Uncommon Candida Species Fungemia among Cancer Patients, Houston, Texas, USA

Potentially resistant bloodstream isolates and increasing use of echinocandins create a need for institutional surveillance for candidemia in cancer patients.

BSIs were diagnosed at a cancer treatment center during January 1998-September 2013. Of 1,395 Candida bloodstream isolates, 79 from 68 patients were uncommon Candida spp. The incidence density of uncommon Candida spp. BSIs and their proportion to all candidemia episodes substantively increased during the study period, and the rise was associated with increasing use of echinocandin antifungal drugs. Thirty-seven patients had breakthrough infections during therapy or prophylaxis with various systemic antifungal drugs for >7 consecutive days; 21 were receiving an echinocandin. C. kefyr (82%), and C. lusitaniae (21%) isolates frequently showed caspofungin MICs above the epidemiologic cutoff values. These findings support the need for institutional surveillance for uncommon Candida spp. among cancer patients. D espite the widespread use of antifungal prophylaxis and the introduction of new antifungal agents, the incidence of candidemia and associated mortality rates among patients with cancer remain relatively unchanged (1). In previous studies (1)(2)(3), >90% of all Candida-associated invasive fungal infections were caused by 1 of 5 Candida spp.: C. albicans, C. glabrata, C. parapsilosis, C. tropicalis, or C. krusei. However, the use of antifungal drugs such as azoles for prophylaxis and echinocandins that are being used more frequently among high-risk populations have been associated with a continuous shift from C. albicans to various non-albicans Candida spp. during the past 2 decades (1,(4)(5)(6)(7)(8)(9). Moreover, uncommon Candida spp. have emerged as causes of nosocomial bloodstream infections (BSIs) in studies of specific Candida spp. Those isolates commonly exhibit decreased in vitro susceptibility to antifungal agents (10)(11)(12)(13)(14)(15).
The epidemiology and clinical features of many uncommon Candida spp. BSIs have not been well characterized. To that end, we evaluated the epidemiologic characteristics, susceptibility patterns, and factors associated with all-cause death among cancer patients who had uncommon Candida spp. BSIs. We also determined whether the increasing frequency of uncommon Candida spp. BSIs in the study cohort correlated with the increased use of specific antifungal agents.

Isolates
In this retrospective study, we examined the clinical microbiology database at the University of Texas MD Anderson Cancer Center (Houston, Texas, USA) to identify blood cultures that were positive for Candida spp. from patients ≥18 years of age during January 1998-September 2013. Candida isolates were grown on Sabouraud dextrose medium (37°C/48 h/200 rpm) and then phenotypically identified by using CHROMagar Candida medium (CHROMagar Company, Paris, France) and VITEK-2 YST (bioMérieux, Marcy l'Etoile, France). The identification methods were not changed during the study period. We excluded unidentified Candida spp. For our analyses, we selected only the first isolate recovered from blood if a patient had several blood cultures drawn that were positive for the same uncommon Candida spp. Antifungal susceptibility was tested by using the Clinical Laboratory Standards Institute broth microdilution reference method (16). The MIC for caspofungin was tested after March 2005 in the center. For uncommon Candida spp., other than for C. guilliermondii, clinical breakpoints are undefined; therefore, isolates that showed MICs higher than the epidemiologic cutoff value (ECV) were considered potentially resistant (17). There was no ECV for C. famata; therefore, those isolates were excluded from susceptibility comparisons.

Data Collection
We retrospectively reviewed the electronic medical records of patients to obtain demographic, clinical, and laboratory data on the day of blood culture collection (Table 1); we also determined 28-day, all-cause mortality rates using a standardized electronic data collection form. Only first episodes of uncommon Candida spp. BSIs per patient were included in survival analyses. The study and a waiver of informed consent for anonymous data collection were approved by the Institutional Review Board of the MD Anderson Cancer Center.

Definitions
An episode of candidemia was defined as signs or symptoms of infection and >1 blood culture that was positive for Candida spp. Episodes were considered to be separate if they occurred ≥1 month apart. Breakthrough candidemia was defined as candidemia in a patient who had undergone therapy or prophylaxis with any systemic antifungal drug for >7 consecutive days before the index blood culture (18).
Neutropenia was defined as an absolute neutrophil count (ANC) of <500/mL, with further stratification at <100. Persistent neutropenia was defined as an ANC of <500 for >7 days. Neutrophil recovery was defined as restoration of the ANC to >500 for >3 consecutive days (18,19). The source of candidemia was considered to be intraabdominal if the patient had undergone abdominal surgery or had gastrointestinal graft-versus-host disease, peritonitis, cholecystitis, or cholangitis.
Catheter-related bloodstream infections were defined as described by Raad et al. (20) as 1) a colony count of blood obtained through the catheter hub that was >5-fold higher than that in blood obtained from a peripheral vein or 2) a catheter tip culture that was positive for Candida spp. The department of pharmacy provided defined daily doses according to the World Health Organization Anatomical Therapeutic Chemical classification system definition (http://www.whocc.no) for echinocandins, azoles, and amphotericin B (ampB) per 1,000 adult inpatient-days during the study period.

