cncr.23466

The Changing Epidemiology of Invasive Candidiasis
Candida Glabrata and Candida Krusei as the Leading Causes of Candidemia
in Hematologic Malignancy
Ray Hachem, MD
Hend Hanna, MD
Dimitrios Kontoyiannis, MD
Ying Jiang, MS
Issam Raad, MD
Department of Infectious Diseases, Infection Con-
trol and Employee Health, The University of Texas
M. D. Anderson Cancer Center, Houston, Texas.
BACKGROUND. The objective of the current retrospective study was to compare
the epidemiology of candidemia and its risk factors in patients who had hemato-
logic malignancies(HM) with those in patients who had solid tumors (ST).
METHODS. The medical and electronic records of all patients with cancer who
had candidemia at the authors’ institution from 1993 to 2003 were reviewed for
demographic data and clinical information, including the use of prophylactic flu-
conazole, the infecting Candida species, and the source of candidemia (catheter-
related vs other apparent sources).
RESULTS. Six hundred thirty-five patients with candidemia were analyzed. C.
glabrata and C. krusei were the leading causes of candidemia in 31% and 24% of
patients with HM, respectively, and in 18% and 2% of patients with ST, respec-
tively (P<.001). A catheter was the source of candidemia in 36% of the patients
with ST and in 12% of the patients with HM (P<.001). Response to antifungal
therapy occurred in 73% of the ST group compared with 49% of the HM group
(P<.001). Multivariate logistic regression analysis revealed that fluconazole pro-
phylaxis was a risk factor for both C. glabrata and C. krusei candidemia. The
analysis also identified neutropenia as a risk factor for all candidemia and cathe-
ter-related infection as a risk factor for C. parapsilosis candidemia.
CONCLUSIONS. The results of this study indicated that C. glabrata and C. krusei
were the leading causes of candidemia in patients with HM. Neutropenia was the
leading risk factor for all candidemia, whereas the catheter was the leading risk
factor for C. parapsilosis candidemia. Cancer 2008;112:2493–9. 2008 American
Cancer Society.
KEYWORDS: epidemiology, candidemia, cancer patient, hematologic malignancies,
solid tumors.
Bloodstream infections caused by Candida species increasingly
have become an important form of invasive candidiasis accom-
panied by consequent rises in related mortality rates and prolonged
hospitalization.
1–5
Although Candida albicans remains the most
common fungal isolate obtained from the blood of patients with
invasive fungal infection, several studies have demonstrated a trend
toward an increased prevalence of infections with non-albicans Can-
dida spp., such as C. glabrata and C. krusei.
6–9
The past 20 years
have seen an increase in the severity of illness in hospitalized
patients, extensive use of invasive medical devices, and wider use of
broad-spectrum antimicrobial agents, all of which have resulted in
an increased incidence of nosocomial fungal infections.
10–12
Furthermore, because of advancements in medical technology, cur-
rently, immunocompromised patients are subjected to more aggres-
sive management of their underlying malignancies with potent
See editorial on pages 23347, this issue.
Address for reprints: Ray Hachem, MD, Depart-
ment of Infectious Diseases, Infection Control
and Employee Health, Unit 402, The University of
Texas M. D. Anderson Cancer Center, 1515
Holcombe Boulevard, Houston, TX 77030; Fax: (713)
745-6839; E-mail: [e-mail]
Received September 12, 2007; revision received
November 16, 2007; accepted November 26,
2007.
ª2008 American Cancer Society
DOI 10.1002/cncr.23466
Published online 15 April 2008 in Wiley InterScience (www.interscience.wiley.com).
2493

chemotherapeutic agents and transplantation, which
lead to more profound neutropenia as well as
damage of the mucosal integrity. These immunocom-
promised patients also are administered more fre-
quent and prolonged antibiotic regimens, which tend
to suppress the natural endogenous bacterial flora.
All of these factors have set the stage for a significant
increase of invasive fungal infections caused by
Candida spp.
3,7,13
Today, Candida spp. rank as the
fourth most common cause of nosocomial blood-
stream infections in the U.S.
14
Candidemia not only
is associated with increased mortality and morbidity
rates but also prolongs hospitalization and increases
medical care costs.
15–17
In the current study, we com-
pared the epidemiology of candidemia and its risk
factors in patients with hematologic malignancies
with that in patients with solid tumors.
MATERIALS AND METHODS
Clinical Characteristics
This retrospective cohort study was conducted at the
University of Texas M. D. Anderson Cancer Center.
Patients with candidemia were identified by search-
ing infection control records and a microbiology lab-
oratory database for all episodes of candidemia that
occurred from January 1, 1993, to January 1, 2003.
