MAJOR ARTICLE

Epidemiology and Risk Factors for Gram-

Positive Coccal Infections in Neutropenia:
Toward a More Targeted Antibiotic Strategy
Catherine Cordonnier,1 Agnès Buzyn,3 Guy Leverger,4 Raoul Herbrecht,5 Mathilde Hunault,6 Roland Leclercq,6
and Sylvie Bastuji-Garin,2 for the Club de Réflexion sur les Infections en Onco-Hématologie
1
Hematology Department and 2Public Health Department, University Paris XII, Henri Mondor Hospital, Créteil, 3Hematology Department, Necker
Hospital, and 4Pediatrics Department, Trousseau Hospital, Assistance Publique–Hôpitaux de Paris, Paris, 5Hautepierre Hospital, Strasbourg,
6
Hematology Department, University Hospital, Angers, and 7Microbiology Department, University Hospital, Caen, France

Downloaded from http://cid.oxfordjournals.org/ by guest on April 7, 2016

The objective of this study was to evaluate the risk of acquiring gram-positive coccal infections in febrile
neutropenic patients and to develop risk indexes for gram-positive and streptococcal infections. This pro-
spective, multicenter study included 513 patients. The prevalence of gram-positive coccal infections was 21%
(14% were staphylococcal infections and 7.8% were streptococcal infections). The mortality rate during the
month after study enrollment was 5%. On multivariate analysis, the occurrence of gram-positive coccal in-
fections was significantly associated with receipt of high-dose cytarabine therapy, proton pump inhibitors,
and gut decontamination with colimycin without glycopeptides and presence of chills. Staphylococcal infection
was significantly associated with use of nonabsorbable colimycin, and streptococcal infection was associated
with diarrhea, use of nonabsorbable antifungals, receipt of high-dose cytarabine, and gut decontamination
with colimycin. The relative risks for streptococcal infection were 2.9, 13.2, and 20.7 in the presence of 1, 2,
and ⭓3 parameters, respectively. Risk factors for staphylococcal and streptococcal infections differ among
neutropenic patients. A simple scoring system for predicting streptococcal infection is proposed.

The increasing mortality rate for patients with febrile drome due to streptococci while avoiding an increase
neutropenia due to gram-positive coccal infections (es- in the prevalence of antibiotic-resistant organisms [4].
pecially streptococcal infections) has been attributed to b-Lactams do not all have adequate activity against
receipt of quinolone prophylaxis, receipt of high-dose gram-positive organisms: cefotaxime and ceftriaxone
cytarabine treatment, and the presence of central ve- are effective against b-lactam–susceptible strains, but
nous catheters [1–3]. The current challenge for phy- they have limited antipseudomonal activity, and cef-
sicians is to choose appropriate antibiotics to protect tazidime, which is commonly used to treat patients with
against septic shock and acute respiratory distress syn- Pseudomonas aeruginosa infection, may not be the best
choice to treat streptococcal bacteremia [5]. The routine
use of glycopeptides remains controversial [6, 7]. Co-
Received 9 April 2002; accepted 30 September 2002; electronically published
agulase-negative staphylococci are the most common
3 January 2003. cause of gram-positive bacteremia [8–11], but staphy-
Financial support: The Club de Réflexion sur les Infections en Onco-Hématologie lococcal bacteremia results in a low mortality rate,
is part of the Institut Maurice Rapin (Paris). This study was supported by a grant
from Laboratoires Roussel (Paris). whereas oral streptococci, which are responsible for up
Reprints or correspondence: Dr. Catherine Cordonnier, Hematology Dept., Henri to 39% of infections in neutropenic hosts [12], have
Mondor Hospital, 94000 Créteil, France (carlcord@club-internet.fr). been associated with mortality rates of 4%–22% [1, 13].
Clinical Infectious Diseases 2003; 36:149–58
 2003 by the Infectious Diseases Society of America. All rights reserved.
We conducted this prospective study to determine the
1058-4838/2003/3602-0003$15.00 prevalence of and risk factors for gram-positive coccal

Gram-Positive Infections and Neutropenia • CID 2003:36 (15 January) • 149

infection in febrile neutropenic patients and to establish a risk bility of streptococci to penicillin and methicillin-resistant staph-
index to aid physicians in their choice of first-line antibiotics ylococci were defined according to the recommendations of the
for the treatment of individual patients. Comité Français de l’Antibiogramme [15].
Classification of patients and definitions. For each febrile
neutropenic episode, patients were classified as having fever of
PATIENTS AND METHODS unknown origin, clinically documented infection, or micro-
biologically documented infection, according to the definitions
Patients. All consecutive patients who had a first episode of the Immunocompromised Host Society [14]. “High-dose
of fever while they had neutropenia were prospectively en- cytarabine” was defined as a dose of ⭓1.5 g/m2. Mucositis was
rolled at 36 French hematology centers during a 2-month
graded according to the World Health Organization score. Den-
period (table 1). Fever was defined as a temperature of
tal status at enrollment was graded as follows: 1, no visible
⭓38.3C once or ⭓38C twice within 8 h. Neutropenia was
teeth; 2, teeth without dental lesions or prior work; 3, presence
defined as a granulocyte count of !500 cells/mL or as a gran-
of crowns, bridges, or implants without lesions; and 4, cavities,
ulocyte count that was expected to decrease to !500 cells/mL
broken teeth, or exposed roots.
⭐48 h after study enrollment because of recent receipt of
Statistical analysis. Prevalences of gram-positive coccal,
chemotherapy. Patients could be included in the study only
streptococcal, and staphylococcal infections were computed by
once. Patients with solid tumors were excluded, except for
use of the total number of included patients as the denominator.
those who had neutropenia after they underwent autologous

