Class A carbapenemases.

One of the most common mechanisms of carbapenem resistance among class A enzymes is the production of KPC β-lactamases. KPCs were initially detected in a clinical isolate in 1996 in North Carolina; since then, 19 variants have been discovered (21–24). KPC has been found in a variety of Enterobacteriaceae, including Klebsiella spp., Escherichia coli, Enterobacter spp., Citrobacter spp., Morganella spp., Serratia marcescens (25–29), Raoultella spp. (30), Kluyvera (31), and Salmonella (32), and in non-Enterobacteriaceae such as Aeromonas (33), Pseudomonas, and Acinetobacter baumannii (34).

Attributable and crude mortality rates for infections caused by bacteria harboring KPCs are higher than those for patients with non-KPC-producing isolates (35); the reason for this increased mortality is still enigmatic. Epidemiological studies suggest that KPC-producing K. pneumoniae isolates belonging to sequence type 258 (ST258) are of two distinct clones and that the clinical behavior of isolates bearing bla_KPC-2 is different from that of isolates carrying bla_KPC-3. Molecular differences between the two clones include aminoglycoside resistance and the ability to form biofilms (36, 37). The molecular reason for this difference in clinical behavior is not yet understood. The prevalences of KPC-producing bacteria vary widely. In one surveillance study, 37% of patients in an intensive care unit (ICU) carried bla_KPC (38). Other studies place its prevalence at between 0 and 5%, depending on the population being surveyed (39, 40).

Some areas of Europe (Greece, Italy, and Poland), South America (Colombia and Argentina), the Middle East (Israel), and North America are areas where KPC is endemic. Recently, cases and localized outbreaks have been linked to importation from areas of endemicity (22, 41). In addition, long-term-care facilities (LTCFs) are rapidly becoming reservoirs for KPC producers (41). Other class A carbapenemases are important in some specific locales, such as GES-5 in Brazil, where it constitutes the main carbapenemase in Enterobacteriaceae (22). SME carbapenemases, also belonging to class A and associated with Serratia marcescens, are quite rare.

Class D carbapenemases.

Another important carbapenemase in Enterobacteriaceae is a class D β-lactamase, OXA-48. This β-lactamase, sometimes referred to as the “phantom menace,” was initially identified in a Turkish patient in 2001 (42–44). For the next 5 years, OXA-48 was not isolated from any other country. In 2008, OXA-48 spread outside Turkey and became prevalent in clinical isolates from Continental Europe, the Middle East, and northern Africa (45, 46). Since then, outbreaks throughout Europe have been reported (45, 47). Most recently, OXA-48 was detected in the United States, Canada, and South Africa (20, 48–50). Many of these reports involve patients previously treated in Middle Eastern and North African countries (51). Nonetheless, an early outbreak of OXA-48-producing K. pneumoniae in England was not linked to known regions of endemicity (52). More concerning, however, was a retrospective analysis that uncovered an outbreak of OXA-48-producing Enterobacteriaceae in a Dutch hospital that had been ongoing for 2 years (53). OXA-48 has disseminated to a wide variety of Enterobacteriaceae species, including Klebsiella spp., E. coli, Citrobacter spp., Serratia marcescens (54–56), Enterobacter spp., Morganella morganii (55), Providencia stuartii (57), Raoultella planticola (56), and Salmonella enterica (51).

OXA-48 is contained in a 61.8-kb self-conjugating IncL plasmid, which likely contributes to its ability to spread in Enterobacteriaceae (22, 58, 59). Other OXA-48-like enzymes with carbapenemase activity in Enterobacteriaceae that either have caused or have the potential to cause outbreaks include OXA-181, OXA-204, OXA-232, and OXA-162 (22, 43, 60–62). Other class D carbapenemases of clinical importance are OXA-23 and OXA-24/40; these carbapenemases are found mainly in Acinetobacter baumannii (63). Recently, some OXA-type carbapenemases have been reclassified based on their hydrolytic activity. To illustrate, once thought to be a carbapenemase, the kinetic profile of OXA-163 resembles more an ESBL than a carbapenemase (62).

Class B carbapenemases.

