Collection of individual AMU records

Cow-calf herds were recruited from the Canadian Cow-Calf Surveillance Network (C3SN). The C3SN was established in 2018 and includes producers from across Canada.²² Herds recruited to the network were to have at least 40 breeding animals, conduct pregnancy testing, and maintain basic calving production records. Herd management, attribute, and productivity data were previously reported.²²

In the spring of 2020, producers were asked if they would be willing to share their 2021 individual animal treatment records. Interested producers were sent a binder containing carbonless treatment records. One copy of the completed records was to be returned to the research team. Producers also had the option to share original AMU records generated by their on-farm digital records system or copies of other handwritten AMU records.

Collection of AMR data

Herd owners willing to share 2021 animal health records were contacted to determine their interest in collecting fecal samples. Due to the logistics of sample shipment, only producers from Western Canada were eligible to participate in the fecal sample portion of the study. Fifty herds were originally recruited from across Western Canada, with individual fecal samples to be collected from 10 calves and 10 cows in the spring and fall of 2021. As previously described,^8,9 the fecal samples were weighed, and 4.0 g was transferred into 50-mL centrifuge tubes containing 1% buffered peptone water. The mixture was vortexed thoroughly and placed in a mixer for 1 hour. The pre-enrichment mixtures were incubated at 35 °C for 18 to 24 hours under ambient atmospheric conditions.

Briefly for E coli, MacConkey agar (Difco Laboratories) was inoculated with 10 μL of the pre-enrichment mixture and incubated at 35 °C for 18 to 24 hours under ambient atmospheric conditions. After incubation, individual lactose-fermenting colonies were then subcultured onto 5% sheep blood agar to obtain pure colonies. The subcultured plates were incubated at 35 °C for 18 to 24 hours under 5% CO₂. The colonies grown on blood agar were identified using matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF MS; Bruker) following manufacturer guidelines. Only isolates that had secure species identification based on a MALDI-TOF MS score > 2 were selected for further analysis.

A selective medium for Enterococcus spp, m-Enterococcus agar (Oxoid; Fisher Scientific), was inoculated with 10 μL of the pre-enrichment mixture and incubated at 35 °C for 48 hours under 5% CO₂ conditions. Up to 6 colonies exhibiting different morphological characteristics (color: pink or red; size: tiny, small, or large) per plate were subcultured on Columbia agar with 5% sheep blood to obtain pure colonies. The subcultured plates were incubated at 35 °C for 18 to 24 hours under 5% CO₂ conditions. The colonies grown on blood agar were identified using MALDI-TOF MS as described for E coli.

One E coli and 1 Enterococcus spp isolate from each sample was analyzed for AMR, with detailed methods and complete susceptibility testing results for all submitting herds published elsewhere.^8,9 Enterococcus spp isolates were selected for antimicrobial susceptibility testing in order of priority as follows: Enterococcus faecalis, Enterococcus faecium, Enterococcus hirae, and then, if none of these were identified, the next most common Enterococcus spp identified on each plate. The range of tested dilutions and Clinical and Laboratory Standards Institute standard breakpoints used to categorize isolates as resistant are reported in Supplementary Tables S1 and S2.

Minimum inhibitory concentrations were measured for E coli using a National Antimicrobial Resistance Monitoring System CMV5AGNF plate for gram-negative bacteria (TREK Diagnostic Systems) and for Enterococcus spp using National Antimicrobial Resistance Monitoring System CMV3AGPF Sensititre plates recommended for determining MICs for gram-positive bacteria (Thermo Fisher Scientific). A bacterial broth (0.5 McFarland turbidity equivalent) was prepared from a pure bacterial isolate using the Sensititre Nephelometer (ThermoFisher Scientific). The dosing broths were prepared by transferring 10 μL of the suspension to 11 mL of a Sensititre Cation Adjusted AutoRead Muller-Hinton Broth w/TES (ThermoFisher Scientific). The dosing broth was inoculated onto a Sensititre plate using a Sensititre AIM Automated Inoculation System (ThermoFisher Scientific). The inoculated plates were incubated at 35 °C for 18 to 24 hours under normal atmospheric conditions. Minimum inhibitory concentrations were interpreted using a BioMic V3 system (Giles Scientific) and a manual mirror box confirmation if necessary to ensure growth was present. The MIC was determined by evaluating the panel for the first well without visible growth. Bacteria were classified as either susceptible or resistant, with intermediate MIC values classed as susceptible.

