Epidemiology of Candida kefyr in Patients with Hematologic Malignancies

Abstract

Candida kefyr is an emerging pathogen among patients with hematologic malignancies (HM). We performed a retrospective study at Johns Hopkins Hospital to evaluate the epidemiology of C. kefyr colonization and infection in HM patients between 2004 and 2010. Eighty-three patients were colonized and/or infected with C. kefyr, with 8 (9.6%) having invasive candidiasis (IC). The yearly incidence of C. kefyr colonization and candidemia increased over the study period (P < 0.01), particularly after 2009. In 2010, C. kefyr caused 16.7% of candidemia episodes. The monthly incidence of C. kefyr was higher during the summer throughout the study. In a cohort of patients with acute myelogenic leukemia receiving induction chemotherapy, risks for C. kefyr colonization included the summer season (odds ratio [OR], 3.1; P = 0.03); administration of an azole (OR, 0.06; P < 0.001) or amphotericin B (OR, 0.35; P = 0.05) was protective. Fingerprinting of 16 isolates by repetitive sequence-based PCR showed that all were different genotypes. The epidemiology of C. kefyr candidemia was evaluated in another hospital in Montreal, Canada; data confirmed higher rates of C. kefyr infection in the summer. C. kefyr appears to be increasing in HM patients, with prominent summer seasonality. These findings raise questions about the effect of antifungal agents and health care exposures (e.g., yogurt) on the epidemiology of this yeast.

INTRODUCTION

Candida kefyr (formerly Candida pseudotropicalis) is a yeast with its teleomorph currently recognized as Kluyveromyces marxianus (1, 2). The latter was first isolated from kefir in 1909 and reported under the obsolete name Saccharomyces fragilis (3). The ecology of this yeast is not fully understood, although it appears to thrive in diverse habitats, including dairy products (4–10).

C. kefyr remains a rare cause of disease. The PATH Alliance Registry, which reported infections in 23 major academic institutions in North America between 2003 and 2008, found C. kefyr as a cause of invasive candidiasis (IC) in only 11 (0.2%) of 5,526 Candida infections (11). However, recent reports suggest that it may be an emerging pathogen in patients with hematologic malignancies (HM) (12). In France, in one hospital, 17% of neutropenic patients with acute leukemia were colonized with C. kefyr (13), whereas in three other hospitals, 4.8% of all Candida isolates from hematology wards were C. kefyr (more than twice as many as in nonhematology wards) (14). In contrast, C. kefyr accounted for only 0.7% of Candida species found in the oral cavity of cancer patients in a United Kingdom regional center (15), and this species was not found among candidal surveillance cultures in patients with HM treated in a Finnish hospital (16).

We observed a relatively high number of C. kefyr cultures reported in HM patients at Johns Hopkins Hospital (JHH), beginning in 2009. Considering the paucity of data and potential emergence of C. kefyr as a significant pathogen in patients with HM, we conducted a retrospective study to characterize the epidemiology of C. kefyr in this patient population. We aimed at describing time trends and specific risk factors for C. kefyr infection.

MATERIALS AND METHODS

Study design and patient populations.

This retrospective observational study was approved by the JHH Institutional Review Board (IRB). We studied three distinct but overlapping cohorts within the study period at JHH spanning between 1 January 2004 and 31 December 2010: (i) all adult (>18 years) patients with HM with at least one culture yielding C. kefyr, for descriptive epidemiology; (ii) all patients admitted to two HM-dedicated wards for intensive chemotherapy having systematic fungal surveillance cultures, for time trend analyses; and (iii) a cohort of acute myelogenous leukemia (AML) patients undergoing induction therapy, for risk factor analysis.

Descriptive epidemiology.