The class B metallo-β-lactamases (MBLs) hydrolyze a broad range of β-lactams, including carbapenems (18). The most widespread MBLs include the NDM, VIM, and IMP family enzymes. Of the MBLs, NDM-1 has emerged as a major cause of concern due to its widespread dissemination (64). NDM-1 was initially detected in a patient of Indian origin in Sweden in 2007 (65). NDM-1 was subsequently found to be widespread in the Indian subcontinent, including in environmental samples (66), and has now been reported in more than 15 countries (67). In the United Kingdom, 52% of 101 patients with NDM-producing isolates collected from 2008 to 2013 reported health care exposure or travel to the Indian subcontinent (68). NDM has spread between different bacterial species, including Enterobacter cloacae, K. pneumoniae, and Escherichia coli (69).

Horizontal spread of NDM has also been described in the clinical setting; in a recent study, four neonates from India acquired an NDM-1-producing E. coli isolate from the environment and developed sepsis (70). Clonal spread of NDM-1-producing isolates has been documented in some regions of India, while spread elsewhere, including to the United Kingdom, likely happened due to a transfer of plasmids (71).

Equally important, IMP- and VIM-producing bacteria have also been found in the United States, Europe (mostly Greece, Italy, and southern France), the Middle East, the Indian subcontinent, Japan, and China (22, 72–74). Outbreaks have occurred throughout the world, as these MBLs spread as part of complicated integrons (42). To illustrate, a recent surveillance study performed in northeastern Ohio uncovered a clinical isolate of Pseudomonas aeruginosa with bla_VIM-2 in a class I integron that was proximal to a Salmonella genomic island (SGI), suggesting recombination between these two bacteria. A detailed analysis of this genetic locus showed multiple resistance and transposing elements that likely resulted in the successful dissemination of this isolate (75).

Culture-based methods.

Culture-based testing is easier to implement, as the necessary equipment and knowledge are already present in the routine microbiology laboratory. These tests also have the potential to detect reduced susceptibility to carbapenems caused by newly emerging mechanisms as long as the mechanism is able to achieve at least a moderate level of resistance.

The CDC screening method addresses, with significant limitations, the need for the detection of low-level resistance (MIC < 2 mg/liter) and the ability to detect low loads of resistant bacteria. This method consists of an enrichment phase where a rectal swab is inoculated into 5 ml of Trypticase soy broth (TSB) in which a disk impregnated with 10 μg of ertapenem or meropenem has been immersed and incubated for 24 h. This broth is then subcultured onto MacConkey agar, where only lactose fermenters are selected. The CDC notes that many laboratories add a meropenem or ertapenem disk to this agar. A limitation of this test is that further testing is needed to determine the species and antimicrobial susceptibility of isolates growing on the agar (138). Furthermore, bacteria other than lactose fermenters that can harbor carbapenemases are routinely missed. Given the increased length of time needed for detection when using methods such as the CDC method, selective agars (see below) have been developed to optimize detection while obtaining results in a shorter time span. More important is that the CDC method will fail to detect the presence of bacteria with low-level resistance unless these bacteria are present in a large inoculum and without competition with other CROs, conditions that are unlikely to happen. Furthermore, small inocula of fully resistant CPOs can be missed if there is a large inoculum of bacteria that have low-level carbapenem resistance through mechanisms other than carbapenemases. Although the CDC broth enrichment method was meant to increase sensitivity, recent reports demonstrate that some of the selective agar methods have a performance that is superior or at least comparable to that of the CDC method, so the delay of an overnight enrichment is not necessary (139–141).

Specialized solid media aim to simplify the detection of CPE. Chromogenic media incorporate chromogenic enzyme substrates (mainly glycosides) that release a pigment when hydrolyzed by bacterial enzymes (142). Antibiotics added to the media make them selective for a particular resistance trait. Chromogenic media have been compared regarding their limit of detection of CPE with different inocula when used for stool screening (80, 139, 143–146). At this time, the currently available media have not been cleared by the U.S. Food and Drug Administration (FDA).

Available chromogenic media that may be used for the detection of carbapenemases include CHROMagar KPC (CHROMagar, France), HardyChrom (Hardy Diagnostics, CA, USA), chromID Carba (bioMérieux, France) (Fig. 1), chromID ESBL (bioMérieux, France), chromID OXA-48 (bioMérieux, France), Colorex KPC (Biomed Diagnostics, OR, USA), RambaChrom KPC (Gibson Bioscience, USA), SpectraCRE (Thermo Diagnostics, USA), and Brilliance CRE (Thermo Diagnostics, USA). Colorex KPC medium consists of medium commercially prepared from dry CHROMagar reagents.

