Antimicrobial use in dairy cattle and its potential impact on antimicrobial resistance in enteric bacteria: sampling, data collection and analytical methods

Abstract

Some researchers have suggested that antimicrobial-use in animals may have an effect on antimicrobial resistance in bacteria isolated from humans. Before this belief can be studied, a possible connection between antimicrobial-use in food-animals and resistance in bacteria isolated from those animals needs to be studied. Some estimates exist on national antimicrobial-use in food-animals. However, trying to relate these estimates to resistance prevalence among bacteria isolated from food-animals across the country may not appropriate. Potential problems that may surface when attempting to study this relationship are explored. Because of these problems, standard methodology to measure antimicrobial-use and resistance needs to be developed. In this study, various methods for data collection on antimicrobial-use and resistance were evaluated, as well as sampling strategies and statistical analysis methods, to study the possible relationship between antimicrobial-use in dairy cattle and resistance among enteric bacteria isolated from those cattle. Potential confounding and interaction factors were evaluated. Measurement of resistance in enteric bacteria was studied testing one or more isolates per animal, the latter resulting in a better estimate of resistance prevalence. Comparison of handwritten and computerized records for data collection on antimicrobial-use resulted in concluding that, currently, computerized records lack an appropriate field code to record detailed data on antimicrobial-use. As for sampling strategies, random sampling stratified across pens was a better strategy than repeatedly sampling a cohort group of treated and non-treated animals. The main reasons for this conclusion were the lack of attrition of studied animals, effective representation of subpopulations of the dairy, easiness and time-effectiveness of the stratified random sampling strategy. Evaluation of statistical analysis methods for the study at hand showed that crude analysis resulted in erroneous conclusions as to the effect of antimicrobial-use on resistance, likely because it ignores the lack of independence between isolates obtained from the same animals on different occasions. A better option was the use of hierarchical multivariate analysis, which revealed that other factors such as age in calves and production group and parity in cows were strongly associated with the probability of isolation of resistant enteric bacteria. New questions that arose during this study are outlined.