Salmonella surveillance in pigs: what oral fluids can do
Picture a finisher barn six weeks out from market. Salmonella surveillance in pigs usually starts with the least pleasant job on the list: catching pigs, taking rectal samples, pooling them. and keeping the chain of identification straight across a building with a thousand animals in it. It is slow, and on most farms it is the reason surveillance does not happen at all.
So the appeal of hanging a rope in a pen and letting pigs chew on it is obvious. A 2026 study tested that idea across 51 pens at three commercial finisher sites, taking paired oral fluid and pooled fecal samples from every pen and running both through the same enrichment, culture and PCR pipeline. The results are useful. and they are a warning about how easily a good sampling idea gets misused.
Surveillance in pigs does not fail because the tests are bad. It fails because sampling costs too much to keep doing. That is what oral fluids are trying to fix.
Why Salmonella in finisher pigs is worth chasing
Salmonella sits in an awkward position. In the live animal it often causes very little that anyone notices. and pigs can shed heavily while looking completely normal. The trouble shows up in the lairage and processing chain, where shedding becomes carcass contamination, and at the regulatory end, where control programs rank herds and apply consequences.
That makes the timing strange. The animals most likely to be carrying the organism at slaughter are the ones about to leave the farm, which is exactly when nobody wants to spend money on testing. Anything that cuts the cost of a finisher stage sample changes the economics of the whole program.
What Salmonella surveillance in pigs looks like with oral fluids
The study design was straightforward and that is part of its value. Fifty one pens across three commercial finisher sites were sampled between February and May 2025. Each pen gave both an oral fluid sample and a pooled fecal sample, 102 samples in total.
| Parameter | Detail |
|---|---|
| Design | Cross sectional, paired sampling |
| Sites | Three commercial finisher sites |
| Pens sampled | 51 |
| Samples | 102 paired oral fluid and pooled fecal samples |
| Period | February to May 2025 |
| Processing | Sequential enrichment, then bacteriological culture and multiplex PCR |
| Serotyping | Positive cultures sent to a national reference laboratory |
Both sample types went through identical handling. Sequential enrichment came first, then culture, then a multiplex PCR designed to pick out the serotypes that matter most in swine and in human disease.
The result nobody expected: raw PCR found nothing
Direct PCR run on unenriched oral fluid and fecal aliquots did not detect Salmonella in a single sample. Every result came back above a cycle threshold of 34, which the team treated as no detection.
That is worth sitting with. The assumption most people carry into molecular testing is that PCR is more sensitive than culture. so if you are going to spend money on a test, spend it there. On raw material from both sample types, PCR produced zero positives. Culture on enriched material found Salmonella in 34 of 51 oral fluid samples and 38 of 51 pooled fecal samples.
Enrichment is not an optional step here. It is the entire test.
Detection rates, method by method
| Method | Oral fluid | Pooled feces |
|---|---|---|
| Culture after enrichment | 66.7% (34/51) | 74.5% (38/51) |
| PCR after enrichment | 65.3% (32/49) | 70.6% (36/51) |
| Direct PCR, no enrichment | 0% (Ct > 34) | 0% (Ct > 34) |
Across all pens, Salmonella was detected in 42 of 51 pens, or 82.3 percent, by at least one of the culture or enrichment PCR methods. Detection differed sharply by site, at 33.3 percent for one site, 31.4 percent for another and 17.6 percent for the third.
Two samples were unavailable for PCR and two enriched samples were lost before testing, which is why the PCR denominators are 49 and 100 rather than 51 and 102.
Sensitivity and specificity are not the same story
The team ran a Bayesian analysis to estimate test performance without assuming either method was the gold standard. That matters, because there is no perfect reference test for Salmonella in live pigs.
| Estimate | Pooled feces | Oral fluids |
|---|---|---|
| Sensitivity | 0.84 (95% CrI 0.72 to 0.93) | 0.71 (95% CrI 0.57 to 0.83) |
| Specificity | 0.84 (95% CrI 0.70 to 0.95) | 0.91 (95% CrI 0.79 to 0.98) |
| Positive predictive value | 0.93 (95% CrI 0.85 to 0.98) | 0.95 (95% CrI 0.88 to 0.99) |
| Negative predictive value | 0.69 (95% CrI 0.52 to 0.84) | 0.57 (95% CrI 0.41 to 0.72) |
| Probability of detection | 81.8% (95% CI 43.3 to 96.4%) | 63.8% (95% CI 23.7 to 90.9%) |
Oral fluids are less sensitive than pooled feces. But they are more specific, and their positive predictive value is actually higher at 0.95 against 0.93. A positive oral fluid result is very likely a real positive.
