Materials and Methods

Study population—A cross-sectional study of a convenience sample of 50 Ohio dairy herds was conducted. Nine veterinarians in private practice were contacted from across the state of Ohio for help in the recruitment of prospective dairy herd study participants. Herds were voluntarily enrolled in the study if dairy herd owners permitted the investigators to collect fecal samples from their cows and if the dairy herd owners or managers would agree to answer survey questions regarding demographic and antimicrobial usage on their farm. Samples were collected from each herd a single time between the summer of 2004 and the spring of 2006.

Sample collection and processing—Fresh fecal samples were collected from all lactating cows in study herds that were milking < 100 cows. In herds that were milking > 100 cows, a maximum of 100 fecal samples were collected from lactating cows. Approximately 25 g of feces was obtained from the rectum of each cow by use of individual palpation sleeves. Samples were placed in individual containers to prevent cross-contamination, transported to the laboratory, and processed on the day of collection.

In the laboratory, selective media were used in an attempt to recover E coli with reduced susceptibility to ceftriaxone and Salmonella spp from each fecal sample. The selection procedure for the detection of E coli with reduced susceptibility to ceftriaxone was a 2-stage process. A 10-g aliquot of feces was placed in 90 mL of nutrient broth^a containing cefoxitin^b (4 μg/mL) and incubated at 37°C for 24 hours. The next day, MacConkey agar^c that contained ceftriaxone^d (8 μg/mL) was inoculated by use of sterile cotton-tipped swabs and incubated for 24 hours. Typical lactose-positive colonies were selected and confirmed as E coli by the indole test, and sample aliquots of tryptic soy broth^e were inoculated with individual colonies and frozen with dimethyl sulfoxide for preservation.

The detection of Salmonella spp was a 3-stage process. A 4-g aliquot of feces was obtained from each sample and placed in 36 mL of tetrathionate broth^f that was supplemented with brilliant green and tergitol.^g This was incubated for 24 hours at 37°C. The next day, 10 mL of Rappaport-Vassiliadis broth^h was inoculated with 100 ML of tetrathionate broth. This was incubated for 24 hours at 42°C. On the third day, sterile cottontipped swabs were used to inoculate xylose lysine desoxycholate 4 agar.ⁱ Red colonies with black centers were considered presumptive positive for Salmonella spp. A single isolate was selected and used to inoculate MacConkey agar that was incubated for 24 hours at 37°C. A single colorless lactose-negative colony was selected and used to inoculate tryptic soy broth for storage and for triple sugar iron and urea biochemical confirmation. Isolates were further confirmed as Salmonella spp by agglutination with Salmonella spp polyvalent antisera.^j

To evaluate whether reduced susceptibility to third-generation cephalosporins was a characteristic of the recovered Salmonella organisms, all Salmonella spp isolates were plated onto MacConkey agar and MacConkey agar supplemented with ceftriaxone (8 μg/mL) to determine whether they had reduced susceptibility to ceftriaxone.

Survey data—At the time of sample collection, dairy herd owners or managers were given a short survey that assessed basic herd demographic information and antimicrobial use on the farm, including ceftiofur. Information was collected on cattle in all stages of production and also on antimicrobials used to treat various types of diseases on each farm. These data were analyzed and used to construct variables to be used in the data analysis.

On the basis of previous data collected in our laboratory, it was estimated that ceftiofur was used on approximately 61% of dairy farms in Ohio.¹⁵ However, in the present study, 88% (44/50) of all study herds were reported to have at least some ceftiofur use, which did not allow for easy comparison of herds that used ceftiofur with herds that did not use ceftiofur. Therefore, after all samples had been collected, an attempt was made to quantify ceftiofur usage for each farm by estimating the proportion of cattle on each farm that are typically treated with ceftiofur during a defined period. To do this, owners of the study herds were contacted by telephone and asked to report the proportion of cows that were treated with ceftiofur in a 3-month period (January 1 through March 31, 2006) according to farm records. This variable was used as the primary risk factor of interest. Of the 50 herds initially enrolled in the study, only 43 herds provided this additional information, and the remaining 7 herds were not included in the final analysis.