Downloaded from http://cid.oxfordjournals.org/ by guest on April 7, 2016

Ninety-five percent CIs were also calculated.
stem cell transplantation.
To identify risk factors, the baseline characteristics of patients
Of the 532 enrolled patients, 14 did not meet the inclusion
who had gram-positive coccal infections (streptococcal and
criteria and 5 had missing data. Of the remaining 513 patients,
staphylococcal infections), including those who had concom-
291 were male and 222 were female. The mean age (SD) was
itant gram-negative infections, were compared with the baseline
40.8  21.7 years; 92 (18%) of the patients were children (age,
characteristics of patients who did not have gram-positive coc-
!15 years). The number of patients included per center was
cal infections. ORs with 95% CIs were calculated separately for
1–48 (median, 11).
each parameter by means of unconditional logistic regression
Neutropenia was most often due to chemotherapy without
models. Age (adulthood or childhood) was included in all of
stem cell transplantation. At the time of study enrollment, two-
the models. Variables with a P value of ⭐.15 on univariate analy-
thirds of the patients had granulocyte counts of !100 cells/mL,
81% had a central venous catheter in place, and 77% were sis were then entered into multivariate logistic models [16]. Mul-
already hospitalized. All patients had blood samples obtained tiple 2 ⫻ 2 analyses were used to assess interaction and confusion.
for ⭓2 cultures for aerobic and anaerobic organisms before When an interaction was found, a composite variable was built.
they began receiving antibiotics. In accordance with interna- The same analyses were done to identify risk factors for strep-
tional guidelines for first-line treatment [14], 91% of the pa- tococcal infection and for staphylococcal infection.
tients received b-lactams, 66% received aminoglycosides, and Independent risk factors determined on multivariate analysis
31% received glycopeptides, usually vancomycin. (P ⭐ .05) were used to create a gram-positive risk index (GPRI)
Data collection. Data that were collected at study enroll- and a streptococcal infection risk index (SRI) by adding the
ment included the following: hematological diagnosis and number of factors for each patient. The GPRI was then com-
status, cause of neutropenia, conditioning regimen for trans- pared between patients with and patients without gram-positive
plantation (when applicable), chemotherapy received during coccal infection, and the SRI was compared between patients
the month before enrollment, all anti-infective drugs taken ⭐7 with and patients without streptococcal infection. To determine
days before study enrollment (including drugs used for gut the risk of infection, the relative risk (estimated with the OR)
decontamination, antifungal agents, and growth factors), ad- for gram-positive coccal infection and the relative risk for strep-
ministration of antiulcer and antacid drugs, location during the tococcal infection were calculated according to the number of
week before study enrollment (laminar air-flow room, single factors present at study enrollment. A formal goodness-of-fit
room, ⭓2-bed room, or outpatient status), and site of the test was used to evaluate the calibration of both models [17],
intravenous catheter (central or peripheral). Clinical data in- and the area under the receiver operating characteristic (ROC)
cluded information regarding the presence of infection focus and curve was calculated to evaluate the discriminatory power of
organ failure. Day 1 was the day of study enrollment. On day the indexes [18]. To assess the predictive value of the GPRI
30 after enrollment, the patient’s clinical status was recorded or and SRI, we compared the scores of patients who had gram-
the cause of death was determined by the investigator at each negative infection with the scores of patients who did not have
center, and the findings were independently reviewed by 2 of the gram-negative infection; patients who had both gram-positive
principal investigators (C.C., A.B, or G.L.). Decreased suscepti- and gram-negative infections were excluded. For the SRI, the

150 • CID 2003:36 (15 January) • Cordonnier et al.

Table 1. Characteristics of 513 hospitalized patients with Table 1. (Continued.)
neutropenia and fever who were investigated for assessment
of risk factors for gram-positive coccal infection.

No. (%) No. (%)

Characteristic of patients Characteristic of patients
Hematologic disease Death 28 (5)
Acute leukemia 261 (51) Cause of death
Lymphoproliferative disease 159 (31) Directly related to initial infectious episode 7 (1)
Chronic myeloid leukemia 24 (5) Initial episode as contributing factor 2 (0.4)
Othera 69 (13) Subsequent infection that was independent from
Cause of neutropenia first infection 5 (1)
Chemotherapy 298 (58) Unrelated to initial or subsequent infection 13 (3)
Stem cell transplantation Unknown 1 (0.2)
Autologous 96 (19) a
For example, aplastic anemia.
b
Allogeneic 52 (10) Eleven patients experienced several types of clinical infection.
c
World Health Organization score of ⭓1.
Spontaneous neutropenia 67 (13) d
Cellulitis, folliculitis, or necrotic lesions.
Granulocyte count at study enrollment, cells/mL e
Four patients had both streptococcal and staphylococcal infections, and
!100 353 (69) 6 patients had both gram-positive coccal and gram-negative infections.

