Accuracy of Six Antimicrobial Susceptibility Methods for Testing Linezolid against Staphylococci and Enterococci

Abstract

A challenge panel of enterococci (n = 50) and staphylococci (n = 50), including 17 and 15 isolates that were nonsusceptible to linezolid, respectively, were tested with the Clinical and Laboratory Standards Institute broth microdilution and disk diffusion reference methods. In addition, all 100 isolates were tested in parallel by Etest (AB Biodisk, Solna, Sweden), MicroScan WalkAway (Dade, West Sacramento, CA), BD Phoenix (BD Diagnostic Systems, Sparks, MD), VITEK (bioMérieux, Durham, NC), and VITEK 2 (bioMérieux) by using the manufacturers’ protocols. Compared to the results of the broth microdilution method for detecting linezolid-nonsusceptible staphylococci and enterococci, MicroScan results showed the highest category agreement (96.0%). The overall categorical agreement levels for VITEK 2, Etest, Phoenix, disk diffusion, and VITEK were 93.0%, 90.0%, 89.6%, 88.0%, and 85.9%, respectively. The essential agreement levels (results within ±1 doubling dilution of the MIC determined by the reference method) for MicroScan, Phoenix, VITEK 2, Etest, and VITEK were 99.0%, 95.8%, 92.0%, 92.0%, and 85.9%, respectively. The very major error rates for staphylococci were the highest for VITEK (35.7%), Etest (40.0%), and disk diffusion (53.3%), although the total number of resistant isolates tested was small. The very major error rate for enterococci with VITEK was 20.0%. Three systems (MicroScan, VITEK, and VITEK 2) provided no interpretations of nonsusceptible results for staphylococci. These data, from a challenge panel of isolates, illustrate that the recent emergence of linezolid-nonsusceptible staphylococci and enterococci is providing a challenge for many susceptibility testing systems.

Linezolid is an oxazolidinone agent that inhibits bacterial protein synthesis by binding to the 50S ribosomal subunit, blocking the formation of the initiation complex (12, 26). It has potent activity against many gram-positive organisms, including methicillin-resistant Staphylococcus aureus, vancomycin-resistant enterococci (5, 9, 26), and several anaerobic species (24). Linezolid is usually bacteriostatic against staphylococci and enterococci (11, 12). Resistance, although rare, has been reported in both staphylococcal and enterococcal isolates (8, 14, 16, 23). The Clinical and Laboratory Standards Institute (CLSI) has defined susceptible (S), intermediate (I), and resistant (R) breakpoints for linezolid against enterococci for both disk diffusion (≥23 mm, S; 21 to 22 mm, I; ≤20 mm, R) and dilution MIC testing (≤2 μg/ml, S; 4 μg/ml, I; ≥8 μg/ml, R) (2, 3). However, only susceptibility breakpoints have been defined for disk diffusion (≥21 mm) and MIC testing (≤4 μg/ml) of staphylococci (because of a lack of linezolid-resistant strains when the breakpoints were developed) (2, 3). Linezolid MICs have been noted to vary among laboratories even when the same testing method is used (12). This may be due to trailing growth, which has been observed for some isolates of both staphylococci and enterococci (7, 12).

In vitro linezolid susceptibility can be determined by disk diffusion, broth microdilution, agar dilution, Etest, and most automated antimicrobial susceptibility testing systems. The Etest method (AB Biodisk, Solna, Sweden) yields MICs that are typically 1 doubling dilution lower than broth microdilution results, probably because the manufacturer recommends reading an endpoint at 90% inhibition instead of complete inhibition (7). Livermore noted that most MICs would be reduced by approximately 1 dilution if this approach of reading endpoints of less than 100% inhibition were adopted for other linezolid MIC testing methods (12). The disk diffusion method was effective in detecting the first reported isolate of S. aureus that was nonsusceptible to linezolid (23). However, because of the paucity of linezolid-resistant enterococci and nonsusceptible strains of staphylococci, there is little information available about the performance of automated methods for testing linezolid. In one study, Qi et al. noted that the linezolid resistance results reported for enterococci by the MicroScan system (with a PC-21 panel) correlated well with the presence of G2576T mutations in domain V of the 23S rRNA gene of the organisms tested. In that same study, however, the Etest and VITEK 2 methods (with an AST GP-61 card) tended to produce false resistance results when compared by using the results of rRNA sequencing as the “gold standard” (18). Problems with false linezolid resistance results for enterococci were also reported by Scheetz et al. (19). Their study of the Etest and disk diffusion methods concluded not only that false linezolid resistance results were a problem but that the results led to greater overall resource use and ultimately greater patient mortality in their hospital (19).

