Interlaboratory Variability of Caspofungin MICs for Candida spp. Using CLSI and EUCAST Methods: Should the Clinical Laboratory Be Testing This Agent?

Isolates.

Echinocandin susceptibility testing of unique clinical isolates of Candida spp. was performed according to the CLSI broth microdilution method at the following medical centers: Virginia Commonwealth University (VCU) Medical Center, Richmond, VA; JMI Laboratories and the University of Iowa, Iowa City, IA; Adolfo Lutz Institute, Araçatuba City, Brazil; Instituto de Medicina Tropical Alexander von Humboldt, Universidad Peruana Cayetano Heredia, Lima, Peru; Unidad de Microbiologia Experimental, Centro de Investigacion, Hospital Universitario La Fe, Valencia, Spain; University of Texas Health Science Center, San Antonio, TX; The University of Alberta, Edmonton, Alberta, Canada; Universidad Autónoma de Nuevo León, Monterrey, Nuevo León, México; Facultat de Medicina, IISPV, URV, Reus, Spain; The Innsbruck Medical University, Division of Hygiene and Medical Microbiology, Innsbruck, Austria; Mycotic Diseases Branch, Centers for Disease Control and Prevention, Atlanta, GA; Canisius Wilhelmina Hospital, Nijmegen, The Netherlands; University of Texas Health Science Center, Houston, TX; Hospital General Universitario Gregorio Marañón, Facultad de Medicina, Universidad Complutense, Madrid, Spain; Fungal Taxonomy Laboratories, Adolfo Lutz Institute, São Paulo City, Brazil; Laboratoire de Santé Publique du Québec, Québec, Canada; and the Mycology Department, Adolfo Lutz Institute, São Paulo, Brazil. These laboratories were assigned codes (numbers 1 to 20 in Tables 1 and 2).

Similarly, EUCAST MICs were determined at the following medical centers: Statens Serum Institut, Copenhagen, Denmark; Karolinska University Hospital, Stockholm, Sweden; Instituto de Salud Carlos III, Madrid, Spain; Institut Pasteur, Centre National de Reference Mycoses and Antifongiques, Paris, France; Rikshospitalet University Hospital, Oslo, Norway; University Hospital of South Manchester, Manchester, United Kingdom; and The Innsbruck Medical University, Innsbruck, Austria. These laboratories were assigned codes (numbers 1 to 7 in Tables 3 and 4).

The total of available CLSI caspofungin MICs for each species was 11,550 for C. albicans (9,655 anidulafungin MICs), 145 for C. dubliniensis, 5,322 for C. glabrata, 262 for C. guilliermondii, 850 for C. krusei, 400 for C. lusitaniae, 4,880 for C. parapsilosis, and 2,958 for C. tropicalis. The total of available EUCAST caspofungin MICs for each species was 2,556 for C. albicans, 948 for C. glabrata, 680 for C. parapsilosis, and 403 for C. tropicalis. Isolates were identified and stored at each medical center using standardized methodologies (13), and they were not evaluated for FKS mutations. In addition, CLSI and EUCAST MICs for one or both quality control (QC) isolates, C. parapsilosis ATCC 22019 and C. krusei ATCC 6258, were obtained each time that testing was performed in each of the participant laboratories (4, 5).

Antifungal susceptibility testing.

As defined in the CLSI and EUCAST broth microdilution standards documents (M27-A3 and EDef 7.2), inclusion criteria for CLSI-derived echinocandin MICs from the participant laboratories required the use of RPMI 1640 broth with 0.2% dextrose, incubation temperatures at 35°C for 24 h, and a prominent inhibition of growth (≥50% reduction relative to the growth control). For EUCAST-derived MICs, the required conditions included RPMI 1640 broth with 2% dextrose, incubation temperatures at 35°C for 24 h, and a spectrophotometric prominent inhibition of growth (optical density [OD] reduction of ≥50%) (4, 7).

In addition, CLSI and EUCAST MICs for QC isolates were obtained following the respective standard conditions.

Survey.