Adult HM patients with at least one culture yielding C. kefyr admitted to JHH Sidney Kimmel Comprehensive Cancer Center (JHSKCCC) or JHH during the study period defined the primary study population. Patients were identified by querying the TheraDoc automated surveillance software (TheraDoc, Salt Lake City, UT) for all cultures positive for C. kefyr during the study period. For patients with HM, detailed demographic (age, ethnicity, and gender), clinical (underlying HM, type of stem cell transplant [SCT] [autologous versus allogeneic], type of chemotherapy [induction versus consolidation versus other], and type of C. kefyr infection [IC versus colonization]), and microbiological (site of C. kefyr isolation and antifungal susceptibility testing data) data were collected.

Time trends in C. kefyr colonization and candidemia.

Per standard of care at our institution, patients with HM admitted to HM-dedicated wards A and C have fungal surveillance cultures performed systematically on throat and rectal swabs or stool specimens on admission and weekly thereafter until their discharge. We calculated the incidence of C. kefyr colonization by dividing the number of patients with at least one surveillance culture positive for C. kefyr by the total number of patients with available surveillance culture data, overall and by year and month for time trend evaluation. A single patient could not contribute to the numerator more than once during a calendar year. To evaluate the proportion of candidemia caused by C. kefyr in this cohort, all blood cultures positive for Candida species collected from patients hospitalized in the HM-dedicated wards were also retrieved. Only the first blood culture for a given species per patient was considered.

To determine if our findings were center specific, we performed an IRB-approved, retrospective observational study in a tertiary center caring for patients with HM in Montréal, Canada (Hôpital Maisonneuve-Rosemont [HMR]). All cultures of clinical samples yielding C. kefyr during the same study period were retrospectively identified through the local microbiology laboratory database, including the time (month and year) of sample collection for each culture. Denominator data were not available, as surveillance cultures were not routinely performed in this hospital.

Risk factors for C. kefyr colonization.

We investigated risk factors of C. kefyr colonization using a previously described cohort of 254 consecutive patients with AML undergoing induction chemotherapy with one of two intensive, multiagent induction regimens given in a timed-sequential manner between 1 January 2005 and 30 May 2010 (17). We performed a nested case-control study comparing patients colonized with C. kefyr (cases, n = 22) with those colonized with any other Candida species (controls, n = 205). Exposure variables included year and month of AML induction therapy initiation, chemotherapy regimen, mucositis (site, degree, and duration), total parenteral nutrition, duration of absolute neutrophil count (ANC) of <500 and <100 cells/mm³ following chemotherapy, and the empirical/targeted antibacterial and antifungal regimens administered during the first 6 weeks after chemotherapy initiation. For cases, only antifungal agents received prior to the first culture yielding C. kefyr were considered.

Institutional practices for patients with HM.

Adult patients with HM receiving intensive chemotherapy were primarily admitted to two HM units (A and C) at the JHSKCCC; occasionally, patients may be admitted to other units based on bed availability and concomitant complications. Acyclovir (or valacyclovir) and norfloxacin were used for herpes prophylaxis and gastrointestinal flora decontamination, respectively, from initiation of induction chemotherapy until ANC reached >100 cells/mm³. Antibacterial prophylaxis included ampicillin (or vancomycin in cases of penicillin allergy), initiated on day 8 of induction chemotherapy until resolution of mucositis; additional broad-spectrum antibacterial agents were used for neutropenic fever. Primary antifungal prophylaxis was not routinely administered during the study period. First-line agents for treatment of fever during neutropenia included liposomal amphotericin B or an echinocandin.

Definitions.

Invasive candidiasis (IC) was defined according to published guidelines (18, 19). Candidal colonization was defined as a surveillance culture growing Candida spp., without concomitant evidence of IC. Neutropenic fever was defined as a single temperature of ≥38.3°C or two episodes of ≥38.0°C at least 2 h apart during neutropenia (ANC of <500 cells/mm³) (20). Mucositis was defined according to clinical symptoms and signs of tissue inflammation involving the oropharynx or/and lower gastrointestinal tract and was graded according to published guidelines (21).

Microbiological analyses. (i) Routine microbiological testing.