Statistical Analysis
We used descriptive statistics to summarize the demographic, clinical, and outcome variables and the in vitro susceptibility data. We compared percentages with the χ 2 test or Fisher exact test if the expected numbers were <5 in >20% of all cells. Poisson regression and the Cochran-Armitage test were used for the trend analysis of the annual BSI incidence densities and the proportions of candidemia caused by uncommon Candida spp., respectively. We also compared BSI incidence densities for 2 time periods-1998-2005 and 2006-2013-using Poisson distribution and test-based methods. The correlation between the annual use of antifungals and time was evaluated by using the Spearman correlation. The associations between the incidence densities of uncommon Candida spp. BSIs and the annual use of antifungals (defined as daily doses per 1,000 patient-days) were evaluated by using Poisson regression.
We used Cox regression analysis to identify factors that were significantly associated with death. Clinically relevant parameters in the univariate analyses (p<0.1) were included at model entry. The full model was reduced to a final model by using a stepwise elimination procedure. The proportional hazards assumption was tested graphically and by building time-dependent variables. Two-tailed p values <0.05 were considered statistically significant. All analyses were done by using SPSS statistical software version 21 (SPSS IBM, Armonk, NY, USA).
The overall incidence of uncommon Candida spp. BSIs and their proportion relative to all episodes of candidemia increased significantly during 1998-2013 (incidence density p<0.0001; proportion p = 0.001) ( Figure 1). The overall incidence density of uncommon Candida spp. BSIs was 3.17 episodes per 100,000 inpatient days, which increased from 1.89 (1998-2005) (Figure 2), whereas annual azole and ampB use did not (data not shown). The increase in incidence density of uncommon Candida spp. BSIs was associated with the continuous increase in echinocandin use (p = 0.0062).

In Vitro Susceptibility
In vitro susceptibility results were available for 57 isolates (Table 3). C. guilliermondii strains exhibited high rates of azole MICs above ECVs (fluconazole, 17%; voriconazole and posaconazole, 24%; Table 3). The 2 species that commonly were positive for caspofungin MICs above ECVs were C. kefyr (82% vs. 17% among other species; p<0.001) and C. lusitaniae (21%) ( Table 3). Caspofungin  Even though ECVs for that species have not been defined, on the basis of ECV and clinical breakpoints for other Candida spp., that isolate could be considered azole/candinnonsusceptible, making it multidrug resistant.