Some of these media are designed to target KPC producers and have markedly decreased sensitivity for mechanisms based on other enzymes, particularly OXA-48 (143, 147). This is specifically addressed with a medium designed for the detection of OXA-48 producers, chromID OXA-48.

Table 4 shows the performance characteristics of different chromogenic media when tested with pure cultures. The specificity varies depending on the type of negative controls used (clinical specimens or known non-carbapenemase-producing but carbapenem-resistant isolates). In addition, although not shown in Table 4, all testing methods had slightly but consistently lower sensitivities for VIM β-lactamases than for other class B β-lactamases (80, 139, 143–146). This may be due to the inclusion of isolates that contained plasmids carrying at most another β-lactamase, rather than isolates with more complex backgrounds that have now become prevalent (148). However, this likely does not hold true for bacteria harboring VIM-containing plasmids that also carry an ESBL or another carbapenemase.

Supercarba agar (Fig. 1) is another specialized medium that incorporates the use of ertapenem (0.5 mg/liter) in addition to cloxacillin in a zinc-supplemented Drigalski lactose agar (149). Ertapenem will select for carbapenem resistance, and cloxacillin is added to inhibit the growth of AmpC producers such as Serratia and Enterobacter species, while zinc enhances the activity of MBLs (149). Different studies have shown a sensitivity of ∼96% and a specificity of 60%. These numbers are similar to those obtained with chromogenic media (143, 149). Those authors recommend the selection of only lactose-fermenting bacteria, limiting the ability of this method to detect carbapenemases in bacteria other than lactose fermenters. Another disadvantage is that the shelf life of the medium is limited to 7 days, a significant obstacle in any routine clinical laboratory (143).

Relative performance of culture methods.

Studies describing different methods for screening of CPE are difficult to compare, and each study has its own limitations and particular variations. Some studies have addressed the detection limit of different commercial assays by using previously characterized CPE isolates. These isolates, however, may not be representative of the population at a specific hospital, and clinical performance in actual practice may vary due to the prevalence of different β-lactamases in different institutions. Meanwhile, other studies have compared the performances of tests in a particular setting, such as hospitals, where there is a particular distribution of resistance mechanisms within the bacterial population. It is difficult to extrapolate the performance of these tests to other clinical settings. In addition, some studies use comparators that are known to perform poorly, which may exaggerate the performance of certain media.

Performance characteristics of the different media used for screening of rectal or perirectal swabs are shown in Table 6. Of the 13 studies mentioned, 9 showed an almost exclusive presence of KPC producers (80, 139, 141, 150–156), while 2 revealed the exclusive presence of NDM producers (157, 158). Only two studies were done at institutions where KPC and VIM producers were reported to coexist (140, 159), and only one study was performed with OXA-48-producing isolates (160). Furthermore, the prevalence of ESBL producers at these locations is not taken into account and could impact the specificity of these screening methods. The different screening systems performed variably on stool specimens compared with pure cultures (Table 4), showing mostly declines of both sensitivity and specificity with stool specimens.

We assert here that the sensitivity of a screening medium corresponds to the sum of sensitivities for each particular mechanism (e.g., OXA-48, KPN, and NDM). If a particular medium is tested in cases where one mechanism is overrepresented, it will have a greater contribution to the calculated sensitivity for the detection of CPE. For instance, consider that medium A has sensitivities of 90% for KPC and 70% for OXA-48. If this medium is tested where 95% of CPE are KPC producers while 5% are OXA-48 producers, the study will show an overall sensitivity for CPE detection of 89%. However, if 70% of CPE are OXA-48 producers and 30% are KPC producers, it will show an overall sensitivity of 76%.

