Comparative Evaluation of Etest and Sensititre YeastOne Panels against the Clinical and Laboratory Standards Institute M27-A2 Reference Broth Microdilution Method for Testing Candida Susceptibility to Seven Antifungal Agents

ABSTRACT To assess their utility for antifungal susceptibility testing in our clinical laboratory, the Etest and Sensititre methods were compared with the Clinical and Laboratory Standards Institute (CLSI) M27-A2 reference broth microdilution method. Fluconazole (FL), itraconazole (I), voriconazole (V), posaconazole (P), flucytosine (FC), caspofungin (C), and amphotericin B (A) were tested with 212 Candida isolates. Reference MICs were determined after 48 h of incubation, and Etest and Sensititre MICs were determined after 24 h and 48 h of incubation. Overall, excellent essential agreement (EA) between the reference and test methods was observed for Etest (95%) and Sensititre (91%). Etest showed an ≥92% EA for MICs for all drugs tested; Sensititre showed a ≥92% EA for MICs for I, FC, A, and C but 82% for FL and 85% for V. The overall categorical agreement (CA) was 90% for Etest and 88% for Sensititre; minor errors accounted for the majority of all categorical errors for both systems. Categorical agreement was lowest for Candida glabrata and Candida tropicalis with both test systems. Etest and Sensititre provided better CA at 24 h compared to 48 h for C. glabrata; however, CA for C. glabrata was <80% for FL with both test systems despite MIC determination at 24 h. Agreement between technologists for both methods was ≥98% for each agent against all organisms tested. Overall, Etest and Sensititre methods compared favorably with the CLSI reference method for determining the susceptibility of Candida. However, further evaluation of their performance for determining the MICs of azoles, particularly for C. glabrata, is warranted.

Fungi have become pathogens of increasing concern over the past several decades, and Candida is now a common cause of nosocomial bloodstream infection in the United States (6). Accordingly, antifungal prophylaxis is routinely used in several populations considered at high risk for candidiasis, and breakthrough infections in patients on such prophylactic regimens are increasingly reported. The breakthrough organisms often have higher MICs of the drug to which they were exposed, and some may exhibit heteroresistance (14)(15)(16) and cross-resistance to one or more additional agents (1,11,33,34). Although fluconazole and, more recently, voriconazole, have been widely used as prophylactic agents, the introduction of newer agents and several supporting recent studies will likely increase the spectrum of agents used for this purpose. Micafungin, a member of the echinocandin class of antifungal agents, has now been cleared by the U.S. Food and Drug Administration (FDA) for prophylactic use against Candida in hematopoietic stem cell transplant recipients (32). Posaconazole has also been approved by the FDA for prophylaxis in patients with graft-versus-host disease and in high-risk neutro-  M716, 2005). The increasing number of antifungal agents now on the market, the widespread use of these agents in prophylactic regimens, and the frequent isolation of Candida species with unpredictable patterns of susceptibility to these agents has led clinicians to seek guidance from the clinical laboratory when selecting antifungal therapy for these complex, infected patients (26).
The Clinical and Laboratory Standards Institute (CLSI; formerly the National Committee for Clinical Laboratory Standards [NCCLS]) has developed and published an approved reference method for broth microdilution testing (CLSI document M27-A2) of Candida species. This method was developed through a consensus process to facilitate agreement among laboratories in determining the susceptibility of Candida species to several antifungal agents (20). This test method was published as a reference standard for use in clinical laboratories in 1997, updated as M27-A2 in 2002, and remains the standard method of susceptibility testing in many clinical laboratories. Unfortunately, this reference method is labor-intensive. Commercial systems are now available to clinical laboratories for determining MICs of various agents against Candida species. In November 2001, the Sensititre YeastOne test panel (Trek Diagnostic Systems, Inc., Westlake, OH) was cleared for commercial sale by the FDA as a new and less labor-intensive method for determining MICs of fluconazole, itraconazole, and flucytosine for yeasts (8). Etest (AB BIODISK, Solna, Sweden), an agar-based predefined concentration gradient method for determining the MICs of various agents, is also available for Candida and could reduce further the effort required for routine susceptibility testing (17,30).
