Evaluation of two commercial carbapenemase gene assays, the Rapidec Carba NP test and the in-house Rapid Carba NP test, on bacterial cultures: Table 1.

Sir, Carbapenemase-producing organisms (CPO) are recognized as one of the biggest problems in infectious diseases today: how do you treat a patient infected with an organism that might no longer be eliminated by antibiotics? Not only do you have to produce susceptibility results, but knowing the mechanism behind the resistance is also crucial, e.g. many of the new antibiotic/inhibitor combinations are effective only against class A enzymes.1 With a long list of enzymes and a broad spectrum of susceptibility phenotypes, this is a big challenge for clinical laboratories. We need methods that are simple to set up and perform, have a low rate of false negatives and are reasonably specific. We compared two commercial molecular assays [Check-Direct CPE (Check-Points Health, Wageningen, the Netherlands) and Xpert Carba-R (Cepheid, Maurens-Scopont, France)] and two phenotypic tests [the in-house Rapid Carba NP test2 and the commercially available version, i.e. Rapidec Carba NP (bioMérieux, Marcy-l'Étoile, France)] using a panel of bacteria with known carbapenemase status, selected among strains sent to our reference laboratory during 2008 to March 2014 from all Finnish clinical laboratories. Of the 94 isolates tested, 57 were CPO (Table 1): 45 Enterobacteriaceae (11 KPC, 15 NDM, 9 OXA-48, 5 OXA-181 and 5 VIM), 9 Pseudomonas aeruginosa (6 VIM and 3 IMP) and 3 Acinetobacter spp. (2 NDM and 1 VIM). They had been confirmed with our in-house PCR (testing for KPC, NDM, OXA-48-like, VIM, IMP, GES and SPM) and UV-spectrophotometric imipenem hydrolysis assay.3 Of the non-CPO, all had been referred to us because of non-WT susceptibility to carbapenems (screening cut-off for Enterobacteriaceae, zone diameters: meropenem <22 mm, imipenem <21 mm, ertapenem <25 mm; and screening cut-off for Pseudomonas, zone diameters: meropenem <24 mm and imipenem <20 mm); 11 Enterobacteriaceae isolates had ambiguous (borderline) results in the hydrolysis assay, but no carbapenemase genes were found with the in-house PCR; the borderline result was assumed to be caused by increased chromosomal AmpC enzyme activity.4 Whole-genome sequences of two similar strains have since been studied and only chromosomal genes found. Nineteen of the non-CPO Enterobacteriaceae had ESBL or transferable ampC genes (tested for CTX-M, SHV, TEM, CMY, DHA, ACC, FOX, MOX and EBC),3 seven P. aeruginosa and one Klebsiella pneumoniae strains had reduced susceptibility to carbapenems but no carbapenemase activity or transferable β-lactamase genes. Performance of the carbapenemase assays NA, not applicable. aThe sixth strain was first positive for NDM, but when repeated it was negative. The gene was determined to be VIM-5 by sequencing. bFive out of five strains; one strain could not be revived from −70°C. cPositive upon retesting; the gene was VIM-2. dThe assay was not powered to detect IMP-15, but one isolate was positive. Performance of the carbapenemase assays NA, not applicable. aThe sixth strain was first positive for NDM, but when repeated it was negative. The gene was determined to be VIM-5 by sequencing. bFive out of five strains; one strain could not be revived from −70°C. cPositive upon retesting; the gene was VIM-2. dThe assay was not powered to detect IMP-15, but one isolate was positive. Check-Direct CPE was performed as outlined in the manufacturer's manual and run on a LightCycler 480 II (Roche, Basel, Switzerland). Xpert Carba-R was performed using the manufacturer's standard protocol and run on a GeneXpert IV. False negative results were retested once. In-house Carba NP was conducted using the Rapid protocol,2 as taught to the participants of the Capacity-Building Workshop of the ECDC EuSCAPE project.5 The Rapidec Carba NP test was done according to the kit instructions, without modifications. For both phenotypic tests, strains were grown on Mueller–Hinton-II agar overnight and results were read after 2 h. The performers and interpreters of all the assays were blinded to the properties of the isolates and each Carba NP assay was independently interpreted by two technicians. No false positives were seen