Assessment of Risk Factors for a High Within-Batch Prevalence of Yersinia enterocolitica in Pigs Based on Microbiological Analysis at Slaughter

Introduction

Yersiniosis is an important foodborne disease responsible for about 7000 confirmed human cases per year in the European Union and around 14,600 infections in young children in the United States each year (EFSA/ECDC, 2013; Scallan et al., 2013). Infections are mainly caused by the human pathogenic biotypes of Yersinia enterocolitica (98.4%) and to a lesser extent, Y. pseudotuberculosis (0.9%). Pork and products thereof are considered the most important infection source. Contamination of the meat occurs mainly by (cross-)contamination during slaughter by pigs harboring Yersiniae in their tonsils or in their feces (Laukkanen et al., 2009). It was suggested to assist the reduction of the contamination risk in pig slaughterhouses by decreasing the number of infected animals at farm level (EFSA, 2011). Decreasing the occurrence of Y. enterocolitica in pigs and ultimately in pork is considered the most important control step toward reducing the number of human infections; therefore, the infection rate of delivered batches prior to slaughter should be known so that appropriate actions can be taken in the slaughterhouse.

Studies of risk factors for the occurrence of Y. enterocolitica at pig-farm level have been performed in several countries, but, except for the study of Wesley et al. (2008) in the United States, these were always based on a limited number of farms (9–15) or pigs per farm sampled (6.1) (Nowak et al., 2006; Laukkanen et al., 2008, 2009; Virtanen et al., 2011; Novoslavskij et al., 2013). So far, common findings are that the use of bedding material, the purchase of piglets originating from more than one farm, daily observation of the presence of a cat in the stables, drinking from nipples, and snout contact between pigs from adjoining pens in the fattening pig unit are risk factors for infection. The use of municipal water, a farrow-to-finish approach, and organic farming acted as protective factors (Laukkanen et al., 2009; Virtanen et al., 2011; Novoslavskij et al., 2013; Vilar et al., 2013). Certain intervention strategies based on these factors could reduce the within-batch prevalence prior to slaughter, but have not been implemented so far.

The aim of the present study is to assess different farm factors influencing the microbiological prevalence of human pathogenic Y. enterocolitica in pigs at slaughter to obtain interventions to reduce this prevalence and that finally decrease the level of Y. enterocolitica (cross-)contamination in the slaughterhouse.

Materials and Methods

In a previous study by Vanantwerpen et al. (2014), tonsils of pigs at slaughter (n=7047) were sampled from January to December 2012. These pigs were originating from 100 batches (a batch is defined as pigs originating from 1 farm delivered to the slaughterhouse on the same day), each batch of which originated from another randomly selected farm (n=100). On average, 70 pigs per batch were sampled. The tonsils of these pigs were analyzed by direct plating on Cefsulodin–Irgasan–Novobiocin agar plates (Schiemann, 1979), whereafter the isolates were confirmed by polymerase chain reaction targeting the ail, ystB, and virF genes (Harnett et al., 1996). Pathogenic Y. enterocolitica serotype O:3 was isolated from tonsils of 2009 pigs (28.5%), originating from 85 farms (15 farms were Y. enterocolitica negative). The within-batch prevalence ranged from 5.1% to 64.4%.

In the present study, shortly before the pigs were slaughtered, each farm was visited and a face-to-face questionnaire was filled out combined with a guided tour on the farm. The Yersinia status of the farm was still unknown and the questioning was always performed by the same investigator. In total, 59 fattening pig herds and 41 farrow-to-finish herds were included. The number of sows in the farrow-to-finish farms varied between 80 and 750, and the size of the fattening pig farms ranged from 297 to 10,500 slaughter pigs. The items in the questionnaire were related to the following aspects: type of farm management and housing system, biosecurity and hygiene measurements, animal management, origin of the drinking water, type of feed and veterinary support (See Supplementary Data; Supplementary Data are available online at www.liebertpub.com/fpd). Production parameters were obtained from the farmer after slaughter. In total, 68 questions were asked to the farmer or evaluated during the visit. Most of the data were binomial (questions answered by yes or no), some categorical (e.g., feed: [1] grain, [2] meal, [3] porridge) or continuous (e.g., daily weight gain). The yearly average Salmonella result, expressed by the ratio of sample to positive, was calculated based on farm-specific official reports.

Excel software and Stata/MP 12.1 (StataCorp., 2011) were used for all analyses. The dependent variable was defined as the infection status of the animals (presence/absence of Y. enterocolitica). Independent variables included categorical, continuous, and binomial farm-level variables. When the correlation coefficient between farm variables was high (r≥|0.7|), only one of the variables was further included in the model. The decision regarding which variable to include depended on the biological plausibility, which ruled out only one item: the number of sows was correlated with the number of slaughter pigs present.

The association between each independent variable and the outcome was first screened using a univariable mixed effect logistic regression with farm as random effect. Variables with p<0.05 in this univariable analysis were retained for the multivariable model also with farm as random effect. In this model, only items that remained significantly associated with Y. enterocolitica carriage in the tonsils (p<0.1) were determined either as risk or as protective factors. All possible interactions between the significant factors of the univariable analysis were evaluated and included in the model when significant (p<0.1).