Statistical analysis

Summary of AMU data—All individual animal AMU data were entered into a commercial spreadsheet, then the data for nursing calves were recoded from producer descriptions, typically the trade name or an abbreviation, to the generic names for antimicrobials. Records of drug use that were not antimicrobials were removed from the analysis. Records were then further categorized by antimicrobial class and then category of importance to human health (category I, very high importance; II, high importance; III, medium importance).²³ Examples of antimicrobial drug classes considered included third-generation cephalosporins, fluoroquinolones, and topical polymyxins (category I); macrolides, sulfadoxine/trimethoprim, penicillins, aminoglycosides (oral bolus), and first-generation cephalosporins (topical; category II); and amphenicols, tetracyclines, and sulfonamides (oral bolus; category III).

All producers providing records were classified based on whether they had used the reported antimicrobials, and then the total number of uses in calves before weaning were summarized for each generic drug for each herd. The cumulative incidence of AMU within herd was determined as the total number of reported uses in calves before weaning as a percentage of the total calves born alive for each generic antimicrobial, each drug class of antimicrobial, each category of importance to human health, and reported use of any antimicrobial.

Assessment of risk factors for AMU in nursing calves—Univariable associations between individual herd attributes and the cumulative incidence of AMU were examined using a series of generalized estimating equations (GEEs) with a logit link function and binomial distribution accounting for the clustering of AMU by herd with a repeated term and exchangeable covariance structure (PROC GENMOD; SAS, version 16.1; SAS Institute Inc). Cumulative incidence was described in the model with the total number of AMU treatments recorded as the numerator and the number of calves born alive as the denominator for each herd. Specific outcomes of interest included cumulative incidence of the use of any antimicrobial, the use of a category I antimicrobial, and the use of macrolides.

The potential risk factors considered included both categorical and continuous variables and were selected from those described elsewhere for the C3SN herds based on previous reports.²² The potential risk factors for AMU examined included the following variables: herds that sold at least some seedstock (1) versus herds that were 100% commercial and sold no seedstock (0), herds that used confined calving systems and did not calve on large pastures (1) versus herds that primarily calved on large pastures (0), count of cattle purchased from previous pregnancy testing to start of calving season (continuous), herd size comparing herds with at least 300 females calved (1) to herds with fewer than 300 females (0), the month calving season started (coded as December/January vs February/March/April vs May or later), the percentage of calves that died before weaning scaled to reflect 10% increments (continuous), and the percentage of calves reported as treated scaled to reflect 10% increments (continuous). Evidence of nonlinearity for the continuous risk factors was assessed by introducing a squared term into the model and was noted if significant.

Any variables where the P value was < .20 were retained and evaluated together in a multivariable model. Variables were eliminated from the multivariable model if they were both not significant in the multivariable model (P > .05) and did not confound other associations of interest in the model. Due to the relatively small number of herds, interaction was not evaluated between significant risk factors due to limited power and resulting issues with model convergence. Results were reported as ORs with 95% CIs.

Evaluating the association between AMU and AMR—Appropriate descriptive statistics were provided for the AMR data for E coli and for Enterococcus spp isolates by species. The next step in the statistical analysis was to examine AMU as a potential risk factor or predictor for AMR. The analysis was restricted to herds that submitted fecal samples for nursing calves in both the spring and fall of 2021 and also provided individual animal treatment records for 2021. Examined associations were limited to comparing the classes of antimicrobials used for treatments (potential risk factors of interest) and corresponding resistance measured to the same antimicrobial class in E coli and then Enterococcus spp isolates from the calf fecal samples (outcome variables of interest).^8,9

Antimicrobial use for nursing calves was summarized by antimicrobial class at the herd level and then examined for association with the AMR results from the fecal samples using GEEs as described below. The associations were examined for 2 time periods: (1) between calving season treatments reported in nursing calves in the winter/spring (January 1 through June 30, 2021) and AMR in the fecal samples collected before July 1, 2021, and (2) between any AMU in calves before weaning in 2021 and AMR in fecal samples collected from calves in the fall near the time of weaning.

