SEARCH

SEARCH BY CITATION

Keywords:

  • Salmonella;
  • pig;
  • factor analysis;
  • herd classification;
  • biosecurity

Summary

  1. Top of page
  2. Summary
  3. Impacts
  4. Introduction
  5. Material and Methods
  6. Results
  7. Discussion
  8. Conclusions
  9. Acknowledgements
  10. Funding
  11. References

Salmonella surveillance-and-control programs in pigs are highly resource demanding, so alternative cost-effective approaches are desirable. The aim of this study was to develop and evaluate a tool for predicting the Salmonella test status in pig herds based on herd information collected from 108 industrial farrow-to-finish pig herds in Portugal. A questionnaire including known risk factors for Salmonella was used. A factor analysis model was developed to identify relevant factors that were then tested for association with Salmonella status. Three factors were identified and labelled: general biosecurity (factor 1), herd size (factor 2) and sanitary gap implementation (factor 3). Based on the loadings in factor 1 and factor 3, herds were classified according to their biosecurity practices. In total, 59% of the herds had a good level of biosecurity (interpreted as a loading below zero in factor 1) and 37% of the farms had good biosecurity and implemented sanitary gap (loading below zero in factor 1 and loading above zero in factor 3). This implied that they, among other things, implemented preventive measures for visitors and workers entering the herd, controlled biological vectors, had hygiene procedures in place, water quality assessment, and sanitary gap in the fattening and growing sections. In total, 50 herds were tested for Salmonella. Logistic regression analysis showed that factor 1 was significantly associated with Salmonella test status (= 0.04). Herds with poor biosecurity had a higher probability of testing Salmonella positive compared with herds with good biosecurity. This study shows the potential for using herd information to classify herds according to their Salmonella status in the absence of good testing options. The method might be used as a potentially cost-effective tool for future development of risk-based approaches to surveillance, targeting interventions to high-risk herds or differentiating sampling strategies in herds with different levels of infection.


Impacts

  1. Top of page
  2. Summary
  3. Impacts
  4. Introduction
  5. Material and Methods
  6. Results
  7. Discussion
  8. Conclusions
  9. Acknowledgements
  10. Funding
  11. References
  • • 
    Salmonellosis is the second most commonly reported foodborne zoonoses in the European Union and pork is considered to be a significant source. In response, a target for reduction of Salmonella in pig production will be set.
  • • 
    Current surveillance of Salmonella in pigs involves intensive and expensive sampling schemes for herd classification. This study presents a potential alternative for herd classification based on herd information.
  • • 
    Herds with poorer biosecurity had a higher probability of testing Salmonella positive compared with herds with good biosecurity. Results suggest that this herd classification approach can be used for risk-based Salmonella surveillance.

Introduction

  1. Top of page
  2. Summary
  3. Impacts
  4. Introduction
  5. Material and Methods
  6. Results
  7. Discussion
  8. Conclusions
  9. Acknowledgements
  10. Funding
  11. References

Over the recent years, salmonellosis has been the second most commonly reported zoonoses in the European Union (EU), accounting for 151 995 recorded human cases in 2007 (European Food Safety Authority, 2009a). Reduction of Salmonella enterica subsp. enterica (Salmonella) prevalence in the pig industry will be set as a target at the EU level possibly in 2011 and it is believed to significantly contribute to the protection of human health. The specific reduction target will be based upon the results of a quantitative microbiological risk assessment on Salmonella in slaughter and breeder pigs1 as well as cost-benefit analyses, all conducted at the EU level. According to the Regulation EC-2160/20032, protection of human health from food-borne zoonotic agents is an issue of paramount importance. Farm-to-fork control programs will probably be needed to ensure a reduction of the prevalence of specified zoonoses and zoonotic agents. Moreover, Member States will have the responsibility to establish effective national control programs adjusted for the country-specific characteristics, including the disease burden and the financial implications for stakeholders. Results of the EU baseline survey on the prevalence of Salmonella in lymph nodes of slaughter pigs showed a wide range of prevalences from 0% to 29% infected pigs among EU countries (European Food Safety Authority, 2008). These findings suggest that country-tailored surveillance-and-control strategies should be designed aiming to achieve the targets in a cost-effective way, assuring human-health protection.

Salmonella surveillance-and-control programs in pigs are already implemented in several countries and even though different approaches have been used, they mostly rely on intensive sampling schemes (Alban et al., 2002; Blaha, 2004; Bengtsson et al., 2009; Cortiñas Abrahantes et al., 2009). Given the high demand for human and financial resources for the maintenance of such surveillance programs, alternative approaches based on cost-effective indicators should be further investigated. Livestock data have previously been used to classify cattle herds according to the risk profile for disease presence (Ortiz-Pelaez and Pfeiffer, 2008). However, to our knowledge no study was conducted to evaluate the usefulness of herd information to classify pig herds according to the Salmonella status.

Herd-level risk factors for Salmonella in pigs have been described in the literature and biosecurity measures are of major significance (Funk and Gebreyes, 2004; Lo Fo Wong et al., 2004; Fosse et al., 2009). These include: controlled access to the herd for workers, visitors and vehicles, general hygiene requirements for people, equipment and facilities, all-in/all-out management and control of potential biological vectors, such as rodents, cats and dogs. Information on Salmonella risk factors might be used to identify high-risk herds for which specific surveillance-and-control strategies might be targeted. This study presents an approach that can be used as part of a risk assessment for herd classification for the development of alternative risk-based surveillance programs for Salmonella in pigs. Such methods may be particularly interesting for countries that have not yet implemented surveillance programmes for Salmonella in pigs. This study aimed at developing a tool for predicting the Salmonella status in Portuguese pig herds based on herd information collected from a questionnaire survey.