Before 2009, surveillance cultures were plated on Sabouraud dextrose agar (SDA, Emmons modification; Becton, Dickinson and Co., Franklin Lakes, NJ) with gentamicin. In 2009, CHROMagar Candida (Becton, Dickinson) replaced SDA as the primary culture medium. Plates were incubated for 7 days. Germ tube-negative yeast isolates were identified with combined morphological (Dalmau plate method) and biochemical (in-house fermentation panel and API 20 C AUX system; bioMérieux, Inc., Durham, NC) assays. Antifungal susceptibility testing was performed when clinically indicated using the concentration gradient diffusion assay (Etest; bioMérieux) for amphotericin B and a commercial microdilution plate method (YeastOne; Trek Diagnostic Systems/Thermofisher Scientific, Oakwood Village, OH) for all other antifungal agents. Clinical and Laboratory Standards Institute (CLSI) document M27–S4 (22) breakpoints were used for Candida albicans (fluconazole and micafungin), and CLSI M27–S3 (23) defined breakpoints were used for Candida spp. (flucytosine). Isolates with an amphotericin B MIC of >1 μg/ml were considered resistant, although there is no accepted clinical breakpoint (24).

(ii) Molecular epidemiology.

Based on a clinical observation of higher numbers of cases of C. kefyr IC in patients with HM, a number of C. kefyr isolates were saved in the microbiology laboratory from 2009 onward for future reference. Genomic DNA fingerprinting was performed using repetitive sequence-based PCR (rep-PCR), adapted from a method for Candida rugosa (25). Briefly, genomic DNA was extracted from logarithmic-phase broth culture yeast cells with the MasterPure yeast DNA purification kit (Epicentre Biotechnologies, Chicago, IL) according to the manufacturer's instructions. PCR was performed with the TaKaRa Ex Taq kit (Clontech, Mountain View, CA) according to the manufacturer's instructions. Oligonucleotides Ca-21 (5′CATCTGTGGTGGAAAGTAAAC-3′) and Ca-22 (5′-ATAATGCTCAAAGGTGGTAAG-3′) were used as primers. Each primer was used at 1.0 μM with 100 ng of genomic DNA, in a 25-μl reaction volume. PCR amplification was carried out with the following conditions: initial denaturation at 94°C for 5 min, followed by 35 cycles at 94°C for 15 s for denaturation, annealing with a ramping temperature rate of 1.5°C per second at 51°C for 30 s, and extension at 72°C for 30 s, and a final extension step at 72°C for 5 min. The amplicons were resolved in a polyacrylamide gel (Novex 6% TBE gel; Life Technologies Inc., Grand Island, NY) at 180 V for 35 min. Candida albicans strain SC5314 and Candida kefyr strain ATCC 4922 were used as controls. Cluster analysis was done with the unweighted pair group method with arithmetic mean (UPGMA) using the software PyElph version 1.4 (http://pyelph.sourceforge.net/), according to published guidelines (26). To control for intergel variability, software parameters were adjusted so that a replicate from each gel (C. kefyr ATCC 4922) would appear identical upon blinded cluster analysis.

Statistical analysis.

Analyses were performed using Stata v.11.1 (StataCorp LP, College Station, TX, 2010). Comparisons between categorical variables were performed using the chi-square or Fisher exact test. Continuous variables were compared with the Student t test. The linear trend in yearly incidences of C. kefyr colonization was tested using Poisson regression. To assess risks for C. kefyr colonization, a multivariate logistic regression model was built in a stepwise fashion using variables with a P value of ≤0.05 in the univariate analysis. Two-sided P values of ≤0.05 were considered significant.

RESULTS

Description of HM patients with C. kefyr.

During the study period, 297 cultures growing C. kefyr were identified in 117 patients, 83 (70.9%) of whom had an underlying HM (Table 1). The majority of patients (57/83; 68.7%) had AML, and most of them (50/57; 87.7%) were undergoing induction or reinduction chemotherapy. A total of 77 (92.8%) patients had at least one surveillance culture yielding C. kefyr. Eight (9.6%) patients had C. kefyr IC; all had at least one surveillance culture positive for C. kefyr within a month of their IC diagnosis. Six (7.2%) patients had other samples positive for C. kefyr: urine (n = 5) and a swab culture from a polymicrobial perirectal abscess (n = 1). The first positive culture for C. kefyr was observed >2 weeks after admission to the hospital in the majority of patients (66; 79.5%).