The most important feature of a laboratory test for susceptibility testing is its ability to produce accurate results consistently within an appropriately rapid turnaround time in order to influence clinical decisions. Initial evaluations of these parameters are often performed in specialty or research laboratories by personnel with expertise in the test application area. Even FDA clearance of a commercial product does not predict how well that product will perform in an individual laboratory under actual testing conditions and with the clinical specimen mix encountered in particular hospital populations. Therefore, before one of the new antifungal susceptibility test devices can be used for clinical testing in place of the CLSI M27-A2 reference standard, a detailed comparative evaluation of its performance should be investigated by the offering laboratory. This local validation and verification of new test devices is a necessary part of the quality assurance program for all clinical laboratories in order to meet accreditation standards as required by the Centers for Medicare and Medicaid Services (4).
This single-center study was designed to compare the Etest and Sensititre antifungal susceptibility test systems with the CLSI M27-A2 reference method for seven antifungal agents in a hospital clinical laboratory as required by regulatory agencies. The present study is one of the first in a clinical laboratory setting to document the performance of these new test systems for seven of the most frequently used antifungal agents.

MATERIALS AND METHODS
Isolates. Duke University Health System Institutional Review Board approval was obtained for the present study. All patients with Candida bloodstream infection at Duke University Medical Center in 2000 and 2001 were identified via a review of the clinical microbiology laboratory database. A new episode of infection was defined as the recovery of the same Candida species more than 14 days after the last positive blood culture or if a new species was found Ն7 days since the last positive blood culture for Candida. A total of 212 initial isolates recovered from unique, consecutive episodes of bloodstream infection were retrieved from frozen (Ϫ70°C) storage for testing. Prior to testing, each isolate was subcultured at least twice on potato-dextrose agar to ensure purity and viability. All isolates were subcultured again 24 h prior to testing.
Definitions. Standard antimicrobial susceptibility testing definitions were used (7,18). For MICs measured with a continuous concentration scale (i.e., Etest) that yielded results falling in between conventional serial twofold dilutions, the next higher dilution was assigned. The results were considered to be in essential agreement when the test result was within two dilutions of the reference value. The percent essential agreement was calculated by using the number of test results in essential agreement as the numerator and the total number of organisms tested as the denominator. The results were considered to be in categorical agreement when the test and reference MICs fell within the same interpretive category (i.e., susceptible, susceptible dose dependent, intermediate, or resistant, depending on the agent tested). Established CLSI interpretive breakpoints were used to tabulate categorical agreement and very major, major, and minor errors for fluconazole, itraconazole, and flucytosine; proposed interpretive breakpoints were used for voriconazole (20,23). The percent categorical agreement was calculated by dividing the number of tests with no category discrepancy by the number of organisms tested. "Very major errors" occurred when the reference method categorized the organism as resistant, but the test method categorized the organism as susceptible (falsely susceptible). The percent very major errors were calculated by using the number of resistant isolates as the denominator. "Major errors" occurred when the reference method categorized the isolate as susceptible, but the test method categorized it as resistant (falsely resistant) and were calculated by using the number of susceptible isolates as the denominator. "Minor errors" occurred when the reference method categorized an organism as susceptible or resistant and the test method categorized it as susceptible dose dependent (or intermediate) or the reference categorized it as susceptible dose dependent (or intermediate) and the test method categorized it as susceptible or resistant. The percent minor errors were calculated by using the total number of organisms tested as the denominator.
Testing. Each Candida isolate was tested with the three methods in parallel as outlined below by two fully certified (American Society of Clinical Pathologists) medical technologists (B. D. Alexander and T. C. Byrne) with training in antifungal susceptibility testing. All seven antifungal agents were tested with each isolate by all test systems with the exception of posaconazole, which was not included on the Senstitre panels. Technologists interpreted results from each test system independent of the other test system results and independent of each other. The results were recorded on paper onto a preset grid, and a database expert (K. L. Smith) subsequently entered these results into an electronic database. Testing was repeated for tests with major and very major errors (for drugs with categorical interpretive criteria available) and for MIC results that did not agree with the reference standard within two dilutions. Discrepancies were resolved through repeat testing of both systems (the test method and reference method) in triplicate.
