Salmonellosis remains one of the most frequent foodborne zoonoses in the EU, with 60,050 human cases (15.7/100,000 inhabitants) in 2021. In the last year, S. Enteritidis, S. Typhimurium, and the monophasic variant of S. Typhimurium (mST) were among the most reported serovars, with the latter two mainly associated with contaminated pork (EFSA and ECDC, 2022).

In contrast to the fowl industry, to date, few EU countries have established National Control Programs (NCPs) against pig salmonellosis. On-farm Salmonella control programs in fattening swine were expected to be initiated in all EU after Regulation EC No. 2160/2003. However, most EU countries did not implement them, likely because they were not considered cost-effective (Anonymous, 2011).

The first comprehensive Salmonella NCP for pigs in Europe was established in Sweden in the 60s (Wierup, 2006) after a major food-borne outbreak in 1953 (Lundbeck et al., 1955), which was followed later by Norway, Finland, and Denmark in 1995 (Mousing et al., 1997; Maijala et al., 2005; Lyngstad et al., 2007). All but Denmark's were focused on eradication and had bacteriological analyses as the keystone. Denmark developed its own NCP based on both bacteriological and serological analyses, and its focus was mainly on Salmonella control. In all cases, the farm-level prevalence was initially low, and strict measures were enforced when Salmonella was found. These measures included the application of economic penalties. Positive results were observed in reducing the overall Salmonella prevalence in pig carcasses but at a high cost (Anonymous, 2011).

After the success of the Scandinavian action plans and along with EU regulation, new NCPs followed suit in other European countries: Germany and United Kingdom in 2002 (Osterkorn et al., 2001; BPEX, 2002; Snary et al., 2010), Ireland in 2003 (Statutory Instrument No. 165/2002), the Netherlands in 2005 (Hanssen et al., 2007), and Belgium in 2007 (Méroc et al., 2012). In general, these programs were similar to the Danish NCP, focusing on control, but they were based mostly on serological analysis of a relatively small number of pigs per batch slaughtered. Thus, pig herds were categorized into three different risk groups: low-risk (I), medium-risk (II), and high-risk herds (III). Category III herds had to undertake farm-specific activities aimed at reducing their Salmonella exposure and, accordingly, their Salmonella seroprevalence. Although no penalties were generally applied, incentives were offered to farmers in some countries, such as to be included in pork quality assurance schemes, particularly, the Qualität und Sicherheit (QS) in Germany, the British Quality Assured Pork (BQAP) in the UK, the Bord Bía Quality Assurance Scheme in Ireland, and the IKB Nederland Varkens in The Netherlands.

Despite these efforts, there is no evidence in the scientific literature of any significant change in swine Salmonella infection reduction in pigs or in human cases related to pork consumption, and the overall Salmonella seroprevalence remains stable in pigs in many of these non-Scandinavian countries (Correia-Gomes et al., 2021). Only Germany recently reported some positive results after more than 20 years since the implementation of its program (Anonymous, 2021). Meanwhile, the United Kingdom suspended its serological monitoring in 2012 (Anonymous, 2012), and Belgium, which also suspended its serological monitoring, only has maintained veterinary advice on the control of pig salmonellosis (Anonymous, 2015).

The overall lack of efficacy and the high cost of the on-farm control of pig salmonellosis, especially for countries with large pig census (Anonymous, 2011; Gavin et al., 2018), suggest the need to revisit these NCPs. Since in the EU, pig salmonellosis is by far a public health problem, not a pig health problem, a change in the programs' main objective would be advisable. Asymptomatic Salmonella-infected pigs commonly arrive at the abattoir for slaughter [European Food Safety Authority (EFSA), 2011], and they are particularly prone to Salmonella shedding (Rostagno et al., 2010). Thus, live pigs, through their feces, are a major source of abattoir environmental Salmonella contamination, likely being the main source of carcass contamination and, consequently, pork and related products (Argüello et al., 2013a; Swart et al., 2016; Marin et al., 2020). Thus, the main objective of a NCP aimed at reducing the incidence of human salmonellosis may be to focus on finding ways to minimize Salmonella contamination in abattoirs, which in the short term, could be more cost-effective than trying to stop the infection within pig farms. Being able to predict the likelihood that a pig will shed Salmonella upon its arrival at the abattoir may be the first step to reaching this objective.

