Comparative Evaluation of Phenotypic and Molecular Methods for Early Detection of Carbapenemase-Producing Gram-Negative Bacteria: Implications for Antimicrobial Stewardship

Abstract

Background: Carbapenem resistance among Gram-negative bacilli now sits at critical priority on the WHO pathogen list, yet head-to-head diagnostic-accuracy data comparing phenotypic and molecular assays remain sparse for resource-limited hospitals.[1] We set out to establish, prospectively, how the modified carbapenem inactivation method (mCIM) and its EDTA-based adjunct (eCIM) perform against polymerase chain reaction (PCR) for carbapenemase detection, and whether faster molecular results translate into measurable changes in prescribing.

Methods: Over a four-and-a-half-month period (7 July–15 November 2024) in the Department of Microbiology, Combined Military Hospital (CMH), Sialkot, Pakistan, 113 consecutive, non-duplicate Gram-negative isolates recovered from hospitalised adults underwent parallel mCIM, eCIM, and multiplex PCR testing for blaKPC, blaNDM, blaOXA-48, and blaVIM, following a STARD-compliant design with PCR as the reference standard. Sensitivity, specificity, predictive values, overall accuracy, and Cohen's kappa were computed for each phenotypic test, and multivariable logistic regression was used to identify independent predictors of carbapenem resistance and in-hospital death.

Results: Twenty-seven isolates (23.9%) were PCR-positive, blaKPC being the commonest determinant (37.0% of positives). mCIM reached 88.5% sensitivity and 96.5% specificity (94.6% accuracy; κ = 0.850); eCIM was less sensitive (63.0%) but perfectly specific (100.0%; κ = 0.720). PCR results returned a median of 19 hours faster than culture (23 vs 40 hours, p<0.001), and this speed advantage was associated with prescribing behaviour: 70.4% of PCR-positive patients had their empirical regimen changed, versus none of the PCR-negative group (p<0.001). Chronic kidney disease (adjusted odds ratio [aOR] 4.44) and infection with Acinetobacter baumannii (aOR 4.73, versus E. coli) independently predicted carbapenem resistance, which in turn carried a substantially higher mortality burden (28.0% vs 4.5%, p=0.002).

Conclusion: mCIM offers dependable, low-cost carbapenemase screening for laboratories without molecular infrastructure, while PCR remains the indispensable tool for gene-level surveillance and was associated with a rapid, quantifiable stewardship benefit once results reached the ward. A tiered algorithm phenotypic screening backed by reflex molecular or eCIM confirmation appears well matched to the diagnostic and resource constraints of similar hospital settings.