INVITED ARTICLE
H E A LT H C A R E E P I D E M I O L O G Y
Robert A. Weinstein, Section Editor
Prophylactic Antifungal Therapy in the Intensive Care
Unit
John H. Rex1 and Jack D. Sobel2
1
Division of Infectious Diseases, Department of Internal Medicine, Center for the Study of Emerging and Re-Emerging Pathogens, University of Texas Medical School,
Houston; and 2Wayne State University School of Medicine, Detroit
Antifungal prophylaxis is regularly used during treatment of patients with some cancers, as subgroups with high rates of
invasive fungal infections are readily identified; for these patients, prophylaxis has been shown to be of value. High-risk liver
transplant recipients also benefit from antifungal prophylaxis. Although the idea of extending this concept to the prevention
of candidal infections in the larger population of critically ill patients who are seen in the intensive care unit (ICU) and who
do not have neutropenia is attractive, implementation of this strategy is difficult because of the widely varying characteristics
of patients in the ICU. Two studies have shown the benefit of such prophylaxis, but the benefit was shown only in selected
groups of patients who had an unusually high risk for invasive candidiasis. Although the concept is sound, broad-scale
implementation of antifungal prophylaxis would be premature and costly, both financially and with regard to resistance and
toxicity. Investigations are needed to define and prove the utility of predictive tools for the identification of patients in the
ICU who would benefit from prophylaxis.
Although there is little doubt that “a stitch in time saves nine,”
knowing where and when to place that stitch is another matter.
With respect to invasive fungal infections, prophylactic therapy
has been extensively studied. Although the use of prophylaxis
was long limited by the toxicity of amphotericin B deoxycholate, the steady introduction during the past 20 years of increasingly safe and well-tolerated alternatives has made prophylaxis both feasible and attractive. Although there may
remain issues for debate [1–3], antifungal prophylaxis is now
regularly used in selected populations of patients with cancer.
As nosocomial fungal infections have steadily risen in frequency
and have become commonplace in populations without cancer
[4, 5], the idea that the success of the use of antifungal prophylaxis in patients with cancer might be extended to patients
in the intensive care unit (ICU) who do not have neutropenia
is tantalizing and is the subject of this review.
Received 4 October 2000; revised 29 November 2000; electronically published 26 March
2001.
Publication of the CID Special Section on Healthcare Epidemiology is made possible by
an educational grant from Pfizer, Inc.
Reprints or correspondence: Dr. John H. Rex, University of Texas Medical School, 6431
Fannin, 1728 JFB, Houston, TX 77030 ([email protected]).
Clinical Infectious Diseases 2001; 32:1191–200
2001 by the Infectious Diseases Society of America. All rights reserved.
1058-4838/2001/3208-0011$03.00
STUDIES OF PATIENTS WITH CANCER
PROVIDE VALUABLE LESSONS AND SHOW
THAT ANTIFUNGAL PROPHYLAXIS IS
POSSIBLE
The outcomes of studies of prophylaxis in patients with cancer
offer many valuable insights into the issues of the use of prophylaxis for patients in the ICU who do not have neutropenia.
In patients with cancer, the principal fungi of concern are Candida species and the various filamentous fungi, especially Aspergillus species. Because candidal infections have been historically predominant, most investigations done to date have
focused on this area. The findings of studies of the use of orally
administered but minimally absorbed agents (such as nystatin
[6], amphotericin B [7], and clotrimazole [8]) among patients
with hematologic malignancies have suggested that these drugs
have variable but generally limited efficacy.
The advent of fluconazole, an antifungal agent that has excellent activity against many Candida species, opened the door
to a large number of studies that collectively demonstrated that
antifungal prophylaxis was entirely possible [1, 3]. However,
these studies have also shown the limitations of prophylactic
strategies. Clear demonstrations of the utility of antifungal prophylaxis have been limited to patients undergoing bone marrow
transplantation. Two placebo-controlled studies of fluconazole,
HEALTHCARE EPIDEMIOLOGY • CID 2001:32 (15 April) • 1191
400 mg/day, given to a total of 656 recipients of bone marrow
transplants (two-thirds of the transplants were allogeneic and
one-third were autologous) demonstrated that prophylaxis reduced the incidence of proven systemic fungal infections from
17% to 4.5% (a 3.7-fold reduction) [9, 10].
These and related results emphasized the importance of the
degree of immunosuppression with regard to the risk for fungal
infection [11, 12], and the issue of immunosuppression was
emphasized in a recent review of these risks [13]. If the risk
of fungal infection is too low, then prophylaxis will not be
warranted because of the potential negative aspects of treatment
(e.g., selection of resistant isolates, drug reactions, and cost).
For example, a placebo-controlled study of fluconazole, 400
mg/day, given to 255 adults who were undergoing chemotherapy for acute leukemia, showed a trend similar to that seen in
the 2 bone marrow transplantation studies: fluconazole reduced
the rate of proven systemic fungal infection from 8% to 4%
[14]. However, this 2-fold reduction in the rate of infection
had a P value of only .3; given the lower event rate in the
placebo-treated patient group (relative to that seen in patients
in the bone marrow transplantation studies), a study ∼2.5 times
larger would have been required to yield a P value of !.05.
Likewise, a study that compared fluconazole therapy, 200 mg/
day, with nystatin suspension, 6,000,000 IU/day, among 109
adults with leukemia found that rates of systemic fungal infections among patients who received fluconazole were reduced
from 11% to 4%; however, the associated P value in this small
study was .15 [15].
The choice of comparison agent is also important. It is presumed that the origin of invasive candidiasis is often in gut
colonization [16, 17]. Although nonabsorbed antifungal agents
would not be presumed to be effective treatment for a fungal
infection, they might reduce the incidence of gut colonization,
and, in at least 1 study [7], they have produced a modest effect
on the rate of development of invasive candidiasis. Such an
effect might have contributed to the findings of Menichetti et
al. [18]. Their study compared the effects of fluconazole, 150
mg/day, with those of oral amphotericin B among 820 adults
who were undergoing chemotherapy for acute leukemia; proven
systemic infections were seen in 2.6% and 2.5% of the fluconazole- and oral amphotericin B–treated patients, respectively.
Either the rate of invasive fungal infections among patients in
the group treated with oral amphotericin B was unexpectedly
low (and, therefore, was not amenable to much further reduction by fluconazole), or oral amphotericin B provided some
protection against invasive candidiasis. Either way, these results
show that placebo-controlled studies are preferred, at least until
the value of prophylaxis is clearly established.
Another concern is the potential for selection/induction of
resistant isolates. Most studies have focused on the use of azoles—most notably, fluconazole—so it is not surprising that
1192 • CID 2001:32 (15 April) • HEALTHCARE EPIDEMIOLOGY
the majority of the reports on this topic have focused on azole
resistance [19]. Although mutation of susceptible isolates into
resistant isolates appears to be uncommon during the relatively
short durations of antifungal therapy given to patients with
cancer [20], selection of inherently less susceptible species (such
as Candida krusei and Candida glabrata) has been described
[21, 22]. Although other factors could be operative [23], prior
antifungal therapy has been strongly linked to infection caused
by these species [24, 25].