Downloaded from http://cid.oxfordjournals.org/ by guest on April 7, 2016

100–499 150 (29)
⭓500 but !500 within 48 h after study
enrollment 10 (2) sidered to be statistically significant. Data were analyzed with
Origin of fever use of BMDP software (University of California, Berkeley).
Unknown 305 (59)
Clinically documented infectionb 40 (8)
Bronchopulmonary focus 20 RESULTS
Mucositisc 10
Causes of fever, prevalence of infection, and mortality rate.
Cutaneous infectiond 11
The causes of fever, the prevalence of infection, and the mor-
Sinusitis 2
Infection with dental focus 8
tality rate are shown in table 1. Of the 168 microbiologically
Microbiologically documented infection 168 (33) documented infections, 147 (87.5%) were documented by
Gram-positive coccal infectione 108 (21) blood culture. The prevalence of streptococcal infection was
Streptococci and enterococci 40 (8) 7.8% (95% CI, 5.5%–10.1%). Of the 24 isolates of oral (viri-
Oral streptococci 24 dans) streptococci isolates (Streptococcus mitis and others), 13
Streptococcus pneumoniae 3 (54%) were susceptible to penicillin, 7 (29%) had decreased
Enterococci 6 susceptibility to penicillin, 1 (4%) was resistant to penicillin,
Other streptococci 7 and 3 had unknown resistance profiles. Four patients had both
Staphylococci 72 (14) streptococcal and staphylococcal infections, and 1 had both
Coagulase-negative staphylococci 52
streptococcal and gram-negative infections. The prevalence of
Staphylococcus aureus 14
staphylococcal infection was 14.0% (95% CI, 11.0%–17.0%).
Nonidentified staphylococci 6
Of the 53 isolates of coagulase-negative staphylococci, suscep-
Gram-negative infectionse 55 (11)
Escherichia coli 30
tibility to methicillin was known for 47; of these, 27 (57%)
Other enterobacteria 11 were resistant to methicillin. Of 14 Staphylococcus aureus iso-
Pseudomonas species 13 lates, methicillin susceptibility status was known for 12; of these
Salmonella species 1 12 isolates, 4 were methicillin resistant. Five patients had both
Other 5 (1) staphylococcal and gram-negative infections. The overall prev-
(continued) alence of gram-positive coccal infection was 21.0% (95% CI,
17.5%–24.6%). Among these patients, 6 also had gram-negative
infections. The overall prevalence of gram-negative infections
scores of patients with and patients without staphylococcal in- was 10.7% (95% CI, 8.0%–13.4%).
fection were also compared; patients who had both strepto- The overall mortality rate during the first 30 days of the
coccal and gram-negative or streptococcal and staphylococcal study was 5%. The mortality rates were 3% for patients with
infections were excluded. fever of unknown origin (9 of 305 patients), 15% for those
Data are presented as mean  SD or as proportions, as ap- with clinically documented infection (6 of 40 patients), and
propriate. All significance tests were 2-tailed. P ⭐ .05 was con- 8% for those with microbiologically documented infection (13

Gram-Positive Infections and Neutropenia • CID 2003:36 (15 January) • 151

 Table 2. Comparisons of patients with and patients without gram-positive coccal infection.

a,b
Percentage of patients with factor Multivariate analysis
Gram-positive No gram-positive P on
Factor (no. of patients coccal infection coccal infection univariate
with missing data) (n p 108) (n p 405) analysisa OR (95 % CI) P
Age !15 years 12 19 .06 1.4 (0.7–2.9) .36
Living condition before day 1c
Home 14 25 — 1
Conventional hospital room 42 41 — 1.5 (0.8–3.0) —
Laminar air-flow room (4) 44 34 .03 1.3 (0.7–2.5) .49
Treatment received before day 1
High-dose cytarabined 33 24 .04 1.7 (1.0–2.8) .05
Growth factors 13 20 .07 0.6 (0.3–1.1) .07
Proton pump inhibitors 20 13 .08 1.7 (1.1–3.8) .05
Antacids (9) 14 8 .08 1.4 (0.7–2.9) .32
Histamine2 antagonists 18 18 .99
Nonabsorbable antifungals 83 72 .04 0.9 (0.6–1.6) .83

Downloaded from http://cid.oxfordjournals.org/ by guest on April 7, 2016

e
Oral colimycin without glycopeptide 34 17 .001 3.0 (1.8–5.0) .0001
f
Trimethoprim-sulfamethoxazole 5 12 .08 0.3 (0.1–1.0) .04
Symptom at day 1
Chills (8) 34 24 .05 1.6 (1.0–2.7) .05
Mucositisg (4) 44 34 .05 1.2 (0.7–1.9) .60
Inflammatory catheter insertion site 7 3 .12 1.9 (0.7–5.3) .21
Diarrhea 21 14 .05 1.5 (0.8–2.6) .22

