Herd-level risk factors associated with fecal shedding of Shiga toxin-encoding bacteria on dairy farms in Minnesota, USA

Abstract

This study aimed to identify herd-level risk factors associated with fecal shedding of Shiga toxin-encoding bacteria (STB) on dairy cattle farms in Minnesota, USA. After adjustment for farm size, risk factors included: use of total mixed ration (TMR) for lactating dairy cows [odds ratio (OR) = 3.0; 95% confidence interval (CI): 1.8 to 5.1], no use of monensin for weaned calves (OR = 4.8, 95% CI: 2.5, 9.3), and no use of decoquinate for preweaned calves (OR = 2.2, 95% CI: 1.4, 3.6). Fecal shedding of STB was more common in small herds (< 100 cows, OR = 3.6, 95% CI: 2.1, 6.2) than in large herds (≥ 100 cows). Herd management factors related to cattle feeding practices were associated with fecal shedding of STB.

Shiga toxin-producing Escherichia coli (STEC) have emerged as significant foodborne pathogens since the 1980’s. These STEC include the important serotype O157:H7, and more than 100 non-O157 serotypes which have been linked to serious illness in humans, including hemolytic colitis and hemolytic uremic syndrome. Non-O157 STEC may be under-diagnosed as foodborne pathogens compared to serotype O157:H7 which have a non-sorbitol fermenting characteristic that makes it easier to detect. Ruminants such as cattle and sheep are likely major reservoirs of STEC that can be linked to human infections (1).

Currently, there is limited research identifying risk factors associated with fecal shedding of STEC on dairy cattle farms. Some of the cited risk factors include season, herd management, age, animal contact, and stress (2–5). For example, increasing prevalence has been described in spring and summer (3). One study demonstrated that feeding corn silage, grain screens, and ionophores to heifers and feeding animal by-products to adults were associated with a higher prevalence of E. coli O157 in cattle (6). A study of Swiss organic and conventional dairy farms reported that manure contamination of feeds with subsequent mixing and distribution, cross-infection of cows, age of animals, and pasture access were identified in a univariate analysis but no significant differences were found in the multivariate analysis between the 2 types of farms (7).

Most herd-level risk factors for STEC on dairy farms have focused on E. coli O157:H7. Additionally, no consistent significant associations have been identified between shedding of STEC and cattle or farm management systems. Furthermore, few STEC prevalence studies have been documented in Minnesota dairy operations, despite the fact that Minnesota ranks 6th in milk production in the United States (8). We have previously published animal-level factors associated with STEC shedding, which included cattle group (preweaned calf), calf age (29 to 56 days old), and farm type (small organic) (2). The objective of this study was to identify herd-level risk factors associated with fecal shedding of Shiga toxin-encoding bacteria (STB), a surrogate for STEC in dairy cattle.

In Minnesota, 28 Holstein dairy herds were enrolled in this study based on farm type (8 organic and 20 conventional farms) and farm size (based on the number of milking and dry cows). Fecal samples were obtained over a 6-month period as part of a multi-state study for Salmonella and Campylobacter infections in dairy cattle (9). Herd selection criteria required each farm to consist of at least 30 milking or dry cows, have at least 90% of cows of Holstein breed, raise their own calves and heifers for replacement animals, ship milk year round, and express willingness to participate.

Conventional farms within approximately 100 miles of the University of Minnesota St. Paul campus that were willing to participate and met the selection criteria were randomly selected. Organic farms, which were certified by independent organic certification agencies, were included in this study based on the herd selection criteria and their desire to participate. Twenty-eight dairy farms were categorized into 4 herd sizes with 30 to 49, 50 to 99, 100 to 199, and ≥ 200 or more milking and dry cows.

A total of 2039 fecal samples was collected from 28 farms up to 3 times at 2-month intervals from April 2001 to September 2001 (average: 1.9 visits per farm). The target number of samples per visit was estimated to detect Salmonella fecal shedding as described in a previous study (9). Approximately, 30, 40, 50, and 55 cattle fecal samples were collected per visit from herds with 30 to 49, 50 to 99, 100 to 199, and ≥ 200 cows, respectively (9).