THE STATISTICAL CHALLENGE
These data show that a firm understanding of statistical principles is critical to establishing the utility of prophylaxis and
to determining its subsequent proper use. If the rate of invasive
fungal infections is high enough, demonstrating the value of
prophylaxis is straightforward. Such a strong and convincing
benefit is important when prophylaxis is weighed against the
inevitable problems of drug resistance and toxicity. The statistical implications of these ideas are shown in the series of
sample-size estimates presented in table 1.
With use of a common set of trial-design specifications, the
grid shows typical sample sizes for a variety of possible trialdesign scenarios. For example, if the group of placebo-treated
patients in a trial were expected to have a 10% infection rate,
Table 1.
Sample-size estimates in relation to event rates.
Treatment
arm A
event
rate
0.5%
1%
Treatment arm B event rate
1%
2% 5% 10% 15% 20% 25% 30%
5065
2%
989 2515
5%
249
332 653
10%
106
121 162 474
15%
66
72
86 160 726
20%
47
50
57
88 219
945
25%
36
37
42
58 112
270 1133
30%
29
30
33
43
72
134
35%
23
24
26
33
50
82
313 1290
151
348 1416
NOTE. The table shows approximate sample sizes required per treatment
arm of a 2-arm study, in relation to the estimated event rates. The sample
size is computed according to the method of Casagrande et al. [26] and uses
these commonly selected trial-design specifications: (1) the ultimate statistical
test will be a 2-sided test comparison of proportions of success (or failure),
(2) the desired statistical significance (P value or a value) will be set at .05,
and (3) the desired power will be 80% (corresponding to a b value of 0.20).
Note also that these values are the theoretical values for the study design. In
actual practice, the values shown in boldface type should be further increased
so that treatment arm B could have at least 5 events. For example, when
going from 5% (treatment arm A) to 0.5% ( treatment arm B), the theoretical
value is 249 per treatment arm. However, one would then see only
249 ⫻ 0.5% p 1.2 events in treatment arm B. In practice, a more robust result
would be produced if 1000 patients were enrolled in each treatment arm, thus
giving an expected absolute number of 5 events in arm B.
and if the treatment were expected to reduce this rate to 5%,
then a study of 948 patients (2 ⫻ 474 patients) would be needed
to detect this treatment benefit 80% of the time.
Table 1 presents several points. First, therapy should reduce
the event rate sharply. Consider comparisons in which the event
rate drops by only 5% (e.g., from 20% to 15%): these designs
require at least several hundred patients per treatment arm. On
the other hand, comparisons in which the rate drops 4-fold to
5-fold (e.g., from 20% to 5%) require much smaller sample
sizes. Second, these effects are magnified as the rates approach
zero: a 2-fold rate of reduction (from 30% to 15%) requires
only 266 patients (2 ⫻ 133 patients), whereas a reduction from
1% to 0.5% requires 10,130 patients (2 ⫻ 5065 patients).
A third and more profound consequence of the association
of large sample sizes with small event rates is that it raises the
issue of clinical relevance. There is no reason why we could
not do a trial with 10,130 patients and demonstrate that a given
therapy in a specific population reduces the rate of invasive
fungal functions from 1% to 0.5%. If the study were well done,
the result would be scientifically and statistically sound. However, it would be almost irrelevant in a medical sense. Because
we would have to treat 200 patients to eliminate 1 infection,
the cost of the therapy (measured in dollars and according to
side effects and induction of resistance) would almost certainly
exceed its value.
Therefore, this statistical analysis defines 2 key criteria for
the successful study and use of prophylaxis. A therapy that
produces a 4-fold to 5-fold reduction in the rate of infection
is highly desirable, both medically and statistically. This therapy
should be targeted toward subgroups with a baseline rate of
infection of ∼10% or more.
ANTIFUNGAL PROPHYLAXIS FOR PATIENTS
IN THE CRITICAL CARE SETTING WITHOUT
NEUTROPENIA: PRINCIPLES
With the aforementioned concepts as a background, let us now
turn to the issues that surround the particular question at hand
and try to determine whether antifungal prophylaxis is either
possible or relevant in the critical care setting. Note that invasive
candidiasis can also be a particular problem for patients who
have undergone liver transplantation [27–29] and that effective
prophylaxis is possible [30–33]. Likewise, prophylaxis might
have value for patients who have undergone pancreas transplantation [34]. However, both of these entities are quite specialized in nature and limited in overall scope. This review
focuses on the relevance of prophylaxis to the much larger
population of patients in the general ICU. (Although neonatal
patients will not be specifically discussed, the concepts for the
patient population in the general ICU are applicable to that
population as well.)
If the population is critically ill patients who do not have
neutropenia, then the target is Candida species. Although
causative agents other than Candida species may be seen, such
fungal infections usually are the cause of a patient’s admission
to the ICU (e.g., for rhinocerebral zygomycosis in a patient
with diabetes) or are related to neutropenia (e.g., invasive aspergillosis after engraftment in a bone marrow transplant recipient). Otherwise, fungal infections in this setting are limited
almost entirely to invasive candidiasis. The subsequent discussion will focus on this fungus.
Prophylaxis versus empirical (or preemptive) therapy for
the febrile patient. It is important to distinguish prophylactic
from empirical therapy. If one is faced with a new febrile illness
in a patient with diabetes who has undergone broad-spectrum
antibiotic therapy, hemodialysis, multiple intravascular catheterizations, and abdominal surgery during a 3-week ICU stay
and who is colonized with Candida albicans at multiple sites,
the time for prophylaxis is long past. This patient has multiple
well-established risk factors for invasive candidiasis [35, 36]
and is a candidate for empirical (also sometimes referred to as
“preemptive”) antifungal therapy.
On the other hand, this patient might have been a candidate
for antifungal prophylaxis at the start of the ICU stay. The
proper, early identification of such patients is the key to establishing the value of this strategy.
What is the rate of incidence of invasive candidiasis among
patients in the critical care setting who do not have neutropenia? As the statistical discussion above makes clear, this
question is fundamental to our ability to apply prophylaxis. To
answer the question, we must make some decisions about the
forms of candidiasis that we will accept as relevant. Candida
species produce diseases that range from trivial (albeit troublesome) mucosal syndromes to acute bloodstream infections
without clear evidence of visceral involvement, to focal involvement of any organ. The presence of mucosal disease might
necessitate therapy, but use of a systemic agent to reduce the
rate of incidence of oral candidiasis does not seem warranted.
Therefore, studies of prophylaxis should focus exclusively on
invasive forms of candidiasis.
Some general diagnostic criteria for invasive fungal infections
have been put forward [37]. Generally speaking, the easiest
syndromes to diagnose are those that involve growth of Candida
species from a sterile body site. Bloodstream infection is the
most common of such presentations. Although one can debate
almost endlessly about the meaning of 1 positive blood culture
result, in comparison with 3 such results, and about the relevance of whether the culture specimen was obtained by means
of peripheral venipuncture or via a pre-existing vascular catheter, it is clear that a report from the microbiology laboratory
about the growth of Candida species in a blood culture should
HEALTHCARE EPIDEMIOLOGY • CID 2001:32 (15 April) • 1193
precipitate investigation and (in the vast majority of instances)
therapy [38].
