You are on page 1of 11

Journal of Applied Microbiology ISSN 1364-5072

ORIGINAL ARTICLE

Comparison of sampling methods to detect Salmonella


infection of turkey flocks
D. Mueller-Doblies1, A.R. Sayers2, J.J. Carrique-Mas1 and R.H. Davies1
1 Food and Environmental Safety Department, Veterinary Laboratories Agency Weybridge, New Haw, Addlestone, Surrey, UK
2 Centre for Epidemiology and Risk Analysis, Veterinary Laboratories Agency Weybridge, New Haw, Addlestone, Surrey, UK

Keywords Abstract
detection, environment, Salmonella, sampling
methods, turkeys. Aims: To compare the efficiency of various sampling methods for detection of
Salmonella in turkey flocks.
Correspondence Methods and Results: In a field study that compared various sampling
Robert H. Davies, Food and Environmental methods one pair of boot swabs taken from the whole turkey house provided
Safety Department, Veterinary Laboratories
suitably sensitive results for fattening and rearing flocks and was no less
Agency Weybridge, New Haw, Addlestone,
Surrey, KT15 3NB, UK.
sensitive than two pairs, each from half the house, tested as a pooled sample.
E-mail: r.h.davies@vla.defra.gsi.gov.uk The sensitivity was further enhanced by adding a dust sample. The dust sample
appeared to be particularly useful in flocks with a low prevalence, especially in
2008 ⁄ 2110: received 10 December 2008, breeding flocks, and was more sensitive than a method which used five pairs of
revised 10 December 2008 and accepted 5 boot swabs per flock. Combined incubation of a boot swab and a dust sample
January 2009 showed no interference between the two sample types and a maximum sensitiv-
ity of detection. Litter samples and commercial sponge drag swabs provided a
doi:10.1111/j.1365-2672.2009.04230.x
lower level of detection.
Conclusions: A single pair of boot swabs taken from the whole house is recom-
mended for routine sampling of commercial rearing or fattening flocks. An
additional dust sample could be added to increase detection in flocks with a
low prevalence or in breeding flocks, but adding an additional pair of boot
swabs would not increase detection compared with a single pair.
Significance and Impact of the Study: This study demonstrates that significant
efficiencies can be made in sampling programmes for detection of Salmonella
in turkey flocks without detracting from the sensitivity. Similar studies are
recommended for other poultry sectors, particularly in chicken breeding flocks.

although a significant decrease was observed over the pre-


Introduction
vious three years (EFSA 2008a).
Although the number of human cases of salmonellosis Most cases of human salmonellosis are attributed to
has declined in the past decade, there were still 14 060 the consumption of eggs and chicken meat (Humphrey
laboratory-confirmed cases of Salmonella infection 1990; Poppe et al. 1991; de Jong and Ekdahl 2006), but
reported in the UK in 2006, thus ranking it among the very little is known about the role of turkeys in the infec-
major causes of foodborne zoonoses. More than half of tion chain, and there are few references to turkey meat as
the reported cases in England and Wales in 2006 were a cause of human salmonellosis (Baggesen et al. 1996).
attributable to S. Enteritidis (7157 cases, 56% of all Relatively little research on the epidemiology and control
cases), and S. Typhimurium was the second most com- of Salmonella in turkeys has been done in Europe.
mon serovar observed (1735 cases, 14% of all cases) Between October 2006 and September 2007, a EU-wide
(Anonymous 2007; EFSA 2008c). In the EU, salmonellosis survey of turkey flocks (SANCO ⁄ 2083 ⁄ 2006) was carried
was the second most commonly reported foodborne zoo- out, providing the first robust data on the prevalence of
nosis in 2006, accounting for 160 649 reported cases, Salmonella infections in turkey flocks across the EU

ª 2009 The Authors


Journal compilation ª 2009 The Society for Applied Microbiology, Journal of Applied Microbiology 107 (2009) 635–645 635
Salmonella detection in turkey flocks D. Mueller-Doblies et al.

