Evaluation of Carbapenemase Screening and Confirmation Tests with Enterobacteriaceae and Development of a Practical Diagnostic Algorithm

Abstract

Reliable identification of carbapenemase-producing members of the family Enterobacteriaceae is necessary to limit their spread. This study aimed to develop a diagnostic flow chart using phenotypic screening and confirmation tests that is suitable for implementation in different types of clinical laboratories. A total of 334 clinical Enterobacteriaceae isolates genetically characterized with respect to carbapenemase, extended-spectrum β-lactamase (ESBL), and AmpC genes were analyzed. A total of 142/334 isolates (42.2%) were suspected of carbapenemase production, i.e., intermediate or resistant to ertapenem (ETP) and/or meropenem (MEM) and/or imipenem (IPM) according to EUCAST clinical breakpoints (CBPs). A group of 193/334 isolates (57.8%) showing susceptibility to ETP, MEM, and IPM was considered the negative-control group in this study. CLSI and EUCAST carbapenem CBPs and the new EUCAST MEM screening cutoff were evaluated as screening parameters. ETP, MEM, and IPM with or without aminophenylboronic acid (APBA) or EDTA combined-disk tests (CDTs) and the Carba NP-II test were evaluated as confirmation assays. EUCAST temocillin cutoffs were evaluated for OXA-48 detection. The EUCAST MEM screening cutoff (<25 mm) showed a sensitivity of 100%. The ETP APBA CDT on Mueller-Hinton agar containing cloxacillin (MH-CLX) displayed 100% sensitivity and specificity for class A carbapenemase confirmation. ETP and MEM EDTA CDTs showed 100% sensitivity and specificity for class B carbapenemases. Temocillin zone diameters/MIC testing on MH-CLX was highly specific for OXA-48 producers. The overall sensitivity, specificity, positive predictive value, and negative predictive value of the Carba NP-II test were 78.9, 100, 100, and 98.7%, respectively. Combining the EUCAST MEM carbapenemase screening cutoff (<25 mm), ETP (or MEM), APBA, and EDTA CDTs, and temocillin disk diffusion on MH-CLX promises excellent performance for carbapenemase detection.

INTRODUCTION

In recent years, the emergence of diverse carbapenemases in members of the family Enterobacteriaceae has become a major challenge for health care systems (1). Carbapenemase-producing bacterial isolates pose a severe clinical problem, as nonsusceptibility to beta-lactams is frequently accompanied by coresistance to additional drug classes, e.g., aminoglycosides or quinolones (2, 3). As a consequence, treatment options for carbapenemase producers are alarmingly limited and often drugs displaying significant side effects need to be administered as a last resort (4).

β-Lactamases are classified according to their functional properties and molecular structure by Ambler and Bush (5, 6). Some of these enzymes also display hydrolytic activity toward carbapenems, e.g., Klebsiella pneumoniae carbapenemase (KPC, Ambler/Bush class A)-, the New Delhi metallo-β-lactamase (NDM-1)-, VIM-, and GIM-type enzymes (all Ambler/Bush class B) or OXA-48 (Ambler/Bush class D). A key characteristic used to discriminate enzymes belonging to different Ambler/Bush classes is responsiveness to specific inhibitors: class A enzymes are inhibited by clavulanic acid and aminophenylboronic acid (APBA), class B enzymes are inhibited by EDTA, and class D enzymes do not respond to any of the inhibitors used in β-lactamase diagnostics (5, 6).

KPC enzymes were first detected in the United States in 1996 and have subsequently spread worldwide (7). In Europe, KPC is endemic in Italy, Greece, Poland, and northwestern England (7). In Central Europe, France, and Spain, other carbapenemases are reported more frequently. NDM-1 is endemic in India, Bangladesh, and Pakistan. In Europe, most of the NDM-1-producing strains are being isolated in Great Britain (8). OXA-48 is endemic in Turkey and Morocco but is increasingly reported from other European countries, mostly in repatriated patients (8, 9). Scandinavian countries, the Netherlands, and other countries such as Switzerland generally report low prevalence rates for all carbapenemases. Thus, rapid and reliable detection of carbapenemases is desirable in order to limit the spread of these enzymes.

Detection of carbapenemase-producing bacteria comprises carrier screening and detection of carbapenemase production in routine antimicrobial susceptibility testing (AST). While chromogenic media are often used for carrier screening, the laboratory strategies for β-lactamase detection in routine AST consist of a screening and a confirmation step (10–14).

