- Email: [email protected]
Risk factors associated with Salmonella enterica serovar typhimurium infection in Danish broiler flocks
ENVIRONMENT AND HEALTH Risk Factors Associated with Salmonella enterica Serovar typhimurium Infection in Danish Broiler Flocks M. N. SKOV,*,1 Ø. ANGEN,*,2 M. CHRIE´L,† J. E. OLSEN,* and M. BISGAARD* *Department of Veterinary Microbiology, and †Department of Animal Science and Animal Health, Division of Ethology and Health, The Royal Veterinary and Agricultural University, Bu¨lowsvej 13, DK-1870 Frederiksberg C, Denmark hatcheries, and with five houses on the farm. An interaction between season and the previously mentioned hatcheries, and a random effect at farm level was also found to be statistically significant. Twelve variables were not found to be associated with S. typhimurium infection: medication, growth promoters, breed of the laying flock, animal density, size of the flock, area of the house, age of the house, geographical location of the farm, observation of beetles, number of days between disinfection and replacement, visual appearance of the bedding, and age of the chickens when they were tested for Salmonella. Three variables (feed mill, slaughterhouse, and Salmonella status of the preceding flock) were not evaluated in the multivariate analysis due to collinearity with other included variables.
(Key words: Salmonella, broiler, epidemiology, risk factors, Salmonella typhimurium) 1999 Poultry Science 78:848–854
Salmonella berta (Olsen et al., 1992), Salmonella enteritidis (Brown et al., 1994), Salmonella tennessee (Christensen et al., 1997), and Salmonella virchow (D. J. Brown et al., The Royal Veterinary and Agricultural University, Denmark, unpublished results) are examples of serotypes that have been successfully eliminated, whereas others, predominantly S. typhimurium, Salmonella infantis, and Salmonella 4:12:b: still have a relatively high prevalence. Prevention of Salmonella contamination of broilers requires detailed knowledge of the most important risk factors associated with its presence in the production system. A previous investigation has identified certain hatcheries and feed mills, size of farms in terms of number of houses, a positive Salmonella status of the preceding flock, and rearing of flocks in the autumn as important risk factors for S. enterica in broilers in Denmark (Angen et al., 1996). Data available at that time, however, did not allow serotype specific risk analysis. By the end of 1994, the detection procedure used to monitor the Salmonella status of the Danish broiler flocks
INTRODUCTION In Denmark, the incidence rate of human Salmonella enterica serovar typhimurium (S. typhimurium) infections has been increasing during the last decade. Next to S. enteritidis, S. typhimurium was the most frequently isolated serotype of S. enterica in 1995 (Anonymous, 1995a). Pork and poultry products are considered to be major sources of human infections with S. typhimurium (Baggesen and Wegener, 1994; Wegener et al., 1994). In order to reduce the level of S. enterica infection in broilers, a national eradication program has been implemented (Bisgaard, 1992). The program has been continuously upgraded, and although relapses have occurred, the program has resulted in decreasing prevalence of S. enterica in broilers (Anonymous, 1995a).
Received for publication August 13, 1998. Accepted for publication February 2, 1999. 1Present address: Danish Veterinary Laboratory, Hangøvej 2, DK˚ rhus N, Denmark. To whom correspondence should be 8200 A addressed: [email protected] 2Present address: Danish Veterinary Laboratory, Bu ¨ lowsvej 27, DK1790 København V, Denmark.
Abbreviation Key: AM = antemortem.
848
Downloaded from http://ps.oxfordjournals.org/ at Kainan University on April 30, 2015
ABSTRACT A retrospective longitudinal study was conducted to identify risk factors associated with Salmonella enterica serovar typhimurium (S. typhimurium) infection in Danish broiler flocks. The data included all broiler flocks slaughtered in 1995, and the epidemiological unit was the individual broiler flock. The S. typhimurium status was determined by microbiological examination of 60 fresh fecal samples. This procedure should detect an infected flock with a probability above 95%, if the prevalence is above 5%, and given that the sensitivity of the test is 100%. Nineteen variables were selected for analysis. Five factors and an interaction term were found significant by multivariate logistic regression analysis. An increased risk for S. typhimurium infection was associated with two parent flocks, one confirmed infected and one suspected of being infected with S. typhimurium, with two of the
849
RISK FACTORS RELATED TO SALMONELLA TYPHIMURIUM INFECTIONS
was changed from analysis of cecal tonsils from 16 broilers per flock to analysis of 60 fresh fecal samples per flock. Results obtained by the improved sampling procedure are also recorded in the antemortem (AM) data base, described by Angen et al. (1996). Using the AM data base, the aim of the present investigation was to identify risk factors specifically associated with S. typhimurium infections in broiler flocks in Denmark.
MATERIALS AND METHODS
The AM Data Base
Salmonella Status of the Parent Flocks During the production period, the parent flocks were screened for Salmonella. Meconium samples from 250 chickens or the intestines from 50 weak or dead in shell chickens from each parent flock were examined every 2nd wk. If one of these samples tested Salmonella positive, the parent flock was suspected of being infected with Salmonella. Sixty hens per subunit (section or house) of the parent flock were subsequently killed and submitted for bacteriological confirmation. If S. enterica was isolated from one or more killed birds, the chickens in the Salmonella-positive sections or houses were confirmed as infected and subsequently killed.
Selection of Variables Salmonella Status of Broiler Flocks The Salmonella status of the flocks was based upon examination of 12 pools each consisting of five fresh fecal samples taken by the farmer at the flock age of approximately 3 wk (2.7 wk in average, SE = 6 d). This sample size will detect an infected flock with a probability > 95%, if the prevalence is above 5%, and given that the
The selection of variables was based on previous studies (Angen et al., 1996). Nineteen variables were initially selected for use in a multivariate analysis (Table 1). In addition to the variables used in the study of Angen et al. (1996), the following new variables were included: breed of the parent flock, medication (yes/no), growth promoters used, identification number of the parent flocks
TABLE 1. Variables selected for initial analysis, including distribution of the continuous variables for 3,839 Danish broiler flocks Variables
Mean (range)
SD
Hatchery Feed mill Season (date of the AM2 inspection) Slaughterhouse of the preceding flock ST3 status of the preceding flock AM district (geographical location of the farm) ID4 number of the parent flock from which the broiler flock was derived (No. 1-127) Breed of the parent flock Medication (yes/no) Name of growth promoters used Beetles observed during production or cleaning of the house (yes/no) Appearance of the bedding (hard/wet/dry or dry and hard) Number of houses on the farm Animal density, broilers/m2 Flock size, number of chickens placed Area of the house, m2 Age of the house, yr of construction Days between disinfection and replacement The age of the broilers when Salmonella tested, d
. . .1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 (1 to 9) 27.4 (4.6 to 49.5) 28,753 (1,1100 to 72,213) 1,202 (151 to 2,800) 1979 (1935 to 1995) 11.7 (1 to 95) 18.7 (1 to 52)
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 3.3 13,708 509 10.4 6.7 6.1
1Not
applicable. = antemortem. 3ST = Salmonella typhimurium. 4ID = identification number of the parent flock. 5Rounded figures. 2AM
Downloaded from http://ps.oxfordjournals.org/ at Kainan University on April 30, 2015
The AM data base consists of flock data collected by veterinarians during their AM inspections, data from the slaughterhouses, and results from Salmonella examination of 60 fecal samples per flock collected at the age of approximately 3 weeks as previously described (Angen et al., 1996). In addition, 60% of the farms (representing 60% of the flocks) participated in the Flock Economy Control Program, in which different parameters related to the production are recorded. These data were also used in the present study.
sensitivity of the test is 100% (Martin et al., 1987). The flock was regarded as S. typhimurium-positive if one or more of a maximum of 60 Salmonella suspect colonies investigated per flock were identified as S. typhimurium. Detection and serotyping of S. enterica were performed according to the procedures recommended by the Nordic Committee on Food Analysis (1991).
