Materials and Methods

Study population and record assessment—Records of 16 randomized clinical trials that were conducted in confined cattle feeding facilities in 7 states from October 26, 1996, to November 15, 2004, were searched to identify feeder calves that received tilmicosin for treatment of BRD (Table 1). Records of various types and breeds of beef feeder calves (6 to 10 months old) that had no history of previous antimicrobial treatment, including metaphylaxis, and met criteria for the study definition of BRD (clinical signs of BRD [CIS, 2 to 5 on a 1 to 5 scale as described elsewhere²²] with a rectal temperature ≥ 40.0°C [104.0°F]) at the time of enrollment were included in the present study. In all trials, DNS samples were collected from all eligible cattle prior to antimicrobial treatment and susceptibility status of the isolates was not known or considered in the treatment decision. In each of the trials, the day of treatment was considered day 0; if the CIS of a calf was not lower on day 3 after treatment, compared with the CIS at enrollment, and rectal temperature was again ≥ 40.0°C, the case was identified as a treatment failure. If the CIS on day 3 was improved but the calf met criteria for BRD again ≤ 21 days after treatment, the case was considered a relapse of BRD. If the calf had no clinical signs of BRD ≤ 21 days after treatment and did not die or require euthanasia because of BRD prior to conclusion of the clinical trial, the case was considered a treatment success. The clinical trials were funded by a drug company^a and were conducted according to the various contractors' established animal care protocols. For analytic purposes in the present study, cases identified as treatment failures and BRD relapses were combined to represent an unsuccessful treatment outcome.

Sample collection and treatment—All calves in each of the trials that met the case definition for BRD were restrained in a cattle chute for DNS sample collection immediately prior to antimicrobial administration (for purposes of the study reported here, the treatment selected was tilmicosin,^b (10 mg/kg [4.5 mg/lb], SC). A DNS sample was collected by use of an 80-cm long, cotton-tipped, guarded swab^c passed via one of the nares through the ventral nasal meatus to the pharyngeal area. After withdrawal into the self-contained transport tip, the tip was snapped off according to the manufacturer's instructions and plugged on each end with plastic plugs provided with the swab. Each sample was labeled and stored on ice or refrigerated at 4°C until transported by courier service (within 24 hours after collection) to microbiology laboratories^d–k near each of the individual clinical trial sites.

Microbiological analysis—At each microbiology laboratory,^d–k samples were placed on 5% sheep blood or 5% bovine blood agar plates. Samples were incubated for ≥ 18 hours at 36 ± 2°C in 5 ± 2% CO₂. Plates were examined for the presence of respiratory pathogens, including colonies of M haemolytica, P multocida, and H somni, by use of standard identification procedures. The identity of each isolate of interest was confirmed by means of Gram staining, oxidase and indole biochemical testing, atmospheric CO enrichment, and growth on MacConkey agar as appropriate. Quality-control strains of representative bacteria were included with each identification test performed.

Following pathogen isolation and identification, samples with positive test results were transported by courier from the laboratory in which the microorganisms were first isolated to one of 2 laboratories,^d,f where MICs of tilmicosin against each isolate were determined. Both laboratories followed CLSI procedures¹¹ for MIC testing in cation-adjusted Mueller-Hinton broth for M haemolytica and P multocida isolates and in veterinary fastidious medium for H somni. Personnel at one microbiology laboratory^d used commercially available microdilution plates,^l and those at the other^f used plates prepared on the premises that contained 100 μL of cation-adjusted Mueller-Hinton broth or veterinary fastidious medium. Inoculated plates that contained various concentrations of tilmicosin (range, 0.5 to 64.0 μg/mL) were incubated aerobically at 36 ± 2°C for 16 to 20 hours, except for H somni–inoculated plates, which were incubated in a 5 ± 2% CO₂ atmosphere at 36 ± 2°C for 16 to 20 hours. Microorganisms included for quality-control purposes included Escherichia coli (ATCC 25922), Pseudomonas aeruginosa (ATCC 27853), Enterococcus faecalis (ATCC 29212), and Staphylococcus aureus (ATCC 29213); these were tested on each sample testing date. To assure proper quality control when the MIC of tilmicosin against H somni isolates was determined, H somni (ATCC 700025) was tested concurrently. Positive growth controls, negative sterility controls, and inoculum counts were also included.

