Analysis of Risk Factors Associated with Salmonella spp. Isolated from U.S. Feedlot Cattle

Abstract

Contamination can occur at a number of stages during farm-to-fork processing. Preharvest intervention is an ongoing area of interest in reduction of risk of foodborne illness. This study examined risk factors associated with detection of Salmonella from cattle in U.S. feedlots. During two visits to 73 feedlots, 25 fresh fecal samples were collected from each of three pen floors. Associations between management and demographic factors and culture status were evaluated using logistic regression. Factors positively associated with culture-positive status included brewers' grains (odds ratio [OR] = 26.35; confidence interval [CI], 10.33–67.20), corn gluten (OR = 10.35; CI, 5.98–17.91), or cottonseed hulls (OR = 8.34; CI, 3.58–19.42) in the ration, and sourcing animals in a pen from multiple herds of origin (OR = 5.17; CI, 2.32–11.51). Factors negatively associated with positive culture status included urea (OR = 0.27; CI, 0.16–0.44), alfalfa, clover, or sorghum silage (OR = 0.31; CI, 0.12–0.79), and antimicrobials of the tetracycline class in the ration (within 2 weeks before sampling, OR = 0.04 and CI, 0.02–0.09; more than 2 weeks before sampling, OR = 0.23 and CI, 0.06–0.80). Since 18.3% of positive samples were on a single operation, a second model was constructed after excluding data from this operation. Three additional variables were retained in the second model, including grain-processing method (OR for dry roll, cracked, or unprocessed grain = 2.99; CI, 1.55–5.75), soybean meal (OR = 2.74; CI, 1.58–4.75), and use of a coccidiostat in the ration (OR for no coccidiostat = 4.50; CI, 2.03–10.01). Considering the increasing use of by-products of the biofuel industry as feeds, further investigation of the association between feeding brewers' grains and corn gluten and Salmonella recovery is warranted.

Introduction

S poradic Salmonella infections have been estimated to account for approximately 1.4 million cases of illness, 15,000 hospitalizations, and 400 deaths annually in the United States (Voetsch et al., 2004). Annual cost estimates range from $0.5 to $2.3 billion (Frenzen et al., 1999; Mead et al., 1999). Infection may develop after consumption of Salmonella-contaminated animal-origin foods or fruits and vegetables. Contamination may occur during production, transport, slaughter, processing, storage, retail, and food preparation (Hogue et al., 1998).

To reduce the likelihood of carcass contamination with Salmonella and other foodborne pathogens, numerous intervention strategies have been implemented in harvest facilities (Reed, 1995). Additionally, there is ongoing interest in preharvest strategies for reducing pathogen load in the gastrointestinal tracts or on the hides of animals (Attenborough and Matthews, 2000; Stephens et al., 2007). To facilitate preharvest intervention, additional understanding of the complex distribution of these pathogens in the feedlot setting is needed.

Salmonella transmission to cattle on-farm can occur in many ways. Feed ingredients such as corn, hay, silage, cottonseed, and other additions may be contaminated by wild birds (Kapperud and Rosef, 1983; Malmqvist et al., 1995; Refsum et al., 2003), mammalian carriers (Davies and Wray, 1995; Malmqvist et al., 1995; Letellier et al., 1999), or surface-water runoff while in the field (Vaessen et al., 1998). Contamination may occur during processing, transport (Fedorka-Cray et al., 1997; Dargatz et al., 2005), or on-site. Ingestion of contaminated surface water (Fossler et al., 2005b) or water from contaminated troughs (Branham et al., 2005) can also result in transmission. Additionally, animal-to-animal transmission may occur (Khaitsa et al., 2007), and management factors are likely to play a role (Fossler et al., 2005b). Knowledge of risk factors affecting Salmonella shedding can assist in planning preharvest strategies for reducing pathogen loads. Contributing to the epidemiologic evidence for relationships between dietary, social, physical, and other feedlot-related characteristics and Salmonella prevalence was an objective of this national study.

Materials and Methods

A stratified random sample of feedlots with ≥1000-head capacity in 12 major cattle-feeding states (Arizona, California, Colorado, Idaho, Iowa, Kansas, Nebraska, New Mexico, Oklahoma, South Dakota, Texas, and Washington) was selected to participate in the study (Dargatz et al., 2003). Data were collected on 520 operations via personal interview from August 16 through September 22, 1999. These states accounted for 95.8% of the cattle on feed in lots with ≥1000-head capacity in the U.S. as of January 1, 1999.