A variety of phenotypic and molecular commercially available and in-house laboratory tests have been described for carbapenemase detection. Molecular techniques comprise endpoint and real-time PCRs, as well as microarray techniques (15–17). Critical diameters and MICs of ertapenem (ETP), meropenem (MEM), and imipenem (IPM) and automated microdilution expert systems have been evaluated as screening methods (14, 18–20). For carbapenemase confirmation, the modified Hodge test is recommended by CLSI and various commercial and in-house combined-disk tests (CDTs) using boronic acid derivatives and EDTA/dipicolinic acid as specific inhibitors have been described (13, 19–25). In 2014, EUCAST published new guidelines for the detection of resistance mechanisms, including carbapenemases, in which a CDT is recommended for carbapenemase confirmation (14, 22, 25). Recently, Nordmann et al. described a new inhibitor-based biochemical assay for carbapenemase detection, the Carba NP test, which has been published in two versions: the Carba NP-I assay provides a positive or negative result (carbapenemase detected/not detected), whereas the Carba NP-II test has been designed to also discriminate among carbapenemase classes A, B, and D (26–29). Apart from the original publications, few studies have systematically evaluated the Carba NP-I test for Enterobacteriaceae, and both the originally published protocol and modified versions were used. Reported sensitivities varied between 72.5 and 100%, whereas specificity was generally reported to be 100% (30–33). Except for its original description, the Carba NP-II assay has been systematically evaluated for Pseudomonas aeruginosa only (30, 31, 34–36).

Several issues of carbapenemase detection remain challenging. (i) Enterobacteriaceae overexpressing AmpC β-lactamases in combination with reduced cell wall permeability compromise the specificity of APBA CDTs, as the inhibitor (APBA) affects both AmpC and carbapenemases (37–44). (ii) Detection of OXA-48 and related enzymes remains problematic, as no specific inhibitor is available. Temocillin resistance was suggested as an indicator of OXA-48 production but not for OXA-48 confirmation (14, 25, 31, 45, 46).

This study aimed to develop a modular diagnostic flow chart suitable for all types of clinical laboratories that integrates various phenotypic screening and confirmation tests for highly sensitive and specific carbapenemase detection.

MATERIALS AND METHODS

Bacterial isolates.

A total of 334 nonduplicate clinical isolates recovered in our laboratory from 2009 until 2014 were included in this study (Table 1). All of the isolates were genetically characterized for the presence of extended-spectrum β-lactamases (ESBLs; TEM-ESBL, SHV-ESBL, and CTX-M types), plasmid-encoded AmpCs, chromosomal ampC promoter/attenuator mutations leading to overexpression (Escherichia coli only), and carbapenemases (16, 47, 48). A total of 142/334 isolates (42.2%) were suspected of carbapenemase production because of nonsusceptibility to ETP and/or meropenem and/or imipenem (intermediate or resistant zone diameters according to EUCAST clinical breakpoints [CBPs]), whereas 193/334 isolates (57.8%) not suspected of carbapenemase production (susceptible to ETP, meropenem, and imipenem) served as a negative-control group.

TABLE 1.

Species identification and β-lactamase genotypes of the isolates studied in this investigation

Species or parameter and no. (%) of isolates	ESBL, AmpC, carbapenemase negative	Production of:
		AmpC	ESBL	KPC	IMI	VIM	NDM	GIM	OXA-48
E. coli
5 (1.5)		+	+
26 (7.8)		+
34 (10.2)			+
45 (13.5)	+
1 (0.3)				+
Total, 111 (33.3)
E. cloacae
59 (17.7)	NAa	+
15 (4.5)	NA	+	+
1 (0.3)	NA	+			+
2 (0.6)	NA	+				+
1 (0.3)	NA	+						+
Total, 78 (23.4)
K. pneumoniae
24 (7.2)	+
22 (6.6)			+
13 (3.9)		+
2 (0.6)		+	+
2 (0.6)			+	+
4 (1.2)				+
1 (0.3)							+
3 (0.9)			+						+
1 (0.3)									+
1 (0.3)						+
Total, 73 (21.9)
E. aerogenes
11 (3.3)	NA	+
4 (1.2)	NA	+	+
1 (0.3)	NA	+				+
Total, 16 (4.8)
K. oxytoca
6 (1.8)			+
4 (1.2)		+
6 (1.8)	+
Total, 16 (4.8)
C. freundii
1 (0.3)	NA	+				+
3 (0.9)	NA	+	+
10 (3.0)	NA	+
Total, 14 (4.2)
H. alvei
5 (1.5)	NA	+
1 (0.3)	NA	+	+
Total, 6 (1.8)
P. mirabilis
1 (0.3)		+					+
1 (0.3)		+	+
2 (0.6)			+
Total, 4 (1.2)
M. morganii
1 (0.3)	NA	+	+
2 (0.6)	NA	+
Total, 3 (0.9)
S. marcescens, 3 (0.9)	NA	+
C. koseri, 2 (0.6)	+
Salmonella spp., 2 (0.6)			+
P. rettgeri, 1 (0.3)			+				+
P. stuartii, 1 (0.3)	NA	+					+
Enterobacter sp., 1 (0.3)	NA	+
Pantoea spp., 1 (0.3)	+
Citrobacter spp., 1 (0.3)	NA	+
Serratia spp., 1 (0.3)	NA	+
Total, 334 (100)	78	178	105	7	1	5	4	1	5
Genotypes (%), 100	23.4	53.3	31.4	2.1	0.3	1.5	1.2	0.3	1.5