Inoculum suspension. Yeast inoculum suspensions were prepared as described in the CLSI M27-A2 document using sterile 0.85% saline (20). The cell density was adjusted with a spectrophotometer by adding sufficient saline to match the transmittance produced by a 0.5 McFarland density standard at a 530-nm wavelength, resulting in a concentration of 1 ϫ 10 6 to 5 ϫ 10 6 cells per ml. This suspension was used to directly inoculate agar plates for Etest and was diluted as directed by CLSI M27-A2 for the broth microdilution procedure. For the Sensititre system, a suspension of yeast in demineralized water was adjusted to match the turbidity of a 0.5 McFarland standard by using a Sensititre nephelometer; 20 l of this suspension was transferred into an 11-ml tube of Sensititre YeastOne inoculum broth, resulting in a concentration of 1.5 ϫ 10 3 to 8 ϫ 10 3 cells per ml. This broth suspension was then used to inoculate the wells of the Sensititre plates.
CLSI M27-A2 broth microdilution method. The CLSI M27-A2 broth microdilution method was performed in accordance with the published guidelines that specify RPMI 1640 medium (Remel, Lenexa, KS) buffered to pH 7.0 with MOPS (morpholinepropanesulfonic acid) (20). Microdilution trays were incubated at 35°C and read visually after 48 h of incubation. For amphotericin B, the MIC endpoint was defined as the lowest concentration with complete (100%) growth inhibition. For the azoles, flucytosine, and caspofungin, an 80% inhibition in growth compared to that of the drug-free growth control was used as the MIC endpoint.
Etest predefined concentration gradient method. The Etest method was performed according to the manufacturer's instructions. Preformed and predefined gradients of the different agents immobilized on individual plastic test strips covered a continuous concentration range across 15 dilutions. A cotton-tipped, sterile swab was used to inoculate the organism from the yeast suspension onto a 150-mm agar plate containing RPMI 1640 medium supplemented with 2% glucose and buffered with MOPS to pH 7.0. After allowing the excess moisture to be fully absorbed into the agar, Etest strips were applied to the inoculated surface. The plates were incubated at 35°C and read at 24 and 48 h. The MIC was read as the lowest concentration at which the border of the elliptical zone of growth inhibition intersected the scale on the test strip. For the azoles and caspofungin an 80% inhibition in growth was used as the MIC cutoff (microcolonies were ignored), and for flucytosine and amphotericin B the MIC endpoint was defined as the lowest concentration with nearly complete (90%) and complete (100%) growth inhibition, respectively.
Sensititre YeastOne colorimetric antifungal susceptibility test method. The Sensititre antifungal susceptibility test method was performed according to the manufacturer's instructions. Plates containing serial twofold dilutions of the antifungal agents across 12 dilutions were inoculated using a suspension of the organism prepared as described above, incubated at 35°C, and read at 24 and 48 h. The MIC was recorded as the lowest concentration of antifungal agent preventing the development of a red color (the first blue well).

VOL. 45, 2007
COMPARING Etest AND Sensititre WITH CLSI M27-A2 METHOD Quality control strains. Quality control was performed for each test method every day the tests were performed. Quality control strains included C. albicans ATCC 90028 (Etest), C. krusei ATCC 6258 (Etest, Sensititre, and CLSI M27-A2), and C. parapsilosis ATCC 22019 (Etest, Sensititre, and CLSI M27-A2). The MIC control ranges used were those established by Barry et al. (2) for the reference method and those provided by the manufacturers for the test devices.
