Anne Hiller*1, 4, Derk Oorburg*2, 4, Henk J. Wisselink3, Conny B. van Solt-Smits3, Bert Urlings2,4, Günter Klein1, Gereon Schulze Althoff4 and Lourens Heres4
1University of Veterinary Medicine, Bischofsholer Damm 15, 30173 Hannover, Germany
2 Animal Sciences Group of Wageningen UR, P.O. Box 338, 6700 AH Wageningen, The Netherlands
3 Central Veterinary Institute of Wageningen UR, P.O. Box 65, 8200 AB-Lelystad, The Netherlands
4 VION Food Group, P.O. Box 380, 5680 AJ Best, The Netherlands
*Both authors contributed equally to this paper. Correspondence:
Anne Hiller – e-mail: anne.hiller@vionfood.com – tel. +49- 211-44033/0 Derk Oorburg – e-mail: derk.oorburg@vionfood.com - tel. +31-411-658555
Abstract
Mycobacterium avium (MA) is a potential food safety hazard in pigs. In a newly developed risk based supply chain meat inspection blood samples of slaughtered pigs are routinely tested for MA antibodies. The integrated control of MA and the verification serology is applied as alternative for the incision of lymph nodes at official meat inspecton. The serological MA prevalence in the tested pig populations was estimated and the validity of the MA-ELISA as applied in this system was evaluated. In the Dutch and German population the number of MA positive samples was 1.0% and 1.7%. Respectively 0.5% and 17.4% of the Dutch and German herds were at risk for having a MA infection. The specificity of the MA-ELISA is high (>98.4%). The probability to recognize a MA infected herd is 35 to 100%. Depending on the number of samples and the applied cut-off the herd sensitivity is between 8 and 100%. It is concluded that the risk based supply chain meat inspection detects and controls MA more actively and improves food safety control as well.
Keywords: Mycobacterium avium; pig; supply chain meat inspection; validation; ELISA
Introduction
Mycobacterium avium subsp. avium and M. avium subsp. hominissuis belong to the Mycobac-terium avium complex (MAC) and are frequently associated with diseases in animals and hu-mans. MAC is an opportunistic pathogen which leads to disseminated infections with in-creased morbidity and mortality, particularly in immune compromised people (Ashford et al., 2001, Falkinham, 1996). MAC infections are reported in 30 to 80% of patients with AIDS (Bermudez et al., 1992). MAC also causes chronic pneumonia in elderly people and cervical lymphadenitis in young children between 0 and 5 years of age (Eriksson et al., 2001, Nylen et al., 2000). In pigs MA can cause lymphadenitis with granulomatous lesions, especially the submaxillary and mesenteric lymph nodes are affected (Thoen, 2006). The main route of in-fection in pigs is via the gastro-intestinal tract (Thoen, 2006). Outbreaks in herds are de-scribed after feeding pigs with mycobacteria contaminated peat, compost, bark mulch and sawdust (Matlova et al., 2005, Trckova et al., 2005). Pigs have been suggested as a vector for transmission of MA towards humans (Komijn et al., 1999, Martin and Schimmel, 2000, Möbius et al., 2006) and therefore MA is considered a hazard in pig meat production, that needs to be controlled in the pork supply chain.
European law (EU/854/2004) prescribes the procedures for traditional meat inspection, which includes the incision of the submaxillary lymph nodes and palpation of the mesenteric lymph nodes. Aim of this legal requirement is to detect mycobacterial infections in pigs at slaughter.
However, the incision of the lymph nodes is characterized by relatively high false positive and false negative results (Komijn et al., 2007, Wisselink et al., 2010). In addition, it can cause cross-contamination with other food safety hazards e.g. salmonella (Hamilton et al., 2002, SCVMPH, 2000). A meat inspection system replacing the routine incision of the sub-maxillary lymph nodes with a procedure having equal effect, but lacking the negative cross contamination aspect, offers therefore the opportunity to improve food safety.
Starting in 2006, a risk based meat inspection system was introduced in six Dutch slaughter-houses and in 2008 in one German slaughterhouse. This system has several characteristics.
