Materials and Methods

Sample collection—This research was a part of a large study investigating the apparent prevalence of and risk factors associated with antimicrobial susceptibility patterns of Escherichia coli and Campylobacter spp isolated from finisher pigs on antimicrobial-free and conventionally managed farms in 8 states in the Midwestern United States. Results regarding AMR patterns in E coli from the study have been reported.¹⁷

The present study included 95 farms in the Midwestern United States, including Iowa (n = 37), Illinois (15), Indiana (5), Michigan (21), Minnesota (8), Nebraska (6), Ohio (2), and Wisconsin (1). Sixty farms were managed under conventional swine farm practices, and 35 farms were considered antimicrobial-free facilities. The production systems that were classified as antimicrobial-free had not used antimicrobial drugs for a minimum of 1 year prior to enrollment in the study. Antimicrobial-free farms were selected from membership lists of 2 cooperatives; conventional farms were selected on the basis of close geographic proximity to the antimicrobial-free farms, or the number of slaughter pigs produced per year.¹⁷ The total number of pigs marketed per year was used as a surrogate for herd size.

Samples of feces were collected from 15 pigs on each farm with the exception of 1 farm, where only 12 pigs were available for sample collection. Collection of feces from individual pigs on farms has been previously described.¹⁷ Briefly, farms were visited once in 2002 or 2003, and samples were collected only from healthy pigs. Approximately 5 g of fresh fecal material/ pig was collected and placed in a tube containing Cary-Blair transport medium.^a The specimens were sent on ice to the National Food Safety and Toxicology Center, Michigan State University, and plated within 48 hours of collection.

Bacterial culture—Approximately 1 g of fecal material/sample from the Cary-Blair transport medium was inoculated onto 1 blood agar plate.^b The inoculated plates were incubated for 48 hours at 42°C in a microaerophilic atmosphere of 5% to 12% CO₂ and 5% to 15% O₂.^c On each plate, 4 Campylobacter-like colonies were identified, if present, and plated on 1 of 4 quadrants on a blood agar plate and incubated. Colonies typical for Campylobacter spp were further characterized by Gram stain results, microscopic appearance, and catalase and oxidase production in accordance with the standard methods at the National Food Safety and Toxicology Center, Michigan State University.¹⁸ Isolates were identified as Campylobacter spp if they were gram negative with a typical curved appearance microscopically, grew at 42°C under microaerophilic conditions, and were positive for catalase and oxidase. Campylobacter jejuni was further characterized by positive results of a hippurate hydrolysis test.¹⁹ Because some species of Campylobacter can be difficult to distinguish, a colony m-PCR assay was also used for isolate identification. Campylobacter isolates were frozen in 2% skimmed milk at −70°C in preparation for antimicrobial susceptibility testing and final identification by the use of the m-PCR procedure.²⁰

Differentiation of Campylobacter spp—Pure cultures of Campylobacter spp were thawed at room temperature (approx 20°C). Subsequently, Campylobacter spp were differentiated by use of a colony m-PCR with slight modifications²⁰; the original PCR assay also identified Campylobacter upsaliensis and Campylobacter fetus subsp fetus. In brief, the m-PCR procedure identified the 23S rRNA from Campylobacter spp, the hipO gene (hippuricase) from C jejuni, and the glyA gene (serine hydroxymethyltransferase) from C coli and from C lari by use of specific primer pairs (Appendix).²⁰ Primers and reagents were used in a 50-μL PCR system. The m-PCR assay mixture contained 1X Taqman buffer, 0.2mM deoxyribonucleotide triphosphate mix, 7.5mM MgCl₂, 0.5μM C lari glyA forward and reverse primers, 0.5μM C jejuni hipO forward and reverse primers, 1.0μM C coli glyA forward and reverse primers, 0.2μM C jejuni 23S rRNA forward and reverse primers, 0.05 U/μL (2.5 units) Taq DNA polymerase,^d and approximately 10⁶ whole bacterial cells. Amplification was achieved by use of a thermocycler^e with an initial denaturation step at 95°C for 6 minutes. An additional denaturation step at 95°C for 30 seconds followed by annealing at 59°C for 30 seconds and polymerization at 72°C for 30 seconds was repeated for 30 cycles. Final extension was carried out at 72°C for 7 minutes. Polymerase chain reaction products were separated on a 1.5% agarose gel at 90 V for 2.25 hours with ethydium bromide (0.5 μg/ mL) added to the Tris, boric acid, EDTA buffer.

