Clostridium difficile Infection Is a Risk Factor for Bacteremia Due to
Vancomycin-Resistant Enterococci (VRE) in VRE-Colonized Patients with Acute
Mary-Claire Roghmann, Robert J. McCarter, Jr.,
Jeanine Brewrink, Alan S. Cross,
and J. Glenn Morris, Jr.
From the Division of Hospital Epidemiology, Department of Medicine,
University of Maryland School of Medicine and Veterans Affairs
Maryland Health Care System; and the Department of Epidemiology
and Preventive Medicine, the Greenebaum Cancer Center, and the
Division of Infectious Disease, Department of Medicine, University of
Maryland School of Medicine, Baltimore, Maryland
A cohort study was conducted in a cancer center to identify risk factors for bacteremia with
vancomycin-resistant enterococci (VRE) in neutropenic cancer patients colonized with VRE. There
were 10 patients with VRE bacteremia among 56 colonized with VRE, of whose charts 51 were
available for review. One hundred percent of patients with VRE bacteremia (10 of 10) vs. 56% of
patients without VRE bacteremia (23 of 41) had acute leukemia (P Å .01, Fisher’s exact test). Four
of the 10 patients with VRE bacteremia had a positive Clostridium difficile toxin assay within 6
days of their first positive VRE blood culture. Both C. difficile infection and antimicrobial (vancomy-
cin and ciprofloxacin) use during VRE colonization were significant risk factors for VRE bacteremia
in univariate analysis. When a Cox proportional hazards model was used to account for differences
in follow-up time, C. difficile infection was the only statistically significant risk factor (risk ratio,
8.2; P Å .007) for VRE bacteremia in VRE-colonized patients with acute leukemia.
Although the enterococcus is not a highly virulent organism
in healthy hosts, it can cause life-threatening infection in immu-
nosuppressed patients, including those with cancer. Neutro-
penic cancer patients are at high risk for bacteremia with organ-
isms that originate from their gastrointestinal tract. Other
investigators have shown that gastrointestinal colonization with
vancomycin-resistant enterococci (VRE) is a risk factor for
bacteremia with VRE in patients with cancer . During 1995,
15% of our neutropenic patients at the Greenebaum Cancer
Center (formerly the University of Maryland Cancer Center)
were colonized with VRE. Because of the high prevalence of
VRE colonization and the resistance of VRE to all bactericidal
antibiotics, we sought to identify potentially alterable risk fac-
tors for VRE bacteremia in cancer patients. We estimated the
incidence of VRE bacteremia and identified risk factors for
VRE bacteremia in neutropenic cancer patients colonized with
VRE in a cohort study.
Greenebaum Cancer Center from January 1992 through June
1996. The facility is a comprehensive cancer center at the
University of Maryland Hospital that provides both inpatient
and outpatient care for patients with acute leukemia, other
hematologic malignancies, and solid tumors. Neutropenic pa-
tients (neutrophil count, õ1 1 109/L) with temperatures of
ú100.4?F and/or signs of infection are admitted to the hospital
rapidly. Appropriate cultures and radiographic procedures are
performed, and broad-spectrum antibiotic therapy is immedi-
Empirical antibiotic regimens varied over the study period, but
lin. Acyclovir is the only antimicrobial agent used for chemo-
prophylaxis. All patients with acute leukemia have a long-term
tunneled central vascular catheter (usually a Hickman catheter)
placed.Patients admittedto thefacility haveweekly rectal surveil-
(õ1 1 109/L). The specimens were plated on media selective for
VRE (colistin nalidixic acid agar supplemented with defibrinated
sheep blood [5%], vancomycin [10 mg/mL], and amphotericin
[1 mg/mL]), in addition to other media .
The study cohort included all inpatients with a rectal surveil-
lance culture positive for VRE from January 1992 through
June 1996 (the study period). VRE-colonized patients were
identified prospectively through the microbiology laboratory
and followed to the endpoints of VRE bacteremia, death, or
the end of the study. VRE bacteremia was defined as one or
more blood culture positive for VRE during the study period.
The duration of follow-up was the number of days from initial
VRE colonization to onset of VRE bacteremia, death, or the
end of the study.