Through both surveillance activities and outbreak investigations, the Centers for Disease Control and Prevention (CDC) and Project ICARE have collected a series of linezolid-resistant enterococci and linezolid-nonsusceptible staphylococci, as defined by the broth microdilution reference method. The goal of this study was to evaluate the other susceptibility testing methods that are most commonly used in the United States for the ability to detect known linezolid-resistant (or -nonsusceptible) organisms and differentiate them from known linezolid-susceptible isolates.

MATERIALS AND METHODS

Bacterial strains.

Twenty-five isolates of each of the organism groups S. aureus (of which 7 were linezolid nonsusceptible); coagulase-negative staphylococci (CoNS), including 10 S. epidermidis isolates (5 nonsusceptible), 3 S. haemolyticus isolates (1 nonsusceptible), 5 S. hominis isolates (2 nonsusceptible), 4 S. lugdunensis isolates, 2 S. simulans isolates, and 1 S. warneri isolate; Enterococcus faecium (10 resistant and 2 intermediate isolates); and E. faecalis (5 resistant isolates) were obtained from the strain collections of the CDC and Project ICARE (6). Linezolid-nonsusceptible isolates were collected from 2001 to 2005. DNA sequence analysis was used to identify the presence of mutations in the central loop of domain V in the 23S rRNA subunit of strains for which two or more susceptibility testing systems disagreed with the results of the broth microdilution reference method. This method was used to resolve discrepant testing results for six nonsusceptible S. aureus, two nonsusceptible S. epidermidis, and two E. faecium isolates (see below). The quality control organisms included on each day of susceptibility testing were S. aureus ATCC 25923 and ATCC 29213 and Enterococcus faecalis ATCC 29212 (4). All quality control results were in control for all of the methods used throughout this study.

Susceptibility testing methods.

The antimicrobial susceptibility profiles of the isolates were determined by the CLSI broth microdilution reference method with cation-adjusted Mueller-Hinton broth (BD Diagnostic Systems [BDDS], Sparks, MD) (2). MIC panels were prepared in house at the CDC and contained linezolid concentrations ranging from 1.0 to 16 μg/ml. Disk diffusion testing was performed as described in CLSI document M2-A9 (3); results for linezolid were read at 18 to 20 h by using both reflected and transmitted light. For this study, only results recorded at 18 to 20 h by using reflected light were used to establish categorical and essential errors (in accordance with CLSI guidelines). For 12 isolates (10 enterococci and 2 staphylococci), results could not be interpreted at 18 to 20 h because of insufficient growth; these results were interpreted at 24 h. Etest procedures were performed by following the manufacturer's guidelines. This included using 90% inhibition of growth as the endpoint for reading the test. Test results were interpreted at 18 to 20 h, except for the 12 isolates noted above, which also showed insufficient growth by the Etest method and had to be interpreted at 24 h. Etest MICs were rounded up to the next higher log₂ dilution to correspond to broth microdilution MICs for the purposes of comparison and analysis. The following MIC cards or panels (manufacturer; software version) were also used in this study: MicroScan WalkAway (Dade, West Sacramento, CA; POS MIC 20A, LabPro 1.61 Alert 1.6), BD Phoenix (BDDS; PMIC/ID-100, V5.15A/V4.11B), VITEK (bioMérieux, Durham, NC; GPS110, R10.10), and VITEK 2 (bioMérieux; GP63, R04.02). All systems were tested according to the manufacturers’ protocols. The results from the VITEK 2 expert system were recorded, but the expert results were not used for error calculations.

All of the systems were tested in parallel with inocula prepared from the same subculture. All of the isolates for which there were discrepancies between the categorical results (i.e., susceptible, intermediate, or resistant) of the test method and the results of the broth microdilution reference method were retested in parallel by broth microdilution and the method in question; however, only the original results were used in error calculations. If two or more systems disagreed with the broth reference result, all of the systems were retested in parallel to determine the reproducibility of the results. For categorical agreement, very major errors were defined as those in which the test method result was susceptible and the reference method result was resistant, major errors were defined as those in which the test method result was resistant and the reference method result was susceptible, and minor errors are defined as those in which either method reported a result as intermediate and the other method reported the result as susceptible or resistant. Very major error rates were calculated by using the 15 nonsusceptible isolates as the denominator for staphylococci and the 15 resistant isolates as the denominator for enterococci. For major error rate calculations, the denominators were 35 for staphylococci and 33 for enterococci. For minor error rate calculations, the total number of organisms tested was used as the denominator, including two E. faecium isolates that were intermediate to linezolid. Essential agreement was calculated for MIC methods by determining the number of test results that were within ±1 doubling dilution of the MIC determined by the reference method.

