Nontyphoidal salmonellosis originates from various food products, including pork, and is a common cause of gastroenteritis in humans. Swine are not clinically affected by infection with most nontyphoidal Salmonella spp, but swine do carry Salmonella spp into abattoirs.1 This has been identified as a source of contamination of pork products. As a result of the association between pork consumption and nontyphoidal salmonellosis, several European countries have instituted surveillance programs and on-farm reduction programs for Salmonella spp.2,3 These programs require interventions to reduce the prevalence of exposure of pigs to Salmonella spp on farms.4
Farms targeted for interventions are identified on the basis of a high prevalence of antibodies against Salmonella spp in finished swine. Consequently, a large amount of data is available from Europe about the prevalence of exposure to Salmonella spp in swine. Knowledge of this prevalence potentially enables the swine industry and associated industries to determine the extent of exposure to Salmonella spp and to evaluate on-farm interventions aimed at reducing the prevalence of Salmonella spp in swine. To our knowledge, similar data are not available to describe the prevalence of antibodies against Salmonella spp in swine in the United States. Therefore, the objective of the study reported here was to provide baseline data about the seroprevalence of antibodies against nontyphoidal Salmonella spp in marketed swine in Iowa.
Materials and Methods
Study population—Producers were designated as lowvolume or high-volume producers on the basis of the number of swine marketed to cooperating abattoirs. Production characteristics of these 2 types of producers differed sufficiently to warrant separate evaluation. Because there is no national register of swine producers in the United States, a list of all producers was not available for random selection of study participants. Therefore, a convenience sample from the MSSP database was used to identify lowand high-volume producers. This convenience sample represented approximately a fourth of the national number of finishing pigs.
The MSSP used carcass sampling at 8 high-capacity slaughter plants in Iowa to assist in the Midwestern efforts for eradication of pseudorabies virus from 2001 to 2005. For the study reported here, carcass samples (meat juice) were selected as an alternative to serum as a medium for antibodies because they could be easily and safely collected, and dependable carcass identification systems were available at line speeds used in the slaughter plants. Although not widely recognized in the United States as a corollary sample to serum, meat juice has been used successfully for the past 10 years in a farm-based Salmonella control program in Denmark to classify exposure of pigs to Salmonella spp.2,4
In the 8 high-capacity slaughter plants in Iowa, samples (50 to 75 g) of diaphragm muscle were collected by USDA–Animal and Plant Health Inspection Service–Veterinary Service employees at the rate of 4 specimens/lot during 1 shift/slaughter plant. Samples of diaphragm muscle were obtained from a carcass and placed in a plastic bag,a and the identification number for that lot of pigs was written on the bag. Samples were frozen overnight and shipped the next day to a laboratory at Iowa State University for processing. Lot identification number, slaughter plant, and harvest date recorded on each sample bag uniquely identified each sample. Samples were received in the processing lab, assigned a laboratory control number, and thawed. The fluid portion of each sample was decanted into numbered tubes, which were submitted for Salmonella assay.
Experiment 1—Selection of low-volume producers from the MSSP was performed by identifying producers who submitted typical lot sizes of between 60 and 180 swine to market at least 4, but not more than 12, times between October and December 2001. For the low-volume producers, a stratified sampling scheme was used. Each lot of swine submitted by a producer represented a stratum, and carcasses were selected from within the stratum for use in testing to detect evidence of pseudorabies virus and Salmonella spp. From January to March 2002, the MSSP sampling scheme was used for collecting samples of diaphragm from 4 carcasses/lot. The anticipated sample size for each low-volume producer was a minimum of 12 (1 lot/4-week period) and a maximum of 48 (1 lot/wk for 12 weeks). Because of the rapid line speed in the slaughter plants, random number allocation methods were not feasible. Therefore, convenience samples were selected from 4 carcasses/lot. Samples were identified on the basis of lot tattoo, slaughter plant, and date and were frozen for transport to the laboratory.
Samples were tested to detect antibodies by use of a commercially available, indirect ELISAb that is based on lipopolysaccharide antigens.5 The test results were reported as a percentage of the OD of the sample, relative to the OD of the positive reference samples. On the basis of the manufacturer's directions, OD values > 40% (ratio of sample to positive control sample = 100%) were considered to have positive results.
An analysis describing the prevalence of pigs, lots, and farms with positive results was conducted with a procedure used to estimate survey populationsc that accounted for the sampling design. The procedure yields estimates of population means and totals from sample survey data. The procedure also yields variance estimates, CIs, and other descriptive statistics. When computing these estimates, the procedure takes into account the design used to select the survey sample, such as a complex survey sample with stratification, clustering, and unequal weighting. The procedure uses the Taylor expansion method to estimate sampling errors of estimators on the basis of complex sample designs. When there are clusters (eg, primary sampling units) in the sample design, the procedure estimates variance from the variation among those clusters. When the design is stratified, the procedure pools variance estimates for the strata to compute an overall variance estimate.