Discussion
Comprehensive population-based registries of candidemia surveillance have documented the shift from human infections with C. albicans to non-albicans species over the past 2 decades (4,21,22). However, institutional surveillance is equally essential. For example, higher rates of echinocandin resistance are reported from oncology and transplantation centers in the United States (23-25) compared with population-based cohorts (4). At the MD Anderson Cancer Center hospital, the incidence of BSIs caused by uncommon Candida spp. and the proportion of those cases relative to all candidemia cases more than doubled during the past 16 years. Uncommon Candida spp. were frequently nonsusceptible to azoles and echinocandins and were commonly associated with breakthrough infections and high mortality rates. Notably, the incidence density for BSIs caused by uncommon Candida spp. was positively associated with the annual use of echinocandins.
Uncommon Candida spp. distributions vary by geographic region, patient population, and antifungal practices. In general, reported frequencies have been <10% among all Candida isolates (21,22,26,27), which is similar to the proportion of uncommon Candida spp. among all Candida BSIs (3.6%) during the first period of our study (1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005) and to that (3.3%) found in another study of cancer patients during 2009-2012 (28). Nevertheless, the proportion of uncommon Candida spp. BSIs relative to all episodes of candidemia in the MD Anderson Cancer Center hospital increased over the years, accounting for 12% of all Candida BSIs reported during 2013 ( Figure  1), which is among the highest proportions reported to date. This striking difference reflects a severely immunocompromised patient population: 75% had hematologic malignancies, compared with 10.7% in the study by Tang et al (28). However, the most crucial determinant of this marked increase in uncommon Candida BSIs is likely the broad use of echinocandins. For example, in the study by Tang et al. (28), 88.8% of cancer patients with candidemia had previously received fluconazole and <2% had received an echinocandin; the opposite was true in our cohort, where almost one third of patients with uncommon Candida spp. fungemia had breakthrough infections while being treated with an echinocandin. Moreover, the incidence density of the uncommon Candida spp. BSIs in our study was positively associated with the increase in treatment with echinocandins.
In previous reports, C. guilliermondii was one of the most commonly isolated uncommon Candida spp. among patients with cancer (9,13,29); C. dubliniensis was common in the outpatient setting (27). Nevertheless, in our study, during the years 2006-2013, C. guilliermondii was not the most common isolate, and the incidence of C. guilliermondii fungemia did not increase substantially over the study period (Figure 2). This finding is in agreement with another study, wherein the increased use of echinocandins was not associated with an increase in the incidence of C. guilliermondii fungemia (30  incidence of C. kefyr, predominantly among patients with hematologic malignancies, is in agreement with the results of another recent report (12), in which the increase was also attributed to the increasing use of the echinocandin drug micafungin. Taken together, those findings highlight the need, at an institutional level, to systematically monitor changes in Candida spp. distribution and the association with the selective pressure from antifungals. The clinical features and outcomes of breakthrough candidemia with uncommon Candida spp. have not been well described. In our study, more than half of all patients with fungemia caused by uncommon Candida spp., and 36 of 51 patients who had hematologic malignancies (70%), had breakthrough infections. On the contrary, in a 1993-1998 candidemia study at our institution in which uncommon Candida spp. were excluded, ≈25% of all patients, and 46% of those with hematologic malignancies, had breakthrough infections (31). Nevertheless, the percentage of breakthrough infections among all Candida spp. BSIs (53%) in a more recent report (32) was almost identical to that in this study of fungemia caused by uncommon Candida spp. (54%). Those differences are further reflective of the changing epidemiologic characteristics of candidemia and the unique features of uncommon Candida spp. breakthrough infections, which seem to affect a more compromised patient population.
A direct comparison between common and uncommon Candida spp. was beyond the scope of this study, but in another report, among candidemic patients with acute leukemia, we observed a trend for higher mortality rates with the same uncommon Candida spp. infections on univariate analysis, but not on multivariate analysis (25). The only independent predictors of death in the study described here were ICU stay, persistent neutropenia, and high APACHE II score (Table 5), confirming that host characteristics are the most powerful predictors of response and should be adequately adjusted for in studies of candidemia outcomes.
We used the ECV to characterize uncommon Candida spp. bloodstream isolates as susceptible or potentially resistant, according to the updated Clinical and Laboratory Standards Institute/EUCAST definitions (17). C. guilliermondii strains exhibited high rates of azole resistance (Table 3), in agreement with the results of previous reports (13,33,34). However, echinocandin resistance among C. guilliermondii bloodstream isolates in our study was uncommon (a MIC >1 mg/L was observed for only 13% of isolates); in contrast, Girmenia et al. reported that a caspofungin MIC >1 mg/L was observed for 67% of C. guilliermondii strains (13). Moreover, the incidence of C. guilliermondii BSIs remained stable during the 16 years of our study ( Figure 1) and was not substantially associated with echinocandin use. On the contrary, the most common species with caspofungin MICs above ECV was C. kefyr (82%); the incidence density for the species increased substantially over time ( Figure 1) and was positively associated with the annual use of echinocandins (p = 0.004), but not azoles or ampB. Dufresne et al. (12) recently reported a similar rate (88%) of micafungin resistance (MIC >0.12 mg/L) in C. kefyr bloodstream isolates in patients with hematologic malignancies, possibly associated with institutional use of micafungin.
Our study has limitations that should be taken into consideration. First, it was a retrospective study from a single cancer center with a small number of episodes caused by individual uncommon Candida spp.; therefore, our observations might not be applicable to different patient groups at risk for uncommon Candida spp. BSIs. Second, uncommon Candida spp. were identified phenotypically, and it is possible that during the study period, some C. dublinensis isolates were identified as C. albicans, underestimating the frequency of that species. It should also be noted that with the introduction of molecular identification, the taxonomy of the Candida genus is in a state of change (35). The recent implementation of internal transcriber section sequencing (http://www.cbs.knaw.nl/ databases, http://www.ncbi.nlm.nih.gov/genbank) and matrix-assisted laser desorption/ionization in mass spectrometry for Candida spp. identification are expected to further advance understanding of the epidemiology and clinical course of serious infections with uncommon Candida spp.
Third, we used in vitro caspofungin MIC alone to define echinocandin resistance, using no data on DNA mutations. However, there is evidence that caspofungin MIC interlaboratory variability may lead to incorrect categorization of susceptibility results (36), and micafungin and   anidulafungin MICs correlate better with the presence of FKS mutations and clinical outcomes (37). Resistance to echinocandins emerges as a result of treatment and has been associated with mutations in FKS 1/2 genes, which encode the target enzyme for this specific class of antifungals, b-D-glucan synthase (24,38,39). In agreement with what we know about more common Candida spp., investigators have recently identified novel and established FKS1 gene mutations in C. kefyr clinical isolates that are associated with in vitro echinocandin resistance (35,39). Still, the spectrum of mutations that predispose patients to antifungal resistance, the role of epigenetic mechanisms, and the virulence of nonsusceptible, uncommon Candida strains (compared with wild-type) remain unknown at present. Therefore, some experts propose the concept of "clinical resistance," which is a composite of factors related to the host, pathogen, and specific antifungal agent (38).
In summary, we observed a marked increase in the frequency of BSIs caused by uncommon Candida spp. in a contemporary series of patients with malignancies; those species were often associated with breakthrough infections and high mortality rates. The positive correlation between the increasing incidence of uncommon, potentially resistant Candida bloodstream isolates and the increasing use of echinocandins underscores the need for institutional surveillance and the rational use of antifungal drugs in cancer patients.