DNA sequence analysis.

DNA sequence data for several of the isolates in this study, including two S. epidermidis isolates, have been reported elsewhere (25). For those isolates not previously sequenced, PCR template DNA was prepared by resuspending one colony from overnight growth on a Trypticase soy agar plate (BDDS) in 100 μl pure water and boiling it gently. The PCR primers, designed to amplify a 690-bp fragment of domain V of the 23S rRNA genes from S. aureus were 5′ TGGGCACTGTCTCAACGA (23) and 3′ ATCCCGGTCCTCTCGTACTA (16). Primers for domain V of the 23S rRNA genes of E. faecium isolates were 5′ GACGGAAAGACCCCATGG and 3′ ACACTTAGATGCTTT (17), which resulted in a 718-bp fragment. Amplification assays included GoTaq Green Master Mix (Promega), each primer at 0.2 μM, and 1 μl of template suspension in a total volume of 50 μl. The reaction mixtures were heated to 95°C for 5 min and then cycled 35 times at 95°C for 1 min, 42°C for 1 min, and 72°C for 2 min, followed by a 5-min incubation at 72°C. PCR products were visualized on agarose gels, and the remaining products were purified with the Invitrogen PureLink PCR Purification Kit. Purified PCR products were sent to the DNA Core Sequencing Center at the University of Illinois at Urbana-Champaign for sequencing. Each PCR product was sequenced from both the 5′ and 3′ ends. Results were analyzed with DNASTAR Lasergene software. Additionally, a visual inspection of the accompanying chromatograms for duplicate peaks representing a mixture of nucleotides at each position was done. No ambiguities were detected. Finally, the consensus sequences from the isolates were aligned for S. aureus and for E. faecium to check for any differences among the isolates.

RESULTS

Categorical and essential agreement.

Fifty isolates of staphylococci and 50 isolates of enterococci were tested for susceptibility to linezolid by broth microdilution, disk diffusion, Etest, MicroScan, Phoenix, VITEK, and VITEK 2. The categorical and essential agreements of the results of the six methods with those of the broth microdilution method are shown in Table 1. Results for each method are detailed below. Isolates for which very major or major errors were reported are listed in Table 2.

TABLE 1.

Error rates, categorical agreements, and essential agreements for 100 isolates of staphylococci and enterococci tested with linezolid

Method (no. reported)	Staphylococci (n = 50)a,b,c		Enterococci (n = 50)b			Overall categorical agreement (%)d	Overall essential agreement (%)d
	No. of very major errors/no. of resistant isolates (%)	No. of major errors/no. of susceptible isolates (%)	No. of very major errors/no. of resistant isolates (%)	No. of major errors/no. of susceptible isolates (%)	No. of minor errors/total no. of isolates (%)
Disk diffusion (100)	8/15 (53.3)	0	1/15 (6.7)	0	2/50 (4.0)	89.0	NAe
Etest (100)	6/15 (40.0)	0	1/15 (6.7)	0	3/50 (6.0)	90.0	92.0
MicroScan (100)	1/15 (6.7)	0	0	0	3/50 (6.0)	96.0	99.0
Phoenixb (97)	4/15 (26.7)	0	0	1/34 (2.9)	5/49 (10.2)	89.6	95.8
VITEKc (99)	5/15 (35.7)	1/34 (2.9)	3/15 (20.0)	0	5/50 (10.0)	85.9	85.9
VITEK 2 (100)	1/15 (6.7)	1/35 (2.8)	0	0	5/50 (10.0)	93.0	92.0

^aThere are no intermediate or resistant breakpoints defined by CLSI for staphylococci for linezolid; thus, there are no minor errors.

^bThe test run on the Phoenix was aborted by the system before reporting results for three isolates (one each of S. epidermidis, S. aureus, and E. faecium) on both initial and repeat testing. A fourth isolate (S. hominis) was reported as showing no growth on the first run but was susceptible (correct result) on repeat testing and was included in error calculations. Therefore, the denominator is 48 for staphylococci, 49 for enterococci, and 97 for total categorical and essential errors.

