Intensive Care Med (1997) 23: 317–325

 Springer-Verlag 1997 ORIGINAL

M.G. Petri1 Epidemiology of invasive mycosis in ICU

J. König2
H.P. Moecke3 patients: a prospective multicenter study
H.J. Gramm4
H. Barkow5 in 435 non-neutropenic patients
P. Kujath6
R. Dennhart7
H. Schäfer8
N. Meyer9
P. Kalmar10
P.Thülig11
J. Müller12
H. Lode1
Paul-Ehrlich Society for Chemotherapy,
Divisions of Mycology and Pneumonia Research

Received: 8 February 1996

Accepted: 30 October 1996
)
H. Lode ( ) ⋅ M. G. Petri
City Hospital Zehlendorf/Heckeshorn,
Free University Berlin, Department of
Chest and Infectious Diseases, Zum
Heckeshorn 33, D-14 109 Berlin, Germany
Fax: + 49(30)8002 2623
Participating Centers:
1
City Hospital Zehlendorf/Heckeshorn,
Department of Chest and Infectious
Diseases
2
Universität Heidelberg, Zentrum zur
Methodischen Betreuung von Therapie-
studien (ZMBT)
3
Allgemeinkrankenhaus Altona/Hamburg,
Anästhesie-Abteilung
4
Universitätsklinikum Steglitz/Berlin,
Klinik für Anästhesiologie und Operative
Intensivmedizin
5
Universitätsklinikum Rudolf-Virchow/
Berlin, Abteilung für Reanimation der
Medizinischen Klinik und Poliklinik
6
Universitätsklinik Würzburg,
Chirurgische Klinik
7
Krankenhaus Nordwest/Frankfurt,
Anästhesie-Abteilung
8
Universitätsklinikum Steglitz/Berlin,
Internistische Intensivstation
9
Ruhr-Universität-Bochum/Herne,
Chirurgische Klinik
10
Universitätskrankenhaus Eppendorf/
Hamburg, Chirurgische Klinik
11
Universitätsklinikum Münster,
Klinik für Anästhesiologie und
Intensivmedizin
12
Universität Freiburg, Institut für
Parasitologie und Mykologie
Abstract Objective: To determine
the epidemiological and clinical sig-
nificance of invasive fungal infections
in non-neutropenic patients in inten-
sive care who stay longer than 10 days
on the intensive care unit (ICU).
Design: Prospective epidemiological
multicenter study over a period of
11 months, based on strict clinical,
bacteriological, serological and his-
tological criteria.
Setting: Six surgical and two medical
ICUs units in five university and two
municipal hospitals.
Patients: 435 non-neutropenic pa-
tients from medical and surgical
ICUs with an ICU stay of more than
10 days.
Measurements and main results: A
new occurrence of invasive mycosis
(3 sepsis/ 4 peritonitis/ 1 disseminated
candidiasis), corresponding to the
protocol conditions with onset after
day 10 in the ICU, was detectable in
2.0 % (95 % confidence interval 0.85
to 3.8 %) of the 409 patients who
could be assessed. Candida species
were identified as an infection-rele-
vant pathogen in all cases. The most
important risk factor for the devel-
opment of an invasive mycosis was
the onset of peritonitis by the day 11
in the ICU (odds ratio 11.3;
p = 0.003). A fungal colonization was
detected in 64 % of patients (Candida
species 56 %, Aspergillus 4 %, and
other fungi). Six of 8 patients with an
invasive mycosis died on the ICU;
ICU mortality in patients with fungal
colonization was 31 % and in noncol-
onized patients 26 %. Serological
tests were not helpful clinically. The
sensitivity was 88 % for the Candida
HAT (haemagglutination test)
(threshold titer > 1:160), 100 % for
the Candida IFT (immunofluores-
cence test) (threshold titer > 1:80),
and 50 % for the Candida Antigen
Test (Candtec Ramco, threshold ti-
ter ≥ 1:8), and the specificity was 26,
6, and 73 %, respectively. The speci-
ficity for the Aspergillus HAT
(threshold titer > 1:10) was 29 %.
Conclusions: Invasive mycoses are
rare in non-neutropenic ICU pa-
tients, even after a longer stay in the
intensive care unit; fungal coloniza-
tion, on the other hand, is frequently
detectable. The mortality of invasive
mycosis – even with systemic antimy-
cotic therapy- was high; the mortality
in patients with fungal colonization
was not significantly increased com-
pared to that in noncolonized pa-
tients. The serological test proce-
dures, Candida HAT, Candida IFT,
and the Candida Ramco Antigen
Test, had a low specificity and were
not helpful in diagnosing relevant in-
vasive mycosis.
Key words Invasive mycosis ⋅ ICU
patients ⋅ Epidemiology ⋅
Prospective multicenter study
318
Introduction
Clinical and postmortem studies describe an increasing
incidence of mycosis worldwide, mainly in leukemia pa-
tients and in transplantation and burn patients, as well
as in patients with solid tumors [1–11, 14, 27]. The pri-
mary fungal species involved here are Candida (particu-
larly C. albicans) and Aspergillus. There have been nu-
merous international investigations on the development
of fungal infections in known groups of patients at risk
but only a few epidemiological studies on the incidence
and significance of relevant mycoses in non-neutropenic
patients. Nosocomial infections are seen more fre-
quently in the intensive care unit (ICU), with an even
higher incidence among surgical ICU patients than in
other areas of the hospital, since nearly all patients are
managed with intravascular catheters, indwelling uri-
nary catheters, mechanical ventilation and other inva-
sive measures [12–17]. Fungi are frequently detected in
cultures of tracheal and wound secretions, stool, urine,
and blood in longstay ICU patients [18–21]. However,