The medical and electronic records of patients with
cancer who had candidemia were reviewed for de-
mographic data and clinical information, including
the use of prophylactic fluconazole and other anti-
fungal agents, the infecting Candida species, the
underlying malignancy, the source of candidemia,
bone marrow transplantation status, the Acute Physi-
ology and Chronic Health Evaluation (APACHE) II
score at the onset of candidemia, intensive care unit
(ICU) stay, neutropenia status, and the use of central
venous catheters, steroids, antibiotics, chemothera-
peutic agents, and hyperalimentation.
The antifungal agents that were used during this
study were predominately lipid formulations of
amphotericin B or fluconazole, depending on charac-
terization of the fungal isolate. However, voriconazole
or echinocandin were not used, because the study
period was completed before their approval at our
center.
Microbiologic Methods
Detection of candidemia and species identification
of isolates were performed in our microbiology labo-
ratory. The presence of Candida spp. in blood iso-
lates was confirmed by using the Vitek YBC system
or the API 20C AUX or API 32ID (BioMerieux) in ac-
cordance with the manufacturer’s instructions. When
necessary, germ tube analysis, morphologic analysis
with cornmeal-Tween 80 agar, and differential growth
testing at 35 8C and 43 8C were performed for confir-
mation. The method used for antifungal susceptibil-
ity testing of the isolates and the interpretative
criteria used for evaluating this susceptibility were in
strict accordance with the recommendations of the
Clinical and Laboratory Standards Institute.
18
The
intravascular segments (catheter tips) of all central
venous catheters that were used in the patient group
were cultured by using the semiquantitative roll-
plate method described by Maki et al.
19
Definitions
On the basis of the guidelines of the Centers for Dis-
ease Control and Prevention and the Infectious Dis-
eases Society of America (IDSA),
20–22
candidemia was
defined as the isolation of a Candida spp. from at
least 1 blood culture in the presence of signs and
symptoms of systemic fungal infection. A case was
considered likely to be catheter-related when the
same Candida species was isolated both from a pe-
ripheral blood culture and from the catheter tip,
yielding >15 colony-forming units by using the semi-
quantitative roll-plate culture method in the absence
of other sources of candidemia, or from a quantita-
tive blood culture that was collected through a cen-
tral venous catheter, yielding 5-fold more colony-
forming units per mm
3
than were yielded in a simul-
taneous culture of blood collected from a peripheral
vein. Secondary candidemia was defined as a blood-
stream infection caused by a Candida species with a
documented concurrent infection or colonization
caused by the same Candida species at a site other
than the catheter in accordance with the guidelines
described by Pittet and Wenzel.
15
Neutropenia was
defined as a neutrophil count <500 neutrophils/
mm
3
. Response was defined as the resolution of all
clinical manifestations (such as fever or chills) of the
candidemia with sterilization of the bloodstream.
Failure to respond was defined as persistence of the
clinical signs and symptoms of the fungal infection,
persistence of blood cultures positive for Candida
spp., development of a Candida infection at a new
site, or a combination of these factors.
Statistical Analysis
The chi-square test and Fisher exact tests were used
to compare categorical variables. Continuous vari-
ables were compared using the Student ttest or the
Wilcoxon rank-sum test, as appropriate. Logistic
regression analysis was used to identify the inde-
pendent risk factors for candidemia caused by differ-
ent species. First, univariate analyses were performed
2494 CANCER June 1, 2008 / Volume 112 / Number 11

to evaluate the predictive effect of each factor alone.
Then, any factor with a univariate Pvalue <.1 was
included in a full multiple logistic regression model.
Final models were determined by using a backward
method for variable selection. All tests were con-
ducted as 2-sided tests at a significance level of
P<.05. Statistical analyses were performed by using
the SPSS software program (version 12.0; SPSS Inc,
Chicago, Ill).
RESULTS
We evaluated 635 episodes of candidemia based on
the definitions described above that occurred over
the 10-year study period. Of these episodes, 354
(56%) occurred in patients with underlying solid
tumors, and 281 (44%) occurred in patients with
underlying hematologic malignancies. Among the
patients with hematologic malignancies, C. glabrata
was the most common infecting Candida species
(31% compared with 18% in patients with solid
tumors; P.001) followed by C. krusei at 24% (com-
pared with 2% in patients with solid tumors;
P.001). In contrast, C. albicans (45%; P.001) and
C. parapsilosis (20%; P5.045) were the most com-
mon infecting Candida spp. in patients with underly-
ing solid tumors. The numbers of infections caused
by C. tropicalis,C. guilliermondii,C. lusitaniae, and
mixed Candida spp. were comparable in the 2
patient groups (Table 1).
