Abstract
This study investigated potential risk factors for the presence of Salmonella on 89 Alberta swine-finishing farms with the use of a questionnaire. Salmonella status was regressed on each fixed effect in a logistic mixed regression model, with farm as the random effect. Eleven variables were significant at the 10% level: farm type, number of square feet per pen, number of pigs per pen, source of feed, ration type, dust control measures, cat presence, reported effective mouse-control measures, time required to be away from pigs before visiting the farm, precautions taken when entering or leaving the farm, and reported use of antimicrobials through water. Three factors remained significant at the 5% level in the multivariable analysis: farm type, ration type, and precautions taken when entering or leaving the farm. Finishing barns at multisite operations or individual grow-to-finish farms had a greater risk of the presence of Salmonella at a single visit than did finishing barns at farrow-to-finish farms. The use of pelleted and wet feed was associated with higher odds of the presence of Salmonella than was the use of meal feed. Farms that required their personnel or visitors to shower before entering and before leaving had increased odds of the presence of Salmonella compared with farms that provided boots and coveralls; no significant difference was observed between the latter category and farms that used boot disinfection. Further work is necessary to better understand the effectiveness of all-in/all-out pig management and disinfection practices in reducing the presence of Salmonella in swine and to evaluate the association with certain types of feed.
Résumé
L’objectif de la présente étude était d’investiguer au moyen d’un questionnaire les facteurs de risque potentiels pour la présence de Salmonella dans 89 fermes porcines de finition en Alberta. Une régression sur le statut pour la présence de Salmonella et chaque effet fixe a été faite dans un modèle de régression logistique mixte, avec la ferme comme effet aléatoire. Onze variables étaient significatives à un niveau de 10 % : le type de ferme, le nombre de pieds carrés par enclos, le nombre de porcs par enclos, la source de nourriture, le type de ration, les mesures de contrôle de la poussière, la présence de chat, des mesures de contrôles des souris reconnues comme efficaces, le délai entre un contact avec des porcs et la visite de la ferme, les mesures préventives prises à l’entrée et à la sortie de la ferme et l’utilisation rapportée d’antimicrobiens administrés dans l’eau de boisson. Trois facteurs sont demeurés significatifs à un niveau de 5 % dans l’analyse multivariée : le type de ferme, le type de ration et les mesures préventives prises à l’entrée et à la sortie de la ferme. Les fermes de finition d’opérations multisites ou les fermes individuelles de type engraissement–finition étaient plus à risque pour la présence de Salmonella lors d’une visite unique que les fermes de finition dans les opérations de type naisseur–finisseur. L’utilisation de nourriture cubée et mouillée était associée avec un risque plus élevé de retrouver du Salmonella que l’utilisation d’un aliment en farine. Sur les fermes où l’on exigeait que le personnel ou les visiteurs prennent une douche avant d’entrer et de quitter, il y avait un risque plus élevé de retrouver du Salmonella comparativement aux fermes qui fournissaient des bottes et des vêtements protecteurs; aucune différence significative n’a été observée entre la dernière catégorie de fermes et celles qui utilisaient une procédure de désinfection des bottes. Du travail supplémentaire est nécessaire afin de mieux comprendre l’efficacité d’une gestion de type tout plein/tout vide et des habitudes de désinfection à réduire la présence de Salmonella chez le porc et pour évaluer l’association avec certains types d’aliments.
(Traduit par Docteur Serge Messier)
Introduction
The consumption of Salmonella-contaminated pork may cause illness in humans (1). In Denmark, the Netherlands, and Germany, an estimated 10% to 15%, 14% to 19%, and 18% to 23%, respectively, of cases of salmonellosis in humans were attributable to pork (2). As a result of growing concern over food safety, and particularly after implementation of the Danish Salmonella surveillance program in swine in the early 1990s (3), other European Union (EU) countries increased their monitoring and control activities related to Salmonella in swine (4–6). Such measures have not yet been widely implemented in North America. A recent report (7) suggests that only 3% of human Salmonella outbreaks of known etiology in the United States were attributable to pork. Nevertheless, Salmonella has become a very important food safety issue for the North American swine industry owing to the potential impact on domestic public health and consumer confidence, as well as the need to remain competitive in export markets (8).