the epidemiological and therapeutic significance of
these findings in terms of a fungal colonization or inva-
sive mycosis are unclear and are constantly debated dur-
ing routine care. Based on strict and recognized defini-
tions, the present study therefore aimed (a) at determin-
ing the incidence of invasive mycoses in non-neutropen-
ic ICU patients after a stay of 10 days (b) and at identi-
fying possible risk factors for colonization and infection.
In view of the advances in serological detection proce-
dures and antigen tests in fungal diagnostics in recent
years [22–24] and the frequent uncritical assessment of
their diagnostic value, we also recorded clinical data
and symptoms, bacteriological and mycological patho-
gen detection, and antibody/antigen titers.
Patients and methods
Study design
The inclusion criterion was a stay of more than 10 days in the ICU.
Patients who were under 18 years of age, who developed a detect-
able invasive mycosis between days 1 and 10 of hospitalization, or
who had a life expectancy of less than 5 days were excluded. A to-
tal of 435 patients who remained in the ICU for more than 10 con-
secutive days were enrolled in the period from 1 July 1989 to 1 June
1990. All 26 patients in one center were excluded because consecu-
tive recruitment could not be confirmed and the data were ques-
tionable (no reported invasive mycoses). Results from 409 patients
in eight ICUs are reported (Steglitz Medical Center, FU Berlin:
medical and surgical ICU; Rudolf-Virchow Medical Center/Wes-
tend, FU Berlin: medical ICU; Northwest Hospital, Frankfurt: sur-
gical ICU; Altona General Hospital, Hamburg: surgical ICU; Uni-
versity Hospital Eppendorf, Hamburg: surgical ICU; Ruhr Univer-
sity/Maria Hospital, Herne: surgical ICU; Würzburg University
Hospital: operative ICU). Follow-up was performed until death
or transfer of the patient from the ICU.
Every study patient was evaluated according to the Acute
Physiology and Chronic Health (APACHE) II score [25] on the
day of admission to the ICU. For assessment of the clinical course,
each patient was evaluated according to the multiple organ failure
(MOF) score [26] at the beginning of the study, and then every
5 days up to day 40 and at 10-day intervals thereafter. In addition,
all ICU measures for recording vital functions were documented
on a daily basis. At the start of the study, a chest X-ray was taken
with radiological follow-up at intervals of at least 5 days. The es-
sential laboratory data were recorded from the start of the study
at the same intervals.
Definitions of an invasive mycosis
Clinical definitions
The diagnosis of a fungal sepsis was confirmed if at least two of
three criteria were fulfilled: (1) two positive blood cultures (exclu-
sion of other pathogens) or fungus detection on catheter tip; (2) se-
vere septic clinical picture (failure of at least one organ, e. g., kid-
ney, lung, coagulation, etc.); (3) septic fungal metastases (e. g.,
eyes, liver, skin, brain).
Histological definition
For proof of an invasive mycosis, there had to be distinct invasive
growth in the tissue with corresponding conclusive evidence in bi-
opsy material, e. g., from the esophagus, lung, liver, peritoneum,
bone, pleura, spleen, colon, or duodenum.
Microbiological or serological definition
The following were required for the microbiological definition of
an invasive mycosis: (a) pathogen detection in primarily sterile me-
dia with an organism count of ≥ 102/ml; (b) pathogen detection in
primarily nonsterile media with > 105 organisms/ml; (c) a fourfold
increase in antibody titers in the serology [Candida haemaggluti-
nation test (HAT), Candida immunofluorescence test (IFT)]; (d)
significant antigen detection (Candida Antigen Test, Candtec
Ramco Lab.: tumor patients ≥ 1:4, ICU patients ≥ 1:8; Cryptococ-
cus Latex Antigen Test: ≥ 1:4).
For a confirmed disseminated candidiasis , the following were
required: (a) Candida detected in a blood culture and evidence of
dissemination with endophthalmitis or hepatorenal involvement,
or (b) Candida isolated in a blood culture with biopsy-proven Can-
dida infection in an otherwise sterile area, or (c) Candida detected
in a biopsy specimen or culture from a deep visceral organ without
a positive blood culture.
A fungal pneumonia was regarded as confirmed in the presence
of an infiltrate persisting for more than 2 days, as well as at least
two of the following criteria: (a) detection of ≥ 102 fungi/ml in the
bronchoalveolar lavage, (b) histological confirmation, or (c) exclu-
sion of other pathogens.
The following could be applied in addition to the above clinical
findings as a basic for assuming an invasive mycosis: mycoserologi-
cal information, an increase in antibodies by at least 4 titer levels,
or a significant antigen provocation test. Every detection of fungi
on mucous membranes, in wounds and in urine was assessed pri-
marily as colonization. Demonstration of counts > 103/ml required
testing for an invasive mycosis according to the above definitions.
Fungus detection in blood cultures called for removal of all intra-
vascular catheters and both bacteriological and mycological exam-
 319

