MIC results and category interpretations for all methods are shown in Table . For the method with BMIC-Difco, which we designated as the standard for our study, 34.9% of the results were in the intermediate category (that is, the isolates were vancomycin-intermediate Staphylococcus aureus [VISA]) and 65.1% were in the susceptible category. The percentage of results in the intermediate category for the other methods ranged from 0% with the Vitek Legacy system to 50.4% with the Phoenix system.
Vancomycin MICs and MIC categories determined by three reference methods and six commercial systems
The differences in MIC results among the eight methods compared with the results obtained with BMIC-Difco are shown in Table , along with the percentage of results within ±1 dilution (essential agreement) of the result obtained with BMIC-Difco. Essential agreement was excellent (98.4 to 100%) for all methods except that performed with the Vitek Legacy system.
Dilution difference of eight test methods compared with result by BMIC-Difco reference method
We compared the ability of the testing methods to determine the correct category interpretation compared with the category interpretations obtained with BMIC-Difco. The comparison and the kinds of discrepancies that occurred are shown in Table . The method with BMIC-BBL was the only one with ≥95% overall category agreement compared with the results obtained with BMIC-Difco. Except for the Vitek Legacy system, the category agreements for the other seven MIC methods varied within a close range of 84.5 to 92.2%. Category agreement between the method with BMIC-Difco and the Vitek Legacy system was only 64.8%. For the discrepancies that occurred, the method with BMIC-BBL, the agar dilution method, the Sensititre system, and the Vitek 2 system tended to categorize more VISA isolates as susceptible than the method with BMIC-Difco; Etest, the MicroScan system, and the Phoenix system categorized more susceptible isolates as VISA. The Sensititre system categorized one VISA isolate as resistant. None of the isolates tested with the Vitek Legacy system yielded MICs of 4 or 8 μg/ml; with that system, 35 of the VISA strains were categorized as susceptible and 10 were categorized as resistant.
Comparison of two reference and six commercial susceptibility testing methods and discrepancies for categorization of 129 S. aureus isolates compared with categorization by BMIC-Difco reference method
The results obtained by retesting of the 43 organisms for which a category discrepancy had occurred in the original study but with updated commercial panels for four of the systems are shown in Table . Again, essential agreement was excellent (97.5 to 100%) compared to the MICs obtained by the method with BMIC-Difco, although there were some differences in the trend toward higher or lower MICs and the kinds of category discrepancies noted. The MicroScan system again tended to categorize susceptible isolates as VISA and the Vitek 2 system tended to categorize VISA isolates as susceptible. The results for the Sensititre system changed from tending to undercall resistance to overcalling resistance. However, 40 of the 43 organisms retested had vancomycin MICs at the breakpoint (2 to 4 μg/ml), making minor errors more likely and complicating the conclusions that can be drawn from the retesting.
Dilution differences of five test methods compared with result by BMIC-Difco reference method for 43 isolates demonstrating category discrepancies in the original study
The vancomycin agar screen test (data not shown) and the disk diffusion test (Fig. ) were both insensitive for the detection of VISA isolates. None of the isolates with vancomycin MICs of ≤2 μg/ml grew on the agar screen plates, but isolates with vancomycin MICs of 8 μg/ml did grow. The results for isolates with vancomycin MICs of 4 μg/ml were variable; only 12 of 36 (33.3%) of these isolates grew on the agar screen plates. All of the VISA isolates, regardless of whether the MIC was 4 or 8 μg/ml, were categorized as vancomycin susceptible by disk diffusion (zone diameters, ≥15 mm) (Fig. ). CLSI recently rescinded the vancomycin disk diffusion breakpoints for S. aureus
and they no longer appear in Table of the CLSI performance standards (4
Scatterplot of vancomycin MICs (μg/ml) determined by broth microdilution with BMIC-Difco compared with vancomycin MICs determined by disk diffusion (mm). CAMHB, cation-adjusted Mueller-Hinton broth.
The results obtained by Etest compared with those obtained with BMIC-Difco are shown in Fig. .
Scatterplot of vancomycin MICs (μg/ml) determined by broth microdilution with BMIC-Difco compared with vancomycin MICs determined by Etest. CAMHB, cation-adjusted Mueller-Hinton broth.
Only a few studies have assessed the accuracy of susceptibility testing methods for detecting VISA strains. In 1998, CDC tested eight VISA strains using Etest, conventional and rapid MicroScan panels, the Sensititre system, the Vitek Legacy system, and the vancomycin agar screen method. All the commercial MIC methods, except the MicroScan rapid panels, gave acceptable results under the CLSI interpretive criteria used before 2006 (8
). The authors reported that the vancomycin agar screen plates prepared in-house did not perform as well as commercially prepared plates. The commercially prepared plates detected all the isolates (n
= 8) with vancomycin MICs of 8 μg/ml and only one of four isolates with MICs of 4 μg/ml. These findings are consistent with the findings of the present study.