To place our analysis in a clinical perspective, we performed a statistical analysis comparing the sensitivities, specificities, and diagnostic odds ratios (DORs) of the different methods used for screening of pure cultures by employing a bivariate random-effects model (161) using the mada package of the R programming language (162, 163). The DOR is the ratio of the odds of the test producing a true-positive result to the odds of it producing a false-positive result. The bivariate random-effects model is a meta-analysis technique for pooling diagnostic performance measures across studies and estimating covariate effects. The corresponding forest plots were generated with ggplot2 (164). Data from methods that did not detect all three carbapenemase classes were excluded. Figures 2 and 3 illustrate the sensitivities, specificities, and DORs for the different media in each study and in aggregate, respectively. Table 5 shows the model-estimated 95% confidence intervals for these parameters. Given the proportion of class B carbapenemase-producing isolates included in these studies, their effect on the estimated pooled performance characteristics is likely disproportionate. The same approach was used to analyze performance on rectal/perirectal swabs (Table 6). Given the low number of specimens for these analyses, we included only data from those studies where KPC was the predominant enzyme (>98%) in the analysis. We excluded methodologies that were not available commercially, except for the CDC protocol. Model-estimated sensitivities, specificities, and DORs with their corresponding 95% confidence intervals are shown in Table 7. Forest plots for the individual and aggregate studies are shown in Fig. 4 and 5.

Analysis of the results of the screening media in pure cultures shows that chromID ESBL, chromID Carba, and Supercarba have similar sensitivities. The results with Brilliance CRE medium and the CDC method with ertapenem overlap the 95% confidence intervals of chromID Carba and chromID ESBL media. Large confidence intervals can be seen with the CDC method, Colorex KPC, and CHROMagar KPC, reflecting the low number of tested isolates and conflicting results. For instance, CHROMagar KPC performed well in some studies (151, 158, 159) but not all studies (149, 155, 156), with sensitivities ranging from 40 to 98%.

Analysis of specificity is more homogeneous among the different methods. There is, however, a tendency for superiority favoring chromID Carba, while the opposite holds for chromID ESBL. This is expected, as the growth of ESBL-producing Enterobacteriaceae is considered a false-positive result when screening for CPE. Supercarba had a wide range of specificities, ranging from 35% to 82% depending on the details of the analysis, which is reflected in its large confidence interval. In this analysis, chromID Carba and Supercarba have a clear advantage in the clinic compared to the other methods. Given the large confidence intervals, these results must be interpreted with caution. Not included in the above-described analysis are data from a study by Hu et al., as it involved only 18 isolates of KPC-producing Enterobacteriaceae (105).

Analyses of medium performance on rectal/perirectal swabs are limited to those studies where KPC was the prevailing enzyme. Unfortunately, there are not enough data for a meaningful comparison of these media under different conditions. Available data with note of the enzyme distribution can be found in Table 6. The sensitivities for detection of KPC in rectal/perirectal swabs show overlapping confidence intervals for all methods, except for the CDC protocol, which is clearly inferior. Specificities also show significant overlap. HardyChrom agar showed the worst specificity, although it had a very large confidence interval, a product of being tested in only one study (141). MacConkey agar with imipenem also performed acceptably in some studies (80, 111, 139, 150), showing sensitivities and specificities as high as 92% and 100%, respectively. Analysis of DORs shows homogeneity for most methods. The overall trend is for the CDC method to be inferior to the others. Improved performance is suggested for SpectraCRE, HardyChrom, and chromID Carba. The confidence intervals for HardyChrom and SpectraCRE, however, are exceedingly large. SpectraCRE was tested in a single study in a Chicago LTCF (156), which likely explains its large confidence interval.

Due to its limited scope, chromID OXA-48 agar was not included in this statistical analysis. In a study by Zarakolu et al., it showed 75% sensitivity when tested against clinical specimens containing OXA-48, with 99.3% specificity. When used in conjunction with chromID Carba, the sensitivity and specificity reached 90.9% and 98.5%, respectively (165).

The overall differences in sensitivity between the media can be explained by the carbapenemase being tested. Most media perform reasonably well with class A enzymes, while performance with class B and D enzymes is more variable. chromID Carba medium performed well both in pure culture and when tested on rectal/perirectal swabs. The Supercarba medium performed well on pure cultures. However, it was not tested with patient specimens. SpectraCRE performed well on rectal/perirectal swabs, although one must be aware of its confidence interval. The CDC method underperformed when tested on pure cultures and clinical specimens. Other methods that were tested, particularly those involving “house-grown” techniques, could not be analyzed with the same rigor, and unless more studies are done, we caution against their use in clinical practice.