From the participating feedlots a convenience sample of 73 feedlots, based on laboratory capacity, was selected for fecal sample collection without regard to Salmonella infection status. Veterinary medical officers visited each feedlot once in the period from October 1999 through March 2000, and again in the period from April 2000 through September 2000. Data collectors were instructed to visit each feedlot approximately 6 months apart. At each visit, the pens of cattle that had been at the feedlot the shortest and the longest amounts of time and a randomly chosen pen were selected for sampling to investigate the frequency of recovery of Salmonella at different stages of feeding. For each pen of cattle sampled, health and management data (Tables 1 and 2) were provided by feedlot managers. Ingredient data were collected for the ration fed at the time of sampling.

Table 1.

Percentage of Samples from Which Salmonella Was Recovered by Statistically Significant Feedlot and Pen-Level Management Variables (All Feedlots)

Variable	Description	p-Value	Number of samples	Number positive	Percent positive	Number negative	Percent negative
Season		<0.01
	Jan–March		3025	86	2.8	2939	97.2
	April–June		2930	205	7.0	2725	93.0
	July–Sept		2450	286	11.7	2164	88.3
	Oct–Dec		1838	73	4.0	1765	96.0
Region		<0.01
North	IA, ID, NE, SD, WA		5138	248	4.83	4890	95.2
South	CA, CO, KS, NM, OK, TX		5105	402	7.87	4703	92.1
Size (one-time feedlot capacity)		<0.01
	Less than 8000 head		5370	418	7.78	4952	92.22
	8000 head or more		4873	232	4.76	4641	95.24
Any dairy animals in pen		0.01
	Yes		924	112	12.1	812	87.90
	No		9319	538	5.8	8781	94.2
Pen density category		0.09
	Less than 167 sqft per animal		2551	215	8.40	2336	91.6
	167 or more sqft per animal		7442	389	5.20	7053	94.8
Mass treatment with injectable abx (anytime before sampling)		<0.01
	Yes		1110	121	10.9	989	89.1
	No or DK^a		9133	529	5.8	8604	94.2
Cumulative percent sick category		0.09
	0 to less than 3 pct sick in pen		5406	274	5.1	5132	94.9
	3 or more pct sick in pen		4763	376	7.9	4387	92.1
Any in pen died since arrival		0.21
	Yes		4101	201	4.9	3900	95.1
	No		6142	449	7.3	5693	92.7
Single herd of origin		0.03
	Yes		1903	62	3.3	1841	96.7
	No or DK^b		8340	588	7.1	7752	93.0
Water trough cleaning category		0.07
	Cleaned less than 1 week ago		4823	259	5.4	4564	94.6
	Cleaned 1 week or more ago		5105	390	7.6	4715	92.4

Includes 149 samples without data.

Includes 941 samples without data.

pct, percent.

Table 2.

Percentage of Samples from Which Salmonella Was Recovered by Statistically Significant Feedlot and Pen-Level Feed and Water-Related Variables (All Feedlots)