Recently, Wootton et al. compared the results of three methods for their ability to differentiate vancomycin-susceptible isolates from VISA and heteroresistant VISA (hVISA) strains: the CLSI vancomycin agar screen method; an agar screen method that uses Mueller-Hinton agar containing 5 μg/ml of teicoplanin; and the macro-Etest, a method that uses vancomycin and teicoplanin Etest strips on brain heart infusion agar seeded with an inoculum suspension equivalent to a 2.0 McFarland standard (10
). These methods were compared to population analysis, which is considered the gold standard for the identification of hVISA. It is difficult to compare the results of their study with the results of our study because their strains were grouped by MIC categories that overlap the current CLSI criteria; i.e., vancomycin-intermediate strains were those for which the MICs were ≥8 μg/ml, heteroresistant strains were those for which the MICs were 1.5 to 4 μg/ml, and susceptible strains were those for which the MICs were ≤1 μg/ml. The success of the macro-Etest method in that study was contingent upon the definition of a VISA isolate as one for which the vancomycin MIC was ≥8 μg/ml.
When the MIC results for a large number of organisms are close to the susceptible or intermediate breakpoint, as is true in this study, it is possible to have excellent essential agreement but relatively poor category agreement. Our data show that when the MICs range from 2 to 8 μg/ml, Etest, the MicroScan system, and the Phoenix system tended to give higher vancomycin MIC results than the broth microdilution reference method; agar dilution, the Sensititre system, and the Vitek 2 system tended to give results lower than those obtained by other methods, even though the essential agreement of all the methods was excellent (Table ). When the organisms responsible for the category errors were retested on newer panels for the MicroScan, Phoenix, Sensititre, and Vitek 2 systems, category discrepancies continued to occur, despite the excellent essential agreement.
Recognition of intermediate vancomycin resistance among S. aureus
isolates can be medium dependent (8
).When we used Mueller-Hinton broth from a different manufacturer with the broth microdilution reference method, the results yielded approximately 5% minor errors (Table ) compared to the results obtained with the gold standard broth. Although subtle differences in the performance of the susceptibility testing methods could be because of differences in the media or the concentrations tested, such as the half intervals of the Etest, all the VISA isolates tested in our study (i.e., isolates with vancomycin MICs of 4 to 8 μg/ml with BMIC-Difco) produced MICs of ≥2 μg/ml by the other test methods. Therefore, to increase the rate of detection of VISA strains in the laboratory, the laboratory could consider testing all S. aureus
isolates with vancomycin MICs of 2 μg/ml by an alternate method. Unfortunately, the vancomycin agar screen test was not very sensitive for detecting strains when the vancomycin MICs were 4 μg/ml. For the users of automated systems, Etest may be a good alternative testing method because it detected 44 of 45 (97.8%) VISA isolates.
In a recent review, Appelbaum (1
) suggested that S. aureus
strains for which the vancomycin MICs are 1 or 2 μg/ml and that are isolated from patients who are failing vancomycin therapy with a glycopeptide should be investigated further. Some isolates with vancomycin MICs in this range have been identified as having vancomycin heteroresistance and have been associated with therapeutic failures (8
). Although it is true that efforts to investigate these isolates further would likely identify more vancomycin-intermediate strains, most clinical laboratories are unaware of poor treatment outcomes that would trigger further investigation. Furthermore, given that the modal vancomycin MIC of S. aureus
isolates in the United States is 1.0 μg/ml, as reported in several surveillance studies (5
) and in additional data at the EUCAST website (http://www.eucast.org
), testing of all strains with vancomycin MICs of 1.0 μg/ml would be burdensome. Limiting of further testing by any of the test methods evaluated in the present study to strains with vancomycin MICs of 2 μg/ml may be more feasible.
In summary, with the exception of the Vitek Legacy system, the performance characteristics of all the MIC susceptibility testing methods, when they were measured by essential agreement, were excellent. However, the Phoenix system, Etest, and the MicroScan system tended to yield MIC results 1 dilution higher than those obtained by the broth reference method; and agar dilution, the Sensititre system, and the Vitek 2 system yielded results that were 1 dilution lower than those obtained by the broth reference method. The Vitek Legacy system gave no MIC results of 4 or 8 μg/ml, and thus, it is difficult to compare the results obtained with the Vitek Legacy system with those obtained by the reference method. The disk diffusion test did not distinguish vancomycin-intermediate strains from vancomycin-susceptible strains, and the vancomycin agar screen lacked sensitivity for strains with MICs of 4 μg/ml. Clinical laboratories may enhance their ability to detect S. aureus isolates with reduced susceptibility to vancomycin by performing further testing (e.g., by the vancomycin Etest) with isolates for which the MICs are 2 μg/ml with one of the commercial systems evaluated in the present study.