To further investigate the possible causes of caspofungin MIC variability, a questionnaire was sent to the laboratories that provided the CLSI data depicted in Tables 1 and 2. The questions were the following: (i) What was the source of the drug? (ii) At what temperature was the drug stored? (iii) Over what time period was the batch of powder used? (iv) What was the solvent used for preparing stock and subsequent dilutions? (v) Was the medium always the CLSI RPMI medium with 0.2% dextrose? (vi) Were the MICs always read at 24 h? (vii) Were MICs for 50% or more growth inhibition? (viii) Were any of the data provided from YeastOne, Etest, or EUCAST methods? (ix) Were the microdilution trays read visually, or was a spectrophotometer used?

Data analysis.

The MIC distributions of each of the species tested in each participating laboratory were first screened for evidence of grossly abnormal distributions, and the presumptive WT modal (most frequent value) MICs were determined (12). In general, for ECV determination, grossly skewed distributions (distributions which had a modal MIC at the lowest or highest concentration tested and/or were bimodal in the presumptive wild-type distribution) should be excluded (CLSI Workshop on Establishing Epidemiological Cutoff Values, Tampa, FL, 22 January 2013), but they were not for the present report. In addition, we documented data only for the less common species when the CLSI MICs originated from a minimum of four laboratories and when the distribution from each laboratory comprised at least 10 MICs for each species.

CLSI caspofungin MIC variability.

In the present study, we observed significant interlaboratory variation in both (i) CLSI caspofungin MIC ranges and (ii) WT modes for most species evaluated. For example, we observed caspofungin MIC ranges of 0.0079 to 0.25 μg/ml and of 0.25 to 2 μg/ml for C. albicans (Table 1) and WT modes spanning five to six dilution steps among most of the species (0.016 to 0.5 μg/ml for C. albicans and C. tropicalis; 0.031 to 0.5 μg/ml for C. glabrata; 0.063 to 1 μg/ml for C. krusei). Wide modal ranges were also present among C. dubliniensis and C. lusitaniae distributions, but this was not the case for C. parapsilosis and C. guilliermondii (mostly 0.5 or 1 μg/ml and 0.25 or 0.5 μg/ml, respectively) (Table 2). On the other hand, minimal interlaboratory variation was observed in the anidulafungin comparative distribution for C. albicans (modes, 0.016 and 0.031 μg/ml) (Table 1), suggesting that the variability is caspofungin specific and not a general echinocandin problem. EUCAST caspofungin MIC variability corroborated CLSI data, where individual laboratories reported MIC ranges of 0.0079 to 0.125 μg/ml and 0.125 to 4 μg/ml for C. albicans (Table 3). EUCAST caspofungin modes were also variable for three of the four common species (0.016 to 0.5 μg/ml for C. albicans; 0.063 to 0.5 μg/ml for C. glabrata; and 0.031 to 0.5 μg/ml for C. tropicalis), while C. parapsilosis modes were constant between participant sites (either 1 or 2 μg/ml) (Table 4). Insufficient anidulafungin EUCAST MIC distributions precluded relevant comparisons between the two agents. It is currently not understood why variability is not a problem among caspofungin MIC distributions for C. parapsilosis or C. guilliermondii since these data were gathered in the same laboratories using the same method as that for the other species. Perhaps the stability of caspofungin MICs for these two species may be due to the higher concentrations of drug in the well and higher MICs. This could also explain why the CLSI modal MICs were in general lower at laboratories using dimethyl sulfoxide (DMSO) as the solvent (Tables 1 and 2), which improves solubility. More studies are needed in order to resolve these conflicting results.