Variable	Description	p-Value	Number of samples	Number positive	Percent positive	Number negative	Percent negative
How grain is processed		0.23
	Steam flaked, ground high moisture (high-moisture grain)		3107	179	5.80	2928	94.2
	Dry roll, cracked, unprocessed whole grain, other (dry grain)		6827	460	6.74	6367	93.3
Barley in ration		<0.01
	Yes		375	1	0.3	374	99.7
	No		9743	648	6.7	9095	93.4
Brewers' grains in ration		<0.01
	Yes		590	183	31.0	407	69.0
	No		9528	466	4.9	9062	95.1
Corn in ration		0.20
	Yes		9461	582	6.2	8879	93.9
	No		657	67	10.2	590	89.8
Whole wheat		<0.01
	Yes		872	9	1.0	863	99.0
	No		9246	640	6.9	8606	93.1
Wheat fines/mids		<0.01
	Yes		499	140	28.1	359	71.9
	No		9619	509	5.3	9110	94.7
Coccidiostat (Amprolium, Decoquinate) in pen ration		0.21
	Yes		748	22	2.9	726	97.1
	No		9370	627	6.7	8743	93.3
Canola meal		0.18
	Yes		175	22	12.6	153	87.4
	No or DK^a		9943	627	6.3	9167	93.7
Chicken waste		<0.01
	Yes		25	1	4.0	24	96.0
	No or DK^a		10093	648	6.4	9445	93.6
Cottonseed whole		0.04
	Yes		491	11	2.2	480	97.8
	No or DK^a		9627	638	6.6	8989	93.4
Cottonseed meal		<0.01
	Yes		1538	165	10.7	1373	89.3
	No or DK^a		8580	484	5.6	8096	94.4
Soybean meal		<0.01
	Yes		2762	96	3.5	2666	96.5
	No or DK^a		7356	553	7.5	6803	92.5
Urea		<0.01
	Yes		6086	201	3.3	5885	96.7
	No or DK^a		4032	448	11.1	3584	88.9
Other proteins		<0.01
	Yes		1613	233	14.5	1380	85.6
	No or DK^b		8505	416	4.9	8089	95.1
Alfalfa, clover, or sorghum hay		<0.01
	Yes		7600	423	5.6	7177	94.4
	No		2518	226	9.0	2292	91.0
Cottonseed hulls		<0.01
	Yes		1059	184	17.4	875	82.6
	No		9059	465	5.1	8594	94.9
Alfalfa, clover, or sorghum silage		<0.01
	Yes		1492	25	1.7	1467	98.3
	No		8626	624	7.2	8002	92.8
Corn gluten		<0.01
	Yes		2900	278	9.6	2622	90.4
	No		7218	371	5.1	6847	94.9
Beet pulp		<0.01
	Yes		325	4	1.2	321	98.8
	No		9793	645	6.6	9148	93.4
Tallow		<0.01
	Yes		2404	91	3.8	2313	96.2
	No		7714	558	7.2	7156	92.8
Potato Waste		<0.01
	Yes		774	2	0.3	772	99.7
	No		9344	647	6.9	8697	93.1
Other byproducts		<0.01
	Yes		2482	262	10.6	2220	89.4
	No		7636	387	5.1	7249	94.9
Any bacitracin use in feed or water		N/A
	Yes		0	0	N/A
	No		10243	650	6.4	9593	93.7
Chlortetracycline in feed or water within 2 weeks before sampling		<0.01
	Yes, within 2 weeks		715	5	0.7	710	99.3
	Yes, but not within 2 weeks		1065	35	3.3	1030	96.7
	No, never		8463	610	7.2	7853	92.8
Chlortet/sulfa in feed or water within 2 weeks before sampling		<0.01
	Yes, within 2 weeks		134	5	3.7	129	96.3
	Yes, but not within 2 weeks		390	1	0.3	389	99.7
	No, never		9719	644	6.6	9075	93.4
Oxytet in feed or water within 2 weeks before sampling		<0.01
	Yes, within 2 weeks		299	14	4.7	285.0	95.3
	Yes, but not within 2 weeks		574	105	18.3	469	81.7
	No, never		9370	531	5.7	8839.0	94.3
Any tetracycline use (chlortet, chlortet/sulfa, tet, oxytet)		0.08
	Yes, within 2 weeks		1223	24	2.0	1199	98.0
	Yes, but not within 2 weeks		2004	140	7.0	1864	93.0
	No, never		7016	486	6.9	6530	93.1
Any virginiamycin use in feed or water		N/A
	Yes		0	0	N/A
	No		10243	650	6.4	9593	93.7

Includes 149 samples without data.

Includes 74 samples without data.

N/A, not available.

Within each pen 25 fresh fecal samples (approximately 25 g) were collected from individual pats off pen floors as previously described (Dargatz et al., 2003). Collection of 25 samples per pen allowed 95% confidence of detecting at least one positive animal if the within pen prevalence was 10% or more, assuming 100% test sensitivity and specificity. Separate tongue depressors were used to collect each sample from the top of a fecal pat into a whirlpak bag, shipped with ice pack for overnight delivery to a single laboratory for immediate culture upon receipt. Data and sample collectors received training before the study to standardize the collection process.

Laboratory techniques

Each sample was evaluated by bacteriologic culture for Salmonella using methods described previously (Wells et al., 2001). Approximately 1 g of feces from each sample was incubated in 9 mL each of GN Hajna broth (Difco Laboratories, Detroit, MI) and tetrathionate broth (Difco Laboratories) at 37°C for 24 and 48 h, respectively. Then, a 100 μL aliquot of each culture was transferred to Rappaport Vassiladis broth R-10 (Difco Laboratories), incubated overnight, and then streaked to brilliant green agar with sulfadiazine (Difco Laboratories) and xylose–lysine tergitol 4 agar (Difco Laboratories). Plates were incubated for 18–24 h at 37°C. Up to four colonies having the typical appearance of Salmonella (white to red opaque colonies surrounded by red zones on brilliant green agar with sulfadiazine, or pink to red colonies with a black center on xylose–lysine tergitol 4 agar) were inoculated into agar slants for biochemical confirmation. Presumptive positive isolates were serogrouped using serogroup specific antisera (Difco Laboratories) and were sent to the National Veterinary Services Laboratory (Ames, IA) for serotyping.

Statistical techniques

All analysis accounted for sampling design, nesting of pens within feedlots, and lack of independence within pens. Continuous variables were graphed to identify natural cut points for creation of categories. When these were not clear, the data were divided into quartiles or halves for univariate analysis (Table 1). The association between factors and Salmonella culture status was evaluated using a chi-square test (SUDAAN version 9.0.1, CROSSTAB procedure). Variables were considered for inclusion in a multivariable logistic regression model (SUDAAN version 9.0.1, LOGISTIC procedure) if the chi-square p-value was <0.25. A backward elimination approach was utilized until all variables remaining in the model were significantly associated with the outcome (p < 0.05). Biologically plausible two-way interactions among remaining main effect variables were examined.