Data analysis. Performance based on categorical (qualitative) and MIC (quantitative) comparisons between the methods (Etest versus CLSI M27-A2 and Sensititre versus CLSI M27-A2) were calculated and reported. Primary analysis included overall percent essential agreement (all organisms and all antifungal agents tested) and overall percent categorical agreement (for the four agents with CLSI MIC breakpoints) at 48 h. Secondary analyses included the percent essential and categorical agreement at 48 h for all organisms with each antifungal agent, each Candida species with all antifungal agents, and each Candida species with each antifungal agent. The percent essential agreements between technologists for each antifungal agent against all organisms tested for both the Etest and Sensititre methods were also calculated.
Acceptable percent essential agreement for MICs was set at Ն90% for each antifungal agent against all organisms tested. An acceptable overall categorical agreement was set at Յ1.5% very major errors and Յ3% major errors. An acceptable percent essential agreement for MICs between technologists was set at Ն95% for each antifungal agent against all organisms tested (18,19).
All isolates tested in the study were included in the final analysis. Descriptive statistics were used to report isolate demographic and susceptibility profiles. Proportions of categorical and essential agreements at 24 h and 48 h were compared by using McNemar's chi-square test. Statistical significance was set at the 5% level. All P values and 95% confidence intervals (CI) were two tailed; 95% CI values for agreement were based on a binomial distribution. Statistical analyses were performed by using Analyze-It software (Analyze-It Software, Ltd., Leeds, United Kingdom) for Microsoft Excel.

RESULTS
A total of 212 Candida isolates, including 94 C. albicans, 38 C. glabrata, 34 C. tropicalis, 31 C. parapsilosis, 5 C. krusei, 8 C. lusitaniae, and 2 C. guillermondii isolates, were tested. Table 1 shows the MICs at which 90% of the isolates tested (MIC 90 ) were inhibited for each drug, and Table 2 displays the categorical distribution for susceptibility for each Candida species to fluconazole, itraconazole, voriconazole, and flucytosine based on the CLSI M27-A2 reference method of testing. Table  3 provides a summary of MIC 90 and MIC ranges as determined by CLSI M27-A2, Etest, and Sensititre test methods at 48 h.
The percent essential and categorical agreements for all organisms with each antifungal agent are summarized in Table  4. Etest demonstrated Ն92% essential agreement for MICs for all drugs tested. Sensititre demonstrated Ն92% essential agreement for MICs for itraconazole, flucytosine, amphotericin B, and caspofungin but 82 and 85% essential agreement for fluconazole and voriconazole, respectively. The categorical agreement was Ն95% for flucytosine and voriconazole with both test systems, but 88 and 83% for fluconazole and 80 and 73% for itraconazole for the Etest and Sensititre systems, respectively.
The percent essential and categorical agreements for each   Candida species with all antifungal agents are summarized in Table 5. Low percent categorical agreements for C. krusei, C. lusitaniae, and C. guillermondii were possibly exaggerated by the relatively small number of these Candida species tested. Otherwise, C. glabrata and C. tropicalis showed the lowest rates of categorical agreement in both test systems; categorical agreement was lowest for C. glabrata to fluconazole and C. glabrata and C. tropicalis to itraconazole for both test systems and C. parapsilosis to itraconazole for Sensititre. The percent essential and categorical agreements at 24 h were compared to percent essential and categorical agreements at 48 h for Etest and Sensititre for the azole antifungals. When organisms with insufficient growth at 24 h were excluded from the analysis, the overall categorical agreement was not significantly different at the 24-and 48-h readings for Etest (P ϭ 0.141); for Sensititre, significantly better categorical agreement was obtained with 24-h readings (P ϭ 0.005) ( Table 6). Similar results were obtained when the 24-h versus the 48-h readings for fluconazole and itraconazole were assessed independently (data not shown). For C. glabrata, there was no significant difference in the percent essential agreement between the 24-and 48-h readings with Etest; however, the categorical agreement was improved at 24 h (P ϭ 0.024). For Sensititre, significantly better essential (P ϭ 0.0001) and categorical (P ϭ 0.005) agreements were achieved with 24 h compared to the 48-h results. For C. tropicalis, reading at 48 h provided significantly better essential agreement than reading at 24 h (P ϭ 0.0001) for Etest, although this provided no improvement in categorical agreement. For the Sensititre system, no significant difference in essential or categorical agreement was observed between the 24-and 48-h readings.
