Comparison of annual and regional variation in multidrug resistance using various classification metrics for generic Escherichia coli isolated from chicken abattoir surveillance samples in Canada

Abstract

Antimicrobial resistance (AMR) and related multidrug resistance (MDR) are important global public health issues. The Canadian Integrated Program for Antimicrobial Resistance Surveillance (CIPARS) conducts surveillance of AMR in enteric bacteria and monitors MDR. However, the analysis of MDR is complicated by the lack of consensus for MDR definitions. The objectives were to describe the most common resistance patterns in generic E. coli isolates from chicken cecal samples and determine the impact of using different MDR metrics for analysis of annual and regional variation in MDR. From 2006 to 2015, 1598 E. coli isolates were collected from chickens at slaughter for CIPARS. Three MDR classification metrics were used: MDR-drug (MDR if the isolate was resistant (R) to ≥3 of the 13 antimicrobials); MDR-cat (MDR if R to ≥3 of the 9 antimicrobials categories); and MDR-class (MDR if R to ≥3 of the 6 antimicrobial classes). The most frequent resistance patterns overall, and by year and region were extracted along with patterns that included resistance to quinolones, and third generation cephalosporins and/or β-lactams with β-lactamase inhibitors. For each MDR metric, mixed logistic regression models, which included random intercepts for abattoir, were fitted to analyze the association between prevalence of MDR, and year and region. Interaction effects between year and region were evaluated. Overall, and in all years and regions, non-resistant was the most common resistance pattern (24.9%, 95% CI 22.8–27.1%). Resistance patterns that included third generation cephalosporins and β-lactams with β-lactamase inhibitors were common. The prevalence of MDR was variable: MDR-class 38.5% (95% CI 36.1–41.0%); MDR-cat 49.4% (95% CI 46.9–51.9%); and MDR-drug 53.3% (95% CI 50.8–55.8%). Based on models fitted with individual fixed effects, significant annual variation in the prevalence of MDR was identified with MDR-drug and MDR-class models. Significant regional variation was identified for all three MDR metric models. Significant interaction effects between year and region were identified with the MDR-drug and MDR-cat multivariable mixed logistic regression models. The interpretation of the association between the prevalence of MDR, and year and region differed depending on the MDR metric used. These results are supportive of the previous concerns that caution must be taken when comparing MDR results between studies. Global consensus is needed for the optimal MDR classification metric for foodborne enteric bacteria AMR surveillance.