SEARCH

SEARCH BY CITATION

Keywords:

  • Salmonella spp;
  • risk assessment;
  • mathematical modeling;
  • foodborne pathogens

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Salmonella enterica serovar Enteritidis (S. Enteritidis) is one of the most prevalent causes for human gastroenteritis and is by far the predominant Salmonella serovar among human cases, followed by Salmonella Typhimurium. Contaminated eggs produced by infected laying hens are thought to be the main source of human infection with S. Enteritidis throughout the world. Although previous studies have looked at the proportion of infected eggs from infected flocks, there is still uncertainty over the rate at which infected birds produce contaminated eggs. The aim of this study was to estimate the rate at which infected birds produce contaminated egg shells and egg contents. Data were collected from two studies, consisting of 15 and 20 flocks, respectively. Faecal and environmental sampling and testing of ovaries/caeca from laying hens were carried out in parallel with (i) for the first study, testing 300 individual eggs, contents and shells together and (ii) for the second study, testing 4000 eggs in pools of six, with shells and contents tested separately. Bayesian methods were used to estimate the within-flock prevalence of infection from the faecal and hen post-mortem data, and this was related to the proportion of positive eggs. Results indicated a linear relationship between the rate of contamination of egg contents and the prevalence of infected chickens, but a nonlinear (quadratic) relationship between infection prevalence and the rate of egg shell contamination, with egg shell contamination occurring at a much higher rate than that of egg contents. There was also a significant difference in the rate of egg contamination between serovars, with S. Enteritidis causing a higher rate of contamination of egg contents and a lower rate of contamination of egg shells compared to non-S. Enteritidis serovars. These results will be useful for risk assessments of human exposure to Salmonella-contaminated eggs.

Impacts

  • This study has estimated the rate at which laying hens infected with Salmonella produce Salmonella-contaminated eggs.
  • The rate of egg shell contamination was higher per infected bird in high prevalence flocks, possibly due to a correlation between high Salmonella prevalence and poor hygiene standards. This means that high prevalence flocks contribute disproportionately to eggs with contaminated shells.
  • The rate of contamination was much higher for shells than for contents, and there were differences between Salmonella serovars, with Salmonella Enteriditis having a higher rate of contents but lower rate of egg shell contamination than other serovars.

Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Salmonella is one of the most prevalent causes for human foodborne illness (Rodrigue et al., 1990). Contaminated eggs produced by infected laying hens are thought to be the main source of human infection (Coyle et al., 1988; Rabsch et al., 2001; Gillespie and Elson, 2005; Gillespie et al., 2005), with the majority of human egg-related outbreaks being attributed to Salmonella Enteritidis (S. Enteritidis) (77.2%), followed at a distance by Salmonella Typhimurium (S. Typhimurium) (3.5%) (EFSA, 2010a,b).

It is important when developing and prioritizing surveillance and risk control programmes to have a good understanding of the risk of contamination of eggs from infected chickens and its variability. There is as yet an incomplete understanding of the risks of Salmonella infection in eggs from infected birds. Studies have looked at the proportion of positive egg shells from infected flocks (Kinde et al., 1996; Davies and Breslin, 2004; Chemaly et al., 2009; Renu et al., 2011). While some studies have been able to correlate the prevalence of infected flocks with the number of environmental sample positives (Wales et al., 2007), the rate of egg contamination per infected bird is still uncertain. Furthermore, there could be important differences between serovars. Of the most prevalent serovars in layer flocks, S. Enteritidis is thought to be the major contributor to human infections via eggs, although eggs are not considered to be an important source of S. Typhimurium for humans in most countries (EFSA, 2010a,b).

Experimental studies have investigated the rate of egg contamination of eggs from birds challenged with different Salmonella serovars. For example, Okamura et al., 2001a,b; recovered only S. Enteritidis and S. Heidelberg from egg shells and contents laid by intravenously inoculated hens (percentages of contaminated eggs 15.8 and 10 respectively). Salmonella Typhimurium, S. Infantis, S. Heidelberg and S. Montevideo inoculated hens did not produce any contaminated eggs (Okamura et al., 2001a). When the same study was repeated by intravenously (IV) inoculating more batches of hens the percentage of contaminated eggs (either on the outer or inner surface of the eggshells or in the egg contents) was 27.6 for S. Enteritidis, 3.1 for S. Typhimurium, six for S. Infantis, 9.4 for S. Montevideo, 4.5 for S. Heidelberg and 4.9 for S. Hadar (only S. Typhimurium and S. Enteritidis were recovered from egg contents) (Okamura et al., 2001b). In these studies, the challenge doses administered to the birds were high (5 × 106 colony forming unit IV) and therefore did not resemble the natural challenge to which the birds are exposed in field conditions.