The risk factors for on-farm subclinical infections with Salmonella in swine have been investigated more intensively in the EU than elsewhere (4,9–13). The type of feed, the purchase of animals from highly contaminated herds, continuous production flow, low health status, poor hygiene, and the presence of rodents are generally considered to be important factors associated with a high prevalence of such Salmonella infections (5,14). Funk and Gebreyes (14) noted that most of the risk-factor studies have been conducted in European countries and that the epidemiologic aspects of Salmonella infections in swine in the EU, as well as the impact and feasibility of interventions, may not be directly applicable to US herds owing to differences in production systems, industry structure, and regulatory organization. This may also be true in Canada, where preharvest (farm-level) epidemiologic studies of Salmonella infections in swine have been limited compared with the EU and the United States. Risk factors for farm-level Salmonella infection in swine herds have been investigated only in Quebec (15), where the use of multiple sources for the purchase of animals was found to correlate with the presence of Salmonella in the herds. The development of efficient strategies to control Salmonella in Canadian swine requires an in-depth understanding of the epidemiologic aspects of Salmonella infection across Canada, as management practices may differ across regions. To the best of our knowledge, no epidemiologic study had investigated on-farm risk factors associated with the prevalence of Salmonella infections in Alberta swine. In 2001, Alberta had the 4th-largest pig inventory in Canada, with 14.5% of Canada’s 13.96 million pigs (16).
The purpose of this study was to investigate farm-level management practices potentially associated with the presence of Salmonella on large Alberta swine-finishing farms.
Materials and methods
A longitudinal study was conducted in Alberta from May to September 2000 to determine the farm prevalence and serovar diversity of Salmonella infections, as well as the risk factors, in large swine-finishing barns. The methods of farm selection, sample collection, and laboratory testing have been described in detail elsewhere (17), as have the Salmonella recovery rates at the sample and the farm level, along with the serovar and phagetype diversity of the isolates (17).
In brief, 10 veterinary practitioners selected 90 swine farms that each marketed at least 2000 pigs per year and whose producer was willing to participate in the study. The barns were visited 3 times, approximately 1 mo apart. A total of 15 samples (5 per visit) of pooled pen feces were collected from finishing pigs on each farm. At each visit, the practitioner randomly selected 5 pens containing finishing pigs that were within 45 to 60 d of shipment to market and collected 5 g of fresh fecal material from each of 5 spots in the pen, for a pool of approximately 25 g.
At the Agri-Food Laboratories Branch of the Food Safety Division of Alberta Agriculture, Food and Rural Development, in Edmonton, standard culture protocols were used to isolate Salmonella from the samples (18). All samples were thoroughly mixed with the use of a sterile spatula. Two 5-g aliquots were selected for enrichment according to specific methods. The 1st aliquot was inoculated into 45 mL of selenite–cystine broth and incubated at 35°C for 24 h; the broth was then inoculated onto XLT4 and Rambach agar plates, incubated at 35°C, and evaluated for typical colonies of Salmonella species at 24 and 48 h. The 2nd aliquot was added to 45 mL of buffered peptone water (BPW) and incubated at 35°C for 24 h. Next, 1 mL of the BPW solution was inoculated into 10 mL of tetrathionate broth supplemented with 0.2 mL of iodine and incubated at 35°C for 24 h. After these initial enrichment steps, the tetrathionate broth was inoculated onto XLT4 agar plates, to be incubated at 35°C for 24 h, and onto modified semisolid Rappaport–Vassiliadis (MSRV) plates, to be incubated at 41.5°C for 20 to 24 h. The “halos” of growth that occurred on the MSRV plates were subsequently streaked onto XLT4 and Rambach plates and incubated at 35°C for 24 h. Colonies suspected to be Salmonella on the XLT4 and Rambach plates were screened with the use of lysine iron agar, triple sugar iron agar, and urease, as well as by means of agglutination with Salmonella poly O and O1 antiserum (Denka Seiken, Tokyo, Japan) or a Salmonella latex agglutination kit (Oxoid, Basingstoke, England). Five presumptive Salmonella colonies were randomly harvested, frozen at −70°C, and then thawed, inoculated, and transferred on Columbia agar slants to the Health Canada Laboratory for Foodborne Zoonoses (LFZ), Office International des Epizooties Reference Laboratory for Salmonellosis, Guelph, Ontario, for serotyping and phagetyping (14).
A questionnaire, available from the 1st author on request, was completed by the owner or operator of the farm with the herd veterinarian at the 2nd visit. One section pertained to the swine operation and covered production system, pig flow, antimicrobial use, and disease status; the other section pertained to the finishing farm and covered farm management and biosecurity practices. Of the 70 questions, 52 (74%) were closed (with Yes/No or preset options for answers), 16 (23%) were semiclosed (asking about number of days, animals, or rooms), and 2 (3%) were open-ended (requiring descriptions). A multisite operation was defined as a corporate enterprise with 2 or more sites. Other operations (farrow-to-finish, farrow-to-wean, individual grow-to-finish, or finishing) referred to owner-operated facilities. The questionnaire was pretested by 3 pork producers, 3 provincial pork extension specialists, and 3 veterinarians familiar with survey technique.
