Pre-harvest Interventions to Reduce the Shedding of E. coli O157 in the Faeces of Weaned Domestic Ruminants: A Systematic Review

Authors

  • J. M. Sargeant,

    1.  Department of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, ON, Canada
    2.  Laboratory for Foodborne Zoonoses, Public Health Agency of Canada, Guelph, ON, Canada
    Search for more papers by this author
  • M. R. Amezcua,

    1.  Department of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, ON, Canada
    2.  Laboratory for Foodborne Zoonoses, Public Health Agency of Canada, Guelph, ON, Canada
    Search for more papers by this author
  • A. Rajic,

    1.  Laboratory for Foodborne Zoonoses, Public Health Agency of Canada, Guelph, ON, Canada
    Search for more papers by this author
  • L. Waddell

    1.  Laboratory for Foodborne Zoonoses, Public Health Agency of Canada, Guelph, ON, Canada
    Search for more papers by this author

J. Sargeant. Centre for Public Health and Zoonoses, and Department of Population Medicine, Ontario Veterinary College, University of Guelph, Guelph, ON N1G 2W1, Canada. Tel.: +519 824 4120 extn 54045; Fax: +519 763 3117;
E-mail: sargeanj@uoguelph.ca

Summary

Our objective was to use formal systematic review methods to evaluate the efficacy of interventions to reduce faecal shedding of Escherichia coli O157 in post-weaned ruminants by increasing animal resistance. The methodology consisted of an extensive search to identify all potentially relevant research, screening of titles and abstracts for relevance to the research question, quality assessment of relevant research, extraction of data from research of sufficient quality, and qualitative summarization of results. The interventions evaluated included probiotics, vaccination, antimicrobials, sodium chlorate, bacteriophages and other feed additives. There was evidence of efficacy for the probiotic combination Lactobacillus acidophilus NP51 (NPC 747) and Propionibacterium freudenreichii and for sodium chlorate in feed or water. The effectiveness of vaccination varied among studies and among vaccine protocols and there was no consistent evidence to suggest that antibiotic use was associated with a decrease in faecal shedding of E. coli O157, or that current industry uses of antimicrobials were associated with increased faecal shedding. There were an insufficient number of studies available to address the effectiveness of bacteriophages and several other feed additives. In general, few of the primary studies evaluated the interventions under commercial housing conditions with a natural disease challenge, there were inconsistencies in the results among study designs and in some cases among studies within study designs, and a relatively large proportion of publications were excluded based on quality assessment criteria. Few studies reported on associations between the proposed intervention and production parameters, such as average daily gain and feed: gain ratio. While the results suggest that some interventions may be efficacious, there are knowledge gaps in our understanding of the efficacy of pre-harvest interventions to increase animal resistance to E. coli O157 that require further targeted research.

Introduction

Escherichia coli O157 are an important public health concern, resulting in illness, hospitalizations and occasionally death (Mead et al., 1999). Ruminants are reservoirs for this pathogen and play an important role in the epidemiology of human infection (Griffin and Tauxe, 1991). Human outbreaks have been associated with the consumption of contaminated food and water and through direct contact with cattle or cattle environments (Armstrong et al., 1996). Therefore, reducing E. coli O157 at the farm level should decrease the risk of human illness through both direct and indirect routes.

There are several approaches to on-farm control that theoretically could be used to reduce faecal shedding of E. coli O157:H7 in cattle. These strategies include reduced exposure through modification of management practices and interventions to increase animal resistance (Besser et al., 2003; Loneragan and Brashears, 2005). Strategies to increase animal resistance include the use of probiotics, vaccines, antibiotics, sodium chlorate, bacteriophages and other feed additives.

There is a large volume of original research investigating the efficacy of these potential interventions. Summarizing this research provides an overview of the current level of evidence for an intervention’s effectiveness as well as highlighting knowledge gaps for future research. However, a traditional narrative review may not provide an unbiased summary of all of the available evidence (Cook et al., 1997; Williams, 1998). Systematic review is a methodology used to synthesize the results of all available research evidence on a topic using strategies that reduce bias and random error (Mulrow, 1994; Cook et al., 1997). These strategies comprise a comprehensive search for all potentially relevant articles, an evaluation of the methodological quality of the relevant literature, and synthesis and interpretation of the results. While systematic reviews to address treatment and intervention topics are common in the human health literature, this approach has seldom been used to determine the efficacy of interventions to improve food safety (Sargeant et al., 2006).

Our objective was to conduct a systematic review to evaluate the evidence from published and unpublished literature on the use of interventions to increase animal resistance to faecal shedding of E. coli O157 in weaned domestic ruminants. The focus of this paper was on intervention efficacy in live animals. Thus, discovery research and specific mechanisms of action for each intervention were not reviewed.

Materials and Methods

The specific research question addressed in this review was ‘Do pre-harvest interventions to increase animal resistance reduce faecal shedding of E. coli O157 in weaned domestic ruminants’? Six intervention types were individually assessed: probiotics, vaccination, antimicrobials, sodium chlorate, bacteriophages and feed additives.

The review consisted of several steps; an extensive search was conducted to identify all potentially relevant research to address each question, the literature identified by the search was screened for relevance, relevant literature underwent quality assessment, data were extracted from research of sufficient quality, and the resulting data were summarized.

Search strategy

The search was conducted in two phases; initially to identify studies pertaining to probiotic use and then to identify studies related to the remaining interventions. The identification of potentially relevant original research began by compiling a comprehensive list of search terms. The search terms related to the three components of the study question: population (live post-weaned ruminants), intervention (probiotics, vaccines, antimicrobials, sodium chlorate, bacteriophages, and other feed additives) and outcome (E. coli O157 in faeces). Search terms within each component were combined using ‘OR’ with components linked using ‘AND’. Due to the low percentage of abstracts that were relevant in the probiotic search, the search terms were modified for the remaining interventions (Table 1). To identify observational studies investigating multiple potential interventions, we performed an additional search using the search terms (risk factor or management) and (cattle) and (coli*).

