Nasopharyngeal (NP) swab samples were collected in 2019 and 2020 from healthy feedlot cattle at multiple randomly selected Canadian beef feedlots as part of a surveillance effort examining AMR in BRD pathogens (M. haemolyica, P. multocida, and H. somni). Once annually, 16 individual cattle were sampled at each participating feedlot at the time of entry processing (arrival timepoint). The same group of cattle, but not necessarily the same cattle, were sampled again when handled later in the feeding period (rehandling timepoint). After recovery of pathogen isolates, antimicrobial susceptibility testing was conducted using the microbroth dilution method. Factors potentially associated with AMR were examined, including sample year, sample quarter, sampling timepoint, animal age, animal weight, BRD risk category, and days on feed (DOF). All animal handling protocols were reviewed and approved by an Animal Care Committee before conducting this research (protocol number FHMS-19031). Inclusion criteria for the surveillance program were developed to allow enrollment of eligible feedlots in proportion to feedlot capacity and the number of fed cattle in the target provinces of Alberta, Saskatchewan, and Ontario, where 90% of feedlots are located.

The inclusion criteria require that enrolled feedlots have a one-time capacity of >1,000 animals, have cattle going directly to slaughter on-site, and have an established veterinarian-client-patient relationship with the veterinarian that enrolls the feedlot. From an anonymized list of feedlots provided by participating veterinarians, eligible feedlots were stratified by size (1,000–5,000 cattle, 5,001–10,000 cattle, 10,001–20,000 cattle, and > 20,000 cattle) and province and then randomly chosen in proportion to that type of feedlot’s contribution to Canada’s overall fed cattle production (See details in Supplementary methods). The final sample was comprised of 21 feedlots in 2019 and 26 feedlots in 2020. Selection was intended to reflect the diversity of cattle sources and production methods among the majority of feedlot cattle produced in Canada.

A single group of cattle assigned to be managed together, at least until the second sampling timepoint, was enrolled for sampling at each feedlot annually. Convenience sampling was used to select each group and the individual cattle at each feedlot. The veterinary practice that enrolled the feedlot was responsible for sample collection. Samples were to be shipped to the laboratory by mid-week to ensure timely processing, therefore, the group of cattle that were sampled depended the laboratory timelines, the schedule of the veterinarian and the availability of arrival cattle at the feedlot,

Sampling occurred at feedlot entry and later when cattle were rehandled for routine management procedures. Sixteen individual animals from each enrolled group were sampled at arrival processing, and 16 individual animals (not necessarily the same animals) from the same enrolled group were sampled at rehandling. The timing for the collection of samples at rehandling was variable across feedlots.

Nasopharyngeal (NP) samples were collected using sterile double-guarded swabs (Sterile Equine Double Guarded Uterine Swab, VetSource Canada Inc., Cambridge, Canada) using a standardized procedure established by the study. The sample collector wore exam gloves that were changed between each animal. Cattle were appropriately restrained in a squeeze chute, and the swabs were passed through the ventral meatus to the pharynx (approximately 10 to 14 inches, depending on the size of the animal), advanced through the guard to collect samples, retracted into the guarded sheath and removed from the nasal passage of the animal. Swab tips were aseptically placed in tubes containing Amies media (Micronostyx, Ottawa, Canada), placed in a cooler, and shipped on ice to Prairie Diagnostic Services (PDS) in Saskatoon, SK, Canada, for isolation of BRD pathogens and susceptibility testing.

NP swabs were plated on blood and chocolate agar plates (Oxoid, Nepean, ON, Canada) and incubated at 35 ± 2°C for 18–24 h in an environment containing 5% CO₂ for isolation of M. haemolytica, P. multocida, and H. somni. Suspected colonies of pathogens were selected based on phenotypic characteristics such as the production of β-hemolysis (M. haemolytica), mucoid colonies (P. multocida), or yellow pigment (H. somni) at 18–24 h of incubation. After 48 h of incubation, plates were examined again with specific emphasis on identifying new H. somni growth. Species identification of suspect colonies was evaluated using the Matrix Assisted Laser Desorption and Ionization Time of Flight (MALDI-TOF) Mass Spectrometry System (Bruker Daltonics Ltd. East Milton, ON, Canada) according to the manufacturer’s operating standards and procedures. One isolate of each organism of interest with confirmed identification was stored per sample in Tryptic Soy Broth (TSB) containing 15% glycerol at −80°C until processed for susceptibility testing.

