Modeling the effect of the number of presumptive colonies selected from agar plate on the recovery and prevalence of E. coli O157:H7 and non-O157 Shiga-toxin producing E. coli from dairy manure.

内容摘要

Although various laboratory protocols exist for the detection of Shiga toxin-producing E. coli (STEC), the number of presumptive colonies selected per sample for confirmation remains a critical yet largely overlooked factor that is has not been standardized between investigators. Our study evaluated how the number of presumptive colonies selected for confirmation per fecal sample influences the probability to detect true positive samples, which when optimized can facilitate more accurate estimates of STEC prevalence in feces while simultaneously minimizing costs associated with confirming excessive numbers of colonies. An experimental trial was first conducted using inoculated concentrations of E. coli O157:H7 (-1.1 to 4.3 log&#x2081;&#x2080; CFU/g) and non-O157 STEC (-1.9 to 4.6 log&#x2081;&#x2080; CFU/g), with up to 45 presumptive colonies selected for qPCR confirmation. In parallel, a dairy fecal longitudinal survey was conducted on 14 dairy farms in California to examine how the number of presumptive colonies selected per sample affected STEC prevalence estimates, using the 45-presumptive colony protocol as the benchmark method. Using colony-based positive predictive values (cPPV) generated from the experimental trial for CT-SMAC, Rainbow Agar O157, and CHROMagar STEC selective agars, the hypergeometric (HG) and binomial (BN) distributions were used to calculate the probability of diagnosing a sample as true positive (PrHG and PrBN, respectively) and the computation of the STEC prevalence for dairy fecal samples as a function of number of confirmed colonies. The experimental trial comprised 126 spiked samples. Results showed that the high cPPV (~100%) values observed for all but the lowest bacterial concentrations (-1.1 log&#x2081;&#x2080; CFU/g) for E. coli O157:H7 generated a&#x202f;>&#x202f;90% probability of true positive detection (PrHG and PrBN) when as few as 3 to 4 presumptive colonies were selected. Selecting only 3-4 presumptive colonies yielded a model-predicted prevalence of ~33%, representing a difference of <5% from the observed prevalence of 37.9%. This difference decreased to <2% when 8 (HG) or 11 (BN) presumptive colonies were selected. In contrast, for non-O157 STEC, &#x2265;38 colonies per sample had to be selected when the maximum concentration of bacteria was &#x2264;1.6 log&#x2081;&#x2080; CFU/g in order to achieve &#x2265;90% detection probability of true positives. Selecting 31 (HG) or 56 (BN) presumptive colonies resulted in a model-predicted prevalence of ~38%, which differed by 5% from the observed prevalence (42.9%). The difference was further reduced to <2% when 38 (HG) or 92 (BN) colonies were selected. In conclusion, given the large impact of colony selection totals on the PrBN or PrHG and STEC prevalence estimates, we recommend that authors report the number of presumptive colonies selected per sample in related future research in order to improve inter-study comparability and accurate interpretation of results from studies evaluating E. coli O157:H7 and especially non-O157 STEC in manure samples.