A producer was considered to have positive results when 1 or more samples had positive results during the 12-week period. Mean prevalence for the lots was used to classify producers in 1 of 4 categories (0, evidence of negligible previous exposure to Salmonella spp; 1, evidence of low amounts of previous exposure to Salmonella spp; 2, evidence of moderate amounts of previous exposure to Salmonella spp; and 3, evidence of widespread previous exposure to Salmonella spp). The seroprevalence cutoff value for each of these categories was adopted from cutoff values described elsewhere (Appendix).2 Seroprevalence classifications varied on the basis of estimated annual farm production, which was extrapolated from the total number of pigs submitted for slaughter at the participating slaughter plants during the 12-week study period.
Experiment 2—High-volume producers were selected from the MSSP on the basis of the following criteria: they marketed > 50,000 swine, researchers had knowledge of the producer's production system, and there was reliability of sample collection (ie, consistently marketed pigs to an MSSPparticipating slaughter plant). Because of budget constraints, the sample size was limited to approximately 2,400 samples by use of a 2-stage cluster scheme. The primary sampling unit (first-stage cluster) was a selected lot within a producer designation, and the sampling unit for the second-stage cluster was a carcass within a selected lot.
From January to March 2002, samples from 1 lot/wk from each of 45 high-volume producers were collected for testing to detect evidence of pseudorabies virus and Salmonella spp. The anticipated sample size for each highvolume producer was 48 samples for the 12-week study. Similar to testing in experiment 1, these samples were tested to detect antibodies by use of a commercially available ELISAb based on lipopolysaccharides antigens.6 The procedure to estimate survey means c was again used to calculate prevalence of antibodies. Similar to the classification used for low-volume producers, the mean seroprevalence of the weekly lots was used to classify the producers (categories 0, 1, 2, or 3) on the basis of evidence of previous exposure to Salmonella spp.
Results
Experiment 1 (low-volume producers)—For the 1,044 low-volume producers, 25,478 samples were collected from 6,299 lots. Mean number of samples collected per producer was 26, with 940 of 1,044 (90%) producers providing between 8 and 57 samples. Maximum number of samples collected from a single producer was 104. Producers with < 10 samples for the 12-week study period were excluded from the producer-level analysis. Of the 1,044 low-volume producers, 967 (93%) were estimated to market < 3,000 swine annually (Table 1).
Frequency distribution of number of herds for categories that describe swine herds as having negligible, low, moderate, or high evidence of exposure to Salmonella spp.
Estimated No. of pigs marketed annually* | No. of herds | Negligible (category 0) | Low (category 1) | Moderate (category 2) | Widespread (category 3) |
---|---|---|---|---|---|
< 200 | 46 | 35 (76) | 10 (22) | NA | 1 (2) |
201–500 | 213 | 148 (70) | 48 (22) | 15 (7) | 2 (1) |
501–1,000 | 313 | 171 (55) | 107 (34) | 27 (9) | 8 (2) |
1,001–2,000 | 286 | 127 (44) | 124 (43) | 32 (11) | 3 (1) |
2,001–3,000 | 109 | 39 (36) | 44 (40) | 25 (23) | 1 (1) |
3,001–5,000 | 69 | 26 (38) | 39 (56) | 4 (6) | 0 (0) |
5,001–20,000 | 8 | 4 (50) | 2 (25) | 2 (25) | 0 (0) |
Total for low-volume producers | 1,044 | 550 (53) | 374 (36) | 105 (10) | 15 (1) |
High-volume producers† | 45 | 5 (11) | 27(60) | 12 (27) | 1 (2) |
All pigs marketed | 1,089 | 555 (51) | 401(37) | 117 (11) | 16 (1) |
Values in parentheses are percentages.
Estimated number of pigs marketed annually was extrapolated from the number of pigs marketed for the 12-week period from January to March 2002.
Defined as producers who met the following criteria: they marketed 50,000 swine/y, researchers had knowledge of the producer's production system, and there was reliability of sample collection (ie, consistently marketed pigs to an MSSP-participating slaughter plant).
NA = Not applicable; category 2 was not used as a designation for this number of pigs because when this number of pigs had a seroprevalence 50%, it was assigned to category 3.