850
SKOV ET AL.
from which the respective broiler flocks originated, and a random effect at the farm level.
Descriptive statistics of the continuous variables not included in the final model are shown in Tables 1 and 2. The prevalence of S. typhimurium in relation to the
Definition of the Data Sets The data base, provided by the Danish Poultry Council, initially contained information on 4,221 flocks, all of which had been visited by the AM veterinarians in 1995. The uni- and bivariate analyses were based on 3,839 flock observations for which the Salmonella status of the individual flocks was known. The multivariate analysis was based on 2,776 flock observations, excluding flocks with missing values.
TABLE 2. Descriptive statistics for variables which were significant (P < 0.20) in the bivariate analysis, but not included in the final logistic regression model; 3,839 Danish broiler flocks, 1995
Variable1
Level
Percentage of flocks
Density
<5 5 to 10 10 to 15 15 to 20 20 to 25 25 to 30 ≥30 <10,000 10 to 20,000 20 to 30,000 30 to 40,000 ≥40,000 <500 500 to 1,000 1,000 to 1,500 ≥1,500
0.0 2.4 0.1 2.1 41.3 53.7 0.6 9.1 21.1 24.7 17.4 27.8 25.0 20.5 23.3 31.2
0.0 6.6 100.0 4.0 10.4 14.3 5.3 9.7 11.9 8.8 16.1 13.8 11.6 10.2 11.9 14.2
<1960 1960 to 70 1970 to 80 1980 to 90 ≥1990 R208 COBB Shaver RossPM Many lines/mix of line – +
35.4 13.1 17.5 19.8 14.1 71.4 14.5 1.2 2.0
10.5 14.3 12.0 13.3 12.9 12.5 8.6 4.6 5.3
11.0 92.5 7.5
17.1 12.2 8.1
Zinc bacitracin 25.8 Virginiamycin 1.0 Flavophospholipol 3.7 24.8 Avilamycin 44.7
9.2 8.3 6.2 14.2 13.7
(%)
Statistical Analysis
Flock size
Area
Construction year
Breed
Medication Growth promoters Avoparcin
RESULTS
Data Sets Statistically significant differences between the flock sizes (t test, P > 0.93) and the Salmonella frequencies (x2 test, P = 0.71) in the original and analyzed data sets were not observed.
Uni- and Bivariate Analysis In 1995, 12.2% of the broiler flocks sampled positive for S. typhimurium. In the bivariate analysis the following variables were shown not to be significantly associated with the presence of this serotype, and were therefore excluded from the multivariate analysis: observation of beetles, days between disinfection and replacement of chickens, and visual appearance of the bedding.
Salmonella test, wk
1 1 to 2 to 3 to ≥4 Parent flock n = AM district n = Slaughterhouse n = Feed mill n = – Lagstyph3 +
2 3 4 125 7 9 10
0.2 21.8 39.5 30.0 8.6 <1 to <1 to <1 to 5 to 87.0 13.1
5.8 23 23 20
12.5 13.8 13.5 9.0 13.1 0 to 0 to 4 to 8 to 8.7 35.7
Downloaded from http://ps.oxfordjournals.org/ at Kainan University on April 30, 2015
A retrospective longitudinal study design was used (Kleinbaum et al., 1982). The individual broiler flock was the unit of concern. Based on uni- and bivariate analyses, variables were selected for the multivariate analysis, with only significant (x2 test, P < 0.20) and independent variables being included. A multivariate logistic regression analysis was performed using the CATMOD procedure in SAS (SAS Institute, 1990). The best model fit was found by a combined forward and backward selection process of adding or subtracting a variable (P < 0.05 for the model being better with the addition or subtraction) or an interaction (P < 0.01) from the model (Hosmer and Lemeshow, 1989). Estimation of the parameters for the main variables and the interaction term in the final model was done with and without a term related to random effect on the house and on the farm level using MLn software (Rasbash and Woodhouse, 1995) for analyzing multilevel statistical models (Goldstein, 1995). This analysis was performed due to lack of independence between flocks produced within the same house and between houses located on the same farm. The random effects (Uj) of the houses or farms were assumed to be normally distributed (with zero mean and variance s2).
S. typhimurium2
60 15 14 21
1Density = number of chickens placed per square meter; Flock size = number of chickens placed in the house; Area = area of house in square meters; Construction year = construction year of the house; Breed = breed of the parent flock; Salmonella test = the age of the chickens when Salmonella tested; Slaughterhouse = slaughterhouse of the preceding flock. 2Salmonella typhimurium prevalence in percentage for the actual covariate level. 3Lagstyph = S. typhimurium status of the preceding flock.
851
RISK FACTORS RELATED TO SALMONELLA TYPHIMURIUM INFECTIONS
different levels of the variables and the corresponding number of flock observations are shown in Tables 2 and 3. Significant association was found among the hatcheries, the feed mills, and the slaughterhouses (data not shown). As collinearity between these variables was suspected to bias the further analysis, only the hatcheries that had the highest odds ratios of these three variables were included in the multivariate analysis. The variable “S. typhimurium status of the preceding flock” was not evaluated in the multivariate analysis due to the association observed between the parent flocks and broiler flocks in the same house.
TABLE 4. Odds ratios and confidence intervals for the interaction term between hatchery and season, included in the multivariate analysis of Salmonella typhimurium infections in Danish broiler flocks, 1995 Hatchery Season Summer Autumn Winter Spring
Multivariate Analysis
Flock percentage2
Adjusted S. odds typhimurium3 ratio
19.5 24.9 19.7 12.1 9.6 6.7 7.5
9.9 9.9 10.2 18.7 17.7 15.1 10.5
14 0.8 1.1 1.5 2.5* 0.8 1.0
. . 0.5 0.7 0.9 1.4 0.4 0.5
96.5 3.5
11.0 43.0
14 2.8*
. . . 1.7 to 4.4
98.8 1.2
12.0 29.6
14 3.2*
. . . 1.4 to 16.4
24.2 25.7 24.9 25.3
15.2 10.5 13.5 9.6
. . . .
8 to 38
4 to 22
C (28.1%1)
D (38.2%1)
14 . . . 1.1 (0.3 to 4.2) 1.9 (0.6 to 6.3) 1.6 (0.5 to 5.5)
1.5 (0.2 to 13.1)2 2.9 (0.5 to 17.6) 2.2 (0.4 to 13.4) . . .3
12.8* (4.7 to 4.1* (1.5 to 13.6* (5.0 to 6.2* (2.3 to
2.5 (0.8 to 4.3* (1.6 to 5.0* (1.8 to 4.2* (1.6 to
34.8) 12.2) 36.9) 16.9)
–7.4) 11.7) 13.5) 11.5)
. . . .
.5 . . .
. . .5
95% Confidence interval
. . . .
. . . .
. to to to to to to
1.3 1.7 2.5 4.4 1.7 1.9
.5 . . .