The mean age, the percentage of patients who
were men, the mean ICU stay during the month
before the onset of candidemia, and the number of
patients who underwent transplantation within the
previous year were similar in the 2 patient groups
(Table 2). Patients with hematologic malignancies
were more critically ill than patients with solid
tumors at the onset of their candidemia, as evi-
denced by higher mean APACHE II scores (mean 
standard deviation: 15.3 4.6 vs 13.4 4.3, respec-
tively; P<.001). Also, more patients with hemato-
logic malignancies than patients with solid tumors
received corticosteroids and prophylactic antifungals
(P<.001). Furthermore, patients with hematologic
malignancies were admitted to the ICU more fre-
quently (46% vs 33%; P5.001), more frequently were
neutropenic (75% vs 20%; P<.001) during the course
of their candidemia, and more of them had persis-
tent neutropenia (41% vs 4%; P<.001). Conversely,
candidemia was catheter-related more often in
patients with solid tumors than in patients with he-
matologic malignancies (36% vs 12%; P<.001). In
addition, 73% of patients with solid tumors had clini-
cal and microbiologic responses to treatment with
antifungal agents compared with 49% of patients
with hematologic malignancies (P<.001).
According to multiple logistic regression analysis,
factors that were predictive of C. albicans candide-
mia were an absence of neutropenia (odds ratio
[OR], 1.56; 95% confidence interval [95% CI], 1.20–
2.44 [P5.047]), the presence of an underlying solid
tumor (OR, 2.50; 95% CI, 1.54-4 [P<.001]), and no
prior use of prophylactic fluconazole (OR, 3.33; 95%
CI, 2.04-5.56; [P<.001]) (Table 3). Antifungal prophy-
laxis with fluconazole and the presence of neutrope-
nia were independent risk factors for C. krusei
candidemia. In addition, antifungal prophylaxis with
TABLE 1
Distribution of Different Candida Species Among Patients With
Hematologic Malignancies and Solid Tumors
Candida species
No. of patients (%)
PHM, n 5281 ST, n 5354
C. albicans 38 (14) 161 (45) <.001
C. glabrata 86 (31) 64 (18) <.001
C. krusei 68 (24) 6 (2) <.001
C. parapsilosis 39 (14) 71 (20) .045
C. tropicalis 27 (10) 30 (9) .620
C. guilliermondii 4 (1) 2 (1) .410
C. lusitaniae 3 (1) 3 (1) .990
Mixed Candida species 16 (6) 17 (5) .620
HM indicates patients with hematologic malignancies; ST, patients with solid tumors.
TABLE 2
Clinical Characteristics of Patients With Candidemia According to
Underlying Disease
Characteristic
No. of patients (%)
PHM, n 5281 ST, n 5354
Sex (men) 161 (57) 182 (51) .830
Mean age SD, y 53.0 17.8 53.3 17.1 .830
Mean APACHE II score SD 15.3 4.6 13.4 4.3 <.001
ICU stay before onset of infection 62 (22) 64 (18) .210
ICU stay during infection 128 (46) 115 (32) .001
Transplantation within previous y 42 (15) 43 (12) .350
Corticosteroid use 118 (42) 75 (21) <.001
Catheter-related candidemia 34 (12) 127 (36) <.001
Concomitant infection 197 (70) 266 (75) .178
Prophylactic fluconazole use 151 (54) 56 (16) .001
Neutropenia during infection 211 (75) 72 (20) <.001
Persistent neutropenia 114 (41) 15 (4) <.001
Clinical and microbiologic
response to antifungals 137 (49) 259 (73) <.001
HM indicates patients with hematologic malignancies; ST, patients with solid tumors; SD, standard
deviation; APACHE II, Acute Physiology and Chronic Health Evaluation II; ICU, intensive care unit;
Concomitant infection, coinfection.
Epidemiology of Candidemia in Cancer/Hachem et al. 2495

fluconazole was a predisposing risk factor for C.
glabrata candidemia, the presence of neutropenia
was an independent risk factor for C. tropicalis can-
didemia, and catheter-related infection was an inde-
pendent risk factor for C. parapsilosis candidemia.
Furthermore, in the late 1990s, there was a shift in
the distribution of Candida spp. with the increase in
the rate of candidemia caused by non-albicans Can-
dida spp., predominately C. glabrata and C. krusei
(Table 4).
