Comparison of Polymyxin B, Tigecycline, Cefepime, and Meropenem MICs for KPC-Producing Klebsiella pneumoniae by Broth Microdilution, Vitek 2, and Etest

Abstract

We report MIC agreement and error rates between broth microdilution (BMD), Vitek 2, and Etest against 48 clinical KPC-producing Klebsiella pneumoniae isolates for polymyxin B, tigecycline, cefepime, and meropenem. Both commercial testing methods were useful for tigecycline testing; Etest provided a conservative estimate of polymyxin B susceptibility. We suggest that laboratories consider the supplemental use of reference BMD or Etest for cefepime and meropenem for susceptibility testing of KPC-producing K. pneumoniae, as Vitek 2 did not provide reliable results.

INTRODUCTION

Carbapenems are considered the broadest-spectrum β-lactam agents and are often required for treatment of severe hospital-acquired infections caused by multidrug-resistant Gram-negative organisms. In recent years, plasmid-mediated serine-based β-lactamases capable of hydrolyzing carbapenems, as well as other β-lactams, have become increasingly prevalent worldwide and are now endemic in geographically diverse areas, including New York City (2, 9, 15). These enzymes, described as Klebsiella pneumoniae carbapenemases (KPCs), are most commonly harbored by K. pneumoniae, and the plasmids that encode KPCs frequently carry resistance genes for other antimicrobial classes, including fluoroquinolones and aminoglycosides (1, 12, 15).

Due to a lack of active antimicrobial agents, therapeutic options are limited for K. pneumoniae isolates expressing KPCs, and optimal treatment is currently unknown. Treatment regimens often consist of various combinations of antibiotics that demonstrate in vitro activity, which might include polymyxin B, tigecycline, and/or a broad-spectrum β-lactam at pharmacodynamically optimized doses (12, 14). Selection of combination regimens may be guided by the results of antimicrobial susceptibility and/or synergy testing, but there is increasing concern that the results of commercially available susceptibility testing assays for KPC-producing K. pneumoniae may indicate false susceptibility for some agents, although the frequency of this is unknown (1, 3, 19). In this study, we compared three susceptibility testing methods (broth microdilution [BMD], Vitek 2, and Etest) to determine the rates of MIC agreement for polymyxin B, tigecycline, cefepime, and meropenem among KPC-producing K. pneumoniae clinical isolates.

MATERIALS AND METHODS

Bacterial isolates.

This substudy was part of a larger multicenter study of hospital-acquired infections caused by extremely drug-resistant Gram-negative bacilli (XDR-GNB), defined as strains resistant to all but one first-line agent (8). From January 2008 to March 2010, case subjects with bacteremia, pneumonia, or urinary tract infections caused by XDR-GNB were enrolled. Infections were validated using National Healthcare Safety Network definitions (13). K. pneumoniae isolates associated with these infections were tested for bla_KPC by PCR as previously described (9).

MIC testing.

MIC testing was performed using BMD (Trek Diagnostics Systems, Inc., Cleveland, OH), Vitek 2 (bioMérieux Inc., Durham, NC), and Etest (bioMérieux Inc.), in accordance with the Clinical and Laboratory Standards Institute (CLSI) recommendations and manufacturers' instructions. Escherichia coli ATCC 25922 was used as the quality control strain for BMD and Etest evaluations. Prior to the performance of each test, all study strains were subcultured from an individual −80°C freezer stock onto tryptic soy agar with 5% sheep blood (BBL-Becton Dickinson, Sparks, MD). BMD was performed using commercially prepared microtiter plates which were stored at −80°C prior to use and read manually to determine MICs. The MIC ranges tested by BMD were 0.25 μg/ml to 1,024 μg/ml for polymyxin B, 0.5 μg/ml to 32 μg/ml for tigecycline, 2 μg/ml to 128 μg/ml for cefepime, and 0.5 μg/ml to 32 μg/ml for meropenem. The AST-GN28 susceptibility card for Vitek 2 was used and included the following drugs at the indicated concentrations: tigecycline at 0.75 μg/ml to 4 μg/ml, cefepime at 2 μg/ml to 32 μg/ml, and meropenem at 0.5 μg/ml to 16 μg/ml. Etest MICs were determined on Mueller-Hinton II agar (BBL-Becton Dickinson) and included MIC test ranges of 0.064 μg/ml to 1,024 μg/ml for polymyxin B, 0.016 μg/ml to 256 μg/ml for tigecycline, 0.016 μg/ml to 256 μg/ml for cefepime, and 0.002 μg/ml to 32 μg/ml for meropenem. Etest MICs were rounded up to the nearest 2-fold BMD dilution to allow comparison of categorical and MIC essential agreement.