A series of univariable associations were estimated between the cumulative incidence for the use of an antimicrobial class and the odds of a corresponding class of AMR in E coli isolates and then isolates from any Enterococcus spp. The GEE used a logit link function and binomial distribution accounting for the clustering of AMR by herd with a repeated term (SAS, version 16.1; SAS Institute Inc). The outcome for each model was a numerator representing the number of isolates with the specific class of AMR in the target organism and a denominator representing the total isolates from each herd. The risk factor of interest was cumulative incidence of AMU for the corresponding class scaled by multiplying by 100 to reflect the equivalent of a 1-in-100-calves-born-at-risk increase in the rate of use. As cumulative incidence was included in the model as a continuous variable, significant associations were tested for the linearity assumption by introducing testing the addition of a squared term within the existing model and reported only if significant. Associations between AMU and the corresponding class of AMR were examined in the regression analysis if 2 or more herds had isolates resistant to that class and 2 or more herds reported use due to failure of GEE convergence with fewer herds. Results were reported as OR and 95% CI.

Summary of 2021 AMU data in nursing calves

Records of AMU in nursing calves (N = 2,045 treatment records) were available for analysis from 54 of 56 herds that provided any individual animal AMU data from a total of 14,304 calves born alive and an overall cumulative incidence of 14.3 antimicrobial treatments/100 calves born at risk. The median number of calves born alive per herd was 201, with a range of 31 to 1,117. No individual calf treatment records were reported by the remaining 2 herds, which instead provided individual AMU records for at least 1 other age class. The drugs most reported as used at least once in nursing calves were the category III antimicrobials florfenicol (N = 39 herds [70%]) and oxytetracycline (N = 36 herds [64%]; Table 1). The most frequently reported category I and category II antimicrobials were third-generation cephalosporins and macrolides, respectively (Table 2).

While most herds used antimicrobials at least once, the frequency of use within these herds was relatively low (Tables 1 and 2). For the herds that reported AMU, the median cumulative incidence of any AMU was 10%, with only 17 of 56 herds (30%) reporting a cumulative incidence of any AMU > 25%, 7 of 56 (13%) reporting > 50%, and 2 of 56 (4%) reporting > 75%. Half of the herds using category I and II antimicrobials treated less than 1.9% and 3.6% of calves, respectively. No herds treated > 25% of their calves with category I antimicrobials, and only 4 herds treated > 25% of calves with category II antimicrobials (1 herd > 50% of calves).

Nearly all herds that provided records and samples (N = 30 of 31 herds [97%]) reported the treatment of at least 1 calf prior to weaning in 2021 (Table 1). Florfenicol was used by 21 (68%) and oxytetracycline was used in calves by 20 (65%) of the sampled herds. Ceftiofur use was reported in calves by 8 of the 31 herds (26%; Tables 1 and 2).

Common reasons for treatment (N = 56 herds; N = 2,045 treatments) included respiratory diseases (30%); gastrointestinal issues, such as diarrhea (15%); navel infections (12%); pinkeye (8%); ear infections (3%); fever (3%); footrot (2%); arthritis (2%); injury (2%); and diphtheria (1%). However, in 20% of treatment records the reason for treatment was not specific (eg, sick or dull [11%]), did not fall into 1 of the noted categories (4%), was reported as other (3%), or was missing (2%).

Among the 56 herds that reported 2,045 antimicrobial treatments in calves in 2021, 70% of reported treatments were administered by SC injection, 16% by IM injection, 6% by oral bolus, and 2% topically. The route of administration was not reported for 124 of 2,045 treatment records (6%). In addition to reason for treatment and route of administration, information was missing from treatment records for 29 of 2,045 animal identifications (1%), 18 of 2,045 treatment dates (1%), and 1,212 of 2,045 calf age at treatment (59%).

Risk factors associated with herd AMU in nursing calves

Factors associated with the cumulative incidence of any AMU in univariable models for the 56 herds providing records included selling at least some seedstock versus 100% commercial and selling no seedstock, not using large pastures for calving and calving in a confined area versus calving in large pastures, at least 300 females calving versus calving fewer than 300 females, increasing percentage of cows assisted at calving, and earlier month calving season started (Table 3). In the final multivariable model, the odds of any AMU were 2.8 times higher (95% CI, 1.2 to 4.5; P = .015) in herds that sold at least some seed stock compared to herds that did not, 2.0 times higher (95% CI, 1.0 to 4.1; P = .045) in herds with at least 300 cows compared to herds with fewer than 300 cows, and 2.5 times (95% CI, 1.5 to 4.0; P < .001) for every 10% increase in percentage of calves assisted at birth.