(EFSA 2008b). This survey found 29Æ3% of fattening chicken houses (broiler, rearing and laying flocks) (Davies
turkey flocks across the EU to be infected with Salmonella and Wray 1996a; Skov et al. 1999; Gradel et al. 2002),
spp., and the EU-wide prevalence was estimated as 30Æ7%, and the results may not necessarily be extrapolated to
with 3Æ8% being positive for either S. Typhimurium or turkeys.
S. Enteritidis and 27Æ3% being positive for other serovars. In this study, we selected 106 turkey flocks in England
In the UK, the prevalence of Salmonella spp. in fattening and compared the following sampling methods – (i) Full
turkeys was estimated as 32Æ2%, with 4Æ6% being positive EU survey protocol (i.e. five pairs of boot swabs cultured
for either S. Typhimurium or S. Enteritidis and 28% as five separate pairs); (ii) one pair of boot swabs taken
being positive for serovars other than S. Typhimurium or from the whole house and cultured as a pair; (iii) two
S. Enteritidis. pairs of boot swabs, each taken from half the house and
In the US, Salmonella spp. were reported to have cultured together; (iv) a litter sample; (v) a commercial
accounted for 42% of outbreaks and 51% of cases of gas- drag swab sample and (vi) a dust sample. From 45 of the
troenteritis due to bacterial pathogens irrespective of the 106 flocks, an additional 60 individual faecal samples
food vehicle, and 4Æ6% of Salmonella outbreaks were were analysed. Results were entered onto a database and
thought to be associated with consumption of turkey meat statistical methods were used to assess which was the
between 1973 and 1987 (Bean and Griffin 1990). A recent most sensitive sampling methodology.
publication from the US found S. Kentucky to be the most
prevalent serovar in turkeys, followed by S. Senftenberg
Material and methods
and S. Muenster (Santos et al. 2007). In another study from
the US, Cox and co-authors sampled individual faecal
Farms and flocks
droppings and found 17Æ7% of female and 31Æ1% of male
turkeys positive for Salmonella (Cox et al. 2000). A total of 106 flocks on 24 farms across England were
As the estimation of prevalence and identification of sampled between February and October 2007. Three
infected flocks in order to apply enhanced control mea- farms were visited more than once. The study focused on
sures depends on the sampling method, it is essential to farms where positive flocks had been identified in the EU
choose a sensitive method of sampling for all stages of turkey survey as well as through the companies’ own sur-
turkey production (Entis 2006; Fletcher 2006), and exten- veillance systems. The flocks that were sampled in the
sive sampling is initially required to define parameters for current study were the follow-on flocks for which the
testing this (Davies et al. 1998; Nayak et al. 2003). At the Salmonella-status was unknown. The 106 flocks included
same time, the costs of the sampling regime have to be 15 breeding flocks, 29 rearing flocks and 62 finishing
kept as low as possible to make it feasible on a large scale. flocks. On average, three to four flocks per farm and visit
There is also a need to define the sensitivity of different were tested.
sampling methods in relation to each other in order to be
able to provide advice to risk managers and the turkey
Sampling
industry. Several sampling methods are currently in use
by the turkey industry and by researchers to detect Boot swab method
Salmonella in occupied turkey houses, but very little work For the sampling methods involving five (‘EU survey
has been done to compare their performance. Sampling method’, ‘BS5’) and two spun olefin-fabric boot swabs
individual birds using cloacal swabs has been widely used (‘two-boot swab’ method, ‘BS2’), each house was divided
by the UK turkey industry, but various other methods to into five and two approximately equal sectors respectively.
sample the environment of the birds are in use, such as All sections of the house were represented in the sampling
litter samples, boot swabs ⁄ sock swabs, drag swabs, in a proportionate way. After walking through any boot-
dust samples ⁄ dust swabs and pooled individual faecal dips and before putting on the boot swabs, a new pair of
droppings. plastic over boots was put on to avoid contact of the disin-
The EU baseline survey of fattening and breeding fectant with the boot swab. For each pair of boot swabs a
turkey flocks (October 2006–September 2007) was based new pair of plastic over boots was used. The surface of the
on sampling occupied houses using five pairs of boot swabs (‘Overshoes’, Bowman and Knights, Norwich,
boot swabs, cultured as five separate pairs. Previous UK) was moistened with sterile distilled water prior to use.
VLA studies have shown that faeces samples alone are Within each chosen sector, at least 100 shuffling steps per
insufficient to demonstrate infection in some commercial pair of boot swabs were taken, ensuring that all parts of
egg laying flocks, and the addition of dust may improve the sector were sampled. For the ‘BS1’ - method, at least
the sensitivity (Carrique-Mas et al. 2008). Most published 100 steps were walked covering a circuit of the whole
studies have compared Salmonella sampling methods in house. Boot swabs were put into sterile plastic pots.

ª 2009 The Authors


636 Journal compilation ª 2009 The Society for Applied Microbiology, Journal of Applied Microbiology 107 (2009) 635–645
D. Mueller-Doblies et al. Salmonella detection in turkey flocks

Drag swab method 90 ml BPW. The 60 individual faeces were each incubated
A commercially available sponge drag swab (5 · 10 cm in 20 ml BPW.
Polywipe Premoistened Sponge Swab with Draw Cord, The combined boot swab ⁄ dust swab sample was
Medical Wire and Equipment, Corsham, UK) was used. incubated in 450 ml BPW.
The drag swab sample was taken by walking around the Selective enrichment of Salmonella was carried out
whole of the house (at least 100 steps) and dragging the using Modified Semi-solid Rappaport-Vassiliadis agar
swab along. The drag swab was then put into a plastic pot. (MSRV), followed by Rambach agar as described previ-
ously (Davies and Wray 1994).
Litter sampling method
Forty pinches of litter were collected from the whole
Serotyping
house into a sterile plastic pot.
Serotyping was performed according to the Kauffmann–
Dust sampling method White–Scheme (Popoff 2001) and phage typing was
Dust, representative of the whole house, was collected performed according to the HPA Colindale Scheme
into a sterile plastic pot using a gloved hand. Dust was (Anderson et al. 1977). One colony per plate, from the
collected from horizontal surfaces such as ledges, pipes, end of the streak, was picked and serotyped.
beams and partitions, but not from feeders or heaters.
Of the 106 visited flocks, 71 flocks were sampled with the
Statistical analysis
whole range of sampling methods, whereas for the remain-
ing 35 flocks, a reduced set of samples was taken. For all The statistical analysis was done on two outcome measures
production types (i.e. breeding, rearing or finishing units) (i) whether or not Salmonella was detected at a flock visit
the same sampling approach was applied. For free range and (ii) the proportion of samples positive. The sampling
flocks, samples were only collected inside the houses. methods were compared using the GLMM (generalized
From 45 of the 106 flocks, 60 individual faecal samples linear mixed models) procedure run in GENSTAT. The
were collected in addition to the samples described above. farm ⁄ house ⁄ visit combinations were specified as random
effects and sampling method as a fixed effect.
Sampling of individual faeces Two models were fitted to the results from 106 flocks
Sixty fresh faeces from the whole house were collected using the six methods (BS1, BS2, BS5, litter sample, drag
individually into sterile faeces specimen pots. swab, dust sample). The first (model 1) included these
methods individually and the second (model 2) combined
Combination sampling method the BS1 and dust sample results into a single outcome,
Sixty-eight flocks were sampled of which 22 proved to be regarding positive as either one or both methods positive.
Salmonella-positive and 46 proved to be negative. One A further model compared the proportions of flocks posi-
pair of boot swabs was collected from the littered floor tive by the individual faecal sampling with the other
area of each flock as described above. Dust was sampled methods.
with a large gauze swab (Kleenex Readiwipe, Robinson All pairs of methods were compared by dividing the
HealthCare, Chesterfield, UK), which had been premois- differences in the predicted means (on the logit scale) by
tened with Buffered Peptone Water. The swab was swept their standard errors and treating the results as a standard
through dusty surfaces, typically beams and ledges, in at normal deviate. The results are given as two-tailed
least 10 different representative parts of the house until it P-values. Because of the multiple comparisons, the
was liberally coated with dust. emphasis is placed on P-values <0Æ01.
Comparisons of the flock type prevalences detected by
BS5 were done by Fisher’s exact test.
Processing and culture of samples
Samples were normally transported to the laboratory
Results
within 6 h and incubated in buffered peptone water
(BPW) at 37C for 18 h. In some cases, samples were
Detection of Salmonella in turkey flocks
stored overnight at 4C before incubation was started.
Each pair of boot swab samples was incubated in The full set of samples was collected from 71 flocks, of
225 ml BPW, except for the two pairs from half houses, which 52 (73Æ2%) tested positive for Salmonella in at least
which were incubated together in 450 ml BPW. Litter and one sample. Of the remaining 35 flocks, where one or more
dust samples were divided and each half was incubated in samples were not taken, 29 (82Æ9%) tested positive for
225 ml BPW. The drag swab sample was incubated in Salmonella. Therefore, we can regard the overall percentage