Assessment of AMR—Antimicrobial susceptibility testing was performed by use of commercially available gradient disk diffusion strips^f according to the manufacturer's instructions. Frozen bacterial isolates were thawed at room temperature, inoculated onto blood agar plates, and incubated at 42°C in a microaerophilic atmosphere^c for a minimum of 48 hours. Typical colonies were selected and subcultured on plates containing trypticase soy agar with 5% sheep blood.^g These plates were incubated under microaerophilic conditions at 42°C for 48 hours. Colonies from subculture were tested as described by Sato et al.²¹ Six antimicrobials were tested: azithromycin (0.016 to 256 μg/mL), erythromycin (0.016 to 256 μg/mL), ciprofloxacin (0.002 to 32 μg/mL), nalidixic acid (0.016 to 256 μg/mL), gentamicin (0.016 to 256 μg/mL), and tetracycline (0.016 to 256 μg/mL). The gradient disk diffusion strips provided 29 possible MIC values for each antimicrobial drug tested. For each antimicrobial drug, there were 15 possible log₂ dilutions on a strip (eg, 0.016 through 256) and intermediate values between each log₂ dilution. Intermediate values between log₂ dilutions were rounded up to the higher log₂ dilution during poststudy data management, as recommended by the manufacturer. Campylobacter jejuni (ATCC 33560²²) and E coli (ATCC 25922) were used as quality control strains. Resistance breakpoint²³ is defined as the MIC at which a bacterial isolate is considered resistant to a particular antimicrobial drug. Resistance breakpoints used by the National Antimicrobial Resistance Monitoring System were adopted.²⁴ The resistance breakpoints were azithromycin (≥ 2 μg/mL), erythromycin (≥ 8 μg/mL), ciprofloxacin (≥ 4 μg/mL), nalidixic acid (≥ 32 μg/mL), gentamicin (≥ 16 μg/mL), and tetracycline (≥ 16 μg/mL).

Statistical analysis—Data regarding Campylobacter isolates, AMR, and farm management factors were compiled in a commercially available database software program.^h Apparent prevalence is reported as a proportion with 95% exact CIs. Results of susceptibility testing are reported as MIC distributions and proportions of resistant and susceptible isolates according to the Clinical and Laboratory Standards Institute guidelines.²⁵ Statistical analysis was performed by use of a commercial software package.ⁱ Logistic regression analysis was used to test the associations between resistant isolates for each of the 6 antimicrobial agents and production method (ie, conventional or antimicrobial-free farms).²⁶ At the individual animal level, a population-averaged logistic regression model involving a generalized model framework with a logit link and binomial error distribution was used to determine the potential association between the proportion of resistance for each of the 6 antimicrobial agents and production method.²⁷ A generalized estimating equation involving an exchangeable working correlation structure and semirobust variance estimator was used to model withinfarm dependence.^26,27 Potential confounding effects by herd size and season were assessed for each antimicrobial agent. Season was defined as winter (January through March), spring (April through June), summer (July through September), and fall (October through December). Herd size was defined as the total number of finisher pigs marketed per year. Herd size was then dichotomized at a cutoff of 2,000 animals. The generalized Wald test was used to test significance (set at a value of P < 0.05) of independent variables in the models. Potential confounding variables were assessed by comparison of the differences in the regression coefficients with and without the presence of the potential confounder in the model. If there was a change of 20% or more, then adjusted measures of association were reported.²⁷ Additionally, a possible dose-dependent relationship between the number of years that antimicrobial drugs were not used on antimicrobial-free farms, and the level of resistance to the 6 antimicrobial drugs was investigated.

Pan-susceptible isolates were defined as those susceptible to all 6 antimicrobial drugs. Multidrug resistance was defined as resistance to 2 or more antimicrobial drugs. We assessed multidrug resistance using 2 approaches: specific and nonspecific MDR patterns. Nonspecific MDR was defined as resistance to any combination of ≥ 2 antimicrobials. Specific MDR was defined as resistance to a specific combination of ≥ 2 antimicrobials (eg, azithromycin-erythromycin-tetracycline).