Factors associated with increasing use of category I antimicrobials of very high importance to human health (eg, third-generation cephalosporins, fluoroquinolones) included at least 300 females calving as compared to fewer than 300, starting to calve before February as compared to calving later, and an increasing percentage of calves that died before weaning (Table 3). In the final multivariable model, the odds of use of category I antimicrobials were 14.4 times higher (95% CI, 2.7 to 76.2; P = .002) in herds with at least 300 cows compared to herds with fewer than 300 cows.

Factors associated with an increasing cumulative incidence of macrolide use included selling at least some seedstock as compared to commercial herds that did not sell seedstock, not using large pastures for calving or calving in a confined area as compared to calving in large pastures, increasing total number of cattle purchased, at least 300 females calving as compared to herds with fewer than 300 females, increasing percentage of cows assisted at calving, and earlier month calving season started (Table 3). In the final multivariable model, the odds of macrolide use were 4.8 times higher (95% CI, 1.7 to 13.7; P = .003) in herds that sold at least some seed stock compared to herds that did not, 2.1 times (95% CI, 1.1 to 3.9; P = .019) for every 10% increase in percentage of calves assisted at birth, and 1.3 times (95% CI, 1.1 to 1.5; P = .002) for every additional animal purchased.

Herd level survey reports from the same year describing the antimicrobial treatment of respiratory disease, diarrhea, and navel infection were also associated with the cumulative incidence of total AMU and use of category I antimicrobials derived from individual treatment records (Table 3).

Summary of matching 2021 AMR data for nursing calves

In total, 646 samples were collected from calves in the 31 herds providing AMU records for 2021. Of the samples collected, 307 were from the spring and 339 from the fall of 2021. The isolation rate for E coli in the selected herds was 100%, with isolates obtained from every sample in the spring and fall of 2021. The overall isolation rate for Enterococcus spp was slightly lower at 93% (604 of 646) and varied by season: 88% (271 of 307) in the spring and 98% (333 of 339) in the fall.

The frequency of AMR in E coli isolates obtained from calves in the spring and fall of 2021 was low, with no resistance to the category I antimicrobials of the highest importance to human health (Table 4). In the spring and fall, 23 and 24 of the 31 herds reported at least 1 isolate resistant to the tested antimicrobials. Resistance to ampicillin, a category II antimicrobial, was observed in a few isolates obtained from calves in both seasons. Most resistance observed in E coli isolates was to category III antimicrobials (ie, those of medium importance to human health, such as tetracycline, sulfisoxazole, and chloramphenicol).

Resistant Enterococcus spp isolates were recovered from all 31 herds that provided fecal samples in the spring and fall of 2021 and individual animal treatment records for 2021. Enterococcus faecalis, Enterococcus casseliflavus, and E hirae isolates were most frequently tested for resistance in the spring, whereas E casseliflavus isolates were most commonly recovered for testing from samples in the fall (Table 5).

Antimicrobial resistance in the spring was observed in every Enterococcus spp and across all categories of importance (Table 5). The most common resistance patterns included lincomycin, daptomycin, and quinupristin/dalfopristin; however, certain Enterococcus spp, such as E faecalis and E casseliflavus, are known for being intrinsically resistant to some of these antimicrobials. Unfortunately, there is a lack of clarity on which of the broad range of Enterococcus spp are potentially intrinsically resistant to certain antimicrobials, thereby making the classification of AMR a challenge in less studied species, such as E casseliflavus. Resistance to tetracycline was the most common acquired AMR (excluding potential intrinsic resistance and uncertain breakpoints) across all species of Enterococcus (45 of 271 [17%]) in the spring.

Similarly, in the fall, the most common resistance profiles observed were also to the antimicrobials for which intrinsic resistance was expected (Table 5). As in the spring, resistance was seen in the fall for every isolated Enterococcus spp as well as across every category of antimicrobial. In contrast to the spring, the most common acquired resistance across all Enterococcus spp (for which there are clear breakpoints) was to ciprofloxacin (42 of 333 [13%]).