The antimicrobial susceptibility of M. haemolytica, P. multocida, and H. somni isolates was determined by broth microdilution on a commercially available bovine panel (BOPO7F; Sensitire; Trek Diagnostic Systems, Cleveland, OH, USA), using Clinical Laboratory Standards Institute (CLSI) Guidelines and standards (CLSI, 2020) for testing and quality control. Control strains used as references in these analyses were M. haemolytica ATCC 33396, P. multocida ATCC 12945, and H. somni ATCC 700025, respectively. Thawed M. haemolytica and P. multocida isolates were regrown on blood agar plates at 37°C for 18–24 h, while H. somni isolates were grown on chocolate agar plates at 37°C with 5% CO₂ for 20–24 h. For each sample and organism, a bacterial suspension reaching a final concentration of McFarland 0.5 was prepared, inoculated onto the BOPO7F plates, and incubated at 37°C with 5% CO₂ for 20–24 h for H. somni or 37°C for 18–24 h for M. haemolytica and P. multocida, respectively.

The minimum inhibitory concentration (MIC) for each drug and isolate was interpreted with reference to the antimicrobial susceptibility breakpoints established by CLSI (2020); isolates were assigned to one of the susceptible, intermediate, or resistant categories on this basis (Supplementary Table S1). For the purposes of analysis and to permit a dichotomous outcome, “intermediate” isolates were categorized as “not resistant” to the respective AMD. CLSI breakpoints for the organisms of interest were not available for clindamycin, gentamicin, neomycin, tiamulin, trimethoprim-sulfamethoxazole, tylosin tartrate, or sulfadimethoxine, and only MIC data were summarized. The MIC₅₀ and MIC₉₀ (minimum concentrations at which 50 and 90% of isolates are inhibited from growing, respectively), MIC range, percentage of resistance, and multiple drug resistance (resistance to ≥3 AMD classes) were summarized for each bacterial species.

The prevalence of resistance among isolates was categorized for each species and AMD: rare (<0.1%), very low (0.1 to 1.0%), low (>1.0 to 10.0%), moderate (>10.0 to 20.0%), high (>20.0 to 50.0%), very high (>50.0 to 70.0%) and extremely high (>70.0%) (European Food Safety Authority European Centre For Disease Prevention Control, 2021). AMDs were classified relative to their importance to human medicine as outlined by the Veterinary Drugs Directorate, Health Canada (Health Canada, 2009). Briefly, AMDs are classified as Very Highly Important in Human Medicine (Category I) when essential for treating serious bacterial infections, and there is no or limited availability of alternative antimicrobials for effective treatment. Antimicrobials classified in Categories II through IV are of decreasing importance to human medicine, with the latter including those that are not currently used to treat infections in humans.

The susceptibility data were exported from a central repository to Stata/IC 14.2 2 (StataCorp LLC, College Station, TX, USA) for statistical analyses, and population-averaged prevalences of resistance to AMD with CLSI breakpoints were calculated to account for the hierarchical population structures (clustering) of isolates obtained from animals within different pens and feedlots. Null binomial response models were fitted using generalized estimating equations (xtgee) with a binary outcome, logit-link function, and exchangeable correlation structures to provide adjusted prevalence estimates with 95% confidence intervals. Separate models were used in estimating resistance prevalence for each of the antimicrobial drugs that were tested. If the GEE model did not converge, the standard error for the raw proportion was adjusted to account for clustering by feedlot. If the prevalence of resistance was 0%, a confidence interval calculator was used to estimate an exact upper confidence interval only (StataCorp LLC, College Station, TX, USA).

UpSet plots were produced to visualize intersections between the number of isolates with unique AMR patterns for the different respiratory pathogens (M. haemolytica, P. multocida, H. somni) at arrival and rehandling timepoints, by animal age class (calf vs. yearling) and BRD risk status (high vs. low). These plots were created in RStudio (R Core Team, 2023; Posit Team, 2023) using the ComplexUpset (Krassowski et al., 2022; Lex et al., 2014), tidyverse (Wickham et al., 2019), and patchwork packages (Pedersen, 2024).