Table 1 Demographic data Surgical patients (%) Medical patients (%)

(n = 310) (n = 99)
Underlying diseases
Trauma (23 %) 30 1
Solid tumors (18 %) 21 9
Diseases of the heart/coronary vessels (14 %) 4 45
Diseases of the gastrointestinal tract (10 %) 13 2
Main admission diagnosis
Trauma (21 %) 27 1
Diseases of the gastrointestinal tract (17 %) 23 0
Diseases of the respiratory tract (13 %) 9 25
Diseases of the heart/coronary vessels (13 %) 5 39
Accompanying illnesses (78 %) 74 89
Diseases of the cardiocirculatory system (36 %) 34 44
Alcohol abuse (14 %) 12 20
Diseases of the liver (without alcohol abuse) (11 %) 13 7
Kidney diseases (9 %) 7 13
Diseases of the gastrointestinal tract (10 %) 12 5
Endocrinological diseases (without diabetes) (5 %) 3 13
Diabetes mellitus (15 %) 13 21
Neurological diseases (10 %) 8 17
Depressed immune status (7 %) 5 15

inations. At an interval of 30 min to 24 h after removal of the cath- achieved. The results are given as adjusted odds ratios and 95 %
eters, new blood cultures were required for exclusion of an invasive confidence intervals [27–29].
mycosis.

Microbiology
At the start of the study (day 11 in the ICU), it was necessary to
perform bacteriological and mycological tests in the local labora-
tory with blood cultures, tracheal secretions, urine specimens, stool
specimens, and, in some cases, wound swabs, drainage, throat
swabs, cerebrospinal fluid, and sputum and to repeat these tests at
5-day intervals to the day of transfer out of the ICU or death.
Serology
The mycoserological investigation required serum and urine sam-
ples from each patient for fungal diagnostic tests 5 days before the
start of the study (day 11) and at 5-day intervals thereafter. The
samples were examined blind by the mycological reference labora-
tory (Institute for Parasitology and Mycology, University of Frei-
burg, J.Müller); Roche indirect Candida HAT, Roche indirect
IFT, Candida Antigen Test (Candtec, Ramco Lab.), indirect Roche
Aspergillus HAT, Cryptococcus Antigen Latex Test. The following
threshold titers were defined as pathognomonic by the mycological
reference laboratory: Candida HAT > 1:160, Candida IFT > 1:80,
Candida Antigen Test ≥ 1:4, Aspergillus HAT > 1:10.
Biostatistics
If not otherwise stated, dependencies among discrete variables
were investigated by cross-tables and the chi-square test. Predis-
posing factors for the development of invasive mycosis were stud-
ied by a logistic model which makes it possible to describe the
role of one risk factor while adjusting for possible interference
from others. Particularly, stratification for center effects is
Results
Patient selection
In the course of the 11-month study period, 435 patients
who fulfilled the inclusion criteria of the study were ad-
mitted to the ICUs. Twenty-six patients were excluded
because of protocol errors (absent or inadequate data),
and 5 patients under 18 years of age (2 patients aged
14, 1 patient aged 15, and 2 patients aged 17 years) re-
mained in the study; 409 patients were therefore consid-
ered in the final evaluation.
Patients
The surgical patients comprised more than 76 % of the
group examined and medical patients 24 % (Table 1).
The ages of the 246 male (60 %) and 163 female (40 %)
patients ranged from 14 to 88 (median 59) years; none
was neutropenic (granulocytes < 1000/ml) during the in-
vestigation. In 407 patients, the APACHE II score was
between 0 and 39 (median 14) on the day of admission
to the ICU; no assessment was done in 2 cases. The
MOF score on the study day 1 (day 11 in the ICU) was
between 0 and 12 (median 3). Of the 409 patients, 95 %
were mechanically ventilated continuously or intermit-
tently (87 % medical ICUs vs 98 % surgical ICUs, 1–
140 days, mean intubation time 23 days). An operation
320

Table 2 Patients with proven invasive mycosis

Patient Age Underlying disease, Primary location Specimens positive Diagnosis Survival Systemic treatment
(years) APACHE II (AP) of invasive mycosis based on:
gender
ES 44, F Resection of colon Sepsis Blood culture, Clinical findings, No Amphotericin B,
(chronic inflamma- abdominal drain microbiology flucytosine
tion) AP 7
SM 35, F Perforation of small Peritonitis Abdominal drain Clinical findings, Yes Fluconazole
intestine AP 2 microbiology
RG 70, M Mesenteric infarc- Peritonitis Liver biopsy Clinical findings, No No
tion AP 26 microbiology
AM 59, F Subarachnoid Sepsis Blood culture, Clinical findings, No Amphotericin B,
hemorrhage AP 8 central venous line, microbiology flucytosine
puncture
RR 34, M Chronic inflamma- Sepsis Blood culture, Clinical findings, No Amphotericin B,
tion of intestine central venous line microbiology flucytosine
AP 17
EW 58, F Complete hyster- Peritonitis Abdominal drain Clinical findings, Yes Amphotericin B,
ectomy AP 10 microbiology flucytosine
GM 58, M Multiple injuries Peritonitis Abdominal drain Clinical findings, No Amphotericin B,
AP 16 microbiology flucytosine
KL 37, F Chronic myeloid Disseminated Autopsy No No
leukemia AP 13 candidiasis