As there is lack of data on the number of contaminated eggs produced by an infected flock (EFSA, 2010a,b), the aim of this study was to estimate the rate at which naturally infected birds of commercial laying flocks produce contaminated egg shells and egg contents. Faecal and environmental sampling and testing of ovaries/oviducts and caeca from 100 hens per flock were carried out in parallel with testing 4000 eggs in pools of six, with shells and contents tested separately and 20 farms included in the study. Previously developed Bayesian methods (Arnold et al., 2010, 2011) were used to estimate the within-flock prevalence of infection from the faecal, environmental and hen post-mortem data, and this was related to the proportion of positive eggs. Serovars in the infected flocks included S. Enteritidis, S. Typhimurium and non- S. Enteritidis/S. Typhimurium, and the intention was to investigate whether there were important differences in the rate of egg contamination between serovars.

Materials and Methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Data

Data were obtained from two studies where individual bird post-mortem examinations, environmental sampling and sampling of eggs were undertaken. All farms in study 1 were known to be infected with S. Enteritidis, whereas all farms in study 2 had been detected as Salmonella positive (not restricted to S. Enteritidis) through the National Control Programme for Salmonella in layers.

Study 1: environmental and individual bird sampling

In study 1 (15 farms), up to 300 birds were selected for the testing of ovaries/oviduct and caeca, with a total of 4418 birds tested across the 15 flocks. Environmental sampling consisted of three different methods: the method used in the EU baseline survey, (EFSA, 2006, 2007), the method used for official sampling in the National Control Programme for Salmonella in laying flocks (‘NCP sampling method’), Defra, 2007) and an in-house method (‘VLA sampling’) (Carrique-Mas et al., 2008). Briefly, the EU baseline survey involved seven tests, consisting of five 200–300 g composite faecal samples or five pairs of boot swabs, each representative of 1/5th of the laying house and 2 × 250 ml dust samples. The NCP sampling method consisted of two tests: one faecal sample formed from 2 × 150 g faecal samples, each representing half of the house and collected from the same locations as the EU baseline survey sampling, and one for 250 ml dust collected from prolific sources of dust throughout the house. The VLA sampling method consisted of up to 40 tests: 10–20 composite faecal samples, each weighing ~25 g and 10–20 dust samples, each weighing ~15 g, from representative point locations across the house, collected with BPW (buffered peptone water) impregnated gauze swabs (Robinson Healthcare 5345, Nottinghamshire, UK).

Study 1: egg sampling

In addition, up to 300 eggs were tested individually, in a method in which the shell and contents from each egg were tested together in 225 ml BPW. The samples in BPW were incubated at 37°C for 16–20 h, and 0.1 ml of incubated broth was inoculated onto modified semisolid Rappaport-Vassiliadis medium (MSRV; Mast DM440D, with addition of 1 mg/ml of novobiocin, Sigma N1628, Nottinghamshire, UK). MSRV plates were incubated at 41.5°C and examined after 24 and 48 ± h for suspect Salmonella growth. From any MSRV plate a 0.1-ml loop was taken and streaked onto Rambach agar (Merck 1.07500.0002, Dorset, UK). Inoculated plates were incubated at 37°C for 24 ± h, and presumptive Salmonella colonies were identified. When Salmonella was identified after the first 24 h on Rambach plates, no further plating from MSRV was carried out.

Study 2: environmental and individual bird sampling

Farms were recruited to obtain a range of Salmonella serovars after being identified via routine surveillance, and a total of 20 farms were recruited. In each farm, a range of environmental samples, one hundred hens and up to 4000 eggs, were collected, with a total of 2000 hens tested and 75011 eggs tested (in pools of six) across the 20 farms. Environmental sampling consisted of (i) the EU baseline survey method, (ii) the NCP sampling method but with five faeces and five dust samples per flock and (iii) the VLA sampling method with 20 faeces and 20 dust samples per flock. Furthermore, 100 individual freshly voided droppings were tested both individually and in pools of five, for which a semiquantitative dilution-enrichment method was used. Briefly, 100 ml of the initial BPW solution of 10 g faecal sample was used to make 10-fold dilution series in BPW to 10−7. The last dilution to test positive was recorded (Wales et al., 2006).