^aNA, not applicable for species naturally producing chromosomally encoded AmpC β-lactamases.

Genetic detection of carbapenemase, ESBL, and ampC genes.

Total DNA was extracted from bacterial colonies with the InstaGene Matrix (Bio-Rad, Reinach, Switzerland) after growth on sheep blood agar medium (bioMérieux, Marcy l'Etoile, France). Genetic detection of carbapenemase genes was done by performing a carbapenemase multiplex PCR (16). For variant analysis, OXA-48 genes were amplified with previously described primers (49). PCR amplicons were sequenced with PCR primers, and sequences were analyzed with GenBank and DNASTAR Lasergene software (DNASTAR Inc., Madison, WI). The AID ESBL line probe assay (AID Autoimmun Diagnostika GmbH, Germany) was used to detect ESBL genes (50). Bacterial isolates were genetically characterized for the presence of plasmid-mediated AmpC-type β-lactamase genes by multiplex PCR (51). Chromosomal ampC promoter mutations of E. coli isolates were analyzed as described previously (52).

Susceptibility testing.

Disk diffusion susceptibility testing was done according to EUCAST recommendations (53). Antibiotic disks and Mueller-Hinton (MH) agar were obtained from Becton Dickinson, Franklin Lakes, NJ. Cloxacillin-supplemented MH (MH-CLX) agar was obtained from Axonlab AG, Baden, Switzerland. Zone diameters were recorded with the Sirweb/Sirscan system (i2a, Montpellier, France). MICs were determined by gradient diffusion (Etest; bioMérieux, Marcy l'Etoile, France) according to the manufacturer's instructions.

CDTs for carbapenemase detection.

CDTs were performed as described elsewhere (19, 24). Sets of two disks each containing IPM (10 μg), MEM (10 μg), or ETP (10 μg) (all from Becton Dickinson) were placed onto MH agar (EDTA CDT) or both MH agar and MH-CLX agar (APBA CDT) plates inoculated with a sample of the test isolate (0.5 McFarland turbidity standard). Immediately after the disks were placed onto the agar, 10 μl of a 29.2-mg/ml (0.1 M) EDTA solution (EDTA CDT) or 10 μl of a 30-mg/ml APBA solution (APBA CDT) was added to one of the two carbapenem disks in each set. The plates were incubated at 35°C for 16 to 20 h, and zone diameters were recorded with the Sirweb/Sirscan system (i2a). Diameter differences of ≥5 mm between the APBA-free and APBA-containing discs or between the EDTA-free and EDTA-containing discs were considered indicative of class A and B carbapenemase production, respectively.

Carba NP-II test.

The Carba NP-II test was performed and interpreted as described previously (26). Reactions were read after 0, 30, 60, and 120 min of incubation. Color changes from red to yellow-orange were interpreted as follows: wells 2 and 4 positive, Ambler class A carbapenemase; wells 2 and 3 positive, Ambler class B carbapenemase; wells 2 to 4 positive, probably Ambler class D carbapenemase; no wells positive, carbapenemase negative; all wells positive, test not interpretable. The Carba NP-II test was performed by experienced personnel, and all discrepant results were additionally repeated at least three times.

Software.

All calculations were done with IBM SPSS statistics software version 20 (IBM Corporation, Armonk, NY) and the Microsoft Excel 2010 software (Microsoft Corporation, Redmond, WA).