TABLE 1.

Characteristics of all adult patients with a hematologic malignancy (HM) and a culture yielding Candida kefyr between 2004 and 2010 (n = 83)

Characteristicd	n (%) (except for age)
Demographic data
Age in yr, mean (range)	48.7 (18–73)
Gender, male	44 (53.0)
Ethnicity
White	52 (62.7)
Black	16 (19.3)
Hispanic	5 (6.0)
Asian	2 (2.4)
Other	8 (9.0)
Underlying disease
Hematologic malignancya
AML	57 (68.7)
ALL	8 (9.6)
Other AL	6 (7.2)
MDS	1 (1.2)
CML	3 (3.6)
MM	2 (2.4)
Lymphoma	6 (7.2)
SCTa
Autologous	1 (1.2)
Allogeneic	3 (3.6)
Chemotherapy type (AML only; n = 57)
Induction	35 (61.4)
Reinduction	15 (26.3)
Consolidation	5 (8.8)
None	2 (3.5)
Clinical manifestation
Colonization	69 (83.1)
Invasive candidiasis	8 (9.6)
Otherb	6 (7.2)
Timingc
<2 wk	17 (20.5)
>2 wk	66 (79.5)

^aExcluding patients with relapsing hematologic malignancy after SCT, undergoing salvage chemotherapy at the time of C. kefyr colonization.

^bOther included the following 6 cases: 5 patients with asymptomatic candiduria and 1 patient with a polymicrobial perirectal abscess.

^cTime between admission to hospital and first culture yielding C. kefyr.

^dAbbreviations: AML, acute myelogenous leukemia; ALL, acute lymphoblastic leukemia; AL, acute leukemia; MDS, myelodysplastic syndrome; CML, chronic myeloid leukemia; MM, multiple myeloma; SCT, stem cell transplant.

There were nine episodes of C. kefyr IC in eight patients (Table 2). Patient 5 had two episodes of C. kefyr IC 1 year apart: the first due to likely gastrointestinal translocation and the second retrieved coincident with a bladder fungus ball. Only one patient had confirmed catheter-related candidemia and was not neutropenic at the time of first positive blood culture. Most episodes occurred between July and September (6/9) or in 2009 to 2010 (6/9). Five episodes were breakthrough infections occurring coincident with antifungal therapy (liposomal amphotericin B, n = 2; micafungin, n = 3).

TABLE 2.

Clinical characteristics of 9 episodes of Candida kefyr invasive candidiasis (8 patients)^a

Patient	Underlying disease (status)/type of chemotherapy	Mucositis	TPN	Central intravenous catheter duration (days)	ANC (at episode/days <100 during prior mo)	Timing post-chemotherapy initiation (days)	Antifungal therapy at time of first positive blood culture	Mo and yr at first positive blood culture	Duration of candidemia (days)	Catheter (day of removal/culture result)	Other foci	Death
1	ALL (pre-B) (de novo)/induction	No	No	28	1,157/16	27	L-Amb	Aug 2004	2	1, Pos	None	No
2	AML (de novo)/induction	Yes	No	18	0/12	18	None	Oct 2005	1	2, Neg	None	No
3	Bilineage leukemia (de novo)/consolidation	No	No	126	0/5	15	None	Sep 2008	1	1, Neg	None	No
4	AML (de novo)/induction	No	No	43	0/15	18	Mica	Jul 2009	2	2, Neg	None	Yes
5	AML (relapsed)/salvage	Yes	No	20	0/20	13	None	Aug 2009	2	2, ND	None	No
	AML (relapsed)/salvage	No	No	None	0/30	58	L-Amb, 5FC	Jul 2010	6		Bladder fungus ball	Yes
6	AML (relapsed)/salvage	No	No	19	0/15	18	None	Oct 2009	3	1, Neg	None	No
7	AML (relapsed)/salvage	No	No	75	0/25	17	Mica	Jan 2010	9	14, Neg	Retina	Yes
8	AML (relapsed)/induction	Yes	Yes	23	0/13	23	Mica	Aug 2010	2	1, Neg	None	No