Results

Fecal samples were collected from 1,422 pigs on antimicrobial-free (n = 35) and conventional (60) swine farms in the Midwestern United States. The number of years that antimicrobial drugs were not used on farms ranged from 1 to 14, with a median of 3 years. The mean number of pigs from farms that were considered antimicrobial-free production systems was 1,262 (range, 150 to 11,000; median, 800), whereas the mean number of pigs from conventional farms was 7,909 (range, 500 to 45,000; median, 4,800; P < 0.001). The proportions of antimicrobial-free and conventional farms evaluated in each season were as follows: winter, 11 of 35 (31%) and 22 of 60 (37%) farms, respectively; spring, 6 of 35 (17%) and 13 of 60 (22%) farms, respectively; summer, 8 of 35 (23%) and 9 of 60 (15%) farms, respectively; and fall, 10 of 35 (29%) and 16 of 60 (27%) farms, respectively.

Apparent prevalence of Campylobacter spp—Culture results were positive for Campylobacter spp for 1 or more pigs on 90 of the 95 (94.7% [95% CI, 88.1% to 98.3%]) farms included in the study (Table 1). Among the 35 antimicrobial-free farms, 33 (94.3% [95% CI, 80.8% to 99.3%]) had 1 or more Campylobacter-positive samples, and among the 60 conventional farms, 57 (95.0% [95% CI, 86.1% to 99.0%]) had 1 or more Campylobacter-positive samples. Across all farms, 512 fecal samples (36.0% [95% CI, 33.5% to 38.6%]) were positive for Campylobacter spp. Among antimicrobial-free farms, 190 of 522 (36.4% [95% CI, 32.3% to 40.7%]) samples were positive for Campylobacter spp. Among conventional farms, 322 of 900 (35.8% [95% CI, 32.6% to 39.0%]) isolates were Campylobacter spp. The herd-level and individual animal–level apparent prevalences were not significantly different between antimicrobial-free and conventional farms. In addition, herd size was not associated with apparent prevalence. The m-PCR assay was performed on 427 of the 512 isolates, and they were identified as C coli (n = 426 [99.6%]) and C jejuni (1 [0.4%]).

Antimicrobial susceptibility—Of the 512 Campylobacter spp isolates, 464 (90.6%) were available for antimicrobial susceptibility testing; these isolates were obtained from 30 of 33 (90.9%) antimicrobial-free farms and 55 of 57 (96.5%) conventional farms that had ≥ 1 pig with positive culture results. Forty-eight (9.4%) samples across all samples were not recoverable after storage at −70°C; the unrecoverable samples included 16 of 190 (8.4%) samples collected from antimicrobial-free farms and 32 of 322 (9.9%) samples collected from conventional farms. Five (3/33 [9.1%] antimicrobial-free farms and 2/55 [3.6%] conventional) farms on which Campylobacter spp were isolated from at least 1 pig had at least 1 sample that was not available for susceptibility testing.

At the farm level, the proportion of farms with 1 or more Campylobacter isolate resistant to azithromycin or to erythromycin was significantly (P < 0.001) higher for conventional farms, compared with antimicrobial-free farms (Table 2). The number of herds with at least 1 ciprofloxacin- or nalidixic acid–resistant isolate was higher for antimicrobial-free farms, compared with conventional farms. Conversely, the individual animal apparent prevalence of resistance to ciprofloxacin or nalidixic acid was greater on conventional farms (Table 3).

The distributions of MICs were bimodal for azithromycin, erythromycin, ciprofloxacin, and nalidixic acid (Table 4). The distributions of MICs for tetracycline was almost uniform across the various dilutions. Across farm type, significantly more Campylobacter isolates had a higher apparent prevalence of resistance to azithromycin, erythromycin, or tetracycline on conventional farms, compared with findings on antimicrobial-free farms (P < 0.001). For the macrolide antimicrobials erythromycin and azithromycin, the MIC₅₀ value for each drug was 256 μg/mL for isolates obtained from conventional farms; for isolates obtained from antimicrobial-free farms, the MIC₅₀ for azithromycin and erythromycin was 0.5 and 2 μg/mL, respectively. The MIC₅₀ values for ciprofloxacin and nalidixic acid did not differ significantly between the 2 production systems, and none of the 464 isolates were resistant to gentamicin.