MIC breakpoints.

Food and Drug Administration (FDA) breakpoints were used to evaluate tigecycline susceptibility (susceptible, ≤2 μg/ml; intermediate, 4 μg/ml; resistant, ≥8 μg/ml) (18). CLSI 2010 breakpoints were used for cefepime (susceptible, ≤8 μg/ml; intermediate, 16 μg/ml; resistant, ≥32 μg/ml) and meropenem (susceptible, ≤1 μg/ml; intermediate, 2 μg/ml; resistant, ≥4 μg/ml) (7). Because CLSI documents do not currently provide interpretive criteria for polymyxin B against Enterobacteriaceae, breakpoints for Acinetobacter baumannii were utilized for analysis of polymyxin B (susceptible, ≤2 μg/ml; resistant, ≥4 μg/ml) (6, 7).

Definitions and classification of MIC results.

MICs obtained from BMD testing were considered the reference standard to which results from Vitek 2 and Etest were compared. MIC agreement between methods was defined as an MIC value within 1 2-fold dilution of the MIC determined by the reference method (5). Error classification was assessed using interpretive criteria for susceptibility (5). Very major errors (VMEs) were identified when an isolate was determined to be susceptible to a given agent by Vitek 2 or Etest but resistant by BMD. Major errors (MEs) were identified when an isolate was determined to be resistant to a given agent by Vitek 2 or Etest but susceptible by BMD. A result was deemed to be a minor error (MiE) when the MIC for a given agent was intermediate by any of the testing methods studied but was determined to be either susceptible or resistant by the other comparative method.

RESULTS

MIC agreement.

The susceptibilities of 48 bla_KPC-positive K. pneumoniae isolates, each from a unique patient, were determined for the four antimicrobial agents of interest using the three assays, as shown in Table 1. Strains were more likely to be susceptible to polymyxin B and tigecycline by BMD than by the other methods. In contrast, susceptibility to cefepime and meropenem was more likely with Vitek 2 than BMD or Etest. Vitek 2 was the only test method for which isolates were determined to be meropenem susceptible (n = 13) or tigecycline resistant (n = 5).

Table 1.

Comparison of interpretative results and MIC₅₀s and MIC₉₀s for study agents and susceptibility testing methods

Agent and assay	No. (%) of isolates			MIC (μg/ml)
	Susceptible	Intermediate	Resistant	50%	90%
Polymyxin Ba
BMD	41 (85)	NAe	7 (15)	≤0.25	8
Etest	30 (63)	NA	18 (37)	2	8
Vitek 2	NA	NA	NA	NA	NA
Tigecyclineb
BMD	43 (90)	5 (10)	0 (0)	1	2
Etest	42 (88)	6 (12)	0 (0)	2	4
Vitek 2	30 (63)	13 (27)	5 (10)	2	4
Cefepimec
BMD	0 (0)	1 (2)	47 (98)	>128	>128
Etest	3 (6)	13 (27)	32 (67)	32	256
Vitek 2	33 (69)	4 (8)	11 (23)	8	≥64
Meropenemd
BMD	0 (0)	0 (0)	48 (100)	32	>32
Etest	0 (0)	1 (2)	47 (98)	32	>32
Vitek 2	13 (27)	13 (27)	22 (46)	2	≥16

^aResults based on CLSI breakpoints for Acinetobacter baumannii (susceptible, ≤2 μg/ml; resistant, ≥4 μg/ml).

^bResults based on FDA breakpoints (susceptible, ≤2 μg/ml; intermediate, 4 μg/ml; resistant, ≥8 μg/ml).

^cResults based on 2010 CLSI breakpoints (susceptible, ≤8 μg/ml; intermediate, 16 μg/ml; resistant, ≥32 μg/ml).

^dResults based on 2010 CLSI breakpoints (susceptible, ≤1 μg/ml; intermediate, 2 μg/ml; resistant, ≥4 μg/ml).

^eNA, not applicable.

The MIC₅₀ values were inconsistent among the testing methods for polymyxin B, cefepime, and meropenem (Table 1). While testing by both BMD and Etest resulted in an MIC₅₀ within the susceptible range for polymyxin B, the BMD MIC₅₀ was several 2-fold dilutions lower than the Etest MIC₅₀ (0.25 μg/ml versus 2 μg/ml, respectively). Tigecycline MIC₅₀ values for all three testing methods were comparable. In contrast, the MIC₅₀ values for both cefepime and meropenem were within the resistant range for BMD and Etest but susceptible for Vitek 2. There was substantially less variability among MIC₉₀ results between the 3 methods for all 4 tested agents.

