66, 61.97	33.69, 47.82
Presumed status (n)
Positive	0	9	30
Negative	0	2	21
Total	0	11	51
Confirmed status (n)
Positive	11	18	50
Negative	54	27	96
Total	65	45	146

^aNumber of sites with the outcome.

^bTotal number of sites with known status in the region.

95% CI — 95% confidence interval.

Figure 3.

Temporal trend of porcine reproductive and respiratory syndrome (PRRS) for the Niagara region.

Spatial analysis

Complete flows regarding exclusion of sites from spatial analysis for the 3 regions are shown in Figure 2. The D-functions estimated from the data did not drift outside the 95% confidence bands constructed from the Monte Carlo simulations; therefore, spatial clustering of PRRSV in swine sites could not be detected for the region of Niagara in August 2013 (P = 0.25), for the region of Watford in October 2013 (P = 0.91), and for the region of Perth in January 2014 (P = 0.11)

Figure 2.

Flow diagram of the porcine reproductive and respiratory syndrome (PRRS) Area Regional Control and Elimination (ARC&E) projects in Ontario.

Cluster detection with the scan statistic method found one cluster where risk of disease was higher and one cluster where risk of disease was lower for both regions of Watford and Perth. For the region of Watford, the high risk cluster had an observed number of 18 cases (expected 10.6 cases, P = 0.013), and the low risk cluster had 0 observed cases (expected 11.0 cases, P = 0.016). For the region of Perth, the high risk cluster had 58 cases (expected 37.8 cases, P < 0.01) and the low risk cluster had 2 cases (expected 13.8 cases, P < 0.01). No clusters were detected in the Niagara region. High risk clusters for the regions of Watford and Perth were mapped using R and are shown with the risk maps in Figure 4. High risk areas were located southwest of the region of Watford and the central-southern area of Perth, along with one smaller area located on the northeastern portion of the polygon (Figure 4). For the Niagara region, a high risk area was located in the eastern part and southwest. The spatial relative risk method was applied in order to detect high risk clusters. This method allowed for detection of clusters for all regions, including for the Niagara region using adaptive bandwidth (P < 0.05). For the regions of Watford and Perth, clusters in the same area were located using the 2 different methods.

Figure 4.

Risk maps and high risk clusters detected by the spatial scan statistic method for the regions of (A) Niagara in August 2013, (B) Watford in October 2013, and (C) Perth in January 2014.

Discussion

To the knowledge of the authors, this is the first study to describe swine sites participating in PRRS ARC&E projects in Ontario and to explore spatial distributions of positive and negative sites from 3 different projects. It is important to note not only the large sample size used, representing approximately 15% of all Ontario swine sites (23), but also the inclusion of 3 regions that contain different pig densities. The average number of pigs per site for the 3 regions was very similar to Ontario’s average, 1238 (23).

Even though biosecurity practices such as use of shower-in and AIAO are highly recommended for infectious disease prevention and control (24), results of this study show that implementation of these practices is not high for the analyzed regions. The percentage of sites with growing pigs managed by AIAO pig flow could have considerable implications for the design of disease control strategies, since for these sites PRRS status could rapidly change with strict age separation and thorough cleaning between batches. In contrast, elimination of infection from continuous flow sites is expected to be more challenging. The relatively high percentage of continuous flow sites, as defined in this study, is a reflection of the demographics of the Ontario swine industry, and it would be unrealistic to attempt to change this over a short time frame. It would therefore be worthwhile to develop and evaluate procedures for elimination of PRRS from continuous flow sites without complete site depopulation. However, a large proportion of swine sites that did not have a shower-in facility reported had at least a Danish entry in place.

Dead stock removal may present an important risk factor for PRRS infection. Regular visits by incoming trucks can be an important source of virus coming from other sites, as trucks may mechanically transport the virus (6). Currently, it is recommended that rendering trucks should not be allowed access to farms; composting or incinerating carcasses on farm would be a more appropriate control measure (24). The most appropriate procedure to manage deadstock, however, also needs to be considered on a herd-by-herd basis. The high percentage of sites using external trucking companies for outgoing pigs is another area in which regional-level collaboration is needed, given the importance of transportation practices for risk of transmission between sites (25).

For the regions of Watford and Perth, approximately 43% and 47% of the sites had a premises owner that was different from the pig owner, representing a challenge when determining who should be notified if needed. This becomes particularly important when the concept of surveillance comes into place: according to the Center for Disease Control and Prevention, surveillance corresponds to “the ongoing systematic collection, analysis, and interpretation of health-related data essential for the planning, implementation, and evaluation of (public) health practice, (…) closely integrated with the timely dissemination of these data to those who need to know” (26).