Statistical analysis—All data analysis was accomplished by use of a commercial statistical package.^k The outcome of interest was recovery of E coli with reduced susceptibility to ceftriaxone at the individual cow level (1 = recovered and 0 = not recovered). No Salmonella spp with reduced susceptibility to ceftriaxone were recovered from the study herds, so no association could be investigated. The association between ceftiofur use and recovery of E coli with reduced susceptibility to ceftriaxone was accomplished by use of the multilevel mixed-effects logistic regression (xtmelogit) command. A mixed-effects logistic regression model was constructed with the herd included in the model as a random effect. Modeling herd as a random effect was used to account for the clustering of cows within farms and the lack of independence of samples within the same herd.

To construct the logistic regression model, the risk factor of primary interest (ie, the proportion of cattle treated with ceftiofur) was entered first into a model that also contained the herd random effect. A forward-selection process was then used to screen for potential confounding variables. Variables that were screened for confounding included number of lactating cows; number of nonlactating cows; number of heifers; number of calves; breeds of cows that were milked; somatic cell count; milk rolling herd average; age of the herd; type of housing for lactating and nonlactating cows; whether the lactating and nonlactating cows had access to pasture; whether any additions had been made to the herd in the previous 12 months; use of nonlactating mastitis treatment; type of antimicrobials used for the treatment of calf scours, pneumonia, and lameness; use of coreantigen vaccines; and whether the herd was vaccinated with Salmonella spp vaccines. Variables were included in the model as confounding variables if their addition changed the coefficient of the primary risk factor by ≥ 15%. Adjusted ORs and the CIs of the ORs were calculated for the variables that remained in the final model.

Results

The mean ± SD herd size was 180 ± 27 lactating cows and ranged from 13 to 810 lactating cows. The mean rolling herd average for milk production for the 50 herds was 9,751 ± 218 kg with a mean somatic cell count of 254,125 cells/mL. Seventy-four percent (37/50) of farms used free-stall barns for housing lactating cows, and 70% (35/50) milked Holstein cows exclusively. Of the 50 herds in the study, 54% (27/50) had a closed herd for the previous 12 months. Eighty-eight percent (44/50) of herd owners reported that they used ceftiofur as an injectable antimicrobial on their farm within the previous 12 months. Study herds had been in existence for a mean of 41.13 years (range, 0.66 to 114.0 years).

A total of 3,840 fecal samples were collected from the 50 dairy herds. Escherichia coli with reduced susceptibility to ceftriaxone was recovered from ≥ 1 of the fecal samples from 92% (46/50) of the herds tested. Escherichia coli with reduced susceptibility to ceftriaxone was recovered from 60.9% (2,338/3,840) of individual fecal samples. The mean herd proportion of cows from which E coli with reduced susceptibility to ceftriaxone was recovered was 55.6% (range, 0% to 100%).

Salmonella spp were recovered from ≥ 1 of the fecal samples from 44% (22/50) of the herds and from 9.9% (382/3,840) of individual fecal samples. The mean herd proportion of cows shedding Salmonella spp in Salmonella-positive herds was 10.2% (range, 0% to 98.3%). All of the Salmonella spp recovered had a minimum inhibitory concentration to ceftriaxone of ≤ 8 μg/mL, indicating that it was unlikely that they possessed the blaCMY-2 gene.

Eighty-six percent (43/50) of the herd owners were successfully contacted to determine the proportion of cattle treated with ceftiofur. Because information needed for the primary risk factor of interest could not be collected, the remaining 7 herds were excluded from the multivariable models. An association was not observed between the proportion of cattle treated with ceftiofur and the recovery of E coli with reduced susceptibility to ceftriaxone (OR, 1.19; P = 0.32; CI, 0.84 to 1.67). Recovery of E coli with reduced susceptibility to ceftriaxone was more likely (OR, 24.96; P = 0.002; CI, 3.17 to 196.68) from cows on farms where Salmonella spp were also recovered. Additionally, the odds of recovery of E coli with reduced susceptibility to ceftriaxone from an individual cow increased by 62% with every 454-kg increase in the rolling herd average of milk production (OR, 1.62; P = 0.004; CI, 1.16 to 2.25).

