Score de candidemia
A bedside scoring system (“Candida score”) for early antifungal
treatment in nonneutropenic critically ill patients with Candida
colonization*
Cristóbal León, MD; Sergio Ruiz-Santana, MD, PhD; Pedro Saavedra, PhD; Benito Almirante, MD, PhD;
Juan Nolla-Salas, MD, PhD; Francisco Álvarez-Lerma, MD, PhD; José Garnacho-Montero, MD;
María Ángeles León, MD, PhD; EPCAN Study Group
The incidence of infections
caused by Candida species in
the critical care setting has
substantially increased in recent
years (1–3). Invasive candidiasis has
been associated with severe sepsis, septic
shock, and multiorgan failure with clinical
characteristics resembling those
caused by bacterial pathogens (4 –7).
Signs of invasive candidiasis might be
apparent early, but the disease is usually
diagnosed late in the course of intensive
care unit (ICU) stay, representing a diagnosis
challenge with an estimated mortality
rate of 40% despite the development
of new antifungal drugs (8).
Different risk factors for invasive candidiasis,
including prior Candida species
colonization, could allow recognition of
patients at highest risk. Such patients
may be potential candidates for preemptive
antifungal therapy. An important
proportion of patients are admitted or
become colonized in the ICU, but only
few subsequently develop systemic candidal
infection (9). Candida species colonization
assessment based on multiplebody-
site screening is now performed
routinely in many ICUs. The value of positive
surveillance cultures and of several
*See also p. 913.
See Appendix for list of EPCAN Study Group participants.
From the Intensive Care Unit (CL), Hospital Universitario
de Valme, Universidad de Sevilla, Sevilla; Intensive
Care Unit, Hospital Universitario Dr. Negrín (SRS),
and Mathematics Department (PS), Universidad de Las
Palmas de Gran Canaria, Las Palmas de Gran Canaria;
Infectious Diseases Unit (BA), Hospital Universitari Vall
d’Hebron, Universitat Autònoma, Barcelona; Intensive
Care Unit (JN, FAL), Hospital Universitari del Mar,
Universitat Autònoma, Barcelona; Intensive Care Unit
(JGM), Hospital Universitario Virgen del Rocío, Universidad
de Sevilla, Sevilla; and Intensive Care Unit (MAL),
Hospital General de Catalunya, Barcelona, Spain.
Supported in part by a grant from Gilead Sciences,
S.L., Madrid, Spain.
The authors declare that they have no competing
interests.
This study was presented in part at the 34th
Critical Care Congress of the Society of Critical Care
Medicine, Phoenix, Arizona, January 15–19, 2005.
Address requests for reprints to: Cristóbal León,
MD, Intensive Care Unit, Hospital Universitario de
Valme, Universidad de Sevilla, Carretera de Cádiz s/n,
E-41014, Sevilla, Spain. E-mail: cleong@telefonica.net
Copyright © 2006 by the Society of Critical Care
Medicine and Lippincott Williams & Wilkins
DOI: 10.1097/01.CCM.0000202208.37364.7D
Objective: To obtain a score for deciding early antifungal
treatment when candidal infection is suspected in nonneutropenic
critically ill patients.
Design: Analysis of data collected from the database of the
EPCAN project, an ongoing prospective, cohort, observational,
multicenter surveillance study of fungal infection and colonization
in intensive care unit (ICU) patients.
Setting: Seventy-three medical-surgical ICUs of 70 teaching
hospitals in Spain.
Patients: A total of 1,699 ICU patients aged 18 yrs and older
admitted for at least 7 days between May 1998 and January 1999
were studied.
Interventions: Surveillance cultures of urine, tracheal, and
gastric samples were obtained weekly. Patients were grouped as
follows: neither colonized nor infected (n 719), unifocal or
multifocal Candida colonization (n 883), and proven candidal
infection (n 97). The odds ratio (OR) for each risk factor
associated with colonization vs. proven candidal infection was
estimated. A logistic regression model was performed to adjust
for possible confounders. The “Candida score” was obtained
according to the logit method. The discriminatory power was
evaluated by the area under the receiver operating characteristics
curve.
Measurements and main results: In the logit model, surgery
(OR 2.71, 95% confidence interval [CI], 1.45–5.06); multifocal
colonization (OR 3.04, 95% CI, 1.45– 6.39); total parenteral
nutrition (OR 2.48, 95% CI, 1.16 –5.31); and severe sepsis (OR
7.68, 95% CI, 4.14 –14.22) were predictors of proven candidal
infection. The “Candida score” for a cut-off value of 2.5 (sensitivity
81%, specificity 74%) was as follows: parenteral nutrition,
0.908; surgery, 0.997; multifocal colonization, 1.112; and
severe sepsis, 2.038. Central venous catheters were not a
significant risk factor for proven candidal infection (p .292).
Conclusions: In a large cohort of nonneutropenic critically ill
patients in whom Candida colonization was prospectively assessed,
a “Candida score” >2.5 accurately selected patients who
would benefit from early antifungal treatment. (Crit Care Med
2006; 34:730–737)
KEY WORDS: Candida colonization; intensive care unit; critically
ill patients; Candida score; preemptive antifungal therapy; invasive
candidiasis
developed colonization indexes to predict
the risk for invasive candidiasis and to
indicate preemptive antifungal therapy is
currently a matter of active investigation
(10). Therefore, the objective of this study
was to obtain a simple scoring system
(named “Candida score”) that may assist
clinicians in differentiating between Candida
species colonization and proven candidal
infection when they are considering
preemptive antifungal treatment for nonneutropenic
critically ill patients.
METHODS
Study Population
This study was performed in the context of
the EPCAN project (Estudio de Prevalencia de
CANdidiasis, Candidiasis Prevalence Study), a
surveillance study of fungal infection and colonization
in critically ill patients (11, 12). A
total of 1,765 patients over the age of 18 yrs
who were admitted for at least 7 days to 73
medical-surgical ICUs in 70 tertiary care hospitals
in Spain between May 1998 and January
1999 were included. The study was approved
by the institutional review board of the participating
centers.
Design
This was a prospective, cohort, observational,
multicenter study. For all patients,
screening cultures for Candida species were
performed on ICU admission and once a week
thereafter until discharge from the ICU or
death. Samples were obtained from tracheal
aspirates, pharyngeal exudates, gastric aspirates,
and urine, as part of the EPCAN surveillance
study. Other samples, taken from peripheral
blood, intravascular lines, feces,
wound exudates, surgical drains, or other infectious
foci, were obtained at the discretion of
the attending physician. Samples were processed
by the different reference clinical microbiology
laboratories of the participating
hospitals according to standard procedures,
including seeding of the samples in Sabouraud
dextrose agar culture medium and Sabouraud
agar with cicloheximide and chloramphenicol
(mycobiotic agar) and incubation at 35°C for 7
days. Blood cultures were processed with an
automated system (BACTEC, Becton Dickinson
Diagnostic Instrument Systems, Paramus,
NJ). Identification of yeasts at the species level
was made with the API 20C, API 32C, or the
YST card of the Vitek system (bioMérieux España,
Madrid, Spain) whenever possible (13).
A case report form was completed for each
patient and data were prospectively included
in the EPCAN database. For the purpose of this
study—that is, to develop the “Candida score”
system for deciding the use of early antifungal
treatment when Candida colonization is diagnosed
in nonneutropenic critically ill patients
—the following data were collected from
the database: age, gender, underlying disease,
reason for ICU admission, concomitant infections,
presence and duration of risk factors for
Candida species colonization and infection,
antifungal treatment, and vital status at discharge
(survival vs. death). Neutropenia was
an exclusion criterion. Severity of illness on
ICU admission was calculated with the Acute
Physiology and Chronic Health Evaluation II
(APACHE II) system (14). According to diagnoses
at the time of ICU admission, patients
were classified as surgical, trauma, or medical.