Minor errors accounted for the majority of all categorical errors for both systems. For C. glabrata and C. parapsilosis, 100% of errors for itraconazole were minor with both test systems; for C. tropicalis, 92 and 91% of errors for itraconazole were minor with Etest and Sensititre, respectively. For C. glabrata, 88 and 96% of the errors for fluconazole were minor for Etest and Sensititre, respectively. Only two very major errors were documented. Both very major errors occurred with the Etest system for different isolates of C. tropicalis: one with itraconazole and one with voriconazole.
Agreement between technologists for both the Etest and the Sensititre methods was 98% or greater for each antifungal agent against all organisms tested. Of the 1,484 total tests

DISCUSSION
The goal of the present study was to validate the performance in our laboratory of two new commercially available systems for antifungal susceptibility testing before implementing the test systems for clinical use. In our hands, the overall percent essential agreement for the Etest and Sensititre systems was Ͼ90%. When discrepancies compared to the CLSI method were found, Etest tended to yield higher MICs for all drugs except caspofungin, whereas Sensititre tended to yield higher MICs for all drugs tested except for caspofungin and flucytosine. The essential agreement was Ͼ90% for all drugs tested, except for fluconazole and voriconazole with the Sen-sititre system. Despite only 85% overall essential agreement for voriconazole with the Sensititre system, the categorical agreement was still 97%. On the other hand, the lack of essential agreement for fluconazole with the Sensititre system did lead to lower categorical agreement (83%) for this drug.
In general, the number of categorical errors was low for both test systems for the four agents with CLSI MIC breakpoints. When categorical errors did occur, minor errors accounted for the majority. We observed two very major (falsely susceptible) errors for the Etest. The purpose for calculating very major errors is to determine whether a test system can detect resistance since failure to do so could place a patient in jeopardy. The criteria for categorical errors used by the FDA in considering a susceptibility test system for clearance specify Յ1.5% very major errors and Յ3% major errors (10). For a clinical laboratory and test system to meet this stringent requirement for very major errors, at least 65 isolates resistant to each drug under study would need to be tested to allow one very major error per drug tested and still have an acceptable error rate of  Ͻ1.5% (7,18). Our cohort included 212 isolates with 56 total resistant results (the numbers of isolates resistant to fluconazole, itraconazole, voriconazole, and flucytosine were 10, 32, 7, and 7, respectively). When 65 isolates resistant to each drug are not available for testing as in our study, experts have suggested that a less stringent criterion be applied (e.g., the total of the very major and major errors should be Ͻ5%) (12). Regardless of the exact criterion used, it is important to examine a collection of challenge strains that contains enough resistant isolates to verify the ability of the new test to detect resistance. We tested a relatively large cohort of isolates with 56 instances of resistance to one or more drugs; the Etest system failed to detect resistance only twice. The Sensititre system did not fail to detect resistance. Importantly, both very major errors for the Etest system were found in testing C. tropicalis against azole antifungals. One potential explanation is related to known limitations of the reference method. Trailing endpoints, a well-described phenomenon with azole antifungals and the CLSI M27-A2 microdilution method, can lead to overinterpretation of the MICs, particularly for C. tropicalis. In six of the seven discrepancies between Etest and CLSI M27-A2 azole results for C. tropicalis, the MIC was lower by Etest. For Sensititre, also a microdilution method, only 1 of the 13 discordant MIC azole results for C. tropicalis was lower than the CLSI M27-A2 result. Thus, it is plausible that improved endpoint discrimination by Etest accounts for its discrepancies with the CLSI reference method.