Since 18.3% (119/149) of positive samples were on a single feedlot, a second model was constructed after excluding the 149 samples from this feedlot. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated for variables in the models. To assess the fit of each model, sensitivity and specificity of the model-predicted outcomes were assessed by treating the observed status as the gold standard. A probability cut point of ≥0.5 was used for predicted outcome; samples with a predicted probability of ≥50% that they were positive were classified as positive.

Results

Descriptive results from the study have been reported elsewhere (Dargatz et al., 2003). For the first model, 10,243 fecal samples from 415 pens from 73 feedlots were included. Salmonella was isolated from 650 samples (6%) from 92 pens (22%) on 37 feedlots (51%). The five most common serotypes included Anatum (n = 195), Montevideo (n = 127), Reading (n = 69), Newport (n = 63), and Kentucky (n = 57). For the second model, 10,094 fecal samples from 409 pens were included. Salmonella was isolated from 531 samples (5.3%) from 86 pens (21.0%) on 36 feedlots (50.0%).

Risk factors

Of the 55 variables considered as potential risk factors associated with detection of Salmonella-positive fecal samples (Tables 1 and 2), 30 were eligible for entry into the first model; 4 variables representing the use of chlortetracycline, chlortetracycline/sulfamethazine, oxytetracycline, and tetracycline in feed or water were combined into a single variable, so 26 candidate variables were evaluated. In addition to tetracycline use, other variables that met the screening criterion included season, region, feedlot size, pen density, mass treatment of cattle in pen with injectable antibiotics, any dairy cattle in pen, percent sick in pen, any deaths in pen, single herd of origin, frequency of water trough cleaning, grain processing method, and the presence or absence of specific ingredients in pen rations, including soybean meal; cottonseed; cottonseed meal; urea; other proteins; brewer's grains; whole wheat; a coccidiostat; alfalfa, clover, or sorghum hay; cottonseed hulls; alfalfa, clover, or sorghum silage; corn gluten; tallow; and other byproducts. Variables for sex of the predominant number of animals in pen, sourcing any cattle in pen from salebarn, entry weight, average weight of cattle in pen at sampling, days on feed at time of sampling, remaining days on feed at time of sampling, whether animals in the pen were dewormed, distance traveled to feedlot, and the ration ingredients and/or additives including sorghum, other concentrates, ionophore, corn silage, other roughage, and tylosin in feed or water were not eligible for model entry due to a chi-square p > 0.25. Variables for use of bacitracin, neomycin, or virginiamycin could not be analyzed because these antimicrobials were not utilized. Use of sulfa antimicrobials could not be analyzed owing to their use on a single operation. The ration variables for wheat fines/mids, chicken waste, barley, potato waste, and beet pulp were excluded from the model because of sparse data, as each of these ingredients was used on less than 10% of feedlots and in only one region. The ration variable for corn was excluded from the model due to sparse data.

Fifteen variables remained in the multivariable model using all observations (Table 3). Samples collected in all other seasons were more likely to have Salmonella detected compared to the January through March season. Samples collected from feedlots in the southern region were also more likely to have Salmonella detected (OR = 3.23; CI, 1.57–6.64) as were samples collected from feedlots with less than 8000-head capacity (OR = 7.99; CI, 3.60–17.71). Other variables that were positively associated with the detection of Salmonella included whether cattle in a pen were from multiple herds of origin (OR = 8.37; CI, 4.22–16.60), and inclusion of cottonseed hulls (OR = 5.19; CI, 1.87–14.41), corn gluten (OR = 7.11; CI, 3.96–12.79), other protein sources (OR = 3.10; CI, 1.78–5.38), brewer's grains (OR = 26.85; CI, 9.89–72.86), or hay (OR = 2.56; CI, 1.28–5.09) in the ration. Factors associated with decreased risk of Salmonella detection included use of a tetracycline-class antimicrobial in feed or water within 2 weeks before fecal sampling (OR = 0.08; CI, 0.02–0.32), inclusion of urea (OR = 0.46; CI, 0.30–0.69), and use of alfalfa, clover, or sorghum silage (OR = 0.35; CI, 0.15–0.83) in the pen ration.

Table 3.

All-Operation and Second Logistic Regression Models for Risk Factors Associated with Isolation of Salmonella spp.