NOTE. Day 1, day of study enrollment.

a
All analyses were age adjusted.
b
Only factors with P ! .15 on univariate analysis were used for multivariate analysis.
c
Reference class was “home.”
d
More than 1.5 g/m2.
e
Reference class was all other categories—that is, neither colimycin nor glycopeptide, glycopeptide without colimycin, and colimycin
plus glycopeptide.
f
OR reflecting negative association.
g
World Health Organization score of ⭓1.

of 168 patients). Death was not directly related to the initial 9 (2%) were related to the initial episode alone (n p 7 ) or were
episode in any of the patients who had fever of unknown origin, associated with a contributory cause (n p 2 ). No relationship
but it was related to the initial episode in 3 (7.5%) of those was found between the number of positive blood culture results
who had clinically documented infection and in 6 (4%) of those and the risk of death.
who had microbiologically documented infections (2 patients Factors associated with gram-positive coccal infections.
with streptococcal infections and 1 patient each infected P. On univariate analysis, previous hospitalization in a laminar
aeruginosa, Escherichia coli, Klebsiella pneumoniae, and Candida air-flow room, treatment with high-dose cytarabine, oral non-
species). Of the 3 patients whose deaths were attributed to absorbable antifungals, or oral colimycin without glycopeptide,
streptococcal infection alone or were associated with another and chills, mucositis, and diarrhea at study enrollment were
contributing factor, 1 was infected with penicillin-susceptible significantly associated with gram-positive coccal infection (ta-
Streptococcus pneumoniae, 1 was infected with an oral Strep- ble 2). Quinolone use (5% among gram-positive coccal infec-
tococcus species, and 1 was infected with an untyped strepto- tions vs. 6% among others), severe neutropenia (71% vs. 69%),
coccal strain. These 3 patients had all been treated with first- underlying disease, dental status, presence of cutaneous lesions
line b-lactams that are active against streptococci (1 patient (7% vs. 8%), and bronchopulmonary focus (6% vs. 10%) did
each was treated with piperacillin, ticarcillin plus clavulanic not influence the occurrence of gram-positive coccal infection.
acid, and cefpirome), and 2 of these patients had also received On multivariate analysis, only administration of high-dose cy-
vancomycin. None of the patients with staphylococcal infec- tarabine, proton pump inhibitors, and colimycin without gly-
tions died as a result of the initial episode. Of the 14 deaths copeptides and the presence of chills were independent risk
that were considered to have been due to an infectious cause, factors for gram-positive coccal infection. Mucositis was no

152 • CID 2003:36 (15 January) • Cordonnier et al.

longer associated with gram-positive coccal infection because gram-positive and gram-negative infections, according to the
of its strong relationship with use of proton pump inhibitors GPRI. The risk for gram-positive infection increased 3.0, 3.9,
and colimycin use alone (P p .04 and P p .001, respectively). 7.0, and 17.5 times when the GPRI was 1, 2, 3, and 4, respec-
Trimethoprim-sulfamethoxazole (TMP-SMZ) use appeared to tively. For each additional point on the GPRI, the OR was 1.85
be a significant protective factor. (95% CI, 1.43–2.38; P ! 10⫺4) for gram-positive infection and
On univariate analysis, use of high-dose cytarabine, use of 1.14 (95% CI, 0.81–1.61; P p .46) for gram-negative infection.
colimycin without glycopeptides, and diarrhea at study enroll- The observed (20.5%) and expected (21.2%) numbers of
ment were significantly associated with streptococcal infection, gram-positive infections were similar; the high P value deter-
and use of colimycin without glycopeptides and inflammation mined with use of the goodness-of-fit test indicated good agree-
of the catheter site were significantly associated with staphylo- ment (calibration) (P 1 .50 ). The area under the ROC curve
coccal infection (table 3). On multivariate analysis, the indepen- was 69%  16%.
dent risk factors for streptococcal infection were use of high- SRI. Each of the 4 independent risk factors for strepto-
dose cytarabine, use of colimycin without glycopeptides, use of coccal infection was assigned a value of 1. The prevalence of
nonabsorbable antifungals, and diarrhea; for staphylococcal in- and relative risks for streptococcal, gram-negative, and staph-
fection, only administration of colimycin without glycopeptides ylococcal infections according to the SRI are shown in table 6.
was found to be significant (table 4). There was a trend for TMP- The risk for streptococcal infection increased 2.9, 13.2, and
SMZ use to protect from staphylococcal infection. 20.7 times when the SRI was 1, 2, and ⭓3 respectively. For
GPRI. Each of the 4 independent risk factors for gram- each additional point on the SRI, the OR was 2.44 (95% CI,

Downloaded from http://cid.oxfordjournals.org/ by guest on April 7, 2016

positive coccal infection was assigned a value of 1 so that the 1.69–3.52; P p 10⫺4) for streptococcal infection, 0.86 (95% CI,
GPRI ranged from 0 (no factor present) to 4 (all factors pre- 0.63–1.19; P p .37) for gram-negative infection, and 1.27 (95%
sent). Table 5 shows the prevalence of and the relative risks for CI, 0.96–1.69; P p .09) for staphylococcal infection. The ob-