. . .5
1House = number of chicken houses on the farm, Parent flock No. 10 =
parent flock confirmed to be infected with S. typhimurium, Parent flock No. 83 = parent flock suspected of being infected with S. typhimurium. 2Percentage of flocks in the study (3839). 3Salmonella typhimurium prevalence in percentage for the actual covariate level (3,839 flocks). 4Reference. 5See Table 4 for odds ratios and confidence intervals for hatchery and season. 6Only minimum and maximum values shown. *P < 0.05.
of flocks in the study (3,839). interval. 3The hatchery had no deliveries of broiler flocks with S. typhimurium in that season. 4Reference. *P < 0.05. 2Confidence
an interaction term between season and hatchery (Table 4). An increased risk for S. typhimurium infection in the broiler flocks was associated with two parent flocks [one bacteriologically confirmed (No. 10) and one suspected of being infected with S. typhimurium (No. 83)], and having five houses on the farm (Table 3). In addition, an increased risk for S. typhimurium infection in the broiler flocks was associated with two of the hatcheries with season (observed as an interaction term between these two variables) (Table 4). The final model was further analyzed by including a random effect at house and farm level. Added separately, the random effect at house (variation between houses su0 = 0.70) and farm level (variation between farms su0 = 0.62) were statistically significant, but when both random effects were included simultaneously only the random effect at farm level (variation between farms su0 = 0.56) was shown to be statistically significant in the model.
DISCUSSION The multivariate epidemiological approach to the study of risk factors for salmonellosis in poultry has so far not received much attention. Henken et al. (1992) showed that feed mills and poor hygiene barriers, including the lack of disinfection tubs, represented risk factors. Hatcheries and feed mills, size of farms in terms of number of houses, a positive Salmonella status of the preceding flock, and rearing of flocks in the autumn have been identified as important risk factors for salmonellosis by Angen et al. (1996). With the exception of two variables (parent flocks and the random effect at the farm level) the five risk factors identified in the present analysis for S. typhimurium were in agreement with those identified by Angen et al. (1996).
Downloaded from http://ps.oxfordjournals.org/ at Kainan University on April 30, 2015
TABLE 3. Prevalence and odds ratios for variables included (P < 0.05) in the final multivariate logistic regression model (including a random effect term at farm level) of Salmonella typhimurium infections in 2,776 Danish broiler flocks, 1995
House 1 2 3 4 5 6 ≥7 Parent flock No. 10 (confirmed) – + Parent flock No. 83 (suspected) – + Season Dec to Feb Mar to May Jun to Aug Sep to Nov Hatchery6 n = 4
B (7.5%1)
1Percentage
In the multivariate logistic regression model, five variables were shown to be associated with S. typhimurium infections in Danish broiler flocks (Table 3) in addition to
Variable1
A (21.3%1)
852
SKOV ET AL.
reflect that houses at the farm are epidemiologically connected (e.g., by the environment surrounding the houses at the farm, level of biosecurity, the feed mill delivering to the farm, or the slaughterhouse collecting the chickens at the farm). It might also indicate differences in management, including differences between advisers connected to the farms. Others have found feed mills (Henken et al., 1992, Angen et al., 1996) and slaughterhouses (Christensen et al., 1994) to represent risk factors, although these investigations were not specific for S. typhimurium. All Danish broiler feed is heat-treated, and all final broiler feed investigated in 1995 sampled negative for S. enterica (Anonymous, 1995b). However, the sensitivity of the Salmonella test used in these analyses has not been evaluated and, consequently, the reliability of the test results is unknown. Further investigations are also needed to evaluate the significance of the slaughterhouses as a source of contamination for the chicken houses and the environment surrounding these houses while collecting the chickens for slaughter. The number of houses on the farm was shown to be part of the multivariate logistic regression model, but only farms having five houses were shown to be significantly associated with S. typhimurium infection in broiler flocks (odds ratio = 2.5). It has not been possible to explain this finding biologically, and considering the relatively few farms having more than five houses (Table 3), it is difficult to investigate if special problems can be related to having more than five houses on a farm. Previous investigations (Angen et al. 1996) have shown that broiler flocks at farms with more than three houses have an increased risk of S. enterica infection. Genetic differences in resistance to mortality caused by S. enterica have been reported by Bumstead and Barrow (1993), but little is known about the resistance to colonization or infection. In the present investigation, the breeding lines used for parent stock were shown not to be significantly associated with S. typhimurium infection. This lack of association might be explained by the fact that Ross 208 was the breeding line of more than 71% of the flocks (Table 2) and, thus, the genetic variability of the flocks was low. Medication has previously been shown to influence infection and the shedding of S. typhimurium (Smith and Tucker, 1975). In the present investigation, these observations could not be confirmed. The significance of avoparcin and other antibiotic growth promoters for colonization of the alimentary tract by Salmonella, and the duration of shedding of these organisms by infected chickens have been described in a number of papers (Smith and Tucker, 1975, 1978, 1980; Linton et al, 1985; Hinton, 1988; Barrow, 1989) but the results have been inconsistent. The use of antibiotic growth promoters is presently a subject of strong debate due to the problems related to cross resistance to antibiotics used in humans. None of the antibiotic growth promoters used was shown to be
Downloaded from http://ps.oxfordjournals.org/ at Kainan University on April 30, 2015
In the present study, two parent flocks (No. 10 and 83) were shown to represent a risk factor for contamination of broiler flocks with S. typhimurium. This finding is not surprising as vertical transmission of S. enterica has been demonstrated on several occasions for certain serovars (Lahellec and Colin, 1985; Mcllroy et al., 1989), underlining the importance of a top-down strategy for the eradiation of S. enterica. One of these “risk parent flocks” (No. 10) was the only parent flock in the present investigation confirmed as being infected with S. typhimurium according to bacteriological investigations. It consisted of six subunits, all located on the same farm, but in separate sections. Epidemiologically, each section was regarded as a separate unit. Only two out of six sections tested positive for S. typhimurium. Parents from these sections were killed late in November, 1994, whereas the other subunits remained in production until June, 1995. These four subunits were shown to represent an important risk factor, indicating that even though parent flocks are kept in separate sections they should all be considered as one epidemiological unit. The other parent flock shown to represent a risk factor was only found suspected of infection but found to be bacteriologically negative on subsequent examination. This observation indicates that the sensitivity of the confirmation procedure used for S. typhimurium might be insufficient, especially in cases with low Salmonella prevalence in both the flock and the individual hen. The present results are in agreement with previous investigations (Chrie´l, 1997) demonstrating an increased risk for S. typhimurium infections associated with receiving chickens from infected parent flocks, independent of the proportion of the flock these chickens represent. In addition, Chrie´l (1997) also found an increased risk was associated with receiving chickens from more than one parent flock. With only one exception (Hatchery D in the summer), receiving chickens from Hatchery C or D was shown to be connected with an increased risk of S. typhimurium infection (Table 4). These hatcheries received eggs from the “risk parent flocks” (Flock 10 was owned by Hatchery C, whereas Flock 83 was owned by Hatchery D) discussed above, but, as seen from Table 4, the hatcheries themselves also represented a risk factor— most likely caused by cross contamination at the hatcheries or because of other Salmonella-positive parent flocks not identified. This finding is in agreement with previous investigations showing that the hatcheries on their own might contribute to the spread of S. enterica (Christensen et al., 1997). The random effect at the house level was significant when included as the only random effect in the model, which is in agreement with the observations reported by Angen et al. (1996). However, inclusion of an additional random effect at farm level rendered only the random effect at the farm level statistically significant. This finding indicates that it is possible to clean and disinfect a house contaminated with S. typhimurium and might
RISK FACTORS RELATED TO SALMONELLA TYPHIMURIUM INFECTIONS
ACKNOWLEDGMENTS This research was supported by The Danish Agricultural and Veterinary Research Council (Grant Number 20-3510-2) and The Ministry of Agriculture and Fisheries (SUN 94-KVL-11).
