We included patients seen at MD Anderson Cancer Center from March 2005 through September 2013, for whom ≥1 blood culture was positive for C. glabrata and who had symptoms, signs, or laboratory findings consistent with infection. We retrospectively reviewed electronic medical records for demographic, clinical, and laboratory data for the day of candidemia (defined as day of blood collection for culture), and we reviewed pharmacy records and clinical notes for previous use of antifungal drugs and cumulative doses. The study was approved by the MD Anderson Cancer Center institutional review board.

Isolation and identification of C. glabrata isolates in blood culture were performed by using standard microbiological procedures (2–4). We determined MICs of fluconazole, voriconazole, posaconazole, caspofungin, and amphotericin B by using the broth microdilution method described in CLSI M27-A documents (10,16), according to prominent (50%) reduction in turbidity and 100% growth inhibition for amphotericin B. In a subset of caspofungin-resistant isolates, we also tested in vitro susceptibility to micafungin and anidulafungin.

Susceptibility to antifungal drugs was defined according to clinical break points for C. glabrata (10,11); isolates for which fluconazole MIC was ≤32 mg/L were considered dose dependent, whereas those for which MIC was ≥64 mg/L were considered resistant. Because clinical break points for voriconazole and posaconazole are undefined, C. glabrata isolates in which MIC was >1 dilution above the epidemiologic cutoff values (0.5 and 2.0 mg/L, respectively) were considered potentially resistant. Caspofungin/anidulafungin and micafungin resistance was defined as MICs ≥0.5 mg/L and ≥0.25 mg/L, respectively (10,11). Strains for which caspofungin or anidulafungin MIC was 0.25 mg/L or micafungin MIC was 0.125 mg/L were classified as intermediate. Resistance to amphotericin B was defined as MIC ≥2 mg/L (10,11).

Continuous variables were compared by using the Student t-test or the Mann-Whitney U criterion for variables that were not normally distributed. Categorical variables were compared by using the χ² test, Fisher exact test, and linear-by-linear associations for trend. Binary and ordinal (after testing the parallel lines assumption) logistic regression analyses were used to identify variables independently associated with fluconazole, caspofungin, and multidrug resistance. Survival curves were compared by using the log-rank test and Cox regression analysis. The proportional hazards assumption was tested graphically and by building time-dependent variables. For univariate analyses, clinically relevant parameters (p<0.1) were included at model entry. Variables were retained in the final model if p<0.05; p values >0.05 but <0.1 were noted as indicating trends. All analyses were performed by using SPSS, version 21 (IBM Corporation, Chicago, IL, USA).

In this contemporary series of cancer patients with C. glabrata fungemia, the rates of in vitro caspofungin resistance and multidrug resistance are among the highest reported to date. By comparing the updated (10) with the previous, non–species-specific CLSI definitions of in vitro susceptibility (16), we found that 90% of caspofungin-intermediate or -resistant C. glabrata bloodstream isolates would have been previously classified as susceptible. Caspofungin resistance was associated with previous exposure to echinocandins, use of TPN, and all-cause mortality rate.

Contrary to previous findings from our institution (3), most patients with C. glabrata fungemia in the series reported here had solid tumors rather than hematologic malignancies. One third of fluconazole-resistant isolates and half of those with decreased susceptibility to caspofungin were isolated from patients with solid malignancies. These results probably reflect an overall increase in solid tumors; however, our findings also confirm that C. glabrata bloodstream infections have become major clinical problems among all patients at risk for candidemia (6,9,14,15,17).

In agreement with previously reported findings, our study indicated that broad use of azoles—mainly voriconazole—and echinocandins was strongly associated with C. glabrata fluconazole and caspofungin resistance (3,5,6,14,15). In our study, 11 C. glabrata isolates were classified as resistant without having had any previous documented exposure to the respective classes of antifungal drugs. This finding is in agreement with a recent report of isolation of 4 C. glabrata FKS mutants from patients who had not received echinocandins (17). Because several factors place cancer patients at risk for candidemia and clinical failure of antifungal drugs (1–5), we sought to identify those clinical factors associated with in vitro resistance. On the basis of our results, we consider it likely that poor host defense mechanisms associated with the presence of hematologic malignancy, myelosuppression, and critical illness are independently associated with resistance.

We also observed an independent association between TPN and caspofungin resistance or multidrug resistance. TPN is an established risk factor for candidemia and a marker of intestinal dysfunction (18). Moreover, TPN causes atrophy of the intestinal mucosa, facilitating microperforations and Candida translocation, and it is associated with thick biofilm formation and catheter-related infections (18,19). Whether our observed association between TPN and caspofungin resistance is reflective of critical illness or whether the above mechanisms also promote the development of resistance remains to be determined.