Of the 25,478 samples collected, 1,863 had positive results for antibodies against Salmonella spp, 23,609 had negative results, and 6 could not be tested because of an insufficient amount of sample material. An estimate of the proportion of samples with positive results for antibodies against Salmonella spp was 7.3% (95% CI, 7.05% to 7.56%; SEM, 0.001). Prevalence estimate for samples with negative results for antibodies against Salmonella spp was 92.7% (95% CI, 92.4% to 92.9%; SEM, 0.001).
Of the 6,299 lots from which samples were collected, 1,191 (18.9%) contained at least 1 sample seropositive for Salmonella spp. In those 1,191 lots, 5,054 samples were tested and 1,863 of those samples had positive results. Within the 1,191 lots with positive results, the mean prevalence of Salmonella-positive samples was 38.97% (95% CI, 40.2% to 37.7%; SEM, 0.600).
Swine marketed by 1,044 low-volume producers were tested in the study, and 494 of those producers had at least 1 sample with positive results. Within those low-volume producers with carcasses that had positive results, mean prevalence of samples with positive results for Salmonella spp was 15.3% (95% CI, 14.0% to 16.5%; SEM, 0.006). The highest percentage of samples with positive results from a single producer was 73%. Among the low-volume producers, the apparent prevalence of herds classified as category 0 was 53% (550/1,044 herds), whereas it was 36% (374 herds) for category 1, 10% (105 herds) for catgory 2, and 1% (15 herds) for category 3 (Table 1).
Experiment 2 (high-volume producers)—For the 45 high-volume producers, 2,486 samples were collected from 629 market lots. Total number of samples collected for each high-volume producer varied from 36 to 84. The 629 lots represented samples obtained from 13,888 lots marketed by the high-volume producers during the study period. We did not retain data on the number of swine in lots from which samples were not collected. However, on the basis of the mean number of pigs in each of the sampled lots, it is estimated that each of the 45 producers marketed between 20,000 and 2,000,000 pigs annually.
Of the 2,486 samples collected, 203 (8.1%; 95% CI, 6.8% to 9.3%) had positive results for antibodies against Salmonella spp. Furthermore, of the 629 lots, 124 had at least 1 sample with positive results for antibodies against Salmonella spp. Among high-volume producers, the apparent prevalence for herds for each category of exposure was 11% (5/45 herds), 60% (27 herds), 27% (12 herds), and 2% (1 herd) for categories 0, 1, 2, and 3, respectively.
Discussion
The apparent pig-level seroprevalence for Salmonella spp was approximately 8% among the lowvolume producers, and the apparent prevalence of lowvolume producers with at least 1 sample with positive results for Salmonella spp was 47.3%. Among the lowvolume producers, the apparent prevalence of herds for category 0 was 53% (550/1,044 herds), whereas it was 36% (374 herds) for category 1, 10% (105 herds) for category 2, and 1% (15 herds) for category 3. Analysis of this distribution suggests widespread exposure to Salmonella spp for swine of low-volume producers but low herd-level exposure among producers.
Authors from several countries have reported the prevalence of herds that are classified on the basis of the degree of exposure to Salmonella spp. Classifications in Denmark include a study2 in which investigators used 4 levels (0, negligible prevalence of antibodies against Salmonella spp; 1, low prevalence of antibodies against Salmonella spp; 2, moderate prevalence of antibodies against Salmonella spp; and 3, unacceptably high prevalence of antibodies against Salmonella spp). Seroprevalence cutoff values for these classifications were designed to ensure that the number of herds for category 2 or 3 did not exceed the bacteriologic and slaughter capacity of the system. In Denmark, herds at level 2 or 3 are required to complete an on-farm microbial culture of fecal samples, and herds at level 3 are restricted to certain slaughter plants and certain time periods for marketing of slaughter hogs. In 1997, the 40% OD cutoff for seropositive swine and an unweighted estimate of herd samples was used to determine that the distribution of Danish herds for levels 0 and 1 combined, 2, and 3 was 93.7%, 3.9%, and 2.3%, respectively.4 The frequency of herds for categories (ie, levels) 2 and 3 was higher in the population of the study reported here, compared with the frequency for the Danish population. It is not sensible to compare the prevalence reported7 for swine herds in the Netherlands with the prevalence for our study population because the cutoff values used to designate samples with positive results differed.
For pig-level analyses, the prevalence of antibodies against Salmonella spp among swine marketed by highvolume producers was approximately 8%, with the apparent prevalence of category 0, 1, 2, and 3 herds at 11% (5/45 herds), 60% (27 herds), 27% (12 herds), and 2% (1 herd), respectively. The high-volume producers in our study were substantially larger than those in other studies.4,7,8 Therefore, no direct comparison group is available to evaluate the apparently greater evidence of exposure to Salmonella spp. Furthermore, swine of the high-volume producers were subject to a sampling scheme and sampling intensity that differed from those used for the swine of low-volume producers in our study, which makes it inappropriate to directly compare the prevalence of herds for the various exposure categories between these populations.