RESULTS

Evaluation of screening parameters for carbapenemase production.

The EUCAST CBP for ETP nonsusceptibility (<25 mm) and the EUCAST-recommended carbapenemase MEM screening cutoff (<25 mm) for carbapenemase production displayed the highest sensitivity of all of the cutoffs evaluated (100%, Table 2). ETP, however, had a lower specificity (62.5%) than MEM (90.7%, Table 2). The CLSI CBP for ETP nonsusceptibility (<22 mm) and the CLSI CBP for MEM nonsusceptibility (<23 mm) displayed lower sensitivity (95.5% for both compounds, Table 2). The EUCAST CBP for IPM nonsusceptibility (<22 mm) had the lowest sensitivity (81.8%), whereas the CLSI CBP for IPM nonsusceptibility (23 mm) had a sensitivity of 90.9%.

TABLE 2.

Performance parameters of screening and confirmation assays and the proposed diagnostic flow chart (see Fig. 2)

Parameter	No. of resultsa					Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)
	TP	FP	TN	FN	Total
Screening cutoffs/CBPs
EUCAST MEM screening (<25 mm)	22	29	283	0	334	100.0	90.7	43.1	100.0
EUCAST ETP I/Rd (<25 mm)	22	117	195	0	334	100.0	62.5	15.8	100.0
EUCAST IPM I/R (<22 mm)	18	16	296	4	334	81.8	94.9	52.9	98.7
EUCAST MEM I/R (<22 mm)	20	18	294	2	334	90.9	94.2	52.6	99.3
CLSI ETP I/R (<22 mm)	21	68	244	1	334	95.5	78.2	23.6	99.6
CLSI IPM I/R (<23 mm)	20	19	293	2	334	90.9	93.9	51.3	99.3
CLSI MEM I/R (<23 mm)	21	20	292	1	334	95.5	93.6	51.2	99.7
CDTs
ETP-BA MH agar	6	10	316	2	334	75.0	96.9	37.5	99.4
IPM-BA MH agar	6	2	324	2	334	75.0	99.4	75.0	99.4
MEM-BA MH agar	7	11	315	1	334	87.5	96.6	38.9	99.7
ETP-BA MH-CLX agar	8	0	326	0	334	100.0	100.0	100.0	100.0
IPM-BA MH-CLX agar	6	0	326	2	334	75.0	100.0	100.0	99.4
MEM-BA MH-CLX agar	7	0	326	1	334	87.5	100.0	100.0	99.7
ETP-EDTA MH agar	10	0	324	0	334	100.0	100.0	100.0	100.0
IPM-EDTA MH agar	7	0	324	3	334	70.0	100.0	100.0	99.1
MEM-EDTA MH agar	10	0	324	0	334	100.0	100.0	100.0	100.0
Carba NP-II testb	15	0	312	4	331	78.9	100.0	100.0	98.7
Carba NP-II testc	15	0	312	7	334	68.2	100.0	100.0	97.8
Proposed algorithm	22	0	312	0	334	100.0	100.0	100.0	100.0

^aTP, true positive; FP, false positive; TN, true negative; FN, false negative.

^bInconclusive results were excluded from the calculation.

^cInconclusive results were considered negative.

^dI/R, intermediate or resistant.

Performance of carbapenemase confirmation tests. (i) CDTs.

The ETP APBA CDT on MH-CLX agar displayed the highest sensitivity and negative predictive value (NPV) for class A carbapenemase detection (100%, Table 2). A specificity of 100% was found for the ETP APBA CDT, the IPM APBA CDT, and the MEM APBA CDT on MH-CLX agar, whereas the same CDTs on conventional MH agar showed specificities of 96.9, 99.4, and 96.6%, respectively (Table 2). Nine of 10 false-positive ETP APBA CDTs on conventional MH agar occurred in species with chromosomal AmpC (6 Enterobacter cloacae, 1 Enterobacter aerogenes, and 2 Hafnia alvei isolates). Nine of 11 false-positive MEM APBA CDTs on conventional MH agar were also found in AmpC-positive species, i.e., 6 E. cloacae isolates, 1 E. aerogenes isolate, 1 H. alvei isolate, and 1 E. coli isolate producing CIT-type plasmid-encoded AmpC. One K. pneumoniae isolate lacking AmpC or an ESBL was borderline positive in both ETP and MEM APBA CDTs on conventional MH agar (5- and 7-mm zone differences, respectively). Another K. pneumoniae isolate producing an ESBL was borderline positive only in the MEM APBA CDT on conventional MH agar (5-mm difference).