ª 2009 The Authors


Journal compilation ª 2009 The Society for Applied Microbiology, Journal of Applied Microbiology 107 (2009) 635–645 637
Salmonella detection in turkey flocks D. Mueller-Doblies et al.

Table 1 Number of Salmonella-positive houses in visited farms with Table 2 Results by serovar
more than one house
Number of Number of positive
Number of visits positive farms ⁄ positive
Serovar flocks companies
1 All houses negative 4Æ5%
3 One house positive 13Æ6% S. Typhimurium (STM) (including 16 8 farms
18 Two or more houses positive 81Æ8% all phagetypes and
14 All houses positive 63Æ6% untypeable isolates)
Total: 22 visits STM DT104 10 5 farms
STM DT120 1 1 farm
STM U302 2 1 farm
of 76Æ4% positive flocks as a lower limit for the percentage STM untypeable 2 1 farm
of positive flocks. Eight out of 15 breeding flocks (53%), 29 O rough:e,h:1,2 1 1 farm
O Rough:e,h:1,5 1 1 farm
of the 29 rearing flocks (100%) and 44 of the 62 finishing
O rough:i:1,2 1 1 farm
flocks (71%) were positive for Salmonella. Senftenberg 5 1 farm
To investigate whether it is likely that a farm where one 4, 12:d:- 1 1 farm
positive house was identified has one or more other posi- Kedougou 12 6 farms
tive houses, results from 97 flocks included in 22 visits were Derby 2 2 farms
analysed and are shown in Table 1. Farms with only one Indiana 2 1 farm
house and visits where only one house was sampled were Saintpaul 1 1 farm
Montevideo 2 1 farm
not included in the analysis. Of the 22 visits, only one farm
Kottbus 14 5 farms
(4Æ5%) had all houses negative for Salmonella. Three farms Newport 33 6 farms
(13Æ6%) had only one positive house; whereas the remain- (3 belonging to
ing 18 farms (81Æ8%) had two or more positive houses on same company)
the date of the visit. Fourteen out of the 22 sampled farms
(63Æ6%) had all houses positive for Salmonella.
Testing individual faeces did not add any more infor- two flocks negative, two flocks positive dust only, one
mation in terms of presence ⁄ absence of Salmonella, as all flock positive litter and dust only, one flock positive BS2
flocks, which had one or more positive individual faecal only; no individual faeces positive).
samples, also had at least one other positive sample (boot
swabs, litter, dust or drag swab) (data not shown).
Sensitivity of the BS1-method combined with a dust
sample
Serotyping of Salmonella isolates
Results from 65 eligible positive flocks were analysed to
From all visits, a total of 13 different serovars was see how many of the positive flocks were detected by the
isolated. Results are shown in Table 2. In 71 of the 81 combination of one pair of boot swabs combined with a
positive flocks (88%), only one serovar was detected, dust sample. Four flocks (6Æ2%) would have been missed
whereas in 10 flocks (12%) more than one serovar was if only these two samples had been taken. As the addition
detected. The most frequently isolated serovars were of a dust sample enhances the chance of detecting flocks
S. Typhimurium (eight farms), S. Kedougou (six farms), with a low prevalence, this method would be particularly
S. Newport (six farms) and S. Kottbus (five farms). suitable for monitoring of breeding flocks.

Sensitivity of the BS5 (EU survey) method Sensitivity of the BS5 – method for different flock types
The results from 77 eligible positive flocks were analysed Eight of 15 breeding flocks (53%), 29 of the 29 rearing
to investigate the detection of positive flocks by the BS5 flocks (100%) and 44 of the 62 finishing flocks (71%) were
method compared to BS2, BS1, dust sample, drag swab positive for Salmonella, indicating a significantly lower
and litter sample. A flock was defined as positive if any prevalence in breeding flocks than in rearing and finishing
one of the samples was positive. If only the BS5 method flocks (P < 0Æ001). To further analyse whether different
had been used, 9 of the 77 positive flocks (11Æ7%) would sampling methods should be used to test different types of
have been missed. Interestingly, four out of the nine flocks, we compared the number of BS5 samples that tested
flocks that would have been missed originated from the positive for the different flock types and the number of
same farm, and this particular farm was a breeding farm flocks from the different production groups (i.e. breeding,
with a low prevalence of Salmonella (six flocks sampled; rearing and finishing flocks) that would have been missed

