Carbapenem resistance in Gram-negative bacilli (GNB) is rapidly increasing and spreading worldwide, with significant clinical, therapeutic and epidemiological repercussions (Paniagua-García et al., 2024). According to data reported by EARS-Net for the EU, the incidence of carbapenem-resistant Klebsiella pneumoniae increased by almost 50% between 2019 and 2022, while the incidence of Pseudomonas aeurginosa (18%–19%) and Acinetobacter baumannii (36%–39%) remained constant during this period (ECDC, 2023). From an epidemiological point of view, the acquisition of plasmid-encoded carbapenemases is the main mechanism of carbapenem resistance (Khaled and Aboshanab, 2020). The most frequently reported belong to molecular classes A (KPC-like), B (VIM-like, IMP-type and NDM-like) and D (OXA-48-like in Enterobacterales). In recent years, there has been a disturbing increase worldwide in the detection of isolates co-producing multiple acquired β-lactamases, such as different carbapenemases, extended-spectrum beta-lactamases (ESBL) and/or plasmid-encoded AmpC-type enzymes. The co-production of multiple carbapenemases is rapidly increasing and has been reported worldwide (Tadese et al., 2022). Genes encoding carbapenemases are often linked to other genetic determinants that cause resistance to other antimicrobials (such as colistin, aminoglycosides and/or fosfomycin, among others) and confer multidrug resistance, extensive drug resistance or pandrug resistance. Therefore, infections caused by isolates co-producing several carbapenemases represent a serious clinical and microbiological problem, as well as a serious global problem for human health due to the limited or no therapeutic alternatives available and the high impact of these infections on morbidity, mortality, and healthcare costs (Bonomo et al., 2017).

The detection of carbapenemase production presents important diagnostic, clinical, and epidemiological challenges. In the laboratory, accurate identification can be difficult due to the type of carbapenemase—such as certain OXA-48-type variants with low hydrolytic activity or low expression levels—and the presence of additional resistance mechanisms like porin loss in AmpC-or ESBL-hyperproducing isolates. These factors can lead to missed or delayed detection, resulting in inappropriate treatment, poorer patient outcomes, and increased mortality. From an epidemiological perspective, underdetection compromises surveillance efforts and delays outbreak containment. High-quality testing is therefore essential—not only to guide effective antimicrobial therapy but also to support infection control and limit the spread of resistance within clinical settings.

The EUCAST algorithm for the detection of carbapenemase production includes susceptibility testing of carbapenems, especially ertapenem (high sensitivity) and meropenem (high specificity), as phenotypic markers for carbapenemase production (The European Committee on Antimicrobial Susceptibility Testing, 2017). These results need to be confirmed by phenotypic and molecular assays due to the limitations of ertapenem in terms of specificity. A wide variety of phenotypic methods are currently available for the detection of the most frequent carbapenemases (Tamma and Simner, 2018).

The main objective of this study was to evaluate the ability of Spanish clinical microbiology laboratories to report reliable carbapenem susceptibility test results and to detect carbapenemase production and/or carbapenemase genes in double carbapenemase-producing (DCP) and single carbapenemase-producing (SCP) isolates.

The multiple carbapenemase-producing isolates that are emerging worldwide have potentially worrying clinical, epidemiological and microbiological implications (Meletis et al., 2015). The present study is, to the best of our knowledge, the first to focus on DCP and SCP isolates to determine whether interpretation of antimicrobial susceptibility results (discrepant MICs and errors) and detection of carbapenemase genes are more difficult in DCP isolates than in SCP isolates.

We noted some relevant findings in our study that may be difficult to compare with the results of previous studies that included carbapenemase-producing GNB isolates because differences between DCP and SCP isolates were not investigated.

For antimicrobial susceptibility test results, the gradient strip method was found to be the least reliable. This method returned unacceptably high percentages of discrepant MICs and very major errors (false susceptible), which is consistent with previously published findings (Lee and Chung, 2015; Markelz et al., 2011). There were significant differences between DCP and SCP isolates, suggesting that the interpretation of antimicrobial test results is less reliable for DCP compared to SCP isolates, with an unacceptably high percentage of discrepant MICs, lower categorical agreement, and errors, particularly very major errors with imipenem, and minor errors with imipenem and meropenem. However, a high number of very major errors were detected for ertapenem in SCP isolates. The reasons for this were not investigated, but may be explained by methodological factors related to inoculum preparation, decreased antimicrobial activity, or interpretation of MIC values, especially when small bacterial subpopulations are growing within the inhibition halo (heteroresistance). Heteroresistant subpopulations can be difficult to detect (Humphries et al., 2018; Fernández-Cuenca et al., 2021; Harino et al., 2013), as has been observed in porin-deficient K. pneumoniae isolates harboring bla_OXA-48, or isolates carrying bla_VIM-type or bla_KPC-type genes (Tato et al., 2010; Nodari et al., 2015; Adams-Sapper et al., 2015; López-Camacho et al., 2019).

When false resistance (major error) to meropenem happens, the problem can be resolved by testing with a second, more reliable antimicrobial susceptibility testing method. False susceptibility to ertapenem and imipenem (very major errors) on the other hand often goes undetected because susceptible results are not routinely checked in the laboratory. It is essential therefore that laboratories have the ability to detect both types of false results when testing carbapenems in order to avoid potentially serious consequences for the management and outcome of patients infected with DCP isolates, as has been described for SCP isolates. These consequences include treatment failure associated with increased mortality and prolonged length of stay (Bartoletti et al., 2022) following false susceptible results, or missing a possible good opportunity to use carbapenems, in accordance with EUCAST recommendations, following false resistance results. Carbapenemase production can usually be inferred from carbapenem susceptibility results, so that it is essential that antimicrobial susceptibility testing is reliable (The European Committee on Antimicrobial Susceptibility Testing, 2017; Markelz et al., 2011; Kanahashi et al., 2021).