Statistical analysis
Questionnaire data were entered into a spreadsheet program (Excel 2000, Microsoft Corporation, Redmond, Washington, USA) and combined with the bacteriologic results. Data were verified by checking each entry against the original hard-copy data and then transferred into the statistical software package Stata Intercooled, version 8 (Stata Corporation, College Station, Texas, USA) for descriptive analyses. All variables were screened with the use of descriptive statistics to check accuracy and to identify those that might be of little value for modeling, such as variables with large numbers of missing observations, low variability, or ambiguous answers (19). Categorical variables that had a small number of observations per category were recategorized according to biologic sense or eliminated (19).
The data were initially split into a set for the 76 farrow-to-finish farms and a set for all 90 participating farms. All variables, including those pertaining to breeding (e.g., average sow number) and nursery phases of production (e.g., average weaning age), were analyzed in the farrow-to-finish data set only, to avoid the elimination of farms that did not have these production phases and to which this type of data was not applicable.
All potential explanatory variables were individually screened by logistic regression with the xtlogit procedure in Stata. Clustering at the farm level was accounted for by including a farm identification variable as a random effect in all regression models (19). When independent variables were highly correlated (r > 0.8), preference was given to the variable with the smallest P-value. The relationship between continuous independent variables and the outcome variable (whether at least 1 of the 5 pooled fecal samples collected during a visit was or was not positive for Salmonella) was assessed by generating a smoothed scatterplot of the logarithmic odds of the outcome variable against the potential explanatory variable (e.g., number of marketed hogs in 1999). If the relationship was not linear, the variables were categorized by means of biologic sense or median values. The variable for veterinary practitioner was believed to contain a complex of factors that could affect management practices and potentially affect the association between the farm factors of interest and the outcome variable. Therefore, this variable was considered a potential confounder in the analyses and forced into all models as a fixed-effect dummy variable.
All potential explanatory variables with a P-value ≤ 0.10 in a random farm-effect model were offered to the full model. This model was reduced by backward elimination by means of generalized estimating equations (GEEs) in Stata, with farm as the group-level identifier and farm visit as the time identifier. Confounding was assessed every time a nonsignificant variable was dropped from the model by comparing the change in the coefficients for the variables remaining in the model. Variables that altered the coefficients for the independent variables of interest by 25% or more were classified as confounding and retained in the model (19). Once the main-effects model was obtained, 2-way interactions were generated and checked for statistical significance and biologic plausibility. The GEE models were run with the use of unstructured correlation, which allows correlations to be freely estimated from the data (20). The models were assessed by investigation of extreme data points, largely guided by delta–beta statistics (19).
Results
One farm was visited only twice for logistic reasons; therefore, there were 269 farm visits in total. Salmonella was detected in 14.3% of all fecal samples. There was at least 1 Salmonella-positive sample from 46 (51%) of the 90 farms and 86 (32%) of the 269 visits.
Several variables were discarded because of low variability (e.g., bedding type and number of pigs purchased at an auction market), large number of missing values (e.g., farm disease status and farm location), or ambiguity in the answer (e.g., pig weight at the beginning of the finishing period and supplier management). The information on farm disease status (16 specified diseases) and farm location was incomplete because 3 veterinary practitioners declined to complete that section of the questionnaire for 24 participating herds for privacy reasons. Table I summarizes the factors from the questionnaire used to analyze the risk of Salmonella presence in the study finishing farms.
Table I.