Our findings also demonstrate the complex nature of the structure of modern swine systems in Canada, raising questions about effective methods of communicating major changes in the disease control regions to all involved stakeholders. The solution to this issue is not simple and requires effort from both producers and premises owners as well as from informers; an effective way of communicating is still under development.

The PRRS ARC&E projects in Ontario are producer-driven and focused on conducting disease surveillance, control, and elimination when warranted; with minimum cost and maximum practicality for producers and veterinarians involved. When data are being utilized for research purposes, the inconsistency of sample collection for laboratory analysis among sites and regions might raise some concerns. Recognition that there is heterogeneity among sites supports the idea of an output-surveillance approach, in which rather than prescribing what surveillance activities must be done, standards prescribe what must be achieved (27). This provides more flexibility with regard to the use of different tests or test combinations, different sample sizes, and different sampling strategies that will be defined according to factors such as historical information and multiple sources of surveillance, always with the aim of increasing efficiency (27). This has been the approach of the on-going surveillance project from which data for this study were generated, with individual samples based on serum and rope samples based on oral fluids both allowed (and yielding similar theoretical confidence). This provides accurate classification and is at the same time practical and more economically effective.

The overall PRRS prevalence found in this study was 37.1%, below that previously reported for Canada (28,29), the United States (30), Mexico (31), and Spain (32). The region of Niagara contributed low site level prevalence since it has had control and elimination strategies previously implemented. With regard to PRRS within-site prevalence, assessments from all regions over time support a mean within-site prevalence greater than 70%. This is in agreement with previous reports that 80% to 95% of the pigs in a site experience seroconversion within 2 to 3 mo of exposure (14).

For the region of Perth, the volunteer cohort used in the current study only represented approximately 50% of the swine sites in the region; therefore, PRRS prevalence estimated for this region is less accurate compared with the prevalence on the other 2 regions. The authors also acknowledge the possibility of misclassifying sites due to inadequate sampling, which is more likely to occur in some segments of sow herds in which control strategies contribute to a very low prevalence of PRRSV.

Existence of spatial clustering should be reflective of area spread under certain conditions. Larochelle et al (8) report area transmission being suspected when PRRSV strains originated from sites belonging to different ownerships located within 3 km of one another. In our study, despite existence of spatial clusters, clustering analysis was not able to identify that distance to positive sites plays a significant role in the PRRS status of neighboring sites. These findings agree with what has been previously reported for Ontario by Rosendal et al (33), but disagree with a study conducted in Quebec that reported that sites located ≤ 2.5 km from the closest pig site were more likely to be positive (29), and could be a reflection of differences in the source population or unreported efforts already taken in disease control.

It is important to emphasize that the spatial scan test aims to detect location of high risk areas, where the risk of disease is elevated relative to the area outside of the cluster. Such elevated risk of disease could be due to several distinct types of exposure. Two very distinct types of exposure that could, at least theoretically, result in such areas are: i) some type of environmental exposure, or (ii) spread of infectious disease between herds; the latter would be indicative of point-to-point interaction and can be measured using tests of spatial clustering such as the D-function. Thus, it can be concluded that high risk areas for PRRS exist, but we were not able to confirm with our approach whether they are result of aggregation between case herds. However, PRRS continues to be the most important disease of North American swine populations, and many producers deliberately choose certain disease status that could influence the results found herein. The authors attempted to adjust for that by running analysis including only sites with sows and continuous flow, but the results did not change. One methodological question that arises is whether these well-described methods of spatial clustering are indeed the most suitable for assessing local spread in highly concentrated industries such as the current swine industry. Further elucidation of this question is needed.

There was surprising complexity of the data collected from producers, which included ownership structure (premises owners, pig owners, contract producers) and related networks from service providers (such as: veterinarians, feed companies, semen and gilt sources, truck companies). Observed data complexity and the absence of spatial dependence in the study findings suggest that other factors may be involved in PRRS area transmission. Further analysis on the relative contribution of network structure and location on PRRS spread are needed for better understanding of PRRSV dynamics and development of efficient control and elimination strategies at a regional level.

In conclusion, the current study describes the demographics, biosecurity practices, and ownership structure for swine sites located in 3 regions participating in PRRS (ARC&E) projects in Ontario. These programs have been well-received by producers, with high participation rates, and are anticipated to expand further. For the Niagara region, coordinated control efforts resulted in a 25% decrease in PRRS prevalence over the last 3 y.

Patterns of disease spread were also investigated and results showed no evidence for spatial dependence in the data. However, other visual representations, such as risk maps and cluster analysis, suggest the presence of high risk areas within regions, which might be an important focus in situations where resource usage must be optimized. They could also be useful as hypothesis generators and in further investigations. Future challenges include sustainability and update of information on a regular basis, and the design of strategies to facilitate decision-making processes when it comes to infectious disease prevention and control at the farm and region levels.