Casanova-Higes et al. (2017) observed that pigs shedding Salmonella at slaughter seroconverted earlier during the fattening period than non-shedder pigs. A subsequent study showed that on-farm serology could, to some extent, help predict the probability of a pig shedding Salmonella at slaughter, thus allowing for the prompt implementation of on-farm and slaughter interventions to reduce the likelihood of abattoir environmental contamination with Salmonella (Mainar-Jaime et al., 2018). However, since these studies were carried out on a small number of pig batches from a single Salmonella-positive farm and with no additional information, their results should be confirmed further.

Thus, the main objective of this study was to assess whether serology and other farm and/or environmental characteristics (Salmonella pen contamination, farm biosecurity, season, etc.) could be used as predictors of Salmonella shedding at the abattoir. By predicting the risk of Salmonella shedding for a given batch of pigs upon their arrival at the abattoir, subsequent carcass contamination could be prevented by implementing both on-farm and abattoir control strategies.

In this study, a total of 1,341 pigs from 30 farms that were willing to participate and were located in Northeast Spain, the largest pig production region in Spain (MAGRAMA, 2021), were selected. Even though pig salmonellosis is considered a public health concern, Salmonella contamination was detected in 43.3% of the farms, a figure comparable to that reported in 2003-2004 in a similar study on the entire country (García-Feliz et al., 2007). More concerning, the proportion of pigs shedding Salmonella at the abattoir was also high (23.6%), with some pig batches reaching up to 80% and with zoonotic S. Typhimurium and the mST being the predominant serotypes. These pigs are likely major sources of carcass contamination (Argüello et al., 2013a; Marin et al., 2020), and finding ways to prevent this shedding should be of utmost importance.

As indicated by the random-effects logistic model, the proportion of pigs shedding Salmonella at the abattoir was strongly related to the presence of Salmonella in the fattening unit from which pigs came (Table 1). This relationship was supported to some extent by the identification through PFGE analyses of genetic matches between Salmonella isolates from pen FF and pig IC samples in most of the analyzed batches (Figure 3), despite the possibility of cross-contamination during transport or lairage (Argüello et al., 2013a). These findings emphasized the role that Salmonella farm contamination plays in Salmonella shedding at slaughter and suggested that Salmonella control should begin at the farm (de Busser et al., 2013).

A thorough biosecurity questionnaire was carried out on all the pig farms to describe their overall biosecurity level. On average, they presented good biosecurity levels. The mean biosecurity score for these farms was 73%, which appears to be somewhat higher than the Spanish national average (68%, from 275 questionnaires) and even higher than the national average for other big European pig-producer countries such as Germany (64%; 180 questionnaires), The Netherlands (69%; 198 questionnaires), Italy (71%; 353 questionnaires), or Ireland (72%; 486 questionnaires). However, the score was lower than that in Belgium (75%; 10,068 questionnaires) (as checked at biocheck.ugent.be on 31 January 2023). However, despite this good level of overall biosecurity, serological results suggested that Salmonella was circulating within most farms, highlighting the difficulties in controlling it at the farm level. On average, 32.9% of the pigs presented high ELISA OD% values (≥40%). Based on these results, one-third of the farms would be classified within the high-seroprevalence category (≥40% seroprevalence), and another one-third would be classified within the medium category (between 20% and 40% seroprevalence), according to the main NCPs. Only six farms showed seroprevalence levels below 10%. These results might also explain to some extent the high proportion of pigs shedding Salmonella at the abattoir, as a significant positive but weak association was observed in the logistic model between pig serology and Salmonella shedding at slaughter (Table 1). Pigs with much higher ELISA OD% values would have somewhat higher odds of shedding Salmonella at slaughter (Kranker et al., 2003; Sørensen et al., 2004; Korsak et al., 2006; Mainar-Jaime et al., 2018).