Material and Methods

  1. Top of page
  2. Summary
  3. Impacts
  4. Introduction
  5. Material and Methods
  6. Results
  7. Discussion
  8. Conclusions
  9. Acknowledgements
  10. Funding
  11. References

Questionnaire survey

A cross-sectional questionnaire study was conducted to collect information from the target population – industrial pig herds in Portugal. In Portugal, herds with ≥20 sows or ≥200 fattening pigs are classified as industrial herds. To obtain as much information as possible within 4 months, a convenience sample of 250 herds was selected. From September to December 2008, questionnaires were distributed to veterinarians working in industrial herds located in Lisbon and Tagus Valley region (source population). Contact to the veterinarians was provided by the Scientific Society of Swine Production (Sociedade Científica de Suinicultura). Additionally, veterinarians were asked to provide contacts of other colleagues that were not affiliated in this Society. A total of 74 veterinarians were contacted but only 50 were willing to participate in the questionnaire survey. Veterinarians received no payment for filling out the questionnaire. The aim of the survey was presented by the first author to all participating veterinarians. A covering letter was presented with the questionnaire, explaining the rationale of the study and assuring the confidentiality of the interviewed, emphasizing that the results would not be used for official purposes and that the responses should disclose the actual practices in the herd and not the intended. Follow-up reminder phone calls were made to the participating veterinarians that did not return questionnaires after 4 weeks.

The questionnaire assembled 77 questions under six main topics: herd characteristics, production parameters, hygiene, feeding, management and health. The full questionnaire in Portuguese and translated into English are available from the corresponding author. Pre-testing to identify misleading questions was performed before using the questionnaire in the study population. This included an informal stage where the questionnaire was presented to nine swine experts. Formal pre-testing was conducted in three pig herds to attain the final version of the questionnaire. The general questionnaire layout was designed to assure an easy comprehension and clear data entry. Question sequence was carefully addressed to avoid the phenomenon of ‘carry over’. For reliability purposes, pairs of questions capturing the same information were included in different parts of the questionnaire (Thrusfield, 2007). One question related to all-in/all-out management was excluded from the analysis because it was unclear if it was implemented at the pen or at the unit level. Sanitary gap was defined as the period of time a pen was left empty, after moving out all pigs and before introducing a new group of pigs.

Salmonella data

Three sources of data were used to establish the Salmonella status of the pig herds in the study
  • 1
     Data from a baseline study on the prevalence of Salmonella in slaughter pigs in the EU were obtained for herds included in the questionnaire survey (= 37 herds). In the baseline study, slaughtered pigs with a live weight between 50 and 170 kg were randomly selected from February to September 2007 at abattoirs accounting for at least 80% of all slaughtered fattening pigs (European Food Safety Authority EFSA, 2008). The number of pigs tested per herd ranged from one to six pigs. Five ileo-caecal lymph nodes weighing at least 15 g were collected from each selected pig.
  • 2
     A cross-sectional field study was conducted from April to July 2009 on 20 of the herds included in the questionnaire survey. Pooled faecal samples were collected from randomly selected pens housing 6- to 10-week old pigs. One pooled sample consisted of 25 g of faecal material collected from five different places of the main dunging areas of the pen. At each pen a minimum of 10 pigs were present. At each herd, 4–10 pooled samples were collected.
  • 3
    Approximately 10 weeks after the pen sampling, blood samples were collected from 10 pigs from the same group where faecal samples were previously collected.

Pooled faecal and blood samples were collected from age groups where shedding and antibody levels have been reported to be at its maximum, respectively 6- to 10-week-old pigs and 16- to 20-week-old pigs (Kranker et al., 2003; Arnold et al., 2005). One herd went out of business during the study period and, hence, blood samples could not be collected. Seven herds were both included in the baseline and the field study. Therefore, Salmonella test results were available on a total of 50 herds.

A herd was considered test positive if (i) at least one lymph node or pooled faecal sample tested positive for Salmonella, and/or (ii) blood samples from at least one of the 10 tested pigs had an optical density percentage (OD%) above 20. A herd was considered test negative if none of the samples tested positive for Salmonella-bacteria or antibodies.

Bacteriological examination

Before analysis, lymph nodes were decontaminated by dipping into absolute alcohol and drying by air. After homogenization, lymph nodes were inoculated in buffered peptone water in dilution 1 : 10 and allowed to incubate for a total of 18 ± 2 h at 37 ± 1°C. For the pooled faecal samples, after homogenization, 25 g were inoculated in buffered peptone water in dilution 1 : 10 and allowed to incubate for a total of 18 ± 2 h at 37 ± 1°C. For lymph nodes and pooled faecal samples, bacteriological examination was then conducted according to the method described in Annex D of ISO 6579:20023: ‘Detection of Salmonella spp. in animal faeces and in environmental samples of the primary production stage’.

Serological examination

Blood samples were allowed to coagulate for 12–24 h at 4°C and centrifuged for 10 min at 4500 g. Serum was stored at −20°C until use. Serum samples were tested for presence of antibodies against Salmonella using Salmotype® Pigscreen ELISA (Svanova Biotech AB, Uppsala, Sweden), according to the instructions of the test kit. Results were given as OD% and a 20 OD% cut-off was used, as recommended by the manufacturer.

Statistical analysis

Questionnaire responses were coded and entered into a Microsoft Access® database. Improved data accuracy was assured by double data entry and implementation of validation criteria for computer data entry. Data analysis was performed in sas® v. 9.1.3. (SAS Institute Inc., Cary, NC, USA).

Only variables related to farrow-to-finish herds (= 108) were further analysed due to the small number of responding herds from other farm types. The number of sows present in the herd was used as an indicator of herd size. Herds with ≤250 sows were classified as small; herds with >250 sows were classified as large. Fisher’s exact test and Mann–Whitney test were used to identify differences between herd size, for dichotomous and continuous variables respectively (< 0.05). Differences regarding herd size between responding and non-participating herds were also evaluated.