Variable	Categories	All operations odds ratio	All operations 95% confidence interval	Final odds ratio	Final 95% confidence interval
Season	Jan–March	1.0		Not in model
	Apr–June	3.69	1.76–7.77
	July–Sept	4.04	2.09–7.81
	Oct–Dec	4.39	1.81–10.64
Region	North (IA,ID, NE, SD, WA)	1.0		Not in model
	South (CA, CO, KS, NM, OK, TX)	3.23	1.57–6.64
Feedlot size (one-time capacity)	Less than 8000 head	7.99	3.60–17.71	Not in model
	8000 head or more	1.0
Single herd of origin	Yes	0.12	0.06–0.24	0.19	0.09–0.43
	No or DK	1.0		1.0
Grain processing method	steam flaked or ground high moisture corn	Not in model		1.0
	dry roll, cracked or unprocessed whole grain			2.99	1.55–5.75
Soy in pen ration	Yes	Not in model		2.74	1.58–4.75
	No
Urea in pen ration	Yes	0.46	0.30–0.69	0.27	0.16–0.44
	No or DK	1.0		1.0
Other protein in pen ration (includes but not limited to sunflower products, feather meal unspecified plant protein, and alfalfa)	Yes	3.10	1.78–5.38	Not in model
	No or DK	1.0
Cottonseed hulls in pen ration	Yes	5.19	1.87–14.41	8.34	3.58–19.42
	No or DK	1.0		1.0
Alfalfa, clover, or sorghum silage in pen ration	Yes	0.35	0.15–0.83	0.31	0.12–0.79
	No or DK	1.0		1.0
Hay in pen ration	Yes	2.56	1.28–5.09	Not in model
	No or DK	1.0
Corn gluten in pen ration	Yes	7.11	3.96–12.79	10.35	5.98–17.91
	No	1.0		1.0
Brewers grains/malt in pen ration	Yes	26.85	9.89–72.86	26.35	10.33–67.20
	No or DK	1.0		1.0
Whole wheat in pen ration	Yes	0.11	0.03–0.35
	No or DK	1.0		Not in model
Cottonseed meal in pen ration	Yes	4.10	2.07–8.12	Not in model
	No or DK	1.0
Tallow in pen ration	Yes	0.37	0.16–0.88
	No or DK	1.0		Not in model
Coccidiostat (Amprolium, Decoquinate) in pen ration	Yes	Not in model		1.0
	No	4.50	2.03–10.01
Tetracycline-class antimicrobials in pen ration	Yes, within 2 weeks before sampling	0.08	0.02–0.32	0.04	0.02–0.09
	Yes, but not before sampling	0.47	0.20–1.10	0.23	0.06–0.80
	No	1.0		1.0

After removing the data for the feedlot with 18.3% of the positive samples, a slightly different model resulted (Table 3). Seven of the variables from the model with the full dataset were retained in the second model. These included having a single herd of origin for cattle in a pen and the following ration ingredients: urea; cottonseed hulls; alfalfa, clover, or sorghum silage; corn gluten; brewers' grains or malt; and tetracycline-class antimicrobials. Three additional variables were added to the model: grain processing method, soybean meal, and use of a coccidiostat in rations. Biologically plausible two-way interactions were examined. Unfortunately, since all of these variables except one were feed related, interaction terms resulted in sparse data for some of the combinations of variables, so these interactions could not be analyzed.

Sensitivity and specificity for the complete dataset model were 18% and 100%, respectively. Within this population, the positive predictive value for the model was 70% and the negative predictive value for the model was 95%. For the second model, sensitivity and specificity were 20% and 99%, respectively. The positive predictive value was approximately 54% and the negative predictive value was approximately 96%.

Discussion

Other research findings are consistent with an increase in likelihood of Salmonella detection in samples from pens with cattle from more than one herd of origin. Change in social groups has been shown to increase fecal shedding of Salmonella in production animals (Callaway et al., 2006); this may be due to the stress that accompanies reestablishing a dominance hierarchy (Morrow-Tesch et al., 1994). Another explanation for this finding is the transmission of Salmonella to previously uninfected cattle once commingling occurs (Khaitsa et al., 2007). This possibility of transmission is consistent with the increase in percent positive samples seen as length of time on feed increased. The potentially stress-related increase in Salmonella in samples from pens with cattle from more than one herd of origin may also occur in other situations where social groups of cattle are mixed, such as in transport to slaughter (Reicks et al., 2007).

In both models, the inclusion of cottonseed hulls in the pen rations was associated with an increase in likelihood of Salmonella detection. Cottonseed-component feeds sampled on the farm (Davis et al., 2003) and at the processor (McChesney et al., 1995) have tested positive for Salmonella in several studies. Feeding whole cottonseed or cottonseed hulls has been associated with increased risk of Salmonella in a previous National Animal Health Monitoring System (NAHMS) feedlot study (Losinger et al., 1997). Physiologically, when fat-supplemented diets are fed, ruminal pH tends to be greater, and volatile fatty acid (VFA) concentration is generally lower (Elliott et al., 1999). Increased production of VFA may more effectively inhibit growth of Salmonella (Mattila et al., 1988), whereas decreased production of ruminal VFAs may be less likely to inhibit growth.