For the individual bird sampling, hens were culled by cervical dislocation on farm and stored at 4°C overnight. At post-mortem examination, ovaries/oviduct and caeca were aseptically removed and cultured separately as 25 g samples in 225 ml of BPW.

Study 2: egg sampling

In each flock, up to 4000 eggs (less if fewer eggs were available) were tested in pools of six, with shells and contents tested separately. The eggs were carefully cracked open without prior disinfection of the shell to maximize recovery from shells. Shells and contents were pooled separately in groups of six for culture and pre-enriched in 225 and 333 ml of BPW, respectively. Culture of eggs was as for study 1.

Serotyping

Salmonella colonies were confirmed by serotyping using the White-Kauffmann–Le Minor typing scheme, and phage typing for S. Enteritidis and S. Typhimurium isolates was performed using HPA typing schemes (Jones et al., 2000).

Statistical methods

The prevalence of Salmonella infection in each flock was estimated from the testing of ova and caeca (in 100 hens) and from the results of environmental sampling using the EU, NCP and VLA sampling methods, plus 100 droppings, using previously developed Bayesian methods (Arnold et al. (2010, 2011) (see supplementary information).

We assumed that the proportion of Salmonella-contaminated egg contents and egg shells (and consequently a mixture of contents and shells as in study 1) (μC, μS), respectively, was proportional to the number of infected birds, so that:

  • display math

where γa is the rate at which infected birds produce infected eggs, π is the within-flock prevalence of infected birds, and where a = S for shells and a = C for contents and a = CS for shells and contents tested together. The fit of the linear model was also compared with a quadratic model for both shells and contents as preliminary analyses had suggested a possible nonlinear relationship between prevalence and egg contamination, that is

  • display math

For the testing of individual eggs, the number of positive eggs was assumed to follow a binomial distribution with P = μa and = the number of eggs. Where eggs were tested in pools of six, it is assumed that there was no dilution effect of testing in pools, that is, it is assumed that there no loss in sensitivity from diluting a Salmonella-contaminated egg with five uncontaminated eggs. The proportion of positive pools of contents/shells is then given by

  • display math

For each pool, there were four possible outcomes, both contents and shells positive, only contents or shells positive, or both contents and shells negative, that is, data were multinomial. The distribution of these four outcomes for each flock followed a multinomial distribution with the following probabilities:

  • display math

where ρ is the covariance between shell and contents contamination and represents a potential increased likelihood of one being contaminated if the other is contaminated. To explore the possibility that the rate of egg contamination would vary between different Salmonella serovars, a model was fitted to the data where the rate of contamination could vary between serovars and compared with the fit of a model assuming no difference between the rate of egg contamination between serovars. Serovars were grouped into (i) S. Enteritidis (ii) S. Typhimurium and (iii) Salmonella other (SO), that is, non- S. Enteritidis/S. Typhimurium.

All model comparisons (linear versus nonlinear and contamination by serovar) were performed by use of the deviance information criterion (DIC) (Speigelhalter et al., 2002), which is a Bayesian analogue of the Akaike information criterion. To assist in the interpretation of the DIC, a DIC weight (wDIC) was calculated for each model being compared, which gives an estimate of the probability that each model is the best model for the data at hand and is calculated according to

  • display math

where ∆DIC was the difference between the model in question and the minimum value of the DIC for the models being compared, and the denominator was the sum of the differences over all the models being compared. The best fitting model out of those compared will be that with the highest DIC weight, and a value close to 1 indicates strong evidence that it is the best model.

All calculations were performed in winbugs 3.1 (Lunn et al., 2000), using a burn-in of 1000 iterations followed by 5000 iterations of the model.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Study 1

All flocks were positive for at least one faeces or dust sample (Table 1), and 8/15 had at least one positive egg. The proportion of eggs positive was highly variable between farms, with one farm having 15% of eggs positive. The linear model of the rate of egg contamination versus the proportion of infected birds appeared to produce a poor fit to the data (Fig. 1), and so a quadratic model was fitted to the data. The deviance information criterion (DIC) indicated strong evidence in favour of the quadratic model of the rate of egg contamination versus infection prevalence (P = 0.999 in favour of the quadratic versus the linear model). The rate of egg contamination was then given by 0.00084π + 0.16 π2 (95% CrI of linear coefficient 0.0005–0.031; 95% CrI of quadratic coefficient: 0.11–0.21).