Only pigs from herds which do not show serological signs of a MA infection are allowed for visual meat inspection without incision of the submaxillary and palpation of the mesenteric lymph nodes. For monitoring MA infections in slaughter pigs, blood samples were collected
during bleeding. Risk categorisation was based in an aggregate set of results of the MA-ELISA. Organs and carcasses are also only visually inspected.
In the present paper the active control of MA in a risk based supply chain meat inspection system in slaughterhouses is evaluated with emphasis on the validity of the applied MA-ELISA (Wisselink et al., 2010). First the risk of MA infections in pigs was estimated, based on the serological monitoring results from this risk based meat inspection. The validity of the MA-ELISA test was evaluated with samples from MA positive and negative herds.
Methods
Collection of samples
At every delivery of a consignment of pigs, blood samples were collected randomly from clinically healthy pigs. Treated test tubes (10 ml) for serum collection with coagulation in-ducer were used. Samples were identified on herd level. Until coagulation, samples were stored at room temperature and then up to analyses at 4°C.
The blood was send to one laboratory that carried out the MA-ELISA. Characteristics of the MA-ELISA and its procedures have been previously described by Wisselink et al. (2010).
The test results were calculated as percentage positivity (PP). Cut-off values of PP 10.6 (Wis-selink et al., 2010) and PP 20 were used. PP 20 was used for evaluations as in the applied supply chain meat inspection MA-ELISA test results were designated positive when PP was between 20 and 50 and strongly positive when PP was greater than 50.
MA-ELISA validation
To evaluate the MA-ELISA under field conditions, pig herds (n=11) with a MA infection were chosen. To identify these herds, a high number of positive serum samples and when available visual meat inspection findings (lymph nodes with granulomatous lesions) were used. To confirm the MA infection status on these farms, fattening pigs (n=22-68) which were nearly ready to slaughter were subjected to an intradermal tuberculin test into the base of the ear with 0.1 ml Avian Tuberculin PPD (25.000 I.U., ASG, Lelystad, The Netherlands).
Evaluation occurred after 36 to 72 hours by checking the injection site for signs of induration
and erythema. In the case that pigs reacted positive in tuberculin skin test, blood serum sam-ples and the submaxillary and mesenteric lymph nodes were collected at slaughter. In the la-boratory, serum samples were stored at -20 °C until analysis and the lymph nodes were exam-ined pathologically for caseous malformations and bacteriologically for MA, as described by Wisselink et al. (2010).
For evaluation of the specificity of the MA-ELISA under field conditions, pig herds (n=8) with only negative samples for MA were selected. Blood samples (50 ml) and the submaxil-lary and mesenteric lymph nodes were collected at slaughter and examined as described above.
Herd sensitivity calculations
The MA-ELISA was used as a herd diagnostic test; therefore the sensitivity as herd diagnostic test was calculated. Herd sensitivity was defined as the probability that a positive herd is di-agnosed positive following the evaluation of an aggregated set of serum samples. The herd sensitivity is the probability that the number of positive samples from positive herds is equal to or above a minimum needed number of positive serum samples.
The probability that from a certain set of serum samples (n) equal or less than a certain num-ber of serum samples (x) are positive can be calculated with the cumulative distribution func-tion of the binomial distribufunc-tion, equafunc-tion (1):
( )
n iHere the prevalence in the population (p) is the real prevalence in the population (Prev) mul-tiplied with the sensitivity (Sens) of the individual test. The probability to have 1 or zero posi-tives out of 36 samples (achieved over 12 batches) could therefore be calculated by equation 2:
The herd sensitivity (Sensherd) is the probability that out of the 36 samples more than 1 sample is positive. The probability to have more than 1 positive sample is one minus Pr (X≤1).
Sensherd = 1 –Pr (X≤1). (3)
The herd sensitivity was calculated for a range of test sensitivities at a range of prevalences as observed in the different validation trials.