Discussion

We found that 88% of herd owners in this study reported using ceftiofur to treat cattle. We previously reported that ceftiofur was used in 61% of Ohio dairy herds,¹⁵ suggesting that its use on dairy farms may have become more common in Ohio in recent years. It is important to note that this estimate was determined through use of a convenience sample of dairy herds and may be different from the true proportion of dairy herd owners using ceftiofur in Ohio. Veterinary use of ceftiofur and other cephalosporins in dairy cattle is a common practice because pharmacokinetic properties of these antimicrobials make the possibility of adulterated milk or meat entering the food chain as a result of human error low. In addition, there is a perception by veterinarians and dairy producers that ceftiofur is effective in the treatment of a wide variety of disease conditions other than and including those that are indicated on its label. For example, veterinarians and producers may use ceftiofur in the treatment of undifferentiated fever or enteric disease thought to be caused by Salmonella spp on the basis of previous experience or on the basis of reports found in the literature. The extralabel use of ceftiofur in the treatment of Holstein bull calves with experimental salmonellosis has been evaluated.¹⁶ These authors found that ceftiofur-treated calves had fewer days with abnormal rectal temperatures and had fewer days of diarrhea, compared with untreated control calves; however, no differences in mortality rate between the 2 groups were detected.

We were able to recover E coli with reduced susceptibility to ceftriaxone from the fecal flora of dairy cows from most (92%) of the herds in the study. The percentage of cows on each farm from which E coli with reduced susceptibility to ceftriaxone was recovered was 55.6%. We have previously reported that approximately 83% of E coli with reduced susceptibility to ceftriaxone were found to contain the β-lactamase gene, blaCMY-2.¹⁵ In this study, we tested a small subset of isolates for the blaCMY-2 gene by use of previously published primers and PCR conditions¹³ and found all had positive results for blaCMY-2. It is possible that other mechanisms of resistance may be responsible for the reduced susceptibility to ceftriaxone found in these isolates, but it is unlikely. These data indicate that the blaCMY-2 gene, which is frequently responsible for resistance to third-generation cephalosporins in E coli and Salmonella spp from cattle,^{13,15,17–19} is present on most commercial dairy farms; however, the frequency with which it occurs cannot be estimated from our data. Use of ceftriaxone in our selection media only allows for detection and not quantification of organisms with reduced susceptibility to ceftriaxone that are suspected to carry the blaCMY-2 gene. However, we believe that in the absence of selective pressure in either the medium or in the cow, the number of organisms with the blaCMY-2 gene is quite low. Because we did not attempt to quantify the number of E coli organisms with reduced susceptibility to ceftriaxone in this study, we cannot draw any conclusions as to their true frequency in the fecal flora of dairy cows.

Detection of Salmonella spp from the dairy herds in the study was similar to another study²⁰ of Salmonella spp prevalence in dairy herds. We were unable to recover Salmonella spp with reduced susceptibility to ceftriaxone from any cows in this study. Thus, our data provide no evidence that the use of ceftiofur in dairy cattle impacts the prevalence of Salmonella spp resistant to expanded-spectrum cephalosporins.

In this study, the proportion of dairy cattle on a farm that were treated with ceftiofur was not associated with the recovery of E coli with reduced susceptibility to ceftriaxone after controlling for the effect of Salmonella spp status of the herds, the rolling herd average for milk production, and the random effects of the herd. However, because we estimated ceftiofur use at the end of the study, it is possible that we either over- or underestimated ceftiofur usage on these farms, which may affect our results. This misclassification bias could have occurred either way, so its effect on our results should be minimal. In our previous study,¹⁵ in which the association between ceftiofur usage on Ohio dairy farms and the recovery of E coli with reduced susceptibility to ceftriaxone was examined, we had similar findings.¹⁵ As in our current study, selective media were used to determine the presence or absence of E coli with reduced susceptibility to ceftriaxone, and the numbers of susceptible and reduced-susceptibility E coli were not quantified. Our current results support our earlier observation that treating a greater number of cows with ceftiofur does not increase the prevalence of E coli with reduced susceptibility to ceftriaxone in the fecal flora of cows in the herd.