The PRINQUAL procedure in sas was used to transform categorical variables by optimal scoring of the categories (Fisher, 1938). Such transformation allowed for using dichotomous variables in the factor analysis (SAS Institute Inc., 2008). After careful inspection, transformed variables and the original interval-scale variables were used in the factor analysis (FACTOR procedure in sas). An iterated principal factor analysis model with orthogonal rotation (option VARIMAX in sas) was used to identify underlying factors that could explain intercorrelation among the different variables. In an orthogonal factor model, the variable loading indicates the correlation between the variable and the underlying factor and it can be used to assess the extent to which a given variable measures the factor (Sharma, 1996). Visual inspection of the correlation matrix and the Kaiser–Meyer–Olkin (KMO) measure of the homogeneity of the variables were used to evaluate sampling adequacy. According to Sharma (1996), high correlations between the variables and a KMO > 0.6 are considered acceptable. The interpretability of the factors and a scree plot were the criteria used for determining the number of factors. Variable loadings (>|0.40|) indicating influential variables in each factor and plausibility according to the authors’ knowledge were the criteria used for interpreting the factors. To evaluate the estimated factor solution, the residual correlation matrix was checked and the root mean square residual (RMSR) measure was calculated. The RMSR and the residual correlations should be as small as possible indicating an appropriate factor solution (Sharma, 1996). Regarding sample size, according to Hatcher (1994) a minimal sample of 100 observations or a minimum of five observations per variables should be used.

Logistic regression analysis was used to assess the association between herd Salmonella status and the factors previously identified in the factor analysis model. The logistic regression model was fitted in sas (GENMOD procedure). The binary-dependent variable was herd Salmonella test result (positive/negative). The following factors were included as independent variables: factor 1 (general biosecurity), factor 2 (herd size) and factor 3 (sanitary gap implementation). The identification of the significant variables and interactions (< 0.05) was performed by backward elimination. Confounding was assessed by adding potential confounding variables to the model and comparing the crude and adjusted estimates. The estimated dispersion parameter was used to assess goodness-of-fit of the final model.

Results

  1. Top of page
  2. Summary
  3. Impacts
  4. Introduction
  5. Material and Methods
  6. Results
  7. Discussion
  8. Conclusions
  9. Acknowledgements
  10. Funding
  11. References

A total of 139 questionnaires were received, corresponding to a response rate of 56%, of which 136 were located in the target region. One questionnaire was left out of the analysis because it was not an industrial herd, according to the official Portuguese definition. Of the 135 questionnaires, 108 were farrow-to-finish herds, 12 farrow-to-weaner herds and 15 farrow-to-grower herds. Descriptive data from the 108 industrial farrow-to-finish pig herds used for further analyses are presented in Table 1.

Table 1.   Distribution of herd dichotomous variables used in the factor analysis model of data from 108 Portuguese industrial farrow-to-finish pig herds surveyed from September to December 2008
VariablesLevelsNo. herdsa (%)
  1. aSome questions were not answered. Thus, not all variables sum to 108.

Other livestock species present in the herdYes72 (66.7)
No36 (33.3)
Dogs/cats present in the herdYes51 (47.2)
No57 (52.8)
Each pen has its own slurry drainYes52 (48.1)
No56 (51.9)
Footbath at the buildings entranceYes40 (37.4)
No67 (62.6)
Shower when entering the herdYes24 (22.2)
No84 (77.8)
Footwear for visitorsYes91 (85.0)
No16 (15.0)
Sanitary gap farrowingYes101 (93.5)
No7 (6.5)
Sanitary gap growingYes100 (92.6)
No8 (7.4)
Sanitary gap fatteningYes82 (75.9)
No26 (24.1)
Clean and disinfect the loading bay after loading the pigsYes77 (72.0)
No30 (28.0)
Dogs/cats have access to the herd clean areaYes39 (36.1)
No69 (63.9)
Dogs/cats have access to the buildingsYes24 (22.2)
No84 (77.8)
Rodent control plan implementedYes83 (76.9)
No25 (23.1)
Insect control plan implementedYes76 (70.4)
No32 (29.6)
Water testedYes62 (59.0)
No43 (41.0)
People dedicated to specific unitsYes41 (38.3)
No66 (61.7)
Exclusively home-bred breedersYes No44 (40.7)
64 (59.3)
Small/weak pigs kept back to mix with younger pigsYes32 (29.9)
No75 (70.1)
Transporting pigs driver enters the herd clean areaYes22 (20.6)
No85 (79.4)

Differences between respondents and herds not participating in the survey were tested. The number of sows in the responding herds was significantly higher than in the non-participating herds (< 0.01). When only responding herds classified as small were compared with non-participants, no difference in herd size was found (= 0.7).

Differences in practices between herd size groups

The following practices were significantly more common in large herds than small herds (< 0.05): people dedicated to specific units; larger number of pigs per pen; larger average number of weaners/sow/year; pigs kept in outdoor units; washing slurry drains after every batch; implementation of sanitary gap in the fattening sector; water testing; the driver transporting pigs does not enter the clean area; transporting vehicles do not unload feed in bulk in the herd clean area; longer average duration of veterinary visits. However, the duration of veterinary visits was not linearly correlated to the number of sows in the herds.

Factor analysis model results

Because of missing observations for some of the variables of interest, six herds were excluded from the analysis. A three-factor model explaining 44.5% of the variation in the data was developed (Table 2). The percentage of variance explained by the factors was 26.4% (factor 1), 9.5% (factor 2) and 8.6% (factor 3).

Table 2.   Rotated factor pattern for a factor analysis model of data from 102 Portuguese industrial farrow-to-finish pig herds surveyed from September to December 2008a
VariablesLoadings
Factor 1Factor 2Factor 3
  1. Factor loadings > |0.40| are highlighted in bold.

  2. aOf the 108 farrow-to-finish herds included in the survey, six were excluded from the factor analysis model because of missing observations.

