INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The rising incidence of serious fungal infections has created an increased demand for reliable methods of in vitro testing of antifungal agents that can assist in their clinical use. The National Committee for Clinical Laboratory Standards (NCCLS) has developed a standardized broth macrodilution method for the testing of Candida spp. and Cryptococcus neoformans which has greatly improved the reproducibility of antifungal susceptibility testing and serves as the "gold standard" by which all new methods are compared (9). To increase the efficiency of testing large numbers of isolates, a broth microdilution adaptation of the reference method was developed and evaluated and gives nearly identical results (1, 5). However, even in the microdilution format, these methods are time-consuming and labor-intensive and have not eliminated the need for simpler and more economical methods of routine testing. Furthermore, several factors can influence the run-to-run, as well as the laboratory-to-laboratory, variability of the test, including variations in the composition and pH of the test medium, variations in the preparation of antifungal drug dilutions, and the concentration of the inoculum (13, 17).

To address the need for simpler and more efficient testing methods, various commercial methods have been developed and evaluated including colorimetric broth microdilution methods (3, 12, 20), breakpoint testing methods (2, 21), and agar diffusion methods (4, 23). Likewise, the PASCO Division of Becton-Dickinson has developed a commercially available broth microdilution panel for the in vitro susceptibility testing of antibacterial agents which has been evaluated on a number of occasions (10, 19, 22).

In this study, we collaborated with Becton-Dickinson to develop and evaluate the first prototype PASCO antifungal susceptibility testing panel. For prototype testing, each panel contained eight antifungal agents (amphotericin B, flucytosine, fluconazole, ketoconazole, itraconazole, clotrimazole, miconazole, and terconazole) and was prepared in advance by Becton-Dickinson and contained 10 serial two fold dilutions of each agent frozen in broth, together with positive and negative control wells. Advance preparation of the plates eliminated the need to prepare media and drug dilutions in-house and to pipette prepared dilutions into individual microtiter plate wells. An additional advantage of the PASCO system included quality control testing of each lot of plates before shipment to our laboratory to reduce run-to-run variability. The PASCO system also uses a novel one-step inoculation system which permits the simultaneous inoculation of all wells of a plate in a single step. This feature reduces the setup time and pipetting errors as well as the potential for cross-contamination of microtitration wells. The ease of use of the PASCO method, along with the potential to eliminate laboratory-to-laboratory variability caused by in-house preparation of media and plates and the flexibility of testing up to eight antifungal agents on a single plate, prompted a side-by-side comparison of the PASCO antifungal system with the standard NCCLS M27-A broth microdilution method for the testing of the antifungal drug susceptibilities of yeasts. The results of this comparison are reported here.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Isolates. Seventy-four recent clinical isolates (14 Candida albicans, 10 Candida glabrata, 10 Candida tropicalis, 10 Candida krusei, 10 Candida dubliniensis, 10 Candida parapsilosis, and 10 Cryptococcus neoformans isolates) obtained from the oral, vaginal, or rectal mucosa or from the cerebrospinal fluid, urine, or bloodstream of human immunodeficiency virus (HIV)-positive or -negative patients from diverse geographic regions were tested. Two reference strains, C. parapsilosis ATCC 99018 and C. krusei ATCC 6258, were included on each day of testing for purposes of quality control. Isolates were subcultured twice on Sabouraud dextrose agar plates (BBL, Cockeysville, Md.) and were incubated at 35°C for 24 h to ensure optimal growth prior to testing.