Of the drugs tested, itraconazole and fluconazole were associated with the lowest rates of overall categorical agreement in both test systems. Furthermore, C. glabrata had high rates of nonconcordance for fluconazole and itraconazole compared to other species, and C. tropicalis had high rates of nonconcordance with itraconazole. These findings are important given that one of the most clinically pertinent and frequent reasons to perform antifungal susceptibility testing is the need to determine the susceptibility of non-albicans Candida species to the azoles. For all discrepant results for C. glabrata and an azole, the MIC was higher by both test systems compared to the reference MIC. This raises the possibility that C. glabrata may be associated with azole resistance detectable by the test systems but undetectable by the reference method. Heteroresistance, or selection of a resistant subpopulation of cells from a population of predominantly susceptible cells, could account for such a phenomenon. Heteroresistance has been described in both Cryptococcus neoformans and C. albicans (15,16,29). The molecular mechanism associated with heteroresistance in C. albicans, i.e., the upregulation of CDR genes responsible for encoding cell wall pumps that infer cross-resistance among azoles, has also been associated with the development of fluconazole resistance in C. glabrata (3,27,28). Further, detection of the heterogenous phenotype in strains of C. albicans was possible by using Etest; heterogenous isolates displayed increased satellite growth in the inner zone of growth inhibition, while homogeneous isolates displayed clearly demarcated zones of inhibition (15). It is therefore feasible that the test methods may be better able to detect heteroresistance than the reference method.
Other investigators have also found lower rates of agreement of C. glabrata and C. tropicalis for azoles when comparing the Etest and Sensititre systems to the CLSI M27-A2 method. Moreover, incubation times have been shown to significantly influence MIC results. Based on the current literature, however, it is not clear whether reading endpoints at 24 h provides better agreement than reading at 48 h for the different species of Candida. For instance, in a study of fluconazole that compared Etest with CLSI M27-A2, there was better essential agreement at 24 h than at 48 h for C. tropicalis and C. albicans, but lower agreement at both times for C. glabrata (30). In a study of posaconazole, C. glabrata and C. tropicalis showed the lowest concordance rates between Etest and CLSI M27-A2 at both times. However, C. glabrata essential agreement was higher at 24 h (72%) than at 48 h (56%), whereas C. tropicalis was higher at 48 h (71%) than at 24 h (39%) (31). In studies comparing the Sensititre system with the CLSI M27-A2 method, agreement at 24 h compared to 48 h also seemed to vary for the different Candida species. In one study that defined essential agreement as within three dilutions of the reference standard, the best performance of the Sensititre system was with 24 h MICs (92 to 100% essential agreement) with the azoles and flucytosine for all of the species tested, with the exception of C. albicans (87 to 90%) (9). In another study, overall agreement between the Sensititre and CLSI M27-A2 methods was better at 24 h (95 to 99% agreement) than at 48 h (80 to 97% agreement) for all species and antifungal agents (posaconazole, ravuconazole, and voriconazole) because Sen- sititre MICs were consistently more than two dilutions higher than the reference MICs, especially for isolates of C. albicans, C. glabrata, and C. tropicalis (8). Finally, in a study comparing the Sensititre to the CLSI M27-A2 method for voriconazole, agreement was Ͻ90% for C. glabrata, despite the 24-h readings (24 (5,13), there is no current CLSI method for MIC determinations. One published study suggested that using RPMI medium, the lowest concentration of caspofungin resulting in a significant diminution of growth below control growth levels as the endpoint, and reading endpoints at 24 h provides the best interlaboratory agreement when testing caspofungin (21). Under these conditions, a high level of discrimination was achieved between strains known to have lowered caspofungin susceptibility (based on molecular evidence of mutations in the FKS 1 gene) and normally susceptible isolates. The range of caspofungin MICs in our cohort of pedigree isolates was Ͻ0.03 to 2 g/ml, a finding consistent with ranges published by other investigators (22). Further, the MIC for the quality control isolates fell almost perfectly within expected limits, despite reading at 48 h. Our study suggests that the new test systems may provide results in concordance with those obtained with RPMI medium by the CLSI M27-A2 method in testing caspofungin.