Diagnostic use of the recovery of Candida species from nonsterile sites or from generally sterile sites that have close proximity to a mucosal surface is problematic. Candida isolates in
urine samples are the prototype for this problem, and careful
investigations (albeit primarily in patients outside of the ICU)
have made it clear that asymptomatic candiduria (1) has few
clinical consequences and (2) follows a natural history that is
only minimally affected by antifungal therapy [39, 40].
Candida isolates in sputum samples are equally problematic.
The true rate of candidal pneumonia among patients who do
not have cancer is not known [41]. Two small autopsy series
that focused primarily on surgical patients revealed that ∼20%
of patients have at least a minor focus of candidiasis in a lung
[42, 43]. Even if present, such lesions appear to be clinically
imperceptible; a recent investigation of a series of 435 prospectively followed patients in the ICU who did not have neutropenia failed to yield even a single convincing case of candidal
pneumonia [44]. Finally, a detailed study that correlated findings of immediate postmortem bronchoalveolar lavage, transbronchial biopsy, and open-lung biopsy demonstrated an 8%
rate of infection with candidal pneumonia [45]. However,
quantitative sputum sampling did not reliably predict the presence of underlying pneumonia. Therefore, although the presence of Candida species at these sites may be relevant as a risk
factor for invasive candidiasis, it will not be useful as a marker
of disease or as an outcome measure.
A third issue is the frequent occurrence of a syndrome that
is strongly suggestive of invasive candidiasis but that lacks clear
microbiological proof of infection. The aforementioned example of a febrile, antibacterial antibiotic–unresponsive, heavily
colonized patient in the ICU is the prototypical scenario. Even
with use of the newer technologies, blood cultures have a sensitivity for invasive candidiasis that is no more than 50%–60%
[46, 47]. Colonization has clearly been established as a major
risk factor (if not the major risk factor) for invasive candidiasis
among these patients [48], and many clinicians would feel compelled to initiate empirical antifungal therapy in such a scenario.
Although a diagnosis based on colonization is intellectually less
satisfying than a diagnosis based on growth in a specimen from
a sterile site, prevention of the need for empirical antifungal
therapy is a valid clinical outcome measure, and it parallels the
findings of prior investigations that have involved patients who
did not have neutropenia.
With this background in mind, the literature on rates of
incidence of invasive candidiasis can be analyzed. Although
nosocomial fungal infection rates of incidence of 7–16 cases
per 1000 discharged patients have been reported by the National Nosocomial Infection Surveillance system for patients in
a variety of critical care settings [4], such surveillance data are
1194 • CID 2001:32 (15 April) • HEALTHCARE EPIDEMIOLOGY
limited by the definitions and processes used to collect the data.
Reports that have focused specifically on invasive candidiasis
have described case rates that have ranged from 35% (in a
series of 111 patients who stayed for ⭓4 days in a trauma ICU
[49]) to as low as 2% (in a series of 435 patients who were
admitted to medical or surgical ICUs [44]).
Differences in diagnostic criteria, along the lines of those discussed in the preceding paragraphs, explain this range of rates.
The 35% rate of infection seen in the study of Mahr et al. [49]
was reached because the authors had used quite liberal definitions, and at least 75% of the reported infections should have
been classified, at best, as “possible” invasive candidiasis. On the
other hand, the 2% rate reported in the study of Petri et al. [44]
was determined on the basis of stringent criteria that allowed for
the diagnosis of only conclusively proven invasive candidiasis. A
study by Pittet et al. of 650 patients in surgical or neonatal ICUs
used similarly stringent rules and showed a 1.7% rate of invasive
candidiasis [48]. Finally, the study by Borzotta and Beardsley
[50] of a series of 459 patients who were admitted to a trauma
ICU for ⭓4 days revealed an overall rate of invasive candidiasis
of 4.3% but a fungemia rate of only 1.5%.
Collectively, the data from studies with the more stringent
rules are similar to those from the National Nosocomial Infection Surveillance system and suggest that a figure of ∼2%
is a good baseline rate with regard to the incidence of invasive
candidiasis in patients who stay in an ICU for more than a few
days. On the other hand, more liberal rules that include presumed forms of invasive candidiasis may yield rates that are
2-fold to 15-fold higher. Although it is hard to deny the need
to make an empirical diagnosis at times, the subjective aspects
of this process must be carefully addressed. Clinical trials will
probably need to use a combination of such rules, but it is
likely that most attention would be paid to the subset of cases
of proven disease.
ANTIFUNGAL PROPHYLAXIS IN PATIENTS
IN THE CRITICAL CARE SETTING WHO DO NOT
HAVE NEUTROPENIA: PRACTICES
As is evident from the earlier discussion of statistical factors,
a baseline rate of incidence of 2% for proven invasive candidiasis makes prophylaxis problematic for several reasons. However, not all patients in the ICU are at equal risk for invasive
candidiasis; this figure of 2% is an average, and some subgroups
have much higher rates of disease. How, then, can we identify
those higher-risk patients? Risk factors for patients both with
and without neutropenia have been studied. Unfortunately, the
amount of data here is surprisingly limited. Although many
investigators have described the characteristics of patients who
have invasive candidiasis, the key risk factors can be identified
only by means of comparison with a control group. Studies
that have included such an analysis are shown in table 2.
Neutropenia, the one factor that consistently makes prophylaxis of value for patients with cancer, is not at all relevant
to patients in the ICU setting. On the other hand, several new
factors become important for patients who do not have neutropenia. Among these factors, use of hyperalimentation and
issues related to the gut are particularly intriguing. Also of
interest is the absence of burn injury as a risk factor in these
studies, despite the association of patient admissions to burn
units with high rates of invasive fungal infections [66]. Likewise,
surgery is probably a risk factor [62], but reports to date have
not teased out the impact of different types of surgery.
Studies of the relevance of candidal colonization also have
been notably limited. Most reports simply show that colonization is a risk factor; the critical issues of site, extent, and
duration of colonization are only beginning to be explored.
Some carefully performed studies have found that colonization
was not a meaningful independent predictor of invasive disease
[62]. Pittet et al. [48] have broken new ground in this subject
area with their work that shows that semiquantitative measures
of density of colonization could be linked to risk. Further work
in this important area is clearly needed. The intriguing issue
of the duration of the period when the patient is at risk also
has been little studied. Craven et al. [67] showed a median
time to onset of nosocomial infection (of any type) of 7–10
days among a large group of patients in the medical and surgical
Table 2.
ICUs, and the time at risk might well be relevant with regard
to fungal infections.