The mean within-batch prevalence was calculated per number of piglet suppliers (three categories: none, one, and more than one supplier) and analyzed using analysis of variance.

Results

After a first evaluation visiting 100 farms, some of the collected items turned out to be unusable due to the low variance between farms, as for instance the use of straw bedding in the nursery or fattening pig unit (only applied by 1 of the 100 farms), the obligated use of a shower for visitors before entering the stables (3%), the use of different shoes when entering the pen with the ill pigs (1%), and the use of rainwater tanks or wells as drinking-water supply (100%). Therefore, in the complete study, items with five or fewer farms applying this item were ruled out, resulting in the exclusion of 19 of 68 items.

Some closely related items were pooled in the univariable analysis (e.g., the item “cleaning–disinfection–emptiness” [CDE] consists of (1) the cleaning and (2) the disinfection of the stable after each rearing round and (3) the time period in which the stable remained empty after the pigs were slaughtered with a cut-off of 3 days). A “proper use of a disinfection bath” consists of (1) whether the bath is placed inside the stable, so it is not protected from weather conditions and one is obliged to step in it; and (2) the bath being cleaned at least every 2 weeks.

Items with more than 15 missing responses were omitted. These were mostly the consequence of a farmer or person responsible for the stable who did not know the answer to our question. As a result, only 36 explanatory items were taken further into account (Table 1).

After the univariable analysis, only seven factors were significant (p<0.05), which were all included in the multivariable analysis. During this analysis, the element “snout contact possible between pens” (p-value_univariable=0.045) was excluded due to a p-value_{multivariable}>0.1. In the final logistic regression model, three risk factors, two protective factors, and one significant interaction were identified (Table 2). Significant risk factors were “the use of a semislatted floor in the fattening pig unit,” “presence of other pig farms in the area (closer than 500 m),” and “the number of piglet suppliers” (p _{multivariable}<0.1). Protective factors were “the proper use of a disinfection bath” and a “bad CDE.” A significant interaction (p _{multivariable}<0.1) was observed between a “bad CDE” and “the presence of pets in the stable,” which led to an increasing prevalence.

The mean within-batch prevalence increased with an increasing number of piglet suppliers. Nevertheless, the mean within-batch prevalence of farrow-to-finish farms (25.9%; n=41 farms) and fattening pig farms with only 1 supplier (25.2%; n=34 farms) was almost equal. When more suppliers were involved, the mean within-batch prevalence increased to 37.9% (i=25) (p<0.1).

Discussion

The present study assessed different risk factors based on microbiology of human pathogenic Y. enterocolitica present in pig tonsils at time of slaughter originating from 100 farms, taking into account clustering per batch (Vanantwerpen et al., 2014). As the prevalence based on bacteriological data at time of slaughter leads to the final carcass contamination (Laukkanen et al., 2009), risk factors related to this prevalence are also affecting this contamination.

Based on a much larger number of farms, as well as pigs examined in the present study, some previous factors could be confirmed, whereas other factors were new or even contradicted previous studies.

In the present study, factors increasing the prevalence at time of slaughter were identified. The first two risk factors, “the use of a semislatted floor in the fattening pig unit” and “presence of other pig farms in the area (<500 m)” probably cannot be remediated immediately. The use of a semislatted floor increases the prevalence based on bacteriological results compared to a fully slatted floor, represented by a 10% increase in mean within-batch prevalence of 36% and 26.6%, respectively. The use of a fully slatted floor was already mentioned as a protective factor by von Altrock et al. (2011) in German pig farms. The influence of the type of floor can be explained by the contact time with pig feces, in which Y. enterocolitica is present. Semislatted floors create a more constant contact with the pig feces (Fukushima et al., 1983; Nielsen et al., 1996; Nesbakken et al., 2006). The second risk factor—“presence of other pig farms in the area (closer than 500 m)”—was never studied before and indicates a transmission route between nearby farms. The origin of this spread, however, is still unknown. It is possible that fecal contamination spreads by fouled boots or cars and is distributed to other farms.

The third risk factor, “the number of piglet suppliers,” differed between farms with none or one supplier, and farms with more suppliers. The maximum number of piglet suppliers in this study was 11 per farm. It is known that incoming piglets, potentially infected on multiplying farms, are a possible infection source and spreading on fattening pig farms (Virtanen et al., 2012). When piglets arrive from different multiplying farms, there is more chance to purchase infected, Yersinia-excreting piglets. Purchasing piglets from more than one farm was also identified as a risk factor in the studies of Nowak et al. (2006) and Vilar et al. (2013). In contrast to the study of Skjerve et al. (1998), which was based on serological results, a farrow-to-finish farm was seen as a protective factor. Nevertheless, the present study indicates an equal risk between fattening farms when purchasing piglets from one supplier and farrow-to-finish farms, which is similar to the study of Vilar et al. (2013).