preceded the ICU stay in 81 % of patients on the surgi-
cal wards and in 17 % on the medical wards. This was
performed between ICU days 2 and 10 in 7 and 14 % of
the patients, respectively, and 5 and 14 %, respectively,
first underwent surgery on or after the day 11 (start of
study). In 96 % of patients, at least one antibacterial
drug was given, and 34 % were given antimycotical
drugs.The most frequent clinical findings documented
were bronchitis (59 %), pneumonia (48 %), urinary tract
infection (21 %), sepsis (20 %), peritonitis and wound
infection (18 % each), and enteritis/colitis (15 %).
Incidence of invasive mycosis
A diagnosis of invasive candidiasis according to the cri-
teria applied was made in 8 of 408 patients (2 %). In 1
further case, there was a preexisting invasive mycosis
(confirmed Candida infection and suspicion of addi-
tional Aspergillus involvement); the diagnosis of an in-
vasive mycosis (4 Candida peritonitis, 3 Candida sepsis,
1 disseminated candidiasis) was based on clinical find-
ings and results of microbiological cultures in 7/8 cases
and on the postmortem findings in 1 case (Table 2). Six
of the 8 patients died in the ICU stay (7, 14, 18, 30, 42,
and 129 days after the start of the study). Six patients re-
ceived systemic antimycotic therapy with amphotericin
B and flucytosine (5) or fluconazole (1). Four of the 6
patients on antimycotic treatment died (all receiving
amphotericin B + flucytosine), as well as the two pa-
tients who were not treated. In 6 of the 8 patients with
demonstrated mycosis, peritonitis was already docu-
mented at the time of ICU admission. Two patients sur-
vived the invasive mycosis and had no signs of a sepsis
during the clinical course. In 1 case, pneumonia findings
concurred with documented Candida sepsis; the pneu-
monia pathogen was not clarified by bronchoalveolar la-
vage or autopsy. The only patient from the medical
ICUs (chronic myelocytic leukemia, CML; minimally
1700 granulocytes/ml) should be regarded as an excep-
tion to the rest of the group, since multiple organ in-
volvement with Candida (lung, liver, spleen, myocar-
dium, thyroid, kidney, and choroid plexus) was only
demonstrated by autopsy (previous positive fungal find-
ings in the tracheal secretions and urine). Fungal sepsis
as the expression of a relevant invasive mycosis was
found in 3 of our patients, all of whom had received
four or more antibiotics before the onset of mycosis.
Candida was the only pathogen detected in time in
blood cultures in 2 patients; a 3rd patient, additionally,
had polymicrobial findings from the abdominal cavity.
All 3 patients died (1 of septic-toxic circulatory failure,
1 of heart failure, and 1 of MOF).
Colonization
Fungal colonization was detected from all media exam-
ined in 64 % of the patients (Candida species 56 %, As-
pergillus species 4 %, and other fungi 14 %; multiple
sites of entry was possible). If only primarily sterile me-
dia are considered, fungal growth was found in 9 %
 321

(Candida species 8 %, Aspergillus species < 1 %). Can- Table 3 Sensivity and specificity of serological findings
dida species was most frequently demonstrated in tra- Invasive mycosis Colonization
cheal secretions (36 %), throat swabs (27 %), urine
Sensivity Specificity Sensivity Specificity
(25 %), stool cultures (11 %), and wound swabs (6 %). (%) (%) (%) (%)
Candida was detected to a markedly less extent in aero-
bic (2 %) and anaerobic (1 %) blood cultures, from Candida HAT
> 1 : 160 88 26 79 31
(peritoneal) drainage (4 %) and catheter tips (3 %). All > 1 : 320 75 47 61 56
patients with invasive Candida infection were previ-
Candida IFT
ously colonized. Aspergillus species could only be culti- > 1 : 80 100 6 95 6
vated to an appreciable extent from tracheal secretions > 1 : 160 75 17 86 21
(3 %); in the other media, detection was either < 1 % or Candida AG
no growth at all could be documented. ≥ 1:4 75 48 53 48
In considering the time course of positive fungal de- ≥ 1:8 50 73 28 74
tection, a fairly constant high level was found; after a
slight increase starting on study day 1, the maximal fre-
quency of 47 % from all media was reached on study
days 10 and 15. Positive blood cultures were first de-
tected as of study day 5; urine cultures showed a gradual
increase in frequency up to study day 25.