Stata/IC 14.2 (StataCorp LLC, College Station, TX, USA) was utilized to summarize the percent recovery (i.e., bacterial isolation rate) by year, month, province, DOF when sampled, and the categories of resistance phenotypes for AMDs. Univariable analysis was performed to identify potential associations with classifying isolates as resistant using a population average logit model with feedlot as a random effect (xtlogit, StataCorp LLC, College Station, TX, USA). Separate models were used in estimating resistance prevalence for each of the antimicrobial drugs that were tested. Only univariable associations were explored; isolates with missing information for the predictor variable were excluded for that analysis only. Variables examined included the province, quarter of the year when the second sampling occurred, sampling timepoint (arrival vs. rehandling), animal age (calf vs. yearling), BRD risk category (high vs. low), and DOF category (20–40, 41–60, 61–80, 81–100, 101–120) at the time the second sample was collected. BRD risk categories were assigned by the supervising veterinarian considering age (calf or yearling), body weight (frequently representative of age), method of procurement (sale barn or ranch direct), amount of commingling before arrival, overall health assessments for the group, vaccination history, and likelihood or history of AMD exposures before arrival. High-risk cattle were typically younger, smaller, and more likely to have experienced commingling than low-risk cattle.

In 2019, 21 feedlots and 21 cattle housing groups (1 group per feedlot) were enrolled, and 608 animals were sampled. Seventeen groups had complete sampling (32 samples per group; 16 at arrival and 16 at rehandling) for a total of 544 samples; 3 groups only had the 16 arrival samples (48 total samples), and 1 group only had 16 samples collected at rehandling. In the 2020 sampling year, 26 feedlots and 26 groups (1 group/feedlot) were enrolled, and 784 cattle were sampled. Twenty-three groups had complete sampling (32 samples per group; 16 at arrival and 16 at rehandling), contributing 736 samples. One group only had the 16 arrival samples collected, and 2 groups only had rehandling samples collected, contributing 32 samples. Weight data were available for 608 animals in 2019 and 736 animals in 2020, with 48 missing values from 3 enrollment groups. Upon arrival, weights ranged from 139.2 to 503.5 kg (mean 337.7 kg, median 337 kg), and at rehandling, weights ranged from 181.4 to 635.9 kg (mean 430.6 kg, median 454.5 kg). Animal age was recorded for 98.9% (1,376/1392) of the sampled animals; 51.1% (704/1376) of these animals were classified as calves (i.e., less than 1 year of age), and 48.8% (672/1376) were classified as yearlings. Forty-seven percent (640/1360) of animals were classified as high-risk for BRD, and 53% (720/1360) were classified as low-risk. Days on feed at the time of rehandling were provided for 76.2% (16/21) of the participating feedlots in 2019 and 84.6% (22/26) in 2020, which ranged from 20 to 104 days, with a mean of 59.8 days and a median of 62 days (Figure 1).

A total of 608 NP swabs were collected between August 2019 and February 2020 (year 1) from 21 enrolled groups, and 784 samples were collected from 26 enrolled groups between June 2020 and February 2021 (year 2; Table 1). Overall recovery of BRD pathogens was highest for P. multocida (27.4%; 381/1392), followed by H. somni (9.0%; 125/1392) and M. haemolytica (8.6%; 119/1392) (Table 1). At arrival sampling, P. multocida, M. haemolytica, and H. somni isolates were recovered from 14.3% (199/1392), 4.1% (57/1392), and 2.2% (30/1392) of calves, respectively. At rehandling, the corresponding prevalences were 13.1% (182/1392), 4.5% (62/1392), and 6.8% (95/1392), respectively. There was variation in the recovery of BRD pathogens by province, which may have partly been attributable to the smaller number of samples and recovered isolates in cattle outside of Alberta (Table 1) and therefore results should be interpreted with caution.

Feedlot level prevalence also varied considerably by organism and year (Table 2). In 2019, 95% (20/21) of feedlots had cattle test positive for at least one of the organisms of interest. M. haemolytica was recovered from 71.4% (15/21) and 76.9% (20/26) of feedlots in 2019 and 2020, respectively. Mannheimia haemolytica was not recovered from any cattle in 2 feedlots in either year. Pasteurella multocida was routinely recovered from cattle in all feedlots, except in 1 feedlot in 2019, where this agent was not recovered from any cattle. In contrast, 47.6% (10/21) and 46.2% (12/26) of feedlots in 2019 and 2020, respectively, had no cattle that were culture-positive for H. somni. Histophilus somni was not recovered from any cattle in 5 of these feedlots in both years.

Across the combined study years, the proportion of samples from which any BRD organism was recovered was highest in September, followed by November and January (Table 3).