^aAbbreviations: AML, acute myelogenous leukemia; ALL, acute lymphoblastic leukemia; TPN, total parenteral nutrition; ANC, absolute neutrophil count; Mica, micafungin; L-Amb, liposomal amphotericin B; 5FC, flucytosine; Pos, positive; Neg, negative; ND, not determined; Jan, January; Jul, July; Aug, August; Sep, September; Oct, October.

Antimicrobial susceptibility testing data were available for 13 isolates from patients with C. kefyr IC. All breakthrough isolates were resistant to the ongoing antifungal therapy. Detailed culture data with MICs (μg/ml) are presented in Table 3.

TABLE 3.

MICs to fluconazole, micafungin, amphotericin B, and flucytosine of 13 C. kefyr isolates from 8 patients with invasive candidiasis

Patient	Specimen	MIC (μg/ml)a
		FLU	MICA	AMB	5FC
1	Stool	0.5		2	0.12
2	Blood	0.25
3	Blood	0.25
4	Blood	≤0.12	≥8
5	Blood	1	0.12	1
	Stool			1
	Urine	128	8	≥2	0.06
	Urine	8		≥2	64
	Blood	8	0.25	≥2	64
6	Blood	0.5	0.25	0.5
7	Blood	0.5	1
	Blood	0.5	1	0.5
8	Blood	0.25	8	≤1

^aAbbreviations: FLU, fluconazole; MICA, micafungin; AMB, amphotericin B; 5FC, flucytosine. Values at or above the level of resistance are marked in bold.

Trends in C. kefyr colonization and candidemia among HM patients.

A total of 1,844 patients admitted to two HM-dedicated wards (A and C) had 12,478 surveillance cultures sent during the study period. From these, 66 patients had 153 cultures yielding C. kefyr, with a positivity rate and incidence of 1.2% and 3.6%, respectively. Seventeen of the 83 HM patients from the primary study patient population were not captured by this analysis: in six, C. kefyr was found in a specimen other than a surveillance culture, and 11 patients were not hospitalized on the two targeted wards at the time at which the C. kefyr-positive specimen was collected. A total of 89 candidemia episodes occurred in this cohort, 7 (7.9%) of which were caused by C. kefyr. One patient with C. kefyr candidemia described in Table 2 was not captured in this analysis because he was hospitalized on another ward at the time of candidemia.

Overall, there was a significant increase in yearly incidence of C. kefyr colonization during the study period (linear trend, P < 0.01; Fig. 1). This trend was mostly driven by 2009, when the incidence was significantly higher than in the rest of the study period (21/354, 5.9%, versus 45/1,445, 3.1%; P < 0.01). If data from 2009 were removed, the statistical significance was lost (P = 0.18). The proportion of candidemia caused by C. kefyr mirrored that of C. kefyr colonization and was the highest in 2009 (2/13, 15.4%) and 2010 (2/12, 16.7%), although the numbers were low (Fig. 1).

FIG 1.

Time trends in Candida kefyr (Ck) colonization and candidemia among patients hospitalized on hematologic malignancy (HM)-dedicated wards between 2004 and 2010; yearly incidence of C. kefyr colonization (open bars) and proportion of candidemia episodes caused by C. kefyr (closed triangles) are shown.

The monthly incidence of C. kefyr colonization in patients with HM from HM-dedicated wards A and C over the study period revealed a strong seasonality, with the highest rates consistently observed between July and September, throughout the study period (Fig. 2A). The most prominent peak occurred in the summer of 2009. The monthly incidence combining all study years is shown in Fig. 2B. Considering the high incidence of C. kefyr colonization in 2009, we compared the monthly incidence with and without data from 2009: results were not significantly driven by 2009 data (P > 0.20; Fisher's exact test). These findings were not isolated to JHH; a similar seasonal distribution was observed at the HMR (n = 37) (Fig. 2C).