Other livestock species present in the herd0.460.090.11
Dogs/cats present in the herd0.580.050.04
Each pen has each own slurry drain−0.490.260.23
Footbath at the buildings entrance−0.490.260.18
Shower when entering the herd−0.440.000.13
Footwear for visitors−0.540.340.14
Sanitary gap farrowing0.120.040.66
Sanitary gap growing0.070.040.53
Sanitary gap fattening0.130.290.16
Clean and disinfect the loading bay after loading the pigs−0.550.390.13
Dogs/cats have access to the herd clean area0.780.020.04
Dogs/cats have access to the buildings0.710.050.04
Rodent control plan implemented−0.510.370.19
Insect control plan implemented−0.450.300.21
No. sows in the herd0.010.740.04
Water tested−0.470.250.29
People dedicated to specific units0.090.640.22
Small/weak pigs kept back to mix with younger pigs0.380.280.22
Transporting pigs driver enters the herd clean area0.39−0.500.07
Breeders replacement0.290.000.04
No. veterinary visits per year0.130.370.19

The three factors can be described as follows (a site with a low loading can be described as the opposite):

  • factor 1:
     scoring high on factor 1 was consistent with poor biosecurity and was characterized by the following practices - other livestock species being present in the herd; dogs/cats being present in the herd; no slurry drain for each pen; no footbaths at the building entrances; no shower when entering the herd; no footwear available for visitors; no cleaning and disinfection of the loading bay after loading the pigs; dogs/cats having access to the herd clean area; dogs/cats having access to the building interior; no rodent control plan implemented; no insect control implemented; no water tested.
  • factor 2:
     scoring high on factor 2 was related to large herd size and included a high number of sows in the herd; people dedicated to specific units; the driver transporting pigs does not enter the clean area in the herd.
  • factor 3:
     scoring high on factor 3 was consistent with having sanitary gap implemented in the farrowing and growing sectors.

Factor 1 and factor 3 were further used as scales to classify herds according to their biosecurity profile (Fig. 1). Factor 2 was included to classify herds according to herd size (Fig. 2).

image

Figure 1.  Results of a factor analysis model presented as a plot of the 102 Portuguese industrial farrow-to-finish pig herds surveyed for herd information, 2008 – factor 1 versus factor 3.

Download figure to PowerPoint

image

Figure 2.  Results of a factor analysis model presented as a plot of the 102 Portuguese industrial farrow-to-finish pig herds surveyed for herd information, 2008 – factor 1 versus factor 2.

Download figure to PowerPoint

Factor analysis model validity

A total of 21 variables were included in the factor analysis model because the number of variables should be related to the sample size and, as a ‘rule of thumb’, a minimum of five observations per variable is generally recommended (Hatcher, 1994). Variables included in the model were selected based on their biosecurity importance in the specific Portuguese scenario. Different models were developed with different sets of variables and herds (data not shown) and no major differences were found suggesting that the model is robust.

For our dataset, a KMO measure of 0.8 suggested that the correlation matrix was appropriate for factor analysis. For our residual matrix, a RMSR of 0.07 was computed, implying an acceptable factor solution (Sharma, 1996). The scree plot along with the interpretability of the factors indicated that a three-factor solution would be acceptable (Fig. 3).

image

Figure 3.  Screen plot of the eigenvalues against the number of factors for the factor analysis model including 102 Portuguese industrial farrow-to-finish pig herds surveyed for herd information, 2008.

Download figure to PowerPoint

Herd classification model

Overall, 59% of the 102 herds included in the factor analysis model had good general biosecurity (interpreted as a loading below zero on factor 1). The median herd size in the smaller herd category was 200 sows (interpreted as a loading below zero in factor 2) and in the larger herds was 350 sows (interpreted as a loading above zero in factor 2). However, it should be noted that the herd loading is a combination of all variables loading on a factor. For that reason, seven large herds with no people dedicated to specific units were plotted in the ‘smaller herds’ category. Approximately the same proportion of smaller and larger herds was assigned to the good and poor general biosecurity category (Fig. 2). In total, 37% of the herds both had a good level of general biosecurity and implemented a sanitary gap period in the farrowing and growing sectors, interpreted as a loading below zero in factor 1 and loading above zero in factor 3 (Fig. 1). Factors 1, 2 and 3 loadings were further used to evaluate the usefulness of the factors as indicators of the herd Salmonella test status.

Of the 37 herds included in the baseline study and in the questionnaire survey from which lymph nodes were collected, 15 (41%) were test positive for Salmonella; from the 20 herds included in the field study from which pooled faecal samples were collected, five (25%) tested positive; from the 19 herds from which blood samples were collected, 18 (95%) were positive. Overall, of the total 50 herds from which Salmonella data were available, 30 herds tested Salmonella positive (60%).

According to the final logistic regression model, factor 1 was significantly associated with the herd Salmonella test status (= 0.04). The higher the score in factor 1, the higher the probability of a herd testing Salmonella positive. There were no statistically significant associations with factor 2 or factor 3 (> 0.05). The scaled deviance and Pearson’ chi-squared statistics were 1.3 and 1.1, respectively, indicating that the model fitted the data well. No interactions or confounding was found in the logistic regression model.

Discussion

  1. Top of page
  2. Summary
  3. Impacts
  4. Introduction
  5. Material and Methods
  6. Results
  7. Discussion
  8. Conclusions
  9. Acknowledgements
  10. Funding
  11. References

Questionnaire survey

In 2007, the Portuguese pig industry comprised a total of 2 374 000 pigs (Statistics Portugal, 2008). Lisbon and Tagus Valley region was chosen for the questionnaire survey because it represented 46% of the total number of pigs in the country, in 2008. In 2008, in Lisbon and Tagus Valley region, industrial pig herds accounted for 52.8% of the total number of herds (= 1285 total number of herds) and 94.0% of the total number of pigs in this region (= 1 122 696 total number of pigs). Thus, this region was assumed to be representative of most industrial pig herds in Portugal. However, it is not unlikely that there are pig herds that differ substantially from these in other parts of the country.

A mailed questionnaire survey was chosen as a result of practical and economical constraints that made it impossible to visit all farms. Furthermore, most of the questions were related to herd routines that could not be directly observed during a herd visit. Apart from the 50 veterinarians that participated in the survey, 74 veterinarians were contacted but were not willing to participate. The most common justifications were: ‘lack of time’, ‘no longer working for pig herds because they went out of businesses’, and ‘working in regions of the country outside the survey area’. Some selection bias in the data collection might have arisen because of the lack of participation from these veterinarians, particularly for those who pointed ‘lack of time’ as the reason for declining participation. This might indicate that the veterinarians that participated in the survey are more aware of the importance of biosecurity than in general.

Difference in herd size found among respondents and non-participating herds might indicate another source of selection bias. Hence, the findings presented in this study might accurately reflect practices in larger herds, which are the ones expected to contribute the most for the infection load of Salmonella to abattoirs and thus the most important to handle with regard to risk.

Herd information

There was large variation between herd characteristics, for instance herd size (counted as number of sows present) was recorded in the range of 40–1200.