Surgical patients were those for whom the
reason of ICU admission was the postoperative
control of an elective or urgent procedure,
trauma patients were those admitted for trauma-
related acute lesions, and medical patients
were those admitted for any other reason.
Only insulin-treated patients were considered
to have diabetes mellitus. Chronic bronchitis
was defined as the presence of a productive
cough or expectoration for 90 days a
year (although on separate days) and for 2
(consecutive) yrs, provided that a specific disorder
responsible for these symptoms was not
present. Chronic liver disease was confirmed
by liver biopsy or signs of portal hypertension,
such as esophageal varices or ascites. Chronic
renal failure was considered in patients requiring
hemodialysis or peritoneal dialysis at the
time of admission to the hospital. Severe heart
failure was defined as grades III and IV of the
New York Heart Association (NYHA) classifi-
cation (15). Other risk factors included the
following: arterial catheter, central venous
catheter, total parenteral nutrition, enteral
nutrition, urinary catheter, antibiotic treatment
(when given within 10 days before ICU
admission), extrarenal depuration procedures
(hemodialysis or continuous hemofiltration),
and use of steroids (a daily dose equivalent to
20 mg prednisone for at least 2 wks or 30 mg
for at least 1 wk before isolation of Candida in
cultures). The development of organ failure,
sepsis, and septic shock was also recorded
(16).
Definitions of Colonization and
Infection
Colonization was defined as the presence of
Candida species in nonsignificant samples obtained
from the oropharynx, stomach, urine,
or tracheal aspirates. Colonization was considered
unifocal when Candida species were isolated
from one focus and multifocal when
Candida species were simultaneously isolated
from various noncontiguous foci, even if two
different Candida species were isolated. Oropharynx
and stomach were considered one site
(digestive focus). Unifocal and multifocal colonization
persistence was defined by at least
two weekly consecutive sets of Candidapositive
cultures. Proven candidal infection
required one of the following criteria: presence
of candidemia, that is, documentation of
one blood culture that yielded a Candida species;
ophthalmic examination consistent with
candidal endophthalmitis in a patient with
clinical sepsis; isolation of Candida species in
significant samples (e.g., pleural fluid, pericardial
fluid) or candidal peritonitis; or histologically
documented candidiasis. Ophthalmic examination
was done routinely for every patient
with sepsis. Candidal peritonitis was defined
by the isolation of Candida species in a peritoneal
sample obtained by laparotomy or percutaneous
puncture in patients with associated
clinical findings, including perforation of
an abdominal organ, dehiscence of an intestinal
suture with peritonitis, severe acute pancreatitis,
or presence of a peritoneal catheter
for dialysis. Catheter-related candidemia was
considered in those patients who had an intravascular
device and one or more positive
cultures of blood samples obtained from the
peripheral vein, clinical manifestations of infection
(e.g., fever, chills, and/or hypotension),
and no apparent source for bloodstream infection
(with the exception of the catheter), as
well as a positive catheter culture, either semiquantitative
(15 colony-forming units [cfu]
per catheter segment) or quantitative (102
cfu per catheter segment), whereby the same
organism (species and susceptibility) was isolated
from a catheter segment and a peripheral
blood sample (17).
Patients were classified into three groups
as follows: neither colonized nor infected, unifocal
or multifocal Candida species colonization
without proven infection, and proven
candidal infection.
Statistical Analysis
To estimate the predictive model that will
allow us to differentiate between Candida species
and proven candidal infection, the crude
odds ratio (OR) for each risk factor associated
with colonization vs. proven candidal infection
was estimated. In order to estimate the
multivariate model, the dataset was subdivided
into two groups: a training set to fit the
model, composed of 65% of the sample, and a
validation set to validate the model, made up
of 35% of the sample (18). Based on the cases
of the training set, a logistic regression (logit)
model was performed to adjust for possible
confounders. Statistically significant variables
in the univariate analysis were included in the
model, and through a stepwise elimination
process, the so-called “Candida score” was obtained.
The discriminatory power of this score
was evaluated by the area under the receiver
operating characteristics (ROC) curve and the
95% confidence interval (CI). Then, a cut-off
value to estimate the diagnostic sensitivity and
specificity in the validation set was selected.
Statistical significance was set at p .05. Data
were analyzed with the SPSS statistical program
(11.5, SPSS, Chicago, IL) for Windows.
RESULTS
Of the initial 1,765 patients included
in the study, 96 (5.4%) were excluded
because of inadequate data collection.
The study population consisted of 1,669
patients, 66.5% men, with a mean (SD)
age of 57.8 (17.2) yrs.
There were 719 (43.1%) patients in
the neither-colonized-nor-infected
group, 67.9% men, with a mean age of
57.5 (17.0) yrs. The median (5th to 95th
percentile) APACHE II score on ICU admission
was 18 (6.6 –33). A total of 239
died, for a mortality rate of 33.2%.
Colonization solely by Candida species
was diagnosed in 883 patients. There
were 577 men and 306 women in this
group, with a mean age of 58.9 (17.0) yrs
and a median APACHE II score of 18
(8 –32.4). Unifocal Candida species colonization
was diagnosed in 388 patients
(43.9%) and multifocal Candida species
colonization in the remaining 495 patients
(56.0%). The overall mortality rate
was 40.2%. There were 103 deaths (mortality
rate, 26.5%) among patients with
unifocal Candida colonization and 252
deaths (mortality rate, 50.9%) among patients
with multifocal colonization.
Proven candidal infection was diagnosed
in 97 patients (5.8%). There were
68 men and 29 women in this group, with
a mean age of 58.5 (16.9) yrs and a median
APACHE II score of 17 (10.6 –30.8).
Fifty-eight patients developed candidemia,
30 peritonitis, 6 endophthalmitis,
and 3 candidemia and peritonitis concomitantly.
Fifty-six patients died, for a
mortality rate of 57.7%. Eighty-five patients
(87.6%) received antifungal treatment.
The median (5th to 95th percentile)
time elapsed between the onset of
proven candidal infection and the beginning
of the antifungal therapy was 12
(0.3–37.8) days. The median (5th to 95th
percentile) APACHE II score at the start
of the antifungal treatment was 18 (4.9–
29.3). Eighteen patients (18.6%) had
catheter-related candidemia, and the
catheter was removed from all of them.
There were no statistically significant
differences in the APACHE II scores between
the groups who were noncolonized,
noninfected, colonized with Candida species,
and infected by Candida species
(Kruskal-Wallis test, p .145). However,
when the risk for death was estimated (Table
1), there were statistically significant
differences between the variable indicating
patient group and the variable indicating
the mortality in the Mantel-Haenszel test
for linear association (p .001).
As shown in Table 2, patients with
candidal infection compared with those
with Candida species colonization alone
showed statistically significant differences
in the following variables: length of
ICU stay, patient category, surgery on
ICU admission, total parenteral nutrition,
extrarenal depuration procedures, unifocal
or multifocal colonization, and severe
sepsis. Central venous catheters were not
found to be a significant risk factor for
proven candidal infection (p .292).
In the logit model adjusted for possible
confounding variables, surgery on
ICU admission, total parenteral nutrition,
multifocal Candida species colonization,
and severe sepsis were independently associated
with a greater risk for proven
candidal infection (Table 3). Through a
stepwise elimination process, the Candida
score was obtained (Table 4). The
discriminatory power of this score, assessed
by the area under the ROC curve
and its main cut-off values, is shown in
Figure 1.
DISCUSSION
This study shows that the new Candida
score allows differentiating between
Candida species colonization and candidal
infection in nonneutropenic ICU patients.