Materials and Methods
Design and patient identification. A prospective study of
neutropenic patients colonized with VRE was performed at the
Received 20 November 1996; revised 15 April 1997.
Grant support: in part by the Centers for Disease Control and Prevention.
Reprints or correspondence: Dr. Mary-Claire Roghmann, V.A. Maryland
Health Care System—Medical Care, 10 North Greene Street, Baltimore, Mary-
Clinical Infectious Diseases
? 1997 by The University of Chicago. All rights reserved.
/ 9c3f$$no6810-20-97 10:48:09cida UC: CID
by guest on July 13, 2011
1057 CID 1997;25 (November) VRE Bacteremia in VRE-Colonized Cancer Patients
Data collection and variable definitions.
clinical databases were reviewed for all study patients’ demo-
graphic information, underlying malignancy, and hospitaliza-
tions. Medical records were then reviewed for potential risk
factors at the time of VRE colonization and during follow-
up, including chemotherapy, neutropenia, Clostridium difficile
infection, and antibiotic use. Neutropenia in the study was
defined as a neutrophil count of õ0.5 1 109/L. C. difficile
infection was defined as either a positive cytotoxicity assay or
a positive EIA for toxin (Cambridge Bioscience, Cambridge,
MA) in conjunction with diarrhea (§3 nonsolid bowel move-
ments a day). Antibiotic use was recorded, including number
of days of therapy with intravenous and oral vancomycin, intra-
venous and oral metronidazole, imipenem, ceftazidime, pipera-
cillin, and intravenous ciprofloxacin, aswell as the total number
of days of antibiotic therapy during the 90 days before the first
positive surveillance culture and during the follow-up period.
All statistical analyses were performed
with SAS statistical software (SAS Institute, Carey, NC). Con-
tinuous variables were compared by Wilcoxon’s rank-sum test,
and categorical variables by Fisher’s exact test or Pearson’s
x2test. Cox’s proportional hazards models were used to ac-
count for variations in follow-up while estimating the associa-
tion of C. difficile infection, vancomycin use, and total antibi-
otic use with VRE bacteremia in VRE-colonized patients.
C. difficile infection, vancomycin use, and total antibiotic use
were modeled as time-dependent variables. It was assumed that
C. difficile infection increased the risk of bacteremia for a total
of 14 days, 7 days before and after the positive C. difficile toxin
significant; however, exact P values are reported.
Computerizedspecific antibiotic therapy. The six bacteremic patients with
neutropenia had a median of 3 days of positive blood cultures.
Two of these patients died while bacteremic with VRE; one had
polymicrobial bacteremia with Stenotrophomonas maltophilia.
The survivors’ neutrophil count recovered within 10 days.
Bacteremia with VRE in cancer patients colonized with VRE.
One hundred percent of patients with VRE bacteremia (10 of
10) vs. 56% of patients without VRE bacteremia (23 of 41)
had acute leukemia (P Å .01, Fisher’s exact test). The patients
with VRE bacteremia were older (mean age, 62 vs. 53 years;
P Å .11). At the time of initial VRE colonization, those patients
with VRE bacteremia were more likely to have had C. difficile
infection (20% [2 of 10] vs. 3% [1 of 41]; P Å .09) and had
had more days of therapy with parenteral vancomycin (mean,
10 vs. 8; P Å .14), oral vancomycin (mean, 6 vs. 3; P Å .09),
and ceftazidime (mean, 7 vs. 3; P Å .10) within the last 90
days than had patients without VRE bacteremia.
During the follow-up period, patients with VRE bacteremia
had a higher rate of C. difficile infection (mean, 38 vs. 1 per
1,000 follow-up days; P Å .01) and had a higher rate of anti-
biotic use, specifically of parenteral vancomycin (mean, 494
vs. 85 days per 1,000 follow-up days; P Å .02) and ciprofloxa-
cin (mean, 215 vs. 51 days per 1,000 follow-up days; P Å .04),
than patients without VRE bacteremia. In addition, patients
with VRE bacteremia had a higher rate of neutropenia (mean,
574 vs. 191 days per 1,000 follow-up days; P Å .06) than
patients without VRE bacteremia.