Summary of 51 variables derived from a questionnaire and assessed as potential risk factors for Salmonella in finishing pigs on 89 large Alberta swine farms visited 3 times during the summer of 2000 (266 visits)
| Section (number of variables) | Variables |
|---|---|
| Production system and pig flow (15) | Type of swine operation (farrow-to-finish or others); pig flow in farrowing unit (continuous or all-in/ all-out by room or barn); pig flow in nursery (continuous or all-in/all-out by pen, room, or barn); pig flow in growers (continuous or all-in/all-out by pen, room, or barn); pig flow in finishers (continuous or all-in/all-out by pen, room, or barn); farm sow number (. 300 or # 300); purchase of replacement gilts in 1999 (Yes/No); number of sources of replacement gilts ($ 2 or 0/1); average weaning age (. 21 or # 21 d); number of nursery barns (. 1 or 1); site of nursery barns (1 or multiple); mixing weaned pigs from $ 2 sources (Yes/No); number of finishing barns (. 1 or 1); mixing finishing pigs from $ 2 sources (Yes/No); and number of pigs marketed in 1999 (. 6200 or # 6200). |
| Antimicrobial and anthelmintic use during the past 12 mo (10) | Reported in-feed antimicrobial use in growers (Yes/No) and finishers (Yes/No); reported use of penicillin, tetracycline, or sulfamethazine (Yes/No) and of carbadox (Yes/No) in weaners; reported in-feed use of penicillin (Yes/No), tetracycline (Yes/No), and tylosin (Yes/No) in growers and finishers; reported antimicrobial use through water (Yes/No) and injection (Yes/No) at the farm level; and reported anthelmintic use (Yes/No) at the farm level. |
| Management practices in the sampled finisher barn (17) | Barn age (. 3 or # 3 y); construction type (conventional solid walls or other); number of rooms per barn (. 2 or # 2); number of pens per room (. 24 or # 24); area per pen (. 130 or # 130 sq ft); number of pigs per pen (. 20 or # 20); mixing pigs during the finishing period (Yes/No); type of pen walls (solid, spindles, or a combination); type of floor (solid, slatted, or partially slatted); manure storage (underfloor pit or others); manure removal (through slats or other); feed source (feed mill or mixing on farm); type of ration (meal, pellets, or wet); water source (well or other); water chlorination (Yes/No); dust control (Yes/No); pressure washing and disinfection (Yes/No). |
| Biosecurity practices in the sampled finisher barn (9) | Mandatory time for visitors away from other pigs (. 24 or # 24 h); boots disinfected, boots and coveralls provided, or shower-in/shower-out; bird access (Yes/No); bird access to the feed storage bins (Yes/No); dog access (Yes/No); cat access (Yes/No); reported effective fly control (Yes/No); reported effective rodent control (Yes/No); other swine operations within 3 mi (Yes/No). |
The annual production of the selected farms represented approximately 25% of market hog production in Alberta. The farms were approximately proportionally representative of major swine production areas in Alberta: 76 (84%) were farrow-to-finish farms, 6 (7%) each were part of a multisite operation or an individual grow-to-finish farm, and 2 (2%) were individual finisher farms. One practitioner contributed only 1 farm to the farrow-to-finish data set, and this observation was excluded from the set to permit convergence of the models. The number of participating farms for the remaining 9 practitioners ranged from 3 to 20, for a total of 89 farms and 266 visits. Table II presents descriptive statistics for the continuous variables.
Table II.
Descriptive statistics for the continuous variables assessed as potential risk factors
| Potential risk factor | Minimum | 1st quartile | Median | 3rd quartile | Maximum |
|---|---|---|---|---|---|
| Average number of sows | 100 | 160 | 300 | 500 | 3000 |
| Number of gilts purchased in 1999 | 0 | 0 | 60 | 194 | 1000 |
| Number of nursery pig barns | 1 | 1 | 1 | 1 | 8 |
| Weaning age (d) | 14 | 21 | 23 | 25 | 28 |
| Number of finishing pig barns | 1 | 1 | 1 | 2 | 5 |
| Numbers in the sampled finishing barn | |||||
| Finishers marketed in 1999 | 2000 | 3300 | 6200 | 9600 | 42 000 |
| Rooms | 1 | 1 | 2 | 5 | 22 |
| Pens per room | 4 | 13 | 24 | 37 | 112 |
| Square feet per pen | 100 | 116 | 130 | 160 | 1800 |
| Pigs per pen | 12 | 14 | 17 | 20 | 150 |
With regard to univariate and multivariable analysis, only the results for the all-farms data set are presented because the results were similar for both data sets. Only 1 variable pertaining to the breeding and nursery production phases was significant in the farrow- to-finish data set (weaning age) but insignificant in the multivariable analysis. The coefficients for the variable representing veterinary practitioner, included in the model as a potential confounder, are not reported because our goal was not to evaluate the effect of this factor but to control for it. Table III shows the 11 potential explanatory variables considered for multivariable analysis. During the backward elimination process, none of the nonsignificant risk factors in the multivariable model had a confounding effect on other variables. Seven variables were not associated with the outcome (P > 0.05) in the final model. Interaction terms were created among the 4 remaining variables and offered to the model simultaneously with the 4 main effects. The interaction of type of swine operation and ration was dropped from the model owing to a collinearity problem, and the other interaction terms were not significant. Table IV shows the final multivariable model. No influential cases were observed with the use of delta–beta statistics.
Table III.