Multifocal colonization, total parenteral
nutrition, surgery as the reason of
ICU admission, and clinical symptoms of
severe sepsis were found to be independent
predictors of systemic candidiasis in
this population. Accordingly, it is possible
to stratify the risk of proven candidal
infection in a large population of critically
ill patients and to select those patients
who will most benefit from starting
antifungal therapy (i.e., early antifungal
administration given to patients with evidence
of colonization in the presence of
multiple risk factors for candidal infection).
An important finding of the study is
that multifocal fungal colonization is really
an independent risk factor of proven
candidal infection in this large cohort of
both medical and surgical critically ill
patients at various centers. In the National
Epidemiology of Mycoses Survey
(NEMIS) study conducted in surgical
ICUs at six sites in the United States (19),
recovery of Candida species in rectal
and/or urine surveillance cultures was
not associated with an increased risk of
candidal bloodstream infections. The fact
that fungal colonization assessment was
based on multiple-site cultures performed
weekly in the present study could
account for the discrepant results, since
only two sites were cultured in the
NEMIS study.
Nosocomial fungal infections in nonneutropenic
critically ill patients are
caused by mainly Candida species. The
proposed definitions of “probable,” “possible,”
and “proven” opportunistic fungal
infections intended for immunocompromised
patients (20) may be unreliable for
nonneutropenic patients (21). The clinical
significance of Candida species colonization
as a determinant risk factor for
invasive candidiasis has been largely recognized,
and recent efforts have been directed
toward developing a predictor for
the diagnosis of systemic infection based
on colonization density. A colonization
index with a 0.5 threshold, defined as the
ratio of the number of culture-positive
Table 1. Risk for death in the study population, according to colonization and infection status (n
1,669)
Patient group
Nonsurvivors
Odds Ratio (95% Confidence
Interval)a
No.
Mortality
Rateb Crude Adjustedc
Neither colonized nor infected, n 719 239 33.2% 1 1
Candida species colonization, n 883
Unifocal, n 388 103 26.5% 1.02 (0.8–1.4) 1.04 (0.8–1.4)
Multifocal, n 495 252 50.9% 1.55 (1.3–2) 1.54 (1.2–1.9)
Candidal infection, n 97 56 57.7% 2.74 (1.8–4.2) 3.2 (2.0–5.0)
aEstimated by logistic regression analysis; bp .001, linear association test; cfor Acute Physiology
and Chronic Health Evaluation (APACHE II) score.
sites to the number of sites cultured, and its corrected version with a 0.4 threshold
(22) have been used as tools to start preemptive
antifungal treatment in ICU patients
(10). In a before/after intervention
study of 2-yr prospective and 2-yr historical-
control cohorts carried out by Piarroux
et al. (10), patients with a corrected
colonization index 0.4 received early
preemptive antifungal therapy, and only
18 cases (3.8%) of proven candidiasis
were diagnosed; all were imported infections.
The incidence of ICU-acquired
proven candidiasis significantly decreased
from 2.2% to 0% (p .001, Fisher test).
The authors concluded that targeted preemptive
strategy may efficiently prevent
acquisition of proven candidiasis in patients
admitted to a surgical ICU. It
should be noted that the Candida score
takes in account other relevant risk factors
of candidiasis, in addition to colonization,
to improve the specificity of the
test.
For patients considered “heavily” colonized
by Candida species, there are no
biological markers that may assist clinicians
in deciding to prescribe or not prescribe
antifungal agents. According to the
results of a survey in medical-surgical
ICUs in France, most of the units showed
a homogeneous antifungal prescription
pattern. Furthermore, most intensivists
interviewed prescribed antifungal treatment
in the presence of multifocal Candida
colonization, clinical signs of sepsis,
and several other risk factors for invasive
candidiasis (23). In agreement with these
Table 3. Results of multivariate analysis: Risk factors for proven candidal infection in 1,669 adult
patients
Variable
Proven
Candidal
Infection
% p Value
Crude Odds
Ratio (95%
Confidence
Interval)
Adjusted Odds Ratio
(95% Confidence
Interval)
Surgery on ICU admission
No 6.9
Yes 16.5 .001 2.69 (1.76–4.10) 2.71 (1.45–5.06)
Total parenteral nutrition
No 2.8
Yes 15.5 .001 6.46 (3.48–11.98) 2.48 (1.16–5.31)
Severe sepsis
No 4.5
Yes 28.8 .001 8.63 (5.49–13.56) 7.68 (4.14–14.22)
Candida species colonization
No 4.2
Yes 12.3 .001 3.20 (1.85–5.53) 3.04 (1.45–6.39)
ICU, intensive care unit.
Table 2. Results of univariate analysis: Risk factors for invasive candidiasis according to colonization and infection status (n 1,669)
Variable
Unifocal or Multifocal Candida
Species Colonization n 883
Proven Candidal
Infection n 97 p Value
Age, yrs, mean (SD) 58.9 (17.0) 58.5 (16.9) .825
Male/female 577/306 68/29 .337
APACHE II score on admission, median (range) 19 (1–67) 17 (6–45) .203
Length of ICU stay, days, median (range) 20 (7–166) 28 (7–138) .001
APACHE II score, no. (%)
15 302 (34.2) 37 (38.5) .137
15–25 408 (46.3) 48 (50.5)
25 173 (19.5) 12 (11.0)
Diagnosis on ICU admission, no. (%)
Medical 449 (50.8) 34 (35.1) .001
Surgical 258 (29.2) 51 (52.6)
Trauma 176 (19.9) 12 (12.4)
Underlying disease, no. (%)
Chronic bronchitis 197 (22.3) 14 (14.4) .073
Diabetes mellitus 136 (15.4) 14 (14.4) .801
Chronic liver disease 40 (4.5) 2 (2.1) .255
Chronic renal failure 44 (5.5) 4 (4.1) .710
Heart failure 40 (4.5) 2 (2.1) .255
Risk factors, no (%)
Broad spectrum antibiotics 866 (98.0) 97 (100) .380
Central venous catheter 873 (98.9) 97 (100) .292
Urinary catheter 870 (98.5) 93 (95.9) .078
Mechanical ventilation 837 (94.8) 92 (94.8) .982
Enteral nutrition 695 (78.7) 68 (70.1) .053
Arterial catheter 666 (75.4) 68 (70.1) .251
Total parenteral nutrition 462 (52.3) 85 (87.6) .001
Corticosteroids 214 (24.2) 22 (22.7) .734
Hemodialysis or continuous hemofiltration 106 (12.0) 29 (29.9) .001
Severe sepsis, no. (%) 156 (17.7) 63 (64.9) .001
Candida species colonization, no. (%) .001
Unifocal 390 (44.1) 17 (17.5)a
Multifocal 493 (55.8) 69 (71.1)a
APACHE II, Acute Physiology and Chronic Health Evaluation II; ICU, intensive care unit.
aEleven patients in the proven candidal infection group did not have previous Candida colonization.
733 Crit Care Med 2006 Vol. 34, No. 3
data, 79% of 135 Spanish intensivists in
45 ICUs reported that they would start
antifungal treatment for nonneutropenic
critically ill patients if clinical signs of
infection and multifocal Candida isolates
were noted (24).
