Contaminated or no-growth inoculated samples were not read and repeated. In addition, quality control strains were run weekly alongside the test samples. For anti-microbials in which the BOPO7F Vet AST Plate dilutions included the established breakpoint, “resistant” status was assigned if the isolate grew in or beyond the breakpoint dilution . For antimicrobials in which the testing plate included only dilutions below the established breakpoint, “non-susceptible” status was assigned and included isolates in the intermediate range according to CLSI guidelines or isolates that grew in the highest dilution available. Resistance or non-susceptible status was only assigned to antimicrobials for which breakpoints were available and for which in-vivo activity and antimicrobial spectrum were applicable. For antimicrobials that were assigned non-susceptible status , it was not possible to establish resistance because the drug dilutions did not reach the threshold breakpoint; hence growth or no growth at or beyond the breakpoint could not be established. Antimicrobial breakpoints used and dilution ranges for the BOPO7F Vet AST Plate can be found in Table 1.Data from the ranch survey, individual animal data, pipp racking and AST results were entered into a spreadsheet and combined using a relational database . Descriptive statistics for ranch demographics and prevalence of resistance or non-susceptibility for antimicrobials with existing breakpoints were prepared. Univariable generalized linear mixed models with a logit link were prepared for the outcome of resistance or non-susceptibility status of isolates to each antimicrobial with available breakpoint data using the GLIMMIX procedure in SAS .
A random effect was added to account for correlation between isolates from the same animal, since 2 isolates from each fecal sample were required for inclusion in the MIC analysis. A second random effect of farm with animal nested within farm was attempted but led to non-positive G matrices and not explored further. The independent variables were created from the questionnaire data on herd demographics, antimicrobial practices, treatment history, and management practices on the farm. Multivariable generalized linear mixed models were attempted by including all variables from the univariable analysis with p<0.2. A multiple factor analysis was conducted for survey data and antimicrobial susceptibility testing results of the 244 E. coli isolates. MFA was conducted to reveal the most important variables that explain the variation in the data set . The dataset consisted of 63 data variables which were organized into 6 groups based on relatedness as follows: herd information: a group of 7 categorical variables specifying farm number, the location of farm, breed distribution, herd size, certification status , type of pasture, and type of production; sampled animals’ life stage and treatment history: a group of 7 categorical variables specifying sampled animal life stage, method of fecal sample collection , date of fecal sample collection, whether animal was treated with antimicrobials, and antimicrobial used for treatment ; antimicrobial resistance group: a group of 8 variables describing AMR for E. coli ; AMR for E. coli; farm antimicrobial use and disease treatment group: a group of 17 categorical variables describing the different injectable and intramammary antimicrobial drugs used in farms and type of treated diseases ; antimicrobial dosing and record keeping practices: a group of 12 variables describing methods used for determining treatment duration and dosage, and information recorded regarding antimicrobial treatment ; and nutrition related factors: a group of 12 categorical variables specifying the provision of byproducts and mineral supplement to calves, and cows .
The groups with loading weights of 0.5 or higher on the first two principal components were retained for interpretation . The percentage of variability contributed by each group of variables to the principal components and the correlation coefficients for the component variables within each group were estimated . Variables within each group with loading weights of ≥0.5 on the first two principal components were also retained for interpretation. The function MFA in FactoMiner package was used to perform the MFA on the dataset. The function get_mfa_var was used to extract the results for the groups and variables. Hierarchical clustering was performed on the MFA principal coordinates using the principal component methods at the animal level . The identified clusters were described based on the variables that contributed the most to the data variability. Both MFA and hierarchical clustering were performed in R software using FactoMineR for the analysis and factoextra for data visualization . MFA analysis was not performed for Enterococcus data due to the limited number of resistant and non-susceptible isolates.A total of 18 cow-calf farms were surveyed and sampled during this study. General descriptive data including the major breed, herd size, pasture type, location, antimicrobial practices, and the number of injectable or oral antibiotics used on farm is shown in Table 2. Other management survey results of interest revealed that most farms had at least one beef quality assurance certified employee, one farm fed byproducts, 7 farms had submitted samples to a diagnostic lab in the past year, and 17 had an established veterinarian-client-patient relationship. Oxytetracycline was the most common antimicrobial used on farm , followed by tulathromycin , florfenicol , sulfas including sulfadimethoxine and sulfamethoxazole , penicillin , enrofloxacin , and ceftiofur . No farm reported using danofloxacin or ampicillin.