Table 1. Summary of the results of environmental sampling, individual bird testing and the testing of eggs from flocks infected with Salmonella Enteritidis (study 1)
House IDIndividual bird resultsEU samplingNCP samplingVLA samplingEggs
Only caeca positiveOva/caeca positiveOnly Ova positiveTotal birds positive (% positive)Negative birdsFaeces positiveDust positiveFaeces positiveDust positiveFaecesDust
Positive/testedPositive/testedPositive /tested
11228949 (16.3)25142119/2011/200/300
20000 (0)29611111/205/200/200
339121465 (22.8)220321114/2018/202/300
40000 (0)30000000/204/200/200
520322678 (26.1)22140009/2011/200/200
63051348 (16.1)25001002/204/201/300
754211 (4.1)25522117/1021/302/300
818301058 (19.7)237321111/2019/202/200
923171454 (18)24621108/1010/100/200
101061812136 (45.3)16442113/108/1011/200
1155102085 (28.3)21532114/109/100/200
121382922189 (63)11152119/1010/1021/300
13663913118 (40.7)172521117/2018/2027/180
14865334173 (59.9)116421117/20-6/200
150000 (0)30002004/106/100/200
image

Figure 1. The fit of a linear model (dotted line) and a quadratic model (solid line) relating the infection prevalence of Salmonella to the observed proportion of positive eggs (shells and contents tested together) on each of 15 farms infected with Salmonella Enteritidis.

Download figure to PowerPoint

Study 2

Of 20 flocks tested, 18 had at least one faeces or dust sample positive (Table 2). Of the 18 positive flocks, 13 had at least one contaminated egg (at least one positive pool for shell or contents) (Table 3). Results from the semiquantification of the pools of five are given in the supplementary information, Table S2).

Table 2. Summary of the results (number of samples found positive) of environmental sampling, individual bird testing from flocks infected with Salmonella (study 2)
Farm IDIndividual bird resultsEU samplingVLA samplingNCP samplingIndividual droppings
Only caeca +ve (n = 100)Only ova +ve (n = 100)Ova/caeca +ve (n = 100)Total birds positiveFaeces +ve (n = 5)Dust +ve (n = 2)Faeces +ve (n = 20)Dust +ve (n = 20)Faeces+ve (n = 5)Dust +ve (n = 5)(n = 100)
  1. ‘-’ Denotes that no samples of the specified type were taken from the farm.

1----220100-
2234835511725223
301011224152
4010120024-3
52011738301714011
62156323081303
7592162265234
800003081303
951062265234
100110113112510-15
1110014010010-3
1200000000--0
13084122082011
1401120000010
1500221071422
1600000004000
1700000044322
1800000000000
191233182218163532
2000002046011
Table 3. The estimated flock infection prevalence of Salmonella Spp (estimated by a Bayesian model applied to environmental sampling data and testing of ova and caeca from individual hens) versus the number of pools of eggs (shells and contents) that were positive for each flock
Farm IDSerovarEstimated flock infection prevalenceEggs
Number of pools of eggsShellsContentsShells and contents
1Enteritidis1.70%666000
2Enteritidis57.85%667901893
3Enteritidis5.45%667310
4Enteritidis1.41%552000
5Enteritidis27.77%291501
6Enteritidis17.34%570020
7Enteritidis14.64%666102
8Enteritidis2.19%666000
9Enteritidis10.06%666101
10Typhimurium19.85%666220
11Typhimurium13.32%6661110
12Typhimurium0.37%666000
13Typhimurium9.41%666100
14Typhimurium1.54%440000
15Typhimurium9.11%666701
16Livingstone1.13%666200
17Virchow2.29%666300
18None isolated0%666000
19