ª 2009 The Authors


638 Journal compilation ª 2009 The Society for Applied Microbiology, Journal of Applied Microbiology 107 (2009) 635–645
D. Mueller-Doblies et al. Salmonella detection in turkey flocks

100
100
All flocks (n = 77) Finishing flocks (n = 40)

40 60 80
40 60 80

Percent (%)
Percent (%)

20
20

0
0
0 1 2 3 4 5 0 1 2 3 4 5
No. pairs of boot swabs positive No. pairs of boot swab positive

100
100
Rearing flocks (n = 29) Breeding flocks (n = 8)

40 60 80
40 60 80

Percent (%)
Percent (%)

20
20

0
0

Figure 1 Number of BS5 samples that tested 0 1 2 3 4 5 0 1 2 3 4 5


positive for the different flock types. No. pairs of boot swabs positive No. pairs of boot swabs positive

by the BS5-method. Results are shown in Fig. 1. When all Salmonella was detected, the within-flock prevalence
flock types are analysed together, 53 of the 77 positive ranged from 1Æ7% to 96Æ7% with a mean of 42Æ9% and
flocks had all five pairs of boot swabs positive for Salmo- a median of 26Æ7%.
nella. Only 9 of the 77 (11Æ7%) flocks were not detected by
any of the five pairs of boot swabs, but were positive by one
Statistical comparisons of sampling methods
of the other sample types, as described above. However, if
the different types of flocks are analysed separately, results Samples that were included in the statistical analysis are
vary markedly between the groups. Four out of the 40 shown in Table 3.
(10%) positive finishing flocks were negative by any of the
five boot swabs, whereas only 1 of the 29 (3Æ4%) positive Model 1
rearing flocks were not detected by any of the five boot
Salmonella detection
swabs. In contrast to that, four of the eight (50%) of posi-
tive breeding flocks were missed by the five pairs of boot There were highly significant differences between the
swabs, indicating that the use of five pairs of boot swabs for sampling methods (Wald test, P £ 0Æ001). The predicted
sampling might not be sufficiently sensitive to detect a rea- means from the fitted model back-transformed to the
sonable percentage of positive breeding flocks. original scale (probabilities) and the P-values for the
method comparisons are given in Table 4a.
The five pairs of boot swabs (BS5) and the dust sample
Individual faeces results
were equally effective and significantly more sensitive than
In 18 flocks (four breeding, two rearing and 12 finish- the drag swab and the litter sample. The single pair of
ing flocks), Salmonella was not detected in individual boot swabs (BS1) was nearly as good as BS5 or dust and
faeces samples. In the remaining 27 flocks (60%), where also differed significantly from the drag swab at

Table 3 Samples included in the statistical


analysis Number of Number of
Number Number of % positive samples positive % positive
Type of visits positive visits visits (a) cultured samples samples (b)

BS1 102 60 58Æ8 102 60 58Æ8


BS2 100 56 56Æ0 100 56 56Æ0
BS5 102 68 66Æ7 510 294 57Æ6
Drag swab 93 38 40Æ9 93 38 40Æ9
Litter 95 50 52Æ6 207 110 53Æ1
Dust 86 56 65Æ1 172 100 58Æ1
All types 578 328 56Æ7 1184 578 48Æ8

ª 2009 The Authors


Journal compilation ª 2009 The Society for Applied Microbiology, Journal of Applied Microbiology 107 (2009) 635–645 639
Salmonella detection in turkey flocks D. Mueller-Doblies et al.

Table 4 Model 1: (a) Salmonella detection and (b) proportion of samples positive: P-values for method comparisons*

P-values for method differences


Proportion of
Method flocks positive BS1 BS2 BS5 Drag swab Dust sample

(a)
BS1 0Æ590 *
BS2 0Æ556 0Æ420 *
BS5 0Æ668 0Æ052 0Æ006
Drag swab 0Æ439 <0Æ001 0Æ007 <0Æ001
Dust sample 0Æ671 0Æ055 0Æ008 0Æ941 <0Æ001
Litter sample 0Æ527 0Æ136 0Æ486 <0Æ001 0Æ046 0Æ001

P-values for method differences


Mean proportion
Method samples positive BS1 BS2 BS5 Drag swab Litter sample

(b)
BS1 0Æ590 *
BS2 0Æ554 0Æ456 *
BS5 0Æ578 0Æ747 0Æ522
Drag swab 0Æ445 0Æ003 0Æ027 <0Æ001
Dust sample 0Æ605 0Æ711 0Æ231 0Æ361 <0Æ001
Litter sample 0Æ493 0Æ046 0Æ150 0Æ003 0Æ267 0Æ002

*Bold values indicate a statistically significant result.

P < 0Æ001. The isolation rate from the single pair of boot Table 5 Model 2: Salmonella detection: P-values for method
swabs was greater than that from two pairs pooled (BS2), comparisons*

but not significantly so. P-values for method differences


Proportion of
Method flocks positive BS1+dust BS2 BS5 Drag swab
Proportion of samples positive
BS1 + dust 0Æ727 *
There were again highly significant differences between BS2 0Æ560 <0Æ001 *
the methods (Wald test, P < 0Æ001). The predicted means BS5 0Æ670 0Æ161 0Æ007
from the fitted model back-transformed to the original Drag swab 0Æ439 <0Æ001 0Æ006 <0Æ001
scale (probabilities) and the P-values for the method Litter sample 0Æ528 <0Æ001 0Æ462 <0Æ001 0Æ044