In this study, neither the overall nor the external biosecurity scores were related to the reduction of Salmonella shedding at the abattoir (Table 1). Although the biosecurity level of farms is, in general, considered to be beneficial to reducing bacterial transmission, it appears that biosecurity cannot reduce Salmonella prevalence by itself (Alarcón et al., 2021; Youssef et al., 2021). The implementation of efficient on-farm Salmonella control measures will depend on farmers' perception of the disease and their motivation to maintain them on an ongoing basis (Fraser et al., 2010; Marier et al., 2016). However, pig salmonellosis is usually asymptomatic, that is, of low concern for farmers and swine production veterinarians.

The predictive model built with these three factors, i.e., serology, farm internal biosecurity, and Salmonella pen prevalence, showed an acceptable ability to predict whether an animal will shed Salmonella at the abattoir. Thus, according to the model, a pig with a predicted probability of shedding Salmonella greater than 26% would have a 71.6% probability of being a true shedder. If the predicted probability was ≤ 26%, the animal would have a 73.6% probability of being a true non-shedder. Therefore, estimating the proportion of animals with a model probability higher than 26% in a given batch of pigs intended for slaughter would allow for the assessment of the overall risk of shedding for that batch. Once the potential risk of Salmonella shedding has been assessed 3-4 weeks before slaughter, stakeholders could act according to the results obtained. At the farm level, control measures could be implemented to minimize the likelihood of Salmonella shedding at slaughter. For instance, the addition of organic acids in food/water could help reduce shedding (de Busser et al., 2009; Argüello et al., 2013b; Lynch et al., 2017; Bernad-Roche et al., results to be published). Additionally, the abattoir, being aware of the risk, could implement mitigation measures such as logistic slaughter (Swanenburg et al., 2001; Hotes et al., 2011) or even the addition of organic acids to the water at lairage (Bernad-Roche et al., 2022), to attempt to reduce the risk of Salmonella shedding. These interventions will only be required on those batches of pigs that present probabilities of Salmonella shedding above an established threshold.

It is interesting to note that in farms with high or medium internal biosecurity scores (i.e., scores >64%) and where no Salmonella-positive pens were detected, no animals could be considered shedders at slaughter (i.e., with predicted probability >26%), regardless of the ELISA OD% values (Figure 5). Likewise, serology would be unnecessary in the case of farms with low internal biosecurity when at least one pen is positive for Salmonella, as all pigs would show high predicted probabilities of shedding the bacterium at slaughter. In this second scenario, the likelihood of a pig becoming infected will be high, since Salmonella should be able to circulate easily among pens in farms with low internal biosecurity (Baptista et al., 2010).

These results emphasized the importance of internal biosecurity in the transmission of Salmonella within the farm. The internal biosecurity questionnaire included factors such as disease management (i.e., vaccination and treatment protocols and frequency of health status assessment), fattening unit management (i.e., all-in/all-out system and pig mix and density), measures between compartments and the use of equipment (i.e., foot baths, cleaning and disinfection after equipment usage, workflow from younger to older pigs, and sharing equipment with other farms), and cleaning and disinfection (i.e., after every production cycle, protocols, and drying after cleaning and disinfection). Cleaning and disinfection along with feed, water, and bedding are considered by experts some of the most important biosecurity measures to control Salmonella in indoor settings (De Lucia and Ostanello, 2020; Galipó et al., 2023). However, more research is required to determine with more certainty the most-effective measures from the human health perspective (Youssef et al., 2021).