It must be emphasized that many of the studies of these selective plates are limited to KPC-producing isolates. Furthermore, the various media evaluated in Tables 4, 6, and 8 are not available in all countries. Therefore, the practical issues of cost and availability affect the choice made by an individual laboratory, which must decide if optimal sensitivity is desired, knowing that additional workup will be required to detect false-positive results if the method has a low specificity.

Studies analyzing the LOD include bacteria with specific genetic backgrounds on pure cultures that may not necessarily represent the backgrounds present in a specific clinical setting (Table 8). The LOD will directly impact the sensitivity of the screening method. Given the abundance of Enterobacteriaceae in stool, it is desirable to inhibit the growth of the carbapenem-susceptible population. However, this inhibition comes at the expense of sensitivity. A relatively low inoculum of a CPE isolate with borderline susceptibility will need to overcome this inhibitor, and the medium would have a higher LOD. On the other hand, adjusting growth inhibitors to obtain a lower LOD would allow for the growth of other bacteria and would decrease specificity.

High-resource settings where health care is already expensive may have a lesser impact on the isolation of more patients and may want to err on the side of higher sensitivities. Furthermore, the medical care provided in high-resource settings tends to be more invasive; therefore, there is a higher cost of missing a colonized patient. On the other hand, lower-resource settings may still benefit from the selection of a method with lower sensitivity that would decrease isolation costs while still having an impact on the local spread of CPOs.

Table 8 summarizes the limits of detection of the different agar screening media. All of the tested media and Supercarba agar performed reasonably well for the detection of class A enzymes (KPC), achieving a LOD in the range of 1 × 10¹ to 1 × 10² CFU/ml (80, 140, 143, 149, 166), except for chromID OXA-48, which, as expected, performs better with class D enzymes (OXA-48) (167). The LOD for class B enzymes on chromogenic media are ∼1 log higher than those for class A enzymes (80, 140, 143, 149). MacConkey agar with disks had a LOD ∼1 to 2 logs higher than those of the other media for both class A and B enzymes (80, 140). chromID OXA-48 showed poor performance for both class A and B enzymes, with a LOD of 1 × 10⁷ CFU/ml (167). Class D enzymes were not tested by all methods. The LOD for Supercarba remained consistently close to 1 × 10¹ CFU/ml. Other methods showed a significant increase in their LOD for class D enzymes. Remarkably, chromID KPC, CHROMagar KPC, the CDC method, and chromID Carba had a LOD up to 6 logs higher than those of more sensitive methods (140, 143, 149). As expected, chromID OXA-48 performed exceptionally well with class D enzymes, with a LOD of 5 × 10¹ CFU/ml (167).

Nucleic acid amplification technology.

Nucleic acid amplification technology (NAAT) detects the presence of a specific gene or genes, in most cases limiting its usefulness to previously characterized determinants. Furthermore, newly emergent variants of previously characterized genes may not be reliably detected. Since various genes can encode different carbapenemases, a broad panel of tests is needed to detect all targets. Because it is not practical to detect every enzyme, these tests have been designed to cover the most common carbapenemases. A challenge for nucleic acid-based testing is DNA extraction from stool. Feces contain PCR-inhibiting substances, and poor results may be obtained due to excessive shearing of DNA (168). Despite these concerns, very good methods are available for extracting DNA from stool, and multiplex molecular assays are routinely performed on stool specimens for gastrointestinal pathogens. It is critical to note that detection of resistant determinants in pure cultures or in specimens where a single organism is expected (such as blood or urine) is significantly easier than detection of the same genes in a more complex specimen such as a stool swab. In addition, epidemiological data such as species information are lost in most assays.

There are several NAAT-based methodologies that may be employed to detect carbapenemase genes in bacterial isolates (169). Theoretically, all of them can be used to screen stool specimens. These methodologies include single- and multiplex endpoint PCRs, loop-mediated isothermal amplification (LAMP), single- and multiplex quantitative real-time PCR (qPCR), and microarrays. Next-generation sequencing (NGS) may be another option, although it is not readily available in most clinical laboratories at this time (169). NGS remains expensive due to high equipment acquisition costs and the need for significant computer processing power and data storage (170).

Regardless of the NAAT-based methodology selected by a laboratory, there are complex regulatory requirements that vary from region to region. Implementation of a laboratory-developed assay involves the determination of the test's performance characteristics. The burden of an involved development-and-validation process may be partially relieved by the use of commercial assays. U.S. Food and Drug Administration (FDA) regulations in the United States are evolving at this time and will likely result in an increased regulatory burden on laboratories in the future.