On the initial visit to each farm, a questionnaire was administered to the farm owner or manager, and at each sampling visit, important changes to farm management were noted, such as changes in herd inventory or rations fed. The questionnaire sought information on animal inventory, herd size, other animal contact, other animals brought into the operation, housing and milking facilities, outside access, maternity housing, bedding types, feed and water system, use of chemicals, calf management and feeding, farm hygiene and disinfection methods, disease occurrence, and manure management.

One gram of fecal sample was aseptically mixed with EC broth, supplemented with novobiocin (20 μg/mL) and incubated at 37°C for 18 to 24 h. Then a loop of EC enrichment culture was streaked on a sorbitol MacConkey (SMAC) agar plate and incubated overnight. Shiga toxin-encoding bacteria (STB) were detected using a polymerase chain reaction (PCR) colony-sweep assay as previously described (2,10,11).

Odds ratios (OR) and 95% confidence intervals (CI) for herd-level management factors were calculated using a logistic regression model. Microsoft Access 2000 was used for database management and SAS software for Windows (Ver. 8.2; SAS Institute, Cary, North Carolina, USA) was used for data analysis. The outcome variable (dependent variable) in our study was the within-herd prevalence of STB by visit (number of cattle with STB/number of cattle tested within herd) across all ages of cattle sampled.

Univariable association of independent variables was done using a generalized estimating equations (GEE) approach to adjust for the correlation of observations within herds, using the logit link and binomial distribution. “Herd” was used as the subject effect in the repeated statement of SAS (Proc GENMOD, repeated subject = herd, independence correlation structure). Potential explanatory variables were selected based on biological plausibility. Factors which were present on either a few farms (< 10% of total number of farms) or most farms (> 90% of total number of farms) were not included in the analysis. Variables with P < 0.2 on the basis of GEE parameter estimate when controlling for farm size (small, < 100 cows versus large, ≥ 100 cows) were included in a multivariable model. Farm size (small, < 100 cows versus large, ≥ 100 cows) variable was used as a covariate and forced into both the initial univariable and multivariable models to control potential confounding effects. Backward selection was used to fit the final model until all variables remaining in the model had P < 0.05 based on the GEE parameter estimate. Forward selection was also performed to allow us to identify the final model based on convergence of similar variables.

Herds were randomly selected using a stratified sampling process to represent both large and small dairy producers. The average herd size in our study was 105 cows (36.9 in organic herd versus 132.2 in conventional herd), which is larger than the average Minnesota herd (74.9 cows per farm according to USDA National Agriculture Statistical Service). Of the 28 dairy farms enrolled, 17 and 11 were categorized into smaller (< 100 cows) and larger farms (≥ 100 cows), respectively. As a measure of milk production, the mean rolling herd average (RHA) milk production was 19 137 (lb/cow/year) on dairy farms.

Shiga toxin-encoding bacteria (STB) were detected from 50 (2.5%) of 2039 fecal samples with 18 (64.3%) of 28 dairy farms having at least 1 positive animal sample across all visits per farm. Twenty-three (44.2%) of 52 farm visits had at least 1 positive sample. Thirty (1.8%) of 1626 fecal samples were STB positive on conventional farms with 11 (55%) of 20 farms having at least 1 positive animal. Twenty (4.8%) of 413 fecal samples from organic farms were STB positive and 7 (87.5%) of 8 farms had at least 1 positive animal. On smaller farms (< 100 cows), 34 (3.7%) of 920 samples were STB positive with 11 of 17 (64.7%) of these farms having at least 1 positive sample. On larger farms (≥ 100 cows), 16 (1.4%) of 1119 samples were positive with 7 of 11 (63.6%) of these farms having at least 1 positive animal. There was no significant seasonal difference in STB fecal shedding. In summer (July, August, September), 24 (2.2%) of 1070 cattle samples were STB positive while 26 (2.7%) of 969 samples in spring (April, May, June) were positive.