Fecal samples for AMR in each herd were collected after all eligible treatment records of AMU for 27 of 31 herds in the spring and 25 of 31 herds in the fall. For the remaining 4 herds in the spring, a median of 86% of the treatments was reported before fecal sample collection, and for the 6 herds in the fall a median of 90% of the treatments was reported before sample collection.

Association between class-specific AMU and AMR

The 4 antimicrobial classes used for treatment (Table 2) and that could be linked to the E coli susceptibility panel included amphenicols, tetracyclines, sulfonamides, and fluroquinolones (Table 4). No fluroquinolone resistance was observed in the E coli isolates, and no significant associations were evident between corresponding class AMU and AMR from the other 3 antimicrobial classes for the E coli isolates obtained from calves in either the spring or fall (Table 6).

The 4 antimicrobial classes on the Enterococcus spp susceptibility panel that corresponded with at least some reported use included amphenicols, tetracyclines, macrolides, and fluoroquinolones (Tables 2 and 5). No significant association was noted between corresponding class AMU and AMR in Enterococcus spp isolates for amphenicols, tetracyclines, or fluoroquinolones (Table 6). However, a small but significant association was observed between the cumulative incidence of macrolide use within herd and the frequency of macrolide resistance. The association was tested for linearity and then rescaled for ease of interpretation such that the odds of macrolide resistance in the Enterococcus spp isolates from the fall increased by 1.54 times (95% CI, 1.26 to 1.89) for every 10% increase in reported total macrolide uses within herds.

Summary of 2021 AMU data in breeding females

While the focus of this study was on AMU and AMR in nursing calves, records of AMU in breeding females (≥ 2 years of age; N = 408 treatment records) were also available for analysis from 45 of the 56 herds that provided any individual animal AMU data.

The drugs most reported as used at least once in breeding females were the category III antimicrobials tetracyclines (N = 23 herds [41%]), category II antimicrobials macrolides (N = 22 herds [39%]) and penicillins (N = 18 herds [32%]), category III florfenicols (N = 11 herds [20%]), and category I third-generation cephalosporins (N = 7 herds [13%]).

The overall frequency of use was low, with AMU reported for 3.8/100 cow years at risk in herds reporting at least 1 treatment (408 of 10,533). For the herds that reported AMU, the median cumulative incidence of any AMU was 2.3%, with only 8 of 45 herds (18%) reporting a cumulative incidence of any AMU > 10%, 2 of 45 (4%) reporting > 25%, and 0 of 45 (0%) reporting > 50%. Ninety percent of the herds using category I antimicrobials treated less than 0.8% of cows. No herds treated > 7% of their cows with category I antimicrobials. Half of 45 herds reporting treating at least 1 breeding female described AMU in less than 1.0% of cows with either category II or III antimicrobials.

Common reasons for treatment (N = 45 herds; N = 408 treatments) included lameness (32%), calving complications (24%), mastitis (16.3%), bovine respiratory disease (BRD; 8.1%), eye disease (8.3%), and other (11.3%). Category I antimicrobials were most frequently used as intramammary infusions to treat mastitis (24 total treatments in 5 herds). The remaining category I uses were reported for calving complication (4 treatments in 1 herd), lameness (2 treatments in 1 herd), and BRD (3 treatments in 2 herds).

Given the significant association between AMU and AMR for macrolides in calves, the use of macrolides in cows was further described to evaluate the importance of the potential exposure risk from cows. Half of herds treated less than 0.1% of cows with macrolides, and 90% of herds treated less than 4.2% of cows with macrolides. Macrolides were used in 12 herds to treat lameness, 8 herds for calving complications, 7 herds for BRD, 6 herds for mastitis, 3 herds for eye disease, and 6 herds for other reasons.

Of the 5 herds with cow AMU records that identified macrolide resistance in calves in the fall, 2 of the herds did not use macrolides in the cows, 2 of the herds used macrolides in less than 1% of cows, and 1 herd used macrolides in 1.4% of cows. There was no evidence of macrolide resistance in the calves for the remaining matching herds providing individual cow AMU data.