Definition and distribution of independent variables selected (P ≤ 0.10) by univariate analysis, after controlling for farm as a random effect, for potential association with Salmonella positivity at a farm visit
| Definition | Level | No. (and %) of positive visits | P-value |
|---|---|---|---|
| Type of swine operation | Farrow-to-finish | 63/225 (28) | 0.03 |
| Other | 22/41 (54) | ||
| Number of square feet per pen | . 130 | 52/134 (39) | 0.07 |
| # 130 | 33/132 (25) | ||
| Number of pigs per pen | . 20 | 29/65 (45) | 0.07 |
| # 20 | 56/201 (28) | ||
| Feed source | Feed mill | 52/98 (53) | 0.0004 |
| Mixed on farm | 33/168 (20) | ||
| Type of ration | Meal | 28/174 (16) | < 0.0001 |
| Pellets | 48/80 (60) | ||
| Wet | 9/12 (75) | ||
| Dust control | Yes | 21/114 (18) | 0.003 |
| No | 64/152 (42) | ||
| Mandatory time away from pigs . 24 h | Yes | 31/63 (49) | 0.018 |
| No | 54/203 (27) | ||
| Presence of cats | Yes | 19/90 (21) | 0.043 |
| No | 66/176 (38) | ||
| Reported effective mouse control | Yes | 75/248 (30) | 0.09 |
| No | 10/18 (56) | ||
| Precautions at entering the sampled finishing barn | Boots disinfected | 1/18 (6) | 0.0001 |
| Boots and coveralls provided | 28/141 (20) | ||
| Shower-in/shower-out | 56/107 (52) | ||
| Reported antimicrobial use through watera | Yes | 43/161 (27) | 0.09 |
| No | 42/105 (40) |
In all production phases over the previous 12 mo.
Table IV.
Final multivariable model of farm factors associated with Salmonella presence on at least 1 of 3 consecutive visitsa
| Variable | Level | bb (standard error) | Odds ratio | 95% confidence interval | P-value |
|---|---|---|---|---|---|
| Type of ration | Pellets | 2.48 (0.52) | 11.94 | 4.22 to 33.11 | < 0.0001 |
| Wet | 2.44 (0.88) | 11.47 | 2.03 to 64.71 | 0.006 | |
| Meal | 0 (0) | 1 | — | — | |
| Type of farm | Otherc | 2.15 (0.69) | 8.58 | 2.20 to 33.11 | 0.002 |
| Farrow-to-finish | 0 (0) | 1 | — | — | |
| Precautions at entering the sampled finishing barn | Boots disinfected | −1.68 (1.22) | 0.18 | 0.016 to 2.03 | 0.17 |
| Shower-in/shower-out | 1.06 (0.42) | 2.89 | 1.25 to 6.61 | 0.012 | |
| Boots and coveralls provided | 0 (0) | 1 | — | — |
Intercept = −5.29.
Logistic regression coefficient.
Finisher pig barns from a multisite operation or individual grow-to-finish barns.
Discussion
The effect of ration type on the odds of a farm visit being Salmonella-positive was the largest association found in this study. Although contrary to the findings in Belgium (13), the increased risk of Salmonella shedding on the farms that fed pelleted feed compared with those that fed meal feed is consistent with the findings reported for some European countries (21,22) and has been hypothesized to be a result of smaller particle size, heat treatment, or the pelleted form (21,22). Funk and Gebreyes (14) noted that only a few peer-reviewed clinical trials investigating the effect of feed form have been published and that further work is necessary to evaluate the mechanism behind the higher prevalence of Salmonella shedding associated with pelleted feeds.
Other investigators have reported that pig herds fed dry diets were at increased risk of a high prevalence of Salmonella shedding compared with herds fed wet diets (12). In the present study, the farms that fed wet feed were at greater risk for Salmonella presence than the farms that fed meal. However, the farms that fed wet feed were small in number (4) and might not be representative of other Alberta farms using this type of feed; therefore, this finding should be interpreted with caution. Wet feeding in Europe often includes a fermentation step or addition of organic acid to prevent feed spoilage (14). Wet feeding without these preservation steps was associated with an increased risk of having a Salmonella-positive culture from pooled fecal samples in pig herds in the Netherlands (4). We did not collect additional information on the characteristics of the wet feed used on the 4 farms in this study.
Continuous pig production is believed to increase the risk for Salmonella infection compared with all-in/all-out pig production (5). In the present study, barns from farrow-to-finish farms had lower odds of a Salmonella-positive farm visit than barns from a multisite operation or individual grow-to-finish farms. Furthermore, there was no difference in Salmonella presence between farms that used continuous flow through various stages of production compared with those that used all-in/all-out practices. Davies et al (23) also reported that in North Carolina fewer farrow-to-finish farms were detected as Salmonella-positive compared with multisite operations with all-in/all-out management of finishing pigs. Similarly, in a study of 3-site production systems with all-in/all-out management in the United States, the prevalence of Salmonella shedding among finishers ranged from 0% to more than 70% (24).