In addition, two new protective factors were identified and both are related to hygienic measurements. The first one, “proper use of a disinfection bath,” decreases the spread between different stables on the same farm, so infections stay more localized. The difference with the “farm clothes available for visitors,” which is not a significant factor, is that the disinfection bath is also used by the farmer, and not only by the farm visitors. In this way, infections coming from outside the farm, and even within the farm but from different stables, are stopped by this disinfection bath. The second protective factor, a “bad CDE,” is more difficult to interpret. Similar results are available for the presence of Salmonella on pig farms. Van der Wolf et al. (2009) reported that the omission of disinfection of pig stables was associated with a lower Salmonella seroprevalence compared to herds that sometimes or always applied disinfection. Moreover, Poljak et al. (2008) showed that increased frequency of cleaning with cold water and disinfection was positively correlated with Salmonella shedding. Moreover, when pets are allowed into the stable, “bad CDE” is a risk factor for a higher prevalence. When pets are allowed in and move between stables, they keep the infection going as they act as carriers and transmitters of Y. enterocolitica (Yanagawa et al., 1978; Fredriksson-Ahomaa et al., 2001; Murphy et al., 2010). The daily presence of a cat with kittens in the stable was identified as a risk factor in the study of Skjerve et al. (1998). The use of an all-in/all-out management was also identified as protective factor in a more limited study of Vilar et al. (2013), and Novoslavskij et al. (2013) reported a low biosecurity level as a risk factor. Nevertheless, in the latter study, the biosecurity factor included other factors than those in the present study.

The use of straw bedding was initially taken into account in the questionnaire, but as it was applied in only one farm, not further included. In contrast, the use of bedding was identified as a risk factor (Skjerve et al., 1998; Laukkanen et al., 2009; Vilar et al., 2013). Moreover, the use of municipal water is considered a protective factor (Virtanen et al., 2011; von Altrock et al., 2011; Vilar et al., 2013), though this factor was excluded, as all farms used rainwater tanks or wells as drinking-water supply. It is an interesting subject to take into account in future studies. A higher production capacity has been associated with a higher prevalence of Y. enterocolitica, due to underlying risk factors; nevertheless, no differentiation was found in the present study (Laukkanen et al., 2009). Feeding factors have also been implicated with the prevalence of Y. enterocolitica. Manual feeding of slaughter pigs and buying feed from a certain company A have been determined as protective factors. Use of commercial feed, presence of meat or bone meal in grower–finisher diet, industrial byproducts in feed, and buying feed from a certain company B have been associated with a higher Y. enterocolitica infection rate (Skjerve et al., 1998; Nowak et al., 2006; Wesley et al., 2008; Virtanen et al., 2011). In the present study, no significant differences were found for feed factors. In the present study, no relation was found with the presence of Salmonella in the pig herds. Von Altrock et al. (2011) insinuated that the Yersinia prevalence is inversely associated with the serological Salmonella status.

In the general picture of pig-producing management, another foodborne pathogen, Salmonella, is also very important. The risk and protective factors for a higher prevalence of Y. enterocolitica on pig farms could be conflicting with factors influencing the prevalence of Salmonella on pig farms. In a study by Vico et al. (2011), the mesenteric lymph nodes of pigs were collected in the slaughterhouse and analyzed for the presence of Salmonella. Risk factors in this study were lack of rodent control programs, fattening pig herds, herds managed by more than one full-time worker, municipal water as drinking-water supply, and relatively long fattening times. Only one factor is similar between the study about Salmonella and the present study: pig herds with one or more piglet suppliers (fattening pig herds) are a risk for a higher prevalence of Y. enterocolitica and Salmonella. Cardinale et al. (2010) performed a risk factor study based on the bacteriological Salmonella status of 60 farms by analyzing fecal samples and socks. The prevalence increased when there was no disinfection at the farrowing stage, when large numbers of cockroaches were present, and when birds were seen in the stable. A lower level of Salmonella was reached when the technical personnel visited the stable less than once a month, when castration of piglets was done after 1 week of age, and when the all-in all-out system was respected. Garcia-Feliz et al. (2009) analyzed fecal samples for the presence of Salmonella. The only two risk factors given in this study were the feeding of pelleted feed and a high production rate. No factors of the two latter studies were similar compared to the present study.

A downside of the used study design (a single cross-sectional study), which is implemented in every risk factor study on the prevalence of Y. enterocolitica (Nowak et al., 2006; Laukkanen et al., 2008, 2009; Virtanen et al., 2011; Novoslavskij et al., 2013), is that alterations due to temporality of the within-batch prevalence or to the causality of the studied farm factors cannot be identified. Further research on the variability of subsequent batches originating from the same farm should be accomplished.

Conclusions

Reducing the number of piglet suppliers, using a disinfection bath properly, and prohibiting the entrance of pets into the stable are factors that are easily implemented to lower the prevalence of Y. enterocolitica in pigs at slaughter. This way, it is possible to decrease the number of infected batches.