Initial screening analysis of potential risk factors associated with STB fecal shedding in cattle was performed (Table 1). After adjustment for herd effects and farm size, the variables of interest (P < 0.2) eligible for the multivariable model were included. The odds of cattle being STB positive on dairy farms were approximately 3 times greater in small herds (< 100 cows) than in large herds (≥ 100 cows).

Table 1.

Potential risk factors for fecal shedding of Shiga toxin-encoding bacteria on dairy farms analyzed in initial screening models

		Herds	Cattle
Variable	Level	number	number tested	number positive	%	OR	P-value
Farm management type	Organic	8	413	20	4.8	2.7	0.005
	Conventional	20	1626	30	1.8	—
Farm sizea	< 100 cows	17	920	34	3.7	2.6	0.016
	> 100 cows	11	1119	16	1.4	—
Season	Spring	—	969	26	2.7	1.3	0.662
	Summer	—	1070	24	2.2	—
Rumensin/monensin for weaned calvesa	yes	7	537	4	0.7	0.2	0.006
	no	21	1502	46	3.1	—
Exclusively whole milk fed to calvesa	yes	10	557	26	4.7	2.2	0.042
	no	18	1482	24	1.6	—
Wash boots or use boot dip after handling all calvesa	yes	3	195	1	0.5	0.2	0.080
	no	25	1844	49	2.7	—
Pasture access for lactating or dry cowsa	yes	8	423	18	6.1	1.9	0.084
	no	20	1616	32	2.5	—
Deccox/decoquinate for preweaned calvesa	yes	17	1419	22	1.6	0.4	0.092
	no	11	620	28	4.5	—
Proteins and concentrates all protected from moisture (roof)a	yes	22	1646	33	2.0	0.6	0.112
	no	6	393	17	4.3	—
Poultry have physical contact with dairy cattlea	yes	5	309	13	4.2	1.6	0.179
	no	23	1730	37	2.1	—
TMR used for lactating dairy cowsa	yes	15	1286	29	2.8	1.7	0.180
	no	13	753	21	2.8	—
Routinely use antibiotics in feed or watera	yes	10	974	13	1.3	0.5	0.186
	no	18	1065	37	3.5	—
Vaccinate with E. coli J5 bacterin	yes	10	963	14	1.5	0.6	0.281
	no	18	1076	36	3.3	—
Any parlor (milking facility)	yes	12	1131	17	1.5	0.7	0.317
	no	16	908	33	3.6	—
Lactating dairy cattle in freestalls	yes	12	1131	17	1.5	0.7	0.317
	no	16	908	33	3.6	—
Brewers by-products for high producing cows at any time throughout the study	yes	14	1292	22	1.7	0.7	0.337
	no	14	747	28	3.7	—
Rolling Herd Average less than 18 000	yes	11	661	24	4.8	1.4	0.377
	no	17	1378	26	1.9	—
Wash hands or use disposable gloves after handling all calves	yes	13	1034	18	1.7	0.7	0.381
	no	15	1005	32	3.2	—
Water rinse only or do not wash	yes	9	561	21	3.7	1.3	0.607
	no	19	1478	29	2.0	—
Calf pen/hutches washed and disinfected	yes	6	616	10	1.8	0.8	0.613
	no	22	1423	40	2.8	—
Preweaned dairy calves in individual pen	yes	9	525	18	3.4	0.8	0.719
	no	19	1514	32	2.1	—
Sick pen or facility separate from lactating cows	yes	6	1506	41	2.7	0.9	0.813
	no	22	533	9	1.7	—
Maternity housing separate from other lactating cows	yes	17	1467	32	2.2	1.1	0.817
	no	11	572	18	3.1	—
Any cow primary source of water is surface	yes	4	360	7	1.9	0.9	0.866
	no	24	1679	43	2.6	—
Ration for close-up dry cows different from far-off dry cows	yes	16	1324	29	2.2	1.0	0.990
	no	12	715	21	2.9	—

OR — odds ratio, TMR — total mixed ration.

^adenote variables (P < 0.2) for initial multivariable analysis.

Farm size (< 100 cows versus ≥ 100 cows) was used as a covariate and forced into the initial logistic regression models to control for potential confounding effects (except for farm type).