Both the ETP and MEM EDTA CDTs displayed 100% sensitivity and specificity for class B carbapenemase detection, whereas the sensitivity of the IPM EDTA CDT was significantly lower (70%, Table 2).

(ii) Carba NP-II test.

The overall sensitivity, specificity, positive predictive value (PPV), and NPV of the Carba NP-II test were 78.9, 100, 100, and 98.7%, respectively (Table 2). The test created some reading problems, resulting in ambiguous results that were treated as follows. One E. aerogenes isolate possessing a bla_VIM gene gave ambiguous results in terms of class assignment (isolate 8 in Fig. 1). After 30 min of incubation, the pattern was consistent with a class B carbapenemase, while after 120 min of incubation, the pattern was consistent with a class D carbapenemase (e.g., OXA-48). For calculation of performance parameters, this isolate was considered carbapenemase positive (Table 3). Three K. pneumoniae isolates coproducing OXA-48 and CTX-M ESBL gave inconclusive results (Table 3). The NP-II patterns were negative for carbapenemase production until 60 min of incubation. After 120 min of incubation, the patterns could still be rated either negative or weakly positive for class A carbapenemases (Fig. 1, isolates 20, 99, and 51; the results were reproduced three times with independent preparations); these isolates were excluded from the calculation of performance parameters. In addition, one OXA-48-producing K. pneumoniae isolate (isolate 19 in Fig. 1) and three NDM-producing isolates of Providencia rettgeri, Providencia stuartii, and Proteus mirabilis gave false-negative NP-II test results (Table 3, isolates 136, 138, and 139, respectively; Fig. 1). One E. cloacae isolate producing a GIM (class B) gave an OXA-48-like pattern (class D, isolate 95 in Fig. 1). For the calculation of sensitivity and specificity, this isolate was considered carbapenemase positive (Table 3).

FIG 1.

Discrepancies between Carba NP-II test results and carbapenemase genotypes.

TABLE 3.

Carbapenemase-positive isolates with characteristics and confirmation test results

Isolate no.	Species	AmpC	ESBL	Carbapenemase		NP-II	CDT (Δdiam [mm])b
				Type	Class		BA on MH agar			BA on MH-CLX agar			EDTA on MH agar
							ETP	IMI	MEM	ETP	IMI	MEM	ETP	IMI	MEM
7	K. pneumoniae			KPC	A	+	7	5	8	8	6	10	0	0	0
29	K. pneumoniae		SHV-ESBL	KPC	A	+	6	5	5	11	7	11	0	1	0
31	K. pneumoniae			KPC	A	+	7	11	7	9	8	9	0	0	0
35	K. pneumoniae			KPC	A	+	4	5	8	7	5	6	0	0	1
37	K. pneumoniae			KPC	A	+	7	7	9	8	4	9	0	0	0
40	E. cloacae	cAmpCa		IMI	A	+	11	13	13	11	8	10	2	1	1
55	E. coli			KPC	A	+	4	2	6	7	3	6	2	0	0
137	K. pneumoniae		CTX-M	KPC	A	+	6	4	4	5	5	2	0	3	0
8	E. aerogenes	cAmpC		VIM	B	+	0	0	0	0	0	0	7	5	8
9	K. pneumoniae			NDM	B	+	0	0	0	2	0	0	16	7	13
17	E. cloacae	cAmpC		VIM	B	+	0	0	0	3	0	0	5	3	5
70	C. freundii	cAmpC		VIM	B	+	0	0	0	0	0	0	5	4	7
82	K. pneumoniae			VIM	B	+	0	0	0	0	0	1	15	17	21
95	E. cloacae	cAmpC		GIM-1	B	+	0	0	0	2	0	2	10	3	10
136	P. rettgeri	cAmpC		NDM	B	−	0	0	0	0	0	0	10	19	19
138	P. stuartii	cAmpC		NDM	B	−	0	0	0	0	0	0	9	13	12
139	P. mirabilis	CIT		NDM	B	−	0	4	0	0	0	0	6	16	6
36	E. cloacae	cAmpC	SHV-ESBL	VIM	B	+	0	0	0	3	0	1	5	6	8
20	K. pneumoniae		CTX-M	OXA-48	D	Inconclusive	0	0	0	3	0	0	0	0	0
51	K. pneumoniae		CTX-M	OXA-48	D	Inconclusive	0	0	0	2	0	3	0	0	0
99	K. pneumoniae		CTX-M	OXA-48	D	Inconclusive	0	0	0	2	0	3	0	0	0
19	K. pneumoniae			OXA-48	D	−	3	0	1	3	0	2	0	0	0

^acAmpC, chromosomally encoded ampC gene.