As an example of the possible use of these types of data, the
study by Wey et al. [35] has generated a regression equation
by use of logistic regression for 4 predictive factors (number
of antibiotics administered, presence of Candida species in a
sample other than a blood sample, performance of hemodialysis, and use of a Hickman catheter). Such data could be
used to directly compute a risk for invasive candidiasis and,
therefore, to permit identification of patients who would be
most likely to benefit from prophylactic therapy. However, the
data of Wey et al. are not directly applicable to many patients
in the ICU, because the case-control selection process included
matching with regard to whether the patient underwent a prior
major surgical procedure. Therefore, risk factors related to surgery could not have been (and were not) identified by these
investigators. Because surgery, especially on the gut, often seems
to be relevant, additional data are needed. Furthermore, many
patients with neutropenia who had Hickman catheters inserted
were included in the dataset, again making this group not directly applicable to patients in the general ICU setting.
Despite these data limitations, 5 attempts [65, 68–71] to apply
these concepts in the ICU setting have been made to date (table
3). Savino et al. [68] enrolled 292 patients from surgical and
trauma ICUs in 4 different treatment arms (i.e., no therapy,
therapy with oral clotrimazole, therapy with oral nystatin, and
therapy with oral ketoconazole). The rate of incidence of invasive
Reports of studies of risk factors for invasive candidiasis.
Study cohort
Patients in the ICU
without neutropenia
Risk factor
Patients with cancer
Granulocytopenia, BMT, type/duration of chemotherapy,
GVHD, extent of mucositis
[16, 51–56]
Colonization with Candida species
[17, 51, 55, 57–59]
[35, 48, 60]
Broad-spectrum antibiotics
[51, 54, 59]
[35, 48, 60, 61]
Hemodialysis or azotemia
Central vascular catheter
Severity of illness
Excluded by definition
[35, 60]
[55]
[35, 60, 61]
[48]
Hyperalimentation
[50, 61–63]
Recurrent or persistent gastrointestinal perforation
[64, 65]
Prior surgery
[62]
Neonatal ICU: gestational age, low Apgar score, length
of stay, shock, H2 blockers, and intubation
[61]
NOTE. Although many investigators have reported the general characteristics of groups of patients who have developed
invasive candidiasis, data from case-control or cohort studies that have assessed the critical risk factors are less abundant.
The data in this table are limited to results from such studies (identified via both MEDLINE searches and reviews of
bibliographies of articles on this topic), and the reported factors are either those identified by means of logistic regression
or, if no logistic regression was performed, those with relatively high ORs in univariate analyses. The studies involved the
use of diagnostic criteria that range from the presence of candidemia to inclusion of various forms of localized invasive
candidiasis. Other than a study of pediatric patients with leukemia in the group of patients with cancer [56], a study of
patients who were in the pediatric ICU that included a group of patients in the ICU without neutropenia [63], and a study
of patients who were in the neonatal ICU [61] that included a group of patients in the ICU without neutropenia, all data
are from studies of adults. BMT, bone marrow transplantation; GVHD, graft-versus-host disease.
HEALTHCARE EPIDEMIOLOGY • CID 2001:32 (15 April) • 1195
Table 3.
Ref.
Data from studies of antifungal prophylaxis in patients in the intensive care unit.
Inclusion criteria
a
b
Exclusion criteria
Invasive candidiasis
[68]
Remaining or expected
to remain in an SICU
for ⭓48 h
Pregnancy, burn,
transplant, yeast
colonization
⫹
[65]
Recurrent or refractory
GI perforation/
leakage
Hemodialysis, peritoneal dialysis
⫹
[69]
Expected to remain in
an SICU for ⭓48 h
plus a risk factord for
candidiasis
Pregnancy, life expectancy !3 mo
[70]
Expected to remain in
an SICU for ⭓72 h
plus mechanical
ventilation
[71]
Expected to remain in
an SICU for ⭓72 h
plus “acutely critically ill”
⫹
BC, peritoneal Cx, or
⫹
Cx results from ⭓3
sites
Study arm (evaluable n)
No therapy (72)
Clotr, 10 mg p.o. q8h (80)
Keto, 200 mg p.o. q.d. (65)
Nys, 2 ⫻ 106 U p.o. q6h (75)
IC rate,
% (n)
(2)
(1)
(1)
(5)
NS
Placebo (20)
Flu, 400 mg iv q.d. (23)
35 (7)
4 (1)
.02c
Documented candidiasis or colonization
with fever
Placebo (59)
Flu, 400 mg p.o./iv q.d. (60)
19 (11)e
13 (8)
NS
Not specified in
abstract
“Severe Candida
species infection”
Placebo (101)
Flu, 100 mg iv q.d. (103)
8.9 (9)
3.9 (4)
.2
Not specified in
abstract
⫹
Placebo (130)
Flu, 400 mg p.o. q.d. (130)
15.3 (20)
8.5 (11)
.07
Peritoneal Cx with
symptoms, and
increasing amounts
of Candida species
Cx result at autopsy,
⫹
BC result at biopsy
3
1
2
7
P
NOTE. BC, blood culture; Clotr, clotrimazole; Cx, culture; Flu, fluconazole; GI, gastrointestinal; IC rate, rate of incidence of invasive candiasis; Keto,
ketoconazole; Nys, nystatin; p.o., by mouth; Ref., reference; SICU, surgical intensive care unit; ⫹, positive.
a
Exclusions for prior antifungal therapy, preexisting fungal infection, likelihood of death within a few days of enrollment, and significant baseline laboratory
abnormalities (e.g., for liver enzyme levels or platelet count) were generally used and are not specifically noted.
b
The principal form of invasive candidiasis diagnosed in the study. Other forms may have also been noted, but these forms are the basis for the key results.
c
An additional case of candidemia was noted in the fluconazole treatment arm; inclusion in the definition of invasive candidiasis raises the P value to .06.
d
Central venous catheter, total parenteral nutrition, mechanical ventilation for 124 h, or broad-spectrum antibiotics.
e
Most cases appear to have been diagnosed on the basis of colonization (usually via a sputum culture; see text for discussion).
candidiasis was only 3% (even though the case definition included patients who were merely colonized at multiple sites).
The 4 treatment arms made the study underpowered for any
given comparison, and (not surprisingly) no differences were
seen between the treatment arms. The exclusion of colonized
patients may have contributed to the low event rate. In addition,
ketoconazole is inconsistently absorbed even in healthier individuals, so the amount of drug delivered in this setting is
uncertain.
In the second study, Eggimann et al. [65] followed up on
the observation [64] that patients with refractory gastrointestinal perforation had candidal peritonitis at significant rates.
Although not all of these patients may have met the criteria
for being critically ill, their general character is similar to that
of many patients in the surgical ICU, and the trial is thus
relevant. These authors then enrolled 43 of these highly selected
patients in a randomized, prospective, double-blind, placebocontrolled trial of prophylaxis with fluconazole, 400 mg/day
[65]. They observed a rate of peritonitis of 35% among the
placebo-treated patients, a rate that fell to 4% among the fluconazole-treated patients (P p .02 ). This striking result is even
more impressive when the sample size is revealed: there were
1196 • CID 2001:32 (15 April) • HEALTHCARE EPIDEMIOLOGY
only 23 patients in the fluconazole treatment arm and 20 in
the placebo arm.