FIG 2.

Seasonal distribution of Candida kefyr between 2004 and 2010: monthly incidence of C. kefyr colonization among patients hospitalized on hematologic malignancy (HM)-dedicated wards at the Johns Hopkins Hospital (A) with combined data from all study years with (empty bars) and without (solid bars) 2009 data (B) and monthly total numbers of positive cultures for C. kefyr at Hôpital Maisonneuve-Rosemont (combined data for all study years) (C).

Risk factors for C. kefyr colonization.

Using a previously reported cohort of 254 consecutive AML patients, we identified 227 patients colonized with at least one Candida species. Twenty-two (8.7%) were colonized with C. kefyr compared to 66 (3.6%) of 1,844 HM patients (P < 0.001); the remaining 205 were colonized with other Candida spp. only. The first C. kefyr-positive surveillance cultures occurred at a median of 18 days (range, 2 to 32) after initiation of intensive chemotherapy. In 19 (of 22; 86.4%) C. kefyr-colonized patients, another Candida species was identified, isolated prior to C. kefyr in 15 (of 19; 78.9%) cases. Results of univariate and multivariate analyses to identify predictors of C. kefyr colonization among these 227 AML patients are presented in Table 4. Warm season (July to September, odds ratio [OR], 3.1; 95% confidence interval [CI], 1.1 to 8.6; P = 0.03) and colonization with >1 Candida species (OR, 6.0; 95% CI, 1.6 to 22.5; P = 0.007) were the most significant predictors for C. kefyr colonization. In contrast, administration of an azole antifungal (OR, 0.06; 95% CI, 0.01 to 0.3; P < 0.001) or amphotericin B (OR, 0.35; 95% CI, 0.01 to 0.99; P = 0.05) protected against C. kefyr colonization.

TABLE 4.

Risk factors for colonization with Candida kefyr in patients with acute myelogenous leukemia (AML) undergoing induction chemotherapy with one of two intensive, multiagent induction regimens given in a timed-sequential manner: (i) AcDVP16 (cytarabine, daunorubicin, and etoposide) and (ii) FLAM (flavopiridol, cytarabine, and mitoxantrone)

Characteristica	No. (%)		Value by analysisd:
			Univariate			Multivariate
	Cases, n = 22	Controls, n = 205	OR	95% CI	P	OR	95% CI	P
Age of >50 yr	13 (59)	124 (60)	0.9	0.4–2.3	0.9
Gender, male	11 (50)	116 (57)	0.8	0.3–1.8	0.56
Ethnicity, Caucasian vs other	18 (82)	166 (81)	1.1	0.3–3.3	0.92
Yr of ≥2009	7 (32)	50 (24)	1.5	0.6–3.7	0.45
Summer (July to September)	12 (55)	45 (22)	4.3	1.7–10.5	0.002	3.1	1.1–8.6	0.03
Regimen, AcDVP16 vs FLAM	15 (68)	148 (72)	0.8	0.3–2.1	0.7
Leukemia from MDS	6 (27)	68 (33)	0.8	0.3–2.0	0.58
ANC of <500, >21 days	17 (77)	155 (76)	1.1	0.4–3.1	0.86
ANC of <500 on presentation	6 (27)	62 (30)	0.9	0.3–2.3	0.77
Mucositis of ≥2nd grade	11 (50)	101 (49)	1.0	0.4–2.5	0.95
Receipt of TPN	3 (14)	47 (23)	0.5	0.1–1.9	0.32
Colonized with >1 Candida species	19 (86)	115 (56)	4.9	1.4–17.3	0.01	6.0	1.6–22.5	0.007
Receipt of L-Amb	11 (50)	148 (72)	0.4	0.2–0.9	0.04	0.35	0.01–0.99	0.05
Receipt of azoleb	2 (9)	144 (70)	0.06	0.01–0.3	<0.0001	0.06	0.01–0.3	<0.001
Receipt of echinocandinc	11 (50)	75 (37)	1.7	0.7–4.2	0.22
Receipt of 5FC	1 (5)	36 (18)	0.2	0.03–1.7	0.15

^aAbbreviations: MDS, myelodysplastic syndrome; ANC, absolute neutrophil count; TPN, total parenteral nutrition; L-Amb, liposomal amphotericin B; 5FC, flucytosine.