comparisons are given in Table 4b. *Bold values indicate a statistically significant result.
There was no evidence of differences among the boot
swab methods (BS1, BS2 and BS5) and the dust sample
and all were superior to the drag swab. BS5 and the
Results for individual faecal samples
dust sample were also significantly greater than the litter
sample. A comparison of all the individual methods using the
GLMM model showed that the proportion of flocks with
Model 2 one or more positive individual faecal samples was signif-
icantly higher (P = 0Æ001) than the proportion positive
Salmonella detection
using the drag swab method (back-transformed means
There were highly significant differences between the 0Æ64 and 0Æ44 respectively). It did not differ significantly
methods (Wald test, P £ 0Æ001). The predicted means from any of the other methods,
from the fitted model back-transformed to the original
scale (probabilities) and the P-values for the method
Results for samples from breeding flocks
comparisons are given in Table 5.
The BS1+dust combination and BS5 were signifi- Table 6 shows the results of sampling 15 breeding
cantly better than BS2 and the litter sample, which flocks. BS5 and BS1 provided equivalent detection
gave similar results. The drag swab again had the worst results but three flocks were only detected by dust
performance. samples.

ª 2009 The Authors


640 Journal compilation ª 2009 The Society for Applied Microbiology, Journal of Applied Microbiology 107 (2009) 635–645
D. Mueller-Doblies et al. Salmonella detection in turkey flocks

(Arnold et al. 2005, 2008). In our study, there was no


Results for combination sampling study
flock that was only identified as positive by taking
Table 7 shows the results of sampling 22 Salmonella-posi- individual faecal samples, as all faeces-positive flocks
tive turkey flocks. Forty-six additional flocks were also were also positive by at least one of the other methods
sampled with negative results. Salmonella was found in all used.
22 positive flocks where boot swabs and dust swabs were Collecting and pooling pinches of litter from several
combined as a single sample, compared with 20 flocks different locations in the house is another way of collect-
where two pairs of boot swabs alone were used and 17 ing faecally-soiled material from a large number of birds,
flocks were a dust swab only was used. but due to the different types of litter used (for example
straw or shavings) this method is difficult to standardize
and sampling of dry straw bedding is difficult. It has also
Discussion
been shown, that direct litter sampling may be inferior to
Various sampling methods are currently used by the drag swabs (Hayes et al. 2000), although this depends on
turkey industry to monitor flocks for the presence of the sampling protocol employed.
Salmonella, but hardly any data are available on the Boot swabs are a convenient alternative to litter sam-
relative sensitivity of these methods. As most studies ples, as they collect faecal material from a larger num-
comparing sampling methods have been carried out in ber of birds than litter samples taken from a limited
chicken houses, these data have to be interpreted with number at point locations. Boot swabs can also be
care and only limited statements can be made about their easily posted to the laboratory and potentially provide a
validity for turkey flocks. In small turkey companies or more standardized sampling method than litter sam-
independent farms, little monitoring is carried out, even pling, which is more likely to be influenced by the dili-
in breeding flocks, and sensitive as well as economic and gence of the sampler. These considerations as well as a
practical methods are required to optimize monitoring theoretical ability to detect a 1% within-flock prevalence
programmes. (Skov et al. 1999) were instrumental in the decision to
In the UK turkey industry, cloacal swabs have been choose five pairs of boot swabs (cultured separately) as
widely used as part of a routine monitoring scheme. a sampling method for the EU Salmonella Baseline
However, they have been shown to be relatively insensi- surveys of broilers, turkeys and laying hens on floor
tive, as they sample only a small amount of faeces (Funk systems (i.e., barns and free-range houses) (Anonymous
et al. 2000; Kotton et al. 2006), and a large number of 2004, 2005, 2006) and for the statutory monitoring of
birds must be tested to compensate for intermittent shed- chicken breeding flocks [Commission Regulation (EC)
ding and for a low within-flock prevalence (Cannon and No. 1003 ⁄ 2005)].
Roe 1982). Sensitivity may be further reduced when indi- An alternative to the boot swabs is the sock model
vidual samples from a number of animals are pooled (Gradel et al. 2002), and it was shown that two pairs
(Arnold et al. 2005, 2008), so alternative methods are of socks, analysed as one sample, are at least as sensi-
needed to get reliable information about the Salmonella- tive as the collection of 60 individual broiler faeces,
status of a flock. analysed as one sample (Gradel et al. 2002). Another
A widely-used approach is environmental sampling of study showed, that five pairs of socks appeared to be
houses, which in general is regarded as more sensitive as effective in detecting Salmonella in chicken flocks as
than sampling a large number of individual birds, pro- 60 faecal pools (each consisting of five individual fae-
vided that appropriate samples are collected and tested ces), whereas one pair of socks proved to be inferior to
(Davies and Wray 1996a). The most common environ- 60 pools of five, particularly in flocks with low preva-
mental samples taken in poultry houses are faeces, litter lence (Skov et al. 1999). In another study, surgical shoe
and dust. Salmonella-positive faeces provide indication of covers resulted in better recovery of Salmonella than
a current infection of the flock, whereas dust may indi- traditional drag swabs (McCrea et al. 2005). In a UK
cate previous infection, which has passed its peak, result- study, sock swabs were found to be more laborious
ing in a low within-flock prevalence at the time of and less sensitive than boot swabs (Food Standards
sampling. Agency 2006).
Collecting individual faeces and culturing them indi- Drag swabs are widely used in the poultry industry in
vidually involves considerable work on the farm as well the USA, and slightly different models are in use. Most
as in the laboratory. Alternatively, individual faeces models which have been described in the scientific litera-
could be cultured as pools, but it is likely that pooling ture consist of large gauze pads, which are dragged along
might result in a loss of sensitivity due to a dilution the floor of the poultry house picking up faeces from a
effect, as described previously for pigs and cattle large number of birds. Different sizes of gauze pads