Good management practices for disease control are recommended as the most practical method of Salmonella control on swine farms (14). Surprisingly, in the present study, farms that required their personnel and visitors to use shower-in/shower-out procedures before entering or leaving the farm had increased odds of the presence of Salmonella compared with farms that provided boots and coveralls; no significant difference was observed between the latter category and the farms that used only boot disinfection. The farms requiring boot disinfection were small in number (6), however, compared with the 2 other categories, and might not be representative of other Alberta farms using this practice. Perhaps some of these effects were masked or accounted for by the other variables in the study or by other factors not measured in this study.
It has been generally accepted, although rarely demonstrated, that improved hygiene due to the implementation of strict cleaning and disinfection protocols and accompanied by all-in/all out pig flow reduces Salmonella contamination within the farm environment and the pig population (5,14). The biologic premise is that the combination of cleaning and disinfection of the facility between groups of pigs and the segregation of age groups decreases the potential for Salmonella exposure and infection of new arrivals (14). In this study, no significant difference in Salmonella shedding was observed among farms that did no cleaning or only scraped the pens or used pressure washing with or without disinfection of barns between batches of finishing pigs. Similarly, Belgian researchers investigating herd-risk factors on 62 farrow-to-finish farms did not find any of the investigated hygiene measures to be significantly associated with Salmonella presence (13). Thus, there is a need to evaluate the effectiveness of the current cleaning, disinfection, and pig flow practices in terms of Salmonella control.
Other studies have reported some significant farm-management factors that were not identified as significant for Salmonella presence in this study. It has been suggested that closed pen separation can prevent Salmonella transmission between groups of pigs (25). Certain types of flooring decrease pig contact with fecal material, resulting in decreased fecal–oral transmission (14). In the United States the prevalence of Salmonella infection was lower in swine housed on fenestrated flooring than in those housed on flush-gutter flooring (23,26), whereas in Belgium a fully slatted floor had a protective effect compared with partially slatted floors (13). In the latter study, the type of pen wall (solid compared with spindle or a combination) and the type of flooring (solid compared with partial or full slatting) were not significantly associated with the odds of Salmonella presence. The main limitation of our study was the small number of sampled farms, which limited the statistical power, and this might be another reason why many risk factors reported as significant in other studies were not found to be significant in this study.
Potential associations between certain types of on-farm antimicrobial use and on-farm Salmonella presence were investigated because some European studies have reported this effect (9,10), although others have not observed it (11). For example, in the Netherlands, the use of tylosin as a growth promoter in the finishing feed was associated with a higher Salmonella seroprevalence in swine compared with the use of other growth promoters (9). In Greece, the risk of seropositivity was 4 times higher in pigs fed a combination of chlortetracycline, procaine penicillin, and sulfamethazine during fattening than in those fed an approved growth promoter or a probiotic (10). In our study, there was no difference in Salmonella presence between farms that reported the use of in-feed antimicrobials in grower and finisher pigs and farms that did not report this use. Similarly, no difference in Salmonella presence was observed between farms that reported in-feed use of carbadox, tylosin, or a combination of chlortetracycline, penicillin, and sulfamethazine in weaners or tylosin in growers/finishers and farms that did not report these uses. Carbadox was removed from the Canadian market for use in swine in Canada in 2001, after the end of this study. Experimental studies have shown that adding tylosin and carbadox to feed as growth promoters in swine does not have an effect on Salmonella Typhimurium shedding (27,28).
According to the Canadian Pork Council’s swine industry statistics (29), Alberta had approximately 2200 swine farms in 2000, approximately 75% of which had fewer than 527 pigs per farm; 18% had 527 to 2652 pigs, and 8% had 2653 or more pigs. The last 2 categories contributed 32.6% and 59.2%, respectively, of Alberta’s annual market hog production in 2000. More than 50% of Alberta farms were farrow-to-finish, and most of the grow-to-finish pigs were produced in the Red Deer, Lethbridge, and Drumheller areas (16). The herds participating in this study represented approximately 25% of the annual market pig production in Alberta and were approximately proportionally representative of major Alberta swine-production areas. Of the 16 Alberta veterinary practitioners specializing in swine practice, 9 participated in the study, including all the veterinarians who work exclusively with swine and 2 mixed-animal practitioners. This indicates good coverage of Alberta swine veterinary practices. Thus, the results of this study should be generally representative of swine farms in Alberta marketing at least 2000 pigs per year. However, some caution must be exercised before generalizing the findings beyond these farms because farms producing fewer than 2000 market pigs per year were not included in the study, the farms were not selected randomly, and farrow-to-finish farms were somewhat overrepresented. Although veterinary practitioners were advised to select swine farms from their client lists randomly and not on the basis of a potential problem with Salmonella shedding or clinical disease, it is possible that at least some farms were included for those reasons.