^bIncludes fluconazole, posaconazole, and voriconazole.

^cIncludes anidulafungin, caspofungin, and micafungin.

^dValues in bold indicate statistical significance.

Molecular epidemiology.

A total of 16 clinical isolates (2009, n = 10; 2010, n = 6) from 14 patients were available for molecular epidemiology analysis. All patients were hospitalized in one of two (or both) HM-dedicated wards, A and C. Genetic relatedness of isolates was determined by DNA fingerprinting using rep-PCR. Computer-assisted cluster analysis of profiles was performed, and the dendrogram derived from the UPGMA clustering algorithm is shown in Fig. 3. All isolates appeared to be different genotypes, clustering into 6 groups of genotypes with a genetic distance of <0.10. Isolates c112 and c115 were recovered from the same patient, almost 1 year apart. Isolates c119 and c120 also originated from a single patient, at two different anatomic sites. c119 and c121 appeared to be closely related and were recovered from different patients within a 2-day interval, potentially suggesting a common source. c113–c128 and c126–c127 are other pairs of isolates that were both genetically and temporally close. There was no pattern linking anatomic origin (i.e., from fungal surveillance culture versus blood) to specific genotypes.

FIG 3.

Cluster analysis of rep-PCR profiles of 16 Candida kefyr isolates from 14 patients from 2009 and 2010. A dendrogram derived from the UPGMA algorithm is presented; scale represents percent similarity between isolates. Band profile, date of collection, and specimen origin are shown for each isolate. Isolates from the same patients are indicated with a gray box. Six groups of genotypes with genetic similarity scores of >0.90 are indicated.

DISCUSSION

This is the largest study describing the epidemiology of C. kefyr in a contemporary cohort of patients with HM. Our findings suggest that a significant proportion of patients with HM, especially patients with AML, are colonized with C. kefyr, with a substantial risk for subsequent bloodstream infection. Colonization and infection follow a seasonal distribution, with higher rates during the warm months of the year.

Our understanding of the epidemiology of C. kefyr has been limited, derived mainly from case reports (12, 27, 28), laboratory-based analyses (13–15, 29–34), and large surveillance registries (11). This body of evidence has suggested that HM patients may be a high-risk group for infection with this organism, albeit C. kefyr is a rare cause of invasive disease. Our study is the first comprehensive review of the epidemiology of C. kefyr. Although we could not confirm that C. kefyr incidence was steadily increasing over the study period, it appears that C. kefyr is emerging as a pathogen of concern in HM patients. We found that patients with HM accounted for 71% of all patients with C. kefyr at our institution and that up to 6% of patients with HM were colonized with C. kefyr. This may, in part, represent detection bias due to routine systematic surveillance cultures performed for HM patients at our institution. In addition, in the last 2 years of the study period, 16% of candidemia episodes were caused by C. kefyr in a subset of HM patients, making it the third most common species retrieved from blood in this group (data not shown). The incidence of C. kefyr colonization was highest among patients with AML undergoing induction chemotherapy (22/254, 8.7%). Colonization with multiple Candida species was an independent risk factor for C. kefyr colonization, which may be a reflection of a “Candida-permitting” host environment in this highly antibiotic-experienced, profoundly neutropenic patient population. C. kefyr has been shown to be less virulent than C. albicans and Candida tropicalis experimentally (35); factors that enable gut colonization with C. kefyr are not yet elucidated.