In our study, almost one third of fluconazole-resistant strains and two-thirds of caspofungin-resistant strains were multidrug resistant. These rates of cross-resistance are significantly higher than those previously reported from multi-institutional registries (20,21) and another tertiary academic hospital (6). Specifically, investigators from Duke University Hospital reported a 25% rate of fluconazole resistance over a 10-year period, which is similar to our rate of 21%. In the same report (6), the overall rate of resistance to at least 1 echinocandin was lower (6.7%) than that found in our study (10.7%), although by 2010 it had increased to 12.7% (6). In another study, 11% of C. glabrata bloodstream isolates were resistant to caspofungin and 18% had FKS mutations (17). Notably, the rates of multidrug resistance determined by the study from Duke (3.5%) (6), the Centers for Disease Control and Prevention SENTRY Antimicrobial Surveillance Program (1%) (20), and another recent multi-institutional study (1%) (21) were substantially lower than the rates of multidrug resistance determined in our study (6.8%). These data document a worrisome trend for concomitant resistance of C. glabrata clinical isolates to azoles and echinocandins, which seems to be more prominent in our population of patients with cancer.

In our study, resistance to fluconazole was highly associated with caspofungin resistance, independent of prior use of antifungal drugs; this finding is in agreement with our institution’s previously reported findings for different Candida species (22). Echinocandin, but not azole, exposure was a significant independent predictor of multidrug resistance. These findings could reflect a worrisome potential for development of multidrug resistance in C. glabrata, a versatile, haploid species (7). In a recent study, serial exposures of a C. glabrata laboratory strain to low-dose micafungin led to the development of a single-point mutation conferring multiazole and echinocandin resistance with preserved virulence (23). Moreover, in an analysis of molecular events leading to echinocandin resistance of C. glabrata isogenic isolates consecutively obtained from a patient receiving chronic TPN, a multidrug-resistant strain emerged after multiple courses of treatment with caspofungin but no previous azole exposure (8). Selective pressure from antifungal drugs, along with other factors, such as chemotherapy (24) and broad-spectrum antibacterial drugs (25), might lead to the expansion of similar phenotypes.

By applying the updated clinical break points to our patient population, we captured a strong and potentially independent correlation of all-cause mortality rates with in vitro caspofungin MICs but not with other factors classically associated with poor outcomes such as advanced age and hematologic malignancy (2,4,5). Although other residual confounders cannot be ruled out, this finding is in agreement with previously reported significant associations between clinical failure of echinocandins and elevated in vitro echinocandin MICs (6,8,14,17). In some animal studies, FKS mutations leading to echinocandin resistance were associated with decreased fitness (8,26). Nevertheless, a recent study that used an immunocompromised murine model of systemic candidiasis showed that caspofungin was ineffective against C. glabrata isolates with MIC ≥1 mg/L (27). Furthermore, investigators have also described the development of compensatory mechanisms that override the decreased virulence resulting from clinical exposure of an FKS mutant C. glabrata isolate to an echinocandin (8). Clinical (8,28) and laboratory (23) strains that exhibit high-level antifungal resistance without decreases in fitness have been described. What remains incompletely characterized are the spectrum of mutations predisposing to azole and/or echinocandin resistance, the role of epigenetic mechanisms, and the virulence of resistant (compared with susceptible) Candida strains in humans. According to our results, lowering the MIC break point for caspofungin resistance in C. glabrata bloodstream isolates to 0.5 mg/L is clinically relevant.

Our study has several limitations. It was a retrospective study performed at a single institution, and our patient population was rather small and selected. Therefore, our observations might not be applicable to different patient groups at risk for serious Candida infections. The number of caspofungin-resistant isolates was small, and we used in vitro caspofungin MIC alone to define echinocandin resistance, without molecular confirmation of underlying mutations. The interlaboratory variability in caspofungin MICs is substantial, (29,30), and there is evidence that micafungin and anidulafungin MICs correlate better with the presence of FKS mutations and clinical outcomes (15). However, testing the micafungin and anidulafungin MICs of available caspofungin-resistant isolates did not change our conclusions. Moreover, our most striking finding was the high percentage of multidrug-resistant C. glabrata isolates. In a previous study (20), 100% of such multidrug-resistant isolates had an FKS mutation; in the study reported here, all multidrug-resistant isolates that were available for testing were resistant to 2 echinocandins. Therefore, we believe that the substantial number of multidrug-resistant strains harbored molecular mechanisms of resistance. It should be noted that the reference for assessing sensitivity and specificity of in vitro MICs has been the presence of mutations within the FKS1 and FKS2 hot spot regions. Nevertheless, there is emerging evidence that non–FKS-related mechanisms might be operative or might predispose to the development of echinocandin resistance and even multidrug resistance (8,23). Recently, a high in vitro caspofungin MIC (≥0.5 mg/L) was shown (17) to have a higher positive predictive value for echinocandin failure than the presence of FKS hot spot mutations, in agreement with our findings and contrary to previously reported findings (6,31).

In summary, the rate of in vitro caspofungin and multidrug resistance of C. glabrata bloodstream isolates in our patient population is, to our knowledge, among the highest reported. Our findings might indicate a worrisome propensity of C. glabrata strains for multidrug resistance in cancer patients and should prompt awareness of the need for good stewardship of antifungal drugs. Prospective, large-scale clinical registries, with molecular data on mutations that confer resistance to antifungal drugs, are needed.