ª 2009 The Authors


Journal compilation ª 2009 The Society for Applied Microbiology, Journal of Applied Microbiology 107 (2009) 635–645 641
Salmonella detection in turkey flocks D. Mueller-Doblies et al.

have been reported (usually between 5 · 5 cm and 10 · Table 6 Breeding flocks tested for Salmonella
10 cm, constructed as a multiple swab assembly), and Farm House ID BS5 result BS1 result Dust result
studies have looked at different saturating ⁄ transport
media, such as skim milk, saline, buffered peptone water 1 A1 ) ) +
or chicken broth (Kingston 1981; Opara et al. 1992, 1994; 1 A2 ) ) +
1 B2 ) ) +
Davison et al. 1995; Byrd et al. 1997; Caldwell et al. 1998;
1 C1 ) ) )
Rolfe et al. 2000; Buhr et al. 2007). However, it is difficult 1 C2 ) ) )
to draw definitive conclusions from these studies as the 1 D* ) ) )
drag swab assemblies and the transport media vary 2 W1R ) ) )
between the different studies. Most of the drag swab 2 W1L ) ) )
models are also quite awkward to assemble, and use in 2 W2R + + )
densely-stocked poultry houses, especially turkey houses 2 W2L ) ) )
3 H4 + + Not sampled
where birds are less willing to move out of the paths of
3 H5 ) ) Not sampled
the sampler, can be difficult. 3 H6 + + Not sampled
In our study, we investigated a different type of drag 3 H7 + + Not sampled
swab which is marketed in the UK. There are no data 4 ) ) Not sampled
available comparing the performance of this type of drag
Positive flocks are marked in bold.
swab with the US style drag swab assemblies, but as it has
*Farm 1, house D was positive only by the BS2 sample.
a comparatively small surface and is also very light, it
may not achieve the same degree of contact with the
faeces as the gauze models and is therefore likely to be Table 7 Results of the comparison of sampling turkey houses using
less sensitive. boot swabs (BS), dust and combined samples – number of positive
Collection of dust is not as widely used as other samples

methods, but as it has been shown that Salmonella can BS+Dust


survive in dust for several months, this method pro- Flock ID BS2 Dust (combined) Serotype
vides a valuable tool for the detection of infections
1 + + + S Derby
which peaked prior to the time of sampling (Davies
2 + + + S Derby
and Wray 1996b). 3 + + + S Derby
Our results show that one pair of boot swabs taken 4 + + + S Derby
from the whole turkey house gives suitably sensitive results 5 + + + S Derby
for fattening and rearing flocks and is very cost-efficient. 6 + + + S Derby
The sensitivity can be further enhanced by adding a dust 7 + + + S Derby
sample, which is usually easy to collect from horizontal 8 + + + S Derby
9 + + + S Derby
surfaces such as partitions, beams and ledges at the same
10 + + + S Derby
time as carrying out the boot swabbing. The dust sample 11 + + + S Derby
seems to be particularly useful in flocks with a low preva- 12 + + + S Derby
lence, for example in breeding flocks, and our results 13 + + + S Derby
show that four of the eight (50%) of all Salmonella- 14 + + + S Derby
positive breeding flocks would have been missed by the 15 ) ) + S Derby
BS5 method, whereas the dust sample detected three 16 + ) + S Derby
17 + + + S Newport
out of four of those flocks that were negative by the BS5
18 + ) + S Derby
sample. 19 ) + + S Derby
Dust can easily be collected in houses with controlled 20 + ) + S Derby
ventilation where copious dust accumulates on extractor 21 + + + S Derby
fan baffles, however, it is not always present in some 22 + ) + S Derby
houses in sufficient amounts, especially in open (pole) total 20 17 22
barns or in houses of very young flocks, but the latter BS2 = Two pairs of boot swabs cultured together.
are not routinely sampled. In other cases, dust is pres-
ent only on high beams and is not easy to sample.
Also some farmers prefer to sample faeces since dust a large surface area using a moist fabric swab. Salmo-
sampling requires the wearing of a face mask and may nella appears to have a relative advantage in dust com-
require the use of goggles on some occasions. In such pared to other Enterobacteriaceae, which do not tend to
situations, the fine layer of dust can be gathered from survive in high numbers in dry conditions for such a

ª 2009 The Authors


642 Journal compilation ª 2009 The Society for Applied Microbiology, Journal of Applied Microbiology 107 (2009) 635–645
D. Mueller-Doblies et al. Salmonella detection in turkey flocks