Endpoint PCR is useful when there is a large quantity of the target gene. Specificity cannot be ensured unless positive results are confirmed by DNA sequencing or hybridization with specific probes. With proper validation, a PCR method can be acceptable. The Hyplex Super Bug ID system (Amplex Biosystems GmbH, Giessen, Germany) for the detection of carbapenemases is based on a multiplex endpoint PCR followed by enzyme-linked immunosorbent assay (ELISA) hybridization (171). Although it has not been tested directly on stool specimens, this NAAT showed 98% sensitivity and specificity for VIM-producing CPE when used on DNA extracted from clinical specimens, including blood, urine, pus, and respiratory samples, from Greece (172). Another multiplex endpoint PCR was developed by Voets et al. and allows the detection of a wide range of resistance genes (173). Some of these multiplex assays were developed by independent laboratories and are not widely available to most clinical laboratories. However, there is great value in demonstrating that these comprehensive assays can be developed.

Microarrays consist of oligonucleotides bound to a solid surface. The target gene of the pathogen is then labeled and hybridized to the immobilized probe. This reaction is then measured with a scanner (169). Microarrays are difficult to standardize (169), and reports describing the use of microarrays to directly screen for β-lactamase genes in stool are not available. Most assays, however, can be used to confirm and characterize the β-lactamase gene from suspicious colonies of a screening culture. These tests have excellent sensitivity and specificity, as shown by a study with 149 previously characterized Enterobacteriaceae that were subjected to a commercial Check-Points microarray assay (Check-Points Health, Wageningen, Netherlands), which was found to have 100% sensitivity and specificity (174). Direct testing of blood cultures also showed 98% concordance between a microarray method and routine microbiological testing (175). The Verigene BC-GN test is a microarray-like detection system. It detects nine genus/species targets and six resistance determinants, including KPC, NDM, OXA, VIM, and IMP, without the need for prior PCR amplification (176). Future studies are needed to determine if microarrays will be used to screen perirectal or stool specimens directly, although this may be hampered by their high cost and the advent of next-generation sequencing.

Loop-mediated isothermal amplification (LAMP) is a modification of conventional PCR where several oligonucleotides that bind to the target gene are incubated at the same temperature with the DNA polymerase. As DNA polymerizes, there is a release of pyrophosphate that can be detected with a fluorescent dye or a compound that will increase the turbidity of the solution (177). This method's advantages include increased sensitivity with low DNA concentrations compared to endpoint PCR, no need for a thermocycler, and simple visualization of the result. LAMP assays can be particularly useful in low-resource settings (169). A LAMP assay for the detection of NDM-1 was successfully used on 336 clinical specimens, including rectal swabs (178). Those investigators found a limit of detection of 10.70 pg/μl of genomic DNA, which would correspond to roughly 1 × 10³ CFU, compared to 1,070 pg/μl (or 1 × 10⁵ CFU) for the endpoint PCR assay used as a comparator in that study. Solanki et al. developed two LAMP assays for the detection of KPC and NDM-1 (179). These assays were able to detect all 48 tested isolates with either NDM or KPC, while endpoint PCR detected only 44. Other studies have found an improved performance of LAMP versus endpoint PCR for microbiological targets other than CPE but not versus real-time assays (180, 181). Therefore, LAMP assays may have a useful role in detecting CPOs, but they are not the most sensitive assay for clinical microbiology laboratories that have access to other types of NAATs.

Real-time or quantitative PCR (qPCR) is based on the coupling of PCR with the detection of the amplified target. Real-time PCR has been used for the screening of CPO using both commercial and “in-house” kits, with the advantage of more rapid results, increased sensitivity, and increased specificity (152, 153, 182, 183). A recent seven-center study in the Netherlands found 100% sensitivity and specificity for a multiplex assay used for the detection of KPC, NDM, VIM, IMP, and OXA-48 in 20 selected laboratory isolates (184).

Many laboratories have experience in using qPCR for the direct screening of stool/rectal swab specimens. Examples of stool/rectal swab testing with qPCR in routine clinical practice include screening for vancomycin-resistant enterococci, group B streptococci, and Clostridium difficile (185–187). The validity of the use of qPCR, including for quantification of KPC carriage loads, was evaluated by Lerner et al. (188). They determined a detection limit of 10 plasmid copies, which we presume is close to 1 × 10¹ CFU/ml.