The rate of MIC agreement between the reference method, BMD, and the comparison methods is shown in Table 2. The highest rate of MIC agreement with BMD occurred with Etest for tigecycline (94%) and meropenem (85%). A relatively high rate of agreement occurred with tigecycline MICs determined by Vitek 2 (73%). In contrast, the lowest rate of MIC agreement with BMD occurred for polymyxin B (19%) and cefepime (21%) Etests and for cefepime (19%) and meropenem (23%) Vitek 2 tests. These findings largely reflect lower cefepime and meropenem MICs found by Vitek 2 and higher polymyxin B MICs identified by Etest than by BMD. It is noteworthy that among the 12 isolates for which Etest determined the polymyxin B MICs to be 4 μg/ml, MIC agreement with BMD was achieved for only 1 isolate (8%). The polymyxin B MICs found by Etest for the other 11 isolates were in the susceptible range, with the majority being very low (≤0.25 μg/ml). However, among the 6 isolates for which Etest determined the polymyxin B MICs to be ≥8 μg/ml, MIC agreement with BMD was achieved for 4 isolates (67%).

Table 2.

MIC agreement between selected testing methods and BMD^a

Testing method	No. (%) of isolates with MIC agreement
	Polymyxin B	Tigecycline	Cefepime	Meropenem
Etest	9 (19)	45 (94)	10 (21)	41 (85)
Vitek 2	NAb	35 (73)	9 (19)	11 (23)

^aMIC within 1 2-fold dilution of MIC obtained by BMD.

^bNA, not applicable.

Error rates.

The error rates between the comparison methods and BMD are shown in Table 3. Etest results yielded 1 VME (2%) and 11 MEs (23%) for polymyxin B. Neither comparison method had interpretation discrepancies that yielded VMEs for tigecycline, but Vitek 2 results produced 5 MEs (10%) and both comparison methods yielded similar rates of minor errors for this agent. Overall, Vitek 2 generated the highest rates of VMEs, which were found for cefepime (67%) and meropenem (27%), while Etest had lower rates of VMEs for cefepime (6%) and meropenem (0%). In contrast, neither method resulted in MEs for these 2 agents.

Table 3.

Incidence of errors for selected testing methods^a

Testing method	No. (%) of isolates with the indicated errors
	Polymyxin B			Tigecycline			Cefepime			Meropenem
	Very major	Major	Minor	Very major	Major	Minor	Very major	Major	Minor	Very major	Major	Minor
Etest	1 (2)	11 (23)	NAb	0 (0)	0 (0)	10 (21)	3 (6)	0 (0)	12 (25)	0 (0)	0 (0)	1 (2)
Vitek 2	NA	NA	NA	0 (0)	5 (10)	12 (25)	32 (67)	0 (0)	5 (10)	13 (27)	0 (0)	13 (27)

^aIncidence of very major, major, and minor errors compared to BMD results.

^bNA, not applicable.

DISCUSSION

Treatment of infections caused by KPC-producing K. pneumoniae is not standardized and is often dependent upon results of antimicrobial susceptibility testing provided by clinical microbiology laboratories. Therefore, it is imperative to elucidate the accuracy of commonly used susceptibility testing methods. To our knowledge, this study represents the largest number of clinical KPC-producing K. pneumoniae isolates tested by multiple antimicrobial susceptibility testing methods for four agents that are used clinically as part of combination regimens. The results of this study call into further question the accuracy of commercially available susceptibility testing assays for KPC-producing K. pneumoniae.

The reference method, BMD, found that 15% of isolates were resistant to polymyxin B, while 37% were resistant to this agent by Etest. Previous reports have also identified the poor concordance of MIC results obtained by Etest for polymyxins and those of BMD for Pseudomonas aeruginosa and A. baumannii (10). This phenomenon may be secondary to poor diffusion of the large polymyxin molecules in agar. Our data indicate that this may be especially relevant for KPC-producing K. pneumoniae at borderline Etest MICs of 4 μg/ml, since reference BMD showed substantially lower MICs for such isolates. This may have significant therapeutic implications in clinical practice, as polymyxin B may be classified inactive, leaving even fewer available treatment options. Confirmation of polymyxin B MICs by BMD may be considered for KPC-producing K. pneumoniae with Etest MICs of 4 μg/ml.