Herd specificity (Specherd) was calculated in the same way. Herd specificity is 1 minus the probability that there are 2 or more false positive test results out of 36 samples. Sensitivity and specificity of the MA-ELISA and the pathological examination of the submaxillary lymph nodes was calculated with the results of the bacteriological examination of the sub-maxillary and mesenteric lymph nodes for MA as golden standard.
Statistics
To test the statistical significance of differences in proportion of positive samples a Chi square test was done.
Results
Serological MA-monitoring in the German and Dutch slaughterhouse(s)
In six Dutch slaughterhouses from January 2007 until June 2010 on 870 slaughter days blood serum samples were taken from 248325 pigs delivered by 4830 herds and examined for anti-MA antibodies. The results showed that 2293 (0.93%) serum samples had a PP in the range of 20 ≤ 50 and 202 (0.08%) had a PP > 50 in the MA-ELISA (Figure 1). From 4817 pig herds, 36 or more samples were collected during this period. Twenty five of these herds (0.5%) had a within herd prevalence above 2 of 36 at a PP > 20, and 3.357 herds (69.7%) at a cut-off value of PP 10.6 (Figure 1). In the supply chain meat inspection herds with 2 or more positive samples are considered at risk for MAA.
Fig. 1. Distribution of Percentage Positivity (PP) in MA-ELISA test results from 248235 blood serum samples collected in three Dutch slaughterhouses from January 2007 until June 2010 and from 57044 blood serum samples collected in one German slaughterhouse from October 2008 until April 2010.
In the German slaughterhouse blood serum samples were taken from 57044 pigs delivered by 1249 herds on 385 slaughter days from October 2008 until April 2010. From these serum samples 906 (1.59%) had a PP in the range of 20 ≤ 50 and 78 (0.14%) had a PP > 50 (Figure 1). From 574 herds, 36 or more samples were collected during this period. One hundred of these herds (17.4 %) had a prevalence above 2 of 36 at a PP > 20, and 562 (97.9%) at a cut-off value of PP 10.6.
The proportion of positive samples in the different months is shown in Figure 2. The propor-tion of positive samples and herds were significantly higher in the tested German populapropor-tion compared to the tested Dutch pigs (Chi-square; p < 0.05).
Fig. 2. Proportion of MA-ELISA positive serum samples in relation to the number of tested se-rum samples in The Netherlands (A) and in Germany (B) in the period 2008-2010.
Follow up investigations at farm level and tuberculination of serological MA positive pig herds
From the 11 farms that were suspected and came under investigation four serological MA positive pig herds in the Netherlands (farm A), Belgium (B) and Germany (D and E) were given follow up with additional testing as tuberculination was positive. The characteristics of
farm A were described by Wellenberg et al. (2010). In farm B the pigs were housed in a closed stable without bedding and all in / all out system was practised. At the breeder stage the piglets were supplied with peat. Farm D and E were fattening farms. The pigs were housed in a closed stable without bedding and all in / all out system was practised. However, the breeders supplying the piglets for both fattening farms supplied peat to suckling piglets and one of them kept the piglets in an open stable with paved surfaces outside. Seven other Dutch serological positive herds that were screened were negative at tuberculination (sample size per herd between 22 and 68).
Sensitivity of the MA-ELISA and the pathological examination
Results of bacteriological examination of the submaxillary and mesenteric lymph nodes from pigs from the four herds, which reacted positive in tuberculin skin test showed that on average 50% (188 of 375) of the pigs were infected with M. avium hominisuis (MAH) (Table 1), the minimum level of infection with MAH was 32% (6 of 19) (farm E) and the maximum level was 66% (38 of 55) (farm D). Granulomatous lesions in the submaxillary lymph nodes were detected by pathological examination in total 22.5% (85 of 378), the minimum prevalence of lesions was 8% (9 of 117) (Farm B) and the maximum was 31% (57 of 184) (Farm A).
The results showed that the sensitivity of the MA-ELISA at the four farms varied between 10 and 67%, when the cut-off value of PP 10.6 was applied and varied between 2.4 and 16.7%
when the cut off value of PP 20 was applied (Table 1). The sensitivity of the pathological ex-amination under laboratory conditions varied between 19.5 and 67% and on average it was 32.6%.