Table 1.   Search terms used to identify potentially relevant literature on the efficacy of pre-harvest interventions for reducing E. coli O157 in the faeces of weaned domestic ruminants by increasing animal resistance
 ProbioticsOther interventions
Intervention terms(Probiotic* or Lactob* or Bifidobac* or Propionibac* or Saccharomyces or ‘competitive exclusion’ or ‘strain mixture’ or ‘direct fed microbial*’ or ‘dietary supplementation’ or ‘lactic acid bacteria’ or ‘lactic acid’ or bacteriocin or fermentation or yeast)(vaccin* or immuniz* or bacterin)
(phage or bacteriophag*)
(ionophore* or monensin or salinomycin or lasalocid)
(chlorate)
(antibiotic* or antimicrobial* or prophylaxis or medication or prevent* or therap* or metaphylaxis)
(tetracycline* or oxytetratcycline* or chlortetracycline*)
(aminoglycoside* or neomycin or streptomycin or kanamycin)
(quinolone* or fluoroquinolon* or enrofloxacin or ciprofloxacin)
(cephalosporin* or ceftiofur)
(penicillin or ampicillin)
(extract* or oil* or plant* or herb*)
Population terms(ruminant* or bovine or cattle or cow* or steer* or beef or calf or calves or heifer or sheep or goat* or lamb* or farm or herd)(ruminant* or bovine or cattle or cow* or steer* or feedlot* or calf or calves or sheep or goat* or lamb*)
Outcome terms(‘Escherichia coli’ or ‘Escherichia coli O157’ or O157 or EHEC or ‘shiga-like toxin’ or ‘Bacteria load’ or ‘bacterial load’ or ‘Bacteria level’ or ‘Bacterial level’ or ‘Bacterial counts’ or ‘Bacteria counts’ or faeces or feces or fecal or faecal or manure or coliform)(O157 or O157:H7 or EHEC or shiga* or vertoxi* or VTEC or enterohemorragic or enterohemorrhagic or enterohaemorragic or enterohemorrhagic)

Literature search

Nine electronic databases were searched for abstracts published between 1980 and 2005: MEDLINE, Current Contents, Cab International, Food Science and Technology, PubMed, Biological Sciences, Agricola, EMBASE and Scopus. The search was limited to words in the title or the abstract and no language restriction was imposed. The probiotic search was conducted in January–February 2005 and the search for other interventions in July 2005. The electronic database search was updated for all intervention types in November 2006.

In addition to the electronic database search, theses and dissertations from Canada, the USA and Europe were searched through Theses Canada Portal (http://www.collectionscanada.ca/thesescanada) and Digital Dissertations UMI Pro-Quest (http://www.proquest.com/products_pq/descriptions/pqdt.shtml) databases. Because searching in these databases was more restricted, search term combinations within components and between components were used. The United States Department of Agriculture Agricultural Research Service (USDA- ARS) (http://ars.usda.gov/research/projects) and the Inventory of Canadian Agri-Food Research ICAR (http://www.icar-irac.ca/pls/icarweb/icarqueryeng) web pages were searched for reports on current research. Internet websites were searched through the advanced search in Google (http://www.google.ca). The search terms used for these sources were limited to combinations of each intervention term with the words ‘cattle’ and ‘O157’ and excluded abstracts containing the words meat, milk, yoghurt, bologna, sausage, ground or pigs.

The proceedings of the International Symposium on Shiga Toxin (Verocytotoxin) Producing E. coli Infections (years 2000 and 2003), the Congress of Research Workers in Animal Diseases (CRWAD) (years 2000 to 2005) and the Nebraska Beef Report, University of Nebraska-Lincoln (2000–2005) were hand searched.

All identified citations with their respective abstracts were uploaded into an electronic systematic review data management program (srs 3.0; TrialStat Corporation, Ottawa, ON, Canada). Duplicate abstracts were removed using an internal program within the software, followed by manual deduplication.

Relevance screening

Abstracts were screened for relevance using three criteria: the abstract described (1) primary (original) research, (2) the use of any of the interventions in live weaned domestic ruminants (beef and dairy cattle, sheep and goats) and (3) the presence or concentration of E. coli O157 in faeces. Two reviewers independently screened each abstract. Both reviewers needed to rate the abstract as relevant to advance to the quality assessment stage. Disagreements between reviewers were resolved by consensus.

Quality assessment

Quality assessment forms were created and pre-tested for randomized controlled trials (RCTs), challenge trials, cohort studies and case–control or cross-sectional studies. Full papers were obtained for all relevant abstracts and underwent quality assessment by two reviewers working independently. As there were a total of four reviewers (the co-authors), quality assessment for a specific paper was not necessarily performed by the same individual as the one who had conducted the relevance screening. Disagreements between reviewers were resolved by consensus. The criteria considered essential to include the study in the review were the use of a control group, use of random allocation to treatment groups (RCT and challenge trials), description of intervention protocols and animal housing sufficient to allow the determination of replicates of the intervention, a description of the statistical methods and reporting of the effect estimate or sufficient data to calculate one, and either a standard deviation, standard error, confidence interval or P-value.

Data extraction and analysis

Two reviewers independently extracted data from each study that passed the quality assessment stage using a standardized form. These data included characteristics of the population, unit of allocation of the intervention, description of the intervention protocols, description of the outcome, statistical analysis (including level of analysis) and results. The results provided varied among study designs and publications. Randomized controlled trials and observational studies reported the outcome as dichotomous, corresponding to the presence or absence of E. coli O157. For these studies, the outcome used for this review was the odds ratio (OR) and 95% confidence intervals (CI). When available, these data were extracted for each treatment comparison. When OR were not reported, they were calculated from the number or proportion of positive and negative experimental units in each treatment group when this information was provided. The outcome used for analyses of the challenges trials was the difference in the mean of the log counts of E. coli O157 among treatment groups. When available, these data were extracted for each treatment comparison. If not reported, this information was calculated from the means of each treatment group when possible. Data were summarized separately by intervention and by study design. When multiple treatment comparisons or multiple trials were included within a publication, data were extracted separately for each treatment comparison or trial. Results were considered to be statistically significant at P ≤ 0.05.

Data pertaining to average daily gain, average feed intake and gain to feed ratio were extracted when reported.

Results

Identification and assessment of relevant literature

After removing duplicate citations, there were 3436, 1359 and 323 abstracts screened for relevance for the initial probiotic, other interventions and updated searches respectively. A total of 101 abstracts passed relevance screening. These abstracts pertained to the use of probiotics, vaccines, ionophores, antibiotics, sodium chlorate, bacteriophages or other feed additives.

