Relative contribution of biological variation and technical variables to zone diameter variations of disc diffusion susceptibility testing.

OBJECTIVES: Disc diffusion is still largely based on manual procedures. Technical variations originate from inoculum preparation, variations in materials, individual operator plate streaking and reading accuracy. Resulting measurement imprecision contributes to categorization errors. Biological variation resembles the natural fluctuation of a measured parameter such as antibiotic susceptibility around a mean value. It is deemed to originate from factors such as genetic background or metabolic state. This study analysed the relative contribution of different technical and biological factors to total disc diffusion variation. METHODS: For calculation of relative error factor contribution to disc diffusion variability, five experiments were designed keeping different combinations of error factors constant. A mathematical model was developed to analyse the individual error factor contribution to disc diffusion variation for each of the tested drug-species combinations. RESULTS: The contribution of biological variation to total diameter variance ranged from 10.4% to 98.8% for different drug-species combinations. Highest biological variation was found for Enterococcus faecalis WT and vancomycin (98.8%) and for penicillinase-producing Staphylococcus aureus and penicillin G (96.0%). Average imprecision of automated zone reading revealed that 1.4%-5.3% of total imprecision was due to technical variation, while materials, i.e. antibiotic discs and agar plates, contributed between 2.6% and 3.9%. Inoculum preparation and manual plate streaking contributed 6.8%-24.8% and 6.6%-24.3%, respectively, to total imprecision. CONCLUSIONS: This study illustrates the relative contributions of technical factors that account for a significant part of total variance in disc diffusion. The highest relative contribution originated from the operator, i.e. manual inoculum preparation and plate streaking. Further standardization of inoculum preparation and plate streaking by automation could potentially increase the precision of disc diffusion and improve the correlation of susceptibility reports with clinical outcome.

Standardization of Operator-Dependent Variables Affecting Precision and Accuracy of the Disk Diffusion Method for Antibiotic Susceptibility Testing.

Parameters like zone reading, inoculum density, and plate streaking influence the precision and accuracy of disk diffusion antibiotic susceptibility testing (AST). While improved reading precision has been demonstrated using automated imaging systems, standardization of the inoculum and of plate streaking have not been systematically investigated yet. This study analyzed whether photometrically controlled inoculum preparation and/or automated inoculation could further improve the standardization of disk diffusion. Suspensions of Escherichia coli ATCC 25922 and Staphylococcus aureus ATCC 29213 of 0.5 McFarland standard were prepared by 10 operators using both visual comparison to turbidity standards and a Densichek photometer (bioMérieux), and the resulting CFU counts were determined. Furthermore, eight experienced operators each inoculated 10 Mueller-Hinton agar plates using a single 0.5 McFarland standard bacterial suspension of E. coli ATCC 25922 using regular cotton swabs, dry flocked swabs (Copan, Brescia, Italy), or an automated streaking device (BD-Kiestra, Drachten, Netherlands). The mean CFU counts obtained from 0.5 McFarland standard E. coli ATCC 25922 suspensions were significantly different for suspensions prepared by eye and by Densichek (P < 0.001). Preparation by eye resulted in counts that were closer to the CLSI/EUCAST target of 10(8) CFU/ml than those resulting from Densichek preparation. No significant differences in the standard deviations of the CFU counts were observed. The interoperator differences in standard deviations when dry flocked swabs were used decreased significantly compared to the differences when regular cotton swabs were used, whereas the mean of the standard deviations of all operators together was not significantly altered. In contrast, automated streaking significantly reduced both interoperator differences, i.e., the individual standard deviations, compared to the standard deviations for the manual method, and the mean of the standard deviations of all operators together, i.e., total methodological variation.