Regarding the types of diseases that had been treated with antimicrobials in the past 12months in any cattle on the farm, 13 farms reported treating infectious bovine keratoconjunctivitis , 13 reported treating bovine respiratory disease, 10 reported treating foot rot, 7 reported treating scours, 6 reported treating wounds, 5 reported treating navel infections, 3 reported treating metritis, and 2 reported treating mastitis. There were 2 farms that reported no antimicrobial use because no disease identified as needing treatment was observed during the past year. Only one farm had routine prophylactic use of antibiotics where all calves received an injection of oxytetracycline between 1week and 1month of age, and all farms that used antimicrobials recorded at least one form of information after antimicrobials were administered such as date, dose, route, withdrawal, and/or product name.In total, fecal samples were collected from 187 animals and plated for growth and recovery of E. coli and Enterococcus isolates. A total of 244 E. coli isolates and 238 Enterococcus isolates were recovered and tested for antimicrobial susceptibility using broth microdilution method. Of the 104 cow samples plated, 50 samples grew at least 2 isolates of E. coli and 50 samples grew at least 2 isolates of Enterococcus. Of the 83 calf samples plated, 72 samplesgrew at least 2 isolates of E. coli and 69 samples grew at least 2 isolates of Enterococcus. Details regarding the number of samples and resulting isolates can be found in Figure 1.The distribution of isolates within various drug dilutions tested for each antimicrobial can be found in Table 3. Resistance or non-susceptible data is only shown for those antimicrobials for which established breakpoints by CLSI were available, including ampicillin, ceftiofur, florfenicol, sulfadimethoxine, tetracycline, and trimethoprim-sulfamethoxazole . Among the 244 E. coli isolates, 88/244 were resistant or non-susceptible to at least one antimicrobial excluding ampicillin, to which all isolates were resistant. Similarly, a large proportion of isolates showed antimicrobial resistance or non-susceptibility to sulfadimethoxine followed by trimethoprim-sulfamethoxazole, while the lowest proportion of isolates showed antimicrobial resistance to ceftiofur. More isolates were classified as non-susceptible to tetracycline than florfenicol. Neither univariable nor multivariable generalized linear mixed models revealed any statistically significant associations between any of the risk factors considered, vertical grow racks including record of antimicrobial therapy with the same antimicrobial in the past 6 months, and resistance or non-susceptible isolate status. Although none of the farm-specific variables captured in this study were significantly associated with differences in resistance or non-susceptibility, there were numerical differences between farms in terms of their antimicrobial resistance profile for E. coli isolates. Specifically, the highest percentage of resistant or non-susceptible isolates for florfenicol , tetracycline , and trimethoprimsulfamethoxazole at the farm level was found on Farm 6, which contributed 14 isolates. Interestingly, Farm 6 did not report the use of any antimicrobials on farm.The distribution of isolates within MICs tested for each antimicrobial can be found in Table 4. Resistance or non-susceptible data is only shown for those antimicrobials for which established CLSI breakpoints were available, including ampicillin, penicillin, and tetracycline . Only a small proportion of the total 238 Enterococcus isolates, 35/238 were resistant or non-susceptible to at least one antimicrobial. Amongst all isolates tested, antimicrobial non-susceptibility was highest to tetracycline, followed by non-susceptibility to penicillin, and lowest resistance to ampicillin. Similar to the statistical models for the E. coli isolates, no significant associations between any of the risk factors and AMR status for Enterococcus isolates was found.The first two principal component dimensions of the multiple factor analysis explained approximately 8.5% of the variability in the data, i.e., 4.4 and 4.1% of the variance for the first and second principal component dimensions, respectively. The MFA analysis of 63 variables identified four components and 16 variables with a correlation coefficient≥0.5 on both first and second dimensions that accounted for 98.7% of the variability in the data . Herd information accounted for 27.7% of the total variability in the data, while antimicrobial dosing and record keeping practices accounted for approximately 25% of the total variability in the data.
Nutrition related factors and farm antimicrobial use and disease treatment accounted for 24.2 and 21.6% of the total variability in the data, respectively . The sampled animals’ life stage and treatment history as well as antimicrobial resistance data were groups of variables where correlation stayed below the threshold of 0.5.Hierarchical clustering was performed on the MFA principal coordinates to aggregate homogeneous clusters. The hierarchical tree suggested clustering into six clusters . The identified clusters were described based on the 16 variables that contributed the greatest to the data variability from the MFA analysis . Cluster 5 represented the majority of sampled animals and ranches . Most animals represented in cluster 5 were on ranches that reported estimation of the dose of antimicrobial drugs based on estimated animal weight, reported recording the date of antimicrobial use , reported feeding free choice minerals to calves , reported cleaning of water troughs , and did not use antimicrobials to treat mastitis . However, 91.8% of animals in cluster 5 were on ranches that also reported that withdrawal periods are not recorded when animals are treated with antimicrobials. Cluster 4 represented two ranches in our study. The farms represented in cluster 4 mentioned that they were not recording the date, route, and withdrawal period of antimicrobial use . One farm represented in cluster 2 mentioned routine use of antimicrobials for prevention of disease and use of antimicrobials for treatment of mastitis. Clusters 1, 3, and 6 represented one herd each. Farms represented in clusters 3 and 6 reported that they were not using antimicrobials for treatment of mastitis and the farm in cluster 6 reported dosing antimicrobials according to veterinarian’s orders. The beef operations located in the coastal range were only represented by clusters 5 and 6. The majority of beef ranches in clusters 5 and 6 reported several antimicrobial stewardship or herd health practices including estimation of the dose of antimicrobials based on estimated animal weight, recording of the date of antimicrobial use, feeding free choice mineral to calves, and cleaning of water troughs once a month, and did not use antimicrobials to treat mastitis in comparison to beef ranches included in clusters 1, 2, 3, and 4. A complete description of the six clusters is available in Supplementary Table S1.Antimicrobial resistance is a global problem , and while much of the attentions is focused on human health implications, the effects of AMR on livestock health may be similar,including treatment failures requiring the use of newer and often more expensive antimicrobials . For our study, the distribution of herd sizes closely represented what has previously been reported for cow-calf operations throughout the state of California . Considering the state’s number of beef cow farms, however, our study included a higher proportion of larger herd sizes for the state, since approximately 77% of beef cow farms in California are reported to have fewer than 100 cows, not including hobby farms with less than 10 cows . Information about the percentage of different breeds, type of production , and/or Age and Source Verified versus conventional) for beef cow-calf herds or specific antimicrobial practices have not been previously reported.