Several duplicate citations were identified during relevance screening, mainly due to conference proceedings, research reports or theses/dissertations subsequently being published in the peer-reviewed literature. In these instances, the peer-reviewed publication was included and the earlier works were excluded. There were six non-English publications, which were excluded, as resources were not available for translation. Conference proceedings generally did not contain sufficient detail for quality assessment and, therefore, the authors were contacted for additional information. Two reports with sufficient detail were obtained and the remaining conference proceedings were excluded.

Forty-four publications underwent quality assessment. A total of 17 trials from 15 publications were excluded at the quality assessment stage. Three studies pertaining to probiotic use were excluded: one challenge trial did not include a control group (Ohya et al., 2000), allocation to treatment group was not explicitly described as random in one challenge trial (Zhao et al., 1998), and the authors of one challenge trial did not explicitly state that allocation to treatment group was random and the grouping of animals relative to treatment allocation was not explicitly described (Tkalcic et al., 2003). Three challenge studies related to vaccine use from two publications were excluded. One of these publications included three trials, one RCT and two challenge trials (Potter et al., 2004). The two challenge trials, included as preliminary studies to test the vaccine potential of secreted proteins, were excluded because it was not explicitly stated that allocation to treatment group was random. The other excluded vaccine publication was a research report of a challenge trial with no description of the method of statistical analysis (Johnson and McEwen, 2000). Two trials, one RCT and one challenge, evaluating ionophore use were excluded because the authors did not explicitly state that allocation to treatment group was random, although blocking on body weight and gender was used (Edrington et al., 2006a) or because treatment was allocated at the pen level with only one pen per treatment group (Van Baale et al., 2004). One challenge trial evaluating both ionophores and the antibiotic Tylosin was also excluded because treatment was allocated at the pen level with only one pen per treatment group (McAllister et al., 2006). An additional challenge trial of antibiotic use was excluded because no effect measure was provided (Lema and Nahashon, 2006). All of the studies evaluating bacteriophages in weaned ruminants were excluded. These comprised four challenge trials reported in three publications, with no explicit statement that allocation to treatment group was random (Bach et al., 2003), or no statement that allocation to treatment group was random and the description of the grouping of animals relative to treatment allocation not explicitly described (Raya et al., 2006; Sheng et al., 2006). Finally, one challenge trial evaluating the use of esculin/esculetin was excluded because the method of statistical analysis was not described (Duncan et al., 2004), a challenge trial evaluating activated charcoal was excluded because there was no statement that allocation to treatment group was random and the description of the grouping of animals relative to treatment allocation was not explicitly described (Knutson et al., 2006), and one RCT evaluating Ascophyllum nodosum was excluded because the grouping of animals relative to treatment allocation was not explicitly described (Braden et al., 2004).

In addition to the quality assessments exclusions, four additional studies were excluded. One RCT and one observational study were excluded because the small number of E. coli O157 positive samples precluded statistical analysis (Van Donkersgoed et al., 2001; Schroeder et al., 2004), one observational study was excluded because the outcome (defined at the herd level) combined data from pre-weaned and post-weaned calves (Garber et al., 1995) and one challenge trial was excluded because the measured outcome was Shiga-toxigenic E. coli (STEC), rather than specifically E. coli O157 (Misawa et al., 2000).

Probiotics

Various probiotic products, product combinations and dosages were evaluated. In all of the studies, probiotic treatments were administered in the feed and were compared with a no-treatment control or a placebo. There were six RCTs reporting 15 treatment comparisons, two challenge trials reporting six treatment comparisons and four observational studies each reporting a single treatment comparison (Table 2). One RCT and three observational studies measured the outcome once (Dargatz et al., 1997; Sargeant et al., 2004; Dewell et al., 2005; Woerner et al., 2006). The remaining RCTs and challenge trials measured the outcome more than once and accounted for the repeated measures in the analysis.

Table 2.   Summary of results for studies evaluating the use of probiotics to reduce faecal shedding of E. coli O157 in weaned domestic ruminants
StudyProbiotic (dose in CFU)Level of analysisSample sizeOR (95% CI)Mean difference from control groupP-valueConclusions
  1. Bolded results were calculated post hoc.