Livingstone

Putten

O_Rough:D:L,W

Agona

36.85%6664621
20Virchow0.44%666000

One flock (ID2) had extremely high levels of faecal contamination of very-poor shell-quality eggs produced by birds that were nearly 3 years old but had not been revaccinated for Salmonella. The birds were also survivors of a Mareks disease outbreak that was followed by death of more that 50% of the birds on the farm. This situation is unrepresentative of normal commercial egg-laying flocks, and the farm ceased trading soon after the visit. In this case, data were compared with and without this flock. The inclusion of this flock in the model fitting process resulted in a very-poor fit with an estimation of the number of infected eggs in the other flocks. Comparisons using the linear model of egg contamination, indicated a predicted rate of 2.1% of infected contents and 4.7% of infected shells when data from flock ID2 were included and 0.24% and 0.53%, respectively, when excluded.

For the flocks excluding ID2, both linear and quadratic models of egg contamination in shells and contents were fitted versus the prevalence. Of these models, the best fitting model (i.e. with the lowest value of the deviance information criterion) was that with a linear relationship between the rate of egg contents contaminated versus the prevalence of infection and with a nonlinear (quadratic) relationship between the rate of shell contamination versus prevalence of infection. However, there was only weak evidence of this model in favour of one where both shells and contents had a quadratic relationship with infection prevalence (DIC weight of P = 0.21 for quadratic model for shells and contents versus P = 0.78 for quadratic model for shells and linear for contents). Of the three models considered, the linear model for both shells and contents was the poorest, with a DIC weight of P < 0.01, indicating strong evidence against this model.

The proportion of infected eggs from infected hens was estimated to be 0.24% for contents (95% CrI 0.13–0.37%), that is, the expected proportion of eggs with Salmonella-positive contents given a within-flock prevalence of π, equals 0.0024 π. For shells, the nonlinear relationship was given by 0.0053 π + 0.045 π2 (95% CrI of linear coefficient 0.0016–0.01; 95% CrI of quadratic coefficient: 0.022–0.077), that is, for very low values of within-flock prevalence (of the order of 1% or less) than the rate of shell contamination will be around 0.53%, approximately double that of contents, but as prevalence increases so the rate of shell contamination will increase more rapidly than contents contamination.

The Bayesian model of egg contents and shell contamination indicated a significant positive correlation between contents and shell contamination, with a mean correlation across all flocks of 0.25 (individual farm results and CrI's given in supplementary information, Table S3). Excluding farm 2, almost half the eggs with contaminated contents also had contaminated shells (six of 14, Table 3), and approximately 1/6th of the eggs with contaminated shells also had contaminated contents (6/82), much higher than would be observed if the contamination of contents and shells was not correlated.

Model comparisons via the DIC indicated that there was very strong evidence that the rate of egg contamination differed between serotypes (DIC weight of model with varying rate of contamination between serotypes = 0.999). Salmonella Enteritidis had the highest rate of contamination of egg contents, 0.32%, but had the lowest rate of contamination of egg shells (Table 4). The median rate of contamination of egg contents was equal for the non- S. Enteritidis serovars at 0.23%, but SO had a higher rate of egg shell contamination than S. Enteritidis or S. Typhimurium. The fit of the model to the data for the number of positive pools of eggs for shells and contents only and both shells and contents is given in Fig. 1. The overall fit of the model to the total number of positive pools in each flock was reasonable [total predicted positives was 105 (excluding flock 2) compared to 102 observed], although there is considerable between-flock variability in the rate at which infected hens produce infected eggs, making precise predictions of the number of positive eggs from each flock problematic (Fig. 2a,b).

Table 4. The estimates of the coefficients determining the rate of Salmonella egg contamination of shells and contents by serovar, from the results of a Bayesian model
SerovarRate of egg contamination
ShellsContents
Linear coefficientQuadratic coefficientLinear coefficient
  1. a

    SO- non-S. Enteritidis/S. Typhimurium.

All0.0053 (0.0016,0.010)0.045 (0.022,0.077)0.0024 (0.0013,0.0037)
S. Enteritidis0.0034 (0.0003,0.0078)0.019 (0.0007,0.052)0.0032 (0.0015,0.0059)
S. Typhimurium0.0094 (0.0023,0.016)0.015 (0.0005,0.081)0.0023 (0.0008,0.0055)
SOa0.025 (0.0084,0.041)0.029 (0.001,0.091)0.0023 (0.0007,0.0055)
image

Figure 2. The fit of the model to the observed number of positive pools of eggs for (a) total positive pools of contents (sum of pools positive for contents only and contents and shells) (b) total positive pools of shells.