NCCLS M27-A broth microdilution method. Testing was performed according to the guidelines of NCCLS document M27-A (9). Analytical grade powders of the eight antifungal agents to be tested either were purchased as authentic powders (ketoconazole and itraconazole; Research Diagnostics, Inc., Flanders, N.J.) or were received as gifts from their respective manufacturers (fluconazole from Pfizer, Inc., Groton, Conn.; clotrimazole from Schering-Plough Research Institute, Kenilworth, N.J.; terconazole from R. W. Johnson Pharmaceutical Research Institute, Raritan, N.J.; flucytosine from Hoffmann-La Roche, Inc., Nutley, N.J.; and amphotericin B from Bristol Myers-Squibb Co., Princeton, N.J.). Stock solutions of fluconazole and flucytosine were prepared in distilled water, and those of amphotericin B, itraconazole, ketoconazole, miconazole, clotrimazole, and terconazole were prepared in 100% dimethyl sulfoxide. Stock solutions of antifungal agents (at 100 times the highest concentration tested) were then diluted with RPMI 1640 medium (with L-glutamine but without bicarbonate; Sigma Chemical Co., St. Louis, Mo.) buffered to pH 7.0 with 0.165 M morpholinopropanesulfonic acid (MOPS; Sigma). The final concentration ranges used were 0.03 to 16 µg/ml for amphotericin B, itraconazole, ketoconazole, miconazole, clotrimazole, and terconazole and 0.125 to 64 µg/ml for fluconazole and flucytosine.

PASCO broth microdilution method. PASCO broth microdilution plates were manufactured by the Pasco Division of Becton-Dickinson (Wheatridge, Colo.) to replicate the specifications for the NCCLS M27-A broth microdilution method (9) with the same antifungal drug powders, solvents, drug concentration ranges, and medium. PASCO microtitration plates were prepared in batches of 100 and were frozen at 70°C. The prepared plates were shipped frozen from the manufacturing plant to our laboratory and were stored at 70°C until used. The plates remained active after storage at 70°C for up to 1 year (unpublished observation). Each PASCO plate contained 10 serial dilutions of eight antifungal drugs arranged in rows as well as a growth control well (without drug) and a purity control well which contained medium but no yeasts.

Analysis of results. MICs determined by the PASCO method were compared with the MICs determined by the NCCLS M27-A method, and both on-scale and off-scale results were included in the analysis. The high off-scale results were converted to the next highest concentration, and the low off-scale MICs were left unchanged. The percent agreement between the methods was defined as the proportion of PASCO MIC results that were within ± 2 twofold serial dilutions of the M27-A reference MIC results. Interpretive breakpoints have been proposed by NCCLS for three of the eight antifungal agents tested in this investigation (9). According to the NCCLS criteria, isolates for which fluconazole MICs are 8 µg/ml are classified as susceptible, while those for which MICs are 64 µg/ml are classified as resistant. Isolates for which MICs are 16 to 32 µg/ml are termed susceptible-dose dependent (S-DD). Isolates for which itraconazole MICs are 0.125 µg/ml are classified as susceptible, while those for which MICs are 0.25 to 0.5 µg/ml are classified as S-DD, and those for which MICs are 1 µg/ml are classified as resistant. Isolates for which flucytosine MICs are 4 µg/ml are classified as susceptible, those for which MICs are 8 to 16 µg/ml are classified as intermediate, and those for which MICs are 32 µg/ml are classified as resistant. To permit comparison between the results of the two methods, the NCCLS breakpoints were applied to both tests for these agents. Major errors were defined as results in which the reference method result was susceptible and the PASCO method result was resistant, while very major errors were defined as results in which the reference method result was resistant and the PASCO method result was susceptible (7). Minor errors were defined as variations in results from resistant to S-DD or S-DD to susceptible between the two methods.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

In vitro susceptibilities of yeast isolates to antifungal agents as determined by PASCO and NCCLS M27-A broth microdilution methods. Table 1 summarizes the in vitro susceptibilities of the 74 yeast isolates tested to eight antifungal agents determined by the PASCO and NCCLS M27-A broth microdilution methods. The data are reported as MIC ranges and the MICs at which 50 and 90% of the isolates are inhibited (MIC₅₀s and MIC₉₀s, respectively). A broad range of MICs was observed for all agents tested. With the exceptions of amphotericin B for all isolates tested and all drugs except fluconazole for C. tropicalis, the PASCO method gave generally higher MIC₅₀s and MIC₉₀s than the NCCLS method. However, the differences between methods were within 1 to 2 serial twofold drug dilutions for greater than 80% of all drug-isolate combinations. MIC₅₀ and MIC₉₀ differences between methods which were greater than two dilutions were most prevalent for C. albicans when it was tested against the azole agents (Table 1). In each experiment, the MICs of the antifungal agents for the two quality control strains were within the accepted limits as established by the NCCLS M27-A method (data not shown), demonstrating that both methods were reproducible.