Serological findings

Two separate groups were examined: (1) all patients

who did not develop an invasive mycosis during the clin-
ical course, and (2) the patients with invasive mycosis,
regardless of the time of onset. In all patients with inva-
sive mycosis, positive antibody findings preceded the
clinical diagnosis. The threshold values defined as
pathognomonic were already exceeded before study
day 1 in 4 cases of the HAT titer and in 6 cases of the
IFT titer. Except for the patient with CML (only IFT),
all patients with invasive mycosis had positive antibody
findings for Candida HAT and IFT, with the titer levels
defined as pathognomonic (HAT > 1:160, IFT > 1:80).
An increase in the Candida Antigen titer of ≥ 1:8 was
observed in 5 of the 9 patients (56 %). Maximal titers at
1:2 were observed in 2 patients and titers of up to 1:4 in
2 others. Taking as a basis the maximal titers during the
clinical course, there was an excess of false-positive
findings (Table 3, Fig. 1) for colonized patients without
invasive mycosis as opposed to noncolonized patients.
Predisposing factors for an invasive mycosis
The most prominent risk factor for the development of
an invasive mycosis was the clinical demonstration of
peritonitis by study day 1 (odds ratio 11.3; p = 0.003).
The only invasive mycosis in which peritonitis was not
already present on admission to the ICU or by day 1 at
the latest occurred in 1 patient with CML and in 1 pa-
tient with status postsubarachnoid hemorrhage (onset
of mycosis only on day 117). In contrast, peritonitis oc-
curred in only 20 % of the remaining patients without a
diagnosis of mycosis (Table 4). With only a slight loss of
Fig. 1 Percentage of patients with or without invasive Candida my-
cosis with positive serological tests
information, the risk factor “peritonitis by ICU day 11”
could be replaced by other characteristics already de-
tectable on admission. Accordingly, the factors “abdom-
inal disease/injury” (odds ratio 5.14; p < 0.03) and “gas-
trointestinal disease/injury” (odds ratio 8.14; p < 0.005)
had prognostic significance for an invasive mycosis com-
parable to that of peritonitis by study day 1. Abdominal
drainage (odds ratio 10.0; p = 0.005) could be consid-
ered as a further significant indicator for invasive myco-
sis. Even if the 2 survivors with invasive mycoses were
excluded from analysis, the factor abdominal drainage
remains significant (p = 0,03). Five of the immunosup-
pressed patients had undergone transplantation (all
with status postrenal transplantation); none was granu-
locytopenic. Three of these patients received antimy-
cotic preventive treatment and none developed mycosis.
If the group of immunosuppressed patients is further
limited to the 6 patients with leukemia or lymphoma
(all without neutropenia), the occurrence of 1 case of
mycosis in this group – despite the small number of
322

Table 4 Risk factors for devel- Characteristics Present No. Incidence Odds ratio
oping invasive mycosis of invasive (95 % confidence
mycosis in 0/00 interval)
Peritonitis before day 11 at ICU No 318 6 11.3
Yes 90 67 (2.24 to 56.9)
p = 0.003
Abdominal drainage No 310 7 10
Yes 98 61 (1.99 to 60.6)
p = 0.005
Abdominal diseases/injuries No 305 10 5.14
(incl. neoplasms, no multiple injuries) Yes 103 49 (1.21 to 21.9)
p < 0.03
Gastrointestinal diseases/injuries No 335 9 8.14
(incl. neoplasms, no multiple injuries) Yes 73 69 (1.90 to 34.9)
p < 0.005

cases – means a significantly increased incidence (odds
ratio 11.3; p < 0.05). The presence of further known risk
factors, such as diabetic metabolism, hepatic cirrhosis,
and severe renal insufficiency, did not cause a significant
increase in the incidence of mycosis. Consideration of
neoplasms also proved to be of no prognostic relevance.
Prognostic factors for mortality
The overall mortality during hospitalization was 33 %
(133/409; including 2 patients who died after transfer
from the ICU) and was significantly influenced by the
following: center, APACHE II score, MOF score on
study day 1, age, and accompanying illnesses like car-
diac insufficiency, hepatic cirrhosis, and diabetes. In the
combined multivariate analysis of these features in the
logistic model, the APACHE II score and diabetes lost
their significant influence, while the other features re-
tained their importance (model 1 in Table 5). If the un-
derlying diseases are grouped with reference to the In-
ternational Classification of Diseases (ICD) 9 classifica-
tion, an influence on mortality was only evident if no ad-
justment was made according to the APACHE II or the
MOF score (models 2–4 in Table 5). Six of the 8 patients
who developed an invasive mycosis with onset after
ICU day 11 died in the ICU. The patient with preexist-
ing mycosis was transferred to a peripheral ward on
ICU day 63. Mortality was significantly increased
among patients with invasive mycosis by study day 1
compared to the other patients; when all 9 cases of my-
cosis were included, the odds ratio was 4.4 (p = 0.038).
The impact of invasive mycosis on mortality was ana-
lyzed in models 5–10 in Table 5. Each of the models
showed a significant increase in mortality related to in-
vasive mycosis. Of 400 patients without invasive myco-
sis, 118 died in the ICU (30 %). In 251 patients (63 %),
at least one positive fungus culture was detected. Mor-
tality for these patients was 31 % (79 of 251) versus
26 % (39 of 149) of the patients without fungal coloniza-
tion (p = 0.26; two-sided chi-square test).
Discussion
Invasive measures of intensive care medicine (e. g., arte-
rial and central venous catheters, bladder catheters, me-
chanical ventilation, etc.) create numerous potential
portals of entry for bacterial and mycotic pathogens. In
addition, treatment with a multiplicity of antibiotics
leads to germ selection [13]. Both factors cause an iatro-
genic increase in the predisposition for infections even
in non-neutropenic ICU patients. Moreover, the under-
lying disease is usually associated with a poor general
condition. As the stay in the ICU grows longer, an in-
crease is observed in fungal detection on mucous mem-
branes as well as in the urine and blood [18, 20, 21]; the
significance of fungal colonization for the development
of a relevant infection and the influence on mortality,
however, is still unclear.
Based on strict clinical and microbiological criteria,
the present investigations revealed a low incidence of
invasive mycoses in non-neutropenic patients (8 pa-
tients with onset after ICU day 11; 2 %, with a 95 % con-
fidence interval of 0.85 to 3.8). These were almost exclu-
sively patients on the surgical ICUs (7/8); 1 case (CML)
was a medical ICU patient (total number of surgi-
cal:medical patients ∼ 3:1). In clinical investigations,
only specific nosocomial clinical entities in connection
with fungi in comparable patient groups or single obser-
vations in different ICUs have been reported [14, 15,
17]. However, a study from Geneva [30] in adult surgical
patients reported identical results (1.7 %). The infection
rates vary both between the different hospitals and be-
tween the wards of individual hospitals. Candida species
was isolated as an infection-relevant pathogen in all in-
vasive mycoses in our study; this agrees with the data in
the literature [13, 31]. A fungal sepsis (monoinfection
 323