In contrast to the two previous scenarios, in the case of any of the other situations (i.e., acceptable internal biosecurity with the presence of Salmonella-positive pens or low internal biosecurity and no Salmonella-positive pens), serology would add useful information to the decision of whether a pig should be considered of risk for Salmonella shedding at the abattoir. As an example, pigs with ELISA values >30% would be associated with abattoir Salmonella shedding if they belong to farms with medium internal biosecurity, along with the presence of Salmonella in 10–20% of the pens. If they belong to a similar farm but with high internal biosecurity, the pigs of risk would be those showing much higher ELISA values (close to 80%). ELISA values would also help interpret the possible shedding status of a pig from a low internal biosecurity farm when no Salmonella-positive pens are detected (Figure 5). Therefore, considering the different scenarios that are possible according to the model, the simplest way to proceed with the estimation of the risk of Salmonella shedding at slaughter would be to start with an assessment of the farms' internal biosecurity and the Salmonella pen prevalence. Depending on the results, additional serological analyses should be carried out on a representative sample of the pigs from the batch.

Although many studies have shown the lack of reliability of indirect ELISA tests for ascertaining the individual Salmonella status of a pig, either infection or shedding (Nollet et al., 2005; Farzan et al., 2007; Gradassi et al., 2015), they could be of relative value when used on groups of pigs (Sørensen et al., 2004; Korsak et al., 2006; Farzan et al., 2007). In this study, a significant but low correlation was observed between the Salmonella seroprevalence of a given batch and the proportion of shedding pigs at the abattoir for that batch (Figure 2), when no other factors were taken into account. This simple approach might not be sufficient for accurately predicting Salmonella shedding at the abattoir.

It appears that the value of serology for predicting shedding depends on the context in which it is used. Similar to the results of a previous study (Mainar-Jaime et al., 2018), the model in this study indicated that higher individual OD% values were related to higher odds of Salmonella shedding at slaughter. This was likely due to the presence of stressful factors such as the transport and waiting (lairage) times that pigs went through, which would favor the shedding among the infected pigs (Duggan et al., 2010; Simons et al., 2016). In addition, other factors, such as the internal biosecurity of the farm and the presence of Salmonella in the farm, would also play a significant role in interpreting ELISA values in this context. Thus, these results might help explain, at least in part, the difficulties that many NCPs face in properly assessing the Salmonella status of pig farms, as many of them are based exclusively on serological results usually obtained from a small and hardly representative number of animals, without considering other factors (Mainar-Jaime et al., 2018; Correia-Gomes et al., 2021).

This study highlighted the importance of the context in which serology is used for pig salmonellosis control. Using it 3-4 weeks prior to sending the pigs to slaughter might be helpful for identifying batches of pigs at risk of shedding Salmonella upon their arrival at the abattoir. In some cases, being aware of the farm's internal biosecurity level and performing a bacteriological sampling of a representative number of pens might be sufficient for estimating the risk of Salmonella shedding for a given batch of pigs ready for slaughter. In others, serology would be required for a more accurate interpretation of the results, but in both situations, an acceptable level of knowledge about the risk of Salmonella shedding for a given batch of slaughter pigs could be achieved. Reducing the likelihood of Salmonella shedding at this stage would be an important step for reducing Salmonella carcass contamination.

An additional advantage of this approach is that Salmonella control would not initially rely on the farmer's work but on the farm's data collection, as farmer's engagement seems to be one of the main obstacles NCPs face, especially when dealing with animal infections of no clinical concern (Fraser et al., 2010; Marier et al., 2016; Alarcón et al., 2021). Moreover, this approach would also allow for a combined farm/abattoir strategy that would likely have cumulative benefits (Swart et al., 2016), as it would make both abattoirs and farmers aware of the risk of the pigs coming to the slaughter. Thus, a more precise characterization of the Salmonella status of pig farms would be obtained from routine sampling, with the collection of proper representative samples, which would also help encourage a good attitude among farmers toward Salmonella control in the short/medium term.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The animal study was approved by Ethical Advisory Commission for Animal Experimentation of the University of Zaragoza (permit no. PI13/20). The study was conducted in accordance with the local legislation and institutional requirements.

RM-J: conceptualization, supervision, project administration, and funding acquisition. RM-J and MB-R: methodology, investigation, and writing-original draft preparation. RM-J, MB-R, IB, and AC-S: formal analysis. MB-R, RM-J, and CM-A: writing-review and editing. All authors have read and agreed to the published version of the manuscript.