Despite its small size, the study is sound and convincing. Its
success depended on administration of a potent therapy (there
was a 10-fold reduction in the rate of infection) to a group of
very-high-risk patients, all of which makes it a good example
of the aforementioned statistical principles. With that, the outcome seems inevitable, at least in retrospect. However, the
strength of this study (the use of carefully selected patients) is
also its limitation, because the results do not apply more
broadly.
In the third study, Ables et al. [69] enrolled patients in a
trauma/surgical ICU who had an anticipated ICU stay of at least
48 h in a randomized, prospective, double-blind, placebo-controlled trial of prophylaxis with fluconazole, 400 mg/day. Prophylaxis was considered a failure if a patient developed documented invasive candidiasis, a syndrome compatible with
invasive candidiasis that was believed to require specific treatment, or a systemic inflammatory response syndrome without
demonstrable bacterial or other etiologies. With 60 evaluable
patients in one treatment arm and 59 in the other, the failure
rates were 22% and 24%, respectively. There were slightly more
failures due to candidiasis in the placebo arm (19% vs. 13%),
but the case definition confuses the results. None of the cases of
candidiasis involved the bloodstream, and the majority of cases
were diagnosed on the basis of positive sputum culture results.
Critical examination of this study shows that it seems to suffer
from both small sample size and lack of use of a stringent case
definition. Because the patients were only minimally selected, an
event rate of no more than 2% for proven invasive candidiasis
might have been expected. With a sample size of 60 patients per
study arm, only 1.2 events (60 ⫻ 0.02 ) would be expected in the
placebo arm. Application of the Poisson estimation for rare
events [72] shows that one would expect to see this one event
only 70% of the time in a sample this small. It is therefore hard
to draw convincing conclusions from this study.
The final 2 studies come close to providing generally applicable approaches to this problem. The study by Garbino et al.
[70] included 220 patients in the medical/surgical ICU who,
either on or after their third day in the ICU, were enrolled in
a randomized, prospective, double-blind, placebo-controlled
trial of prophylaxis with fluconazole, 100 mg/day. All patients
in this study were also undergoing mechanical ventilation and
selective digestive tract decontamination with nonabsorbed
antibacterial agents. The incidence of documented invasive candidal infections was reduced from 8.9% to 3.9% by treatment
with fluconazole, but this result did not reach statistical significance (P p .2). A significant reduction in the frequency and
intensity of candidal colonization was observed, however. The
encouraging trend observed in this study thus shows the potential for prophylaxis to be effective, but the low event rate
causes the results to be statistically nonsignificant. As suggested
by the data in table 1, a sample size of ∼500 patients per study
arm (for a total of ∼1000 patients) would have been required
to give the study a power of 80%.
Pelz et al. [71] enrolled 260 acutely critically ill patients with
predicted durations of stay of at least 3 days in the ICU at The
Johns Hopkins University Hospital, Baltimore. The patients
were randomized to receive fluconazole, 400 mg/day given enterally, or placebo, starting as soon as possible after ICU admission. With use of a very strict case definition for invasive
candidiasis that required microbiological proof of infection
(colonization, no matter the extent, was not used), the rate of
incidence of proven invasive candidiasis was 15.3% in the placebo arm and only 8.5% in the treatment arm (P p .07 by use
of the x2 test, but P ! .01 in a time-to-event analysis). These
results are impressive and clearly show the potential for prophylaxis. The use of oral therapy is also intriguing: a companion
pharmacokinetic study shows reasonable blood levels [73], and
(as discussed above) it is possible that the local effects in the
gut added to the systemic effects of this therapy.
However, generalization of these results requires a good un-
derstanding of the nature of the enrolled patients, especially in
light of the fact that the 15.3% rate of development of invasive
candidiasis among the placebo-treated patients is comparable
to the 15.8% rate in a previous study of patients who were
undergoing bone marrow transplantation [9]. Given that the
enrollment criterion for the study (staying or being predicted
to stay at least 3 days in the ICU) is so similar to that for
studies with much lower rates of invasive candidiasis, are there
other factors at play?
Analysis of the available data suggests that this is so. On the
basis of comparison with the patients enrolled in the study of
Garbino et al. [70] and according to data obtained from the
published abstracts and their authors (D. Pittet and P. Lipsett,
personal communications), the patients enrolled by Pelz et al.
were older (mean age, ∼65 vs. ∼55 years), were more severely
ill (mean Acute Physiology and Chronic Health Evaluation version III [APACHE III] score of ∼65 vs. APACHE II score of
∼21), more often had recently undergone surgery (190% vs.
∼60%), more often had diabetes (∼25% vs. ∼12%), more often
had severe liver dysfunction or cirrhosis (∼20% vs. ∼2%, with
many of the patients of Pelz et al. being candidates for liver
transplantation), and stayed in the ICU and thus continued
receiving the study therapy for a longer duration (mean duration of therapy, ∼14 days vs. ∼7 days).
Duration of study enrollment may be a factor of unappreciated importance—70% of the observed infections in the study
by Pelz et al. were seen during or after the second week of
therapy. An additional subtle indication of the difference in the
ICU populations comes from examination of the relative proportions of the patients in the ICU who were enrolled in the
study. Pelz et al. found that 38% of the patients admitted to
their ICU were study candidates (and the study enrolled 21%
of the admitted patients), whereas Garbino et al. found that
!5% of their admitted patients were eligible for their study.
Likewise, the patients studied by Pelz et al. were measurably
more ill than were the ICU populations studied by Petri et al.
(median APACHE II score, 14; median age, 59 years [44]),
Ables et al. (median APACHE II score, ∼18; mean age, ∼45
years [69]), and Borzotta and Beardsley (mean APACHE II
score, ∼16; mean age, ∼40 years [50]). The possibility that the
observed results are principally due to outcomes in a subset of
patients (e.g., the cirrhosis/liver transplantation candidates) also
cannot be excluded, and a better understanding of the patient
group(s) that drove the end result awaits more complete publication of the data. However, this study makes clear the potential for prophylaxis: if a suitable group can be identified,
prophylaxis can and will be of value. With regard to the unusually ill patient population in the ICU studied by Pelz et al.,
selection appears straightforward. On the other hand, the published experience of other ICUs shows that most other ICU
HEALTHCARE EPIDEMIOLOGY • CID 2001:32 (15 April) • 1197
populations are not quite as ill and that a highly selective approach to therapy would be required.
CONCLUSIONS AND FUTURE DIRECTIONS
As this review emphasizes, antifungal prophylaxis appears beneficial for selected critically ill patients who do not have neutropenia. Antifungal prophylaxis has been shown to be effective
and useful for carefully chosen patients with persistent intestinal
perforation or who undergo liver transplantation or, possibly,
pancreatic transplantation. Additional studies are needed to
identify other subpopulations that might benefit. The studies
by Garbino et al. [70] and Pelz et al. [73] make it clear that
such demonstrations are entirely possible. Consistent with these
observations, a multivariate analysis in a multicenter observational study showed that antifungal prophylaxis was independently associated with a decreased risk for fungal infection
[62]. However, routine clinical use of antifungal prophylaxis
will require thoughtful analysis on the part of the physician.