This study confirms that feed composition and structure may be associated with Salmonella presence in pigs. We agree with Funk and Gebreyes (14) that further work is necessary to evaluate the mechanism behind this association.
Acknowledgments
We thank all of the swine veterinary practitioners and producers from Alberta involved in this study and acknowledge the financial support of the Food Safety Division, Alberta Agriculture, Food and Rural Development, and the Government of Canada, through the Western Economic Development Fund.
References
- 1.Fedorka-Cray PJ, Gray JT, Wray C. Salmonella infections in pigsSalmonella. In: Wray C, Wray A, editors. Domestic Animals. London, England: CAB International; 2000. pp. 191–207. [Google Scholar]
- 2.Quantitative assessment of the sources of human salmonellosis attributable to pork. In: Bahnson PB, editor. Proceedings of the 3rd International Symposium on the Epidemiology and Control of Salmonella in Pork; 1999 Aug 4–7; Washington (DC). Urbana Champaign: Biomedical Communications Center, University of Illinois; 1999. pp. 200–205. [Google Scholar]
- 3.Wegener HC, Hald T, Lo Fo Wong D, et al. Salmonella control programs in Denmark. Emerg Infect Dis. 2003;9:774–780. doi: 10.3201/eid0907.030024. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.van der Wolf PJ, Bongers JH, Elbers ARW, et al. Salmonella infections in finishing swine in the Netherlands: bacteriological herd prevalence, serogroup and antibiotic resistance of isolates and risk factors for infection. Vet Microbiol. 1999;67:263–275. doi: 10.1016/s0378-1135(99)00054-1. [DOI] [PubMed] [Google Scholar]
- 5.Nielsen B. Pork safety — a world overview. Proceedings of the 17th Congress of the International Pig Veterinary Society; Ames (IA). 2002 June 2–5; 2002. pp. 121–135. [Google Scholar]
- 6.Rowe TA, Leonard FC, Kelly G, et al. Salmonella serotypes present on a sample of Irish pig farms. Vet Rec. 2003;153:453–456. doi: 10.1136/vr.153.15.453. [DOI] [PubMed] [Google Scholar]
- 7.Hedberg CW. The role of pork as vehicle for confirmed food-borne disease outbreaks in the United Sates. Proceedings of the Pork Quality and Safety Summit; 2002 June 18–20; Des Moines (IA). 2002. pp. 159–166. [Google Scholar]
- 8.Funk J. Pre-harvest food safety diagnostics for Salmonella serovars. Part 1: Microbiological culture. J Swine Health Prod. 2003;11:87–90. [Google Scholar]
- 9.van der Wolf PJ, Wolbers WB, Elbers ARW, et al. Herd level husbandry factors associated with the serological Salmonella prevalence in finishing pig herds in the Netherlands. Vet Microbiol. 2001;78:205–219. doi: 10.1016/s0378-1135(00)00294-7. [DOI] [PubMed] [Google Scholar]
- 10.Leontides LS, Grafanakis E, Genigeorgis C. Factors associated with the serological prevalence of Salmonella enterica in Greek finishing swine herds. Epidemiol Infect. 2003;131:599–606. doi: 10.1017/s0950268803008732. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Lo Fo Wong DMA, Dahl J, Stege H, et al. Herd-level risk factors for subclinical Salmonella infection in European finishing-pig herds. Prev Vet Med. 2004;62:253–266. doi: 10.1016/j.prevetmed.2004.01.001. [DOI] [PubMed] [Google Scholar]
- 12.Beloeil PA, Fravalo P, Fablet C, et al. Risk factors for Salmonella enterica subsp. enterica shedding by market-age pigs in French farrow-to-finish herds. Prev Vet Med. 2004;63:103–120. doi: 10.1016/j.prevetmed.2004.01.010. [DOI] [PubMed] [Google Scholar]
- 13.Nollet N, Maes D, De Zutter L, et al. Risk factors for the herd level bacteriologic prevalence of Salmonella in Belgian slaughter pigs. Prev Vet Med. 2004;65:63–75. doi: 10.1016/j.prevetmed.2004.06.009. [DOI] [PubMed] [Google Scholar]
- 14.Funk J, Gebreyes WA. Risk factors associated with Salmonella prevalence on swine farms. Swine Health Prod. 2004;12:246–251. [Google Scholar]
- 15.Risk factors associated with the presence of Salmonella in swine herds in Quebec. In: Bahnson PB, editor. Proceedings of the 3rd International Symposium on the Epidemiology and Control of Salmonella in Pork; 1999 Aug 4–7; Washington (DC). Urbana Champaign: Biomedical Communications Center, University of Illinois; 1999. pp. 165–168. [Google Scholar]
- 16.Alberta Agriculture, Food and Rural Development. [(accessed June 8, 2005)];Alberta’s Pork Production Industry. 2003 Available at: www1.agric.gov.ab.ca/ $department/deptdocs.nsf/all/agdex3016.