In order to ensure general applicability of the study results, our selection of isolates for testing included a spectrum of Candida species typically isolated in a clinical microbiology laboratory and mirrored national and international epidemiological data (35). Moreover, a broad range of MICs was also exhibited by the isolates from blood cultures. For example, 50% of the C. glabrata blood isolates demonstrated an MIC of Յ8 g/ml (susceptible), 42% demonstrated an MIC of 16 to 32 g/ml (susceptible dose dependent), and 8% demonstrated an MIC of Ն64 g/ml (resistant) to fluconazole. Nonetheless, 66% (25 of 38) of C. glabrata isolates tested were resistant to itraconazole, whereas only 8% (3 of 38) tested resistant to fluconazole based on current interpretive categorical breakpoints. Given drug utilization at our hospital, one might have expected fluconazole resistance to exceed itraconazole resistance for C. glabrata if this finding were attributable to treatment-induced resistance. This paradox has been observed in other epidemiologic studies as well (22). In a large study including 949 clinical isolates of C. glabrata, all of the 78 strains that were resistant to fluconazole were also resistant to itraconazole (25). Conversely, of the 567 isolates found to be susceptible to fluconazole, only 4% were susceptible to itraconazole (inhibited by itraconazole at an MIC of Յ0.12 g/ml). Whether this finding reflects a unique mechanism for resistance to itraconazole that does not impart resistance to fluconazole in C. glabrata, inappropriate thresholds for categorical classifications of resistance for itraconazole, or inability of the reference method to detect resistance to fluconazole in C. glabrata is uncertain, but we suspect one of the latter.
In order to comply with the recommended guidelines for verification studies that use a reference method that is known to be an imperfect standard, discrepant test results were resolved by retesting in triplicate both the new test system and the reference method. Initial discrepancies that resolved with repeat testing included three major errors (all for flucytosine) with Etest and five major errors (two for fluconazole, two for itraconazole, and one for voriconazole) with Sensititre. All major error resolutions were due to changes in initial results with the Etest or Sensititre systems. Using initial test results for Etest, 0.9% major errors occurred compared to 0.5% major errors when repeat test results were used. Using initial test results for Sensititre, 1.4% major errors occurred compared to 0.7% major errors when repeat test results were used. Thus, using the initial test results, major errors were still within the acceptable overall categorical agreement for major errors (3%) with both test systems.
Seven very major error discrepancies (two with fluconazole, five with voriconazole) resolved with repeat testing for Etest and six very major error discrepancies (one with fluconazole, five with voriconazole) resolved with repeat testing for Sensititre. All very major error resolutions were due to changes in the initial CLSI reference method result. Interestingly, all but one of the very major discrepancies that resolved for both test systems were with the same isolates. This highlights known limitations of the current CLSI standard and supports the importance of repeating discrepancies in triplicate with both systems when performing verification studies for the clinical laboratory.
Clinical laboratories are playing an increasingly important role in antifungal drug selection by performance of susceptibility testing of Candida isolated from infected patients. It is essential that the antifungal drug testing systems used clinically provide accurate results when compared to the reference standard. The most important use of antifungal susceptibility testing in the clinical laboratory is to detect resistant isolates, and both Etest and Sensititre performed well in this regard. The methods also performed well for some of the newer antifungal agents including voriconazole, posaconazole, and caspofungin. Furthermore, interobserver variability between the two tech- nologists for Etest and Sensititre methods was minimal, which shows that intralaboratory variability should be minimal with appropriate training. However, our study underscores the influence of incubation time on MIC results, especially with the azole agents. The Sensititre system provided better overall categorical agreement with the reference standard when there was sufficient growth for the results to be read at 24 h rather than at 48 h. For Etest, reading at 24 h also provided better categorical agreement for C. glabrata. However, despite improved agreement by reading at 24 h, categorical agreement for C. glabrata remained disappointingly low (Ͻ80% for fluconazole) with both test systems. Fortunately, the majority of errors encountered with the new test systems were minor, and both tended to overcall resistance for C. glabrata and thereby yield a conservative result for patient care. Nonetheless, because assessing fluconazole activity against C. glabrata is commonly needed clinically, further evaluation of both test systems (and the CLSI reference method and breakpoints) is warranted.