long period (Haysom and Sharp 2003). This makes the References
detection of Salmonella in dust easier than in faeces
Anderson, E.S., Ward, L.R., Saxe, M.J. and de Sa, J.D.
samples where there is normally large, and potentially
(1977) Bacteriophage-typing designations of Salmonella
overwhelming competitive flora, which must be effec- typhimurium. J Hyg (Lond) 78, 297–300.
tively suppressed by selective enrichment and plating Anonymous (2004) Baseline study on the prevalence of Salmo-
before Salmonella can be identified. As Salmonella has nella in laying flocks of Gallus gallus in the EU – Technical
been reported to survive in poultry houses for at least specifications. SANCO ⁄ 34 ⁄ 2004 Rev 3. http://ec.europa.eu/
53 weeks in dust (Davies and Wray 1996b), a dust food/food/biosafety/salmonella/tech_spec_sanco-34-2004_
sample can provide an indication of an earlier infection rev-3_en.pdf (accessed 26 June 2008).
which has become undetectable by faecal monitoring. Anonymous (2005) Baseline study on the prevalence of Salmo-
However, faeces may provide a more accurate indica- nella in broiler flocks of Gallus gallus in the EU – Techni-
tion of substantial current infection, which may be cal specifications. SANCO ⁄ 1688 ⁄ 2005 Rev 1. http://
more relevant to contamination of carcasses at slaughter ec.europa.eu/food/food/biosafety/salmonella/tech_spec_
and hence to public health. sanco-1688-2005_rev1_en.pdf (accessed 26 June 2008).
Our data show further, that the drag swab and the lit- Anonymous (2006) Baseline study on the prevalence of Salmo-
ter sample are less effective for the detection of Salmo- nella in broiler flocks of turkeys in the EU –Technical
nella, and both methods only seem to be sensitive enough specifications. SANCO ⁄ 2083 ⁄ 2006. http://ec.europa.eu/
for flocks with a relatively high prevalence. It is not clear food/food/biosafety/salmonella/tech_spec_sanco-2083-
why the litter sample is less effective, but it might be due 2006_rev1_en.pdf (accessed 26 June 2008).
to the effect of straw, which provides a sample with less Anonymous (2007) Zoonosis report United Kingdom for
available surface area for faecal contamination than 2006. London, UK: DEFRA Publications.
wood-shavings, which provides more effective litter sam- Arnold, M.E., Cook, A. and Davies, R. (2005) A modelling
approach to estimate the sensitivity of pooled faecal sam-
ples (data not shown).
ples for isolation of Salmonella in pigs. J R Soc Interface 2,
The statistical analysis of our data showed that combin-
365–372.
ing one pair of boot swabs (BS1) and a dust sample
Arnold, M.E., Ellis-Iversen, J., Cook, A.J., Davies, R.H.,
detected more Salmonella-positive houses than any of the
McLaren, I.M., Kay, A.C. and Pritchard, G.C. (2008)
individual methods, but the difference compared to BS5 Investigation into the effectiveness of pooled fecal samples
was not statistically significant. However, the combination for detection of verocytotoxin-producing Escherichia coli
of one pair of boot swabs and a dust sample proved the O157 in cattle. J Vet Diagn Invest 20, 21–27.
easiest and fastest way to sample a house effectively, and Baggesen, D.L., Wegener, H.C. and Christensen, J.P. (1996)
it was also cost-efficient, as only two cultures were Typing of Salmonella enterica serovar Saintpaul: an out-
needed. The two samples can also be combined as shown break investigation. APMIS 104, 411–418.
in the combination sampling study. Similar studies, which Bean, N.H. and Griffin, P.M. (1990) Foodborne disease
indicate the possibility of combining sample types, have outbreaks in the United States, 1973–1987: pathogens,
also been carried out in broiler farms (Food Standards vehicles, and trends. J Food Prot 53, 804–817.
Agency 2006). The high rate of isolation of Salmonella Buhr, R.J., Richardson, L.J., Cason, J.A., Cox, N.A. and
from single pairs of boot swabs combined with dust in Fairchild, B.D. (2007) Comparison of four sampling
flocks where dust samples were negative supports the methods for the detection of Salmonella in broiler litter.
previous findings of equivalence of one pair and two pairs Poult Sci 86, 21–25.
of boot swabs. Byrd, J.A., Corrier, D.E., DeLoach, J.R. and Nisbet, D.J. (1997)
Comparison of drag-swab environmental protocols for the
isolation of Salmonella in poultry houses. Avian Dis 41,
Acknowledgements 709–713.
The authors would like to thank the collaborating turkey Caldwell, D.J., Hargis, B.M., Corrier, D.E. and DeLoach, J.R.
(1998) Frequency of isolation of Salmonella from protec-
companies and farmers for their help in carrying out this
tive foot covers worn in broiler houses as compared to
study; the Salmonella lab team at VLA Weybridge for
drag-swab sampling. Avian Dis 42, 381–384.
excellent technical assistance and help with farm visits;
Cannon, R.M. and Roe, R.T. (1982) Livestock Disease Surveil-
staff from VLA regional labs for help with farm visits;
lance: A Field Manual for Veterinarians. Canberra: Austra-
Felicity Clifton-Hadley (VLA Weybridge) for valuable
lian Bureau of Animal Health.
discussions and critical reading of the manuscript. This Carrique-Mas, J.J., Breslin, M., Sayers, A.R., McLaren, I.,
work was funded by the Department for Environment, Arnold, M. and Davies, R.H. (2008) Comparison of envi-
Food and Rural Affairs (Defra) under project code ronmental sampling methods for detecting Salmonella in
OZ0328.

ª 2009 The Authors


Journal compilation ª 2009 The Society for Applied Microbiology, Journal of Applied Microbiology 107 (2009) 635–645 643
Salmonella detection in turkey flocks D. Mueller-Doblies et al.