Bacteremia with VRE in leukemia patients colonized with
Because VRE bacteremia occurred exclusively in VRE-
colonized patients with acute leukemia, factors such as anti-
biotic use could be associated with acute leukemia, not neces-
sarily VRE bacteremia. For this reason, patients with acute
leukemia were compared with regardto the presence or absence
of subsequent VRE bacteremia (table 1). C. difficile infection
and antibiotic use (specifically of parenteral vancomycin) dur-
ing colonization remained statistically significant risk factors.
However, antibiotic use and C. difficile infection within the 90
days prior to colonization were not significantly different.
Because all VRE-colonized patients with VRE bacteremia
had acute leukemia, only patients with acute leukemia were
used to identify risk factors for infection, with use of Cox’s
proportional hazards models to account for differences in fol-
low-up times (table 2). C. difficile infection was associated
with a statistically significant eightfold increase in the risk of
VRE bacteremia in VRE-colonized patients with acute leuke-
mia. When adjusted for age, the risk was diminished but re-
mained statistically significant.
The time relationships between the positive C. difficile tests
and VRE bacteremia in the five patients with both are shown
in figure 1. Patient 5 does not contribute to the cited risk
ratio because the positive C. difficile test preceded the VRE
bacteremia by ú7 days. Although vancomycin use and total
antibiotic use increased the risk of VRE bacteremia, they were
not statistically significant risk factors and were unaffected by
Outcome of bacteremia with VRE in cancer patients.
were 10 cases of VRE bacteremia among 56 VRE-colonized
patients during the study period. Ninety-four percent of the
charts (51 of 56) were available for review. Sixty-five percent
of the VRE-colonized population (33 of 51) had acute leuke-
mia, 19% (10) had other hematologic malignancies, 10% (5)
had solid tumors, and 6% (3) had other nonmalignant condi-
During the study 10 patients developed VRE bacteremia
within 1–326 days (1, 5, 10, 14, 15, 18, 19, 85, 117, and 326
days) of their first positive VRE rectal culture. The incidence
rate of VRE bacteremia in VRE-colonized patients was 7.4 per
10,000 patient follow-up days (10 bacteremias during 13,521
follow-up days), or 63 per 10,000 neutropenic follow-up days
(10 bacteremias during 1,582 days of neutropenia). Although
initial VRE colonization was detected during a period of neu-
tropenia, four of the 10 patients (40%) were not neutropenic
at the time of VRE bacteremia.
All four of the patients without neutropenia had only a single
culture positive for VRE, and three of four defervesced without
/ 9c3f$$no6810-20-97 10:48:09 cidaUC: CID
by guest on July 13, 2011
1058 Roghmann et al.CID 1997;25 (November)
nized with vancomycin-resistant enterococci (VRE), according to the
presence or absence of subsequent VRE bacteremia.
Characteristics of patients with leukemia who were colo-
vancomycin-resistant enterococci (VRE) bacteremia in VRE-colo-
nized patients with acute leukemia.
Cox’s proportional hazards analysis of risk factors for
VariableRR 95% CI
P valueVRE colonization
(n Å 10)
(n Å 23)
C. difficile infection
C. difficile infection
Age in years, mean
Possible risk factors at time of initial
Chemotherapy in preceding
90 d, %
C. difficile infection in preceding
90 d, %
Antibiotics used in preceding
90 d, mean no. of days
Hospitalization in preceding 90 d,
mean no. of days
Possible risk factors during follow-
Days of neutropenia per 1,000
days of follow-up, mean
Episodes of C. difficile infection
per 1,000 days of follow-up,
Days of antibiotic therapy per
1,000 days of follow-up,
Follow-up,†mean no. of days
Hospitalization during follow-up,
mean no. of days
204 .21 2.60.5–13.5.25
adjustment for age. The dataset was not large enough to adjust
for more than one time-dependent covariable at a time.