Table 3. Comparisons of patients with streptococcal infection, patients with staphylococcal infection, and control
subjects, on univariate analyses.

a
Percentage of patients with factor P
Streptococcus- Staphylococcus- Streptococcus- Staphylococcus-
infected infected Control infected infected
patients patients subjects patients vs. patients vs.
Factor (n p 40) (n p 72) (n p 405) control subjects control subjects
Age !15 years 10 13 19 .12 .14
Prior stay in laminar
air-flow room 44 44 34 .20 .13
Therapy received
High-dose cytarabineb 42 28 24 .02 .44
Growth factor 14 12 20 .32 .08
Proton pump inhibitors 22 16 13 .16 .54
Antacids 17 13 8 .12 .18
Nonabsorbable antifungals 94 78 72 .06 .57
Colimycin without
glycopeptide 39 31 17 .001 .005
Trimethoprim-sulfamethoxazole 6 4 12 .47 .10
Symptom
Granulocyte count of
!100 cells/mL 83 62 69 .06 .25
Chills 33 34 24 .29 .12
Mucositisc 44 44 34 .22 .11
Inflammatory catheter
insertion site 3 9 3 .90 .04
Diarrhea 31 16 14 .01 .58

NOTE. Control subjects were persons who had neither streptococcal nor staphylococcal infection.
a
All analyses were age adjusted; only factors with P ! .15 on univariate analysis were used in multivariate analyses.
b
More than 11.5 g/m2.
c
World Health Organization score of ⭓1.

Gram-Positive Infections and Neutropenia • CID 2003:36 (15 January) • 153

 Table 4. Comparisons of patients with streptococcal infection, patients with
staphylococcal infection, and control subjects, on multivariate analysis.

Streptococcus- Staphylococcus-
infected patients infected patients
vs. control subjects vs. control subjects
Factor OR (95% CI)a P OR (95% CI)a P
Prior stay in a laminar
air-flow room 1.2 (0.6–2.5) .35 1.4 (0.8–2.4) .14
Therapy received
High-dose cytarabineb 2.2 (1.1–4.6) .04 — —
c
Growth factor — — 0.5 (0.2–1.1) .08
Antacids 1.6 (0.6–4.4) .40 1.4 (0.6–3.3) .20
Nonabsorbable antifungals 4.4 (1.0–19.1) .02 — —
Colimycin without glycopeptide 2.8 (1.3–6.0) .01 2.6 (1.4–4.7) .01
Trimethoprim-sulfamethoxazole — — 0.3 (0.1–1.1)c .06
Symptom
Granulocyte count of
!100 cells/mL 1.4 (0.5–3.6) .48 — —

Downloaded from http://cid.oxfordjournals.org/ by guest on April 7, 2016

Chills — — 1.6 (0.9–2.8) .11
Mucositisd — — 1.3 (0.7–2.4) .14
Inflammatory catheter
insertion site — — 2.7 (0.9–7.6) .08
Diarrhea 2.5 (1.1–5.5) .03 — —
a
Multivariate ORs were age adjusted; only factors with P ! .15 on univariate analysis were
used in multivariate analyses.
b
More than 1.5 g/m2.
c
OR reflecting negative association.
d
World Health Organization score of ⭓1.

served (7.8%) and expected (7.2%) numbers of streptococcal Bacteremia due to viridans streptococci was detected in 4.7%
infections (P 1 .40) were similar. The area under the ROC curve of our patients and in up to 31% of patients described in other
was 76%  15%. series [12]. Our overall mortality rate of 5% was consistent
with that noted in antibiotic trials that involve patients with
febrile neutropenia, and our mortality rate for patients infected
DISCUSSION
with coagulase-negative staphylococci, which approached 0%,
This prospective epidemiologic study of 513 febrile neutropenic was similar to that reported in most other series [8, 19, 22,
patients allowed us to establish predictive scores to optimize 23]. The mortality rate associated with a-hemolytic strepto-
first-line treatment for febrile neutropenic patients with gram- coccal bacteremia is 4%–18% [1, 2, 12, 13, 24–26], and, al-
positive coccal and streptococcal infections. The strengths of though low mortality rates have been reported [27], it is gen-
this study included prospective collection of data, unbiased erally recognized that streptococcal bacteremia in neutropenic
patient recruitment and independent review, classification of patients may progress rapidly to septic shock and acute res-
patients according to outcome measures, and accurate mea- piratory distress syndrome, which require appropriate man-
surement of risk factors. Our study population was represen- agement to avoid irreversible septic complications [2, 5, 28,
tative of patients with febrile neutropenia hospitalized in French 29]. Previous retrospective studies of patients with bacteremia
hematology centers, because 95% of the centers in France par- due to viridans streptococci have identified profound neutro-
ticipated in the study. Furthermore, 96% of the consecutive penia, use of high-dose cytarabine therapy, mucositis, and use
patients were included in the analysis, which was largely because of histamine2 antagonists, antacids, quinolones, and TMP-SMZ
of the convenient 2-month study period and 1-month follow- as risk factors [1, 2, 3, 5, 8, 12, 13]. However, these findings
up period. have never been prospectively validated.
In our study, the prevalence of gram-positive coccal infection Our prospective study revealed that receipt of high-dose cy-
(21%) was similar to that reported in prospective therapeutic tarabine, proton pump inhibitors, and colimycin without gly-
trials [19–21], and the prevalence of streptococcal infection copeptides and presence of chills were independent risk factors
(7.8%) was in the lower range compared with other studies. for gram-positive coccal infections. Independent risk factors

154 • CID 2003:36 (15 January) • Cordonnier et al.