Table 5 Prognostic factors for Characteristics Evaluated characteristic(s) Odds ratio (95 %
non-survival confidence interval)
p-value
Model 1. Centers, age, MOF (day 11 Age 1.36
in ICU), cirrhosis of liver, diabetes, (1.17 to 1.58)
cardiac insufficiency p < 0.0001
MOF (day 11 at ICU) 1.81
(1.58 to 2.09)
p < 0.0001
Cirrhosis of liver 4.87
(1.50 to 15.8)
p < 0.01
Cardiac insufficiency 2.65
(1.21 to 5.78)
p = 0.015
Diabetes mellitus 1.76
(0.81 to 3.78)
p = 0.15
Model 2. Underlying diseases (grouped Effect of underlying diseases p = 0.019
according to ICD score), centers
Model 3. Underlying diseases (grouped Effect of underlying diseases p = 0.54
according to ICD score), MOF (day 11 at
ICU), centers
Model 4. Underlying diseases (grouped Effect of underlying diseases p = 0.10
according to ICD score), APACHE II
score, centers
Model 5. Invasive mycosis during ICU stay Invasive mycosis 4.39
(p = 0.04)
Model 6. Invasive mycosis during ICU stay, Invasive mycosis 9.3
APACHE II score, centers (p = 0.006)
Model 7. Invasive mycosis during ICU stay, Invasive mycosis 12.2
APACHE II score, centers, age score, (p = 0.005)
cardiac insuff., cirrhosis of liver, renal
diseases, neurologic diseases, neoplasms,
gastrointestinal diseases
Model 8. Invasive mycosis during ICU stay, Invasive mycosis 13.9
MOF (day 11in ICU), centers (p = 0.002)
Model 9. Invasive mycosis, centers, age Invasive mycosis 14.0
score, MOF (day 11in ICU), cirrhosis of (p = 0.0023)
liver, diabetes, cardiac insufficiency
Model 10. Invasive mycosis during ICU Invasive mycosis 20.7
stay, centers, age score, MOF (day 11 (p = 0.0007)
in ICU), cirrhosis of liver, diabetes,
cardiac insufficiency

with Candida species) as the expression of a relevant in-
vasive mycosis was found in 3 of our patients. All 3 pa-
tients died. Not in every case, however, does fungemia
actually prove to be a disseminated mycosis. It may be
assumed that Candida can be detected in blood (and ur-
ine) from immunocompetent patients as evidence of a
temporary asymptomatic fungemia without the exis-
tence of a severe mycosis [32, 33]. In 4 patients in our
study, blood cultures showed a single positive fungus
culture without signs of sepsis. Two patients were trans-
ferred to peripheral wards in improved status. Two pa-
tients had ≥ 3 negative blood cultures. The fact that pos-
itive blood cultures were only found in 50 % of the pa-
tients with invasive mycosis can be ascribed to the late
dissemination or to the known problems of Candida de-
tection in blood cultures [34, 35], since circulating fun-
gus cells are largely caught in the peripheral capillaries
[23]. In the analysis of our study with respect to the risk
of an invasive mycosis, peritonitis by ICU day 11 was as-
sessed as the prognostically most important factor (odds
ratio 11.3; p = 0.003), but this could be replaced, with
only slight informational loss, by factors already recog-
nizable on the day of admission: “abdominal disease/in-
jury” (odds ratio 5.14; p < 0.03) and “gastrointestinal
324