Agreement between PASCO and NCCLS M27-A broth microdilution antifungal drug susceptibility test methods. Table 2 compares the agreement between the MIC results for the two broth microdilution methods. The overall agreements between the two methods for all 74 yeast isolates tested were 91% for fluconazole, 89% for amphotericin B and ketoconazole, 85% for itraconazole, 80% for flucytosine, 77% for terconazole, 66% for miconazole, and 53% for clotrimazole. The overall agreement between the two methods for all eight antifungal agents tested was 80% for four of the seven yeast spp. tested (C. glabrata, C. krusei, C. dubliniensis, and C. neoformans). Comparisons of test results for miconazole and clotrimazole gave the lowest percent agreement (66 and 53%, respectively), and thus, when the data for these two agents were removed from the analysis, comparisons of test results for all seven species tested against the remaining six antifungal agents resulted in 80% agreement (more specifically, 85% agreement). Poor agreement (30%) with terconazole was observed only for C. krusei isolates. The highest overall agreement was observed for all drugs against C. neoformans (90%), whereas the lowest overall agreement among the seven species tested was observed for C. tropicalis and C. albicans (66 and 69%, respectively). However, when the data for miconazole and clotrimazole were removed from the analysis, agreement was improved to 80% for both species. Interestingly, the agreement between the two methods for the azole agents was considerably higher for the newly described Candida sp., C. dubliniensis, than for its close phylogenetic and phenotypic relative, C. albicans (Table 2).

Discrepancies between test methods by antifungal agent and by organism. Table 3 details the susceptibility interpretation discrepancies between the two methods for the three drugs with NCCLS-defined breakpoints (flucytosine, fluconazole, and itraconazole). Very major errors, in which the reference method classified an isolate as resistant and the PASCO method classified it as susceptible, occurred for only three drug-organism pairs (1.3%): fluconazole with one C. albicans isolate and one C. tropicalis isolate and flucytosine with one C. neoformans isolate (Table 3). Major errors in which the reference method classified the isolates as susceptible and the PASCO method classified them as resistant occurred for 12 organism-drug pairs (5.4%). The greatest number of major errors (7 of the 12) occurred when C. albicans isolates were tested with azole drugs (Table 3). The greatest percentage of minor errors was noted for itraconazole (five of the seven instances), and the MIC determined by the PASCO method was higher in four of the five cases (Table 3). The remaining two minor errors were observed for flucytosine against C. neoformans, in which the MIC by the reference method was interpreted as susceptible and the MIC by the PASCO method was interpreted as S-DD (Table 3). Overall, the results indicated that there were only 15 (6.7%) major or very major errors and 7 (3%) minor errors between methods for all organisms tested against flucytosine, fluconazole, and itraconazole, and the MICs determined by the PASCO method were generally higher in these cases.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The present evaluation was the first to compare the PASCO broth microdilution antifungal susceptibility test panel to the reference NCCLS M27-A method. The PASCO method differs from the reference procedure in the method of plate inoculation in that it uses a novel one-step procedure which simultaneously transfers organisms equally to each well. Other facets of the PASCO procedure, including composition of the medium, drug dilutions, incubation time and temperature, and criteria of endpoint determination were identical to those of the reference procedure. Although individual laboratories can make antifungal panels in batches and freeze them in advance, the commercial production of the PASCO panels eliminates the need for individual laboratories to acquire antifungal drugs, prepare media, make drug dilutions, and fill plates, thereby reducing the possibility of error at each of these steps and theoretically increasing the reproducibility of test results from run to run as well as from laboratory to laboratory. Furthermore, a commercially available system is highly desirable to the clinical microbiological laboratory in which a rapid and reproducible testing method that produces reliable results is needed.

