Quality of evidence used for the management of antimicrobial resistance in Australian animals

Abstract

Integral to the success of surveillance programs is the quality of the measurement systems used to collect data. However, the performance of the measurement systems used to evaluate antimicrobial resistance and antimicrobial use is poorly defined. This thesis, therefore, examines the quality of evidence arising from the phenotypic assays and questionnaires used in the surveillance of animals. The performance of disc diffusion was evaluated to determine its fitness-of-purpose as a source of data for clinical decision-making and surveillance. Zone diameter and minimum inhibitory concentration values obtained from the first Australia-wide prevalence studies of clinical Escherichia coli and Staphylococcus pseudintermedius were used to estimate the accuracy of disc diffusion relative to broth microdilution. Conventional measures of test accuracy were described, including diagnostic sensitivity, specificity, and area-under-the-receiver-operating characteristic (ROC) analysis. For most antimicrobials evaluated, disc diffusion was accurate at predicting the resistance of clinical E. coli and S. pseudintermedius that could otherwise be determined by broth microdilution. The assay performed strongly for ciprofloxacin and ceftiofur, and less favourably for amoxicillin-clavulanic acid, cephalothin, and cefoxitin. For S. pseudintermedius and oxacillin, the accuracy of broth microdilution was moderately better than disc diffusion relative to mecA real-time PCR. The precision of disc diffusion was investigated in a test-retest study using a linear mixed-model to estimate intra- and inter-laboratory agreement. Agreement was measured as repeatability (r) and reproducibility (R). The precision of disc diffusion was generally satisfactory for most antimicrobial agents, including ceftiofur (r=4.9mm, R=5.8mm) and gentamicin (r=4.9mm, R=5.4mm). However, the extent of variation in ampicillin (r=4.6mm, R=6.5mm) and trimethoprim-sulfamethoxazole (r=6.6mm, R=7.2mm) was of some concern. The management of antimicrobial resistance is aided by the collection of data on the use of antimicrobial agents via questionnaires or other survey tools. In this thesis, the Australian beef feedlot sector was used as a case study to examine a common survey method in which multi-stakeholder engagement is expected, often leading to methodological constraints in survey design. Here, a mailed questionnaire was used to obtain information on antimicrobial use in beef feedlots. The response rate was 16.1%. For those responding to the survey, the use of antimicrobials was found to be appropriate for the purpose indicated, and there was a strong preference for drugs of low importance to human health. While the low response rate dictates that inferences could only be weakly extended to the broader beef feedlot population, the data was of value in informing the development of antimicrobial stewardship guidelines and acted as a staging position for further research into antimicrobial use in other animal sectors. However, more reliable methods of survey delivery should be considered for the on-going collection of antimicrobial use data at the farm-level. Overall, this thesis concludes that for E. coli and S. pseudintermedius, susceptibility data from disc diffusion or broth microdilution generated in veterinary laboratories can contribute to national surveillance programs. This information, coupled with data from surveys of antimicrobial use at the farm-level, will be of substantial benefit to efforts aimed at managing antimicrobial resistance in animals.