DISCUSSION
The current study demonstrated a shift from C. albi-
cans to non-albicans Candida spp. as the predomi-
nant infecting species in patients with cancer who
had candidemia. Specifically, the majority of the
cases of candidemia in this study were caused by
non-albicans Candida spp. (69%), and more than
half of the patients had underlying solid tumors
(56%). C. glabrata caused most of the cases of candi-
demia in patients with hematologic malignancies
(31%), whereas C. albicans was the most frequent
infecting species in patients with solid tumors (46%).
Candida spp. have become among the most
common organisms causing nosocomial bloodstream
infections over the past 20 years.
3,14
The National
Nosocomial Infections Surveillance System reports
on trends in nosocomial primary bloodstream infec-
tions in the U.S. From 1980 to 1989, 124 National
Nosocomial Infections Surveillance System hospitals
reported more than 25,000 primary bloodstream
infections. During that period, the rates of blood-
stream infections caused by Candida spp. increased
significantly.
23
By the 1990s, Candida spp. had
become the fourth most common organisms causing
nosocomial bloodstream infections.
14
Although at
one time C. albicans was the most frequently isolated
Candida species among patients with candidemia,
over the past few decades, authors have documented
increases in the rates of candidemia caused by non-
albicans Candida spp., predominantly C. glabrata
and C. krusei, especially in critically ill and immuno-
compromised patients with malignancies.
7,23–30
Among the factors that have been proposed to
explain this shift in the distribution of the Candida
spp. responsible for candidemia is the prevalent use
of fluconazole as a prophylactic antifungal agent,
especially in patients with hematologic malignancies
and recipients of bone marrow transplantation.
29,31
Furthermore, several investigators have postulated
that the widespread use of fluconazole would have
selected Candida spp. That intrinsically are resistant
or less sensitive to fluconazole, such as C. glabrata
and C. krusei.
24,26,29,32,33
Investigators have identified
several other risk factors for candidemia over the
years, including the use of central venous catheters,
neutropenia, and the use of broad-spectrum antibio-
tics and corticosteroids. In agreement with the find-
ings from studies in hospitals other than ours,
31,34
the current study demonstrated that patients with
candidemia caused by a non-albicans Candida spe-
cies received prophylactic antifungal agents before
the onset of their infections more frequently than
patients with candidemia caused by C. albicans (51%
vs 15%; P<.001). Fluconazole was the most com-
mon prophylactic antifungal agent received before
the onset of candidemia: It was the agent adminis-
tered in 82% of the cases caused by non-albicans
Candida spp.
More patients with solid tumors than patients
with hematologic malignancies had catheter-related
candidemia, and C. parapsilosis was the most com-
TABLE 3
Multiple Logistic Regression Analysis of Independent Variables
Predisposing Patients to Candidemia Caused by Different Species
Candida species Independent risk factor OR 95% CI
C. albicans No neutropenia 1.560 1.200–2.440
No fluconazole prophylaxis 3.330 2.040–5.560
Presence of solid tumor 2.500 1.540–4.000
C. tropicalis Neutropenia 2.325 1.287–4.202
C. glabrata Fluconazole prophylaxis 2.041 1.361–3.060
C. krusei Fluconazole prophylaxis 5.260 2.922–9.468
Neutropenia 5.378 2.696–10.727
C. parapsilosis Catheter-related candidemia 2.470 1.587–3.845
OR indicates odds ratio; 95% CI, 95% confidence interval.
TABLE 4
Distribution of Different Candida Species Causing Candidemia Among
Patients With Hematologic Malignancies During 2 Periods: 1988-1992
Versus 1993-2003*
Candida species
No. of patients (%)
P1988–1992, n 5230 1993–2003, n 5281
C. albicans 79 (34.4) 38 (13.5) <.0001
C. glabrata 28 (12.2) 86 (30.6) <.0001
C. krusei 17 (7.4) 68 (24.2) <.0001
C. parapsilosis 33 (14.4) 39 (13.9) .88
C. tropicalis 53 (23.0) 27 (9.6) <.0001
C. guillermondii 2 (0.9) 4 (1.4) .7
C. lusitaniae 3 (1.3) 3 (1.1) >.99
Mixed Candida species 12 (5.2) 16 (5.7) .81
* Data from Pittet 1996
34
based on a study on the epidemiology of Candidemia in 1988-1992 at the
University of Texas M. D. Anderson Cancer Center. Fluconazole was introduced in 1989 and came
into heavy prophylactic use after 1993.
2496 CANCER June 1, 2008 / Volume 112 / Number 11

mon non-albicans cause of candidemia in the former
group. C. parapsilosis is known to adhere to the sur-
face of catheters through a fibrin sheath and, thus,
cause candidemia. In general, in a patient with can-
cer who has an indwelling central venous catheter
and candidemia, the catheter should be suspected as
the source of the infection. In addition, catheter-
related candidemia may be associated with serious
complications, such as septic thrombosis, endocardi-
tis, and meningitis.