Randomized controlled trials
 L. acidophilus NP51 (NPC 747) + P. freudenreichii
  Woerner et al., 2006Bovamine Rumen Culture (L. acidophilus NP747 + P. freudenreichii) (2 × 1010)Per cent positive among treatment groupsThree pens (approx. 200 head, 25 sampled) per treatment group0.19 (0.08–0.42) Individual level, no adjustment for pen effect >0.05 (pen level)Not significant
  Younts-Dahl et al., 2004L. acidophilus NP51 (109) + P. freudenreichii (109)Binomial response within pen 12 pens of five animals per treatment group0.42 (0.23–0.79) 0.007Favours probiotic
Younts-Dahl et al., 2005L. acidophilus NP51 (109) + P. freudenreichii (109)Binomial response within pen12 pens of five animals per treatment group0.23 (0.12–0.44) <0.001Favours probiotic
  Younts-Dahl et al., 2005L. acidophilus NP51 (108) + P. freudenreichii (109)Binomial response within pen12 pens of five animals per treatment group0.34 (0.19–0.62) <0.001Favours probiotic
  Younts-Dahl et al., 2005L. acidophilus NP51 (107) + P. freudenreichii (109)Binomial response within pen12 pens of five animals per treatment group0.37 (0.21–0.68) 0.002Favours probiotic
 L. acidophilus NP51 (NPC 747) + L. acidophilus NP 45 (NPC 750) + P. freudenreichii
  Younts-Dahl et al., 2004L. acidophilus NP51 (109) + L. acidophilus NP45 (106) + P. freudenreichii (109)Binomial response within pen 12 pens of five animals per treatment group0.71 (0.35–1.43) 0.337Not significant
 Younts-Dahl et al., 2004L. acidophilus NP51 (106) + L. acidophilus NP45 (106) + P. freudenreichii (109)Binomial response within pen 12 pens of five animals per treatment group1.10 (0.54–2.26) 0.789Not significant
 Younts-Dahl et al., 2005L. acidophilus NP51 (109) + L. acidophilus NP45 (106) + P. freudenreichii(109)Binomial response within pen12 pens of five animals per treatment group0.66 (0.37–1.17) 0.15Not significant
 L. acidophilus NP51 (NPC 747)
  Folmer et al., 2003 (trial 1)L. acidophilus NPC 747 (109) +  pen cleaningPen level (at least one positive)18 pens of eight animals in each of probiotic and control groups 0.7 (0.13–3.7) pen level 0.3Not significant
  Folmer et al., 2003 (trial 2)L. acidophilus NPC 747 (109) +  change of dietPen (at least one positive)18 pens of eight animals in each of probiotic and control groups0.25 (0.05–1.18) pen level 0.1Not significant
  Peterson et al., 2005aL. acidophilus NPC 747 (109) Pen24 pens (total 224 animals) per treatment group, two sampling periods combined(Treated steers 35% less likely to shed E. coli O157) 0.002Favours probiotic
  Brashears et al., 2003L. acidophilus NPC 747 (109)Individual, pen effect controlled 12 pens of five animals per treatment group0.51 (0.32–0.82) 0.006Favours probiotic
 L. acidophilus NP45 (NPC 750)
  Folmer et al., 2003 (trial 1)L. acidophilus NPC 750 (109) +  pen cleaning Pen (at least one positive)18 pens of eight animals in each of probiotic and control groups 0.21 (0.02–2.06) pen level 0.3Not significant
  Folmer et al., 2003 (trial 2)L. acidophilus NPC 750 (109) +  change of dietPen (at least one positive)18 pens of eight animals in each of probiotic and control groups0.25 (0.05–1.18) pen level 0.1Not significant
  Brashears et al., 2003L. acidophilus NPC 750 (109)Individual, pen effect controlled12 pens of five animals per treatment group0.70 (0.45–1.1) 0.12Not significant
 Challenge trials
  Lema et al., 2001L. acidophilusIndividual, no pen controlTwo pens of three animals per treatment group −0.1>0.05Not significant
Lema et al., 2001Streptococcus faeciumIndividual, no pen controlTwo pens of three animals per treatment group −2.1 <0.05Favours probiotic
Lema et al., 2001S. faecium + L. acidophilusIndividual, no pen controlTwo pens of three animals per treatment group −1.2<0.05Favours probiotic
Lema et al., 2001S. faecium + L. acidophilus + L. casei + L. fermentum + L. plantarumIndividual level, no pen controlTwo pens of three animals per treatment group −3.3<0.05Favours probiotic
Schamberger et al., 2004Colicin E7 E. coli (107) (Comparison between groups in period 1)Individual level, room effect controlledFour pens of two animals per treatment group, three sampling periods −0.70.12Not significant
Schamberger et al., 2004Colicin E7 E. coli (108) (Comparison between groups in period 3)Individual level, room effect controlledFour pens of two animals per treatment group, three sampling periods −0.40.30Not significant
Schamberger et al., 2004Colicin E7 E. coli [Cross-over comparison – group 1 (107) versus 3 (108)]Individual level, room effect controlledFour pens of two animals per treatment group, three sampling periods −1.10.001Favours probiotic
Schamberger et al., 2004Colicin E7 E.coli [Cross-over comparison – group 2 (107) versus 3 (108)]Individual level, room effect controlledFour pens of two animals per treatment group, three sampling periods 0.050.93Not significant
 Observational studies
Dargatz et al., 1997Probiotics in current diet (yes/no)Pen level (at least one positive), controlled for farm100 feedlots, four pens per feedlot, 30 samples per pen  >0.25 (univariable)Not significant
Garber et al., 1999Probiotics in feed (yes/no)Farm level (at least one positive)91 herds, variable number sampled per herd3.01 (1.02–8.89) farm level 0.04 (univariable), 0.18 (multivariable)Not significant
Sargeant et al., 2004Probiotics in current ration (yes/no)Individual level controlling for pen and farm73 feedlots, 691 pens, 15 sampled per pen1.02 (0.75–1.39) pen level, univariable >0.2 (univariable)Not significant
Dewell et al., 2005Probiotics in feed (yes/no)Pen level, control of farm not described15 pens from 12 feedlots, 30 samples per pen2.69 (1.72–4.20) pen level, univariable <0.001 (univariable), >0.05 (multivariable)Not significant

All of the RCTs were conducted in beef cattle on single research farms in the USA with the exception of one trial that was conducted on a single commercial beef feedlot (Woerner et al., 2006). Treatment allocation in all trials was at the pen level. The individual level prevalence of E. coli O157 ranged from 12% to 24.5%, but was not reported at the individual animal level in one publication (Folmer et al., 2003). Four of the five treatment comparisons evaluating a combination of Lactobacillus acidophilus NP747 (NP51) and Propionibacterium freudenreichii reported a significantly lower prevalence of E. coli O157 in treated cattle, regardless of the dosage used. None of the studies evaluating L. acidophilus NPC 750 (NP45) alone or in combination with L. acidophilus NPC 747 (NP51) and P. freudenreichii reported a significant odds ratio although, a numerical reduction was reported for all but one trial. Two of the four trials evaluating L. acidophilus NPC 747 (NP51) reported a significant odds ratio, although all four reported a numerical reduction.

Both of the challenge trials were conducted in single research farms in the USA with treatment allocation at the pen level. Reduction in the level of E. coli O157 was reported when Streptococcus faecium or combinations of S. faecium with other lactic acid bacteria were used (Lema et al., 2001). Colicin E7 E. coli showed significance in one of two cross-over comparisons (Schamberger et al., 2004).

All of the observational studies used a cross-sectional design and were conducted in the USA. One study was conducted at the farm level in dairy cows (Garber et al., 1999), with the remainder conducted at the pen level in commercial beef feedlots. The individual-level prevalence of E. coli O157 in the observational studies ranged from <5.5% to 24.7%, but was not reported for one study (Dargatz et al., 1997). Probiotic use was not significantly associated with E. coli O157 in multivariable models in any of the observational studies.

Growth performance was reported in three RCTs (Brashears et al., 2003; Folmer et al., 2003; Peterson et al., 2005a) and one challenge trail (Lema et al., 2001). None of the studies reported negative effects on performance. Average daily gain and gain to feed ratio were significantly higher for the probiotic treated group in one challenge trial (Lema et al., 2001).