^cOne S. aureus isolate did not grow repeatedly in the VITEK card. Therefore, the denominator is 49 for staphylococci and 99 for total categorical and essential errors.

^dSee Materials and Methods for definitions of categorical and essential agreement.

^eNA, not applicable.

TABLE 2.

Organisms for which very major or major categorical errors were reported by testing methods compared to the results of broth microdilution

Laboratory no.	Species	Reference linezolid MIC (σg/ml) (interpretation)	Mutation in 23S rRNA gene	Test system(s) showing errorsa	Comment
3213	S. aureus	16 (NS)e	G2576T	D, E, V, V2	VMc errors
1909	S. aureus	4 (S)	None	V, V2	Major errors
3487	S. aureus	16 (NS)	G2576T	D, E, P, V	VM errors
3036	S. aureus	8 (NS)	G2576T	D, E, P, V	VM errors
5721	S. aureus	8 (NS)	None	D, E, P, V	VM errors
1458	S. aureus	16 (NS)	G2576T	E, V	VM error, NG (VITEK)d
8053	S. aureus	16 (NS)	NDb	V	VM error
7338	S. epidermidis	>16 (NS)	C2534T	D, E, M	VM errors
0538	S. epidermidis	>16 (NS)	C2534T	D, P	VM errors
0960	S. haemolyticus	>16 (NS)	ND	D	VM error
0908	S. epidermidis	16 (NS)	ND	D	VM error
0006	E. faecium	8 (R)	ND	V	VM error
3419	E. faecium	16 (R)	G2576T	D, E, V	VM errors
8766	E. faecium	16 (R)	ND	V	VM error
3217	E. faecalis	≤1 (S)	ND	P	Major error

^aD, disk diffusion; E, Etest; M, MicroScan; P, Phoenix; V, VITEK; V2; VITEK 2.

^bND, not determined.

^cVM, very major.

^dThe isolate that showed no growth (NG) was not considered an error in error rate calculations.

^eNS, nonsusceptible.

Disk diffusion.

The categorical agreement of disk diffusion and broth microdilution testing was 89.0%. Eight very major errors were observed with staphylococci (four S. aureus and four CoNS strains), including 3213, 3487, 3036, 5721, 7338, 0538, 0960, and 0908), for a rate of 53.3%; one additional very major error was observed with an E. faecium strain (3419) (Table 2). For staphylococci, using transmitted light and measuring the innermost edge of the zone of inhibition would have reduced the very major error rate to 20.0% since four CoNS strains (7338, 0538, 0960, and 0908) and one S. aureus strain (3036) yielded zone diameters in the nonsusceptible range when read in this fashion (with an average decrease in zone diameter of 4 mm). Four S. aureus isolates produced nonsusceptible results on retesting with reflected light; a fifth nonsusceptible isolate would have been detected by using transmitted light. No major errors (i.e., false resistant results) were noted when using transmitted light to read the zone diameters of staphylococcal isolates. Two E. faecium isolates produced minor errors. Using transmitted light did not improve the interpretation of enterococcal results.

Etest.

The endpoint for the Etest method was 90% inhibition of growth when read by using reflected light, as defined in the package insert. With this endpoint, the Etest method showed categorical and essential agreement levels of 90.0 and 92.0%, respectively, compared to the results of the broth microdilution reference method. Etest had the highest percentage of very major errors for staphylococci (40.0%) among all of the MIC-based methods. S. aureus strains 3213, 3487, 3036, 5721, and 1458 and S. epidermidis strain 7338 all produced very major errors (Table 2), as did one E. faecium strain (3419). Two of the very major errors (strains 3036 and 7338) resolved after retesting by Etest (i.e., yielded nonsusceptible results). None of the very major errors were obtained with isolates that gave light growth. Three minor errors, all with E. faecium strains, were also observed.

MicroScan.

Among the automated systems, MicroScan showed the highest level of categorical and essential agreement (96.0 and 99.0%, respectively) compared to the results of the broth microdilution reference method. A single very major error (false susceptible result) was noted for an S. epidermidis isolate (7338 with a documented mutation of C2534T) that was initially reported by MicroScan as linezolid susceptible (MIC = 2 μg/ml) but was reported as resistant on retesting (MIC, >4 μg/ml) (Table 2). This was also the only essential error noted for the MicroScan system. Three minor errors, all with E. faecium isolates, varied within the nonsusceptible category (i.e., all results for both the MicroScan and reference methods were either intermediate or resistant), so no linezolid-nonsusceptible organisms were undetected. Version 1.61 of the MicroScan software did not provide an interpretation of nonsusceptible results for staphylococci (i.e., the interpretation field was blank on the reports). Category interpretations (susceptible, intermediate, or resistant) were provided for all results for enterococci.