Shower when entering the herd and the presence of footbaths at the entrance of the buildings were not systematically implemented (22% and 37% of the herds, respectively). Footbaths can be effective in reducing the spread of Salmonella and other pathogens within the herd if manure is previously removed and boots are soaked for the time recommended on the disinfectant label (Amass et al., 2000).

Dogs, cats and other livestock species were commonly present in the surveyed herds, suggesting that farmers are not aware of the potential role of animals as biological vectors of many diseases, including Salmonella (Funk et al., 2001a; Barber et al., 2002).

Almost all herds had a private water source (deep water; data not shown). Still, water was not tested in 41% of the herds, indicating that in these herds water quality was not addressed as a key component of herd biosecurity. Both ground and surface water can become contaminated with Salmonella and other bacteria, chemicals and heavy metals posing a risk to pig health (Ahmed et al., 2009).

Closed herd management was practiced in 41% of the herds; the rest of the herds mainly purchased breeders from only one source herd. These are considered to be effective measures reducing the risk of Salmonella introduction by incoming infected stock (Lo Fo Wong et al., 2004).

Transporting pig drivers entered the clean area of the herd in 21% of the herds, pointing out a potential route of Salmonella introduction.

Participants were asked to draw a figure of the farm housing facilities (data not shown). Hereby, it was clear that, in most of the herds, buildings and unit location did not allow for a one-way flow of pigs and all-in/all-out management. This situation might be explained by the evolution of the pig industry in Portugal. A large number of farms have evolved from small family businesses and expanded progressively into industrial herds. This scenario poses a challenge for the implementation of strict biosecurity measures in Portuguese pig herds.

Differences in practices between herd size groups

For most of the variables collected in the questionnaire no difference was found regarding herd size. Still, some variables indicated that biosecurity measures in larger herds were better implemented than in smaller herds (e.g. people dedicated to specific units, implementation of sanitary gap in the fattening sector; water testing; controlled access of drivers and vehicles entering the herd). If smaller herds (20 ≤ sows < 40) had been included in the survey, more differences between herds in the study were expected to be found. Larger herds kept pigs in open buildings more frequently than smaller herds. This might be explained by the need to expand. Because of economical constraints some of the new pig premises (particularly for weaners) were built as open buildings.

Salmonella data

No surveillance program for Salmonella is currently in place in Portugal. It was therefore necessary to make use of survey data collected for other purposes. Three different sources of data were available and used to determine the herd Salmonella test status, reflecting different exposure and infection statuses. Available data were used as an indicator of Salmonella at herd level and not at individual animal level.

Overall, 30 herds (60%) from which Salmonella data were available tested positive for Salmonella. Taking into account the lack of test sensitivity this is in agreement with the baseline results for Portugal that estimated 23% (95% CI: 19–28) prevalence of slaughter pigs infected with Salmonella measured by lymph node bacteriology at slaughter (European Food Safety Authority (EFSA), 2008). In Portuguese breeding holdings, 46% (95% CI: 39%–54%) prevalence of Salmonella positive herds was estimated in another EU baseline survey (European Food Safety Authority, 2009b). In a study on the occurrence of Salmonella in 101 Portuguese pigs, Salmonella was identified on 13% of the carcasses tested and in 19% of the ileocolic lymph nodes (Vieira-Pinto et al., 2005). The relatively high proportion of Salmonella test positive herds found in both in this study and other studies might be explained by the lack of specific mitigation measures for Salmonella in Portuguese herds, such as liquid feeding and the use of acidified feed. Feed-related factors have been systematically reported to be important risk factors for Salmonella spread in pig herds. Liquid feeding compared with dry-feed is associated with a significant decrease of Salmonella prevalence (van der Wolf et al., 1999; Beloeil et al., 2004; Lo Fo Wong et al., 2004; Bahnson et al., 2006; Farzan et al., 2006). According to van Winsen et al. (2001) fermented feed reduces the levels of Enterobacteriaceae in the gastrointestinal tract of pigs. The addition of organic acids to feed or water has been found to be associated with lower within-herd Salmonella prevalence among finishing pigs (van der Wolf et al., 2001a,b; Creus et al., 2007).

Hence, most herds in Portugal are likely to be infected with Salmonella, but only herds with high levels of infection are likely to test positive with the fairly insensitive test methods and sample sizes available in this study. Thus, if a herd tested positive for Salmonella it was considered to be a high-risk herd and a medium-risk herd if none of the samples tested positive for Salmonella-bacteria or antibodies.

Isolation of Salmonella from lymph nodes is believed to reflect long-term exposure at herd level, but might also indicate infection during transport and lairage, while positive faecal samples indicate active excretion. Bacteriological techniques for Salmonella are reported to have very high specificity (up to 100%) but sensitivity is low (Funk et al., 2000). In this study, pooled faecal samples were collected from pigs aged 6–10 weeks when shedding is reported to be at its maximum, aiming to increase herd sensitivity (Kranker et al., 2003; Arnold et al., 2005).

ELISA detects specific antibodies against Salmonella and therefore it indicates past or recent exposure and different serologic stages. Blood samples were collected when antibody levels have been reported to peak at herd level (Beloeil et al., 2003). Salmotype® ELISA test sensitivity and specificity at individual animal level was reported to be 87% and 99%, respectively (Harris, 2003).

Among Salmonella test negative herds, some might have been misclassified. More herds would be expected to be positive if more samples had been collected. To improve herd test sensitivity, more sampling of those herds that were only in the baseline study would have been desirable. Of the 37 herds that were included in the baseline study, only one pig was tested in 18 herds. Of these, four herds tested positive and three were also included in the field study. It is likely that 11 herds might have been misclassified (false-negative). Furthermore, the time difference between the baseline study and the field study and questionnaire survey may have led to herd misclassification. We included the baseline study results to increase sample size, assuming that herds often present long-term Salmonella infections (van der Wolf et al., 2001c; Baptista et al., 2009) and thus were likely to still be infected in 2008 if they were positive in the baseline study in 2007.