disease/injury” (odds ratio 8.14; p < 0.005). Regarding
an increase in the risk of mycosis, nonlinear correlations
must also be expected, since the ICU stay may be pro-
longed in cases with a moderately severe general condi-
tion but will tend to be short in those in a particularly
poor condition and thus a higher mortality. The results
were confirmed in the logistic model with significant
odds ratios in different stratification models (odds ratios
4.3–24.7). Despite high significance and extreme odds
ratios, it must be kept in mind that the interpretation is
based on only 8 cases with invasive mycosis.
Other studies using uni- and multivariate analyses for
the risk of invasive mycoses in medical and surgical pa-
tients describe different risk factors with varying refer-
ence and significance: previous antibiotic treatments,
isolation of Candida species in several body regions
apart from the blood, duration of mechanical ventila-
tion, previous hemodialysis, use of catheters, azotemia,
transfer from another hospital, diarrhea, and candiduria
[5, 13, 36]. Positive fungus findings from the abdominal
cavity were detectable in another 12 patients in our
study without confirmation of invasive mycosis (only
once vs 2 or more in patients with invasive mycosis);
these findings were assumed to be a result of contamina-
tion or suture dehiscence. The generally recognized risk
factor “immunosuppression” was less significant in the
present study (odds ratio 1.64; p = 0.65). An explanation
would be the low proportion of immunosuppressed pa-
tients. The immunosuppressed patient group in our
study frequently received no systemic antimycotic pro-
phylaxis and had only a negligible increased incidence
compared to the rest of the patients.
Colonization
Fungal colonization in our study was of a higher order of
magnitude at 43 % in tracheal secretions and 28 % in ur-
ine (largely Candida species in each case) then in the in-
vestigation of 101 consecutively registered, ventilated
ICU patients (mainly elective heart surgery and medical
emergencies) in Finland [21]. These authors found 16 %
in tracheal secretions (all Candida growth in tracheal se-
cretions) and in 8 % in urine 105 colony-forming units/
ml of urine) and described a correlation between the in-
tubation period and microbial colonization. The higher
colonization rates we observed could presumably be at-
tributed to the different patient groups. An increased
mortality was observed for colonized (31 %) compared
to noncolonized (26 %) patients, but the difference was
not significant (p = 0.25; two-sided chi-square test). In
our study, the ICU stay was markedly longer, with a
mean of 33 days, for patients with fungal colonization
(identical for nonsurvivors and survivors) than for those
without colonization (mean 21 days; nonsurvivors
20 days, survivors 22 days).
Serology
The measurement of so-called significant titers prior to
the onset of invasive mycosis may be a helpful indica-
tion for early recognition and thus initiation of early
treatment. With the commercially available test proce-
dures we used and the threshold titers we selected as
pathognomonic (Candida HAT > 1:160, Candida
IFT > 1:80, Candida Antigen Test (Ramco) ≥ 1:4), how-
ever, this was not possible because of the numerous
findings that were false-positive according to the defini-
tion. Determination of the Aspergillus HAT for diagno-
sis of an invasive aspergillosis appears to be unimpor-
tant in the presence of a low specificity (threshold titers
1:10; specificity 29 %); this agrees with the reports of
other investigators [22, 37].
Clinical findings and relevant culture results must
therefore continue to form the basis for the diagnosis
of invasive mycosis and initiation of antimycotic therapy
in non-neutropenic ICU patients.

References
1. Horn R, Wong B, Kiehn TE, Armstrong
D (1985) Fungemia in a cancer hospital:
changing frequency, earlier onset and
results of therapy. Clin Infect Dis 7:
646–655
2. Beck-Sague CM, Jarvis WR (1993) The
national nosocomial infections surveil-
lance system: secular trends in the epi-
demiology of nosocomial fungal infec-
tions in the United States 1980–1990. J
Infect Dis 167: 1247–1251
3. Meunier-Carpentier F, Kiehn TE, Arm-
strong D (1981) Fungemia in the immu-
nocompromised host: changing pattern,
antigenemia, high mortality. Am J Med
71: 363–370
4. Richet HM, Andremont A, Tancrede C,
Pico JL, Jarvis WR (1991) Risk factors
for candidemia in patients with acute
lymphocytic leukemia. Clin Infect Dis
13: 211–215
5. Wey SB, Mori M, Pfaller MA, Woolson
RF, Wenzel RP (1988) Hospital-ac-
quired candidemia: the attributable
mortality and excess length of stay.
Arch Intern Med 148: 2642–2645
6. Schaberg DR, Culver DH, Gaynes RP
(1991) Major trends in the microbial
etiology of nosocomial infection. Am J
Med 91[Suppl 3B]:72s–75s
7. Rex JH, Bennett JE, Sugar AM, Pappas
PG, van der Horst CM, Edwards JE,
Washburn RG, Scheld WM, Karchmer
AW, Dine AP, Levenstein MJ, Webb
CD, Candidemia Study Group (1994)
A randomized trial comparing flucon-
azole with amphotericin B for the treat-
ment of candidemia in patients without
neutropenia. N Engl J Med 331: 1325–
1330