The patients enrolled by Pelz et al. appear to belong to an
unusual risk group that is not typical of other ICU settings,
and it may also be that the majority of the benefit was received
by a minority of the patients. The patients enrolled by Garbino
et al. appear to be closer to representing a typical group of
patients in the ICU, but some form of additional patient selection would appear to be needed. Overall, it would appear
likely that only a fraction of the patients in any given ICU
would be candidates for prophylaxis, and careful patient selection will be required to maximize the benefit of such therapy.
In the recently released guidelines of the Infectious Diseases
Society of America–Mycoses Study Group for treatment of fungal infections, we and our coauthors discouraged routine use
of antifungal prophylaxis in the general ICU setting [74]. This
position is consistent with that previously recommended by the
British Society for Antimicrobial Chemotherapy [75] and in a
consensus statement of a group of international clinicians [76].
This review illustrates both the strengths and the weaknesses
of the data behind these conclusions. The work of Pelz et al.
[73], Garbino et al. [70], and Eggiman et al. [65] makes it clear
that patient selection is the key.
The existing data regarding risk factors need to be updated
in the current practice environment, with a particular focus on
understanding the significance of fungal colonization. The relevant regression coefficients would then be used to define truly
useful patient selection criteria. Studies to generate such data
have recently been completed, and their analysis may answer
significant parts of these questions. Studies in the neonatal ICU
may be particularly easy to implement, because the uniform
presence of the risk factors related to prematurity creates a
relatively homogeneous population with a significant rate of
disease. Although measurements of colonization rates, includ1198 • CID 2001:32 (15 April) • HEALTHCARE EPIDEMIOLOGY
ing types of sites and extent of colonization, are of value in the
determination of the predictive value of colonization when selecting patients at risk for invasive candidiasis, current knowledge does not justify performance of routine surveillance cultures, especially in light of the enormous logistical difficulties
and costs attached to such testing.
The principal negative aspects of prophylaxis will be selection
of resistant strains and drug-related toxicity. These negative
consequences are anything but trivial and should not be underestimated. None of the study reports published to date have
addressed such events, but larger-scale use over longer periods
of time than in a typical clinical trial would probably be required to observe the true rate of such events. Despite this,
both possibilities are more than theoretical and lead to challenging questions about choice of drug and route of therapy
for prophylaxis.
All currently available antifungal agents have the potency
required to produce a good clinical effect. However, they do
differ with regard to the secondary features of toxicity and
resistance. The azole antifungal agents are very safe, but selection for resistant species is relatively common. The polyenes
have significant risk of toxicity but low rates of clinical resistance. Agents of the echinocandin class (e.g., caspofungin,
FK463, and LY303366 [77]) appear quite safe, but the risk for
selection of clinically resistant strains is not yet known.
The process of defining the utility of antifungal prophylaxis
would also be simplified by the availability of better tools for
the diagnosis of invasive candidiasis. Our current techniques
(most notably, blood culture) are, at best, blunt instruments
with limited sensitivity. Being forced to rely solely on definite
cases that have been diagnosed with available tools surely makes
the task of assessing the value of prophylaxis more difficult. It
is unfortunate that serodiagnostic strategies for candidiasis (e.g.,
PCR and antigen detection) remain problematic, and a convincing tool has yet to emerge.
Acknowledgment
We thank Jeanette Y. Lee for statistical assistance.
References
1. Bow E, Laverdiere M, Lussier N, Rotstein C, Ioannou S. Antifungal
prophylaxis in neutropenic cancer patients: a meta-analysis [abstract
LM-88]. In: Program and abstracts of the 37th Interscience Conference
on Antimicrobial Agents and Chemotherapy. Washington, DC: American Society for Microbiology, 1997.
2. Bohme A, Karthaus M, Hoelzer D. Antifungal prophylaxis in neutropenic patients with hematologic malignancies: is there a real benefit?
Chemotherapy 1999; 45:224–32.
3. Gøtzsche PC, Johansen HK. Meta-analysis of prophylactic or empirical
antifungal treatment versus placebo or no treatment in patients with
cancer complicated by neutropenia. BMJ 1997; 314:1238–44.
4. Beck-Sagué CM, Jarvis WM, the National Nosocomial Infections Sur-
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
veillance System. Secular trends in the epidemiology of nosocomial
fungal infections in the United States, 1980–1990. J Infect Dis 1993;
167:1247–51.
Kao AS, Brandt ME, Pruitt WR, et al. The epidemiology of candidemia
in two United States cities: results of a population-based active surveillance. Clin Infect Dis 1999; 29:1164–70.
DeGregorio MW, Lee WMF, Ries CA. Candida infections in patients
with acute leukemia: ineffectiveness of nystatin prophylaxis and relationship between oropharyngeal and systemic candidiasis. Cancer
1982; 50:2780–4.
Ezdinli EZ, O’Sullivan DD, Wasser LP, Kim U, Stutzman L. Oral amphotericin for candidiasis in patients with hematologic neoplasms: an
autopsy study. JAMA 1979; 242:258–60.
Cuttner J, Troy KM, Fumaro L, Brenden R, Bottone EJ. Clotrimazole
treatment for prevention of oral candidiasis in patients with acute
leukemia undergoing chemotherapy: results of a double-blind study.
Am J Med 1986; 81:771–4.
Goodman JL, Winston DJ, Greenfield RA, et al. A controlled trial of
fluconazole to prevent fungal infections in patients undergoing bone
marrow transplantation. N Engl J Med 1992; 326:845–51.
Slavin MA, Osborne B, Adams R, et al. Efficacy and safety of fluconazole prophylaxis for fungal infections after bone marrow transplantation: a prospective, randomized, double-blind study. J Infect Dis
1995; 171:1545–52.
Guiot HFL, Fibbe WE, van’t Wout JW. Prevention of invasive candidiasis by fluconazole in patients with malignant hematological disorders and a high grade of Candida [abstract LM-33]. In: Program and
abstracts of the 36th Interscience Conference on Antimicrobial Agents
and Chemotherapy. Washington, DC: American Society for Microbiology, 1996.
Walsh TJ, Hiemenz J, Pizzo PA. Editorial response: evolving risk factors
for invasive fungal infections: all neutropenic patients are not the same.
Clin Infect Dis 1994; 18:793–8.
Prentice HG, Kibbler CC, Prentice AG. Towards a targeted, risk-based,
antifungal strategy in neutropenic patients. Br J Haematol 2000; 110:
273–84.
Winston DJ, Chandrasekar PH, Lazarus HM, et al. Fluconazole prophylaxis of fungal infections in patients with acute leukemia: results
of a randomized placebo-controlled, double-blind, multicenter trial.
Ann Intern Med 1993; 118:495–503.
Young GAR, Bosly A, Gibbs DL, Durrant S. A double-blind comparison
of fluconazole and nystatin in the prevention of candidiasis in patients
with leukaemia. Eur J Cancer 1999; 35:1208–13.
Rotstein C, Bow EJ, Laverdiere M, Ioannou S, Carr D, Moghaddam
N. Randomized placebo-controlled trial of fluconazole prophylaxis for
neutropenic cancer patients: benefit based on purpose and intensity
of cytotoxic therapy. Clin Infect Dis 1999; 28:331–40.