Both methods share the same weakness. The negative predictive values are 0.69 for feces and 0.57 for oral fluids. A negative result does not give you much confidence that a pen is clean. and for oral fluids it gives you barely more than a coin flip.
The serotype data is the quiet win
Fifty nine Salmonella positive cultures were submitted for serotyping. The oral fluid isolates matched the fecal isolates closely within each site, including the serotype that matters most.
| Site | Sample type | Serotypes recovered |
|---|---|---|
| Site 1 | Pooled feces (n=14) | Rissen 50%, I 4,[5],12:i:- 36%, Agona 14% |
| Site 1 | Oral fluid (n=14) | Rissen 43%, I 4,[5],12:i:- 36%, Agona 21% |
| Site 2 | Pooled feces (n=17) | I 4,[5],12:i:- 100% |
| Site 2 | Oral fluid (n=14) | I 4,[5],12:i:- 100% |
| Site 3 | Pooled feces (n=7) | Senftenberg 57%, Johannesburg 43% |
| Site 3 | Oral fluid (n=4) | Senftenberg 75%, Johannesburg 25% |
At one site every isolate, from both sample types, was I 4,[5],12:i:-, the monophasic variant of Typhimurium that has become one of the more concerning serotypes in swine and human medicine. Oral fluid picked that up without a single rectal sample.
That changes what oral fluids are for. They are not just a cheaper yes or no. They are a cheap route to the serotype that is actually circulating, which is the information a control decision rests on.
Where the two methods agree, and where they do not
Agreement between oral fluid and pooled fecal results was moderate at best.
| Comparison | Cohen’s kappa | On the whole concordance |
|---|---|---|
| Oral fluid culture vs fecal culture | 0.373 (95% CI 0.094 to 0.617) | 72.5% (95% CI 58.8 to 84.3%) |
| Oral fluid PCR vs fecal PCR | 0.434 (95% CI 0.167 to 0.689) | 77.6% (95% CI 65.3 to 87.8%) |
| Oral fluid culture vs oral fluid PCR | 0.867 (95% CI 0.697 to 1.000) | 93.9% (95% CI 85.7 to 100.0%) |
| Fecal culture vs fecal PCR | 0.805 (95% CI 0.557 to 1.000) | 93.9% (95% CI 85.7 to 100.0%) |
Read the first two rows against the last two. Within a sample type, culture and PCR agree almost perfectly. Between sample types, agreement drops to fair. The two samples are not measuring the same thing at pen level, even though a logistic regression found no statistically significant difference in on the whole detection probability between them at P = 0.13.
That apparent contradiction has a simple explanation. The methods detect a similar number of positives across the whole population. but not the same pens. For herd level screening that is fine. For deciding that one specific pen is negative, it is not.
What this means for a monitoring program
Use oral fluids to find problems and characterize them. The labor saving is real, the welfare case is real, and the serotype data matches what fecal sampling produces.
Do not use them to declare a pen free. With a negative predictive value of 0.57 at the prevalence seen here, a negative result is weak evidence. Showing freedom needs a different design, more samples, or a more sensitive matrix.
Never skip enrichment. Direct PCR on raw material produced zero detections from 102 samples. Any protocol promising a fast answer from unenriched material deserves suspicion.
Sample the stage you intend to act on. The study sampled finishers, which is what determines what arrives at the plant.
Expect site level variation. Detection ran from 17.6 percent at one site to 33.3 percent at another in the same system, so a single site result does not describe the whole operation.
Limits of the evidence
Fifty one pens at three sites is a modest sample, and all of it came from one production system in one region over about four months. Nothing here should be read as a seasonal or geographic average.
The estimated true prevalence was high, with a global posterior estimate of 0.70. Predictive values depend on prevalence. In a low prevalence herd the positive predictive value falls and the negative predictive value rises. so these numbers do not transfer directly to a herd that has been working on control for years.
Samples were lost. Two samples were unavailable for PCR and two enriched samples were destroyed before testing, which weakens the PCR comparisons slightly.
The credible intervals are wide, and the sensitivity intervals for oral fluids and feces overlap.
This study also compared two pen level sample types only. Neither was measured against individual animal sampling, which remains the most sensitive approach and the least practical one.
Frequently asked questions
What is oral fluid testing in pigs?