^bUnderlined values indicate a positive test result.

(iii) Temocillin testing on MH-CLX agar.

Nineteen representative carbapenem-nonsusceptible isolates were tested for temocillin zone diameters and MICs on MH agar and MH-CLX agar as indicators of the presence of OXA-48. Five isolates harbored bla_OXA-48 genes, nine isolates were bla_OXA-48 gene negative but showed overexpression of chromosomally encoded AmpC, and five isolates harbored ESBL-encoding genes (but not bla_OXA-48, Table 4). All OXA-48 producers showed high-level temocillin resistance on both MH agar and MH-CLX agar (median diameter, 6 mm; median MIC, >1,024 mg/liter; Table 4). Five out of nine AmpC hyperproducers displayed temocillin zone diameters of <11 mm on MH agar (EUCAST screening cutoff for OXA-48-like enzymes) (14). On MH-CLX agar, the median temocillin zone diameter of the AmpC hyperproducers increased by 7 mm (corresponding to a median Etest-determined MIC decrease of 2 dilution steps, Table 4), and the five EUCAST OXA-48 screening false-positive isolates became true negatives. The median temocillin zone diameters and gradient diffusion MICs of the five ESBL producers were not altered by the use of MH-CLX agar rather than conventional MH agar. The median temocillin zone diameter and MIC were 11 mm and 32 mg/liter, respectively, on both media (Table 4). The only false-positive temocillin-based OXA-48 screening result originated from a CTX-M type ESBL-producing K. pneumoniae isolate displaying a temocillin zone diameter of 10 mm and a MIC of 64 mg/liter on both MH agar and MH-CLX agar.

TABLE 4.

Temocillin critical zone diameters and MICs for confirmation of OXA-48-like carbapenemases

IDa or parameter	Species	ESBL	AmpC	Carbapenemase	Temocillin zone diam (mm)		Temocillin MIC (mg/liter)
					MH agar	MH-CLX agar	MH agar	MH-CLX agar
19	K. pneumoniae	−	−	OXA-48	6c	6	1,024	1,024
20	K. pneumoniae	+	−	OXA-48	6	6	1,024	1,024
51	K. pneumoniae	+	−	OXA-48	6	6	1,024	1,024
99	K. pneumoniae	+	−	OXA-48	6	6	1,024	1,024
36	E. cloacae	+	cAmpCb	VIM	6	8	1,024	128
16	H. alvei	−	cAmpC		6	11	128	32
18	E. cloacae	−	cAmpC		9	17	32	16
5	E. cloacae	−	cAmpC		10	21	32	8
27	E. cloacae	−	cAmpC		10	12	32	32
25	H. alvei	−	cAmpC		10	21	32	4
26	E. cloacae	−	cAmpC		11	18	32	8
2	E. cloacae	−	cAmpC		12	16	32	16
125	E. cloacae	−	cAmpC		14	22	16	4
1	E. aerogenes	−	cAmpC		16	20	8	8
39	K. pneumoniae	+	−		10	10	64	64
60	E. coli	+	−		11	11	32	32
38	P. mirabilis	+	−		11	11	32	32
130	K. pneumoniae	+	−		14	13	16	16
128	K. pneumoniae	+	−		18	17	8	8
Median values
OXA-48-positive isolates					6	6	1,024	1,024
AmpC overexpression					10	18	32	8
ESBL					11	11	32	32

^aID, isolate identification number.

^bcAmpC, chromosomally encoded AmpC β-lactamase.

^cUnderlined values indicate a positive test result.

Genetic characterization of isolates.

A total of 23 carbapenemase genes were detected in 22 Enterobacteriaceae isolates, i.e., 7 bla_KPC, 1 bla_IMI, 4 bla_VIM, 4 bla_NDM, 1 bla_GIM, and 4 bla_OXA-48 genes; 1 isolate coproduced VIM and OXA-48 enzymes (Tables 1 and 2). Seventy-eight (23.4%) of the isolates studied were genetically negative for ESBL, AmpC, and carbapenemases; 178 (53.3%) of the isolates produced an AmpC β-lactamase (including those species with chromosomally encoded AmpC, i.e., E. cloacae, E. aerogenes, Citrobacter freundii, H. alvei, Morganella morganii, Serratia marcescens, and P. stuartii; Table 1) (54); 105 (31.4%) of the isolates produced an ESBL (Table 1).