Another in-house qPCR for NDM-1 found a limit of detection of 1 × 10¹ to 3 × 10¹ CFU/ml of stool, compared to 2 × 10¹ to 1 × 10² CFU/ml for chromID ESBL and 2 × 10¹ to 4 × 10³ CFU/ml for CHROMagar KPC (189). An additional study found limits of detection with endpoint PCR of 1 × 10⁴ to 1 × 10⁵ CFU/ml for KPC and 1 × 10³ CFU/ml for NDM (141). Using an in-house qPCR assay, Naas et al. found a limit of detection for OXA-48 of 1 × 10¹ to 1 × 10² CFU/ml in stool for qPCR, compared to 1 × 10¹ to 1 × 10² CFU/ml for Supercarba and 2 × 10¹ to 3 × 10² CFU/ml for chromID ESBL (182). A comparison between agar screening and qPCR for KPC showed 100% sensitivity for the qPCR assay, compared to 77% for culture methods (155). Overall, the limits of detection for single-gene assays (152, 155, 182, 190) tend to be lower than those for multiplex assays (141, 191).

Commercial assays for molecular multiplex CPO detection include Check-Direct CPE, Check-MDR Real Time (Check-Points Health), Hyplex SuperBug ID (Amplex Biosystems), Eazyplex SuperBug CRE (Amplex Biosystems), and Xpert MDRO (Cepheid). Check-MDR Real Time consists of an oligonucleotide probe that binds to the target sequence (VIM, NDM, KPC, and OXA-48), to a pair of universal primers, and to a molecular beacon. Real-time PCR amplifies only the bound target sequences at the same time that the molecular beacon emits fluorescence to measure amplification. The manufacturer has established a limit of detection of <5 copies per reaction. Testing of pure cultures showed 100% sensitivity and specificity (192). The manufacturer, however, recommends the use of pure cultures, which is clearly not the method used to screen perirectal swabs directly. Check-Direct CPE is a real-time assay using probe detection chemistry. It has a limit of detection of 5 copies per reaction. Using “spiked” stool specimens, Check-Direct CPE was able to detect a bacterial inoculum of 10³ to 10⁵ CFU/ml, with less sensitivity for KPC (191). NucliSENS EasyQ KPC (bioMérieux) is another real-time assay that uses molecular probes. This assay was compared to chromID ESBL with ertapenem disks by using surveillance specimens. Although a limit of detection was not determined, the assay showed 93% sensitivity (193). The SuperBug CRE system is a multiplex LAMP system that is able to detect KPC, VIM, NDM, and OXA-48 (and some variants) in addition to the ESBLs CTX-M-1 and CTX-M-9. On pure cultures, Eazyplex SuperBug CRE was able to correctly identify all 139 Enterobacteriaceae isolates (194). However, when used against a panel of 82 Acinetobacter species isolates, it produced 5 false-positive results (195). The Xpert MDRO assay has been used to detect CPO directly from rectal and perirectal swabs. The assay was able to detect KPC, NDM, and VIM with 100% sensitivity and 99% specificity using 328 discarded perirectal, rectal, or stool samples from two U.S. hospitals and one Spanish hospital (183).

Table 9 summarizes the molecular methods that have been used on clinical specimens. Some of the reports show excellent sensitivity and specificity for molecular assays; however, other studies show a broad range for the LOD of specific carbapenemases, with numbers comparable to those for agar-based methods. Limitations include the cost and the inability to detect new or unanticipated carbapenemases. Pooling of specimens for initial screening, followed by confirmatory testing of positive samples, may be an option to contain costs, especially in low-prevalence settings, but investigations are needed to see if the loss of sensitivity is too great. The cost-effectiveness of the approach will also depend on the prevalence of CPOs at an individual institution, which will determine the number of specimens that would require follow-up testing when a positive pooled result is obtained, because all individual specimens in the pool would need to be retested. The concern about detecting new, not-yet-described, carbapenemases will need to be addressed by constant vigilance in updating targets in a chosen assay; if a laboratory reports a new carbapenemase in the local geographic region, or when a medical center treats a large volume of international patients, adjustments will need to be made.