Gastrointestinal colonization with VRE was identified as a
strong risk factor for bacteremia with VRE in cancer patients
in a previous case-control study . Our study sought to iden-
tify risk factors for VRE bacteremia in VRE-colonized cancer
patients with use of a cohort design. In our study and others
[1, 3, 4], the majority of the risk among VRE-colonized cancer
patients appeared to be for patients with acute leukemia, a
finding suggesting that risk factors for infection are also associ-
ated with acute leukemia or its treatment. Patients with VRE
bacteremia and acute leukemia were also more likely to have
had C. difficile infection after VRE colonization. In addition,
they had a higher rate of antibiotic use, especially of vancomy-
assay and onset of VRE bacteremia in VRE-colonized patients with
acute leukemia. Each block equals 1 hospital day. The asterisk (*)
indicates the day of positive C. difficile toxin assay. B indicates the
first positive blood culture for VRE. Patient 5 does not contribute to
the risk ratio because the positive C. difficile test preceded the onset
of VRE bacteremia by ú7 days.
Time relationship between positive C. difficile toxin
* Fisher’s exact or Wilcoxon’s rank-sum test.
†Days from VRE colonization until onset of VRE bacteremia, death, or end
/ 9c3f$$no6810-20-97 10:48:09cida UC: CID
by guest on July 13, 2011
1059CID 1997;25 (November) VRE Bacteremia in VRE-Colonized Cancer Patients
cin, during follow-up. However, when adjustments were made
for different follow-up times, only C. difficile infection was
significantly associated with VRE bacteremia.
The association between C. difficile infection and VRE bacter-
emia may be causal but could also serve as a marker of increased
with increased morbidity, such as chemotherapy-induced mucosi-
tis, may be responsible for the increased VRE bacteremia. Never-
theless, the association between C. difficile infection and VRE
bacteremia is biologically plausible. Investigators of case series
of patients with C. difficile infection and acute leukemia have
reported concurrent bacteremia involving organisms typically
found in gastrointestinal flora [5, 6].
Metronidazole, the current drug of choice for C. difficile
infection, has been associated with VRE bacteremia in other
studies of VRE bacteremia in cancer patients [1, 3]. Metronida-
zole use could be a surrogate marker for unreported C. difficile
infection in the other studies. Conversely, metronidazole could
promote VRE growth through inhibiting the anaerobic flora,
thus increasing the risk of VRE bacteremia.
In our study, the use of metronidazole (or imipenem) was
not significantly different between the patients with bacteremia
and those without bacteremia. In two of our five patients with
VRE bacteremia and C. difficile infection, the C. difficile infec-
tionwas diagnosedafteror within1 dayofthe VREbacteremia,
which does not support this particular hypothesis.
Although this study included all of the VRE-colonized pa-
tients at our center over 3 years, it was a small study. C. difficile
infection was associated with VRE bacteremia in only 40% of
the patients. Other risk factors such as chemotherapy-induced
mucositis probably contribute to the risk of VRE bacteremia
in cancer patients. However, unlike mucositis, C. difficile infec-
tion is a potentially alterable risk factor, and if the association
is causal, preventing C. difficile infection could reduce the
incidence of VRE bacteremia in patients with acute leukemia.
1. Edmond MB, Ober JF, Weinbaum DL, et al. Vancomycin-resistant Entero-
coccus faecium bacteremia: risk factors for infection. Clin Infect Dis
2. Morris Jr JG, Shay DK, Hebden JN, et al. Enterococci resistant to multiple
antimicrobial agents, including vancomycin: establishment of endemicity
in a university medical center. Ann Intern Med 1995;123:250–9.
3. Shay DK, Maloney SA, Montecalvo M, et al. Epidemiology and mortality
risk of vancomycin-resistant enterococcal bloodstream infections. J Infect
4. Montecalvo MA, Horowitz H, Gedris C, et al. Outbreak of vancomycin-,
ampicillin-, and aminoglycoside-resistant Enterococcus faecium bacter-
emia in an adult oncology unit. Antimicrob Agents Chemother 1994;38:
5. Milligan DW, Kelly JK. Pseudomembranous colitis in a leukaemia unit: a
report of five fatal cases. J Clin Pathol 1979;32:1237–43.
6. Rampling A, Warren R, Bevan P, Hoggarth C, Swirsky D, Hayhoe F.
Clostridium difficile in haematological malignancy. J Clin Pathol 1985;
/ 9c3f$$no68 10-20-97 10:48:09cida UC: CID
by guest on July 13, 2011