Statistical analysis—Data were combined into a single electronic data set^m and validated. Data were then imported into a commercially available software package for analysis.ⁿ Only data for calves treated with tilmicosin after DNS sample collection from which a single, well-defined isolate of M haemolytica or P multocida was recovered were included in the statistical analysis. When ≥ 2 phenotypically distinct isolates of M haemolytica or ≥ 2 phenotypically distinct isolates of P multocida were recovered from a sample, or 1 isolate each of M haemolytica and P multocida was recovered from a sample, data for that calf were excluded from the analysis. Because a relatively small number of samples met the inclusion criteria, isolates for which the MIC of tilmicosin was 0.5, 1.0, or 2 μg/mL were combined into a single category (MIC ≤ 2 μg/mL). The MICs of tilmicosin against all isolates were then log₂ transformed. Initial descriptive statistics, such as distribution of M haemolytica and P multocida among calves from the various clinical trials, were generated and inspected prior to model development. Various generalized linear mixed models were constructed to evaluate the association between treatment success (binary response variable) and susceptibility of the recovered organism to tilmicosin; in separate models, log MIC was treated as either a classification variable or continuous variable. In further models, recovery of an isolate categorized as not susceptible (ie, intermediate or resistant [MIC of tilmicosin, ≥ 16 μg/mL; yes vs no]) and recovery of a resistant isolate (MIC of tilmicosin, ≥ 32 μg/mL; yes vs no) were modeled as independent variables. A random effect of clinical trial was included in the model because outcome may have been clustered within a trial. The choice of an unstructured within-trial covariance matrix was made on the basis of model fit criteria (ie, the nested model with the lowest Akaike information criterion was selected). Intercept-only models were constructed to calculate overall probability of treatment success after accounting for within-trial clustering. A model with outcome as the dependent variable and pathogen as the independent variable was constructed. Where appropriate (treating MIC of tilmicosin as a class variable with a significant P value), odds of treatment success were compared. Odds of treatment success and 95% CIs were back-transformed to a proportion and graphed.^o Models were constructed according to the pathogen that was recovered (M haemolytica or P multocida). Histophilus somni—positive samples were excluded from the analysis because of sparse data.

Population attributable fraction,²³ an estimate of the proportion of total treatment failures attributable to antimicrobial susceptibility of the organism, was calculated. Model-adjusted relative risks (from models of M haemolytica and P multocida) were calculated from model-adjusted, back-transformed risk of treatment failure, and PAF was calculated by use of the following equation:

View Expanded

where p is the proportion of all calves from which a resistant isolate was recovered or, in separate calculations, the proportion of calves in which the isolate was categorized as not susceptible (ie, intermediate or resistant).

Results

A total of 1,237 M haemolytica (n = 878) and P multocida (359) isolates were collected from 1,100 calves with clinical signs of BRD during 16 clinical trials. Histophilus somni was isolated from only 29 (2.6%) calves that had clinical signs of BRD; this was the only isolate recovered in 17 cases, whereas in the other 12, H somni was isolated in combination with either M haemolytica or P multocida. Some interyear and intertrial variation was observed for relative numbers of M haemolytica to P multocida isolates. Overall, 71.0% of the isolates were identified as M haemolytica; the lowest M haemolytica: P multocida ratio detected was 43.2:56.8, and the highest ratio was 96.7:3.3. Of the 1,100 calves, samples from 976 were included in the analysis (ie, yielded a single, well-defined bacterial isolate). Of these 976 samples, 745 yielded a single M haemolytica isolate (Table 2) and 231 yielded a single P multocida isolate (Table 3).

Minimum inhibitory concentration distributions were variable among trials. The overall MIC of tilmicosin required to inhibit growth of 50% of M haemolytica and P multocida organisms was 4 μg/mL, whereas the MIC required to inhibit growth of 90% of organisms was 8 μg/mL for M haemolytica and 16 μg/mL for P multocida. On the basis of results of tilmicosin testing, recovered M haemolytica and P multocida organisms were categorized as susceptible, not susceptible, or resistant (MICs, ≤ 8 μg/mL, ≥ 16 μg/mL, or ≥ 32 μg/mL, respectively). Only 6 of 745 (0.8%) M haemolytica and 16 of 231 (6.9%) P multocida isolates were phenotypically resistant to tilmicosin (Tables 2 and 3). Most (42/57 [73.7%]) of the M haemolytica isolates categorized as either not susceptible or resistant were obtained during 2 trials (1 trial in Nebraska in 2002 and 1 in Texas in 2003). Reference strains were all within acceptable quality-control ranges for each testing period.