Guiot HFL, Fibbe WE, van’t Wout JW. Risk factors for fungal infection
in patients with malignant hematologic disorders: implications for empirical therapy and prophylaxis. Clin Infect Dis 1994; 18:525–32.
Menichetti F, Del Favero A, Martino P, et al. Preventing fungal infection
in neutropenic patients with acute leukemia: fluconazole compared
with oral amphotericin B. The GIMEMA Infection Program. Ann Intern Med 1994; 120:913–8.
Rex JH, Rinaldi MG, Pfaller MA. Resistance of Candida species to
fluconazole. Antimicrob Agents Chemother 1995; 39:1–8.
Marr KA, White TC, van Burik J-AH, Bowden RA. Development of
fluconazole resistance in Candida albicans causing disseminated infection in a patient undergoing marrow transplantation. Clin Infect Dis
1997; 25:908–10.
Wingard JR, Merz WG, Rinaldi MG, Johnson TR, Karp JE, Saral R.
Increase in Candida krusei infection among patients with bone marrow
transplantation and neutropenia treated prophylactically with fluconazole. N Engl J Med 1991; 325:1274–7.
Wingard JR, Merz WG, Rinaldi MG, Miller CB, Karp JE, Saral R.
Association of Torulopsis glabrata infections with fluconazole prophy-
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
laxis in neutropenic bone marrow transplant patients. Antimicrob
Agents Chemother 1993; 37:1847–9.
White MH. Editorial response: the contribution of fluconazole to the
changing epidemiology of invasive candidal infections. Clin Infect Dis
1997; 24:1129–30.
Abi-Said D, Anaissie E, Uzun O, Raad I, Pinzcowski H, Vartivarian S.
The epidemiology of hematogenous candidiasis caused by different
Candida species. Clin Infect Dis 1997; 24:1122–8.
Nguyen MH, Peacock JE Jr, Morris AJ, et al. The changing face of
candidemia: emergence of non–C. albicans species and antifungal resistance. Am J Med 1996; 100:617–23.
Casagrande JT, Pike MC, Smith PG. An improved approximate formula
for calculating sample sizes for comparing binomial distributions. Biometrics 1978; 34:483–6.
Karchmer AW, Samore MH, Hadley S, Collins LA, Jenkins RL, Lewis
WD. Fungal infections complicating orthotopic liver transplantation
[with discussion]. Trans Am Clin Climatol Assoc 1994; 106:38–48.
Hadley S, Samore MH, Lewis WD, Jenkins RL, Karchmer AW, Hammer
SM. Major infectious complications after orthotopic liver transplantation and comparison of outcomes in patients receiving cyclosporine
or FK506 as primary immunosuppression. Transplantation 1995; 59:
851–9.
Collins LA, Samore MH, Roberts MS, et al. Risk factors for invasive
fungal infections complicating orthotopic liver transplantation. J Infect
Dis 1994; 170:644–52.
Kung N, Fisher N, Gunson B, Hastings M, Mutimer D. Fluconazole
prophylaxis for high-risk liver transplant recipients. Lancet 1995; 345:
1234–5.
Tollemar J, Hockerstedt K, Ericzon BG, Jalanko H, Ringden O. Liposomal amphotericin B prevents invasive fungal infections in liver
transplant recipients: a randomized, placebo-controlled study. Transplantation 1995; 59:45–50.
Linden P, Kramer DJ, Mazariegos G, et al. Low-dose amphotericin B
for the prophylaxis of serious Candida infections in high-risk liver
recipients [abstract J47]. In: Program and abstracts of the 36th Interscience Conference on Antimicrobial Agents and Chemotherapy. Washington, DC: American Society for Microbiology, 1996.
Winston DJ, Pakrasi A, Busuttil RW. Prophylactic fluconazole in liver
transplant recipients: a randomized, double-blind, placebo-controlled
trial. Ann Intern Med 1999; 131:729–37.
Benedetti E, Gruessner AC, Troppmann C, et al. Intra-abdominal fungal infections after pancreatic transplantation: incidence, treatment,
and outcome. J Am Coll Surg 1996; 183:307–16.
Wey SB, Mori M, Pfaller MA, Woolson RF, Wenzel RP. Risk factors
for hospital-acquired candidemia: a matched case-control study. Arch
Intern Med 1989; 149:2349–53.
Fraser VJ, Jones M, Dunkel J, Storfer S, Medoff G, Dunagan WC.
Candidemia in a tertiary care hospital: epidemiology, risk factors, and
predictors of mortality. Clin Infect Dis 1992; 15:414–21.
Ascioglu S, De Pauw B, Bille J, et al. Analysis of definitions used in
clinical research on invasive fungal infections: consensus proposal for
new, standardized definitions [abstract 1639]. In: Program and abstracts of the 39th Interscience Conference on Antimicrobial Agents
and Chemotherapy. Washington, DC: American Society for Microbiology, 1999.
Edwards JE Jr, Bodey GP, Bowden RA, et al. International conference
for the development of a consensus on the management and prevention
of severe candidal infections. Clin Infect Dis 1997; 25:43–59.
Sobel JD, Kauffman CA, McKinsey D, et al. Candiduria: a randomized,
double-blind study of treatment with fluconazole and placebo. Clin
Infect Dis 2000; 30:19–24.
Kauffman CA, Vazquez JA, Sobel JD, et al. Prospective multicenter
surveillance study of funguria in hospitalized patients. Clin Infect Dis
2000; 30:14–18.
Rodriguez LJ, Anaissie EJ, Rex JH. Pneumonia due to Candida species.
In: Sarosi GA, Davies SF, eds. Fungal disease of the lung. 3d ed. Philadelphia: Lippincott Williams & Wilkins, 2000:115–22.
HEALTHCARE EPIDEMIOLOGY • CID 2001:32 (15 April) • 1199
42. Bernhardt HE, Orlando JC, Benfield JR, Hirose FM, Foos RY. Disseminated candidiasis in surgical patients. Surg Gynecol Obstet 1972; 134:
819–25.
43. Gaines JD, Remington JS. Disseminated candidasis in the surgical patient. Surgery 1972; 72:730–6.
44. Petri MG, König J, Moecke HP, et al. Epidemiology of invasive candidiasis in ICU patients: a prospective multicenter study in 435 nonneutropenic patients. Intensive Care Med 1997; 23:317–25.
45. El-Ebiary M, Torres A, Fabrega N, et al. Significance of the isolation
of Candida species from respiratory samples in critically ill, nonneutropenic patients. Am J Respir Crit Care Med 1997; 156:583–90.
46. Berenguer J, Buck M, Witebsky F, Stock F, Pizzo PA, Walsh TJ. Lysiscentrifugation blood cultures in the detection of tissue-proven invasive
candidiasis: disseminated versus single-organ infection. Diagn Microbiol Infect Dis 1993; 17:103–9.
47. Rex JH, Walsh TJ, Anaissie EA. Fungal infections in iatrogenically
compromised hosts. Adv Intern Med 1998; 43:321–71.