The incidence of C. kefyr colonization varied during the study period, with a peak observed in 2009. Antibiotic prophylaxis, neutropenic diet, infection control practices, and frequency of colonization cultures remained the same during the study period. However, two major changes occurred in 2009 at our institution that could have potentially affected the higher rates of C. kefyr colonization observed. First, a new differential colorimetric medium for yeast recovery was introduced in the microbiology laboratory in early 2009. Replacing the traditional Sabouraud dextrose agar (Emmons modified with antibiotics; with recognition of mixed cultures based on different colony morphologies) could have led to increased recognition of colonization (36). Candida kefyr appears as creamy, glossy colonies, similar to many other Candida species, on SDA and appears as more distinct, pinkish colonies on CHROMagar. However, C. kefyr was not more likely to be found in mixed yeast cultures after 2009 than it was before (data not shown), making this explanation not likely to account for the entire observation.

The other change that occurred during 2009 was replacement of caspofungin with micafungin in the institutional antimicrobial formulary. It is conceivable that exposure to micafungin might have selected for micafungin-resistant organisms among the gut colonization flora. In support of this hypothesis is the observation that three of nine C. kefyr candidemia episodes occurred in patients while receiving micafungin, with all infections caused by organisms that had high MICs. More detailed study of these isolates (J. F. Staab, D. Neofytos, P. Rhee, C. Jimenez-Ortigosa, S. X. Zhang, D. S. Perlin, and K. A. Marr, submitted for publication) documented variable MICs to different echinocandins. Also, other investigators have observed C. kefyr resistance to echinocandins developing rapidly under treatment (37). Although our risk factor analysis identified administration of azoles or amphotericin B to be protective against C. kefyr colonization, it appears that C. kefyr has the propensity to develop resistance to all classes of antifungals, based on the MICs observed in the IC C. kefyr isolates in our study. More data are required to make more meaningful conclusions regarding the clinical implications of these variable susceptibility patterns.

We observed a strong seasonal pattern in C. kefyr occurrence, which was location (institution) and year independent, and found the summer season to be a significant predictor of C. kefyr colonization in a multivariable logistic regression model. This observation may contribute to a better understanding of the pathophysiology and mode of acquisition of this yeast. One can hypothesize that the seasonal pattern could reflect rates of colonization-exposure or even the likelihood of contamination of food products, but the potential route of acquisition remains unclear. In a recent report, two out of three HM patients with C. kefyr candidemia had consumed dairy products (12). Indeed, yogurt and other milk products were offered to HM patients at our institution during the study period. As K. marxianus is the main yeast associated with yogurt spoilage (38–40), which is directly influenced by storage temperature (41), it is plausible that cold chain disturbance that occurs in the summer could affect the concentration of C. kefyr in these products.

We observed that most patients became colonized with C. kefyr after at least 2 weeks of hospitalization, raising the possibility of a nosocomial acquisition or selection by treatment, such as with antifungal exposure. Many variables can impact colonization by Candida species; for instance, the use of metronidazole for intestinal decontamination has been shown to increase yeast colonization in hematopoietic stem cell transplant recipients (42). Based on our molecular analyses, isolates appear dissimilar; we cannot confirm that nosocomial or horizontal transmission has occurred or make any conclusions regarding specific patterns of acquisition.

There are several limitations to this study. The case-control study included a small number of patients and represented only a subset of all cases. Microbiological analysis was also limited to a subset of isolates. Molecular epidemiology of C. kefyr isolates has never been reported; therefore, the discriminatory power of the method (rep-PCR) for this organism is not well defined. However, the observation of interpatient isolate discrimination concurrent with similarity within pairs of isolates recovered from the same patient supports the validity of the technique. Unfortunately, many exposure variables—such as food acquisition—were not available and could neither be introduced in the risk-factor analysis nor complement the fingerprinting data.

We conclude that C. kefyr emerged as an important pathogen in patients with HM at our institution before 2010, particularly in patients with AML undergoing induction chemotherapy. The epidemiology of this organism is characterized by a strong seasonality that suggests potential acquisition by exogenous exposure. Our findings raise questions about the effect of different antifungal agents and health care exposures (e.g., yogurt) on the epidemiology of C. kefyr in patients with HM. More studies are required to further elucidate the epidemiology of this yeast pathogen.