 Table 5. Prevalences and relative risks of gram-positive coccal and
gram-negative infections, according to gram-positive risk index (GPRI).

Gram-positive Gram-negative
coccal infection infection
No. of Infection Infection
GPRI patients rate, % OR (95% CI) rate, % OR (95% CI)
a
0 195 10.3 1 8.3 1
b
1 206 25.2 3.0 (1.7–5.2) 9.4 1.1 (0.6–2.3)
2 98 30.6 3.9 (2.1–7.3) 14.3 1.8 (0.8–3.9)
3 11 36.4 7.0 (1.3–18.6) 0 —
4 3 66.7 17.5 (1.5–203.0) 0 —

NOTE. The GPRI represents the no. of parameters among 4 that emerged from
multivariate analysis comparing patients with and patients without gram-positive coccal
infection for which P ⭐ .05 (receipt of high-dose cytarabine [11.5 g/m2], proton pump
inhibitors, or colimycin without glycopeptide and presence of chills). A weight of 1 was
assigned to each independent parameter.
a
Two patients with both gram-positive coccal and gram-negative infections were
excluded from analysis that compared patients with and patients without gram-negative
infection.

Downloaded from http://cid.oxfordjournals.org/ by guest on April 7, 2016

b
Four patients with both gram-positive coccal and gram-negative infections were
excluded from analysis that compared patients with and patients without gram-negative
infection.

for streptococcal infections were use of high-dose cytarabine, due to the high incidence of mucositis (34%) in our control
colimycin without glycopeptides, and nonabsorbable anti- population. The low percentage of patients receiving quino-
fungals and the occurrence of diarrhea. Although a retrospec- lones in our series may have precluded us from showing the
tive study suggested that high-dose cytarabine therapy causes impact of these antibiotics on the occurrence of gram-positive
gut damage and favors the translocation of streptococci [13], coccal or streptococcal infection, as has been reported elsewhere
no significant association between receipt of high-dose cytar- [1, 2]. Although severe neutropenia has been significantly as-
abine and mucositis of any grade was observed in our series. sociated with infection due to viridans streptococci by means
On our multivariate analysis, oropharyngeal mucositis, which of comparison with control subjects who had other gram-
has previously been identified as a risk factor for streptococcal positive infections [2], we observed only a trend on univariate
infection [2, 3, 27], was not associated with gram-positive coc- analysis in our series. However, the high rate of severe neutro-
cal or streptococcal infection. This difference may have been penia (69%) in our control population—including gram-

Table 6. Prevalences and relative risks of streptococcal, gram-negative, and staphylococcal in-
fections, according to streptococcal infection risk index (SRI).

Streptococcal infection Gram-negative infection Staphylococcal infection

No. of Infection Infection Infection
SRI patients rate, % OR (95% CI) rate, % OR (95% CI) rate, % OR (95% CI)
a
0 86 1.2 1 8.1 1 4.7 1
1 212b 3.3 2.9 (0.3–24.1) 13.7 1.8 (0.8–4.3) 15.6 3.7 (1.3–10.9)
2 164a 13.4 13.2 (1.7–100.0) 9.1 1.1 (0.4–2.9) 14.7 3.5 (1.2–10.5)
c
3 45 20.0 21.3 (2.6–174.8) 6.6 0.8 (0.2–3.3) 11.6 2.7 (0.7–10.5)
4 6 16.7 17.0 (0.9–313.5) 0 — 33.3 10.0 (1.4–73.0)
≥3 51c 19.6 20.7 (2.6–168.4) 5.9 0.7 (0.2–2.9) 14.3 3.4 (0.9–12.2)

NOTE. The SRI represents the no. of parameters among 4 that emerged from multivariate analysis comparing patients
with and patients without streptococcal infection for which P ⭐ .05 (receipt of high-dose cytarabine [11.5 g/m2], colimycin
without glycopeptide, or nonabsorbable antifungals and presence of diarrhea). A weight of 1 was assigned to each
independent parameter.
a
One patient with both streptococcal and staphylococcal infections was excluded from analysis that compared patients
with and patients without staphylococcal infection.
b
One patient with both streptococcal and gram-negative infections was excluded from analysis that compared patients
with and patients without gram-negative infection.
c
Two patients with both streptococcal and staphylococcal infections was excluded from analysis that compared patients
with and patients without staphylococcal infection.