Phoenix.

The Phoenix system showed a category agreement of 89.6% with the broth microdilution reference method, although the overall essential agreement was higher (95.8%). Testing runs were aborted twice by the system prior to completion for three organisms (one isolate each of S. epidermidis, S. aureus, and E. faecium), all of which were susceptible to linezolid by the broth reference method, so no results were available for those organisms. An S. hominis isolate (also linezolid susceptible) failed to grow in the Phoenix test panel during initial testing but did grow on subsequent testing. The Phoenix system reported four linezolid-nonsusceptible isolates of staphylococci, including three S. aureus isolates (5721, 3036, and 3487) and one S. epidermidis isolate (0538), as susceptible, resulting in a very major error rate of 26.7% for staphylococci (Tables 1 and 2). Three of the four very major errors were corrected by Phoenix on retesting. There was a single major error reported for E. faecalis isolate 3217 (reference MIC, ≤1 μg/ml; Phoenix MIC, >4 μg/ml) and additional minor errors for three E. faecalis and two E. faecium isolates; the results for the three E. faecalis isolates were false susceptible results. There were three essential errors with staphylococci and one with an enterococcus. The interpretation given by the Phoenix system for staphylococci with nonsusceptible linezolid results was “N” (i.e., not susceptible).

VITEK.

The VITEK system produced the largest number of linezolid testing errors among the automated systems in this study. The overall categorical and essential agreements were both 85.9%. There were five very major errors for S. aureus isolates, including strains 3213, 3487, 3036, and 5721, organisms for which there were problems on multiple systems (Table 2), and strain 8053. There were also three very major errors with E. faecium isolates (3419, 8766, and 0006). One S. aureus isolate (1458) failed to grow twice in the VITEK cards, although it grew in all of the other automated systems. One additional susceptible S. aureus isolate (1909) was reported on original testing and retesting as nonsusceptible (MIC = 8 μg/ml). There were six minor errors with enterococci, most of which were nonsusceptible (i.e., either intermediate or resistant) by broth microdilution and reported as susceptible by VITEK. The VITEK system also had the largest number of essential agreement errors, primarily among staphylococci that were correctly reported categorically as susceptible but for which the linezolid MICs were reported as 4 μg/ml instead of ≤1 μg/ml.

VITEK 2.

The VITEK 2 system demonstrated a categorical agreement of 93.0% and an essential agreement of 92.0% with the results of broth microdilution testing. The single very major error with S. aureus isolate 3213 (documented mutation of G2576T) (MIC = 4 μg/ml) was corrected by VITEK 2 on retesting (MIC, >8 μg/ml). A single major error was observed with S. aureus isolate 1909, which had no documented mutation by sequence analysis (VITEK 2 MIC, >8 μg/ml; reference MIC = 4 μg/ml); this error was reproduced on retesting (data not shown). Five minor errors were observed with enterococci (two E. faecium and three E. faecalis strains); only two of the errors (both with susceptible E. faecalis isolates) were resolved on retesting. There were eight essential errors noted with the VITEK 2 system—five with staphylococci and three with enterococci. VITEK 2 also provided no interpretation of nonsusceptible results for staphylococci. Of concern, the VITEK 2 advanced expert system recommended reporting all of the linezolid-nonsusceptible results for staphylococci as susceptible. However, the expert results were not used in error calculations for this study.