REFERENCES Angen, Ø., M. N. Skov, M. Chrie´l, J. F. Agger, and M. Bisgaard, 1996. A retrospective study on salmonella infection in Danish broiler flocks. Prev. Vet. Med. 26: 223–237.
Anonymous, 1995a. Annual report on zoonoses in Denmark 1995. Danish Zoonosis Centre. Danish Veterinary Laboratory, Copenhagen, Denmark. Anonymous, 1995b. Annual report on animal feed in Denmark in 1995. Danish Plant Directorate, Lyngby, Denmark. Baggesen, D. L., and H. C. Wegener, 1994. Phage types of Salmonella enterica ssp. enterica serovar typhimurium isolated from production animals and humans in Denmark. Acta Vet. Scand. 35:349–354. Barrow, P. A., 1989. Further observations on the effect of feeding diets containing avoparcin on the excretion of salmonellas by experimentally infected chickens. Epidemiol. Infect. 102:239–252. Bisgaard, M., K. Haaning, and G. Velling, 1982. Poultry-borne salmonellosis. Prophylactic measures in the broiler production including bacteriological investigations before and after slaughter. Pages 227–231 in: XIV Nordic Veterinary Congress. Copenhagen, Denmark. Bisgaard, M., 1992. A voluntary salmonella control programme for the broiler industry, implemented by the Danish Poultry Council. Int. J. Food Microbiol. 15:219–224. Brown, J. D., D. L. Baggesen, H. B. Hansen, H. C. Hansen, and M. Bisgaard, 1994. The characterization of Danish isolates of Salmonella enterica serovar enteritidis by phage typing and plasmid profiling: 1980–1990. APMIS 120:208–214. Bumstead, N., and P. A. Barrow, 1993. Resistance to Salmonella gallinarum, S. pullorum, and S. enteritidis in inbred lines of chickens. Avian Dis. 37:189–193. Chrie´l, M., 1997. Hierarchical, crossclassified modelling of Salmonella typhimurium in Danish Broilers. Pages 163–171 in: Proceeding of the Dutch/Danish Symposium on Animal Health and Management Economics. Copenhagen, Denmark. Christensen, J. P., D. J. Brown, M. Madsen, J. E. Olsen, and M. Bisgaard, 1997. Hatchery-borne Salmonella enterica serovar Tennessee infections in broilers. Avian Pathol. 26:155–168. Christensen, J. P., M. N. Skov, K. H. Hinz, and M. Bisgaard, 1994. Salmonella enterica serovar Gallinarum biovar gallinarum in layers: epidemiological investigations of a recent outbreak in Denmark. Avian Pathol. 23:489–501. Goldstein, H., 1995. Multilevel Statistical Models, second edition. Institute of Education, University of London, London, U.K. Henken, A. M., K. Frankena, J. O. Goelema, E.A.M. Graat, and J.P.T.M. Noordhuizen, 1992. Multivariate epidemiological approach to salmonellosis in broiler breeder flocks. Poultry. Sci. 71:838–843. Hinton, M., 1988. Salmonella colonization in young chickens given feed supplemented with the growth promoting antibiotic avilamycin. J. Vet. Pharmacol. Therap. 11: 269–275. Hosmer, D. W., and S., Lemeshow, 1989. Applied Logistic Regression. John Wiley and Sons, New York. NY. Kleinbaum, D. G., L. L. Kupper, and H. Morgenstern, 1982. Epidemiological Research. Van Nostrand Reinhold Co., New York, NY. Lahellec, C., and P. Colin, 1985. Relationship between serotypes of salmonella from hatcheries and rearing farms and those from processed poultry carcasses. Br. Poult. Sci. 26: 179–186. Linton, A. H., Z.A.M. Al-Chalaby, and M. H. Hinton, 1985. Natural subclinical salmonella infection in chickens: A potential model for testing the effect of various procedures on salmonella shedding. Vet. Rec. 6:361–364.
Downloaded from http://ps.oxfordjournals.org/ at Kainan University on April 30, 2015
associated with increased detection of S. typhimurium. However, as all broiler flocks in Denmark at the time of the investigation were given antibiotic growth promoters the significance in relation to colonization and excretion of Salmonella might be difficult to evaluate (since February 1998 there has been a voluntary cessation of the use of antibiotic growth promotors in Denmark). Excretion of S. enterica has been shown to decrease with age (Bisgaard et al., 1982). However, the age of the chickens was not a significant variable in the present investigation (Table 2). The risk analysis for S. enterica in Danish broiler flocks during the years 1992 to 1993 (Angen et al., 1996) was based on bacteriological examination of cecal tonsils from 16 broilers from each flock (20% prevalence level). The present study was based on bacteriological examination of 60 fresh fecal samples from each flock (5% prevalence level). Comparison of the two different methods in 320 broiler flocks revealed a larger serotype diversity in the flocks related to the increased number of samples being examined. However, the investigation also showed that certain serotypes were more frequently detected with one of the methods, i.e., more flocks were found infected with S. typhimurium on examination of fecal samples, whereas more flocks were found infected with S. enteritidis when investigating the cecal tonsils (M. Chrie´l, The Royal Veterinary and Agricultural University, Denmark, unpublished results). These differences might be related to differences in the pathogenesis (e.g., invasiveness) of the different serotypes. In conclusion, the present investigation has shown that multivariate logistic regression analysis is a valuable tool in outlining risk factors associated with S. typhimurium infections in broilers. The results strongly indicated that parent flocks infected with S. typhimurium, including confirmed flocks and flocks suspected of being infected, as well as the hatcheries receiving eggs from these flocks, represented significant risk factors. This conclusion underlines the significance of a topdown strategy for the eradication of S. typhimurium. In addition, the significance of the random effect related to the farm indicated that all houses at the farm seemed to be exposed to similar external risk factors.
853
854
SKOV ET AL.
Martin, S. W., A. H. Meek, and P. Willeberg, 1987. Veterinary Epidemiology. Principles and Methods, Iowa State University Press, Ames, IA. Mcllroy, S. G., R. M. McCracken, S. D. Neill, and J. J. O’Brien, 1989. Control, prevention and eradication of Salmonella enteritidis infection in broiler and broiler breeder flocks. Vet. Rec. 125:545–548. Nordic Committee on Food Analysis, No. 71. 1991. c/o VTT Bio-og Livsmedelsteknik PB 1500, Fin-02044 VTT, Finland. Olsen, J. E., M. Sørensen, D. J. Brown, K. Gaarslev, and M. Bisgaard, 1992. Plasmid profiles as an epidemiological marker in Salmonella enterica serovar Berta infections. APMIS 100:221–228. Rasbash, J., and G. Woodhouse, 1995. MLn Command reference Version 1.0. Institute of Education, University of London, London, U.K.