Recommendations for starting antifungal
treatment for nonneutropenic
critically ill patients have also recently
been reported in the literature (25–27). A
predictive rule based on known risk factors,
including colonization, that allow us
to differentiate between Candida species
colonization and proven candidal infection
would help clinicians more accurately
than colonization alone to select
those ICU patients who would benefit
from early antifungal treatment. Paphitou
and associates (28) and Ostrosky-
Zeichner and colleagues (29) have proposed
prediction rules for invasive
candidiasis following a retrospective multicenter
study in 12 ICUs at nine hospitals
in the United States and Brazil. The bestperforming
rule required a combination
of at least one “major” and at least two
“minor” risk factors among several
known for candidal infection in patients
staying in the ICU for at least 48 hrs and
who were expected to stay for 2 more
days. Their clinical prediction rule identified
ICU patients with a 10% risk of
invasive candidiasis, but validation of this
instrument is pending. Fungal coloniza-
Table 4. Calculation of the Candida score: Variables selected in the logistic regression model
Variable
Coefficient
()
Standard
Error Wald 2
p
Value
Multifocal Candida species colonization 1.112 .379 8.625 .003
Surgery on ICU admission .997 .319 9.761 .002
Severe sepsis 2.038 .314 42.014 .000
Total parenteral nutrition .908 .389 5.451 .020
Constant 4.916 .485 102.732 .000
ICU, intensive care unit.
Candida score .908 (total parenteral nutrition) .997 (surgery) 1.112 (multifocal
Candida species colonization) 2.038 (severe sepsis). Candida score (rounded) 1 (total
parenteral nutrition) 1 (surgery) 1 (multifocal Candida species colonization) 2 (severe
sepsis). All variables coded as follows: absent, 0; present, 1.
Figure 1. A, receiver operating characteristics (ROC) curve and area under the ROC curve (AUC) for assessing the discriminatory power of the Candida score.
B, cut-off values for the ROC curve.
734 Crit Care Med 2006 Vol. 34, No. 3
tion was not included in this prediction
rule proposed by Ostrosky-Zeichner et al.
(29) because it was derived from the
NEMIS study results (19). The Candida
score presented here could therefore be
considered to be more reliable, given the
weight of fungal colonization in the
pathogenesis of candidiasis.
DuPont and co-workers (30) carried
out a retrospective systematic review of
surgical intensive care patients, with a
prospective follow-up in France. A scoring
system was proposed with the following
risk factors: female gender, upper
gastrointestinal origin of peritonitis, cardiovascular
failure, and use of antibiotics.
A grade C score, defined as the presence
of three qualifiers, was associated with a
sensitivity of 84% and specificity of 50%
for the detection of yeasts in the peritoneal
fluid of patients with peritonitis. The
main drawbacks of this study included its
single-center setting and its potential application
restriction to surgical patients.
We used the EPCAN database, which is
a large cohort of nonneutropenic ICU patients
for whom, among other goals, Candida
colonization and invasive candidiasis
were studied prospectively. As previously
reported (21), the rate of proven candidal
infection found in the present study was
low, but the mortality rate was high. The
mortality rate increased significantly according
to the patient group, that is, with
unifocal colonization (26.5%), multifocal
colonization (50.9%), and candidal
proven infection (57.7%). Although colonization
does not define infection, these
data support the well-known role of Candida
colonization as a key factor in the
decision to start early antifungal treatment
for ICU patients.
The new Candida score was based on
the respective predictive value of previously
reported risk factors. In addition to
multifocal Candida species colonization,
three other risk factors were found to be
significant predictors of proven candidal
infection in the logistic regression model:
use of total parenteral nutrition, surgery
on ICU admission, and clinical manifestations
of severe sepsis. The respective
weight of colonization and these risk factors
as shown in the Candida score allowed
us to reliably differentiate between
Candida species colonization and proven
candidal infection. Although central venous
catheters are repeatedly described
as major risk factors for proven hematogenous
candidiasis (19, 31), in this large
dataset from a prospective multicenter
study, venous catheters were not signifi-
cant predictors of proven candidal infection.
The medical literature is flooded with
complicated prediction rules and scores
(32–37), and there is a need to have available
bedside easy-to-remember scores
that would make daily tasks easier for
clinicians. The simplified version of this
score, after rounding up to 1 the weight
for total parenteral nutrition, surgery, or
multifocal Candida species colonization
and up to 2 the weight for clinical severe
sepsis, is a quite simple ready-to-use prediction
rule. With a cut-off value of 2.5,
that it to say, with a sensitivity of 81%
and a specificity of 74%, we shall only
need the presence of sepsis and any one of
the three other remaining risk factors or
the presence of all of them together except
sepsis in order to consider starting
antifungal treatment for one particular
patient. Finally, the Candida score also
identifies critically ill patients with
proven candidiasis: patients with a score
2.5 are 7.75 times as likely to have
proven infection (risk ratio 7.75; 95%
CI, 4.74 –12.66) than patients with a Candida
score up to 2.5.
Therefore, the easy rule of thumb for
prescribing antifungals, according to a
Candida score 2.5, will allow more effi-
cient selection of patients who indeed will
benefit from the increasing number of
available antifungal drugs (38) and, at the
same time, more adequate prevention of
the development of new resistant species
due to an excess of inappropriate and
potentially detrimental antifungal treatments
(39). Assessment with the Candida
score should be performed at the time of
ICU admission and any time candidiasis is
suspected.
CONCLUSIONS
A new score, the Candida score, which
was calculated according to data collected
in the EPCAN database (in which all cases
of Candida species colonization and
proven candidal infection were prospectively
recorded), is an easy-to-remember
bedside prediction rule. A score 2.5 will
help intensivists select patients who will
benefit from early antifungal administration.
Finally, although the Candida score
contributes to predicting proven candidal
infection, the benefits of preemptive (prophylactic
or empirical) antifungal therapy
remain to be determined.
ACKNOWLEDGMENT
We thank Marta Pulido, MD, for editing
the manuscript.
REFERENCES
1. Vincent JL, Anaissie E, Bruining H, et al:
Epidemiology, diagnosis and treatment of
systemic Candida infection in surgical patients
under intensive care. Intensive Care
Med 1998; 24:206–216
2. Sobel JD, Vazquez J: Candidiasis in the intensive
care unit. Semin Respir Crit Care
Med 2003; 24:99–111
3. Nolla-Salas J, Sitges-Serra A, Leon-Gil C, et
al: Candidemia in non-neutropenic critically
ill patients: Analysis of prognostic factors and
assessment of systemic antifungal therapy.
Intensive Care Med 1997; 23:23–30
4. Eggimann P, Garbino J, Pittet D: Epidemiology
of Candida species infections in critically
ill non-immunosuppressed patients. Lancet
Infect Dis 2003; 3:685–702
5. Charles PE, Doise JM, Quenot J, et al: Candidemia
in critically ill patients: Difference of
outcome between medical and surgical patients.
Intensive Care Med 2003; 29:
2162–2169
6. Debusk CH, Daoud R, Thirumoorthi MC, et
al: Candidemia: current epidemiologic characteristics
and a long-term follow-up of the
survivors. Scand J Infect Dis 1994; 26:
697–703
7. Leleu G, Aegerter P, Guidet B, for Collège des
Utilisateurs de Base de Données en Réanimation:
Systemic candidiasis in intensive care
units: A multicenter matched-cohort study. J
Critical Care 2002;17:168–175
8. Ostrosky-Zeichner, L: Prophylaxis or preemptive
therapy of invasive candidiasis in the
intensive care unit? Crit Care Med 2004; 32:
2552–2553
9. Álvarez-Lerma F, Palomar M, León C, et al:
Colonización y/o infección por hongos en
unidades de cuidados intensivos: Estudio
multicéntrico de 1.562 pacientes. Med Clin
(Barc) 2003; 121:161–166
10. Piarroux R, Grenouillet F, Balvay P, et al:
Assessment of preemptive treatment to prevent
severe candidiasis in critically ill surgical
patients. Crit Care Med 2004; 32:
2443–2449
11. Alvarez-Lerma F, Nolla-Salas J, Leon C, et al,
A score 2.5 will
help intensivists
select patients who
will benefit from early antifungal
administration.
treatment in nonneutropenic critically ill patients with Candida
colonization*
Cristóbal León, MD; Sergio Ruiz-Santana, MD, PhD; Pedro Saavedra, PhD; Benito Almirante, MD, PhD;
Juan Nolla-Salas, MD, PhD; Francisco Álvarez-Lerma, MD, PhD; José Garnacho-Montero, MD;
María Ángeles León, MD, PhD; EPCAN Study Group
The incidence of infections
caused by Candida species in
the critical care setting has
substantially increased in recent
years (1–3). Invasive candidiasis has
been associated with severe sepsis, septic
shock, and multiorgan failure with clinical
characteristics resembling those
caused by bacterial pathogens (4 –7).