commercial laying flocks in the UK. Lett Appl Microbiol. Gradel, K.O., Andersen, J. and Madsen, M. (2002)
47, 514–519. Comparisons of sampling procedures and time of
Cox, N.A., Stern, N.J., Craven, S.E., Berrang, M.E. and sampling for the detection of Salmonella in Danish
Musgrove, M.T. (2000) Prevalence of Campylobacter and infected chicken flocks raised in floor systems. Acta Vet
Salmonella in the cecal droppings of turkeys during Scand 43, 21–30.
production. J Appl Poultry Res 9, 542–545. Hayes, J.R., Carr, L.E., Mallinson, E.T., Douglass, L.W. and
Davies, R.H. and Wray, C. (1994) Evaluation of a rapid cul- Joseph, S.W. (2000) Characterization of the contribution
tural method for identification of salmonellas in naturally of water activity and moisture content to the population
contaminated veterinary samples. J Appl Bacteriol 77, 237– distribution of Salmonella spp. in commercial poultry
241. houses. Poult Sci 79, 1557–1561.
Davies, R.H. and Wray, C. (1996a) Determination of an Haysom, I.W. and Sharp, K. (2003) The survival and recovery
effective sampling regime to detect Salmonella enteritidis of bacteria in vacuum cleaner dust. J R Soc Health 123,
in the environment of poultry units. Vet Microbiol 50, 39–45.
117–127. Humphrey, T.J. (1990) Public health implications of the infec-
Davies, R.H. and Wray, C. (1996b) Persistence of Salmonella tion of egg-laying hens with Salmonella enteritidis phage
enteritidis in poultry units and poultry food. Br Poult Sci type 4. World’s Poultry Sci. J. 46, 5–13.
37, 589–96. de Jong, B. and Ekdahl, K. (2006) Human salmonellosis
Davies, R.H., Breslin, M., Bedford, S. and Wray, C. (1998) in travellers is highly correlated to the prevalence of
Observations on Salmonella contamination of turkey Salmonella in laying hen flocks. Euro Surveill 11,
farms. Proceedings of the first International Symposium E060706.1.
on Turkey Diseases. Hafez, H.M. ed. Germany: Free Kingston, D.J. (1981) A comparison of culturing drag swabs
University Berlin. and litter for identification of infections with Salmonella
Davison, S., Benson, C.E. and Eckroade, R.J. (1995) Compari- spp. in commercial chicken flocks. Avian Dis 25, 513–
son of environmental monitoring protocols for the 516.
detection of Salmonella in poultry houses. Avian Dis 39, Kotton, C.N., Lankowski, A.J. and Hohmann, E.L. (2006)
475–479. Comparison of rectal swabs with fecal cultures for detec-
EFSA (2008a) The Community Summary Report on Trends tion of Salmonella typhimurium in adult volunteers. Diagn
and Sources of Zoonoses, Zoonotic Agents, Antimicrobial Microbiol Infect Dis 56, 123–126.
Resistance and Foodborne Outbreaks in the European McCrea, B.A., Norton, R.A., Macklin, K.S., Hess, J.B.
Union in 2006. http://www.efsa.europa.eu/EFSA/efsa_ and Bilgili, S.F. (2005) Recovery and genetic
locale-1178620753812_1178671312912.htm (accessed 26 similarity of Salmonella from broiler house drag
June 2008). swabs versus surgical shoe covers. J Appl Poult Res 14,
EFSA (2008b) Report of the Task Force on Zoonoses Data 694–699.
Collection on the Analysis of the baseline survey on the Nayak, R., Kenney, P.B., Keswani, J. and Ritz, C. (2003)
prevalence of Salmonella in turkey flocks, part A. The Isolation and characterisation of Salmonella in a turkey
EFSA Journal 134, 1–91. production facility. Br Poult Sci 44, 192–202.
EFSA (2008c) Trends and Sources of Zoonoses and Zoonotic Opara, O.O., Mallinson, E.T., Tate, C.R., Carr, L.E., Miller,
Agents in humans, foodstuffs, animals and feedingstuffs in R.G., Stewart, L., Kelleher, C., Johnston, R.W. et al. (1992)
2006. http://www.defra.gov.uk/animalh/diseases/zoonoses/ The effect of exposure, storage times, and types of holding
pdf/Version%20sent%2020%20June%20Report_2006_ media on the drag-swab monitoring technique for Salmo-
with_human.pdf (accessed 26 June 2008). nella. Avian Dis 36, 63–68.
Entis, P. (2006) Salmonella and food safety: the influence of Opara, O.O., Carr, L.E., Tate, C.R., Miller, R.G., Mallinson,
methodology on outcome. J AOAC Int 89, 503. E.T., Stewart, L.E. and Joseph, S.W. (1994) Evaluation of
Fletcher, D.L. (2006) Influence of sampling methodology on possible alternatives to double-strength skim milk used to
reported incidence of Salmonella in poultry. J AOAC Int saturate drag swabs for Salmonella detection. Avian Dis 38,
89, 512–516. 293–296.
Food Standards Agency (2006) To make recommendations on Popoff, M.Y. (2001) Antigenic Formulas of the Salmonella
the best practical procedures to sample and test poultry Serovars. Paris, France: WHO Collaborating Centre
flocks for salmonella. Final report on project B15003. for Reference and Research on Salmonella, Institut
http://www.foodstandards.gov.uk/science/research/research- Pasteur.
info/foodborneillness/eggsresearch/b15programme/b15pro- Poppe, C., Irwin, R.J., Messier, S., Finley, G.G. and Oggel, J.
jects/b15003/ (accessed 26 June 2008). (1991) The prevalence of Salmonella enteritidis and
Funk, J.A., Davies, P.R. and Nichols, M.A. (2000) The effect of otherSalmonella sp. among Canadian registered
fecal sample weight on detection of Salmonella enterica in commercial chicken broiler flocks. Epidemiol Infect 107,
swine feces. J Vet Diagn Invest 12, 412–418. 201–211.

ª 2009 The Authors


644 Journal compilation ª 2009 The Society for Applied Microbiology, Journal of Applied Microbiology 107 (2009) 635–645
D. Mueller-Doblies et al. Salmonella detection in turkey flocks

Rolfe, D.L., Riemann, H.P., Farver, T.B. and Himathongkham, commercial North Carolina turkey farms. J Food Prot
S. (2000) Drag swab efficiency factors when sampling 70, 1328–1333.
chicken manure. Avian Dis 44, 668–675. Skov, M.N., Carstensen, B., Tornoe, N. and Madsen, M.
Santos, F.B., Dsouza, D.H., Jaykus, L., Ferket, P.R. and (1999) Evaluation of sampling methods for the detection
Sheldon, B.W. (2007) Genotypes, serotypes, and of Salmonella in broiler flocks. J Appl Microbiol 86,
antibiotic resistance profiles of Salmonella isolated from 695–700.

ª 2009 The Authors


Journal compilation ª 2009 The Society for Applied Microbiology, Journal of Applied Microbiology 107 (2009) 635–645 645

You might also like