Vaccination

The four studies evaluating vaccine efficacy were RCTs (Table 3) and were conducted using beef cattle in the USA or Canada. Two of the studies were conducted in single research feedlots (Potter et al., 2004; Peterson et al., 2005b), one was conducted in a single commercial feedlot (Woerner et al., 2006) and one was conducted in nine commercial feedlots (Van Donkersgoed et al., 2005). Vaccines were administered individually to animals grouped in pens, with the exception of Peterson et al. (2005b) who used both within pen and external controls. Three of the RCTs used one to three doses of Type III secreted protein vaccine (Potter et al., 2004; Peterson et al., 2005b; Van Donkersgoed et al., 2005). One study reported the use of two doses of an experimental bacterin that included a proprietary dual adjuvant system and immune dominant antigens of E. coli O157:H7 including intimin and lipopolysaccharides (Woerner et al., 2006). The individual animal prevalence of E. coli O157 ranged from 14.5% to 45% in the US trials. In the Canadian trial, the mean pen level prevalence was 5% with variability among pens (Van Donkersgoed et al., 2005). Three studies measured the outcome more than once and accounted for this in the analysis (Potter et al., 2004; Peterson et al., 2005b; Van Donkersgoed et al., 2005).

Table 3.   Summary of results for studies evaluating the use of vaccines to reduce faecal shedding of E. coli O157 in weaned domestic ruminants
StudyVaccineLevel of analysisSample sizeOR (95% CI)P-valueConclusion
  1. Bolded results were calculated post hoc.

Randomized controlled trials
 One dose of vaccine
Peterson et al., 2005bType III secreted proteinsIndividual (multiple treatment groups per pen and external control pens), pen controlled60 pens of eight animals, two per treatment group within pen, 12 pens (total 128 animals) as external controls0.94 (comparison to non-treated controls within pen) <0.05 compared with external control pensFavours vaccine among pens
 Two doses of vaccine
Woerner et al., 2006Experimental bacterin (Fort dodge, IA)Per cent positive among treatment groupsThree pens (approx. 200 head, 25 sampled) per treatment group0.21 (0.09–0.45) Individual level, no adjustment for pen effect>0.05Not significant
Peterson et al., 2005bType III secreted proteinsIndividual (multiple treatment groups per pen and external control pens), pen controlled60 pens of eight animals, two per treatment group within pen, 12 pens (total 128 animals) as external controls0.82 (comparison to non-treated controls within pen) <0.05 compared with external control pensFavours vaccine among pens
Van Donkersgoed et al., 2005Type III secreted proteinsPen, feedlot effect controlledNine feedlots, total 109 pens per treatment group, 30 samples/ pen at each sampling time >0.25Not significant
 Three doses of vaccine
Potter et al., 2004Type III secreted proteins Individual, pen effect controlled12 pens of eight animals per treatment group 0.04Favours vaccine
Peterson et al., 2005bType III secreted proteinsIndividual (multiple treatment groups per pen and external control pens), pen 60 pens of eight animals, two per treatment group within pen, 12 pens (total 128 animals) as external controls0.59 (comparison to non-treated controls within pen) 0.06 within pen
<0.05 compared with external control pens
Favours vaccine among pens

Two studies reported a significant vaccine effect (Peterson et al., 2005b; Potter et al., 2004) although numerical reduction were reported in all studies. No significant difference in average daily gain was found in the studies reporting this parameter (Potter et al., 2004; Peterson et al., 2005b).

Antimicrobials

The antimicrobial products evaluated included ionophores administered in the feed, ractopamine administered in the feed and antibiotics administered in the feed or by injection (Table 4). Antimicrobial use studies were based on one of two premises: to evaluate their use as a potential intervention to reduce faecal shedding of E. coli O157 or to investigate whether current industry antimicrobial use practices were associated with faecal shedding.

Table 4.   Summary of results for studies evaluating the use of antimicrobials to reduce faecal shedding of E. coli O157 in weaned domestic ruminants
StudyAntimicrobialLevel of analysis and control of clusteringSample size OR (95% CI)Mean differenceP-valueConclusion
  1. Bolded results were calculated post hoc.