SAS Institute, 1990. SAS Stat User’s Guide. Version 6, Fourth Edition. SAS Institute Inc., Cary, NC. Smith, H. W., and J. F. Tucker, 1975. The effect of feeding diets containing permitted antibiotics on the feacal excretion of Salmonella typhimurium by experimentally infected chickens. J. Hyg. Camb. 75:293–301. Smith, H. W., and J. F. Tucker, 1978. The effect of antimicrobial feed additives on the colonization of the alimentary tract of chickens by Salmonella typhimurium. J. Hyg. Camb. 80: 217–231. Smith, H. W., and J. F. Tucker, 1980. The virulence of salmonella strains for chickens: their excretion by infected chickens. J. Hyg. 84:479–488. Wegener, H. C., D. L. Baggesen, and K. Gaarslev, 1994. Salmonellosis in Denmark 1988–1993. APMIS 102:521–525.
Downloaded from http://ps.oxfordjournals.org/ at Kainan University on April 30, 2015
(Key words: Salmonella, broiler, epidemiology, risk factors, Salmonella typhimurium) 1999 Poultry Science 78:848–854
Salmonella berta (Olsen et al., 1992), Salmonella enteritidis (Brown et al., 1994), Salmonella tennessee (Christensen et al., 1997), and Salmonella virchow (D. J. Brown et al., The Royal Veterinary and Agricultural University, Denmark, unpublished results) are examples of serotypes that have been successfully eliminated, whereas others, predominantly S. typhimurium, Salmonella infantis, and Salmonella 4:12:b: still have a relatively high prevalence. Prevention of Salmonella contamination of broilers requires detailed knowledge of the most important risk factors associated with its presence in the production system. A previous investigation has identified certain hatcheries and feed mills, size of farms in terms of number of houses, a positive Salmonella status of the preceding flock, and rearing of flocks in the autumn as important risk factors for S. enterica in broilers in Denmark (Angen et al., 1996). Data available at that time, however, did not allow serotype specific risk analysis. By the end of 1994, the detection procedure used to monitor the Salmonella status of the Danish broiler flocks
INTRODUCTION In Denmark, the incidence rate of human Salmonella enterica serovar typhimurium (S. typhimurium) infections has been increasing during the last decade. Next to S. enteritidis, S. typhimurium was the most frequently isolated serotype of S. enterica in 1995 (Anonymous, 1995a). Pork and poultry products are considered to be major sources of human infections with S. typhimurium (Baggesen and Wegener, 1994; Wegener et al., 1994). In order to reduce the level of S. enterica infection in broilers, a national eradication program has been implemented (Bisgaard, 1992). The program has been continuously upgraded, and although relapses have occurred, the program has resulted in decreasing prevalence of S. enterica in broilers (Anonymous, 1995a).
Received for publication August 13, 1998. Accepted for publication February 2, 1999. 1Present address: Danish Veterinary Laboratory, Hangøvej 2, DK˚ rhus N, Denmark. To whom correspondence should be 8200 A addressed: [email protected] 2Present address: Danish Veterinary Laboratory, Bu ¨ lowsvej 27, DK1790 København V, Denmark.
Abbreviation Key: AM = antemortem.
848
Downloaded from http://ps.oxfordjournals.org/ at Kainan University on April 30, 2015
ABSTRACT A retrospective longitudinal study was conducted to identify risk factors associated with Salmonella enterica serovar typhimurium (S. typhimurium) infection in Danish broiler flocks. The data included all broiler flocks slaughtered in 1995, and the epidemiological unit was the individual broiler flock. The S. typhimurium status was determined by microbiological examination of 60 fresh fecal samples. This procedure should detect an infected flock with a probability above 95%, if the prevalence is above 5%, and given that the sensitivity of the test is 100%. Nineteen variables were selected for analysis. Five factors and an interaction term were found significant by multivariate logistic regression analysis. An increased risk for S. typhimurium infection was associated with two parent flocks, one confirmed infected and one suspected of being infected with S. typhimurium, with two of the
849
RISK FACTORS RELATED TO SALMONELLA TYPHIMURIUM INFECTIONS
was changed from analysis of cecal tonsils from 16 broilers per flock to analysis of 60 fresh fecal samples per flock. Results obtained by the improved sampling procedure are also recorded in the antemortem (AM) data base, described by Angen et al. (1996). Using the AM data base, the aim of the present investigation was to identify risk factors specifically associated with S. typhimurium infections in broiler flocks in Denmark.
MATERIALS AND METHODS
The AM Data Base
Salmonella Status of the Parent Flocks During the production period, the parent flocks were screened for Salmonella. Meconium samples from 250 chickens or the intestines from 50 weak or dead in shell chickens from each parent flock were examined every 2nd wk. If one of these samples tested Salmonella positive, the parent flock was suspected of being infected with Salmonella. Sixty hens per subunit (section or house) of the parent flock were subsequently killed and submitted for bacteriological confirmation. If S. enterica was isolated from one or more killed birds, the chickens in the Salmonella-positive sections or houses were confirmed as infected and subsequently killed.
Selection of Variables Salmonella Status of Broiler Flocks The Salmonella status of the flocks was based upon examination of 12 pools each consisting of five fresh fecal samples taken by the farmer at the flock age of approximately 3 wk (2.7 wk in average, SE = 6 d). This sample size will detect an infected flock with a probability > 95%, if the prevalence is above 5%, and given that the
The selection of variables was based on previous studies (Angen et al., 1996). Nineteen variables were initially selected for use in a multivariate analysis (Table 1). In addition to the variables used in the study of Angen et al. (1996), the following new variables were included: breed of the parent flock, medication (yes/no), growth promoters used, identification number of the parent flocks
TABLE 1. Variables selected for initial analysis, including distribution of the continuous variables for 3,839 Danish broiler flocks Variables
Mean (range)
SD
Hatchery Feed mill Season (date of the AM2 inspection) Slaughterhouse of the preceding flock ST3 status of the preceding flock AM district (geographical location of the farm) ID4 number of the parent flock from which the broiler flock was derived (No. 1-127) Breed of the parent flock Medication (yes/no) Name of growth promoters used Beetles observed during production or cleaning of the house (yes/no) Appearance of the bedding (hard/wet/dry or dry and hard) Number of houses on the farm Animal density, broilers/m2 Flock size, number of chickens placed Area of the house, m2 Age of the house, yr of construction Days between disinfection and replacement The age of the broilers when Salmonella tested, d
. . .1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 (1 to 9) 27.4 (4.6 to 49.5) 28,753 (1,1100 to 72,213) 1,202 (151 to 2,800) 1979 (1935 to 1995) 11.7 (1 to 95) 18.7 (1 to 52)
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 3.3 13,708 509 10.4 6.7 6.1
1Not
applicable. = antemortem. 3ST = Salmonella typhimurium. 4ID = identification number of the parent flock. 5Rounded figures. 2AM
Downloaded from http://ps.oxfordjournals.org/ at Kainan University on April 30, 2015
The AM data base consists of flock data collected by veterinarians during their AM inspections, data from the slaughterhouses, and results from Salmonella examination of 60 fecal samples per flock collected at the age of approximately 3 weeks as previously described (Angen et al., 1996). In addition, 60% of the farms (representing 60% of the flocks) participated in the Flock Economy Control Program, in which different parameters related to the production are recorded. These data were also used in the present study.
sensitivity of the test is 100% (Martin et al., 1987). The flock was regarded as S. typhimurium-positive if one or more of a maximum of 60 Salmonella suspect colonies investigated per flock were identified as S. typhimurium. Detection and serotyping of S. enterica were performed according to the procedures recommended by the Nordic Committee on Food Analysis (1991).
850
SKOV ET AL.
from which the respective broiler flocks originated, and a random effect at the farm level.