DISCUSSION

Resistance to linezolid, although rare, has been recognized in both staphylococci (15, 16, 23) and enterococci (1, 10). The resistant strains reported in the literature were detected by either Etest or disk diffusion. However, problems with false linezolid resistance results obtained with Etest and VITEK 2 have been reported (18), prompting several manufacturers of automated susceptibility testing systems to issue alerts regarding the detection of linezolid-resistant organisms. We also noted that several isolates of staphylococci and enterococci that were sent to either the CDC or Project ICARE for confirmation of a linezolid-nonsusceptible phenotype produced indistinct zones of inhibition when tested by both the disk diffusion and Etest methods, making interpretation of the test results difficult. Depending on where the endpoints for the agar-based methods were read, the results could be interpreted as either susceptible or nonsusceptible, particularly for the staphylococci. While enterococci (both E. faecalis and E. faecium) typically produced lighter growth than the staphylococci on disk diffusion plates, the zones of inhibition tended to be more distinct for the enterococci than for the staphylococci. For this study, the results recorded at 18 to 20 h by using reflected light were used to establish categorical and essential errors (following CLSI guidelines). However, in order to interpret results for 12 organisms, which produced very light growth, it was necessary to extend the incubation time to 24 h. Interestingly, only two of those organisms were staphylococci (one S. haemolyticus strain and one S. simulans strain) and both were linezolid susceptible by all of the test methods. Similarly, 8 of the 10 enterococcal results that were read at 24 h did not result in errors even though two of the isolates were resistant. The remaining two E. faecium isolates showed minor errors. Thus, in this study, light growth was not strongly associated with erroneous results. For the results that were interpretable at 18 to 20 h, using transmitted instead of reflected light to read disk diffusion tests would have reduced the very major error rate for staphylococci from 53.3% to 20.0% but would have had little impact on the testing of enterococci, since most of the enterococci produced zones of inhibition with clear endpoints. No specificity problems (i.e., false resistant results) were noted in this study when using transmitted light to read plates; however, this procedure would need to be investigated prospectively in a multilaboratory study before it could be recommended as a routine procedure for reading linezolid zone diameters. We did not assess the effect of using transmitted light to read the Etest results in this study. However, it is possible that reading the inner zone would increase the sensitivity of the test. The effect on specificity of reading Etest results in this manner would have to be determined.

Overall, the most critical problem with linezolid susceptibility testing in this convenience sample of isolates was the inability to detect several of the nonsusceptible isolates of staphylococci. A total of 25 (8.4%) very major errors were observed among the 297 results reported for staphylococci for the six test methods evaluated. This included 18 S. aureus errors, 6 S. epidermidis errors, and 1 S. haemolyticus error. The establishment of intermediate and resistant breakpoints for staphylococci (e.g., designating an MIC of 8 μg/ml as intermediate and ≥16 μg/ml as resistant) would not rectify this problem of high very major errors since these breakpoints would still fail to capture several of the nonsusceptible strains in this study. On the other hand, it would be difficult to lower the breakpoints for linezolid to reclassify 4 μg/ml (which currently is in the susceptible range) as intermediate unless data were available that demonstrated a lack of clinical efficacy at this MIC breakpoint. Thus, optimizing the reading of these tests, particularly disk diffusion and Etest, is necessary to improve the accuracy of linezolid results for staphylococci.

Of the automated systems, MicroScan showed the highest overall category agreement in this study compared to the results of broth microdilution. This is consistent with the report of Qi et al., who showed a high correlation between the presence or absence of mutations in the 23S rRNA gene and the phenotype reported by the MicroScan system (18). VITEK 2 also had high category agreement, provided that the advanced expert system, which reported all linezolid-nonsusceptible staphylococci as susceptible, was not used. Yet, neither system provided a categorical result for strains classified as nonsusceptible, leaving this space on the report blank. This is potentially very confusing for prescribers.

S. aureus strain 5721 was an enigma in this study. While no mutations in the 23S rRNA gene were detected by DNA sequence analysis, the isolate consistently produced a broth microdilution MIC of 8 μg/ml (nonsusceptible) and therefore was classified for the purposes of this study as a nonsusceptible isolate. There was a consistent 90% decrease in growth at 4 μg/ml, with 100% inhibition occurring at 8 μg/ml. MicroScan and VITEK 2 also repeatedly classified this isolate as nonsusceptible, while all of the other systems consistently gave borderline susceptible results (e.g., Etest produced an MIC of 4 μg/ml with a clearly demarcated ellipse, and disk diffusion showed a zone diameter of 22 mm with no haze discernible by transmitted light). This isolate may have a novel mechanism of resistance, such as the recently described Cfr rRNA methyltransferase system, as opposed to mutations in the 23S rRNA gene (13). Further investigations are in progress.

In summary, several of the linezolid-nonsusceptible strains of staphylococci and enterococci in our challenge panel of isolates were difficult to detect with many of the currently available susceptibility testing systems. The problem of not detecting resistance, particularly in staphylococci, was much greater than the problem of overcalling resistance. Thus, testing of linezolid against staphylococci must be added to the growing list of challenges for antimicrobial susceptibility testing methods (20-22). Further studies of the agar-based methods (disk diffusion and Etest) are needed to better define the optimal endpoints for interpreting results of testing for linezolid against staphylococci and enterococci.