Seven of 9 eligible independent variables were eliminated from the multivariable model during the backwards selection procedure. The same variables were added through forward selection. After adjusting for farm size as well as controlling for correlation of observations within herds from different sampling times, 3 variables were significantly associated with STB shedding on dairy herds (Table 2).

Table 2.

Risk factors associated with fecal shedding of Shiga toxin-encoding bacteria on dairy farms in multivariable logistic regression model

Variable	Level	OR (95% CI)
Farm size	Small (< 100 cows)	3.6 (2.1–6.2)
	Large (≥ 100 cows)
Monensin (Rumensin) for weaned calves	No	4.8 (2.5–9.3)
	Yes
Decoquinate (Deccox) for preweaned calves	No	2.2 (1.4–3.6)
	Yes
TMR used for lactating dairy cows	Yes	3.0 (1.8–5.1)
	No

OR (95% CI) — odds ratio (95% confidence interval), TMR — total mixed ration. The criteria used for variables to remain in the final multivariable model were the significance of independent variables at P < 0.05 based on GEE parameter estimate. Farm size (< 100 cows versus ≥ 100 cows) was forced into the final multivariable model to control for potential confounding effects.

In our study, the GEE approach was used to analyze risk factors for the clustered data and to adjust for correlation of observations within herds, which has been increasingly used in other risk factor studies (3,9,12). A unique feature of this herd-level longitudinal risk factor study was the use of a screening PCR assay to identify STB.

Small farms (< 100 cows) in this study were associated with increased STB shedding. Because all organic farms enrolled in this study were small farms, we have a limited ability to make comparisons between organic and conventional farms. Although a previous Swiss study showed no significant difference in the prevalence of STEC between organic and conventional farms (7), it can be speculated that organic farms may have higher prevalence of STEC because of their restrictions on the use of manure for fertilization of crops and pasture. It is assumed that as herd size increases, there is a greater chance for cattle to come in contact with the pathogen due to practices associated with large herd sizes such as greater concentration of animals and feeding practices that present greater and more wide spread opportunities for fecal contamination of feeds. We, however, demonstrated that smaller herds, irrespective of whether they were organic or conventional, were more likely to shed STB. It is likely due to larger farms having more personnel and placing greater emphasis on biosecurity practices to reduce the risk of pathogens compared with practices on smaller farms (13). Since farm size has previously been assumed to be a risk factor (14,15), it was included as a covariate in both univariable and multivariable models to eliminate potential confounding. Clearly, additional research is needed to clarify this apparent difference between organic and conventional farms with respect to STEC shedding.

It is interesting that identified risk factors were related to feeding practices. This included the feeding of a TMR as an independently associated factor. Most herds (89.3%) in our study participated in dairy herd improvement association (DHIA) programs and more than half (53.6%) of all herds fed TMR to lactating cows. Cattle on farms feeding a TMR to lactating dairy cows were 3 times more likely to shed STB compared with those on farms not feeding a TMR.

Monensin (Rumensin) appeared to have a protective effect, with cattle being less likely to shed STB in growing heifers. A similar observation was made in a recent experimental dietary study in feedlot cattle, demonstrating that cattle fed monensin at 44 mg/kg of feed had less fecal shedding of E. coli O157:H7 compared with cattle fed 33 mg/kg of feed (16). Monensin is used in cattle production mainly for growth promotion and anticoccidial activity (17). It was not approved for use in lactating cows at the time of this study, but it is currently approved for such use in the United States.

We also found use of decoquinate (Deccox) for preweaned calves had a protective effect. Decoquinate, a quinolone derivative, is commonly used for the prevention of coccidiosis in calves. Both of these ionophores affect Gram-positive bacteria (17). The alteration in gut flora and subsequent impact on Gram-negative gut bacteria are unclear and further experimental clinical research is needed.

In conclusion, this study identified multiple factors related to cattle feeding practices associated with STEC fecal shedding. Examination of these factors in prospective controlled studies would more effectively assess their effectiveness in reducing occurrence of STEC on dairy farms.