Download figure to PowerPoint

In Fig. 3 is shown the proportion of positive pools versus the within-flock prevalence for each of the 19 flocks, which highlights the between-flock variability in the rate at which contaminated eggs are produced. There is an approximate linear relationship between the estimated infection prevalence and the prevalence of positive eggs except for the two flocks with very high egg prevalence – with the two highest prevalence farms included in the model (Farm 2 and Farm 15) a quadratic model for the rate at which infected hens produce infected eggs produced a significantly better fit to the data than a linear model (as was found for study 1 flocks).

image

Figure 3. The fit of the Bayesian model relating the infection prevalence of Salmonella to the observed number of positive pools of eggs on each farm for a) shells (quadratic model) and b) contents (linear model).

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

The estimation of the rate of egg contamination versus infection prevalence has been possible in this article due to extensive sampling on flocks, with a range of environmental samples and post-mortem examinations on culled hens, and prior knowledge of the sensitivity of these sampling methods from previous studies (Arnold et al., 2010, 2011). This has meant that the infection prevalence in each flock has been able to be estimated with relative precision, especially due to the inclusion of the individual-level data from the testing of ovaries/caeca in spent hens. Previous studies have found a correlation between the number of positive environmental samples and the proportion of eggs positive in a flock (Kinde et al., 1996; Davies and Breslin, 2004; Chemaly et al., 2009; Renu et al., 2011), which is likely to be due to both the number of environmental samples and the proportion of eggs positive being correlated with the within-flock infection prevalence.

The results in this study will be useful for the estimation of the risk of human exposure to Salmonella in eggs, for which the prevalence of contaminated eggs in Salmonella-infected flocks is a key input parameter. Without estimates of the rate of egg contamination per infected bird, such risk assessments have to rely on using observed prevalence estimates from egg sampling from infected flocks (e.g. EFSA, 2010b). However, the proportion of contaminated eggs depends upon the within-flock prevalence, which could vary between countries and could also be affected by control measures. An understanding of the contamination rate per infected bird allows differences in within-flock prevalence distributions between different countries to be directly accounted for. It also allows a more accurate assessment of the impact of changes to the within-flock prevalence distribution on the proportion of contaminated eggs, especially for the contamination of shells, which has been shown to have a nonlinear relationship with within-flock prevalence.

The nonlinear relationship between the prevalence and the rate of egg shell contamination may be due to a correlation between high Salmonella prevalence and hygiene standards, meaning that in the high prevalence farms there is also a greater risk per infected bird of faecal contamination of egg shells than in low prevalence ones. It is an important result for risk assessment because it suggests that high prevalence farms contribute disproportionately to eggs with contaminated shells, that is, the rate at which eggs are contaminated will be much higher in the high prevalence farms. However, while the finding of a nonlinear relationship for the rate of egg shell contamination versus infection prevalence was found in both study 1 and study 2, there were relatively few high prevalence farms in both studies. It would therefore be useful if further investigations of high prevalence flocks could be undertaken, although this may be problematic as within-flock prevalences of Salmonella found in the present study were low; this is possibly because the NCP in layer flocks has influenced farms with high prevalence to take action against Salmonella and to improve farm hygiene and pest control, leading to fewer high prevalence flocks.

The much higher rate of contamination of shells compared to contents is consistent with findings from other surveys and studies (e.g. Wilson et al., 1998; Davies and Breslin, 2004; Little et al., 2007, 2008). The ratio between the proportion of shells and contents contaminated with Salmonella in previous studies has been variable, with findings ranging from an approximately 5-fold greater prevalence in shells compared to contents (Little et al., 2008) up to 15-fold greater prevalence in shells (Little et al., 2007). The present study indicates that the relative difference between the rate of contamination on shells and in contents will vary according to the within-flock prevalence in the flock from which the eggs are sampled. This provides some explanation for the variability in the ratio of shells to contents positive, as the proportion of shells positive will be greatly influenced by the number of eggs sampled from high prevalence flocks.