Signs of invasive candidiasis might be
apparent early, but the disease is usually
diagnosed late in the course of intensive
care unit (ICU) stay, representing a diagnosis
challenge with an estimated mortality
rate of 40% despite the development
of new antifungal drugs (8).
Different risk factors for invasive candidiasis,
including prior Candida species
colonization, could allow recognition of
patients at highest risk. Such patients
may be potential candidates for preemptive
antifungal therapy. An important
proportion of patients are admitted or
become colonized in the ICU, but only
few subsequently develop systemic candidal
infection (9). Candida species colonization
assessment based on multiplebody-
site screening is now performed
routinely in many ICUs. The value of positive
surveillance cultures and of several
*See also p. 913.
See Appendix for list of EPCAN Study Group participants.
From the Intensive Care Unit (CL), Hospital Universitario
de Valme, Universidad de Sevilla, Sevilla; Intensive
Care Unit, Hospital Universitario Dr. Negrín (SRS),
and Mathematics Department (PS), Universidad de Las
Palmas de Gran Canaria, Las Palmas de Gran Canaria;
Infectious Diseases Unit (BA), Hospital Universitari Vall
d’Hebron, Universitat Autònoma, Barcelona; Intensive
Care Unit (JN, FAL), Hospital Universitari del Mar,
Universitat Autònoma, Barcelona; Intensive Care Unit
(JGM), Hospital Universitario Virgen del Rocío, Universidad
de Sevilla, Sevilla; and Intensive Care Unit (MAL),
Hospital General de Catalunya, Barcelona, Spain.
Supported in part by a grant from Gilead Sciences,
S.L., Madrid, Spain.
The authors declare that they have no competing
interests.
This study was presented in part at the 34th
Critical Care Congress of the Society of Critical Care
Medicine, Phoenix, Arizona, January 15–19, 2005.
Address requests for reprints to: Cristóbal León,
MD, Intensive Care Unit, Hospital Universitario de
Valme, Universidad de Sevilla, Carretera de Cádiz s/n,
E-41014, Sevilla, Spain. E-mail: cleong@telefonica.net
Copyright © 2006 by the Society of Critical Care
Medicine and Lippincott Williams & Wilkins
DOI: 10.1097/01.CCM.0000202208.37364.7D
Objective: To obtain a score for deciding early antifungal
treatment when candidal infection is suspected in nonneutropenic
critically ill patients.
Design: Analysis of data collected from the database of the
EPCAN project, an ongoing prospective, cohort, observational,
multicenter surveillance study of fungal infection and colonization
in intensive care unit (ICU) patients.
Setting: Seventy-three medical-surgical ICUs of 70 teaching
hospitals in Spain.
Patients: A total of 1,699 ICU patients aged 18 yrs and older
admitted for at least 7 days between May 1998 and January 1999
were studied.
Interventions: Surveillance cultures of urine, tracheal, and
gastric samples were obtained weekly. Patients were grouped as
follows: neither colonized nor infected (n 719), unifocal or
multifocal Candida colonization (n 883), and proven candidal
infection (n 97). The odds ratio (OR) for each risk factor
associated with colonization vs. proven candidal infection was
estimated. A logistic regression model was performed to adjust
for possible confounders. The “Candida score” was obtained
according to the logit method. The discriminatory power was
evaluated by the area under the receiver operating characteristics
curve.
Measurements and main results: In the logit model, surgery
(OR 2.71, 95% confidence interval [CI], 1.45–5.06); multifocal
colonization (OR 3.04, 95% CI, 1.45– 6.39); total parenteral
nutrition (OR 2.48, 95% CI, 1.16 –5.31); and severe sepsis (OR
7.68, 95% CI, 4.14 –14.22) were predictors of proven candidal
infection. The “Candida score” for a cut-off value of 2.5 (sensitivity
81%, specificity 74%) was as follows: parenteral nutrition,
0.908; surgery, 0.997; multifocal colonization, 1.112; and
severe sepsis, 2.038. Central venous catheters were not a
significant risk factor for proven candidal infection (p .292).
Conclusions: In a large cohort of nonneutropenic critically ill
patients in whom Candida colonization was prospectively assessed,
a “Candida score” >2.5 accurately selected patients who
would benefit from early antifungal treatment. (Crit Care Med
2006; 34:730–737)
KEY WORDS: Candida colonization; intensive care unit; critically
ill patients; Candida score; preemptive antifungal therapy; invasive
candidiasis
developed colonization indexes to predict
the risk for invasive candidiasis and to
indicate preemptive antifungal therapy is
currently a matter of active investigation
(10). Therefore, the objective of this study
was to obtain a simple scoring system
(named “Candida score”) that may assist
clinicians in differentiating between Candida
species colonization and proven candidal
infection when they are considering
preemptive antifungal treatment for nonneutropenic
critically ill patients.
METHODS
Study Population
This study was performed in the context of
the EPCAN project (Estudio de Prevalencia de
CANdidiasis, Candidiasis Prevalence Study), a
surveillance study of fungal infection and colonization
in critically ill patients (11, 12). A
total of 1,765 patients over the age of 18 yrs
who were admitted for at least 7 days to 73
medical-surgical ICUs in 70 tertiary care hospitals
in Spain between May 1998 and January
1999 were included. The study was approved
by the institutional review board of the participating
centers.
Design
This was a prospective, cohort, observational,
multicenter study. For all patients,
screening cultures for Candida species were
performed on ICU admission and once a week
thereafter until discharge from the ICU or
death. Samples were obtained from tracheal
aspirates, pharyngeal exudates, gastric aspirates,
and urine, as part of the EPCAN surveillance
study. Other samples, taken from peripheral
blood, intravascular lines, feces,
wound exudates, surgical drains, or other infectious
foci, were obtained at the discretion of
the attending physician. Samples were processed
by the different reference clinical microbiology
laboratories of the participating
hospitals according to standard procedures,
including seeding of the samples in Sabouraud
dextrose agar culture medium and Sabouraud
agar with cicloheximide and chloramphenicol
(mycobiotic agar) and incubation at 35°C for 7
days. Blood cultures were processed with an
automated system (BACTEC, Becton Dickinson
Diagnostic Instrument Systems, Paramus,
NJ). Identification of yeasts at the species level
was made with the API 20C, API 32C, or the
YST card of the Vitek system (bioMérieux España,
Madrid, Spain) whenever possible (13).
A case report form was completed for each
patient and data were prospectively included
in the EPCAN database. For the purpose of this
study—that is, to develop the “Candida score”
system for deciding the use of early antifungal
treatment when Candida colonization is diagnosed
in nonneutropenic critically ill patients
—the following data were collected from
the database: age, gender, underlying disease,
reason for ICU admission, concomitant infections,
presence and duration of risk factors for
Candida species colonization and infection,
antifungal treatment, and vital status at discharge
(survival vs. death). Neutropenia was
an exclusion criterion. Severity of illness on
ICU admission was calculated with the Acute
Physiology and Chronic Health Evaluation II
(APACHE II) system (14). According to diagnoses
at the time of ICU admission, patients
were classified as surgical, trauma, or medical.