Herd classification model

Most of the variables which had loaded on factor 1 and factor 3 have been previously described as risk factors for Salmonella in pig herds. Still, these variables might also be used as indicators of general awareness and attitude towards biosecurity, as it has been previously suggested (Lo Fo Wong et al., 2004). Accordingly, logistic regression analysis showed that herds with poorer biosecurity presented a significantly higher probability of testing Salmonella positive (high-risk herds), whereas herds with good biosecurity were less likely to be positive for Salmonella, but should still be considered medium-risk herds. Study results also showed that herd size (factor 2) was not significantly associated with the herd Salmonella status. Sanitary gap implementation allows for all-in/all-out management which is believed to be an important practice to prevent cross-contamination and implement proper cleaning and disinfection between batches. However, in this study implementation of sanitary gap in the growing and fattening sectors (factor 3) did not seem to be an important factor for Salmonella. This may partly be due to a small sample size but might also be explained by lack of efficient cleaning and disinfection procedures and residual Salmonella contamination of pig facilities, as it was previously described (Funk et al., 2001b). Moreover, in the surveyed herds, detergent was only used in less than 19% of the farms (data not shown), which might significantly reduce the efficacy of disinfection.

Univariable logistic regression was also used to evaluate the association of questionnaire survey variables with the herd Salmonella status (data not shown). Of all the variables evaluated only ‘dogs/cats have access to the herd clean area’ was found to be significant (= 0.005). This strongly suggests that multiple biosecurity variables applied simultaneously are necessary to prevent introduction and spread of Salmonella. A standard regression model would require a much larger dataset to study the concurrent effect of several variables. However, that was not the aim of the study. This study aimed at evaluating the usefulness of available knowledge about Salmonella risk factors to classify herds according to their Salmonella test status. Factor analysis is a statistical technique that can be used to develop scales to measure unobservable features, such as biosecurity (Sharma, 1996). In presence of multicollinearity, which is often the case with biosecurity-related variables, factor analysis can be used to identify underlying factors that best account for the correlation among the variables. The eigenvalue-greater-than-one rule is one of the criteria that might be used to identify the factor solution. In this study, it suggested that eight factors should be extracted, accounting for 72% of the variation. However, the eight factors could not be interpreted and so it was decided to keep the three factors supported by the scree plot and the interpretability of the resulting factors. Still, the main objective of factor analysis is not to account for the total variation of the data but to explain the multicollinearity among the variables, through the identification of factors. Along with other parameters used to assess model validity we found this three factor solution to be suitable to understand and evaluate biosecurity in Portuguese pig herds and to use in further analysis. Based on the scores of the resulting factor analysis model, herds were plotted against the factors. Visual inspection was conducted to identify outliers and influential herds and the raw questionnaire data were checked. The analysis was conducted by excluding these herds and no significant changes were observed. Therefore, all 102 herds were kept in the final dataset.

Study results indicate that upgrading general biosecurity in Portuguese industrial pig herds could improve the Salmonella herd status, improving both the ability to lower the within-herd prevalence and the ability to keep out infection. Furthermore, it is largely accepted that interventions for Salmonella control will also decrease the prevalence of other pathogens.

The non-compliance with biosecurity measures might be attributed to several reasons, such as lack of motivation (Ellis-Iversen et al., 2010). A study conducted in the United Kingdom revealed that farmers are not taking systematic actions regarding biosecurity and appear not to rely on official sources of information (Heffernan et al., 2008). The same study also suggests that actions taken at local level with a high commitment of farmers are expected to be more effective. The implementation of a quality certification scheme giving enrolled farmers a higher value for their products and assuring them access to external markets could also act as an important driver. Consequently, some European countries have already developed standards for the production of pigs and pork, including welfare, food safety and traceability requirements, accredited to the European Product Certification Standard EN45011 (Danish Meat Association, 2007; Assured British Pigs, 2008).

Despite data constraints, results presented in this study indicate that herd information about biosecurity might be used as part of a risk assessment to classify herds according to Salmonella risk. This could be used as part of a control program for Salmonella in pigs, targeting interventions to high-risk herds or differentiating sampling and control options between medium and high-risk herds. Still, further studies should be conducted on a larger number of herds to gain more insight on the potential value of the tool presented here. Annual collection of data describing herd characteristics and biosecurity practices might be a cost-effective indicator of the Salmonella herd status, compared with traditional surveillance approaches. In line, annual collection of samples will also be required for EU member states – including Portugal – to document whether the EU target for Salmonella in pigs and/or pork is reached or not.

Conclusions

  1. Top of page
  2. Summary
  3. Impacts
  4. Introduction
  5. Material and Methods
  6. Results
  7. Discussion
  8. Conclusions
  9. Acknowledgements
  10. Funding
  11. References

We suggest an approach to classify pig herds according to their Salmonella status based on herd information about biosecurity measures. A three-factor model explaining 44.5% of the correlation among different herd variables enabled to classify Portuguese farrow-to-finish pig herds. Two factors were interpreted as being related to general biosecurity and sanitary gap implementation and one factor was interpreted as being related to herd size.

Overall, no clear differences were found between small and large herds regarding the implementation of biosecurity measures. Despite Salmonella data constraints, results suggest that herds with poorer biosecurity status presented a higher risk for Salmonella than the remaining herds. The classification model presented in this study might be used as a cost-effective indicator of the Salmonella status in pig herds. Moreover, it might be used to prioritise surveillance activities or differentiating sampling and control methods in different risk groups.

Footnotes
  • 1

    Scientific opinion on ‘Quantitative Microbiological Risk Assessment on Salmonella in slaughter and breeder pigs’ (EFSA-Q-2006-176).

  • 2

    Regulation (EC) No 2160/2003 of the European Parliament and of the Council of 17 November 2003 on the control of Salmonella and other specified food-borne zoonotic agents. Official Journal of the European Union, L 325.

  • 3

    International Organization for Standardization (ISO) 6579:2002 – Microbiology of food and animal feeding stuffs – Horizontal method for the detection of Salmonella spp.

