Evaluation of an expanded antibiotic resistance gene panel on prediction of antimicrobial susceptibility results for Gram-negative bacteria in blood cultures.

The QIAstat-Dx BCID Panels (RUO) ("QIAstat," QIAGEN, Hilden, Germany) for identification of 13 Gram-negative bacteria and 18 antimicrobial resistance (AMR) gene groups was evaluated. The study was conducted in two phases; in phase 1, analytical performance was evaluated against 154 challenge isolates against whole genome sequencing data. In this phase, sensitivity and specificity of organism identification calls were 153/154 (99.3%) and 1,748/1,749 (99.8%), respectively. For AMR genes, sensitivity was 434/435 (99.8%) and specificity was 2,334/2,337 (99.9%). One false-negative bla_IMP, one false-positive bla_CTX-M, and two false-positive aac-6'-lb detections were noted in this challenge set of organisms. In phase 2, 101 clinical blood culture isolates of Gram-negative rods were evaluated by the multiplexed PCR versus reference broth microdilution, for the ability of identification combined with AMR genes to predict final susceptibility results. Negative predictive values were 92.8% for ampicillin resistance (100% for Escherichia coli), 93.4% for ceftriaxone, 97.4% for ceftazidime, and 98.7% for cefepime. In constrast, negative predictive values for current standard of care (identification plus detection of bla_CTX-M) ranged from 56.5% to 88.8%. This study demonstrated additive value of additional beta-lactamase genes for bacteria isolated from blood cultures.

Importance: Prediction of Gram-negative bacteria resistance through detection of resistance genes is complex. This study evaluated a novel, direct-from-blood or bacterial isolate multiplexed PCR for the detection of 17 resistance genes, and evaluated the prediction of antimicrobial susceptibility.