Inclusion of season as a potential confounder in the statistical model did not change the association between production system and AMR. Inclusion of herd size in the model changed the overall effect of production method (conventional and antimicrobial-free) on AMR prevalence by more than 20%; therefore, herd size was considered a confounder. To account for the confounding effect of herd size on the model, herd size was forced into each model for all 6 antimicrobial drugs. In addition, herd size was added to the models investigating nonspecific and specific resistance patterns. However, there was no significant interaction between herd size and production system type. In the present study, nondifferential misclassification was unlikely since the culture and MIC methods were equivalent for antimicrobial-free and conventional farms. In either case, the effects of nondifferential misclassification would likely bias the estimates of association in this study toward a null (a more conservative P value).

As the number of years that an antimicrobial-free production scheme was implemented on a farm increased, there was a significant (P = 0.002 for the first 2 years then P < 0.001 for years 3 to 15) and consistent decrease by year in the proportion of isolates that were resistant to azithromycin or erythromycin (Table 5). Resistance to tetracycline did not decrease consistently as the duration of antimicrobial-free production increased, but after 3 years, the number of resistant strains was significantly (P < 0.001) less, compared with the number of resistant strains on conventional farms. On antimicrobial-free farms on which antimicrobial drugs had not been used for 6 or more years, the apparent prevalence of resistance to tetracycline was 40% less than that of conventional farms; the apparent prevalences of resistance to azithromycin and erythromycin were each 83% less than that of conventional farms.

Ten specific resistance patterns to 2 or more of 5 antimicrobial drugs were identified (none of the isolates were resistant to gentamicin; Table 6). The most common pattern was resistance to azithromycin, erythromycin, and tetracycline, which was significantly (P < 0.001) higher on conventional farms than on antimicrobial-free farms. The proportion of pan-susceptible isolates was higher on antimicrobial-free farms (42.5% [95% CI, 35.1% to 49.0%]), compared with the proportion on conventional farms (7.9% [95% CI, 4.8% to 11.1%]; Figure 1). Across both production systems, 1 isolate was resistant to 4 (azithromycin, erythromycin, nalidixic acid, and tetracycline) antimicrobial drugs, and 1 isolate was resistant to 5 (azithromycin, erythromycin, nalidixic acid, tetracycline, and ciprofloxacin) antimicrobial drugs.

View Full Size

Nonspecific MDR among 464 Campylobacter isolates from 35 antimicrobial-free (black bars) and 60 conventional (white bars) swine farms.

Citation: Journal of the American Veterinary Medical Association 236, 2; 10.2460/javma.236.2.201

• Download Figure
• Download figure as PowerPoint slide

Discussion

In the present cross-sectional study, Campylobacter spp were isolated from approximately a third of samples collected on both conventional and antimicrobial-free swine farms. This is within the previously reported range of Campylobacter spp apparent prevalence among finishing pigs (16% to 100%),^5,6,28 including findings of 1 study²⁹ that compared prevalence in antimicrobial-free and conventional production systems (53% and 55.8%, respectively). A similar study²¹ in cows also did not identify a significant difference in prevalence between the 2 production systems. Often, shedding of pathogens is greater in larger herds³⁰; however, in our study, Campylobacter spp apparent prevalence was not associated with herd size. In addition, 95% of all farms had at least 1 Campylobacter-positive pig, which suggests that Campylobacter spp are widespread. The results of the present study further emphasize that pigs are common reservoirs for Campylobacter spp, regardless of production system and herd size.

At the farm level, resistance of Campylobacter spp to azithromycin or erythromycin for 1 or more individual pigs/farm was detected on most of the conventional farms, yet resistance to each of these macrolides was detected on approximately half as many antimicrobial-free farms. The lack of recent exposure to macrolides may have contributed to the lower number of antimicrobial-free farms with resistance to macrolides because of a reduction in selective pressure. Resistance to macrolides may confer a fitness cost (a decrease in the ability of a bacterium to compete with other bacteria in the environment) that would cause bacteria that acquire additional resistance genes to become less fit.¹⁴ Tetracycline resistance was evident on almost all conventional and antimicrobial-free farms (7% difference). This may result from mutations that confer resistance without reducing the fitness of the bacteria, or from environmental persistence of plasmid-mediated resistance genes associated with Campylobacter spp resistance to tetracycline.^14,31 Resistance to ciprofloxacin or nalidixic acid was rare and was evident on more antimicrobial-free than conventional farms; this may have resulted from unidentified mechanisms, possibly including adaptation of resistant strains or the presence of efflux pumps.³² These efflux pumps limit access of antimicrobial drugs to their targets by actively pumping out these molecules.³³