Overall probability of successful treatment was 58.1% for calves that had a single M haemolytica isolate and 60.2% for calves that had a single P multocida isolate. After adjustment for within-trial clustering, model-adjusted estimates of the probability of treatment success were 61.0% and 63.6% for those calves from which either an M haemolytica or P multocida isolate was recovered, respectively. However, there was no significant difference between likelihood of treatment success and pathogen type recovered (P = 0.27).

For individual calves from which an M haemolytica isolate was recovered, no association was detected between MIC of tilmicosin and treatment response either on a linear scale (OR, 1.08; 95% CI, 0.87 to 1.34; P = 0.47) or as a classification variable (P = 0.45; Figure 1). Furthermore, we did not detect an association between treatment outcome and recovery of a susceptible isolate, compared with recovery of an isolate that was not susceptible (OR, 0.55; 95% CI, 0.27 to 1.08; P = 0.08; Figure 2). Additionally, no association was detected (OR, 0.40; 95% CI, 0.06 to 2.59; P = 0.34) between treatment outcome and recovery of a resistant isolate, compared with recovery of an isolate that was not resistant (ie, < 32 μg/mL). The proportion of treatment failures attributable to recovery of an M haemolytica isolate that was categorized as resistant or not susceptible (ie, population attributable fraction) was 0.2% and 0.5%, respectively.

View Full Size

Probability of treatment success determined in a retrospective cohort study of 976 feeder calves with clinical signs of BRD from which a single isolate of Mannheimia haemolytica (black bars) or Pasteurella multocida (white bars) was recovered from DNS samples. Isolates were categorized according to MIC of tilmicosin prior to administration of the drug (10 mg/kg [4.5 mg/lb], SC). Error bars represent upper 95% CI. Numbers above the error bars represent number of isolates within each category.

Citation: Journal of the American Veterinary Medical Association 239, 1; 10.2460/javma.239.1.129

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Probability of treatment success determined in the 976 calves in Figure 1 with isolates of M haemolytica (black bars) or P multocida (white bars) categorized on the basis of results of tilmicosin testing as susceptible (MIC ≤ 8 μg/mL), not susceptible (MIC ≥ 16 μg/mL), or resistant (MIC ≥ 32 μg/mL). See Figure 1 for remainder of key.

Citation: Journal of the American Veterinary Medical Association 239, 1; 10.2460/javma.239.1.129

• Download Figure
• Download figure as PowerPoint slide

For calves from which a single P multocida isolate was recovered, no association was detected between MIC of tilmicosin and treatment outcome either on a linear scale (OR, 0.86; 95% CI, 0.61 to 1.19; P = 0.36) or as a classification variable (P = 0.64; Figure 1). Furthermore, as observed in the analysis of M haemolytica isolates, we did not detect an association between recovery of an isolate that was not susceptible (OR, 1.51; 95% CI, 0.54 to 4.25; P = 0.43) or was resistant (OR, 1.08; 95% CI, 0.28 to 4.13; P = 0.91) with treatment outcome (Figure 2).

Discussion

Results of the study reported here indicated M haemolytica was the most commonly (745/976 [76.3%]) isolated microorganism from calves with clinical signs of BRD in cases for which a single isolate was recovered. Furthermore, the distribution ratio of M haemolytica to P multocida did not appear to vary substantially among years or among various breeds and types of beef feeder cattle included in the trial populations of calves. This observation differs somewhat from the finding of Welsh et al²⁴ that the ratio of P multocida to M haemolytica increased between 1998 and 2002. In the 16 clinical trials from which data for the present study were obtained, the number of clinical infections associated with a single isolate was greater for M haemolytica (745/878 [84.9%]) than for P multocida (231/359 [64.3%]), possibly indicating P multocida was more likely to be opportunistic or a copathogen rather than a primary pathogen. This observation is consistent with those of other studies⁹ in that P multocida produced less severe pulmonary lesions, compared with M haemolytica, and that P multocida could be isolated from pneumonic lungs in which it was not the primary pathogen.

In the clinical trials that provided records for use in the present study, all calves were treated with tilmicosin without prior knowledge of the susceptibility status of the bacterial isolates. The distribution of MICs of tilmicosin did not appear to differ between M haemolytica and P multocida isolates. Furthermore, the prevalence of tilmicosin-resistant M haemolytica (6/745 [0.8%]) and P multocida (16/231 [6.9%]) isolates from pretreatment DNS specimens was low, indicating the use of macrolides in cattle over several decades has not generated a reservoir of macrolide-resistant BRD-associated pathogens detectable by use of nonselective surveillance methods.