Among the isolates tested, 73.6% (95% CI 63.3–81.9%) of M. haemolytica, 67.4% (95% CI 55.2–77.6%) of P. multocida, and 62.1% (95% CI 47.0–75.2%) of H. somni isolates were susceptible to all tested AMDs with an available CLSI breakpoint. The percentage of isolates susceptible to all AMD tested decreased between arrival and rehandling for M. haemolytica [arrival 86.2% (95% CI 69.8–94.4%), rehandling 60.2% (95% CI 44.7–73.8%)] and P. multocida [arrival 80.4% (95% CI 64.4–90.3%) rehandling 49.6% (95% CI 34.0–65.3%)]. Conversely, the percentage of H. somni isolates susceptible to all AMD tested increased between arrival 46.7% (95% CI 12.2–84.6%) and rehandling 65.1% (95% CI 52.3–76.0%).

Resistance to the macrolide class of antimicrobials was the most common resistance phenotype detected in M. haemolytica isolates (Figure 2 and Supplementary Figure S1). All isolates (n = 119) were susceptible to ceftiofur and florfenicol and exhibited very low resistance prevalence (0.1 to 1%) to penicillin and spectinomycin. A low prevalence of resistance (>1.0 to 10.0%) was identified for ampicillin, danofloxacin, and enrofloxacin (Figure 2). Moderate prevalence (>10.0 to 20.0%) or high prevalence (>20.0 to 50.0%) of resistance was seen for tildipirosin, gamithromycin, tulathromycin, and tilmicosin at both sampling timepoints (Figure 2). Overall, 2.4% (95% CI 0.8–7.1%) of the isolates were resistant to ≥3 AMD classes; 3.9% (95% CI 1.3–11.4%) at arrival, and 1.6% (95% CI 0.2–10.1%) at rehandling. The most common resistance pattern at rehandling was gamithromycin-tetracycline-tildipirosin-tilmicosin-tulathromycin (6.9%; 8/119).

Based on the classification of animal age that was supplied by the supervising veterinarians, there was a marked difference in the resistance patterns between calves and yearlings (Figure 3). At arrival, for M. haemolytica, all isolates recovered from yearlings were pansusceptible (susceptible to all antimicrobials tested), whereas in calves, 20.0% (7/35) of isolates were resistant to ≥1 AMD class. At rehandling, however, only 25.9% (7/27) of isolates from yearlings were resistant to ≥1 AMD class compared to 45.7% (16/35) of isolates recovered from calves.

When M. haemolytica isolates were stratified by BRD risk category, isolates from low-risk cattle were all susceptible at arrival. In contrast, 21.2% (7/33) of the isolates recovered from high-risk cattle were resistant to ≥1 AMD class (Figure 4). At rehandling, 21.8% (7/32) of the isolates recovered from low-risk cattle were resistant to ≥1 AMD class, whereas 53.3% (16/30) of isolates from high-risk cattle were resistant.

Resistance to Category I antimicrobials was detected in less than 10% of the total isolates; resistance in this category was most often linked to the fluoroquinolone class (Figure 5 and Supplementary Table S2). Resistance prevalence for ceftiofur was very low (0.1–1.0%) in P. multocida isolates, whereas the resistance prevalences for enrofloxacin, penicillin, florfenicol, ampicillin, and danofloxacin were all low (>1 to 10%, Figure 5). Moderate (10–20%) or high (20–50%) resistance prevalences were identified for spectinomycin, tulathromycin, gamithromycin, tildipirosin, and tetracycline at both sampling times (Figure 5). Multidrug resistance (≥3 AMD classes) was identified in 21.3% (95% CI 12.5–33.9%) of all the isolates. Multidrug resistance increased between arrival (10.7, 95%CI 3.3–29.3%) and rehandling (36.5, 95% CI 23.8–51.5%). A single isolate recovered at rehandling was resistant to 5 AMD classes, including betalactams, fluoroquinolones, phenicols, macrolides, and aminocyclitols and (phenotypic resistance ampicillin-ceftiofur-danofloxacin-enrofloxacin-florphenicol-gamithromycin-spectinomycin-tildipirosin-tulathromycin). The most common pattern among all resistant isolates was gamithromycin-spectinomycin-tetracycline-tildipirosin-tulathromycin (13.9%, 53/381).