48. Pittet D, Monod M, Suter PM, Frenk E, Auckenthaler R. Candida
colonization and subsequent infections in critically ill surgical patients.
Ann Surg 1994; 220:751–8.
49. Mahr CC, Fildes JJ, Becker EJ, et al. The alarming rate of fungal
recovery in critically ill trauma patients with unresolved sepsis. Crit
Care Med 1995; 23(Suppl 1):A154.
50. Borzotta AP, Beardsley K. Candida infections in critically ill trauma patients: a retrospective case-control study. Arch Surg 1999; 134:657–64.
51. Richet HM, Andremont A, Tancrede C, Pico JL, Jarvis WR. Risk factors
for candidemia in patients with acute lymphocytic leukemia. Rev Infect
Dis 1991; 13:211–5.
52. Goodrich JM, Reed EC, Mori M, et al. Clinical features and analysis
of risk factors for invasive candidal infection after marrow transplantation. J Infect Dis 1991; 164:731–40.
53. Bow EJ, Loewen R, Cheang MS, Schacter B. Invasive fungal disease in
adults undergoing remission-induction therapy for acute myeloid leukemia: the pathogenetic role of the antileukemic regimen. Clin Infect
Dis 1995; 21:361–9.
54. Schwartz RS, Mackintosh FR, Schrier SL, Greenberg PL. Multivariate
analysis of factors associated with invasive fungal disease during remission induction therapy for acute myelogenous leukemia. Cancer
1984; 53:411–9.
55. Karabinis A, Hill C, Leclerq B, Tancrède C, Baume D, Andremont A.
Risk factors for candidemia in cancer patients: a case-control study. J
Clin Microbiol 1988; 26:429–32.
56. Wiley JM, Smith N, Leventhal BG, et al. Invasive fungal disease in
pediatric acute leukemia patients with fever and neutropenia during
induction chemotherapy: a multivariate analysis of risk factors. J Clin
Oncol 1990; 8:280–6.
57. Sandford GR, Merz WG, Wingard JR, Charache P, Saral R. The value
of fungal surveillance cultures as predictors of systemic fungal infections. J Infect Dis 1980; 142:503–9.
58. Aisner J, Schimpff SC, Sutherland JC, Young VM, Wiernik PH. Torulopsis glabrata infections in patients with cancer: increasing incidence
and relationship to colonization. Am J Med 1976; 61:23–8.
59. Marr KA, Seidel K, White TC, Bowden RA. Candidemia in allogeneic
blood and marrow transplant recipients: evolution of risk factors after
the adoption of prophylactic fluconazole. J Infect Dis 2000; 181:309–16.
60. Bross J, Talbot GH, Maislin G, Hurwitz S, Strom BL. Risk factors for
nosocomial candidemia: a case-control study in adults without leukemia. Am J Med 1989; 87:614–20.
61. Saiman L, Ludington E, Pfaller M, et al. Risk factors for candidemia
in neonatal intensive care unit patients. Pediatr Infect Dis J 2000; 19:
319–24.
1200 • CID 2001:32 (15 April) • HEALTHCARE EPIDEMIOLOGY
62. Blumberg HM, Jarvis WR, Wenzel RP, the NEMIS Study Group. Risk
factors for Candida bloodstream infection in surgical intensive care
units: NEMIS prospective multicenter study [abstract 102]. In: Program
and abstracts of the 36th Annual Meeting of the Infectious Diseases
Society of America. Alexandria, Virginia: Infectious Diseases Society
of America, 1998.
63. MacDonald L, Baker C, Chenoweth C. Risk factors for candidemia in
a children’s hospital. Clin Infect Dis 1998; 26:642–5.
64. Calandra T, Bille J, Schneider R, Mosimann F, Francioli P. Clinical
significance of Candida isolated from peritoneum in surgical patients.
Lancet 1989; 2:1437–40.
65. Eggimann P, Francioli P, Bille J, et al. Fluconazole prophylaxis prevents
intraabdominal candidiasis in high-risk surgical patients. Crit Care Med
1999; 27:1066–72.
66. Fridkin SK, Jarvis WR. Epidemiology of nosocomial fungal infections.
Clin Microbiol Rev 1996; 9:499–511.
67. Craven DE, Kunches LM, Lichtenberg DA, et al. Nosocomial infection
and fatality in medical and surgical intensive care unit patients. Arch
Intern Med 1988; 148:1161–8.
68. Savino JA, Agarwal N, Wry P, Policastro A, Cerabona T, Austria L.
Routine prophylactic antifungal agents (clotrimazole, ketoconazole,
and nystatin) in nontransplant/nonburned critically ill surgical and
trauma patients. J Trauma 1994; 36:20–6.
69. Ables AZ, Blumer NA, Valainis GT, Godenick MT, Kajdasz DK, Palesch
YY. Fluconazole prophylaxis of severe Candida infection in trauma and
postsurgical patients: a prospective, double-blind, randomized, placebo-controlled trial. Infect Dis Clin Pract 2000; 9:169–75.
70. Garbino J, Lew D, Romand J-A, Auckenthaler R, Pittet D. Prevention
of severe Candida infections in nonneutropenic, high-risk, critically ill
patients: a randomized, double-blind, placebo-controlled trial [abstract
58.004]. In: Program and abstracts of the 9th International Congress
on Infectious Diseases (Buenos Aires). Previously published in part as:
Garbino J, Lew D, Romand J-A, Auckenthaler R, Suter P, Pittet D.
Fluconazole prevents severe Candida species infections in high-risk
critically ill patients: a randomized, double-blind, placebo-controlled
study [abstract LM-23b]. In: Program and abstracts of the 37th Interscience Conference on Antimicrobial Agents and Chemotherapy.
Washington, DC: American Society for Microbiology, 1997.
71. Pelz RK, Hendrix CW, Swoboda SM, Merz WG, Lipsett PA. A doubleblind placebo controlled trial of prophylactic fluconazole to prevent
Candida infections in critically ill surgical patients [abstract 43]. Crit
Care Med 2000; 27(Suppl):A33.
72. Zar JH. Biostatistical analysis. 2d ed. Englewood Cliffs, NJ: PrenticeHall, 1984.
73. Pelz RK, Hendrix CW, Swoboda SM, Merz WG, Lipsett PA. Fluconazole
blood concentrations after enteral administration in critically-ill surgical patients exceed most Candida minimal inhibitory concentrations
in a double-blind, placebo-controlled trial in which fluconazole prevented candidal infections [abstract 41]. Crit Care Med 2000;
27(Suppl):A33.
74. Rex JH, Walsh TJ, Sobel JD, et al. Practice guidelines for the treatment
of candidiasis. Clin Infect Dis 2000; 30:662–78.
75. British Society for Antimicrobial Chemotherapy Working Party. Management of deep Candida infection in surgical and intensive care unit
patients. Intensive Care Med 1994; 20:522–8.
76. 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; 24:206–16.
77. Arikan S, Rex JH. New agents for treatment of systemic fungal infections. Emerging Drugs 2000; 5:135–60.