Surgical patients were those for whom the
reason of ICU admission was the postoperative
control of an elective or urgent procedure,
trauma patients were those admitted for trauma-
related acute lesions, and medical patients
were those admitted for any other reason.
Only insulin-treated patients were considered
to have diabetes mellitus. Chronic bronchitis
was defined as the presence of a productive
cough or expectoration for 90 days a
year (although on separate days) and for 2
(consecutive) yrs, provided that a specific disorder
responsible for these symptoms was not
present. Chronic liver disease was confirmed
by liver biopsy or signs of portal hypertension,
such as esophageal varices or ascites. Chronic
renal failure was considered in patients requiring
hemodialysis or peritoneal dialysis at the
time of admission to the hospital. Severe heart
failure was defined as grades III and IV of the
New York Heart Association (NYHA) classifi-
cation (15). Other risk factors included the
following: arterial catheter, central venous
catheter, total parenteral nutrition, enteral
nutrition, urinary catheter, antibiotic treatment
(when given within 10 days before ICU
admission), extrarenal depuration procedures
(hemodialysis or continuous hemofiltration),
and use of steroids (a daily dose equivalent to
20 mg prednisone for at least 2 wks or 30 mg
for at least 1 wk before isolation of Candida in
cultures). The development of organ failure,
sepsis, and septic shock was also recorded
(16).
Definitions of Colonization and
Infection
Colonization was defined as the presence of
Candida species in nonsignificant samples obtained
from the oropharynx, stomach, urine,
or tracheal aspirates. Colonization was considered
unifocal when Candida species were isolated
from one focus and multifocal when
Candida species were simultaneously isolated
from various noncontiguous foci, even if two
different Candida species were isolated. Oropharynx
and stomach were considered one site
(digestive focus). Unifocal and multifocal colonization
persistence was defined by at least
two weekly consecutive sets of Candidapositive
cultures. Proven candidal infection
required one of the following criteria: presence
of candidemia, that is, documentation of
one blood culture that yielded a Candida species;
ophthalmic examination consistent with
candidal endophthalmitis in a patient with
clinical sepsis; isolation of Candida species in
significant samples (e.g., pleural fluid, pericardial
fluid) or candidal peritonitis; or histologically
documented candidiasis. Ophthalmic examination
was done routinely for every patient
with sepsis. Candidal peritonitis was defined
by the isolation of Candida species in a peritoneal
sample obtained by laparotomy or percutaneous
puncture in patients with associated
clinical findings, including perforation of
an abdominal organ, dehiscence of an intestinal
suture with peritonitis, severe acute pancreatitis,
or presence of a peritoneal catheter
for dialysis. Catheter-related candidemia was
considered in those patients who had an intravascular
device and one or more positive
cultures of blood samples obtained from the
peripheral vein, clinical manifestations of infection
(e.g., fever, chills, and/or hypotension),
and no apparent source for bloodstream infection
(with the exception of the catheter), as
well as a positive catheter culture, either semiquantitative
(15 colony-forming units [cfu]
per catheter segment) or quantitative (102
cfu per catheter segment), whereby the same
organism (species and susceptibility) was isolated
from a catheter segment and a peripheral
blood sample (17).
Patients were classified into three groups
as follows: neither colonized nor infected, unifocal
or multifocal Candida species colonization
without proven infection, and proven
candidal infection.
Statistical Analysis
To estimate the predictive model that will
allow us to differentiate between Candida species
and proven candidal infection, the crude
odds ratio (OR) for each risk factor associated
with colonization vs. proven candidal infection
was estimated. In order to estimate the
multivariate model, the dataset was subdivided
into two groups: a training set to fit the
model, composed of 65% of the sample, and a
validation set to validate the model, made up
of 35% of the sample (18). Based on the cases
of the training set, a logistic regression (logit)
model was performed to adjust for possible
confounders. Statistically significant variables
in the univariate analysis were included in the
model, and through a stepwise elimination
process, the so-called “Candida score” was obtained.
The discriminatory power of this score
was evaluated by the area under the receiver
operating characteristics (ROC) curve and the
95% confidence interval (CI). Then, a cut-off
value to estimate the diagnostic sensitivity and
specificity in the validation set was selected.
Statistical significance was set at p .05. Data
were analyzed with the SPSS statistical program
(11.5, SPSS, Chicago, IL) for Windows.
RESULTS
Of the initial 1,765 patients included
in the study, 96 (5.4%) were excluded
because of inadequate data collection.
The study population consisted of 1,669
patients, 66.5% men, with a mean (SD)
age of 57.8 (17.2) yrs.
There were 719 (43.1%) patients in
the neither-colonized-nor-infected
group, 67.9% men, with a mean age of
57.5 (17.0) yrs. The median (5th to 95th
percentile) APACHE II score on ICU admission
was 18 (6.6 –33). A total of 239
died, for a mortality rate of 33.2%.
Colonization solely by Candida species
was diagnosed in 883 patients. There
were 577 men and 306 women in this
group, with a mean age of 58.9 (17.0) yrs
and a median APACHE II score of 18
(8 –32.4). Unifocal Candida species colonization
was diagnosed in 388 patients
(43.9%) and multifocal Candida species
colonization in the remaining 495 patients
(56.0%). The overall mortality rate
was 40.2%. There were 103 deaths (mortality
rate, 26.5%) among patients with
unifocal Candida colonization and 252
deaths (mortality rate, 50.9%) among patients
with multifocal colonization.
Proven candidal infection was diagnosed
in 97 patients (5.8%). There were
68 men and 29 women in this group, with
a mean age of 58.5 (16.9) yrs and a median
APACHE II score of 17 (10.6 –30.8).
Fifty-eight patients developed candidemia,
30 peritonitis, 6 endophthalmitis,
and 3 candidemia and peritonitis concomitantly.
Fifty-six patients died, for a
mortality rate of 57.7%. Eighty-five patients
(87.6%) received antifungal treatment.
The median (5th to 95th percentile)
time elapsed between the onset of
proven candidal infection and the beginning
of the antifungal therapy was 12
(0.3–37.8) days. The median (5th to 95th
percentile) APACHE II score at the start
of the antifungal treatment was 18 (4.9–
29.3). Eighteen patients (18.6%) had
catheter-related candidemia, and the
catheter was removed from all of them.
There were no statistically significant
differences in the APACHE II scores between
the groups who were noncolonized,
noninfected, colonized with Candida species,
and infected by Candida species
(Kruskal-Wallis test, p .145). However,
when the risk for death was estimated (Table
1), there were statistically significant
differences between the variable indicating
patient group and the variable indicating
the mortality in the Mantel-Haenszel test
for linear association (p .001).
As shown in Table 2, patients with
candidal infection compared with those
with Candida species colonization alone
showed statistically significant differences
in the following variables: length of
ICU stay, patient category, surgery on
ICU admission, total parenteral nutrition,
extrarenal depuration procedures, unifocal
or multifocal colonization, and severe
sepsis. Central venous catheters were not
found to be a significant risk factor for
proven candidal infection (p .292).
In the logit model adjusted for possible
confounding variables, surgery on
ICU admission, total parenteral nutrition,
multifocal Candida species colonization,
and severe sepsis were independently associated
with a greater risk for proven
candidal infection (Table 3). Through a
stepwise elimination process, the Candida
score was obtained (Table 4). The
discriminatory power of this score, assessed
by the area under the ROC curve
and its main cut-off values, is shown in
Figure 1.
DISCUSSION
This study shows that the new Candida
score allows differentiating between
Candida species colonization and candidal
infection in nonneutropenic ICU patients.
