This study aimed to improve understanding of the epidemiology of PRRSV infection by estimating the herd prevalence of PRRSV in England and by identifying possible risk factors for active PRRSV infection. Despite some limitations, the findings from this study agree with previous findings of risk factors studies and provide some new evidence of the factors involved in the epidemiology of PRRSV infection. The number of recruited farms representing 65 000 sows (approx. 14% of the total sow population in the UK in 2008), their geographic distribution and type of farms, suggest a good representation of English farrow-to-finish farms in this study. Further, when compared with a previous study carried out in England by Evans et al. (2008); farms of similar size were recruited.
The estimated herd prevalence of active PRRSV infection (35.1%) indicates continuous virus circulation or recent virus introduction on a number of farms despite efforts to control the infection. This finding is similar to that of a previous study carried out in England in 2003-2004, where the evidence of virus presence and transmission on the farm was seen in 25 non-vaccinated herds (32.8%) with seropositive young stock . Including the adult pigs in the sample of the same study, in total 41 of 76 non-vaccinated herds (53.9%) were seropositive. A similar result of 56% PRRSV seropositive herds was obtained through a diagnostic service offered to veterinary surgeons, carried out in Great Britain in 2001-2003 .
Sample size, sensitivity and specificity of the ELISA and PCR test and the possibility of selection bias needs to be considered when interpreting the results of this study. The initial recruitment of farms was conducted through the PCV2 vaccination program which could have resulted in the recruitment of a higher proportion of farms with more health problems than average and thus slightly overestimated herd prevalence of active PRRSV infection in England. However, to minimize the impact of this selection bias, 20% of farms were recruited through pig veterinarians and these farms were believed to have very few if any health problems. This study only included farrow-to-finish farms, and therefore results cannot be extrapolated to other types of farms. On the farms were growing pigs are reared on multiple sites, close contacts between the sites were observed and since these farms are continuously occupied by growing pigs of different ages, they are more likely to favour endemic PRRSV infection compared to breeding farms where young susceptible pigs are removed, or compared to single age all-in, all-out rearing units. The contact structure of the herds has previously been found to be an important factor for maintaining the virus within a farm .
Classification of a farm as PRRSV positive was based on growing and finishing pigs only. Therefore no vaccinated animals were included in the case definition. Testing pigs on seropositive vaccinating farms by PCR confirmed active circulation of PRRSV in these herds and reduced misclassification bias. Classification as PRRSV positive based on one positive ELISA only was made on five non-vaccinated farms. On these farms, it was a grower pig testing positive in the ELISA test, with finishers testing seronegative. Antibodies in these younger pigs but not in older finishers could indicate a longer than normal persistence of maternal antibodies [36, 37] rather than exposure to active infection and this may have resulted in misclassification of PRRS status on these five farms. Correlation of antibodies and virus (Table 2) on vaccinated farms is in agreement with the findings reported previously where number of seropositive animals increases with age and prevalence of virus decreases with age . Further, sensitivity (Se) and specificity (Sp) of the diagnostic tests used needs to be considered. The Sp of the Biobest in-house ELISA was validated as being greater than 95% and should result in no more than 5% of false positives on non-vaccinated farms. Less than 100% Se could result in some infected farms being missed; however, as results were interpreted at herd-level, the overall Se improved. The lack of PCR testing on all farms could also have resulted in misclassification bias. Ideally, a case definition would be based on the same laboratory test for all farms and would include use of both ELISA and PCR for evaluation of active PRRSV infection . Further, the detection of the virus on all sampled farms would be more indicative of recent virus infection than the use of antibody ELISA test. Due to financial constraints this was not feasible here. However having ELISA positive animals among both growers and finishers on non-vaccinated farms was considered sufficient evidence of recent virus circulation on these farms. This assumption is in agreement with findings from a previous study, where seropositive unvaccinated young stock was considered to indicate virus presence on that farm .
Taking into account manufacturer information on sensitivity (Se) and specificity (Sp) of the ELISA test (both > 95%) a farm would be correctly classified if the herd prevalence was 60% or higher (Free Calc, Version 2), which is realistic considering estimates of within herd prevalence found by others [3, 34, 38]. On the other hand, the detection of virus on vaccinated farms may be limited by the sample size used as a lower within herd prevalence is expected due to vaccination. Variability in virus prevalence between eight to 30% in growing pigs reported in the study by Duinhof et al, (2011), could have resulted in our sample size being insufficient to detect the virus and thus obtaining false negative results. Accordingly, vaccinated farms which were seropositive but virus negative were classified as dubious and were excluded from further analysis.
This study identified a high proportion (> 85%) of positive farms in North Yorkshire and East Anglia. This finding is similar to a previous report from England with regional variation of seropositive farms apparent in Yorkshire (82%) and East Anglia (56%)  . Both regions contain significant proportions of the pig herds in England and together account for more than half of the overall pig population. There are pig dense areas within both regions with pig units in close geographical proximity to each other. There may also be close relationships between these units sharing the same supply chains for pigs, feed and pig vehicles, which increases the risk of transmission of infection between units [23, 27, 35, 39]. In addition, aerosol transmission of the virus from infected herds over several kilometres, 4.7 km  and up to 9.1 km  under certain meteorological conditions, has also been demonstrated.