Ionophores
 Randomized controlled trial
Lefebvre et al., 2005Monensin (Rumensin) in feedIndividual prevalence, control of pen and repeated measures not describedTwo pens of five animals per treatment group0 prevalence in control group >0.05 across all time periodsNot significant
 Challenge trial
Edrington et al., 2003aMonensin (Rumensin) in feedIndividual level analysis (housed individually)Four animals per treatment group 0.99 cfu (log10)/g, Compared with non-treated controls, pooled over time>0.05Not significant
Edrington et al., 2003aLaidlomycin propionate in feedIndividual level analysis (housed individually)Four animals per treatment group 1.31 cfu (log10)/g, Compared with non-treated controls, pooled over time>0.05Not significant
 Observational studies
Dargatz et al., 1997Ionophores in current diet (yes/no)Pen level (at least one positive), controlled for farm100 feedlots, four pens per feedlot, 30 samples per pen0.70 (0.40–1.22) pen-level, univariable 0.21Not significant
Herriott et al., 1998Ionophores (Monensin or lacalocid) in feed (yes/no)Farm level36 herds, approximately 60 samples per herd per sampling time5.95 (1.03–34.41) farm-level, univariable 0.11Not significant
Herriott et al., 1998Monensin in feed (yes/no)Farm level36 herds, approximately 60 samples per herd per sampling time0 in one cell 0.10Not significant
Herriott et al., 1998Lasalocid in feed (yes/no)Farm level36 herds, approximately 60 samples per herd per sampling time2.41 (0.42–13.83) farm-level, univariable 0.70Not significant
Sargeant et al., 2004Ionophores in feed (yes/no)Individual level, controlled for pen and farm73 feedlots, 680 pens, 15 samples per pen3.31 (1.19–9.20) pen-level, univariable >0.05Not significant
Dewell et al., 2005Ionophores in feed (yes/no)Pen control for farm not described.15 pens from 12 feedlots, 30 samples per pen0.73 (0.4–1.3) univariable 0.31 (univariable)Not significant
 Ractopamine HCl
Edrington et al., 2006bRactopamine (20 mg/head) orally for 28 daysIndividual level10 animals per treatment, housed in one pen  0.0006 (overall)Favours ractopamine
Edrington et al., 2006bRactopamine (200 mg) in feed for 28 daysIndividual pen floor samples, control of pen not describedNine pens per treatment (approximately 100 animals per pen, 30 sampled)  0.05Favours ractopamine
Antibiotics in feed
 Randomized controlled trial
Woerner et al., 2006Neomycin sulphatePer cent positive among treatment groupsThree pens (approx. 200 head, 25 sampled) per treatment group0 prevalence in treated group   
 Challenge trial
Edrington et al., 2003aBambermycinIndividual level analysis (housed individually)Four animals per treatment group 1.22 cfu (log10)/g, Compared with non-treated controls, pooled over time>0.05Not significant
 Observational studies
Dargatz et al., 1997Tetracycline in current diet (yes/no)Pen level (at least one positive), controlled for farm100 feedlots, four pens per feedlot, 30 samples per pen  >0.25 univariableNot significant
Dargatz et al., 1997Antibiotic other than tetracycline in current diet (yes/no)Pen level (at least one positive), controlled for farm100 feedlots, four pens per feedlot, 30 samples per pen  >0.25 univariableNot significant
Sargeant et al., 2004Antibiotics in feed or water at any time in production (yes/no)Individual level, controlled for pen and farm73 feedlots, 689 pens, 15 samples per pen0.55 (0.36–0.85) multivariable 0.006 multivariableFavours antibiotics in feed
Dewell et al., 2005Antibiotics in feed (yes/no)Pen control for farm not described15 pens from 12 feedlots, 30 samples per pen0.5 (0.33–0.78) univariable 0.002 (univariable)
>0.05 multivariable
Not significant
Injectable antibiotics
 Observational study
Sargeant et al., 2004Metaphylaxis- injectable antibiotics at any time during productionIndividual level, controlled for pen and farm73 feedlots, 679 pens, 15 samples per pen2.14 (1.41–3.23) multivariable <0.001Against injectable antibiotics
Lefebvre et al., 2005OxytetracyclineIndividual prevalence, control of pen and repeated measures not describedTwo pens of five animals per treatment group0 prevalence in control group >0.05 across all time periodsNot significant
 Ecological studies related to antibiotic use
Galland et al., 2001Hospital versus home pensPen, control for farm not described, results presented at individual sample levelFour feedlots, 30 pens per sampling period, five samples per pen0.87 (0.25–3.02) individual-level, univariable >0.05Not significant
Dunn et al., 2004Comparison of before and after pre-conditioning that included oxytetracyline in feedIndividual level408 animals0 positive after pre-conditioning <0.01Favours pre-conditioning
Kuhnert et al., 2005Conventional versus organic farmsIndividual level, control of farm, farm type included as forced variable60 farms of each type, 5–13 cows sampled (depending on farm size)1.05 (0.74–1.51) multivariable, all STEC  0.8 (all STEC)
>0.05 (E. coli O57:H7)
Not significant

Ionophore use was evaluated in one RCT, one challenge trial (two ionophore comparisons) and four cross-sectional observational studies comprising six comparisons. The RCT was conducted in feedlot cattle in a single US research herd (Lefebvre et al., 2005) and the challenge trial was conducted in lambs in a single US research flock (Edrington et al., 2003a). The outcome in these studies was measured more than once, with control for the repeated measures described in the challenge trials. All four of the cross-sectional studies of ionophore use were conducted in the USA, with treatment allocation at the pen-level (Dargatz et al., 1997; Sargeant et al., 2004; Dewell et al., 2005) or farm-level (Herriott et al., 1998). All were conducted in commercial feedlots, with the exception of Herriott et al. (1998), which was conducted in commercial dairy herds. The individual-level prevalence of E. coli O157 in the observational studies ranged from <5.5% to 24.7%, but was not reported for one study (Dargatz et al., 1997). None of the studies found a significant association between ionophore use and faecal shedding of E. coli O157. Associations between ionophore use and growth performance were not reported in any of the studies.

Ractopamine was evaluated in two RCTs in feedlot cattle reported in the same publication (Edrington et al., 2006b). The first trial evaluated the product in individually treated and control cattle housed in the same pen. Repeated sampling was used and controlled in the analysis. The second trial used in-feed treatment at the pen level. The individual level prevalence of E. coli O157:H7 in the control group in the first trial was 41.8% and the overall prevalence in the second trial was 21.0%. Although both trials reported a significant decrease with treatment, the second trial noted an increase in the percentage of samples positive for Salmonella in the treated group.

Antibiotics in feed were evaluated in one RCT, one challenge trial and three cross-sectional observational studies. All studies were conducted in commercial feedlots in the USA with treatment at the pen-level, except for the challenge trial, which was conducted in individually housed lambs (Edrington et al., 2003a). The RCT evaluated the use of neomycin sulphate in one commercial feedlot (Woerner et al., 2006). The individual-level prevalence of E. coli O157 in control animals was 45.8%. The challenge trial evaluated the use of bambermycin in individually housed lambs (Edrington et al., 2003a). The observational studies evaluated the use of tetracycline in the current diet or other antibiotics in the current diet (Dargatz et al., 1997), the use of antibiotics (type unspecified) in the feed or water at any time during production (Sargeant et al., 2004) or a binary variable for current antibiotic use (type unspecified) (Dewell et al., 2005). The individual-level prevalence of E. coli O157 ranged from 10.2% to 24.7% in the observational studies, but was not reported from one study (Dargatz et al., 1997). Only one observational study found a statistically significant reduction in faecal shedding of E. coli O157 associated with antibiotic use (Sargeant et al., 2004), although the author of the RCT noted a numeric reduction and low statistical power (Woerner et al., 2006). Associations between antibiotic use in feed and growth performance were not reported in any of the studies.

The use of injectable antibiotics (type unspecified) as metaphylaxis was evaluated in one RCT in a US research herd (Lefebvre et al., 2005) and one cross-sectional observational study in commercial feedlots in the USA (Sargeant et al., 2004). The latter study identified a positive association between metaphylaxis and E. coli O157 in multivariable analysis. Associations between metaphylaxis and growth performance were not reported.