There were some assumptions made that might be inaccurate and therefore affect the validity of the actual rates of contamination. Firstly, it was assumed that the sensitivity of the method used to detect Salmonella in eggs was 100%, as it employs a pre-enrichment stage and a highly effective means of selective enrichment and plating (Carrique-Mas & Davies 2008, Carrique-Mas et al., 2008). It is possible that this could be lower, which would result in an underestimation of the rate of contaminated eggs. It was also assumed that there would be no impact of pooling of six eggs on the likelihood of detection. This has been assumed in other studies (e.g. Kelly et al., 2009; Cowling et al., 1999), and the pooling of 40 eggs at AHVLA has been found to have no impact on detection compared to individual eggs (R.H. Davies, unpublished observations), and therefore, the pooling of a small number of eggs will be unlikely to have a large impact on the likelihood of successful Salmonella culture.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

This study was funded by the UK Department for Environment, Food and Rural Affairs, project OZ0332. We are grateful to AHVLA laboratory staff who carried out the testing of the large numbers of samples involved in the study.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  • Arnold, M. E., J. J. Carrique-Mas, I. McLaren and R. H. Davies, 2010: Estimation of the sensitivity of environmental sampling for detection of Salmonella Enteriditis in commercial egg-laying flocks relative to the within-flock prevalence. Epidemiol. Infect. 138, 330339.
  • Arnold, M. E., J. J. Carrique-Mas, I. McLaren and R. H. Davies, 2011: A comparison of pooled and individual bird sampling for detection of salmonella in commercial egg laying flocks. Prev. Vet. Med. 99, 176184.
  • Carrique-Mas, J., M. Breslin, A. Sayers, I. McLaren, M. Arnold and R. Davies, 2008: Comparison of environmental sampling methods for detecting Salmonella in commercial laying flocks in the UK. Lett. Appl. Microbiol. 47, 514519.
  • Carrique-Mas, J. J., and R. H. Davies, 2008: Bacteriological detection of Salmonella enteriditis in eggs: a review. Rev. Sci. Tech. 27, 657664.
  • Chemaly, M., A. Huneau-Salaun, A. Labbe, C. Houdayer, I. Petetin and P. Fravalo, 2009: Isolation of Salmonella enterica in laying-hen flocks and assessment of eggshell contamination in France. J. Food. Protect. 72, 20712077.
  • Cowling, D. W., I. A. Gardner and W. O. Johnson, 1999: Comparison of methods for estimation of individual-level prevalence based on pooled samples. Prev. Vet. Med. 39, 211225.
  • Coyle, E. F., S. R. Palmer, C. D. Ribeiro, H. I. Jones, A. J. Howard, L. Ward and B. Rowe, 1988: Salmonella enteritidis phage type 4 infection: association with hen's eggs. Lancet 2, 12951297.
  • Davies, R. and M. Breslin, 2004: Observations on Salmonella contamination of eggs from infected commercial laying flocks where vaccination for Salmonella enterica serovar Enteritidis had been used. Avian Pathol. 33, 133144.
  • Defra, 2007: UK national control programme for Salmonella in layers (gallus gallus).
  • EFSA, 2006: Preliminary report on the analysis of the baseline study on the prevalence of Salmonella in laying hen flocks of Gallus Gallus. EFSA J. 81, 171.
  • EFSA, 2007: Report of the Task Force on Zoonoses Data Collection on the Analysis of the baseline study on the prevalence of Salmonella in holdings of laying hen flocks of Gallus gallus. EFSA J. 97, 184.
  • EFSA, 2010a: Scientific Opinion on a quantitative estimation of the public health impact of setting a new target for the reduction of Salmonella in laying hens. EFSA J. 8, 1546.
  • EFSA, 2010b: Quantitative risk assessment of Salmonella Enteritidis in shell eggs in Europe. EFSA J. 8, 1588.
  • Gillespie, I., and R. Elson, 2005: Successful reduction of human Salmonella Enteritidis infection in England and Wales. Euro. Surveill. 10, E051117 051112.
  • Gillespie, I. A., S. J. O'Brien, G. K. Adak, L. R. Ward and H. R. Smith, 2005: Foodborne general outbreaks of Salmonella Enteritidis phage type 4 infection, England and Wales, 1992–2002: where are the risks? Epidemiol. Infect. 133, 795801.
