Multi-centre evaluation of a phenotypic extended spectrum β-lactamase detection guideline in the routine setting

This study aimed to evaluate the routine setting performance of a guideline for phenotypic detection of extended spectrum β-lactamases (ESBLs) in Enterobacteriaceae, recommending ESBL confirmation with Etest or combination disc for isolates with a positive ESBL screen test (i.e. cefotaxime and/or ceftazidime MIC >1 mg/L or an automated system ESBL warning). Twenty laboratories submitted 443 Enterobacteriaceae with a positive ESBL screen test and their confirmation test result (74% Escherichia coli, 12% Enterobacter cloacae, 8% Klebsiella pneumoniae, 3% Proteus mirabilis, 2% Klebsiella oxytoca). Presence of ESBL genes was used as reference test. Accuracy of local phenotypic ESBL detection was 88%. The positive predictive value (PPV) of local screen tests was 70%, and differed per method (Vitek-2: 69%, Phoenix: 68%, disc diffusion: 92%), and species (95% K. pneumoniae-27% K. oxytoca). A low PPV (3%) was observed for isolates with automated system alarm but third-generation cephalosporin MICs <2 mg/L. Local ESBL confirmation had a PPV and negative predictive value (NPV) of 93% and 90%, respectively. Compared with centrally performed confirmation tests, 7% of local tests were misinterpreted. Combination disc was more specific than Etest (91% versus 61%). Confirmation tests were not reliable for P. mirabilis and K. oxytoca (PPV 33% and 38%, respectively, although NPVs were 100%). In conclusion, performance of Etests could be enhanced by education of technicians to improve their interpretation, by genotypic ESBL confirmation of P. mirabilis and K. oxytoca isolates with positive phenotypic ESBL confirmation, and by interpreting isolates with a positive ESBL alarm but an MIC <2 mg/L for cefotaxime and ceftazidime as ESBL-negative.