Our analysis of the PASCO antifungal susceptibility testing system showed that the overall agreement between the two methods was 80% for five of the eight drugs tested (amphotericin B, flucytosine, fluconazole, ketoconazole, and itraconazole). The greatest number of major and very major errors (9 of 15) occurred when C. albicans or C. tropicalis isolates were tested against the azole drugs. C. albicans and C. tropicalis are the most common Candida spp. to exhibit trailing growth, and therefore, the observed discrepancies in MIC results between the two test methods were likely due to the difficulties of consistently reading visual MIC endpoints for isolates which produce trailing growth in the presence of azole antifungal agents. In addition, there was better agreement between the methods for C. dubliniensis than for C. albicans when they were tested with the azole antifungal agents. Interestingly, although C. dubliniensis is almost identical to C. albicans by most phenotypic characteristics, we observed no trailing growth of this organism in any azole agent tested by either method. While we have no explanation for this observation, perhaps the distinction may provide additional insight into differences between these two species.

Agreement of the PASCO method with the NCCLS M27-A method was unsatisfactory only for clotrimazole and miconazole (53 and 66%, respectively) and terconazole with C. krusei isolates (30%). A possible explanation for the diminished agreement with these agents is the reduced solubility of the drugs in the testing medium which may interfere with the reproducibility of the MIC results. The NCCLS M27-A document does not include these antifungal agents, and thus, it may be that modifications of both methods are necessary to achieve meaningful results for these drugs. Clotrimazole, miconazole, and terconazole were all included in the prototype antifungal drug panel to facilitate antifungal drug susceptibility testing of vaginal isolates to be tested as part of a multicenter study of HIV-positive patients with vulvovaginal candidiasis (B. A. Arthington-Skaggs, T. Desai, Y. Ankrah, and C. J. Morrison, Abstr. 38th Intersci. Conf. Antimicrob. Agents Chemother., abstr. J-123, p. 486, 1998). Since the NCCLS reference method has not been optimized for the testing of these agents, it is possible that the higher MIC results for these agents observed by the PASCO method may not be due to a technical problem but may better reflect previous azole use by the HIV-positive patient populations from which the majority of yeast isolates were obtained for this study. Whether or not the MICs determined by the PASCO method are more clinically relevant for these antifungal agents remains to be determined.

Other differences between the two methods included higher MICs by the reference method compared to the MICs by the PASCO method for amphotericin B for most species tested and for C. tropicalis for all agents tested except fluconazole, although the majority of differences were within the acceptable variation of ±2 two-fold serial dilutions. In the case of amphotericin B, this discrepancy may be because by the PASCO method the plates are read from the tops of the wells by use of a light box designed to accommodate the larger PASCO microtiter plate, whereas by the reference method the plates are read from the bottoms of the wells with the aid of a magnifying mirror. This technical variation may cause the reader to miss minute amounts of growth in the PASCO wells which may be detected in the reference plates. For C. tropicalis, because the MICs obtained by the reference method were uniformly higher than those obtained by the PASCO method (except for fluconazole), it is speculated that the PASCO inoculator may have inaccurately delivered a lower initial inoculum to the microtiter plate wells compared to the inoculum delivered by the reference method. This could occur if the PASCO inoculator tips, which fill with 5 µl of cell suspension by capillary action, were affected by some unknown surface characteristics of C. tropicalis. Further analyses are needed to fully understand this phenomenon.

The establishment, by NCCLS, of a standardized broth macrodilution reference method for in vitro susceptibility testing of antifungal agents against Candida spp. and C. neoformans has enabled better analysis of the in vitro correlation of MIC data with clinical outcome and has permitted interpretive breakpoints to be proposed for flucytosine, fluconazole, and itraconazole (9). Tests based on the more convenient microdilution format have also been introduced and evaluated, and published comparisons have demonstrated good agreement between the original NCCLS broth macrodilution method and microdilution adaptations of it (1, 5). However, for these methods to be useful, the results should provide a reliable prediction of the response to treatment for humans with infections. In particular, a high MIC should correlate with therapeutic failure (16). Numerous reports have demonstrated that the ability of antifungal drug susceptibility testing to predict clinical outcome differs from agent to agent and depends on the patient population studied (6). For instance, high fluconazole MICs are often predictive of therapeutic failure in HIV-positive patients with oral candidiasis (14, 18) but do not necessarily correlate with the clinical outcome in patients with candidemia (15). The predictive value for other antifungal agents is even less clear, but a number of investigations have reported that high or rising amphotericin B MICs for isolates of Candida spp. recovered during prolonged treatment with this agent correlate with therapeutic failure (8, 11).

