Concordance Between Multiplex-PCR Resistance-Gene Detection and MIC-Based Susceptibility in Gram-Negative Clinical...

Background/Objectives: Rapid molecular detection of antimicrobial resistance determinants has the potential to optimize antimicrobial therapy before conventional antimicrobial susceptibility testing becomes available. However, the clinical utility of these assays depends on the concordance between detected resistance genes and phenotypic susceptibility. This study aimed to evaluate the genotype-phenotype concordance of carbapenemase and β-lactamase genes detected by the Unyvero multiplex-PCR platform with minimum inhibitory concentration (MIC)-based susceptibility profiles obtained using the phenotypic AST comparator method in Gram-negative clinical isolates. Methods: A total of 117 non-duplicate Gram-negative clinical isolates recovered from hospitalized patients at a Romanian tertiary-care hospital were analyzed. Resistance determinants, including blaKPC, blaOXA-48, blaOXA-23, blaCTX-M, blaTEM, together with aminoglycoside, fluoroquinolone, and sulfonamide resistance markers, were identified using the Unyvero multiplex-PCR system. Phenotypic antimicrobial susceptibility testing was performed using the VITEK^®2 automated system according to EUCAST criteria. Genotype-phenotype concordance was evaluated using sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and Cohen's κ coefficient. Results: Overall, 78.8% (78/99) of isolates were classified as non-susceptible. blaOXA-48 was the most frequently detected carbapenemase gene (n = 12), followed by blaKPC (n = 9) and blaOXA-23 (n = 3). blaKPC demonstrated a sensitivity of 100% for predicting imipenem resistance in Enterobacterales, with moderate agreement (κ = 0.58), whereas blaOXA-23 showed complete diagnostic performance (100% sensitivity and PPV) for carbapenem resistance in Acinetobacter baumannii. In contrast, all blaOXA-48-positive Enterobacterales remained phenotypically susceptible to carbapenems, indicating poor genotype-phenotype concordance. Likewise, blaCTX-M, gyrA83/87, and qnrA/B/S showed limited diagnostic performance for corresponding phenotypic resistance. Conclusions: Phenotype-genotype concordance varies across resistance determinants and Gram-negative bacterial species. Detection of blaKPC and blaOXA-23 may support early optimization of antimicrobial therapy, whereas blaOXA-48 should be interpreted together with conventional phenotypic antimicrobial susceptibility testing. Integrating rapid molecular diagnostics with phenotypic susceptibility testing may improve antimicrobial stewardship and optimize the management of multidrug-resistant Gram-negative infections.