At the animal level, the highest apparent prevalences of AMR were to erythromycin, azithromycin, and tetracycline. This finding was similar to the results of other studies.^6,29 Resistances to the macrolide antimicrobial drugs (azithromycin and erythromycin) were approximately 70% higher on conventional than antimicrobial-free farms. In addition, conventional farms had a higher proportion of isolates resistant to high concentrations of macrolides (MIC ≥ 256 μg/mL). One explanation for the high prevalence of macrolide resistance may be the use of tylosin, a macrolide, which is approved for use for growth promotion and therapeutic purposes in swine.^34,35 Antimicrobial-free farms that lack exposure to macrolides might be expected to eliminate the high-concentration–resistant strains first.

The high level of erythromycin and azithromycin resistance on conventional farms is of concern because erythromycin and azithromycin are currently the most common antimicrobial treatments for Campylobacter infections in humans.^11,36 Erythromycin and azithromycin resistances result from a chromosomal mutation of the ribosome 23S rRNA genes or genes encoding ribosomal proteins L4 and L22, not from horizontally acquired genes from other bacteria.^34,37 Erythromycin and azithromycin resistances are rare in C jejuni but common among C coli strains, particularly among isolates from pigs and pig offal.³⁸ The higher prevalence of resistance to erythromycin in pigs has not been fully explained; however, Engberg et al³⁷ suggest this may be due to a generally higher frequency of mutations conferring resistance among C coli, or due to greater selective pressure resulting from prior use of antimicrobial agents. In addition, C coli also has an efflux pump system that contributes to acquired resistance to macrolides.³⁹

The antimicrobial drug with the highest apparent prevalence of resistance on antimicrobial-free and conventional farms was tetracycline (49% and 75%, respectively). The high apparent prevalence of tetracycline resistance is most likely due to the presence of the genetic determinant tet(O) on transferable plasmids that prevent tetracycline from binding to the ribosome, as well as the presence of efflux pumps.^40,41 The most common mechanism involves the plasmid encoded tet(O) gene, which produces a ribosomal protection protein that confers resistance by preventing tetracycline from binding to the ribosome.^2,40 Tet(O) is commonly found in a variety of bacteria in farming environments³¹ and in pig samples, regardless of prior antimicrobial usage.⁴² Comparison of C jejuni and C coli isolates derived from humans⁴⁰ established that all tet(O) genes among C coli were chromosomally related, rather than carried by plasmids as is the case for C jejuni. If tet(O) genes are chromosomally related among C coli derived from swine, then this is an important distinction that should be further investigated in food animals, because C coli is the predominant subspecies in swine. Also if tet(O) genes are chromosomally related in C coli derived from swine, this may then explain epidemiological differences between swine and other food animals. In addition to transferable plasmids, the multidrug efflux pump CmeABC contributes to intrinsic and acquired resistance.^39,43 The multiple and complex resistance mechanisms of tetracycline are a likely explanation for the high proportion of resistant isolates and the broad characteristic MIC values observed for tetracycline.

The present and previous studies^29,44 have identified resistance of Campylobacter spp to ciprofloxacin in both conventional and antimicrobial-free farming systems in North America. The presence of fluoroquinolone resistance in both production systems is of particular concern because this class of drugs was not approved for use in swine production at the time of our study.⁴⁵ In addition, in a study⁴⁶ of chickens, fluoroquinolone-resistant Campylobacter spp colonized and persisted in chickens as efficiently as susceptible strains in the absence of fluoroquinolone antimicrobials. The gyrA and parC genes are responsible for production of DNA gyrase and toposiomerase IV, the proteins that are targets for fluoroquinolones. Campylobacter spp do not produce toposiomerase; hence a single mutation in gyrA gene can cause a high level of resistance to fluoroquinolones (≥ 32 μg/mL).^46,47 Furthermore, the most frequently reported mechanism of resistance to fluoroquinolones is the target mutation of the gyrA gene; at least 4 unique point mutations in the gyrA gene of the fluoroquinolone-resistant mutants, resulting in high and intermediate levels of resistance of Campylobacter spp to the fluoroquinolones, have been reported.^34,48 In addition, the CmeABC efflux pump is associated with fluoroquinolone resistance in Campylobacter spp.^46,49 In the present study, isolates resistant to ciprofloxacin and nalidixic acid were distributed between 4 antimicrobial-free farms and 1 conventional farm that, combined, had 11 pigs with ciprofloxacin-resistant Campylobacter spp. Hence, ciprofloxacin resistance among Campylobacter spp appears to be present only on certain farms. Furthermore, it is unknown how long ciprofloxacin-resistant Campylobacter organisms have been present on the antimicrobial-free farms in our study. In poultry, resistance of Campylobacter spp to fluoroquinolones persisted for at least 4 years after cessation of antimicrobial usage.⁵⁰ In our study, the data are insufficient to make inferences regarding exposure and resistance. Further studies should concentrate on examination of risk factors that might be expected to promote the presence or persistence of ciprofloxacin resistance on swine farms.