Tigecycline was active against the majority of isolates by BMD, which confirms previous reports of susceptibility of KPC-producing K. pneumoniae to this agent (2, 4). Likewise, as in a prior surveillance study of Enterobacteriaceae, observed tigecycline MICs measured by Etest tended to be slightly higher (by 1 doubling dilution) than those measured by BMD (17). The current study also provides previously unreported data regarding the accuracy of the commercial Vitek 2 system. Although susceptibility results from Vitek 2 compared to those from Etest resulted in a lower rate of MIC agreement with BMD and a somewhat higher rate of MEs, Vitek 2 did provide a good approximation of tigecycline activity. These results demonstrate the utility of both Etest and Vitek 2 in determining tigecycline susceptibilities for KPC-producing K. pneumoniae.

Cefepime was chosen as a test agent against this group of KPC-producing K. pneumoniae due to our frequent observance that Vitek 2 reported such strains to be susceptible. However, in this study, BMD identified high-level resistance in all but one isolate; both the MIC₅₀ and MIC₉₀ by BMD were >128 μg/ml. We speculate that our Vitek 2 results can be attributed to a low inoculum, as has previously been described for testing carbapenems against KPC-producing K. pneumoniae as well as cephalosporins against other types of broad-spectrum β-lactamases (3, 16). Updates of Vitek 2 software and antibiotic cards may improve results, but as of the writing of this paper, we were unaware whether the new software recently approved in the United States will explicitly address and resolve testing issues with cefepime. Etest provided a conservative estimate of susceptibility for these isolates and may be considered an alternative to BMD for cefepime MICs. Our results with BMD contradict those previously reported by Bratu et al., in which 70% of KPC-producing K. pneumoniae isolates (n = 96) were categorized susceptible or intermediate to cefepime by agar dilution or BMD (2). Further molecular characterization of KPCs might elucidate these seemingly contradictory results.

Meropenem was inactive against this group of KPC-producing K. pneumoniae by BMD and by Etest, while Vitek 2 classified 27% of isolates susceptible. The false susceptibility by Vitek 2 occurred despite taking into account the recent CLSI breakpoint revisions that are meant to facilitate identification of carbapenem-resistant Enterobacteriaceae by automated systems. Bulik et al. recently reported similar findings, with approximately 24% of 46 tested KPC-producing K. pneumoniae isolates reported to be falsely susceptible by Vitek 2 (3). Their data indicated that software updates to the Vitek 2 system eliminated these testing issues and did not result in any very major or major errors. Therefore, until laboratories routinely using Vitek 2 are able to update their Vitek 2 automated systems, we recommend manual identification of meropenem MICs with another method for suspected or proven KPC-producing K. pneumoniae isolates that test meropenem susceptible. Given the high-level resistance to both cefepime and meropenem conferred by the KPC gene, it is prudent for clinicians to avoid therapy with these agents until further clinical data support their use.

We acknowledge several limitations to the present study. First, we studied only 48 isolates and did not determine clonality or KPC type. Previous studies have described the prevalence of a single clone (type A) within the New York City area and other centers in the eastern United States (9, 20). A separate analysis of isolates from our institution also identified type A to be the predominant circulating clone, with a minority of isolates representing up to 6 other unrelated clones (11; Phyllis Della-Latta, personal communication). Thus, the isolates studied in the present work are likely characteristic of KPC-producing K. pneumoniae isolates in other medical centers and geographic locales. It is also very likely that our isolates primarily harbor bla_KPC-2 and bla_KPC-3. Several reports have identified KPC-2 and KPC-3 to be the most common KPC types in the New York City area during the study period (9). Similarly, unpublished data from our institution also note the prevalence of KPC-2 (72%) and KPC-3 (28%) among KPC-producing K. pneumoniae (Phyllis Della-Latta, personal communication). Second, reproducibility studies were not performed by conducting duplicate susceptibility tests, but our results confirm previous reports. Finally, updated Vitek 2 software was not available at the time that this work was conducted, so we were not able to assess any effects that this software change might have on susceptibility results.

In conclusion, as multidrug-resistant Gram-negative pathogens are increasingly the cause of health care-acquired infections, it is imperative that reliable antimicrobial susceptibility testing for these isolates be available and performed. As recent CLSI breakpoints have been lowered for select β-lactam agents, clinicians may become even more dependent on 2nd-line agents such as polymyxin B and tigecycline. Both commercial testing methods are useful for tigecycline testing, and Etest provided a conservative estimate of polymyxin B susceptibility. We suggest that laboratories consider supplemental use of reference BMD or Etest for cefepime and meropenem for KPC-producing K. pneumoniae susceptibility testing, as Vitek 2 did not provide reliable results for these agents. Future studies should explore the applicability of these results for KPC-producing K. pneumoniae in other settings and the accuracy of changes to the Vitek 2 system.