In both models, corn gluten and brewers' grains in pen rations were associated with an increase in likelihood of Salmonella detection. Although we did not differentiate between wet and dry corn gluten feed, total ruminal VFAs have been shown to decrease linearly with increasing levels of wet corn gluten feed (Sindt et al., 2002), and growth of Salmonella may be inhibited in the presence of high concentrations of VFA (Mattila et al., 1988). Feeding brewers' grains was associated with an increased risk of Salmonella detection in fecal samples from the NAHMS Dairy’96 study (Kabagambe et al., 2000). Wet brewers' grains have a high moisture content that may promote Salmonella growth (Preston, 1998). Due to the magnitude of association between these variables and the outcome variable, more work is indicated to understand risk associated with feeding by-products from the biofuel industry. A recent study found no association between feeding wet corn distillers' grains and pen-level prevalence of Salmonella spp. (Jacob et al., 2008).

In both models, urea and silage in the pen rations were associated with a decrease in likelihood of Salmonella detection, as was the use of tetracyclines in feed and water. Urea's protective effects may be due to replacement of another protein component with greater likelihood of contamination. Well-fermented silage has a high lactic acid content; lactic acid has an antimicrobial effect on Salmonella in vitro (De Keersmaecker et al., 2006). In a Netherlands dairy study, grass supplemented with silage was shown to have an apparent protective effect against Salmonella Dublin infection when compared with grass-only feed (Vaessen et al., 1998). While tetracycline use may be associated with a decreased prevalence of salmonellae, prudent antimicrobial use is advised (Fossler et al., 2005a). Although tetracycline use may decrease the overall prevalence of salmonellae, those that remain may be more resistant to tetracyclines. This is of concern not only from animal health and food safety perspectives but perhaps from an ecologic perspective as well.

Factors that did not remain in the second model included season, region, and feedlot size, as well as other protein, hay, whole wheat, cottonseed meal, and tallow in the pen rations.

Season has been associated with Salmonella shedding in dairy cattle in a number of studies; shedding tends to be more common in the summer months (Evans, 1996; Kabagambe et al., 2000; Wells et al., 2001; Fossler et al., 2005a; Davison et al., 2006). Additionally, survival of Salmonella in the environment may be increased in the summer months, although soil composition, moisture, and temperature fluctuation appear to play a role (Holley et al., 2006; Semenov et al., 2007). While climate may play a role in Salmonella shedding, and heat stress may be a factor that contributes to increased shedding, the two regions that were compared during the study both included plains states with somewhat similar climates.

Factors present in the second model that were not significant in the full dataset model included grain-processing method, soybean meal, and a coccidiostat in pen ration. That grain-processing method stayed in the second model is consistent with findings from a Colorado feedlot study in which a higher percentage of dry-corn samples than high-moisture corn samples tested positive for Salmonella bacteria (4.0% vs. 0.6%, respectively) (Dargatz et al., 2005). Generally, high-moisture corn is more digestible than dry-rolled corn (Ladely et al., 1995; Archibeque et al., 2006). As noted above, higher digestibility can result in increased production of VFAs, which may more effectively inhibit growth of Salmonella. Soybean-meal feed samples have tested positive for Salmonella in several feed-sample studies (McChesney et al., 1995; Kidd et al., 2002).

Coccidiosis is an opportunistic disease that can develop in conjunction with other stress factors. As both Salmonella shedding and coccidiosis tend to be associated with physiologic stress, the presence of both agents may act synergistically to increase excretion of Salmonella; therefore, the use of a coccidiostat may play a protective role. Such an effect has been demonstrated experimentally in poultry (Kosugi et al., 1986). An alternative explanation is that the protective effect of coccidiostat use seen in the second model may be a reflection of other management differences between pens with Salmonella and those without. Because ionophores are labeled in the use of coccidiosis, one might expect to see a similar effect of ionophores in rations, but an in vitro study suggests that ionophore feeding has minimal effect on ruminant Salmonella populations (Edrington et al., 2003).

The low sensitivity and positive predictive value of both models indicate that other factors than those examined here are involved in predicting the presence of Salmonella in feedlot pen samples. The lack of sensitivity may be due to the presence of other factors that were not measured. Also, it is possible that the sensitivity of the sampling and testing procedure resulted in misclassification of truly positive samples (and pens). For example, testing environmental samples from each pen may have been a more efficient way to identify fecal culture-positive pens than using fresh samples, which were generally representative of individual animals (Warnick et al., 2003). Misclassification would make it more difficult to identify factors associated with the presence of Salmonella but would increase the specificity of the model. Pens in each feedlot were sampled once in the period from October 1999 through March 2000 and once in the period from April through September 2000, intervals during which a number of pen-level changes can occur, whereas a longitudinal approach with more frequent sampling might be of higher sensitivity. The variability in digestibility of particular ration components may play a role in Salmonella replication. Data on previous exposure to various ration components was not collected. As such, previous exposures to those that have longer-term effects would not be detected in this analysis. Additionally, recent dietary changes would not be detected in this analysis. Gastrointestinal flora can be affected by many factors, including dietary changes and other stressors (Rostagno et al., 2009). Further study is needed to understand the roles of various management and feed-related factors in the shedding of Salmonella in feedlot cattle.