Serologic testing is designed to detect historical and recent exposure of animals to Salmonella spp, and no criterion-referenced standard is available to determine exact sensitivity and specificity. However, estimates for this test system yielded a sensitivity of 50% to 85% and a specificity > 95%.2 Obviously, estimates of sensitivity and specificity vary with the choice of cutoff value, which consequently affects the inferences that can be reached for any study. Furthermore, serologic tests to detect Salmonella spp should not be confused with microbial culture of fecal samples to detect Salmonella spp, which is a measure of current Salmonella status.
Investigators have used OD10%,2 OD20%,6 and OD40%4 as cutoff values to distinguish between animals with positive and negative results. The manufacturer of the test kit used in the study reported here recommended a cutoff value of OD40%, and investigators in 1 study4 reported data for which they used an OD40% cutoff value as the designation for seropositive results for Salmonella spp. Consequently, we used these values. We used an unweighted method to combine information about the lots among producers, similar to the method reported in another study,4 in favor of the weighted serologic index described by other investigators.2 We also defined a negative producer as one that had no samples with positive results, whereas other investigators9 have designated negative herds as those that had 2 or fewer samples with positive results. Arguments can be made for the numerous permutations of cutoff values, weighting system, and definition of negative herd status. However, because of the lack of a criterion-referenced standard and subsequent inconsistency among the numerous reports on Salmonella spp in swine, it is important to resist overinterpretation of differences of results among studies.
The objective of the study reported here was to establish baseline information about prevalence of Salmonella spp in finishing swine marketed in Iowa. For this purpose, we chose to use a cutoff value of OD40% for each pig to designate samples with a positive result and a summation of unweighted prevalence estimates to combine lots to determine producer-level estimates. Furthermore, we defined a negative producer as one with no samples that had positive results. Use of various OD cutoff values for status of each pig, various methods of calculating prevalence, and other definitions of a positive herd would result in differing estimates, but for our purposes of describing a baseline, those estimates would not necessarily be better or more relevant. The effect of various cutoff values for serologic tests on the misclassification of Salmonella status has been extensively discussed.10
For proper interpretation of the data reported here, it was important to clarify what is meant by a producer. In the MSSP, the system from which these data were collected, a producer is the entity (company or person) receiving payment for the marketed swine; thus, we retained this definition. This limitation necessitates we infer that differences among producers are attributable to variables based at the production site. For producers of > 5,000 market pigs, this means that deliveries from > 1 production site are almost certainly incorporated into the prevalence estimate. In this way, our study differs markedly from those reported by European researchers in which sample collection was clearly differentiated on the basis of site of production. Furthermore, because of confidentiality requirements for the MSSP, it was not possible for us to contact producers to refine information on herd size or location. However, the information available did enable us to describe the prevalence of Salmonella-positive producers as determined on the basis of results for finishing swine marketed in Iowa.
Finally, the importance of antibodies against Salmonella spp to the safety of pork products must be addressed. At the animal level, the detection of antibodies against Salmonella spp is a poor indicator of carriage of Salmonella organisms into the slaughter environment where there may be contamination of carcasses.9,11,12 Additionally, the association between herdlevel seroprevalence of antibodies against Salmonella spp, carcass contamination, and subsequent public health risks is less clear. The association between seroprevalence and microbial culture of fecal samples from swine has been reported,13 but the real outcome of interest should be disease in humans. In Denmark, the incidence of infections in humans attributable to Salmonella spp has decreased concurrent with the implementation of the Danish Salmonella control program. This finding is likely confounded by the many concurrent changes that have taken place for production and processing of pork in Denmark.
An alternative to studying risk of disease in humans is to evaluate the contamination of pork. We are unaware of any studies that directly correlate seroprevalence of lots with contamination of pork carcasses. However, investigators in 1 study12 reported that at the herd level, the odds of carcass contamination increased by 1.3 to 1.5 for each increase of 10% in herd seroprevalence. Direct application to the evaluation of the health risk to humans is problematic without knowledge of the underlying odds of carcass contamination from herds or the association between carcass contamination and health risk to humans.
Seroprevalence of antibodies against Salmonella spp in finishing swine marketed in Iowa was determined for the study reported here. We are unaware of any equivalent studies that report the prevalence of Salmonella spp in the US swine industry. Apparent exposure to Salmonella spp as measured by seropositive reactions appears to be evident in < 10% of pigs marketed in Iowa. Most swine marketed by low-volume producers had negligible or low evidence of exposure to Salmonella spp. However, a higher percentage of lots and carcasses with positive results for Salmonella spp were detected for swine marketed by high-volume producers.
