bla IMP-4 -associated imipenem susceptibility discordance in the Klebsiella pneumoniae complex: characterization and a phenotype-guided confirmatory workflow

ABSTRACT Automated antimicrobial susceptibility testing (AST) may generate carbapenem minimum inhibitory concentration (MIC) results differing from reference methods, complicating recognition of carbapenemase-producing Klebsiella pneumoniae complex isolates. We characterized carbapenem susceptibility discordance in a retrospective discovery cohort of 133 carbapenemase-producing isolates collected in Shenzhen (2004–2017) by comparing imipenem (IPM) and meropenem (MEM) MICs obtained by VITEK 2 and broth microdilution (BMD). Confirmed discordance was defined as repeat-confirmed raw MIC differences of ≥2 log 2 dilutions. Features were assessed by whole-genome sequencing, mCIM/eCIM, and exploratory phenotypic assays. Cross-card reproducibility was assessed using an alternative VITEK 2 AST card. A separate retrospectively enriched cohort of 166 K. pneumoniae isolates (2018–2024) was used to evaluate a phenotype-guided reflex-testing workflow. Twenty-nine isolates showed confirmed carbapenem AST discordance, including 28 with IPM discordance. Of these, 27 had raw VITEK 2 IPM MICs at least 2 log₂ dilutions higher than those obtained by BMD, whereas one was reported as resistant after system adjustment despite a low raw MIC. All discrepant isolates carried bla IMP-4 as the sole detected carbapenemase gene, had relatively low carbapenem MICs by BMD, and were distributed across 14 sequence types. Cross-card retesting reproduced the direction of IPM elevation, although the magnitude varied. In a combined internal analysis set of 275 isolates, the phenotype-guided workflow identified unflagged ≥2 log₂ MIC discrepancies with 81.4% sensitivity (95% CI, 67.4%–90.3%) and 92.2% specificity (95% CI, 88.1%–95.0%). These findings define a repeat-confirmed bla IMP-4 -associated, IPM-predominant discordance between VITEK 2 and BMD and support phenotype-guided reflex-testing to identify unflagged MIC discrepancies and prioritize confirmatory testing in routine laboratories. IMPORTANCE Clinical microbiology laboratories rely on automated systems to support antimicrobial susceptibility interpretation, but these systems may not always recognize method-dependent carbapenem susceptibility discordance. In this study, we show that a specific group of bacteria carrying the bla IMP-4 gene can produce potentially misleading carbapenem susceptibility results, particularly for imipenem, when tested by automated methods. These discrepancies may not be flagged by the instrument and could therefore be overlooked in routine practice. By combining routine laboratory data with genetic analysis and additional testing, we identify a recognizable pattern associated with these hidden discrepancies and examine a phenotype-guided confirmatory workflow to support their laboratory recognition. Our findings highlight the importance of targeted confirmatory testing and provide a framework for local validation of laboratory-actionable carbapenem susceptibility discrepancies, thereby supporting more reliable AST interpretation and antimicrobial resistance surveillance.