Oral fluid testing uses a rope hung in a pen that pigs chew on, collecting saliva and oral exudate. The sample is then enriched, cultured and tested by PCR. In a 2026 study of 51 pens, oral fluids detected Salmonella in 66.7 percent of pens by culture, against 74.5 percent for paired pooled fecal samples.
Is oral fluid as accurate as fecal sampling?
Not quite. Estimated sensitivity was 0.71 for oral fluids against 0.84 for pooled feces, so oral fluids miss more true positives. Specificity went the other way, at 0.91 for oral fluids against 0.84 for feces, and the positive predictive value was slightly higher for oral fluids at 0.95 against 0.93.
Can PCR replace culture for pig Salmonella testing?
Not directly. PCR run on raw, unenriched oral fluid and fecal samples detected nothing at all, with all results above a cycle threshold of 34. After enrichment, PCR detected the organism in 65.3 percent of oral fluid samples and 70.6 percent of fecal samples. Enrichment is the step that makes the test work.
What does a negative oral fluid result mean?
Less than most people assume. The estimated negative predictive value was 0.57 for oral fluids and 0.69 for pooled feces at the prevalence seen in this study. Roughly, a negative oral fluid result left a substantial chance the pen was still positive, so it is a poor basis for declaring freedom from Salmonella.
How common was Salmonella in the study pens?
Salmonella was detected in 42 of 51 pens, or 82.3 percent, using at least one of the culture or enrichment PCR methods. That is high, and the Bayesian estimate of true prevalence was 0.70. Results varied by site, from 17.6 percent at one site to 33.3 percent at another in the same system. The published summary also cites 86.3 percent when every method is counted, a discrepancy worth noting.
Which serotypes were found?
Serotyping of 59 positive cultures returned Rissen, I 4,[5],12:i:-, Agona, Senftenberg and Johannesburg. At one site every isolate from both sample types was I 4,[5],12:i:-, the monophasic variant of Typhimurium. Another site returned Rissen and Agona alongside it, and a third returned Senftenberg and Johannesburg. Oral fluid isolates matched fecal isolates closely within each site.
Why does the monophasic Typhimurium variant matter?
I 4,[5],12:i:- is a monophasic variant of Salmonella Typhimurium that has become one of the more concerning serotypes in both swine production and human disease. In this study it accounted for 100 percent of isolates at one site, from both oral fluid and fecal samples, which is the kind of finding that changes a control priority.
How much labor does oral fluid sampling save?
The study did not measure labor time directly, so no figure can be quoted. The practical case is that a rope can be deployed and collected without handling individual animals, which removes the catching and restraint that make pooled fecal sampling expensive across large finisher sites. In this work 51 pens produced 102 paired samples, and the oral fluid half required no animal handling at collection.
Should oral fluids replace fecal sampling entirely?
No. Agreement between the two was only fair, with a Cohen’s kappa of 0.373 for culture and 0.434 for PCR. Within a sample type, culture and PCR agreed almost perfectly at kappa values of 0.867 and 0.805. The two matrices detect similar totals but not the same pens. Oral fluids suit herd level screening, but for a pen level clearance decision the disagreement matters too much.
When should finisher pigs be sampled?
This study sampled finishers between February and May, the stage closest to slaughter and the stage that determines what arrives at the processing plant. Testing earlier tells you about earlier stages. The relevant point is to sample the stage whose result you intend to act on. Sampling close to market also captures the window when shedding carries the most downstream consequence.
Does pen level prevalence predict carcass contamination?
This study did not follow pigs to slaughter, so no link to carcass contamination can be claimed from it. What it did establish is that pen level detection in finishers was common and varied substantially between sites within one production system. Detection ran from 17.6 percent at one site to 33.3 percent at another, so a single site reading is a poor basis for predicting what leaves the farm.
What are the main limitations of this research?
Fifty one pens at three sites in one production system over four months. Prevalence was high at an estimated 0.70, so predictive values will shift in lower prevalence herds. Four samples were lost to testing. The Bayesian estimates rely on modelling assumptions, and neither method was compared against individual animal sampling.
Where this leaves the decision
Oral fluids are a good screening tool and a bad clearance tool. They find Salmonella, they identify the serotype, and they cost far less labor than catching pigs for pooled fecal samples. What they will not do is reassure you that a pen is clean. Used as a way to find and characterize infection, they are a genuine improvement on the status quo. Used as a way to prove absence, they will let infection through.
*This article summarizes published research for informational purposes. It does not constitute veterinary or regulatory advice. Always follow local regulations and the product label when using veterinary medicines.