with the first three assays (Table 1), but the Rapidec Carba NP test surprisingly produced 14 false positives. On the other hand, there were only 4 false negatives among the OXA-48-like positive isolates, compared with 12 with the in-house protocol. Others have also experienced difficulties with OXA-48 and seen false positives with the Rapidec Carba NP test.6,7 The reason we observed this many false positives is probably due to our negative panel, which was not random, but hand-picked to be extra challenging (most had some kind of β-lactamase activity, not defined as carbapenemase). The species and mechanisms of the false positives were more or less a cross-section of the negative panel: eight ESBL positive (1 ESBL-SHV Escherichia coli, 7 CTX-M: 2 K. pneumoniae, 5 E. coli), one CMY-positive Proteus mirabilis, two Enterobacter cloacae and one Klebsiella oxytoca with borderline results in the hydrolysis assay, and the K. pneumoniae without transferable genes and negative hydrolysis (ertapenem disc 23 mm). Interpretation was also more difficult with the commercial kit, but interpreting orange-red wells as negative, to get fewer false positives, would have given more false negatives. pH might be the fundamental problem with OXA-48-like enzymes: Studentova et al.8 showed that bicarbonate added to test media significantly enhanced enzyme activity, i.e. a drop in pH, such as happens in a positive Carba NP test, can act to inhibit OXA-48. The two technicians had the same Carba NP results for all but one OXA-181-carrying isolate; this was interpreted in favour of the positive result. In-house Carba NP surprisingly did not detect one KPC isolate and we found no explanation for this: the strain was a non-mucous ST258 K. pneumoniae of Greek origin that was fully resistant (meropenem MIC, 8 mg/L; imipenem MIC, 16 mg/L). It was also weakly positive with the Rapidec Carba NP test. Hands-on time with Xpert Carba-R was much less than with the other methods and it allowed testing directly from a transport swab and also produced the fastest results (<1 h). Check-Direct CPE required isolated DNA and took ∼2.5 h. The Rapidec Carba NP test took up to 2 h and 40 min from colony to result; the in-house version took slightly longer, depending on how far in advance the reagents had been prepared. Xpert Carba-R covers only the IMP-1 subgroup and should not detect IMP-15; however, it did detect one of our three isolates. The version we tested did not cover OXA-181 and none was detected; an updated version including this enzyme was recently released. There were some problems with VIM genes, with seemingly no relation to which version was present (Table 1); VIM-2 and VIM-5 belong to different subgroups.9 Otherwise, Xpert Carba-R detected everything it was specified to detect, as did Check-Direct CPE (see the manufacturers' specifications for full lists). In our opinion, a clinical laboratory could choose any of the three methods: the choice will be based on existing infrastructure. Laboratories already using the GeneXpert module for other tests could easily buy Carba-R cartridges, while those having access to a real-time PCR machine could choose Check-Direct CPE. Those unable to afford either could set up the Carba NP or any of the commercial variants using the same principle. The same caveat applies to all: suspected false negatives, and also positives, should be confirmed by a reference laboratory; we found, as did Findlay et al.,10 that our in-house PCR was the most sensitive method of all. In combination with a hydrolysis assay and WGS, there is the capacity to find also new or rare enzymes. However, for routine, fast diagnosis, on-site preliminary testing for carbapenemases would be an important addition, since it reduces pressure on isolation facilities and—more importantly—probably also lowers the threshold for testing. This study was supported by internal funding. None to declare. A speaker's fee at a GeneXpert symposium was declined (M. Ö.). We thank Toni Huovinen, Heli Laaksonen, Minna Lamppu, Tuula Rantasalo and Mari Virta for performing the assays. We also wish to thank Triolab (representing Check-Points BV) and Immuno Diagnostic Oy (representing Cepheid) for providing proprietary machinery, support and discounted kits.