Descriptive statistics of the continuous variables not included in the final model are shown in Tables 1 and 2. The prevalence of S. typhimurium in relation to the
Definition of the Data Sets The data base, provided by the Danish Poultry Council, initially contained information on 4,221 flocks, all of which had been visited by the AM veterinarians in 1995. The uni- and bivariate analyses were based on 3,839 flock observations for which the Salmonella status of the individual flocks was known. The multivariate analysis was based on 2,776 flock observations, excluding flocks with missing values.
TABLE 2. Descriptive statistics for variables which were significant (P < 0.20) in the bivariate analysis, but not included in the final logistic regression model; 3,839 Danish broiler flocks, 1995
Variable1
Level
Percentage of flocks
Density
<5 5 to 10 10 to 15 15 to 20 20 to 25 25 to 30 ≥30 <10,000 10 to 20,000 20 to 30,000 30 to 40,000 ≥40,000 <500 500 to 1,000 1,000 to 1,500 ≥1,500
0.0 2.4 0.1 2.1 41.3 53.7 0.6 9.1 21.1 24.7 17.4 27.8 25.0 20.5 23.3 31.2
0.0 6.6 100.0 4.0 10.4 14.3 5.3 9.7 11.9 8.8 16.1 13.8 11.6 10.2 11.9 14.2
<1960 1960 to 70 1970 to 80 1980 to 90 ≥1990 R208 COBB Shaver RossPM Many lines/mix of line – +
35.4 13.1 17.5 19.8 14.1 71.4 14.5 1.2 2.0
10.5 14.3 12.0 13.3 12.9 12.5 8.6 4.6 5.3
11.0 92.5 7.5
17.1 12.2 8.1
Zinc bacitracin 25.8 Virginiamycin 1.0 Flavophospholipol 3.7 24.8 Avilamycin 44.7
9.2 8.3 6.2 14.2 13.7
(%)
Statistical Analysis
Flock size
Area
Construction year
Breed
Medication Growth promoters Avoparcin
RESULTS
Data Sets Statistically significant differences between the flock sizes (t test, P > 0.93) and the Salmonella frequencies (x2 test, P = 0.71) in the original and analyzed data sets were not observed.
Uni- and Bivariate Analysis In 1995, 12.2% of the broiler flocks sampled positive for S. typhimurium. In the bivariate analysis the following variables were shown not to be significantly associated with the presence of this serotype, and were therefore excluded from the multivariate analysis: observation of beetles, days between disinfection and replacement of chickens, and visual appearance of the bedding.
Salmonella test, wk
1 1 to 2 to 3 to ≥4 Parent flock n = AM district n = Slaughterhouse n = Feed mill n = – Lagstyph3 +
2 3 4 125 7 9 10
0.2 21.8 39.5 30.0 8.6 <1 to <1 to <1 to 5 to 87.0 13.1
5.8 23 23 20
12.5 13.8 13.5 9.0 13.1 0 to 0 to 4 to 8 to 8.7 35.7
Downloaded from http://ps.oxfordjournals.org/ at Kainan University on April 30, 2015
A retrospective longitudinal study design was used (Kleinbaum et al., 1982). The individual broiler flock was the unit of concern. Based on uni- and bivariate analyses, variables were selected for the multivariate analysis, with only significant (x2 test, P < 0.20) and independent variables being included. A multivariate logistic regression analysis was performed using the CATMOD procedure in SAS (SAS Institute, 1990). The best model fit was found by a combined forward and backward selection process of adding or subtracting a variable (P < 0.05 for the model being better with the addition or subtraction) or an interaction (P < 0.01) from the model (Hosmer and Lemeshow, 1989). Estimation of the parameters for the main variables and the interaction term in the final model was done with and without a term related to random effect on the house and on the farm level using MLn software (Rasbash and Woodhouse, 1995) for analyzing multilevel statistical models (Goldstein, 1995). This analysis was performed due to lack of independence between flocks produced within the same house and between houses located on the same farm. The random effects (Uj) of the houses or farms were assumed to be normally distributed (with zero mean and variance s2).
S. typhimurium2
60 15 14 21
1Density = number of chickens placed per square meter; Flock size = number of chickens placed in the house; Area = area of house in square meters; Construction year = construction year of the house; Breed = breed of the parent flock; Salmonella test = the age of the chickens when Salmonella tested; Slaughterhouse = slaughterhouse of the preceding flock. 2Salmonella typhimurium prevalence in percentage for the actual covariate level. 3Lagstyph = S. typhimurium status of the preceding flock.
851
RISK FACTORS RELATED TO SALMONELLA TYPHIMURIUM INFECTIONS
different levels of the variables and the corresponding number of flock observations are shown in Tables 2 and 3. Significant association was found among the hatcheries, the feed mills, and the slaughterhouses (data not shown). As collinearity between these variables was suspected to bias the further analysis, only the hatcheries that had the highest odds ratios of these three variables were included in the multivariate analysis. The variable “S. typhimurium status of the preceding flock” was not evaluated in the multivariate analysis due to the association observed between the parent flocks and broiler flocks in the same house.
TABLE 4. Odds ratios and confidence intervals for the interaction term between hatchery and season, included in the multivariate analysis of Salmonella typhimurium infections in Danish broiler flocks, 1995 Hatchery Season Summer Autumn Winter Spring
Multivariate Analysis
Flock percentage2
Adjusted S. odds typhimurium3 ratio
19.5 24.9 19.7 12.1 9.6 6.7 7.5
9.9 9.9 10.2 18.7 17.7 15.1 10.5
14 0.8 1.1 1.5 2.5* 0.8 1.0
. . 0.5 0.7 0.9 1.4 0.4 0.5
96.5 3.5
11.0 43.0
14 2.8*
. . . 1.7 to 4.4
98.8 1.2
12.0 29.6
14 3.2*
. . . 1.4 to 16.4
24.2 25.7 24.9 25.3
15.2 10.5 13.5 9.6
. . . .
8 to 38
4 to 22
C (28.1%1)
D (38.2%1)
14 . . . 1.1 (0.3 to 4.2) 1.9 (0.6 to 6.3) 1.6 (0.5 to 5.5)
1.5 (0.2 to 13.1)2 2.9 (0.5 to 17.6) 2.2 (0.4 to 13.4) . . .3
12.8* (4.7 to 4.1* (1.5 to 13.6* (5.0 to 6.2* (2.3 to
2.5 (0.8 to 4.3* (1.6 to 5.0* (1.8 to 4.2* (1.6 to
34.8) 12.2) 36.9) 16.9)
–7.4) 11.7) 13.5) 11.5)
. . . .
.5 . . .
. . .5
95% Confidence interval
. . . .
. . . .
. to to to to to to
1.3 1.7 2.5 4.4 1.7 1.9
.5 . . .
. . .5
1House = number of chicken houses on the farm, Parent flock No. 10 =
parent flock confirmed to be infected with S. typhimurium, Parent flock No. 83 = parent flock suspected of being infected with S. typhimurium. 2Percentage of flocks in the study (3839). 3Salmonella typhimurium prevalence in percentage for the actual covariate level (3,839 flocks). 4Reference. 5See Table 4 for odds ratios and confidence intervals for hatchery and season. 6Only minimum and maximum values shown. *P < 0.05.
of flocks in the study (3,839). interval. 3The hatchery had no deliveries of broiler flocks with S. typhimurium in that season. 4Reference. *P < 0.05. 2Confidence
an interaction term between season and hatchery (Table 4). An increased risk for S. typhimurium infection in the broiler flocks was associated with two parent flocks [one bacteriologically confirmed (No. 10) and one suspected of being infected with S. typhimurium (No. 83)], and having five houses on the farm (Table 3). In addition, an increased risk for S. typhimurium infection in the broiler flocks was associated with two of the hatcheries with season (observed as an interaction term between these two variables) (Table 4). The final model was further analyzed by including a random effect at house and farm level. Added separately, the random effect at house (variation between houses su0 = 0.70) and farm level (variation between farms su0 = 0.62) were statistically significant, but when both random effects were included simultaneously only the random effect at farm level (variation between farms su0 = 0.56) was shown to be statistically significant in the model.