Three ecological studies evaluated the use of antibiotics. The ecological studies compared faecal shedding of E. coli O157 in hospital pens where cattle have frequently been treated with injectable antibiotics versus home pens in commercial US feedlots (Galland et al., 2001), before and after pre-conditioning with a ration that contained oxytetracycline in pastured pre-feedlot beef cattle in the USA (Dunn et al., 2004), and in conventional farms versus organic dairy farms in Switzerland (Kuhnert et al., 2005). The individual-level prevalence of E. coli O157 in these studies ranged from 0.19% to 4.6%. Only the study by Dunn et al. (2004) reported a significant association.

Sodium chlorate

One RCT and three challenge trials evaluated the use of sodium chlorate in feed or water (Table 5). All were conducted on single research farms in the USA and involved cattle (Callaway et al., 2002; Anderson et al., 2005) or sheep (Callaway et al., 2003; Edrington et al., 2003b). Treatment allocation in all studies was at the individual level. In the RCT, the individual-level prevalence of E. coli O157 pre-treatment was 55%. All studies showed a significant reduction in prevalence or level of faecal shedding with the exception of the water treatments in the RCT and the lowest feed dosage in one challenge trial, which resulted in a numeric, but not statistically significant, reduction. In the RCT, feed intake was not significantly different between treated and control animals (Anderson et al., 2005).

Table 5.   Summary of results for studies evaluating the use of sodium chlorate to reduce faecal shedding of E. coli O157 in weaned domestic ruminants
StudyChlorate treatmentLevel of analysisSample sizeMean differenceP-valueConclusion
Randomized controlled trial
 Anderson et al., 20052 × 4 factorial
(1) Water: no treatment, 2500 ppm active chlorate ion 12 h treatment
(2) Feed: no treatment, 0.01% BW chlorate for 1 or 5 days, 0.05% BW for 1 day
Individual level analysis (housed individually)Eight animals per treatment group <0.001 (main effect, feed)
>0.05 (main effect, water)
Favours sodium chlorate in feed
Not significant (water)
Challenge trails
 Callaway et al., 20022.5 mm KNO3 + 100 mm NaClO3  + potassium nitrate for 24 h in waterIndividual levelFour animals per treatment groupApproximately 3 log reduction in CFU/g <0.05Favours sodium chlorate in water
 Callaway et al., 20032.5 mm KNO3 + 100 mm NaClO3 for 24 h in waterIndividual levelSeven animals per treatment groupApproximately 4 log reduction in CFU/g <0.05Favours sodium chlorate in water
 Edrington et al., 2003b24 h feeding of low level of chlorate productIndividual levelFive animals per treatment group1 log CFU/g (24 h post-treatment)>0.05Not significant
 Edrington et al., 2003b24 h feeding of medium level of chlorate productIndividual levelFive animals per treatment group2 log CFU/g (24 h post-treatment)<0.05Favours sodium chlorate in feed
 Edrington et al., 2003b24 h feeding of high level of chlorate productIndividual levelFive animals per treatment group2 log CFU/g (24 h post-treatment)<0.05Favours sodium chlorate in feed

Other feed additives

A single challenge trial evaluated the use of orally administered polyclonal anti-Escherichia coli O157:H7 chicken egg yolk antibodies at various levels in sheep (two pens of three animals per treatment group) (Cook et al., 2005). Sheep were treated three times and faecal samples were collected over a 9-week period. Averaged across all sampling times, faecal shedding was lower in the high and medium level treatment groups.

Discussion

The results of this systematic review suggest that there is evidence of efficacy for some pre-harvest interventions aimed at increasing animal resistance to E. coli O157 in post-weaned ruminants, in particular for the probiotic combination L. acidophilus NP51 (NPC 747) and P. freudenreichii and for sodium chlorate in feed or water. Results for vaccination were not consistent among studies and there was little evidence to suggest that antibiotic use was associated with a decrease in faecal shedding of E. coli O157, or that current industry uses of antimicrobials were associated with increased faecal shedding. However, there were few RCTs evaluating the evidence of effectiveness of any of the interventions under realistic field conditions, there were considerable inconsistencies in the results among study designs and in some cases among studies within study designs, and a relatively large proportion of publications were excluded based on quality assessment criteria.

Multiple study designs applicable to live animal research were included in this review to allow consideration of the totality of research evidence for each intervention. Our review excluded laboratory studies related to the development of interventions and proof of concept testing in laboratory models or non-bovine species. While these types of studies are an important, and often essential, part of development of an intervention and initial validation testing, they do not address the efficacy of the intervention in the species of interest. Thus, the number of studies identified for the different interventions, and the type of study designs used, may in part be due to the stage of development of the intervention. Interventions for which there were few studies evaluating efficacy in live animals might prove to be efficacious with further development and testing. Research conducted in pre-weaned ruminants was also excluded. There are considerable differences in the digestive system, nutrition and housing of pre-weaned ruminants, which may be associated with the efficacy of interventions. Additionally, effectiveness of an intervention in post-weaned ruminants may be more germane to public health.

Therefore, the study designs included in this review were RCTs, challenge trials and observational studies, including ecological designs. Different study designs differ in the level of evidence that they provide (Cockcroft and Holmes, 2003). Experimental designs included challenge trials, in which animals were inoculated with E. coli O157 and RCTs, in which the disease challenge (exposure to E. coli O157) was natural. A further distinction may be based on housing conditions. The challenge trials were conducted on research farms with animals housed individually or in small groups. The RCTs were either conducted on research farms, generally with animals housed in small groups or under commercial conditions. While each of these experimental approaches is appropriate to investigate efficacy of the intervention under the conditions of the experiment, both the type of disease challenge and the housing conditions have important implications to assessing effectiveness in the field. Challenge trials and RCTs using small numbers of animals per group provide an efficient means of evaluating preliminary evidence for effectiveness and may be necessary prior to regulatory approval of an intervention for animals entering the food chain. An additional advantage of challenge trials is that they allow for quantification of pathogen concentration, generally through the use of antibiotic resistant strains of E. coli O157, rather than testing for the presence or absence of the pathogen, as was the case for the natural disease challenge studies. This allows an evaluation of the ability of an intervention to reduce faecal shedding, without necessarily having to eliminate faecal shedding or reduce shedding below detectable levels. A disadvantage, however, is that laboratory adapted strains may not represent the behaviour of wild-type strains. Therefore, while challenge trials and small group RCTs are an essential step, the strength of evidence for effectiveness of an intervention under field conditions would be greatest for an RCT conducted under commercial conditions, and there is a need to conduct these types of studies when development of the intervention is sufficient to allow their use.