Using a multivariable logistic regression model, several risk factors for PRRSV infection were identified. A number of exposure variables related to management practices and biosecurity such as: type of ventilation for lactating sows and weaners; type of lighting for lactating sows; presence of cattle and poultry; and number of farm workers was found to be associated with PRRS farm status in the univariable analysis (Table 3). However, due to strong collinearity (p<0.01) with other exposures they were not retained for further analysis.
Farms using live virus vaccine were more likely to be PRRSV positive compared to non-vaccinated farms. This is not unexpected since farms which vaccinate are those which are likely to have experienced PRRS problems in the past and indeed, the majority of farmers of infected farms indicated that they believe to be infected. Also PRRS vaccination does not provide sterile immunity or prevent infection, but prevents clinical disease and reduces virus replication [42, 43], thus detection of PRRSV in nearly all vaccinating herds was not surprising. Based on PCR results alone it was not possible to determine whether the virus detected was the live vaccine or a field strain, in particular considering that studies elsewhere have detected vaccine virus in both vaccinated and non-vaccinated pigs . This was observed mostly with vaccine virus related to the North American (NA) genotype [44–47]. On the other hand, only limited transmission of the European genotype vaccine virus could be demonstrated [44, 48]. Considering the circulation of only European genotype of PRRSV in the UK and the limited transmission of European genotype vaccine virus, the presence of virus in growers and finishers was considered more likely to reflect a field virus challenge, although virus sequencing would be necessary to confirm this. Results should therefore be interpreted with caution as occasional vaccine virus circulation may have resulted in misclassification of herds.
Overall, the risk factors identified were not surprising and support known routes of virus transmission [23, 27, 35, 39]. They further agree with previous findings from England and elsewhere where herd size and pig density were found to be associated with increased risk of PRRSV infection and thus highlight the importance of biosecurity measures in preventing incursion of PRRSV [3, 32, 33]. Two identified risk factors for infection, collection of dead pigs and increased pig density, suggest that direct transmission from the pigs, or indirect from vehicles and people involved in their handling is an important factor for infection. Farms weaning piglets at the age of ≥ 28 days (28-45 days of age) had lower odds of being PRRSV positive compared to those weaning slightly earlier (21-27 days of age). Further analysis, after accounting for production type, found an association between weaning age and herd size. Farms weaning at the age of 28 days and later were more likely to be smaller herds (OR=0.3, 95%CI: 0.1-0.6, p<0.01). Previous studies suggested that in smaller herds, PRRSV is more likely to fade out [3, 35, 39]. This could be used to explain the association of the weaning age with farms’ PRRSV status observed in this study.
Two risk factors, pig density and vaccination with live virus, were used for the development of the stochastic model to evaluate the detection of an infected farm through passive pig disease surveillance. Farmers’ perception as to whether they believed their farm was infected by PRRSV was used to estimate the proportion of farms showing clinical signs suspicious of PRRS. The use of this information could have resulted in overestimation of this parameter since clinical signs were not seen by farmers on all of these farms. However, as clinical signs were identified to be the most sensitive input parameter, more robust collection of the data such as mortality records, growth performance and respiratory signs suspicious of PRRS in growing pigs and reproduction performance in breeding stock should be performed in the future. Furthermore, the AHVLA pig disease surveillance system supports diagnosis of any pig disease and the presence of clinical signs is an important factor prompting submission of material for diagnosis, thus presence of clinical signs was found to be most important for detecting PRRSV infected farms. Lack of clinical signs and poor recognition of clinical signs  could contribute to PRRSV infection remaining undetected on farms with active infection. Instituting PRRS-specific control measures could be delayed on such farms and they may therefore remain a source of infection to other pig units. The lower probability of detection of infection in farms using live vaccination compared to those without live vaccination (Table 5) is likely to reflect the reduction of clinical signs due to vaccination [34, 43]. While non-significant in the multivariable analysis, the presence of concurrent pathogens on PRRSV positive farms detected in the univariable analysis is important as it affects the morbidity and severity of clinical signs such as respiratory disease and ill thrift [26, 49, 50]. Submissions to AHVLA under passive surveillance diagnosed as PRRS tend to be from severe or unusual disease outbreaks or from herds breaking through natural or vaccine immunity . Such submissions provide important material for monitoring the PRRSV diversity and it has been confirmed that the virus has become more genetically diverse since the introduction into the UK .
Improved knowledge of herd prevalence and geographical distribution of PRRSV in England as well as an understanding of risk factors is important for more effective control of PRRSV. Given the estimated herd prevalence of PRRSV and the results from the scenario tree model, surveillance needs to be enhanced to support control of PRRSV. Knowledge of the risk factors for PRRSV infection could be used to develop more cost effective targeted surveillance which would include diagnosis of infection as well as disease and thus help current control programmes in England.