Caspofungin MIC interlaboratory variability was evident as early as 2002 with reported MIC ranges for C. albicans as low as 0.031 to 1 μg/ml and as high as 0.25 to 4 μg/ml; similar discrepancies were observed for C. tropicalis (14). Earlier caspofungin range variability was attributed to the different panels of isolates that were evaluated in each study, but it does not seem plausible with respect to the use of WT strains. Considerable interlaboratory variability of caspofungin MIC data was also reported in the original study that identified the optimum testing conditions for this agent and Candida spp. (overall agreement, 64.7 to 100%; agreement per species was 84 to 100% for C. albicans, 85.3% for C. glabrata, 41.4 to 85% for C. parapsilosis, and 41 to 85% for C. tropicalis; species data were not included in that report) (14). These authors concluded that interlaboratory agreement needed to be improved, which was corroborated in the present study. Caspofungin MIC variability has been previously reported by the EUCAST group as being dependent upon drug lot-to-lot inconsistency in potency and/or stability (10). Using different caspofungin lots, MICs were repeatedly determined for six ATCC strains resulting in wide caspofungin MIC ranges (e.g., 0.016 to 0.25 μg/ml for the same strain of C. albicans); results were similar for control isolates of the other species evaluated, with the exception of MICs for the C. krusei ATCC 6258 (0.25 to 1 μg/ml) and C. parapsilosis ATCC 22019 (1 to 2 μg/ml) isolates.

The agreement was acceptable (96 to 100%) in most laboratories regarding the CLSI recommended caspofungin MIC range (0.25 to 1 μg/ml) for the QC C. parapsilosis ATCC 22019 strain (5) but unacceptable in three laboratories (84 to 94%); most discrepancies occurred with MICs above the high end of the QC range, 2 μg/ml. Modes ranged from 0.25 to 1 μg/ml in all laboratories except one. There was more caspofungin modal variability for the QC isolate C. krusei ATCC 6258 (mode range, 0.125 to 1 μg/ml), but the agreement was acceptable (95 to 100%) in most laboratories regarding the recommended range of 0.125 to 1 μg/ml (5); the exceptions were MIC ranges in two laboratories (<0.125 to 2 μg/ml). The less pronounced overall variation for these two QC strains than for the clinical isolates suggests that these two control strains may not be sensitive enough to detect performance variations when this agent is tested. Similar concerns have been raised by EUCAST (10). The CLSI plans to address this problem in the near future by collecting new data and using DMSO as the solvent for the selection of more appropriate control strains (15).

Caspofungin powder.

All 17 laboratories used caspofungin powder sourced from Merck in the United States and other countries; 7 of the 17 laboratories (laboratories 4, 7, 10, 14, 15, 16, and 17) used a single lot. Discrepancies were noted in the powder storage temperature, where some laboratories stored their powder at either 4° to 5°C (laboratories 7, 15, 16, and 17) or −20°C (laboratories 6 and 14) and the other 11 laboratories stored their powder at −70°C or below. This variation in storage conditions did not appear to explain the variations in MICs among the participating laboratories (Table 2). While most laboratories used their batch of powder for ≤1 year, laboratories where data were gathered throughout ≥7 years used only two or three lots of caspofungin (laboratories 2, 3, 19, 20); Merck's recommendation is 2 years. However, overall modes from the latter group of laboratories were among the lower values; e.g., the modal MICs for C. albicans were 0.016 and 0.031 μg/ml (Tables 1 and 2). In addition, while we observed some bimodal distributions, especially for C. albicans, this type of abnormal distribution was observed only in data from laboratory 3 and not from laboratory 2, 19, or 20. Therefore, using the powder longer than 2 years, or any other related powder issues, did not explain the problem. Unfortunately, sufficient data on drug lot numbers were not available, so we were unable to determine that lot-to-lot variability caused the problem, as previously reported (10).

Solvents for preparation of caspofungin stock solutions.

Until 2012, the recommendation of the CLSI was the use of water as the solvent for the preparation of caspofungin stock solutions (16). Four (laboratories 4, 17, 18, and 20) of the 17 laboratories used DMSO as the solvent, and laboratory 2 used both solvents. As shown in Tables 1 and 2, MICs derived using water versus DMSO as the solvent could not account for all of the variability in the modes. It should be noted, however, that laboratories using DMSO as the solvent tended to have MIC distributions with lower modal MIC values for each species than the laboratories using water as the solvent, suggesting that DMSO helps in avoiding loss of potency and higher MICs. Therefore, although the use of DMSO versus water is not the sole explanation for the variability shown among the participating laboratories, it appears to play a role. Accordingly, the new and revised version of the document (M27-S4) lists DMSO as the solvent for echinocandins and fluconazole (5); EUCAST also strongly recommends the same for echinocandin susceptibility testing.

