Relative performance of antimicrobial susceptibility assays on clinical Escherichia coli isolates from animals

The assessment of antimicrobial resistance in bacteria derived from animals is often performed using the disc diffusion assay. However broth-microdilution is the preferred assay for national antimicrobial resistance surveillance programs. This study aimed to evaluate the accuracy of disc diffusion relative to broth-microdilution across a panel of 12 antimicrobials using data from a collection of 994 clinical Escherichia coli isolates from animals. Disc diffusion performance was evaluated by diagnostic sensitivity, specificity, likelihood ratio pairs and receive-operating characteristic (ROC) analysis. Data was dichotomised using CLSI susceptible and resistant clinical breakpoints. In addition, disc diffusion breakpoints produced using diffusion Breakpoint Estimation Testing Software (dBETS) were evaluated. Analysis revealed considerable variability in performance estimates for disc diffusion susceptible and resistant breakpoints (AUC ranges: 0.78-0.99 and 0.92-1.0, respectively) across the panel of antimicrobials. Ciprofloxacin, tetracycline, and ampicillin estimates were robust across both breakpoints, whereas estimates for several antimicrobials including amoxicillin-clavulanic acid, cefoxitin and gentamicin were less favourable using susceptible breakpoints. Overall performance estimates were moderately improved when dBETS susceptible breakpoints were applied. For most antimicrobials, disc diffusion was accurate at predicting resistance of clinical E. coli from animals that could otherwise be determined by broth-microdilution. While disc diffusion is suboptimal for assessing the proportion of fully susceptible isolates for some drugs, sensitivity and specificity estimates provided here allow for the use of standard formula to correct this. For this reason, disc diffusion has applicability in national surveillance provided the performance of the assay is taken into account.