An apparent dose-response effect was observed for the duration of antimicrobial-free production (1 to 14 years). The gradual wane in azithromycin and erythromycin resistances over time was expected because their resistance mechanisms have a chromosomal linkage and would only be transmitted vertically. Following mutation, there is often a fitness deficit of the bacteria conferred by resistance⁵¹; therefore, susceptible strains may become more predominant over time in the absence of antimicrobial pressure. However, in our study, tetracycline resistance had a threshold decline after 3 years, or in other words, tetracycline resistance did not decline until a farm was antimicrobial free for 3 or more years. The large variety of mechanisms of tetracycline resistance among Campylobacter spp isolates may explain why there was only a 40% decrease in tetracycline resistance on farms that were antimicrobial free for ≥ 6 years, compared with findings on conventional farms; in contrast, an 80% decrease in erythromycin resistance and an 83% decrease in azithromycin resistance was detected between those farm types. Ciprofloxacin and nalidixic acid did not have a sufficient number of resistant isolates to detect a pattern.

Considering the predominant mechanism of resistance for each antimicrobial tested, the resistance patterns detected in the present study were expected. However, we compared 2 production methods at a single point in time, so the assumption was made that antimicrobial-free farms had AMR prevalences similar to those on conventional farms prior to the cessation of antimicrobial use. Although caution is needed in making inferences about a true dose effect, these patterns can serve to generate hypotheses regarding why resistance to some antimicrobials but not to others appears to change over time.

Multidrug resistance was common in the present study. In 3 other studies,^6,28,29 the most common MDR in C coli in pigs was the combination of erythromycin, nalidixic acid, and tetracycline. In our study, this combination was also present on conventional farms (0.34% of total MDR combinations). Two isolates were resistant to 4 or 5 antimicrobial agents, including erythromycin, ciprofloxacin, and tetracycline, which may be used to treat human infections. Multidrug resistance in Campylobacter spp is most commonly due to the presence of multidrug efflux pumps, which contribute to the intrinsic resistance of Campylobacter spp to a broad range of structurally unrelated antimicrobial agents.^6,38,43 As previously noted, resistances of Campylobacter spp to fluoroquinolones and macrolides result from mutations of the gyrA or 23S rRNA gene, respectively. In a recent review, Payot et al⁴⁸ concluded that the CmeABC efflux system works synergistically with these mutations to confer high-level resistance to fluoroquinolones and macrolides. However, in the present study, only 2 of 17 (11.8%) Campylobacter isolates that were resistant to fluoroquinolones were also resistant to erythromycin or azithromycin. At present, the mechanisms of MDR in Campylobacter spp are still incompletely understood; however, it appears that the role of efflux pumps should be a focus of further research in this area.

Campylobacteriosis in humans is primarily associated with consumption of food animal products.⁵² Intuitively, removal of antimicrobials from a production system should decrease AMR. In the study reported here, decreased AMR to erythromycin, azithromycin, and tetracycline was observed on antimicrobial-free farms. However, one issue with cross-sectional studies is that the rate of decrease in resistance cannot be directly quantified. In our study, the assumption was made that prior to cessation of antimicrobial use on antimicrobial-free farms, the proportions of Campylobacter spp resistant to the antimicrobials tested were the same as the proportions on conventional farms. The cessation of antimicrobial use is a major production change, the benefits of which have yet to be fully examined. The changes in risk factors associated with this production change may inherently affect the outcome. For example, antimicrobial-free farms are typically small and may use different management procedures that may affect risk factors differently than on conventional farms. Results of the present study suggest that AMR is greater on conventional farms; long-term prospective studies are indicated to examine whether these differences persist, and to compare specific risk factors in conventional farming environments with antimicrobial-free farms that lack antimicrobial selection pressure.