ABBREVIATIONS
MSSP | Market Swine Surveillance Program |
OD | Optical density |
CI | Confidence interval |
Whirlpak, Nasco Sampling Products Inc, Fort Atkinson, Wis.
Herdcheck ELISA, IDEXX Laboratories Inc, Westbrook, Me.
Proc SURVEYMEANS, version 9.01, SAS Institute Inc, Cary, NC.
References
- 1.↑
Larsen ST, McKean JD & Hurd HS, et al. Impact of commercial preharvest transportation and holding on the prevalence of Salmonella enterica in cull sows. J Food Prot 2003;66:1134–1138.
- 2.↑
Alban L, Stege H, Dahl J. The new classification system for slaughter-pig herds in the Danish Salmonella surveillance-and-control program. Prev Vet Med 2002;53:133–146.
- 3.
Blaha TH. Interventional measures: implementing a Salmonella monitoring programme for pork in Germany, in Proceedings. 5th Int Symp Epidemiol Control Foodborne Pathogens Pork 2003;200–202.
- 4.↑
Mousing J, Jensen PT & Halgaard C, et al. Nation-wide Salmonella enterica surveillance and control in Danish slaughter swine herds. Prev Vet Med 1997;29:247–261.
- 5.↑
Camitz A, Holmquist G & Ballagi A, et al. HerdChek Salmonella antibody ELISA for serological monitoring of Salmonella infection in swine, in Proceedings. 4th Int Symp Epidemiol Control Foodborne Pathogens Pork 2001;505–508.
- 6.↑
Carstensen B, Christensen J. Herd size and sero-prevalence of Salmonella enterica in Danish swine herds: a random-effects model for register data. Prev Vet Med 1998;34:191–203.
- 7.↑
van der Wolf PJ, Elbers AR & van der Heijden HM, et al. Salmonella seroprevalence at the population and herd level in pigs in The Netherlands. Vet Microbiol 2001;80:171–184.
- 8.
Stege H, Christensen J & Nielsen JP, et al. Prevalence of subclinical Salmonella enterica infection in Danish finishing pig herds. Prev Vet Med 2000;44:175–188.
- 9.↑
Botteldoorn N, Heyndrickx M & Rijpens N, et al. Salmonella on pig carcasses: positive pigs and cross contamination in the slaughterhouse. J Appl Microbiol 2003;95:891–903.
- 10.↑
Steinbach G, Blaha T, Methner U. Estimating the prevalence of Salmonella spp. in swine herds—influence of sensitivity and specificity of Salmonella detection. Zentralbl Veterinarmed [B] 2002;49:438–444.
- 11.
Hurd HS, McKean JD & Griffith RD, et al. Estimation of the Salmonella enterica prevalence in finishing swine. Epidemiol Infect 2004;132:127–135.
- 12.↑
Sorensen L, Alban L & Nielsen B, et al. The correlation between Salmonella serology and isolation of Salmonella in Danish pigs at slaughter. Vet Microbiol 2004;101:131–141.
- 13.↑
Lo Fo Wong DM, Dahl J & van der Wolf PJ, et al. Recovery of Salmonella enterica from seropositive finishing pig herds. Vet Microbiol 2003;97:201–214.
Appendix
Seroprevalence cutoff values* used to classify swine herds as having negligible, low, moderate, or high evidence of exposure to Salmonella spp.5
Estimated No. of pigs marketed annually† | Negligible (category 0) | Low (category 1) | Moderate (category 2) | Widespread (category 3) |
---|---|---|---|---|
<200 | 0 | >0–50 | NA | >50 |
201–500 | 0 | > 0–25 | > 25–50 | > 50 |
501–1,000 | 0 | > 0–23 | > 23–50 | > 50 |
1,001–2,000 | 0 | > 0–20 | > 20–50 | > 50 |
2,001–3,000 | 0 | > 0–17 | > 17–50 | > 50 |
3,001–5,000 | 0 | > 0–17 | > 17–50 | > 50 |
> 5,000 | 0 | > 0–17 | > 17–33 | > 33 |
Seroprevalence cutoff value is the percentage of the herd that is positive for anti-Salmonella antibodies.
Estimated number of pigs marketed annually was extrapolated from the number of pigs marketed for the 12-week period from January to March 2002.
NA = Not applicable; category 2 was not used as a designation for this number of pigs because when this number of pigs had a seroprevalence 50%, it was assigned to category 3.