Gram-Positive Infections and Neutropenia • CID 2003:36 (15 January) • 155

negative bacteremia, which is usually associated with severe neutropenic patients, in line with international recommenda-
neutropenia [30]—compared with the 20% rate in the control tions, the presence of gram-negative infection may have limited
population described by Elting et al. [2] may have precluded impact on the choice of initial treatment. More accurate iden-
our ability to show any significant relationship between severe tification of specific risk factors for streptococci could facilitate
neutropenia and streptococcal infection. In our study, although evaluation of the benefit of glycopeptides or other anti–gram-
the presence of chills at the time of study enrollment was sig- positive drugs in first-line treatment compared with results of
nificantly associated with gram-positive coccal infection (table previous trials [22, 23, 36]. As determined with our GPRI, the
2), this was not the case when streptococcal infections and risk for gram-positive coccal infection increased 3-fold for pa-
staphylococcal infections were analyzed separately. The most tients with 1 risk factor at the onset of febrile neutropenia, and
likely explanation for this difference is the lower power of our it increased 17.5-fold for patients with all 4 risk factors. Sim-
series of patients with streptococcal infection (table 3) com- ilarly, the relative risk for streptococcal infection increased from
pared with that in the study of Elting et al. [2]. 2.9 to 20.7 according to the number of risk factors present.
Gram-positive coccal infections may have a gut origin. Such Because the GPRI was not relevant for gram-negative infection,
drugs as histamine2 antagonists, antacids, and sucralfate select and because the SRI was moderately relevant for staphylococcal
patients with gut lesions and may favor streptococcal bacter- infection, these 2 indexes should allow reasonably accurate pre-
emia [2] by increasing the gastric pH and so favor microbial diction of the occurrence of gram-positive coccal infections in
growth and translocation from the gut. In our experience, the general and of streptococcal infections in particular. We are
prospectively validating these 2 scores among a new cohort of

Downloaded from http://cid.oxfordjournals.org/ by guest on April 7, 2016

use of antiulcer drugs, especially proton pump inhibitors, was
an independent risk factor for gram-positive coccal infection neutropenic patients. Use of these indexes could improve the
but was not significant for streptococcal infection, possibly be- treatment of neutropenic patients and help investigators design
cause of the lower power of this series. antibacterial trials to specifically address the question of the
Gut decontamination with use of nonabsorbable antibiotics best strategy to manage gram-positive coccal infections and
(mainly colimycin without glycopeptides) was a risk factor for streptococcal infections in this population.
both streptococcal and staphylococcal infections in our series.
Because of the low number of enterococcal infections (n p 6),
INVESTIGATORS AND PARTICIPATING
this effect is not explained by the effect of colimycin on en-
CENTERS
terococci overgrowth in the gut, as illustrated by animal models
[31]. Although quinolone prophylaxis is seldom used in France, Investigators and participating centers (all of which are in
most patients receiving chemotherapy for acute leukemia and France) are as follows: A. Auvrignon, Hôpital Trousseau, Paris;
almost all allogeneic stem cell transplant recipients are given Bendahmane, Institut Gustave Roussy, Villejuif; C. Bergeron,
nonabsorbable antibiotics (usually colimycin and aminoglyco- Hôpital Sud, Rennes; C. Berthou, Centre Hospitalo-Universi-
sides) during the neutropenic phase. Consequently, approxi- taire (CHU) Augustin Morvan, Brest; M. Boasson, Centre Hos-
mately one-half of the patients in our study had received non- pitalier Regional Universitaire (CHRU), Angers; O. Bouscary,
absorbable antibiotics. Previous studies from countries in which Hôpital Cochin, Paris; P. Chastagner, Hôpital de Brabois, Van-
gut decontamination is not performed could not identify this doeuvre Les Nancy; L. Collet, Institut Paoli Calmettes, Mar-
risk factor for gram-positive coccal infections [1–3]. seille; P. Cony-Makhoul, Hôpital du Haut-Leveque, Pessac; J.
Similarly, use of nonabsorbable antifungals, namely polyenes, Beaune, Hôpital Henri Mondor, Créteil; Th. De Revel, Hôpital
was associated with an increased risk of streptococcal infection du Val de Grâce, Paris; M. Delain, CHRU Bretonneau, Tours;
in our study. Several studies have reported bacteremia to be M. Macro, CHU, Caen; B. Desablens, CHRU, Amiens; M. C.
associated with antifungal prophylaxis, and a recent large study Escande, Institut Curie, Paris; N. Fegueux, CHR Lapeyronie,
showed a relationship between use of both absorbable and non- Montpellier; B. Girier, Hôpital Saint-Louis, Paris; H. Guy, Hôp-
absorbable antifungals and bacteremia [30, 32–35]. Taken to- ital du Bocage, Dijon; D. Guyotat, Hôpital Nord, Saint-Etienne;
gether, these findings argue in favor of the role of oral polyenes Ph. Henon, Hôpital du Has, Mulhouse; R. Herbrecht and I.
in gram-positive coccal growth. In contrast with a previous Zix-Kieffer, Hôpital de Hautepierre, Strasbourg; M. Hunault,
report, which showed that TMP-SMZ use increased the risk of Hôtel Dieu, Paris; H. Laurichesse, Hôpital de Clermont-
streptococcal infection [2], we found that TMP-SMZ was as- Ferrand; O. Lortholary, Hôpital Avicennes, Bobigny; D. N’Daw,
sociated with a decreased incidence of gram-positive coccal Hôpital Paul Brousse, Villejuif; G. Michel, Hôpital de La Ti-
infection, but this was mainly the result of a decreased risk for mone, Marseille; N. Parquet, Hôpital Saint-Louis, Paris; A. M.
staphylococcal infection. Peny, Centre Francois Baclesse, Caen; Y. Perel, Hôpital Pelle-
Because b-lactams are widely used for empirical therapy for grin, Bordeaux; C. Rose, Hôpital Huriez, Lille; L. Sutton, Pitié

156 • CID 2003:36 (15 January) • Cordonnier et al.