Acknowledgements

  1. Top of page
  2. Summary
  3. Impacts
  4. Introduction
  5. Material and Methods
  6. Results
  7. Discussion
  8. Conclusions
  9. Acknowledgements
  10. Funding
  11. References

Sociedade Científica de Suinicultura is acknowledged for providing veterinarians contacts. Mário Melo (Faculdade de Medicina Veterinária) and Fragoso de Almeida (Direcção de Serviços Veterinários da Região de Lisboa e Vale do Tejo) are acknowledged for providing data and discussing issues related to the survey design and biosecurity in Portuguese pig herds. Andrea Cara D’Anjo (Direcção Geral de Veterinária) is acknowledged for providing the baseline study data. Solange Gil (Faculdade de Medicina Veterinária) is acknowledged for conducting serological testing. Sérgio Gouveia is acknowledged for collecting the samples in the pig herds. The authors also thank all veterinarians that participated in this study and Idalina Camões for helping with the computer data entering.

Funding

  1. Top of page
  2. Summary
  3. Impacts
  4. Introduction
  5. Material and Methods
  6. Results
  7. Discussion
  8. Conclusions
  9. Acknowledgements
  10. Funding
  11. References

Faecal and blood samples bacteriological and serological testing was supported by the Interdisciplinary Centre of Research in Animal Health (Project CIISA 83/2010), Faculdade de Medicina Veterinária, TULisbon, Lisboa, Portugal.