Table 1
Test characteristics of the applied MA-ELISA validated with animals from four positive farms.
1 Positive when M. avium bacteria were detected by bacteriological examination on submaxillary lymph nodes and/ or mesenteric lymph nodes.
2 Positive when granulomatous lesions were seen in the submaxillary lymph nodes during pathological examina-tion in laboratory.
3 Partly published by Wisselink et al. 2010
4 Partly published by Hiller et al. 2010 Se = Sensitivity, Sp = Specificity
PPV = Positive Predictive Value, NPV = Negative Predictive Value +ve = positive
B = Belgium, GE = Germany, NL = The Netherlands
Specificity of the MA-ELISA
Eight farms in the Netherlands with low risk status for MA infections according to the nega-tive serology were selected for sampling to determine the specificity of the MA-ELISA and the pathological examination for granulomatous lesions. From 293 pigs the submaxillary and mesenteric lymph nodes were tested bacteriologically negative for MA. Specificity of the
MA-ELISA at a cut-off value of PP 10.6 was 92.5% and at a cut-off of PP 20 was 100% (95%
CI: 98,4% - 100). The specificity of the pathological examination under laboratory conditions was 97% (Table 2).
Table 2
Results of serological, pathological and bacteriological examinations for M. avium infections on farms categorised at ‘low’ risk for a M. avium infection.
Pig Pathology2 Bacteriology1 Pathology Bacteriology PP
1 Positive when M. avium bacteria were detected by bacteriological examination on submaxillary lymph nodes and/ or mesenteric lymph nodes.
2 Positive when granulomatous lesions were seen in the submaxillary lymph nodes during pathological examina-tion.
-ve = negative, +ve = positive Sp = Specificity
Herd sensitivity calculations
When the observed range of MA-ELISA test sensitivities (4-60%) and the observed range for bacteriological prevalence of MA bacteria at herd level (30-65%) were applied for herd sensi-tivity calculations, the probability to have at least one positive serological sample varies be-tween 35 and 100% (Table 3). This is the probability that positive herds were recognized with the serological test.
The probability to obtain two or more serum samples positive was 8-100%, depending on the bacteriological prevalence (Table 3). This is the probability to become a high risk herd.
Table 3
Herd sensitivity of the MA-ELISA in a range of observed test sensitivities and bacteriological prevalences of M. avium at 1 or more and 2 or more positive samples for a positive herd diag-nosis.
probability ≥1 out of 36 positive blood serum samples
30% 35% 67% 81% 89% 97% 100% 100%
40% 44% 77% 89% 95% 99% 100% 100%
50% 52% 84% 94% 98% 100% 100% 100%
60% 58% 89% 97% 99% 100% 100% 100%
probability ≥2 out of 36 positive blood serum samples
30% 8% 29% 52% 64% 85% 98% 100%
40% 14% 42% 68% 79% 94% 100% 100%
50% 19,4% 54% 79% 89% 98% 100% 100%
60% 25% 64% 87% 94% 99% 100% 100%
Calculations of the apparent herd prevalence (AP) showed that at a cut-off of PP 20 (herd sensitivity and herd specificity set at 20% and 98.5% respectively) and at a low real preva-lence the AP did not decrease below 1.5%. These were false positives resulting from the pre-sumed 98.5% specificity, based on an aspre-sumed test specificity of 99.5%. The results from the
Dutch population indicate that the herd specificity of the test was higher than 98.5%, because the measured herd prevalence in the Dutch herds was only 0.5%.
Discussion
M. avium subsp. avium and M. avium subsp. hominissuis are relevant food safety risks in pigs (Komijn et al., 1999, Martin and Schimmel, 2000, Möbius et al., 2006). To put emphasis on the control of these pathogens in pig husbandry, in pig slaughterhouses in The Netherlands and in Germany a risk based supply chain meat inspection was introduced as an alternative system for meat inspection. In this paper we evaluated this newly developed supply chain meat inspection. The results show, that serological screening for MA infections had the ca-pacity to identify MA positive herds, with bacteriological examination for MA in lymph nodes as reference. The number of MA positive samples was 1.01% and 1.73% respectively in the Dutch and German pig population. These results are comparable to the prevalence of granulomatous malformations in lymph nodes, as other studies found 1.85% (Fischer, 1999), 0.89% (Meyer et al., 2007) and 0.48% (Lücker et al., 1997) lymph nodes with mycobacterial pathology, of which about one third showed to be bacteriological positive for MA.
