Background: To monitor the prevalence of antimicrobial resistance (AMR), methods for interpretation of susceptibility phenotypes of bacteria are needed. Reference limits to declare resistance are generally based on or dominated by data from human bacterial isolates and may not reflect clinical relevance or wild type (WT) populations in livestock or other hosts.
Methods: We compared the observed prevalence of AMR using standard and bespoke interpretations based on clinical breakpoints or epidemiological cut-offs (ECOFF) using gram positive (Staphylococcus aureus) and gram negative (Escherichia coli) bacteria from sheep as exemplars. Isolates were obtained from a cross-sectional study in three lowland sheep flocks in Scotland, and from a longitudinal study in one flock in Norway. S. aureus (n = 101) was predominantly isolated from milk or mammary glands whilst E. coli (n = 103) was mostly isolated from faecal samples. Disc diffusion testing was used to determine inhibition zone diameters, which were interpreted using either clinical breakpoints or ECOFF, which distinguish the bacterial wild type population from bacteria with acquired or mutational resistance to the compound of interest (non-wild type). Standard ECOFF values were considered as well as sheep-specific values calculated from the data using Normalized Resistance Interpretation (NRI) methodology.
Results: The prevalence of AMR as measured based on clinical breakpoints was low, e.g. 4.0% for penicillin resistance in S. aureus. Estimation of AMR prevalence based on standard ECOFFs was hampered by lack of relevant reference values. In addition, standard ECOFFS, which are predominantly based on human data, bisected the normal distribution of inhibition zone diameters for several compounds in our analysis of sheep isolates. This contravenes recommendations for ECOFF setting based on NRI methodology and may lead to high apparent AMR prevalence. Using bespoke ECOFF values based on NRI, S. aureus showed non-wild type for less than 4% of isolates across 13 compounds, and ca. 13% non-wild type for amoxicillin and ampicillin, while E. coli showed non-wild type for less than 3% of isolates across 12 compounds, and ca. 13% non-wild type for tetracyclines and sulfamethoxazole-trimethoprim.
Conclusion: The apparent prevalence of AMR in bacteria isolated from sheep is highly dependent on interpretation criteria. The sheep industry may want to establish bespoke cut-off values for AMR monitoring to avoid the use of cut-offs developed for other host species. The latter could lead to high apparent prevalence of resistance, including to critically important antimicrobial classes such as 4th generation cephalosporins and carbapenems, suggesting an AMR problem that may not actually exist.