The second major objective of our study was to assess the ability to detect carbapenemase production in SCP and DCP isolates. In principle, a higher rate of detection of positive results would be expected among DCP isolates than SCP isolates on the grounds that, regardless of gene expression or gene copy number, the amount of carbapenemase produced in DCP isolates is likely to be higher due to the presence of two carbapenemase-encoding genes. In our study, this hypothesis was true for lateral flow immunochromatography and the carbapenem inactivation assay, but not for the double-disk assay, hydrolysis-based assays and chromogenic media. This is a very important finding and suggests that lateral flow immunochromatography or carbapenem inactivation assays should be included in the algorithms currently used for detection of carbapenemases. Double-disk inhibition assays and hydrolysis-based assays returned unexpectedly high rates of false negative results in both DCP and SCP isolates. There are several possible explanations for this. First, use of the double-disk inhibition assay without the temocillin disk is highly sensitive, but has low specificity for phenotypic detection of bla_OXA-48. Second, methodological problems (such as inoculum preparation at low bacterial concentrations) or the interpretation of results (e.g., color changes using hydrolysis-based assays with colorimetric detection) (Kanahashi et al., 2021). Third, the type of carbapenemase produced, especially bla_OXA-48, due to the low catalytic efficiency of this group of carbapenemases, as reported previously (Österblad et al., 2014), and IMP-type carbapenemases, especially bla_IMP-8, which is consistent with other studies showing the potential difficulty of detecting this heterogeneous group of class B carbapenemases (Jenkins et al., 2020; Findlay et al., 2015).

In our study, the presence of an ESBL or type of ESBL (bla_CTX-M-15 or bla_SHV-12) did not affect the detection of carbapenemase production, regardless of whether or not a carbapenemase was co-produced. Nevertheless, this finding should be interpreted with caution due to the small number of ESBL-producing isolates included. Furthermore, the possibility cannot be ruled out. The presence of ESBLs in isolates with significant efflux pump overexpression and lacking major porins involved in carbapenem resistance may hinder the detection of carbapenemase production, especially with certain carbapenemase gene variants with weak hydrolyzing activity (Hamzaoui et al., 2018). Further experiments with laboratory-derived mutants would help to elucidate the true role of ESBL production in carbapenemase detection.

In our study, genotypic tests showed a higher overall sensitivity for DCP isolates than phenotypic assays, which was consistent with many previous studies (Findlay et al., 2015; Baeza et al., 2019). However, the overall concordance for carbapenemase gene detection was lower in DCP isolates than in SCP isolates, mainly due to the poor detection of bla_VIM-1 and bla_IMP-8 genes. In the case of isolates co-producing bla_VIM-1 and bla_OXA-48, we did not exclude the possibility of loss of the plasmid harboring bla_VIM-1, as most assays failed to detect this gene. Loss of plasmid genes encoding carbapenemases or other types of β-lactamases has been reported previously and appears to be rare (Hopkins et al., 2018). A recent study found that the plasmid loss rate in bla_OXA-like-producing Klebsiella spp. and Enterobacter cloacae was 4.3% and 12.4% (Mahazu et al., 2022).

Of the assays that were most frequently performed, the least reliable were the LAMP Eazyplex® SuperBug CRE assay, followed by the mRT-PCR Xpert Carba-R assay, which showed lower concordance with the reference in DCP versus SCP isolates. This was an unexpected result and may be related to certain carbapenemase genes being more difficult to detect with these assays than others, as has been reported for some IMP-type carbapenemase variants, including bla_IMP-8 (Kanahashi et al., 2021; Huang et al., 2022; Lowe et al., 2020). Yu-Tsung et al. tested a collection of carbapenemase-producing isolates, and the only negative result they obtained with both NG-CARBA 5 and Xpert Carba-R was with a K. pneumoniae isolate co-producing VIM-1 and IMP-8 (Huang et al., 2022).

This study has several minor limitations. First, the small number of isolates, although this is not usual for this type of study. Second, the source of discrepant MICs and errors was not investigated, as this was not an objective of the study. Third, the role of other mechanisms of resistance to carbapenems (such as AmpC-hyperproduction and/or absence of porins and/or overexpression of efflux pumps), which could better explain some of the discrepancies observed, but these resistance mechanisms were not characterized in the isolates selected for this study. Fourth, we did not request information on the working protocols or algorithms used at each center for the detection of carbapenemase production and it cannot be ruled out that some centers used a different method for detection of carbapenemase production with the reference isolates than with isolates routinely collected in their laboratories.

In conclusion, our study demonstrates that discrepancies in MICs and errors are related to the AST system used, especially the gradient strip, and to the carbapenem used, and result in unacceptable very major errors and major errors. For the detection of carbapenemase production, phenotypic assays are more sensitive for DCP isolates than for SCP isolates; the most sensitive assays were lateral flow immunochromatography and the carbapenem inactivation assay, and the genes most challenging for phenotypic detection of carbapenemase production were bla_OXA-48 and the two IMP-type carbapenemases, bla_IMP-8 and bla_IMP-23, especially using double-disk inhibition assays and hydrolysis-based assays. Detection of certain carbapenemase genes, especially bla_VIM-1 and bla _IMP-8, in DCP isolates is unreliable using the LAMP assay, the Eazyplex® SuperBug CRE, and the mRT-PCR Xpert Carba-R.