DISCUSSION The multivariate epidemiological approach to the study of risk factors for salmonellosis in poultry has so far not received much attention. Henken et al. (1992) showed that feed mills and poor hygiene barriers, including the lack of disinfection tubs, represented risk factors. Hatcheries and feed mills, size of farms in terms of number of houses, a positive Salmonella status of the preceding flock, and rearing of flocks in the autumn have been identified as important risk factors for salmonellosis by Angen et al. (1996). With the exception of two variables (parent flocks and the random effect at the farm level) the five risk factors identified in the present analysis for S. typhimurium were in agreement with those identified by Angen et al. (1996).
Downloaded from http://ps.oxfordjournals.org/ at Kainan University on April 30, 2015
TABLE 3. Prevalence and odds ratios for variables included (P < 0.05) in the final multivariate logistic regression model (including a random effect term at farm level) of Salmonella typhimurium infections in 2,776 Danish broiler flocks, 1995
House 1 2 3 4 5 6 ≥7 Parent flock No. 10 (confirmed) – + Parent flock No. 83 (suspected) – + Season Dec to Feb Mar to May Jun to Aug Sep to Nov Hatchery6 n = 4
B (7.5%1)
1Percentage
In the multivariate logistic regression model, five variables were shown to be associated with S. typhimurium infections in Danish broiler flocks (Table 3) in addition to
Variable1
A (21.3%1)
852
SKOV ET AL.
reflect that houses at the farm are epidemiologically connected (e.g., by the environment surrounding the houses at the farm, level of biosecurity, the feed mill delivering to the farm, or the slaughterhouse collecting the chickens at the farm). It might also indicate differences in management, including differences between advisers connected to the farms. Others have found feed mills (Henken et al., 1992, Angen et al., 1996) and slaughterhouses (Christensen et al., 1994) to represent risk factors, although these investigations were not specific for S. typhimurium. All Danish broiler feed is heat-treated, and all final broiler feed investigated in 1995 sampled negative for S. enterica (Anonymous, 1995b). However, the sensitivity of the Salmonella test used in these analyses has not been evaluated and, consequently, the reliability of the test results is unknown. Further investigations are also needed to evaluate the significance of the slaughterhouses as a source of contamination for the chicken houses and the environment surrounding these houses while collecting the chickens for slaughter. The number of houses on the farm was shown to be part of the multivariate logistic regression model, but only farms having five houses were shown to be significantly associated with S. typhimurium infection in broiler flocks (odds ratio = 2.5). It has not been possible to explain this finding biologically, and considering the relatively few farms having more than five houses (Table 3), it is difficult to investigate if special problems can be related to having more than five houses on a farm. Previous investigations (Angen et al. 1996) have shown that broiler flocks at farms with more than three houses have an increased risk of S. enterica infection. Genetic differences in resistance to mortality caused by S. enterica have been reported by Bumstead and Barrow (1993), but little is known about the resistance to colonization or infection. In the present investigation, the breeding lines used for parent stock were shown not to be significantly associated with S. typhimurium infection. This lack of association might be explained by the fact that Ross 208 was the breeding line of more than 71% of the flocks (Table 2) and, thus, the genetic variability of the flocks was low. Medication has previously been shown to influence infection and the shedding of S. typhimurium (Smith and Tucker, 1975). In the present investigation, these observations could not be confirmed. The significance of avoparcin and other antibiotic growth promoters for colonization of the alimentary tract by Salmonella, and the duration of shedding of these organisms by infected chickens have been described in a number of papers (Smith and Tucker, 1975, 1978, 1980; Linton et al, 1985; Hinton, 1988; Barrow, 1989) but the results have been inconsistent. The use of antibiotic growth promoters is presently a subject of strong debate due to the problems related to cross resistance to antibiotics used in humans. None of the antibiotic growth promoters used was shown to be
Downloaded from http://ps.oxfordjournals.org/ at Kainan University on April 30, 2015
In the present study, two parent flocks (No. 10 and 83) were shown to represent a risk factor for contamination of broiler flocks with S. typhimurium. This finding is not surprising as vertical transmission of S. enterica has been demonstrated on several occasions for certain serovars (Lahellec and Colin, 1985; Mcllroy et al., 1989), underlining the importance of a top-down strategy for the eradiation of S. enterica. One of these “risk parent flocks” (No. 10) was the only parent flock in the present investigation confirmed as being infected with S. typhimurium according to bacteriological investigations. It consisted of six subunits, all located on the same farm, but in separate sections. Epidemiologically, each section was regarded as a separate unit. Only two out of six sections tested positive for S. typhimurium. Parents from these sections were killed late in November, 1994, whereas the other subunits remained in production until June, 1995. These four subunits were shown to represent an important risk factor, indicating that even though parent flocks are kept in separate sections they should all be considered as one epidemiological unit. The other parent flock shown to represent a risk factor was only found suspected of infection but found to be bacteriologically negative on subsequent examination. This observation indicates that the sensitivity of the confirmation procedure used for S. typhimurium might be insufficient, especially in cases with low Salmonella prevalence in both the flock and the individual hen. The present results are in agreement with previous investigations (Chrie´l, 1997) demonstrating an increased risk for S. typhimurium infections associated with receiving chickens from infected parent flocks, independent of the proportion of the flock these chickens represent. In addition, Chrie´l (1997) also found an increased risk was associated with receiving chickens from more than one parent flock. With only one exception (Hatchery D in the summer), receiving chickens from Hatchery C or D was shown to be connected with an increased risk of S. typhimurium infection (Table 4). These hatcheries received eggs from the “risk parent flocks” (Flock 10 was owned by Hatchery C, whereas Flock 83 was owned by Hatchery D) discussed above, but, as seen from Table 4, the hatcheries themselves also represented a risk factor— most likely caused by cross contamination at the hatcheries or because of other Salmonella-positive parent flocks not identified. This finding is in agreement with previous investigations showing that the hatcheries on their own might contribute to the spread of S. enterica (Christensen et al., 1997). The random effect at the house level was significant when included as the only random effect in the model, which is in agreement with the observations reported by Angen et al. (1996). However, inclusion of an additional random effect at farm level rendered only the random effect at the farm level statistically significant. This finding indicates that it is possible to clean and disinfect a house contaminated with S. typhimurium and might
RISK FACTORS RELATED TO SALMONELLA TYPHIMURIUM INFECTIONS
ACKNOWLEDGMENTS This research was supported by The Danish Agricultural and Veterinary Research Council (Grant Number 20-3510-2) and The Ministry of Agriculture and Fisheries (SUN 94-KVL-11).
REFERENCES Angen, Ø., M. N. Skov, M. Chrie´l, J. F. Agger, and M. Bisgaard, 1996. A retrospective study on salmonella infection in Danish broiler flocks. Prev. Vet. Med. 26: 223–237.