Other testing parameters.

Although each contributing laboratory was requested to provide only CLSI data and not a mixture of data from different methodologies (commercial methods included), our survey included specific questions (e.g., medium dextrose content, spectrophotometric versus visual MICs, and 50% or more growth inhibition). We received caspofungin MIC data from 19 laboratories; however, based on answers to the survey, data from 2 laboratories (8 and 13 used EUCAST medium) were not included in this report. Caspofungin MICs from the remaining 17 laboratories were determined according to the recommended CLSI testing conditions listed above (see “Antifungal susceptibility testing”); MICs were read visually in 15 of the 17 laboratories, but variation was similar to that for the EUCAST data (all automated spectrophotometer readings), suggesting that subjectivity in endpoint reading was not the source of variability.

New caspofungin CBPs and MIC interpretation.

Before 2004, caspofungin MICs for Candida spp. were obtained using a variety of testing conditions (17). Between 2008 and 2009, standard guidelines for caspofungin MIC determination (50% or more growth inhibition at 24 h) were described in CLSI documents M27-A3 and M27-S3 (4, 16). In 2008, the CLSI established a susceptible CBP (≤2 μg/ml) for echinocandins and all Candida spp. (16). Molecular studies identified the echinocandin antifungal activity target, the protein Fksp, glucan synthase, encoded by three FKS genes. In addition, elevated echinocandin MICs for clinical strains of Candida were associated with specific mutations in FKS genes and therapeutic failure or breakthrough infection (18–21). Using FKS1 mutant strains, Garcia-Effron et al. (22, 23) demonstrated that although caspofungin MICs of >2 μg/ml captured almost 100% of FKS mutant strains, the MICs that differentiated these mutants from WT strains were lower (>0.5 μg/ml for C. albicans and >0.25 μg/ml for C. glabrata) for both anidulafungin and micafungin. By 2011, caspofungin species-specific ECVs were defined, and CBPs were adjusted for six Candida spp.: C. glabrata (susceptible, ≤0.125 μg/ml; resistant, ≥0.5 μg/ml); C. albicans, C. tropicalis, and C. krusei (susceptible, ≤0.25 μg/ml and resistant ≥1 μg/ml); and C. parapsilosis and C. guilliermondii (susceptible, ≤2 μg/ml; resistant, ≥8 μg/ml) (5, 6, 11).

What is the impact in the clinical setting of new and lower species-specific CLSI caspofungin breakpoints? They have created a problem in some laboratories. As shown in Tables 1 and 2 (also according to some contributors' comments), the variability in caspofungin MIC distributions will result in reporting too many WT strains as nonsusceptible (or major errors) for most of the common species with the exception of C. parapsilosis. EUCAST has already recommended the use of anidulafungin MICs as markers for the echinocandins and avoiding the use of caspofungin MIC results for clinical decision making (8, 10). It is expected that the CLSI will reach a similar conclusion and, in addition to the projected revision of MIC ranges for the QC isolates and the selection of more adequate strains, that the issue of caspofungin variability will be addressed by this organization in the near future. CBPs are the same for all three echinocandins and each of the common species except for the lower endpoint for C. glabrata and micafungin. Therefore, for the time being, either micafungin or anidulafungin MIC endpoints can be used as predictors of susceptibility or resistance of Candida spp. to caspofungin.

In summary, something unusual is going on with caspofungin testing. Although we have a large number of caspofungin MICs from different geographical areas, it is evident that caspofungin ECVs cannot be defined using available MIC results. Routine testing or reporting of CLSI caspofungin MICs for Candida is not recommended, especially in the clinical setting; in the meantime, testing one of the other two echinocandins (anidulafungin or micafungin) could provide the desired susceptibility result. Available ECVs have been established based on data from a single laboratory, and interpretation of MICs using the new CBPs or ECVs will result in some laboratories reporting too many isolates as resistant or non-WT isolates, mainly among C. glabrata, C. krusei, and, to a certain extent, among C. albicans and C. tropicalis, which could result in less effective therapy.