References

  1. Top of page
  2. Summary
  3. Impacts
  4. Introduction
  5. Material and Methods
  6. Results
  7. Discussion
  8. Conclusions
  9. Acknowledgements
  10. Funding
  11. References
  • Ahmed, W., S. Sawant, F. Huygens, A. Goonetilleke, and T. Gardner, 2009: Prevalence and occurrence of zoonotic bacterial pathogens in surface waters determined by quantitative PCR. Water Res. 43, 49184928.
  • Alban, L., H. Stege, and J. Dahl, 2002: The new classification system for slaughter-pig herds in the Danish Salmonella surveillance-and-control program. Prev. Vet. Med. 53, 133146.
  • Amass, S. F., B. D. Vyerberg, D. Ragland, C. A. Dowel, C. D. Anderson, J. H. Stover, and D. J. Beaudry, 2000: Evaluating the efficacy of boot baths in biosecurity protocols. J. Swine Health Prod. 8, 169173.
  • Arnold, M. E., A. Cook, and R. Davies, 2005: A modelling approach to estimate the sensitivity of pooled faecal samples for isolation of Salmonella in pigs. J. R. Soc. Interface 2, 365372.
  • Assured British Pigs, 2008: Assured British Pigs Standard. Assured British Pigs, Surrey, UK. Available at: http://www.assuredpigs.co.uk/pigs/.
  • Bahnson, P. B., P. J. Fedorka-Cray, S. R. Ladely, and N. E. Mateus-Pinilla, 2006: Herd-level risk factors for Salmonella enterica subsp. enterica in U.S. market pigs. Prev. Vet. Med. 76, 249262.
  • Baptista, F. M., L. Alban, A. K. Ersbøll, and L. R. Nielsen, 2009: Factors affecting persistence of high Salmonella serology in Danish pig herds. Prev. Vet. Med. 92, 301308.
  • Barber, D. A., P. B. Bahnson, R. Isaacson, C. J. Jones, and R. M. Weigel, 2002: Distribution of Salmonella in swine production ecosystems. J. Food Prot. 65, 18611868.
  • Beloeil, P. A., C. Chauvin, K. Proux, N. Rose, S. Queguiner, E. Eveno, C. Houdayer, V. Rose, P. Fravalo, and F. Madec, 2003: Longitudinal serological responses to Salmonella enterica of growing pigs in a subclinically infected herd. Prev. Vet. Med. 60, 207226.
  • Beloeil, P. A., P. Fravalo, C. Fablet, J.-P. Jolly, E. Eveno, Y. Hascoet, C. Chauvin, G. Salvat, and F. Madec, 2004: Risk factors for Salmonella enterica subsp. enterica shedding by market-age pigs in French farrow-to-finish herds. Prev. Vet. Med. 63, 103120.
  • Bengtsson, B., U. Carlsson, E. Chenais, H. Eriksson, O. Fossum, D. Gavier-Widen, G. Hallgren, A. Hellström, C. Hultén, E. Lahti, S. S. Lewerin, J. Malmsten, K. P. Waller, K. De Verdier, P. Wallgren, S. Widgren, and E. Ågren, 2009: Surveillance and Control Programs. Domestic and Wild Animals in Sweden 2008. SVA National Veterinary Institute. Available at: http://sva.eu/upload/pdf/Tj%C3%A4nster%20och%20produkter/Trycksaker/Surveillance-08-web.pdf.
  • Blaha, T., 2004: Up-to-date information from the German QS Salmonella monitoring and reduction programme. Dtsch. Tierarztl. Wochenschr. 111, 324326.
  • Cortiñas Abrahantes, J., K. Bollaerts, M. Aerts, V. Ogunsanya, and Y. Van der Stede, 2009: Salmonella serosurveillance: different statistical methods to categorise pig herds based on serological data. Prev. Vet. Med. 89, 5966.
  • Creus, E., J. F. Perez, B. Peralta, F. Baucells, and E. Mateu, 2007: Effect of acidified feed on the prevalence of Salmonella in market-age pigs. Zoonoses Public Health 54, 314319.
  • Danish Meat Association, 2007: Danish Standard: Danish Quality Guarantee. Danish Meat Association, Copenhagen, Denmark. Available at: http://dma.bladrekatalog.dk/Quality_Guarantee/.
  • Ellis-Iversen, J., A. J. C. Cook, E. Watson, M. Nielen, L. Larkin, M. Wooldridge, and H. Hogeveen, 2010: Perceptions, circumstances and motivators that influence implementation of zoonotic control programs on cattle farms. Prev. Vet. Med. 93, 276285.
  • European Food Safety Authority (EFSA), 2008: Report of the task force on zoonoses data collection on the analysis of the baseline survey on the prevalence of Salmonella in slaughter pigs, part A. EFSA J. 135, 1111.Available at: http://www.efsa.europa.eu/en/scdocs/doc/zoon_report_ej135_finslpigs_en,3.pdf.
  • European Food Safety Authority (EFSA), 2009a: The community summary report on trends and sources of zoonoses, zoonotic agents, antimicrobial resistance and foodborne outbreaks in the European Union in 2007. EFSA J. 223, 1313.Available at: http://www.efsa.europa.eu/en/scdocs/doc/zoonoses_report_2007,3.pdf.
  • European Food Safety Authority (EFSA), 2009b: Analysis of the baseline survey on the prevalence of Salmonella in holdings with breeding pigs, in the EU, 2008, Part A: Salmonella prevalence estimates. EFSA J. 7, 193.Available at: http://www.efsa.europa.eu/en/scdocs/doc/1377.pdf.
  • Farzan, A., R. M. Friendship, C. E. Dewey, K. Warriner, C. Poppe, and K. Klotins, 2006: Prevalence of Salmonella spp. on Canadian pig farms using liquid or dry-feeding. Prev. Vet. Med. 73, 241254.
  • Fisher, R. A.., 1938: Statistical Methods for Research Workers, 10th edn. Oliver & Boyd, Edinburgh, UK.
  • Fosse, J., H. Seegers, and C. Magras, 2009: Prevalence and risk factors for bacterial food-borne zoonotic hazards in slaughter pigs: a review. Zoonoses Public Health 56, 429454.
  • Funk, J., and W. A. Gebreyes, 2004: Risk factors associated with Salmonella prevalence on swine farms. J. Swine Health Prod. 12, 246251.
  • Funk, J. A., P. R. Davies, and M. A. Nichols, 2000: The effect of fecal sample weight on detection of Salmonella enterica in swine feces. J. Vet. Diagn. Invest. 12, 412418.
  • Funk, J. A., P. R. Davies, and W. Gebreyes, 2001a: Risk factors associated with Salmonella enterica prevalence in three-site swine production systems in North Carolina, USA. Berl. Munch. Tierarztl. Wochenschr. 114, 335338.
  • Funk, J. A., P. R. Davies, and M. A. Nichols, 2001b: Longitudinal study of Salmonella enterica in growing pigs reared in multiple-site swine production systems. Vet. Microbiol. 83, 4560.
  • Harris, I. T., 2003: Serologic basis for assessment of subclinical Salmonella infection in swine: part 1. J Swine Health Prod. 11, 247251.
  • Hatcher, L.., 1994: Exploratory Factor Analysis. A Step-by-Step Approach to Using the SAS System for Factor Analysis and Structural Equation Modeling. SAS Institute, Cary, USA.
  • Heffernan, C., L. Nielsen, K. Thomson, and G. Gunn, 2008: An exploration of the drivers to bio-security collective action among a sample of UK cattle and sheep farmers. Prev. Vet. Med. 87, 358372.
  • Kranker, S., L. Alban, J. Boes, and J. Dahl, 2003: Longitudinal study of Salmonella enterica aerotype Typhimurium infection in three Danish farrow-to-finish swine herds. J. Clin. Microbiol. 41, 22822288.
  • Lo Fo Wong, D. M., J. Dahl, H. Stege, P. J. Van Der Wolf, L. Leontides, A. Von Altrock, and B. M. Thorberg, 2004: Herd-level risk factors for subclinical Salmonella infection in European finishing-pig herds. Prev. Vet. Med. 62, 253266.
  • Ortiz-Pelaez, A., and D. U. Pfeiffer, 2008: Use of data mining techniques to investigate disease risk classification as a proxy for compromised biosecurity of cattle herds in Wales. BMC Vet Res. 4, 24.
  • SAS Institute Inc., 2008: SAS/STAT 9.2 User’s Guide. SAS Institute, Cary, USA.
  • Sharma, S.., 1996: Applied Multivariate Techniques, John Wiley & Sons, Inc., New York, USA.
  • Statistics Portugal, 2008: Statistical Yearbook of Portugal 2007. Instituto Nacional de Estatística, Lisboa, Portugal, Available at: http://www.ine.pt/ngt_server/attachfileu.jsp?look_parentBoui=11797845&att_display=n&att_download=y.
  • Thrusfield, M., 2007: Veterinary Epidemiology. Blackwell Science, Oxford, UK.
  • Vieira-Pinto, M., P. Temudo, and C. Martins, 2005: Occurrence of Salmonella in the ileum, ileocolic lymph nodes, tonsils, mandibular lymph nodes and carcasses of pigs slaughtered for consumption. J. Vet. Med. B 52, 476481.
  • Van Winsen, R. L., B. A. P. Urlings, L. J. A. Lipman, J. M. A. Snijders, D. Keuzenkamp, J. H. M. Verheijden, and F. Van Knapen, 2001: Effect of fermented feed on the microbial population of the gastrointestinal tracts of pigs. Appl. Environ. Microbiol. 67, 30713076.
  • Van Der Wolf, P. J., J. H. Bongers, A. R. W. Elbers, F. M. M. C. Franssen, W. A. Hunneman, A. C. A. Van Exsel, and M. J. M. Tielen, 1999: Salmonella infections in finishing pigs in The Netherlands: bacteriological herd prevalence, serogroup and antibiotic resistance of isolates and risk factors for infection. Vet. Microbiol. 67, 263275.
  • Van Der Wolf, P. J., F. W. Van Schie, A. R. Elbers, B. Engel, H. M. Van Der Heijden, W. A. Hunneman, and M. J. Tielen, 2001a: Administration of acidified drinking water to finishing pigs in order to prevent Salmonella infections. Vet. Q. 23, 121125.
  • Van Der Wolf, P. J., W. B. Wolbers, A. R. Elbers, H. M. Van Der Heijden, J. M. Koppen, W. A. Hunneman, F. W. Van Schie, and M. J. Tielen, 2001b: Herd level husbandry factors associated with the serological Salmonella prevalence in finishing pig herds in The Netherlands. Vet. Microbiol. 78, 205219.
  • Van Der Wolf, P. J., D. M. Lo Fo Wong, W. B. Wolbers, A. R. Elbers, H. M. Van Der Heijden, F. W. Van Schie, W. A. Hunneman, P. Willeberg, and M. J. Tielen, 2001c: A longitudinal study of Salmonella enterica infections in high-and low-seroprevalence finishing swine herds in the Netherlands. Vet. Q. 23, 116121.