The rising incidence of multidrug resistant organisms has made interpretation of antibiograms more challenging for Microbiologists. Assistance through use of interpretive algorithms in automated platforms such as the Vitek® 2XL has been valuable. The comparative performance of these systems has not been well documented. The expert systems (ES) of the BD Phoenix M50 (PHX) and the Vitek 2XL (V2) was tested against a collection of organisms with confirmed resistances (molecular +/- phenotypic).
Isolates were tested simultaneously on both platforms. The instruments interpretation of resistance mechanisms was evaluated based on the ability to accurately classify a number of key resistance mechanisms including extended spectrum β-lactamases (ESBL), acquired AmpC β-lactamases (AmpC), carbapenem resistant Enterobacteriaceae (CRE), vancomycin resistant Enterococci (VRE), glycopeptide non-susceptible Staphylococci and MRSA (Methicillin resistant Staphylococcus aureus).
From 200 Gram negative bacilli tested, there were 158 isolates which harbored 178 designated acquired resistances. For specificity, 42 isolates were included that had no acquired resistance mechanisms detected either phenotypically or genotypically. The sensitivity for the ES corresponding to the reference genotype/phenotype was (PHX 90%, V2 77%), although the error rate was higher when analysed as a proportion of total tests (PHX 91%, V2 78%).
100 Gram positive cocci were tested. For 30 VRE, the sensitivity of the ES to correctly classify vanA & vanB was 100% for both systems. For low MIC vanB, the sensitivity was 30% for PHX, 10% for V2, however the limitation of the PHX system was its inability to differentiate between vanA & vanB. Both ES correctly classified methicillin resistance across the variety of Staphylococci tested. The sensitivity of the V2 ES to alert to possible hVISA/VISA was 53.3% compared to 0% on the PHX.
Conclusion: The performance of the ES was difficult to compare due to different levels of sophistication of the reporting algorithm, however, the V2 had superior performance, with a greater specificity. Both systems were designed to maximise sensitivity and should be considered screening algorithms only.