Tilmicosin breakpoints for M haemolytica have been applied since 1996. In conjunction with clinical experience, these are intended to be used as a guide in selection of appropriate antimicrobial treatment for microbial infections. Typical expectations in the field are that calves from which isolates with lower MICs are recovered are more likely to have a favorable clinical response to treatment, compared with calves from which an isolate with a higher MIC is recovered. However, in the analyses of the data reported here, MICs of tilmicosin were of limited predictive value in terms of clinical outcome of calves from various populations of cattle treated with this drug. This observation is similar to those reported in other studies in which antimicrobial susceptibility testing was of limited value in terms of predicting therapeutic outcome to treatment of mastitis with oxytetracycline²⁰ or pirlimycin hydrochloride.²¹

Although there appeared to be some evidence of greater treatment success against M haemolytica isolates classified as susceptible to tilmicosin, compared with those determined to be not susceptible (ie, categorized as either intermediate or resistant), this difference was not statistically significant (P = 0.08). The comparison of treatment outcome and recovery of a resistant versus not resistant (ie, categorized as either intermediate or susceptible) isolate was associated with a much larger P value (ie, P = 0.34); however, this comparison lacked statistical power as there were only 6 isolates that fell into the resistant category. Wide CIs (Figures 1 and 2) indicate this lack of precision and statistical power. Although the observed relationship is biologically plausible, the evidence supporting it is weak and may be subject to type I error; a potential association between MIC and treatment outcome warrants further studies that have greater statistical power to more appropriately test this hypothesis regarding treatment outcomes for M haemolytica–associated BRD. On the other hand, there was clearly no evidence for an association between the MIC of tilmicosin against P multocida and treatment outcome in models that had substantially greater power.

Clinical breakpoints of an antimicrobial against a particular bacterial pathogen are one piece of information for the clinician to consider when selecting an antimicrobial for treatment of a disease condition. However, it is important to recognize that clinical breakpoints are not intended to be predictive of the clinical outcome for an individual calf with BRD because of the multiplicity of host-pathogen-antimicrobial interactions.

Possible explanations for the failure to detect an association between the MIC of tilmicosin against isolated organisms and treatment outcome for calves reported in these trials were that isolates recovered in pure culture from a DNS were not good indicators of the microbial population within lung tissue of calves with BRD or that calves with clinical signs of BRD may have had a mixed microbial population in the lungs. Thus, classification of the population as resistant or susceptible on the basis of results of a single sample culture may result in erroneous conclusions. If this misclassification is nondifferential, then the data were by definition biased toward the null hypothesis. Another possibility is that use of the CIS method resulted in some amount of non-differential misclassification (ie, calves that no longer had clinical signs of BRD classified as treatment failures or calves that had clinical signs classified as treatment successes), which would also bias the data toward the null hypothesis, and as previously discussed, lack of statistical power may have been an important consideration. Although at least 1 study³ of calves with clinical signs of respiratory disease revealed an association between recovery of respiratory tract pathogens from nasal swabs and recovery of the same pathogens from transtracheal swabs of the same animal, additional investigations should be performed in this area.

The population attributable fraction, a measure of the proportion of all observed treatment failures attributable to classification of an M haemolytica isolate recovered from a DNS as either not susceptible or resistant on the basis of results of tilmicosin testing, was 0.5% and 0.2% for the 2 categories, respectively. In other words, the vast majority of treatment failures were likely attributable to factors other than macrolide resistance among isolates recovered from the DNS. Although these estimates were likely biased somewhat toward 0% because of nondifferential misclassification, a better understanding of other potential variables associated with treatment failure, especially those related to calves' intrinsic mechanisms to resist or overcome BRD, could help to improve predictions of treatment outcome.

One important factor is the ability of the calf to effectively respond immunologically to growing microbial populations within the lower respiratory tract. If a calf was immunocompromised because of factors such as stress, poor nutritional status, concurrent infection with a viral pathogen, or lack of prior exposure to the pathogen, the clinical response would likely have been poor regardless of the susceptibility of the pathogen to the antimicrobial used. Alternatively, if the immune system was competent and the calf minimally stressed, the clinical response may have been favorable even if the pathogen was resistant to the antimicrobial used. Other factors (not evaluated in the present study) that may have a role in determining treatment success or failure include the timing of treatment initiation relative to development of BRD. If treatment was initiated after the development of severe lung disease, results were likely to be poor regardless of the susceptibility of the associated pathogen. It may also have been possible that other unmeasured factors resulted in differential misclassification or confounding and biased the measures of effect in an uncertain direction in relation to the null hypothesis.