  • Jones, Y. E., I. M. McLaren, C. Wray. 2000: Laboratory Aspects of Salmonella. Salmonella in Domestic Animals. CAB International, Eds C. Wray and A. Wray.
  • Kelly, L. A., L. Murchie, B. Xia, P. Whyte, and R. H. Madden, 2009: Probabalistic model for contamination of egg dishes with Salmonella spp made from shell eggs produced on the Island of Ireland. Int. J. Food Microbiol. 135, 187192.
  • Kinde, H., D. H. Read, R. P. Chin, A. A. Bickford, R. L. Walker, A. Ardans, R. E. Breitmeyer, D. Willoughby, H. E. Little, D. Kerr, and I. A. Gardner, 1996: Salmonella enteritidis, phase type 4 infection in a commercial layer flock in southern California: bacteriologic and epidemiologic findings. Avian Dis. 40, 665671.
  • Little, C. L., S. Walsh, L. Hucklesby, S. Surman-Lee, K. Pathak and Y. Gatty, 2007: Survey of Salmonella contamination of non-United Kingdom-produced raw shell eggs on retail sale in the northwest of England and London, 2005 to 2006. J. Food. Protect. 70, 22592265.
  • Little, C. L., J. R. Rhoades, L. Hucklesby, M. Greenwood, S. Surman-Lee and F. J. Bolton, 2008: Survey of Salmonella contamination of raw shell eggs used in food service premises in the United Kingdom, 2005 through 2006. J. Food. Protect. 71, 1926.
  • Lunn, D. J., A. Thomas, N. Best, and D. Spiegelhalter, 2000: WinBUGS - a Bayesian modelling framework: concepts, structure, and extensibility. Stat. Comput. 10, 325337.
  • Okamura, M., Y. Kamijima, T. Miyamoto, H. Tani, K. Sasai and E. Baba, 2001a: Differences among six Salmonella serovars in abilities to colonize reproductive organs and to contaminate eggs in laying hens. Avian Dis. 45, 6169.
  • Okamura, M., T. Miyamoto, Y. Kamijima, H. Tani, K. Sasai and E. Baba, 2001b: Differences in abilities to colonize reproductive organs and to contaminate eggs in intravaginally inoculated hens and in vitro adherences to vaginal explants between Salmonella enteritidis and other Salmonella serovars. Avian Dis. 45, 962971.
  • Rabsch, W., H. Tschape and A. J. Baumler, 2001: Non-typhoidal salmonellosis: emerging problems. Microbes Infect. 3, 237247.
  • Renu, Y., V. Tripathi and R. P. Sing, 2011: Salmonella occurrence in chicken eggs and environmental samples and their sero-prevalence in laying hens. Indian J. Anim. Sci. 81, 10871088.
  • Rodrigue, D. C., R. V. Tauxe and B. Rowe, 1990: International increase in Salmonella enteritidis: a new pandemic? Epidemiol. Infect. 105, 2127.
  • Spiegelhalter, D. J., N. G. Best, B. P. Carlin, and A. van der Linde, 2002: Bayesian measures of model complexity and fit. J. R. Statist. Soc. B 64, 583640.
  • Wales, A., M. Breslin and R. Davies, 2006: Semiquantitative assessment of the distribution of Salmonella in the environment of caged layer flocks. J. Appl. Microbiol. 101, 309318.
  • Wales, A., M. Breslin, B. Carter, R. Sayers and R. Davies, 2007: A longitudinal study of environmental Salmonella contamination in caged and free-range layer flocks. Avian Pathol. 36, 187197.
  • Wilson, I. G., J. C. Heaney and G. G. Powell, 1998: Salmonella in raw shell eggs in Northern Ireland: 1996–7. Communicable Disease and Public Health 1, 156160.

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
FilenameFormatSizeDescription
zph12038-sup-0001-TableS1-S3-FigS1.docWord document331K

Figure S1. A comparison of the observed proportion of eggs positive in study 1 (where contents and shells were tested together) versus the proportion of eggs positive in study 2 (where shells and contents were cultured separately).

Table S1. Description of the priors used in the Bayesian model determining the rate of egg contamination relative to infection prevalence and their source.

Table S2. The final 10-fold dilution to be positive for Salmonella for each of 20 pools of five faeces samples, taken from 20 flocks for which 10 g was initially mixed in 100 ml of BPW solution.

Table S3. Covariance and correlation between egg contents and egg shell contamination from a study of 20 farms (study 2).

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.