8. McGowan JE Jr (1985) Changing etiol-
ogy of nosocomial bacteremia and fun-
gemia and other hospital-acquired in-
fections. Clin Infect Dis 7[Suppl
3]:S357–S370
9. Bodey GP (1966) Fungal infections
complicating acute leukemia. J Clin Ep-
idemiol 19: 667–687
10. Hart PD, Russell E Jr, Remington JS
(1969) The compromised host and in-
fection: II. deep fungal infection. J In-
fect Dis 120: 169–189
11. Sandford GR, Merz WG, Wingard JR,
Sarache P, Saral R (1980) The value of
fungal surveillance cultures as predic-
tors of systemic fungal infections. J In-
fect Dis 142: 503–509
12. Wenzel RP, Osterman CA, Donowitz
LG, Hoyt JW, Sand MA, Martone WJ,
Peacock JE Jr, Levine JI, Miller GB Jr
(1981) Identification of procedure-re-
lated nosocomial infections in high-risk
patients. Clin Infect Dis 3: 701–707
13. Maki DG (1989) Risk factors for noso-
comial infection in intensive care – de-
vices vs. nature and goals for the next
decade. Arch Intern Med 149: 30–35
14. Craven DE, Kunches LM, Lichtenberg
DA, Kollisch NR, Barry A, Heeren
TC, McCabe WR (1988) Nosocomial
infections and fatality in medical and
surgical intensive care unit patients.
Arch Intern Med 148: 1161–1168
15. Daschner F (1985) Nosocomial infec-
tions in intensive care units. Intensive
Care Med 11: 284–287
16. Chandrasekar PH, Kruse JA, Mathews
MF (1986) Nosocomial infection among
patients in different types of intensive
care units at a city hospital. Crit Care
Med 14: 508–510
17. Emmerson AM (1990) The epidemiol-
ogy of infections in intensive care units.
Intensive Care Med 16 [Suppl 3]:S197–
S200
18. Garvey BM, McCambley JA, Tuxen DV
(1989) Effects of gastric alkalization on
bacterial colonization in critically ill pa-
tients. Crit Care Med 17: 211–216
19. Füssle R, Biscoping J, Michaelis G, Zei-
ler D, Sziegoleit A (1991) Intensive care
patients: a high risk population for Can-
dida infections. Clin Intensive Care 2:
76–80
20. Füssle R, Biscoping, Zeiler D, Michae-
lis G, Sziegoleit A (1990) Mikro-
biologische Betreuung von beatmeten
Intensivpatienten: Möglichkeiten für
Diagnostik und Therapie von pulmo-
nalen Infekten. Anästh Intensivther
Notfallmed 25: 491–496
21. Pellinen TJ, Valtonen VV, Luomanmäki
K, Sivonen A, Virtanen KS (1983) The
microbial colonization due to medical
devices in intensive care patients with
special emphasis on C. albicans. Ann
Clin Res 15: 62–65
22. Burnie JP (1991) Developments in the
serological diagnosis of opportunistic
fungal infections. J Antimicrob Che-
mother 28 [Suppl A]:23–33
23. Müller J (1990) Laboratory diagnosis of
deep-seated candidosis. Mycoses 33
[Suppl 1]:7–13
24. Müller J, Kappe R, Kubitza D, Fessler
R, Scheidecker I (1987) The incidence
of deep-seated mycoses in Freiburg im
Breisgau. Mycoses 31[Suppl 1]:9–28
25. Knaus WA, Draper EA, Wagner DP,
Zimmermann JE (1985) APACHE II:
a severity of disease classification sys-
tem. Crit Care Med 13: 818–829
26. Goris RJA, te Boekhorst TPA, Nuy-
tinck JKS, Gimbrere JSF (1985) Multi-
ple-organ failure. Arch Surg 120: 1109–
1115
325
27. SAS Institute (1989) SAS/STAT user’s
guide, 4th ed, vol 2, Version 6
28. Kalbfleisch JD, Prentice RL (1980) The
statistical analysis of failure time data.
Wiley, New York
29. Stone HH, Kolb LD, Currie CA, Ge-
heber CE, Cuzzell JZ (1974) Candida
sepsis: pathogenesis and principles of
treatment. Ann Surg 179: 697–711
30. Pittet D, Monod M, Suter P, Frenk E,
Auckenthaler R (1994) Candida coloni-
zation and subsequent infections in crit-
ically ill patients. Ann Surg 220: 751–
758
31. Gaines JD, Remington JS (1972) Dis-
seminated candidiasis in the surgical pa-
tient. Surgery 72: 730–736
32. Bodey GP, Anaissei EJ (1989) Chronic
systemic candidiasis. Eur J Microbiol
Infect Dis 8: 855–857
33. Gemeinhardt H, Deicke P (1989)
Mykosen des Gefässsystems. In: Ge-
meinhardt H (ed) Endomykosen. Gus-
tav Fischer, Stuttgart pp 327–329
34. Bennett JE (1978) Diagnosis and man-
agement of candidiasis in the immuno-
suppressed host. Scand J Infect Dis
16[Suppl]:63–67
35. Klein JJ, Watanakunakorn C (1979)
Hospital-acquired fungemia: its natural
course and clinical significance. Am J
Med 67: 51–58
36. Bross J, Talbot GH, Maislin G, Hurwitz
S, Storm BL (1989) Risk factors for
nosocomial candidemia: a case control
study in adults without leukemia. Am J
Med 87: 614–620
37. Young RC, Bennett JE (1971) Invasive
aspergillosis: absence of of detectable
antibody response. Am J Respir Crit
Care Med 104: 710–716