Evaluation of two automated systems for colistin susceptibility testing of carbapenem-resistant Acinetobacter baumannii clinical isolates

Background: Colistin is commonly needed for the treatment of infections due to carbapenem-resistant Acinetobacter baumannii (CRAB) and the determination of its in vitro activity is obviously important. However, the accurate routine antimicrobial susceptibility testing (AST) of colistin is still challenging. The only acceptable method for colistin AST is broth microdilution (BMD); disc and gradient diffusion assays are inappropriate and the performance of semi-automated systems has not been validated. Objectives: In the present study, two commonly used semi-automated systems were evaluated for colistin AST of contemporary CRAB clinical isolates. Methods: A total of 117 single-patient CRAB isolates collected randomly during 2015 from distinct tertiary hospitals located throughout Greece were tested. Colistin MICs were determined using the semi-automated systems Phoenix100 and Vitek2 and also agar dilution (AD), compared with the reference BMD. Results: Colistin resistance rates for Phoenix100/Vitek2/AD/BMD were 15.4%/16.2%/35.9%/24.8%. The essential/categorical agreement rates were as follows: Phoenix100, 91.5%/88.9%; Vitek2, 88.9%/89.7%; and AD, 93.2%/87.2%. Alarmingly high rates of very major errors (VMEs) were observed for Phoenix100 (41.4%) and Vitek2 (37.9%), while major errors (MEs) were limited (1.1% by both systems); VMEs were much more common for isolates with MICs of 2 mg/L than for isolates with MICs of <= 1 mg/L, as determined by automated methods. AD produced considerably higher colistin MICs, yielding MEs of 15.9%. Conclusions: Colistin resistance of A. baumannii is greatly underestimated by Phoenix100/Vitek2, potentially leading to inappropriate colistin administration. Colistin AST results by automated systems within the susceptible range, particularly those at the susceptibility breakpoint (2 mg/L), need to be validated by BMD.