Footnotes

Disclosure Statement

No competing financial interests exist.

References

Archibeque

, Miller

, Freetly

et al. Feeding high-moisture corn instead of dry-rolled corn reduces odorous compound production in manure of finishing beef cattle without decreasing performance. J Anim Sci, 2006; 84:1767–1777.

Attenborough

, Matthews

. Food safety through the meat supply chain. Symp Ser Soc Appl Microbiol, 2000; 29:144S–148S.

Branham

, Carr

, Scott

et al. E. coli O157 and Salmonella spp. in white-tailed deer and livestock. Curr Issues Intest Microbiol, 2005; 6:25–29.

Callaway

, Morrow

, Edrington

et al. Social stress increases fecal shedding of Salmonella typhimurium by early weaned piglets. Curr Issues Intest Microbiol, 2006; 7:65–71.

Dargatz

, Fedorka-Cray

, Ladely

et al. Prevalence and antimicrobial susceptibility of Salmonella spp. isolates from U.S. cattle in feedlots in 1999 and 2000. J Appl Microbiol, 2003; 95:753–761.

Dargatz

, Strohmyer

, Morley

et al. Isolation of Escherichia coli and Salmonella enterica from feed ingredients in cattle feedlots and their antimicrobial susceptibility patterns. Foodborne Path Dis, 2005; 2:341–347.

Davies

, Wray

. Mice as carriers of Salmonella enteritidis on persistently infected poultry units. Vet Rec, 1995; 137:337–341.

Davis

, Hancock

, Rice

et al. Feedstuffs as a vehicle of cattle exposure to Escherichia coli O157:H7 and Salmonella enterica. Vet Micro, 2003; 95:199–210.

Davison

, Sayers

, Smith

et al. Risk factors associated with the salmonella status of dairy farms in England and Wales. Vet Rec, 2006; 159:871–880.

10.

De Keersmaecker

, Verhoeven

, Desair

et al. Strong antimicrobial activity of Lactobacillus rhamnosus GG against Salmonella Typhimurium is due to accumulation of lactic acid. FEMS Microbiol Lett, 2006; 259:89–96.

11.

Edrington

, Callaway

, Varey

et al. Effects of the antibiotic ionophors monensin, lasalocid, laidlomycine propionate and bambermycin on Salmonella and E. coli O157:H7 in vitro. J Appl Microbiol, 2003; 94:207–213.

12.

Elliott

, Drackley

, Beaulieu

et al. Effects of saturation and esterification of fat sources on site and extent of digestion in steers: digestion of fatty acids, triglycerides, and energy. J Anim Sci, 1999; 77:1919–1929.

13.

Evans

. A case control study of multiple resistant Salmonella Typhimurium DT104 infection of cattle in Great Britain. Cattle Pract, 1996; 4:259–266.

14.

Fedorka-Cray

, Hogg

, Gray

et al. Feed and feed trucks as sources of Salmonella contamination in swine. Swine Health Prod, 1997; 5:189–193.

15.

Fossler

, Wells

, Kaneene

et al. Cattle and environmental sample-level factors associated with the presence of Salmonella in a multi-state study of conventional and organic dairy farms. Prev Vet Med, 2005a. 67:39–53.

16.

Fossler

, Wells

, Kaneene

et al. Herd-level factors associated with isolation of Salmonella in a multi-state study of conventional and organic dairy farms I. Salmonella shedding in cows. Prev Vet Med, 2005b. 70:257–277.

17.

Frenzen

, Riggs

, Buzby

et al. Salmonella cost estimate updated using FoodNet data. Food Rev, 1999; 22:10–15. www.ers.usda.gov/publications/foodreview/may1999. 2005 April 25(Online.)

18.

Hogue

, White

, Heminover

. Pathogen reduction and hazard analysis and critical control point (HACCP) systems for meat and poultry. USDA. Vet Clin North Am Food Anim Pract, 1998; 14:151–164.

19.

Holley

, Arrus

, Ominski

. Salmonella survival in manure-treated soils during simulated seasonal temperature exposure. J Environ Qual, 2006; 35:1170–1180.

20.

Jacob

, Fox

, Narayanan

et al. Effects of feeding wet corn distillers grains with solubles with or without monensin and tylosin on the prevalence and antimicrobial susceptibilities of fecal foodborne pathogenic and commensal bacteria in feedlot cattle. J Anim Sci, 2008; 86:1182–1190.

21.