Anonymous, 1995a. Annual report on zoonoses in Denmark 1995. Danish Zoonosis Centre. Danish Veterinary Laboratory, Copenhagen, Denmark. Anonymous, 1995b. Annual report on animal feed in Denmark in 1995. Danish Plant Directorate, Lyngby, Denmark. Baggesen, D. L., and H. C. Wegener, 1994. Phage types of Salmonella enterica ssp. enterica serovar typhimurium isolated from production animals and humans in Denmark. Acta Vet. Scand. 35:349–354. Barrow, P. A., 1989. Further observations on the effect of feeding diets containing avoparcin on the excretion of salmonellas by experimentally infected chickens. Epidemiol. Infect. 102:239–252. Bisgaard, M., K. Haaning, and G. Velling, 1982. Poultry-borne salmonellosis. Prophylactic measures in the broiler production including bacteriological investigations before and after slaughter. Pages 227–231 in: XIV Nordic Veterinary Congress. Copenhagen, Denmark. Bisgaard, M., 1992. A voluntary salmonella control programme for the broiler industry, implemented by the Danish Poultry Council. Int. J. Food Microbiol. 15:219–224. Brown, J. D., D. L. Baggesen, H. B. Hansen, H. C. Hansen, and M. Bisgaard, 1994. The characterization of Danish isolates of Salmonella enterica serovar enteritidis by phage typing and plasmid profiling: 1980–1990. APMIS 120:208–214. Bumstead, N., and P. A. Barrow, 1993. Resistance to Salmonella gallinarum, S. pullorum, and S. enteritidis in inbred lines of chickens. Avian Dis. 37:189–193. Chrie´l, M., 1997. Hierarchical, crossclassified modelling of Salmonella typhimurium in Danish Broilers. Pages 163–171 in: Proceeding of the Dutch/Danish Symposium on Animal Health and Management Economics. Copenhagen, Denmark. Christensen, J. P., D. J. Brown, M. Madsen, J. E. Olsen, and M. Bisgaard, 1997. Hatchery-borne Salmonella enterica serovar Tennessee infections in broilers. Avian Pathol. 26:155–168. Christensen, J. P., M. N. Skov, K. H. Hinz, and M. Bisgaard, 1994. Salmonella enterica serovar Gallinarum biovar gallinarum in layers: epidemiological investigations of a recent outbreak in Denmark. Avian Pathol. 23:489–501. Goldstein, H., 1995. Multilevel Statistical Models, second edition. Institute of Education, University of London, London, U.K. Henken, A. M., K. Frankena, J. O. Goelema, E.A.M. Graat, and J.P.T.M. Noordhuizen, 1992. Multivariate epidemiological approach to salmonellosis in broiler breeder flocks. Poultry. Sci. 71:838–843. Hinton, M., 1988. Salmonella colonization in young chickens given feed supplemented with the growth promoting antibiotic avilamycin. J. Vet. Pharmacol. Therap. 11: 269–275. Hosmer, D. W., and S., Lemeshow, 1989. Applied Logistic Regression. John Wiley and Sons, New York. NY. Kleinbaum, D. G., L. L. Kupper, and H. Morgenstern, 1982. Epidemiological Research. Van Nostrand Reinhold Co., New York, NY. Lahellec, C., and P. Colin, 1985. Relationship between serotypes of salmonella from hatcheries and rearing farms and those from processed poultry carcasses. Br. Poult. Sci. 26: 179–186. Linton, A. H., Z.A.M. Al-Chalaby, and M. H. Hinton, 1985. Natural subclinical salmonella infection in chickens: A potential model for testing the effect of various procedures on salmonella shedding. Vet. Rec. 6:361–364.
Downloaded from http://ps.oxfordjournals.org/ at Kainan University on April 30, 2015
associated with increased detection of S. typhimurium. However, as all broiler flocks in Denmark at the time of the investigation were given antibiotic growth promoters the significance in relation to colonization and excretion of Salmonella might be difficult to evaluate (since February 1998 there has been a voluntary cessation of the use of antibiotic growth promotors in Denmark). Excretion of S. enterica has been shown to decrease with age (Bisgaard et al., 1982). However, the age of the chickens was not a significant variable in the present investigation (Table 2). The risk analysis for S. enterica in Danish broiler flocks during the years 1992 to 1993 (Angen et al., 1996) was based on bacteriological examination of cecal tonsils from 16 broilers from each flock (20% prevalence level). The present study was based on bacteriological examination of 60 fresh fecal samples from each flock (5% prevalence level). Comparison of the two different methods in 320 broiler flocks revealed a larger serotype diversity in the flocks related to the increased number of samples being examined. However, the investigation also showed that certain serotypes were more frequently detected with one of the methods, i.e., more flocks were found infected with S. typhimurium on examination of fecal samples, whereas more flocks were found infected with S. enteritidis when investigating the cecal tonsils (M. Chrie´l, The Royal Veterinary and Agricultural University, Denmark, unpublished results). These differences might be related to differences in the pathogenesis (e.g., invasiveness) of the different serotypes. In conclusion, the present investigation has shown that multivariate logistic regression analysis is a valuable tool in outlining risk factors associated with S. typhimurium infections in broilers. The results strongly indicated that parent flocks infected with S. typhimurium, including confirmed flocks and flocks suspected of being infected, as well as the hatcheries receiving eggs from these flocks, represented significant risk factors. This conclusion underlines the significance of a topdown strategy for the eradication of S. typhimurium. In addition, the significance of the random effect related to the farm indicated that all houses at the farm seemed to be exposed to similar external risk factors.
853
854
SKOV ET AL.
Martin, S. W., A. H. Meek, and P. Willeberg, 1987. Veterinary Epidemiology. Principles and Methods, Iowa State University Press, Ames, IA. Mcllroy, S. G., R. M. McCracken, S. D. Neill, and J. J. O’Brien, 1989. Control, prevention and eradication of Salmonella enteritidis infection in broiler and broiler breeder flocks. Vet. Rec. 125:545–548. Nordic Committee on Food Analysis, No. 71. 1991. c/o VTT Bio-og Livsmedelsteknik PB 1500, Fin-02044 VTT, Finland. Olsen, J. E., M. Sørensen, D. J. Brown, K. Gaarslev, and M. Bisgaard, 1992. Plasmid profiles as an epidemiological marker in Salmonella enterica serovar Berta infections. APMIS 100:221–228. Rasbash, J., and G. Woodhouse, 1995. MLn Command reference Version 1.0. Institute of Education, University of London, London, U.K.
SAS Institute, 1990. SAS Stat User’s Guide. Version 6, Fourth Edition. SAS Institute Inc., Cary, NC. Smith, H. W., and J. F. Tucker, 1975. The effect of feeding diets containing permitted antibiotics on the feacal excretion of Salmonella typhimurium by experimentally infected chickens. J. Hyg. Camb. 75:293–301. Smith, H. W., and J. F. Tucker, 1978. The effect of antimicrobial feed additives on the colonization of the alimentary tract of chickens by Salmonella typhimurium. J. Hyg. Camb. 80: 217–231. Smith, H. W., and J. F. Tucker, 1980. The virulence of salmonella strains for chickens: their excretion by infected chickens. J. Hyg. 84:479–488. Wegener, H. C., D. L. Baggesen, and K. Gaarslev, 1994. Salmonellosis in Denmark 1988–1993. APMIS 102:521–525.
Downloaded from http://ps.oxfordjournals.org/ at Kainan University on April 30, 2015