DISCUSSION

Screening parameters for carbapenemases.

Disk diffusion critical diameters have been reported to display high sensitivity for the detection of carbapenemases (13, 20). This study found 100% sensitivity for the EUCAST critical MEM diameter (<25 mm) with a comparably high specificity of 90.7% (Table 2). ETP screening with the EUCAST CBP for nonsusceptibility (<25 mm) also showed high sensitivity (100%) but low specificity (62.5%, Table 2). Thus, our results confirm the current EUCAST recommendation (15). The CLSI CBPs for ETP nonsusceptibility (<22 mm) and MEM nonsusceptibility (<23 mm) displayed lower sensitivity than the current EUCAST recommendation (95.5%, Table 2). On the basis of the findings of this study, carbapenemase screening with MEM is recommended, whereas the use of IMP as a screening drug is discouraged (IMP sensitivities with the EUCAST <22-mm-diameter and CLSI <23-mm-diameter CBPs, 81.8 and 90.9%, respectively). Since automated microdilution AST reportedly lacks sensitivity and specificity because of the antibiotic panel composition and drug concentrations tested (18, 55), disk diffusion critical MEM diameters promise the best performance for carbapenemase detection among all of the techniques evaluated. In addition, disk diffusion is cheap, simple, and widely implemented by many laboratories for routine AST.

Carbapenemase confirmation tests.

The modified Hodge test, which is recommended by CLSI for carbapenemase confirmation, is cheap and, in principle, simple to perform (23). However, it displays significant investigator dependence, practical interpretation is technically demanding, it cannot distinguish among the different carbapenemase classes, and it reportedly has low specificity because of AmpC β-lactamase overproduction and decreased permeability, e.g., caused by porin loss (13, 20, 55). The problem of discriminating carbapenemase activity from AmpC and impermeability is well known both for species possessing a chromosomal AmpC (e.g., Enterobacter spp., Citrobacter spp., or H. alvei) and for producers of plasmid-encoded AmpC, in particular, K. pneumoniae (39, 44, 56). Even E. coli overproducing AmpC because of mutations in the promoter/attenuator region and/or showing mutations in the active center of the enzyme, resulting in an extended-spectrum AmpC phenotype, displays carbapenem nonsusceptibility (41, 43). The same pattern accounts for ESBL producers in combination with porin loss (37, 38). AmpC and ESBL production interferes not only with carbapenemase screening but also with APBA CDT confirmation for class A carbapenemases (14, 19, 57). False-positive results occur, as APBA is an inhibitor of not only class A carbapenemases but also AmpC β-lactamases. To improve the specificity of APBA CDTs, MEM/CLX disks are used to check for AmpC interference (indirect approach) (13, 14, 20, 22, 25). However, on the basis of the current EUCAST algorithm, class A carbapenemases in isolates coproducing AmpC may be missed, as synergy of MEM with both CLX and APBA is interpreted as AmpC and porin loss (14). A recent study found two E. cloacae isolates overproducing AmpC but also producing KPC and other class A enzymes that would have been misclassified by this approach (13). Other investigators pointed out that MEM-MEM/CLX zone diameter differences are relatively lower in AmpC hyperproducers coexpressing a class A carbapenemase (i.e., mean difference, 1 mm) than in AmpC hyperproducers without a class A carbapenemase (mean difference, 5 mm) (22). Another study, however, described MEM-MEM/CLX zone diameter differences of 6 to 7 mm and 0 to 7 mm for AmpC-hyperproducing E. cloacae producing class A carbapenemases and AmpC hyperproducers devoid of carbapenemases, respectively (13). Thus, the discriminative power of relative MEM-MEM/CLX diameter differences may be insufficient. In addition, classification based on the relative degree of MEM-MEM/CLX diameter differences is difficult to standardize and requires significant expertise. The present study of 178 (53.3%) AmpC-producing isolates shows that APBA CDTs performed on MH-CLX agar reliably detect class A carbapenemases with increased specificity (100%) because of suppression of AmpC activity (Table 2). The approach is simple to interpret, as it uses a single critical zone diameter difference (5 mm) and it can be integrated in one step with ESBL confirmation testing on the same MH-CLX agar plate (48).