Kabagambe

, Wells

, Garber

et al. Risk factors for fecal shedding of Salmonella in 91 U.S. dairy herds in 1996. Prev Vet Med, 2000; 43:177–194.

22.

Kapperud

, Rosef

. Avian wildlife reservoir of Campylobacter fetus subsp. jejuni, Yersinia spp., and Salmonella spp. in Norway. Appl Env Microbiol, 1983; 45:375–380.

23.

Khaitsa

, Kegode

, Bauer

et al. A longitudinal study of Salmonella shedding and antimicrobial resistance patterns in North Dakota feedlot cattle. J Food Prot, 2007; 70:476–481.

24.

Kidd

, Rossignol

, Gamroth

. Salmonella and other enterobacteriaceae in dairy cow feed ingredients: antimicrobial resistance in western Oregon. J Env Health, 2002; 64:9–16.

25.

Kosugi

, Baba

, Fukata

et al. Effects of cage contamination with coccidia and salmonella on acute salmonellosis in young chickens. Avian Dis, 1986; 30:313–318.

26.

Ladely

, Stock

, Goedeken

, Huffman

. Effect of corn hybrid and grain processing method on rate of starch disappearance and performance of finishing cattle. J Anim Sci, 1995; 73:360–364.

27.

Letellier

, Messier

, Pare

et al. Distribution of Salmonella in swine in Quebec. Vet Micro, 1999; 67:299–306.

28.

Losinger

, Garber

, Smith

et al. Management and nutritional factors associated with the detection of Salmonella spp. from cattle fecal specimens from feedlot operations in the United States. Prev Vet Med, 1997; 31:231–244.

29.

Malmqvist

, Jacobsson

, Haggblom

et al. Salmonella isolated from animals and feedstuffs in Sweden during 1988–1992. Acta Vet Scand, 1995; 36:21–39.

30.

Mattila

, Frost

, O'Boyle

. The growth of Salmonella in rumen fluid from cattle at slaughter. Epidemiol Infect, 1988; 101:337–345.

31.

McChesney

, Kaplan

, Gardner

. FDA survey determined Salmonella contamination. Feedstuffs, 1995; 13:20–23.

32.

Mead

, Slutsker

, Dietz

et al. Food-related illness and death in the United States. Emerg Infect Dis, 1999; 5:840–842.

33.

Morrow-Tesch

, McGlone

, Salak-Johnson

. Heat and social stress effects on pig immune measures. J Anim Sci, 1994; 72:2599–2609.

34.

Preston

. What's the feed value? Typical composition of feeds for cattle, 1998. Beef, 1998; FC2–FC15.

35.

Reed

. FSIS explains benefits of HACCP. J Am Vet Med Assoc, 1995; 207:284.

36.

Refsum

, Vikoren

, Handeland

et al. Epidemiologic and pathologic aspects of Salmonella Typhimurium infection in passerine birds in Norway. J Wildl Dis, 2003; 38:64–72.

37.

Reicks

, Brashears

, Adams

et al. Impact of transportation of feedlot cattle to the harvest facility on the prevalence of Escherichia coli O157:H7, Salmonella, and total aerobic microorganisms on hides. J Food Prot, 2007; 70:17–21.

38.

Rostagno

. Can stress in farm animals increase food safety risk? Foodborne Pathog Dis, 2009; 6:767–776.

39.

Semenov

, van Bruggen

, van Overbeek

et al. Influence of temperature fluctuations on Escherichia coli O157:H7 and Salmonella enterica serovar Typhimurium in cow manure. FEMS Micobiol Ecol, 2007; 60:419–428.

40.

Sindt

, Drouillard

, Thippareddi

et al. Evaluation of finishing performance, carcass characteristics, acid-resistant E. coli and total coliforms from steers fed combinations of wet corn gluten feed and steam-flaked corn. J Anim Sci, 2002; 80:3328–3335.

41.

Stephens

, Loneragan

, Karunasena

, Brashears

. Reduction of Escherichia coli O157 and Salmonella in feces and on hides of feedlot cattle using various doses of a direct-fed microbial. J Food Prot, 2007; 70:2386–2391.

42.

Vaessen

, Veling

, Frankena

et al. Risk factors for Salmonella Dublin infection on dairy farms. Vet Q, 1998; 20:97–99.

43.

Voetsch

, Van Gilder

, Angulo

et al. FoodNet estimate of the burden of illness caused by nontyphoidal Salmonella infections in the United States. CID, 2004; 38,Suppl 3:S127–S134.

44.

Warnick

, Kaneene

, Ruegg

et al. Evaluation of herd sampling for Salmonella isolation on midwest and northeast US dairy farms. Prev Vet Med, 2003; 60:195–206.

45.

Wells

, Fedorka-Cray

, Dargatz

et al. Fecal shedding of Salmonella spp. by dairy cows on farm and at cull cow markets. J Food Prot, 2001; 64:3–11.