Multifocal colonization, total parenteral
nutrition, surgery as the reason of
ICU admission, and clinical symptoms of
severe sepsis were found to be independent
predictors of systemic candidiasis in
this population. Accordingly, it is possible
to stratify the risk of proven candidal
infection in a large population of critically
ill patients and to select those patients
who will most benefit from starting
antifungal therapy (i.e., early antifungal
administration given to patients with evidence
of colonization in the presence of
multiple risk factors for candidal infection).
An important finding of the study is
that multifocal fungal colonization is really
an independent risk factor of proven
candidal infection in this large cohort of
both medical and surgical critically ill
patients at various centers. In the National
Epidemiology of Mycoses Survey
(NEMIS) study conducted in surgical
ICUs at six sites in the United States (19),
recovery of Candida species in rectal
and/or urine surveillance cultures was
not associated with an increased risk of
candidal bloodstream infections. The fact
that fungal colonization assessment was
based on multiple-site cultures performed
weekly in the present study could
account for the discrepant results, since
only two sites were cultured in the
NEMIS study.
Nosocomial fungal infections in nonneutropenic
critically ill patients are
caused by mainly Candida species. The
proposed definitions of “probable,” “possible,”
and “proven” opportunistic fungal
infections intended for immunocompromised
patients (20) may be unreliable for
nonneutropenic patients (21). The clinical
significance of Candida species colonization
as a determinant risk factor for
invasive candidiasis has been largely recognized,
and recent efforts have been directed
toward developing a predictor for
the diagnosis of systemic infection based
on colonization density. A colonization
index with a 0.5 threshold, defined as the
ratio of the number of culture-positive
Table 1. Risk for death in the study population, according to colonization and infection status (n
1,669)
Patient group
Nonsurvivors
Odds Ratio (95% Confidence
Interval)a
No.
Mortality
Rateb Crude Adjustedc
Neither colonized nor infected, n 719 239 33.2% 1 1
Candida species colonization, n 883
Unifocal, n 388 103 26.5% 1.02 (0.8–1.4) 1.04 (0.8–1.4)
Multifocal, n 495 252 50.9% 1.55 (1.3–2) 1.54 (1.2–1.9)
Candidal infection, n 97 56 57.7% 2.74 (1.8–4.2) 3.2 (2.0–5.0)
aEstimated by logistic regression analysis; bp .001, linear association test; cfor Acute Physiology
and Chronic Health Evaluation (APACHE II) score.
sites to the number of sites cultured, and its corrected version with a 0.4 threshold
(22) have been used as tools to start preemptive
antifungal treatment in ICU patients
(10). In a before/after intervention
study of 2-yr prospective and 2-yr historical-
control cohorts carried out by Piarroux
et al. (10), patients with a corrected
colonization index 0.4 received early
preemptive antifungal therapy, and only
18 cases (3.8%) of proven candidiasis
were diagnosed; all were imported infections.
The incidence of ICU-acquired
proven candidiasis significantly decreased
from 2.2% to 0% (p .001, Fisher test).
The authors concluded that targeted preemptive
strategy may efficiently prevent
acquisition of proven candidiasis in patients
admitted to a surgical ICU. It
should be noted that the Candida score
takes in account other relevant risk factors
of candidiasis, in addition to colonization,
to improve the specificity of the
test.
For patients considered “heavily” colonized
by Candida species, there are no
biological markers that may assist clinicians
in deciding to prescribe or not prescribe
antifungal agents. According to the
results of a survey in medical-surgical
ICUs in France, most of the units showed
a homogeneous antifungal prescription
pattern. Furthermore, most intensivists
interviewed prescribed antifungal treatment
in the presence of multifocal Candida
colonization, clinical signs of sepsis,
and several other risk factors for invasive
candidiasis (23). In agreement with these
Table 3. Results of multivariate analysis: Risk factors for proven candidal infection in 1,669 adult
patients
Variable
Proven
Candidal
Infection
% p Value
Crude Odds
Ratio (95%
Confidence
Interval)
Adjusted Odds Ratio
(95% Confidence
Interval)
Surgery on ICU admission
No 6.9
Yes 16.5 .001 2.69 (1.76–4.10) 2.71 (1.45–5.06)
Total parenteral nutrition
No 2.8
Yes 15.5 .001 6.46 (3.48–11.98) 2.48 (1.16–5.31)
Severe sepsis
No 4.5
Yes 28.8 .001 8.63 (5.49–13.56) 7.68 (4.14–14.22)
Candida species colonization
No 4.2
Yes 12.3 .001 3.20 (1.85–5.53) 3.04 (1.45–6.39)
ICU, intensive care unit.
Table 2. Results of univariate analysis: Risk factors for invasive candidiasis according to colonization and infection status (n 1,669)
Variable
Unifocal or Multifocal Candida
Species Colonization n 883
Proven Candidal
Infection n 97 p Value
Age, yrs, mean (SD) 58.9 (17.0) 58.5 (16.9) .825
Male/female 577/306 68/29 .337
APACHE II score on admission, median (range) 19 (1–67) 17 (6–45) .203
Length of ICU stay, days, median (range) 20 (7–166) 28 (7–138) .001
APACHE II score, no. (%)
15 302 (34.2) 37 (38.5) .137
15–25 408 (46.3) 48 (50.5)
25 173 (19.5) 12 (11.0)
Diagnosis on ICU admission, no. (%)
Medical 449 (50.8) 34 (35.1) .001
Surgical 258 (29.2) 51 (52.6)
Trauma 176 (19.9) 12 (12.4)
Underlying disease, no. (%)
Chronic bronchitis 197 (22.3) 14 (14.4) .073
Diabetes mellitus 136 (15.4) 14 (14.4) .801
Chronic liver disease 40 (4.5) 2 (2.1) .255
Chronic renal failure 44 (5.5) 4 (4.1) .710
Heart failure 40 (4.5) 2 (2.1) .255
Risk factors, no (%)
Broad spectrum antibiotics 866 (98.0) 97 (100) .380
Central venous catheter 873 (98.9) 97 (100) .292
Urinary catheter 870 (98.5) 93 (95.9) .078
Mechanical ventilation 837 (94.8) 92 (94.8) .982
Enteral nutrition 695 (78.7) 68 (70.1) .053
Arterial catheter 666 (75.4) 68 (70.1) .251
Total parenteral nutrition 462 (52.3) 85 (87.6) .001
Corticosteroids 214 (24.2) 22 (22.7) .734
Hemodialysis or continuous hemofiltration 106 (12.0) 29 (29.9) .001
Severe sepsis, no. (%) 156 (17.7) 63 (64.9) .001
Candida species colonization, no. (%) .001
Unifocal 390 (44.1) 17 (17.5)a
Multifocal 493 (55.8) 69 (71.1)a
APACHE II, Acute Physiology and Chronic Health Evaluation II; ICU, intensive care unit.
aEleven patients in the proven candidal infection group did not have previous Candida colonization.
733 Crit Care Med 2006 Vol. 34, No. 3
data, 79% of 135 Spanish intensivists in
45 ICUs reported that they would start
antifungal treatment for nonneutropenic
critically ill patients if clinical signs of
infection and multifocal Candida isolates
were noted (24).