Observational studies also were included in this review. Observational studies are conducted in the species of interest in the natural environment and without investigator control of allocation to treatment group (Martin et al., 1987; Dohoo et al., 2003). Thus, they tend to have high external validity (generalizability) compared with experimental studies using laboratory-adapted bacterial strains or in populations selected based on specific exclusion/inclusion criteria. The observational studies identified in this review were cross sectional or ecological. These designs are prone to confounding bias, it is often difficult to prove the temporal relationship between intervention and outcome, and, for ecological studies the potential for ecological fallacies must be considered. Therefore, these study designs are generally used for hypothesis generating rather than hypothesis testing. In this review, few of the observational studies identified significant associations between the interventions and faecal shedding of E. coli O157, even for interventions where challenge trials and RCTs tended to find an association. One possible explanation for these differences is the nonspecific definitions of intervention used in many of the observational studies. For instance, probiotic use was defined as ‘probiotics in ration’ as a dichotomous variable in the observational studies without differentiation as to specific product, dosage or length of administration prior to sampling. Cross-sectional studies may be best suited for time-invariant factors (Dohoo et al., 2003) or may require more specific information to be collected if they are to be of use in hypothesis generation or testing for on-farm food safety interventions.

There were some differences among studies using the same study design in the estimate of the effect measure and occasionally in the overall conclusion for specific interventions. This could be a function of different sample sizes (and therefore different statistical power), differences in the population under study, differences in the product or product administration protocol or, in the case of RCTs and observational studies, differences in the prevalence of E. coli O157 (natural variation or related to different laboratory testing procedures). Although most of the studies either provided odds ratios (RCTs, observational studies) or least square means (challenge studies) or the data to calculate them, there were differences among studies in the level of analysis (individual versus pen versus farm) or the sampling protocols and measures contributing to the calculation of least square means. For these reasons, meta-analysis, a statistical combination of the data to provide an overall summary measure of effect, was not performed.

Methodological concerns were identified for several studies in the quality assessment stage. The most common issue was no explicit statement that allocation to treatment group was random in RCTs and challenge trials, although in some of these studies blocking by potentially confounding variables was reported. Non-random allocation to treatment group, or inadequate randomization methods, may results in biased estimates of treatment effect (Chalmers et al., 1983; Colditz et al., 1989; Miller et al., 1989; Schulz et al., 1995; Kunz and Oxman, 1998; McKee et al., 1999). It is possible that random allocation was performed, but not explicitly reported. Studies in the medical literature have shown that failure to report randomization and to describe the method used for randomization is not uncommon in RCTs (Altman and Dore, 1990; Schulz et al., 1994). Concerns related to the quality of reporting of RCTs in the medical literature led to the development of the CONSORT (Consolidated Standards for Reporting of Trials) statement, a checklist and flow diagram describing items necessary for comprehensive and complete reporting of RCTs (Begg et al., 1996; Moher, 1998). Several major medical journals, such as BMJ, JAMA and Lancet, require authors to follow the CONSORT recommendations when submitting RCTs for publication (Altman, 1996), resulting in improvements in the quality of reports of RCTs in these journals (Moher et al., 2001). Use of guidelines in food safety research may assist authors, reviewers and editors to ensure complete reporting of essential design features for RCTs.

The CONSORT statement also includes a recommendation to report the method of allocation concealment (blinding) (Begg et al., 1996). We did not include blinding of treatment allocation as a criterion in the quality assessment, because almost none of the studies reported on this criterion. This may be related to the outcome assessment being conducted using objective laboratory testing, often not performed directly by the investigator. Nonetheless, there is empirical evidence in the literature that inadequate concealment of treatment allocation in RCTs yields larger estimates of treatment effects compared with studies reporting adequate concealment of allocation (Colditz et al., 1989; Schulz et al., 1995). Explicitly blinding treatment allocation, and reporting such, should be a consideration in future trials.

The potential for publication bias is of concern in any review of the literature. There is a bias towards not publishing negative trials (Counsell, 1997; Thornton and Lee, 2000) and many abstracts are never published as full papers (Scherer et al., 1994). Searches of electronic databases may identify only half of all relevant research studies (McManus et al., 1998). Unpublished studies may reflect the results of more recent research. However, unpublished studies have not been subject to peer-review and, while peer-review may be imperfect, it does provide some ascertainment of study quality (Thornton and Lee, 2000). The inclusion of poor-quality trials in reviews can bias the conclusions (Khan et al., 1996). In this review, we attempted to identify all relevant literature, both published and unpublished. However, the majority of unpublished research did not contain sufficient detail to allow critical appraisal of the study methods. Attempts to obtain additional detail from the authors of unpublished data for this review were time-intensive and resulted in little additional information. Reasons for not providing additional information included the preliminary nature of the research and proprietary or pending patent issues (data not shown). Thus, publication bias is a potential concern in this review. However, providing explicit search criteria and inclusion/exclusion criteria provides transparency in the methods for locating and including primary research.

Most of the studies included in this review did not include information on cattle performance. The studies that did report performance found either no association or improved performance with the intervention. This information is useful in the decision-making process; cost may be a consideration when deciding between effective interventions and positive associations with performance may increase uptake of an intervention. This information is particularly important when trials are conducted under commercial conditions and, therefore, these measures should be included, when possible, in these types of studies.

In summary, there is some evidence for efficacy of interventions aimed at increasing animal resistance to faecal shedding of E. coli O157. However, there was a paucity of studies evaluating the interventions under commercial housing conditions with a natural disease challenge. Numerous study designs contribute to initial development and evaluation of potential interventions. However, once development of an intervention is sufficiently advanced, there is a need to conduct high quality RCTs or prospective cohort studies and to publish the results in a public forum with sufficient detail of the methods to allow interpretation of the results and replication of study methods, if necessary. Including information on cattle performance would provide useful supplementary information for selecting intervention strategies.

Acknowledgements

Funding was provided by the Laboratory for Foodborne Zoonoses, Public Health Agency of Canada. The authors thank Janet Harris, Lindsay Downey and Sandra Connors for library and technical assistance.

Ancillary