- 17.Rajiæ A, Keenliside J, McFall ME, et al. Longitudinal study of Salmonella species in 90 Alberta swine finishing farms. Vet Microbiol. 2005;105:47–56. doi: 10.1016/j.vetmic.2004.10.005. [DOI] [PubMed] [Google Scholar]
- 18.Sorensen O, Van Donkersgoed J, McFall M, Manninen K, Gensler G, Ollis G. Salmonella spp. shedding by Alberta beef cattle and the detection of Salmonella spp. in ground beef. J Food Prot. 2002;65:484–491. doi: 10.4315/0362-028x-65.3.484. [DOI] [PubMed] [Google Scholar]
- 19.Dohoo I, Martin W, Stryhn H. Veterinary Epidemiology Research. Charlottetown (PEI): AVC; 2003. pp. 335–372. [Google Scholar]
- 20.Hoffman J. Generalized Linear Models — an Applied Approach. Boston (MA): Pearson Education; 2004. pp. 149–167. [Google Scholar]
- 21.The effect of feeding pellets, meal and heat treatment on the Salmonella prevalence of finishing pigs. In: Bahnson PB, editor. Proceedings of the 3rd International Symposium on the Epidemiology and Control of Salmonella in Pork; 1999 Aug 4–7; Washington (DC). Urbana Champaign: Biomedical Communications Center, University of Illinois; 1999. pp. 308–312. [Google Scholar]
- 22.The effect of feeding non-heat treated, non-pelleted feed compared to feeding pelleted, heat-treated feed on Salmonella prevalence of finishing pigs. In: Bahnson PB, editor. Proceedings of the 3rd International Symposium on the Epidemiology and Control of Salmonella in Pork; 1999 Aug 4–7; Washington (DC). Urbana Champaign: Biomedical Communications Center, University of Illinois; 1999. pp. 313–316. [Google Scholar]
- 23.Davies PR, Morrow WEM, Jones FT, Deen J, Fedorka-Cray PJ, Harris IT. Prevalence of Salmonella in finishing swine raised in different production systems in North Carolina, USA. Epidemiol Infect. 1997;119:237–244. doi: 10.1017/s095026889700784x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Funk JA, Davies PR, Gebreyes WA. Risk factors associated with Salmonella enterica prevalence in three-site production systems in North Carolina, USA. Berl Munch Tierarztl Wochenschr. 2001;114:335–338. [PubMed] [Google Scholar]
- 25.Spread of Salmonella infections in pens and between pens. In: Monetti PG, Vignola G, editors. Proceedings of the 14th International Pig Veterinary Society Congress; Bologna, Italy. 1996 July 7–10; Bologna, Italy: Press Point Abbiategrasso; 1996. p. 172. [Google Scholar]
- 26.Davies PR, Morrow WEM, Jones FT, Deen J, Fedorka-Cray PJ, Grey JT. Risk of shedding Salmonella organisms by market-age hogs in a barn with open-flush gutters. J Am Vet Med Assoc. 1997;210:386–389. [PubMed] [Google Scholar]
- 27.Baggesen DL, Wingstrand A, Carstensen B, Nielsen B, Aarestrup FM. Effects of the antimicrobial growth promoter tylosin on subclinical infection of pigs with Salmonella enterica serotype Typhimurium. Am J Vet Res. 1999;60:1201–1205. [PubMed] [Google Scholar]
- 28.Ebner PD, Mathew AG. Effects of antibiotic regimens on the fecal shedding patterns of pigs infected with Salmonella Typhimurium. J Food Prot. 2000;63:709–714. doi: 10.4315/0362-028x-63.6.709. [DOI] [PubMed] [Google Scholar]
- 29.Canadian Pork Council. [(accessed 2004 May 7)];Statistics. Available: www.cpc-ccp.com/stats.html.