A half percent of the Dutch herds and 17.4% of the German herds were tested positive with more than two out of 36 analysed blood samples. In the first place these figures show, that MA infection in pigs occur at a low level. Additionally they show that the prevalence of MA infections differs across populations. There was a higher level of positive samples and herds in the German population compared to the Dutch. At German breeder herds, peat is more fre-quently supplied (data not shown), which possibly explains the differences.
Also at three out of the four confirmed MA positive farms, there was an apparent relation with peat, that is usually used as a feed supplement (Matlova et al., 2005, Trckova et al., 2005). After intervention at the farms and the abolishment of supplying peat, serological re-sponses at the farms decreased (results not shown).
The applied MA-ELISA was validated on confirmed MA positive farms, resulting from the serological surveillance. The validation results showed, that the sensitivity of an individual test was low, i.e. varying sensitivities were found with an average of 4.3%. Nevertheless, it was shown that approximately 20% of bacteriological positive herds can be identified when
36 blood samples are tested and at least two samples need to be positive above PP 20 in the ELISA. An improvement of the MA-ELISA test sensitivity seems achievable, as in the former experimentally studies (Wisselink et al., 2010) and in some of the field farms these higher sensitivities were found. The low average sensitivity might be due to presence of infections with other MA serotypes (Dvorska et al., 2004) that have insufficient cross-immunity toward the antigens used in the test. Additional antigens could be added to the MA-ELISA test to improve its performance.
Comparing the sensitivities of the MA-ELISA to the inspection of the submaxillary lymph nodes after incision within the traditional meat inspection, one can conclude the traditional post mortem inspection has a low sensitivity for detecting MA affected carcasses as well (Brown and Neumann, 1979). A recent review showed that during meat inspection in Ger-many malformations in relation to mycobacteria were reported in 0.22% of slaughtered car-casses (Federal Statistical Office of Germany, 2007; according to BfR report). Caseous mal-formations in porcine lymph nodes and sometimes in kidneys, liver and spleen can be caused by mycobacteria (Thoen, 2006, Offermann et al., 1999, van Ingen et al., 2010), but most of them originate from Rhodococcus equi infections (Dvorska et al., 1999, Komijn et al., 2007, Pate et al., 2004). Recent studies showed lymph nodes without any lesions to harbour MA (Brown and Neumann, 1979, Dvroska et al., 1999). Henceforth the incision of submaxillary lymph nodes in the traditional meat inspection appears to be a sensitive and a non-specific test. Whereas it was shown that the non-specificity of the MA ELISA test was high, i.e.
close to 100%.
Furthermore, there are important advantages of the risk based supply chain meat inspection compared to the traditional meat inspection. Cross contamination of salmonella due to inci-sion is prevented (Peel and Simmons, 1978, SCVMPH, 2000) as lymph nodes do not have to be incised. Within the supply chain meat inspection system much more effort is done to con-trol MA with increased biosecurity standards and follow-up at farms. Prevention of infection with MA in swine that takes place at farm level has not been an active constituent in the tradi-tional meat inspection, which is an end of line control only.
It can be concluded that a population-wide screening for the presence of MA antibodies is capable of identifying pig populations that are at higher risk for MA infection. The validation
results of the applied ELISA indicate that positive farms will not in all cases be identified at first instance. However, on the farms that are identified, MA is actively combated from enter-ing the food chain. Positive MA test results are reported back to the pig producers, and,
results of the applied ELISA indicate that positive farms will not in all cases be identified at first instance. However, on the farms that are identified, MA is actively combated from enter-ing the food chain. Positive MA test results are reported back to the pig producers, and,