Recommendations for starting antifungal
treatment for nonneutropenic
critically ill patients have also recently
been reported in the literature (25–27). A
predictive rule based on known risk factors,
including colonization, that allow us
to differentiate between Candida species
colonization and proven candidal infection
would help clinicians more accurately
than colonization alone to select
those ICU patients who would benefit
from early antifungal treatment. Paphitou
and associates (28) and Ostrosky-
Zeichner and colleagues (29) have proposed
prediction rules for invasive
candidiasis following a retrospective multicenter
study in 12 ICUs at nine hospitals
in the United States and Brazil. The bestperforming
rule required a combination
of at least one “major” and at least two
“minor” risk factors among several
known for candidal infection in patients
staying in the ICU for at least 48 hrs and
who were expected to stay for 2 more
days. Their clinical prediction rule identified
ICU patients with a 10% risk of
invasive candidiasis, but validation of this
instrument is pending. Fungal coloniza-
Table 4. Calculation of the Candida score: Variables selected in the logistic regression model
Variable
Coefficient
()
Standard
Error Wald 2
p
Value
Multifocal Candida species colonization 1.112 .379 8.625 .003
Surgery on ICU admission .997 .319 9.761 .002
Severe sepsis 2.038 .314 42.014 .000
Total parenteral nutrition .908 .389 5.451 .020
Constant 4.916 .485 102.732 .000
ICU, intensive care unit.
Candida score .908 (total parenteral nutrition) .997 (surgery) 1.112 (multifocal
Candida species colonization) 2.038 (severe sepsis). Candida score (rounded) 1 (total
parenteral nutrition) 1 (surgery) 1 (multifocal Candida species colonization) 2 (severe
sepsis). All variables coded as follows: absent, 0; present, 1.
Figure 1. A, receiver operating characteristics (ROC) curve and area under the ROC curve (AUC) for assessing the discriminatory power of the Candida score.
B, cut-off values for the ROC curve.
734 Crit Care Med 2006 Vol. 34, No. 3
tion was not included in this prediction
rule proposed by Ostrosky-Zeichner et al.
(29) because it was derived from the
NEMIS study results (19). The Candida
score presented here could therefore be
considered to be more reliable, given the
weight of fungal colonization in the
pathogenesis of candidiasis.
DuPont and co-workers (30) carried
out a retrospective systematic review of
surgical intensive care patients, with a
prospective follow-up in France. A scoring
system was proposed with the following
risk factors: female gender, upper
gastrointestinal origin of peritonitis, cardiovascular
failure, and use of antibiotics.
A grade C score, defined as the presence
of three qualifiers, was associated with a
sensitivity of 84% and specificity of 50%
for the detection of yeasts in the peritoneal
fluid of patients with peritonitis. The
main drawbacks of this study included its
single-center setting and its potential application
restriction to surgical patients.
We used the EPCAN database, which is
a large cohort of nonneutropenic ICU patients
for whom, among other goals, Candida
colonization and invasive candidiasis
were studied prospectively. As previously
reported (21), the rate of proven candidal
infection found in the present study was
low, but the mortality rate was high. The
mortality rate increased significantly according
to the patient group, that is, with
unifocal colonization (26.5%), multifocal
colonization (50.9%), and candidal
proven infection (57.7%). Although colonization
does not define infection, these
data support the well-known role of Candida
colonization as a key factor in the
decision to start early antifungal treatment
for ICU patients.
The new Candida score was based on
the respective predictive value of previously
reported risk factors. In addition to
multifocal Candida species colonization,
three other risk factors were found to be
significant predictors of proven candidal
infection in the logistic regression model:
use of total parenteral nutrition, surgery
on ICU admission, and clinical manifestations
of severe sepsis. The respective
weight of colonization and these risk factors
as shown in the Candida score allowed
us to reliably differentiate between
Candida species colonization and proven
candidal infection. Although central venous
catheters are repeatedly described
as major risk factors for proven hematogenous
candidiasis (19, 31), in this large
dataset from a prospective multicenter
study, venous catheters were not signifi-
cant predictors of proven candidal infection.
The medical literature is flooded with
complicated prediction rules and scores
(32–37), and there is a need to have available
bedside easy-to-remember scores
that would make daily tasks easier for
clinicians. The simplified version of this
score, after rounding up to 1 the weight
for total parenteral nutrition, surgery, or
multifocal Candida species colonization
and up to 2 the weight for clinical severe
sepsis, is a quite simple ready-to-use prediction
rule. With a cut-off value of 2.5,
that it to say, with a sensitivity of 81%
and a specificity of 74%, we shall only
need the presence of sepsis and any one of
the three other remaining risk factors or
the presence of all of them together except
sepsis in order to consider starting
antifungal treatment for one particular
patient. Finally, the Candida score also
identifies critically ill patients with
proven candidiasis: patients with a score
2.5 are 7.75 times as likely to have
proven infection (risk ratio 7.75; 95%
CI, 4.74 –12.66) than patients with a Candida
score up to 2.5.
Therefore, the easy rule of thumb for
prescribing antifungals, according to a
Candida score 2.5, will allow more effi-
cient selection of patients who indeed will
benefit from the increasing number of
available antifungal drugs (38) and, at the
same time, more adequate prevention of
the development of new resistant species
due to an excess of inappropriate and
potentially detrimental antifungal treatments
(39). Assessment with the Candida
score should be performed at the time of
ICU admission and any time candidiasis is
suspected.
CONCLUSIONS
A new score, the Candida score, which
was calculated according to data collected
in the EPCAN database (in which all cases
of Candida species colonization and
proven candidal infection were prospectively
recorded), is an easy-to-remember
bedside prediction rule. A score 2.5 will
help intensivists select patients who will
benefit from early antifungal administration.
Finally, although the Candida score
contributes to predicting proven candidal
infection, the benefits of preemptive (prophylactic
or empirical) antifungal therapy
remain to be determined.
ACKNOWLEDGMENT
We thank Marta Pulido, MD, for editing
the manuscript.
REFERENCES
1. Vincent JL, Anaissie E, Bruining H, et al:
Epidemiology, diagnosis and treatment of
systemic Candida infection in surgical patients
under intensive care. Intensive Care
Med 1998; 24:206–216
2. Sobel JD, Vazquez J: Candidiasis in the intensive
care unit. Semin Respir Crit Care
Med 2003; 24:99–111
3. Nolla-Salas J, Sitges-Serra A, Leon-Gil C, et
al: Candidemia in non-neutropenic critically
ill patients: Analysis of prognostic factors and
assessment of systemic antifungal therapy.
Intensive Care Med 1997; 23:23–30
4. Eggimann P, Garbino J, Pittet D: Epidemiology
of Candida species infections in critically
ill non-immunosuppressed patients. Lancet
Infect Dis 2003; 3:685–702
5. Charles PE, Doise JM, Quenot J, et al: Candidemia
in critically ill patients: Difference of
outcome between medical and surgical patients.
Intensive Care Med 2003; 29:
2162–2169
6. Debusk CH, Daoud R, Thirumoorthi MC, et
al: Candidemia: current epidemiologic characteristics
and a long-term follow-up of the
survivors. Scand J Infect Dis 1994; 26:
697–703
7. Leleu G, Aegerter P, Guidet B, for Collège des
Utilisateurs de Base de Données en Réanimation:
Systemic candidiasis in intensive care
units: A multicenter matched-cohort study. J
Critical Care 2002;17:168–175
8. Ostrosky-Zeichner, L: Prophylaxis or preemptive
therapy of invasive candidiasis in the
intensive care unit? Crit Care Med 2004; 32:
2552–2553
9. Álvarez-Lerma F, Palomar M, León C, et al:
Colonización y/o infección por hongos en
unidades de cuidados intensivos: Estudio
multicéntrico de 1.562 pacientes. Med Clin
(Barc) 2003; 121:161–166
10. Piarroux R, Grenouillet F, Balvay P, et al:
Assessment of preemptive treatment to prevent
severe candidiasis in critically ill surgical
patients. Crit Care Med 2004; 32:
2443–2449
11. Alvarez-Lerma F, Nolla-Salas J, Leon C, et al,
A score 2.5 will
help intensivists
select patients who
will benefit from early antifungal
administration.
posted by Dr. José Antonio Valiente at 3:41 p. m.
0 comments
