OBJECTIVE: To evaluate various sampling strategies for potential use in measuring prevalence of antimicrobial susceptibility in cattle. SAMPLE POPULATION: 500 isolates of non-type-specific Escherichia coli (NTSEC) isolated from the feces of 50 cows from 2 dairy farms (25 cows/farm and 10 isolates/cow). PROCEDURES: Diameters of inhibition zones for 12 antimicrobials were analyzed to estimate variation among isolates, cows, and farms and then used to determine sampling distributions for a stochastic simulation model to evaluate 4 sampling strategies. These theoretic sampling strategies used a total of 100 isolates in 4 allocations (1 isolate from 100 cows, 2 isolates from 50 cows, 3 isolates from 33 cows, or 4 isolates from 25 cows). RESULTS: Analysis of variance composition revealed that 74.2% of variation was attributable to isolates, 18.5% to cows, and 7.3% to farms. Analysis of results of simulations suggested that when most of the variance was attributable to differences among isolates within a cow, culturing 1 isolate from each of 100 cows underestimated overall prevalence, compared with results for culturing more isolates per cow from fewer cows. When variance was not primarily attributable to differences among isolates, all 4 sampling strategies yielded similar results. CONCLUSIONS AND CLINICAL RELEVANCE: It is not always possible to predict the hierarchical level at which clustering will have its greatest impact on observed susceptibility distributions. Results suggested that sampling strategies that use testing of 3 or 4 isolates/cow from a representative sample of all animals better characterize herd prevalence of antimicrobial resistance when impacted by clustering.
OBJECTIVE: To evaluate various sampling strategies for potential use in measuring prevalence of antimicrobial susceptibility in cattle. SAMPLE POPULATION: 500 isolates of non-type-specific Escherichia coli (NTSEC) isolated from the feces of 50 cows from 2 dairy farms (25 cows/farm and 10 isolates/cow). PROCEDURES: Diameters of inhibition zones for 12 antimicrobials were analyzed to estimate variation among isolates, cows, and farms and then used to determine sampling distributions for a stochastic simulation model to evaluate 4 sampling strategies. These theoretic sampling strategies used a total of 100 isolates in 4 allocations (1 isolate from 100 cows, 2 isolates from 50 cows, 3 isolates from 33 cows, or 4 isolates from 25 cows). RESULTS: Analysis of variance composition revealed that 74.2% of variation was attributable to isolates, 18.5% to cows, and 7.3% to farms. Analysis of results of simulations suggested that when most of the variance was attributable to differences among isolates within a cow, culturing 1 isolate from each of 100 cows underestimated overall prevalence, compared with results for culturing more isolates per cow from fewer cows. When variance was not primarily attributable to differences among isolates, all 4 sampling strategies yielded similar results. CONCLUSIONS AND CLINICAL RELEVANCE: It is not always possible to predict the hierarchical level at which clustering will have its greatest impact on observed susceptibility distributions. Results suggested that sampling strategies that use testing of 3 or 4 isolates/cow from a representative sample of all animals better characterize herd prevalence of antimicrobial resistance when impacted by clustering.
Authors: Angela B Bosman; Jaap A Wagenaar; Jaap Wagenaar; Arjan Stegeman; Hans Vernooij; Dik Mevius Journal: PLoS One Date: 2012-09-07 Impact factor: 3.240