Previous evaluations of commercially available antifungal drug susceptibility test methods have included colorimetric broth microdilution methods (Sensititre, Westlake, Ohio, and ASTY, Kyokuto Pharmaceutical Industrial Co., Ltd.) (3, 12, 20), a noncolorimetric broth breakpoint testing system (FUNGITEST; Sanofi Diagnostics Pasteur, Paris, France) (2, 21), and an agar diffusion system (E-test; AB Biodisk, Solna, Sweden) (4, 23). The results suggest that the Sensititre and ASTY colorimetric broth microdilution methods are suitable alternatives for routine antifungal susceptibility testing of Candida spp. and C. neoformans, showing excellent agreement (80%) with the NCCLS reference method for amphotericin B, flucytosine, fluconazole, and itraconazole (3, 12, 20). The FUNGITEST breakpoint testing system is a simplified method in which each drug (amphotericin B, flucytosine, fluconazole, ketoconazole, and itraconazole) is tested at only two concentrations to distinguish resistant isolates from susceptible ones. While this test offers attractive features to the busy clinical microbiology laboratory, comparative evaluations with the NCCLS reference method have produced equivocal results. One study found FUNGITEST to be an unacceptable alternative method on the basis of its misclassification of azole-S-DD and azole-resistant isolates of several Candida spp. as susceptible (2). A second evaluation reported that the MICs determined with the FUNGITEST were in good agreement with the MICs determined by the NCCLS M27-A method, ranging from 100% agreement for amphotericin B to 76.7% agreement for itraconazole when testing was done with Candida spp. and C. neoformans (21). The agar dilution-based format of the E-test system has been shown to be suitable for routine use with Candida spp. tested with amphotericin B or flucytosine but has been shown to be less reliable for the azoles and isolates that appear to demonstrate acquired resistance (23).

Each of these commercial methods reduces the time and labor needed to perform antifungal susceptibility testing by providing microtitration plates prefilled with media and drugs at the proper concentration or, in the case of the E-test, impregnated into ready-to-use gradient strips. These steps increase test reproducibility by providing a single source of quality-controlled materials for use by different laboratories. The PASCO method offers these same advantages as well as the additional advantages of flexibility for testing of up to eight antifungal drugs on a single plate, a unique single-step inoculation system which significantly reduces the setup time, and MIC results which are in good agreement with those of the NCCLS M27-A broth microdilution method for most of the commonly prescribed antifungal drugs tested against the most medically important yeasts. With the exception of miconazole and clotrimazole and of terconazole with C. krusei isolates, the agreement (80%) between the PASCO MICs and the reference MICs was comparable to that achieved by other commercial tests evaluated against the M27-A reference method. Results of our work presented here suggest that until NCCLS MIC ranges are established for in vitro testing of terconazole, miconazole, and clotrimazole, a comparison of the agreement between the reference method and the PASCO method for these drugs is difficult to interpret. Therefore, at this time, it is recommended either that these drugs be removed from the PASCO panel and replaced with other agents or that current panels be tested with additional isolates under different conditions to improve the agreement with the reference method.

In conclusion, this investigation has demonstrated that the PASCO method is a suitable alternative to the NCCLS M27-A broth microdilution method for the testing of amphotericin B, flucytosine, fluconazole, ketoconazole, itraconazole, and terconazole against a variety of yeasts. It is less labor-intensive and much simpler to perform than the NCCLS method. Further work is needed to assess whether the in vitro PASCO test results for clotrimazole and miconazole correlate better with clinical outcome than those obtained by the NCCLS M27-A method. The PASCO method is as capable of detecting azole-resistant isolates as the NCCLS method but, in general, gave MICs one or two concentrations higher than those obtained by the NCCLS method. Our results suggest that the PASCO method is suitable for use in the routine testing of the susceptibilities of Candida spp. and C. neoformans isolates to commonly prescribed antifungal drugs currently in clinical use.