Most P. multocida isolates recovered from calves and yearlings were pansusceptible at arrival sampling (Figure 6). However, among resistant isolates, the phenotypic patterns were more variable in calves and generally included more AMD classes regardless of sampling timepoint. The most common pattern detected in calves included resistance to macrolides, tetracyclines, and aminocyclitols (Figure 6).

When comparing low and high-BRD-risk cattle, the high-risk results demonstrate a similar pattern to that observed in calves, while the low-risk cattle tended to have patterns similar to those detected in yearlings (Figure 7).

Histophilus somni isolates were most commonly resistant to tetracyclines, followed by macrolides (Figure 8 and Supplementary Table S3). All H. somni isolates (n = 125) were susceptible to ceftiofur, enrofloxacin, and florfenicol. A low prevalence (>1.0 to 10.0%) of resistance was observed for penicillin, ampicillin, tildipirosin, gamithromycin, and spectinomycin at arrival (Figure 8). Moderate resistance prevalences (10–20%) were observed for gamithromycin, tildipirosin, and tulathromycin at rehandling, and moderate or high resistance prevalence was observed for tetracycline at both sampling times (Figure 8). Two of the 125 H. somni isolates (1.6%) were resistant to 4 AMD classes, including betalactams, macrolides, aminocyclitols, and tetracycline. Resistance to ≥3 AMD classes was identified in 3.4% (95% CI 0.7–14.8%) of the isolates. Overall, 3.3% (95% CI 0.5–19.6%) of arrival isolates and 4.5% (95% CI 1.1–16.7%) of rehandling isolates were resistant to ≥3 AMD classes.

Isolates recovered from yearling cattle at arrival were most commonly pansusceptible (Figure 9). The two resistant arrival H. somni isolates collected from yearlings were each resistant to a single AMD class. Of the isolates recovered from calves at arrival, only 37.5% (9/24) were pansusceptible. At rehandling, resistant isolates recovered from either yearling cattle or calves tended to be resistant to several AMD classes. However, in general, isolates obtained from yearling cattle at rehandling were resistant to fewer AMD’s than isolates obtained from calves at the same timepoint (Figure 9).

All H. somni isolates recovered from low-risk cattle were pansusceptible at arrival (Figure 10), and most (80%, 21/26) isolates were also pansusceptible when recovered at rehandling. In high-risk cattle, only 38% (10/26) of the arrival isolates were pansusceptible, while 59% (41/69) were pansusceptible at rehandling. Resistant isolates from low-risk cattle at rehandling only demonstrated resistance to 1–2 AMD classes. In contrast, isolates recovered from high-risk cattle more commonly had resistance to ≥2 AMD classes and had greater diversity in resistance patterns regardless of sampling time (Figure 10).

Risk factors associated with the identification of resistance varied between BRD pathogens. Only 3 of the 6 risk factors explored were statistically significantly associated with AMR, including sampling timepoint (arrival vs. rehandling), animal age category (calf vs. yearling), and DOF category (20–40, 41–60, 61–80, 81–100, 101–120). In general, isolates recovered at arrival were less likely to be resistant than those recovered later in the feeding period. For the M. haemolytica and P. multocida isolates, there was significantly less resistance to all of the tested macrolides with CLSI breakpoints at arrival than at rehandling (Table 4). Among isolates recovered at arrival, for P. multocida only, resistance to ampicillin, danofloxacin, enrofloxacin, tetracycline, and spectinomycin was less likely compared to isolates obtained at rehandling (Table 4). No significant differences were noted between the two timepoints for H. somni.

Significant differences in resistance were detected for P. multocida and H. somni isolates obtained from calves vs. yearlings (Table 5). The odds of resistance to tulathromycin were about 12 times greater in H. somni isolates obtained from yearlings (OR = 12.0, 95% CI = 1.03–140.4, p = 0.05; Table 5). Similarly, P. multocida isolates from calves were more likely to be resistant to gamithromycin, spectinomycin, tetracycline, tildipirosin, and tulathromycin than isolates obtained from yearlings (Table 5). P. multocida, isolates obtained from samples collected between 81 and 100 DOF were less likely to be resistant to spectinomycin than isolates from samples collected between 20 and 40 DOF (Table 6). Isolates from samples obtained between 61 and 80 DOF or between 81 and 100 DOF were less likely to be resistant to tetracycline, tildipirosin, and tulathromycin than isolates from samples collected between 20 and 40 DOF.