In the present study, the Carba NP-II test showed an overall sensitivity of 78.9% for carbapenemase detection and an NPV of 98.7% (Table 2). Our results closely parallel those of a recent study that found a sensitivity of 72.5% and an NPV of 69.2% for the Carba NP-I test. The difference in NPV is well explained by the different prevalences of carbapenemase producers in the study populations, i.e., 6.6% (n = 22) in this study and >45% (n = 145) in the study of Tijet et al. (31). Other investigators found higher Carba NP-I test sensitivities when using different types of protocols (32, 33). Our data confirm ambiguities in the reading of the Carba NP-I/II test, in particular, for OXA-48-producing isolates that tend to produce inconclusive or false-negative results (Fig. 1, isolates 19, 20, 51, and 99) (31). If the inconclusive OXA-48 results in Fig. 1 had been rated negative (only a slight color change was visible after 120 min of incubation), the sensitivity would have been 68.2% (Table 2). If it had been rated positive, the three ambiguous OXA-48 results would have been consistent with a class A carbapenemase pattern, most likely because of the simultaneous presence of a CTX-M-type ESBL (class A enzyme), which may be responsible for the weak color change in wells 2 and 4 after 120 min of incubation and which is inhibited by tazobactam in well 3 (Fig. 1). False-negative Carba NP results have also been described for mucoid colonies, e.g., of P. rettgeri, P. stuartii, or P. mirabilis isolates (29, 31). Negative results were attributed to difficulties in protein extraction, species-specific traits, or the influence of the agar type and ion content of the Carba NP test (30, 31, 36). Besides OXA-48 producers, false-negative results in the present study were also obtained with nonmucoid isolates of P. rettgeri, P. stuartii, and P. mirabilis producing NDM enzymes. All of the tests of these isolates were repeated three times by using the standard protocol and additionally performed with colonies grown on various agar media from different manufacturers, i.e., MH agar (Becton Dickinson), MH-CLX agar (Axonlab), Columbia sheep blood agar, MacConkey agar (bioMérieux), and Uriselect4 agar (Bio-Rad). Despite reports that the Carba NP I test performed better with Columbia sheep blood agar and Uriselect4 agar, the results obtained with these isolates remained false negative with all of the medium types, pointing to species-specific issues related to Providencia and Proteus isolates and low sensitivity for OXA-48 enzymes (34). Other investigators recently found higher sensitivity and specificity for the detection of OXA-48 (28). In summary, because of the higher NPV, the Carba NP-II test may perform better in a low-prevalence environment (i.e., our study) than in high-prevalence settings such as those investigated by Tijet et al. (31). However, the issues of false-negative OXA-48 producers and species-specific false-negative results because of the unknown impact of different genetic backgrounds need to be further analyzed.

The phenotypic detection of OXA-48-like carbapenemases remains challenging. EUCAST recommends indirect OXA-48 confirmation by decreased zone diameters or increased MICs for temocillin (<11 mm and >32 mg/liter, respectively) to exclude ESBLs in combination with porin loss in cases where both APBA and EDTA CDTs are negative (14). Temocillin MICs, however, are not recommended to discriminate AmpC overproduction combined with porin loss from OXA-48, as temocillin MICs are variable in this setting, resulting in poor specificity. By suppressing potential AmpC activity, temocillin disk diffusion testing or MIC determination by a gradient diffusion method on MH-CLX agar can help to clearly increase the specificity of temocillin-based OXA-48 screening without compromising sensitivity (Table 4).

In summary, a combination of the EUCAST MEM carbapenemase screening cutoff (<25 mm) and ETP (or MEM) APBA and EDTA CDTs plus temocillin disk diffusion (or gradient diffusion-based MIC determination) on MH-CLX agar promises excellent performance in carbapenemase detection. The proposed diagnostic flow chart (Fig. 2) would have resulted in 100% sensitivity, specificity, PPV, and NPV in the study population. This algorithm is simple, easy to use, cost efficient, and applicable in the majority of clinical microbiology laboratories.

FIG 2.

Proposed diagnostic flow chart for carbapenemase detection. (Superscript 1) MEM can be used alternatively with slightly lower sensitivity. (Superscript 2) A carbapenem resistance phenotype is most likely due to a combination of AmpC and/or ESBL overexpression and decreased permeability, e.g., because of porin deficiency.