21
More than a decade ago, Abi-Said et al described
the epidemiology of candidiasis in cancer patients
between January 1, 1988 and December 31, 1992 at
our institution, the University of Texas M.D. Ander-
son Cancer Center. C. albicans was the leading cause
of candidemia followed by C. tropicalis and C. para-
psilosis, even in patients with hematologic malignan-
cies.
25
The changing epidemiology observed in this
patient population over time is consistent in the
comparisons outlined in Table 4. A significant de-
crease in the frequency of C. albicans and C. tropica-
lis was noted with a simultaneous significant
increase in C. glabrata and C. krusei with the wide
use of fluconazole in patients with leukemia and
lymphoma and in hematopoietic stem cell transplan-
tation recipients during the 1990s and 2000s (Tables
3,4).
This may raise another question regarding the
newer IDSA guidelines of recommending posacona-
zole for prophylaxis in patients undergoing allo-
geneic hematopoietic stem cell transplantation and
in high-risk acute leukemia patients. On the basis of
2 studies by Ullman et al and Cornely et al.
35,36
it was
demonstrated that posaconazole prophylaxis reduced
all-cause mortality, fungal-related mortality, and
invasive fungal infection compared with fluconazole.
This new generation of azoles, when used for pro-
phylaxis, may have an effect on Candida epidemiol-
ogy that needs to be determined.
Why the use of fluconazole did not decrease the
rate of C. parapsilosis candidemia, although the orga-
nism usually is susceptible to this antifungal agent,
is a question that may be answered by the finding
that C. parapsilosis is associated specifically with
catheter-related candidemia (Table 3). Several stud-
ied have demonstrated that C. parapsilosis organisms
adhere to the surface of indwelling central venous
catheters by embedding themselves in a biofilm
layer; and, in that biofilm matrix, they become resist-
ant to conventional antifungal agents, particularly
fluconazole.
37–40
Candidemia is associated with significant mor-
bidity and mortality. Reported mortality rates have
ranged from 30% to 75% in several published studies
of patients with and without cancer.
1,2,10,41,42
In the
current study, response to antifungal therapy was sig-
nificantly better among patients with solid tumors
than among patients with hematologic malignancies
(P.001). Regardless of the types of antifungal ther-
apy used, the worse response in patients with hema-
tologic malignancies may have been caused in part
by the severity of the infection and the status of the
underlying disease. Patients with hematologic malig-
nancies had higher APACHE II scores, were admitted
to the ICU more frequently (P5.001), were neutro-
penic more frequently during the course of their can-
didemia, and had persistent neutropenia more often
(P<.001). In addition, the overall response was bet-
ter in patients who did not receive prior antifungal
prophylaxis regardless of the treatment modality.
Hence, responses to antifungal therapy are multifac-
torial and depend on the host factors, underlying dis-
ease, severity of immunosuppression, and presence
of indwelling devices.
This study had some limitations because of its
retrospective cohort design, which did not allow
assessment of other potentially important differences
between the hematologic malignancy and solid tu-
mor groups, such as the presence of mucositis and
myocitis, the type and duration of previous or con-
comitant antibacterial-based therapy, dissemination,
and underlying disease activity. In addition, we may
not have detected patients with prior colonization or
infection with Candida spp.
In conclusion, compared with patients who had
solid tumors, patients who had hematologic malig-
nancies were at greater risk for candidemia caused
by non-albicans Candida spp., the most common of
which were C. glabrata and C. krusei. This finding
was associated with prior use of fluconazole as a
prophylactic antifungal agent. In addition, patients
who had solid tumors had better responses to anti-
fungal therapy than patients who had hematologic
malignancies; and, among all of the patients in the
study, the response was better in patients who had
not received prior prophylaxis with fluconazole.
REFERENCES
1. Pappas PG, Rex JH, Lee J, et al. A prospective observational
study of candidemia: epidemiology, therapy and influences
on mortality in hospitalized adults and pediatric patents.
Clin Infect Dis. 2003;37:634–643.
2. Wey SB, Mori M, Pfaller MA, Woolson RF, Wenzel RP. Hos-
pital-acquired candidemia. The attributable mortality and
excess length of stay. Arch Intern Med. 1988;148:2642–2645.
3. Asmundsdottir LR, Erlendsdottir H, Gottfredsson M.
Increasing incidence of candidemia: results from a 20-year
nationwide study in Iceland. J Clin Microbiol. 2002;40:
3489–3492.
Epidemiology of Candidemia in Cancer/Hachem et al. 2497