Although we were unable to detect an association between level of ceftiofur use on the farm and the recovery of E coli with reduced susceptibility to ceftriaxone, we did observe a strong association between the recovery of E coli with reduced susceptibility to ceftriaxone and the Salmonella spp status of the farm. We found that the odds of recovering E coli with reduced susceptibility to ceftriaxone were much greater on farms in which we detected Salmonella spp, compared with farms in which we did not detect it. The most likely explanation for this association is that herds with a high prevalence of Salmonella spp infections are likely to use ceftiofur for treatment. However, we found no association between ceftiofur usage and the recovery of E coli with reduced susceptibility to ceftriaxone. Again, misclassification of ceftiofur usage on each farm may have biased our results toward no association, but we believe we were just as likely to overestimate rather than underestimate ceftiofur usage on each farm. We hypothesize that farms which are positive for Salmonella spp may have different herd characteristics and management practices, other than the use of ceftiofur, which may select for E coli with reduced susceptibility to ceftriaxone. Additionally, we found an association between the milk production of the herd and the recovery of E coli with reduced susceptibility to ceftriaxone. It is possible that herd-level factors, such as population density, that are associated with Salmonella spp infections may also be associated with the dissemination of E coli with reduced susceptibility to ceftriaxone in the flora of cows.

There are few reports on studies investigating the prevalence of E coli with reduced susceptibility to third-generation cephalosporins on dairy farms. In our study, the percentage of herds in which E coli with reduced susceptibility to third-generation cephalosporins was recovered was greatly increased, compared with results of the study¹⁵ on Ohio dairy farms. Both studies used selective media in the laboratory for the detection of the third-generation cephalosporins. We chose to measure E coli in our study because others have proposed that these commensal organisms may be a reservoir for resistance for Salmonella spp and other enteric pathogens.¹⁹ Bacteria with this resistance phenotype are commonly hidden by the susceptible commensal flora in the absence of selective pressure, whether in the medium or in the animal. Had we not used selective media, it is likely that we would not have been able to detect these organisms with resistance to third-generation cephalosporins among all the susceptible E coli.

Although resistance to third-generation cephalosporins does seem to be increasing, especially in E coli, caution should be exercised when evaluating the clinical implications of this finding, especially in terms of Salmonella spp. Although we found a high percentage of E coli with reduced susceptibility to ceftriaxone, none of the isolates of Salmonella spp were found to have reduced susceptibility to ceftriaxone in this study. This may mean that although E coli may serve as reservoirs for antimicrobial-resistance genes, including blaCMY-2, the selective processes that result in their selection and amplification may be different from Salmonella spp; and thus they may not be the best model for resistance to third-generation cephalosporins in Salmonella spp.

Conflicting evidence exists as to whether individual animal treatment with ceftiofur selects for E coli with reduced susceptibility to ceftriaxone. We found no association between the recovery of E coli with reduced susceptibility to ceftriaxone and the treatment history of an individual animal in our previous study¹⁵; however, Jiang et al²¹ and Lowrance et al¹⁴ showed that individual animal treatment does result in a transient increase in the number of E coli organisms with reduced susceptibility to ceftriaxone. Individual animal treatment records were not obtained in this study, so the effect of an individual cow's treatment on the recovery of E coli with reduced susceptibility to ceftriaxone could not be evaluated.

Our finding that the percentage of herds from which E coli with reduced susceptibility to third-generation cephalosporins were recovered was increased over our previous study,¹⁵ however, is concerning. This may mean that the blaCMY-2 gene conferring reduced susceptibility to ceftriaxone is becoming more widespread and that it maybe more available for transfer to pathogenic organisms, such as Salmonella spp, that are frequently recovered from dairy herds.