Salpétriere, Paris; M. Tiab, Hôpital de l’Hôtel Dieu, Nantes; A. 16. Breslow NE, Day NA, eds. Statistical methods in cancer research. Vol
I. The analysis of case-control studies. International Agency for Re-
Buzyn-Lévy, Hôpital Necker, Paris; P. Vic, Hôpital Huriez, Lille. search on Cancer (IARC) scientific publication no. 32. Lyon, France:
IARC, 1980.
17. Lemeshow S, Hosmer DW Jr. A review of goodness of fit statistics for
Acknowledgements use in the development of logistic regression models. Am J Epidemiol
1982; 115:92–106.
We thank Jean-Pierre Ghanassia (Agence Medicom, Paris) 18. Hanley J, McNeil B. The meaning and use of the area under a receiver
and Martine Rozenbaum (Laboratoires Hoescht-Marion-Rous- operating characteristic ROC curve. Radiology 1982; 143:29–36.
sel, Paris), for help in design and management of the study, 19. Del Favero A, Menichetti F, Martino P, et al. A multicenter, double-
blind, placebo-controlled trial comparing piperacillin-tazobactam with
and Isabel Cunningham (Beth Israel Hospital, New York), for and without amikacin as empiric therapy for febrile neutropenia. Clin
help in finalization of the manuscript. Infect Dis 2001; 33:1295–301.
20. Cordonnier C, Pico JL, Gardembas M, et al. Cefepime/amikacin
versus ceftazidime/amikacin as empirical therapy for febrile epi-
References sodes in neutropenic patients: a comparative study. Clin Infect Dis
1997; 24:41–51.
1. Kern W, Kurrle E, Schmeiser T. Streptococcal bacteremia in adult pa- 21. Cometta A, Zinner S, De Bock R, et al. Piperacillin-tazobactam plus
tients with leukemia undergoing aggressive chemotherapy: a review of amikacin versus ceftazidime plus amikacin as empiric therapy for fever
55 cases. Infection 1990; 18:138–45. in granulocytopenic patients with cancer. Antimicrob Agents Che-
2. Elting LS, Bodey GP, Keefe BH. Septicemia and shock syndrome due mother 1995; 39:445–52.
to viridans streptococci: a case-control study of predisposing factors. 22. European Organization for Research and Treatment of Cancer
Clin Infect Dis 1992; 14:1201–7. (EORTC) International Antimicrobial Therapy Cooperative Group

Downloaded from http://cid.oxfordjournals.org/ by guest on April 7, 2016

3. Bochud PY, Calandra T, Francioli P. Bacteremia due to viridans and the National Cancer Institute of Canada-Clinical Trials Group.
streptococci in neutropenic patients: a review. Am J Med 1994; Vancomycin added to empirical combination antibiotic therapy for
97:256–64. fever in granulocytopenic cancer patients. J Infect Dis 1991; 163:
4. Jones R. Contemporary antimicrobial susceptibility patterns of bac- 951–8.
terial pathogens commonly associated with febrile patients with neu- 23. Ramphal R, Bolger M, Oblon D, et al. Vancomycin is not an essential
tropenia. Clin Infect Dis 1999; 29:495–502. component of the initial empiric treatment regimen for febrile neu-
5. Elting LS, Rubenstein EB, Rolston KVI, Bodey GP. Outcomes of bac- tropenic patients receiving ceftazidime: a randomized prospective
teremia in patients with cancer and neutropenia: observations from study. Antimicrob Agents Chemother 1992; 36:1062–7.
two decades of epidemiological and clinical trials. Clin Infect Dis 24. Cohen J, Donnely JP, Worsley AM, et al. Septicemia caused by viridans
1997; 25:247–59. streptococci in neutropenic patients with leukemia. Lancet 1983; 2:
6. Hugues WT, Armstrong D, Bodey GP, et al. Guidelines for the use of
1452–4.
antimicrobial agents in neutropenic patients with unexplained fever.
25. Henslee J, Bostrom B, Weisdorf D, et al. Streptococcal sepsis in bone
Clin Infect Dis 1997; 25:551–73.
marrow transplant patient. Lancet 1984; 1:393.
7. Feld R. Vancomycin as part of initial empirical antibiotic therapy for
26. Villablanca JG, Steiner M, Kersey J, et al. The clinical spectrum of
febrile neutropenia in patients with cancer: pros and cons. Clin Infect
infections with viridans streptococci in bone marrow transplant. Bone
Dis 1999; 29:503–7.
Marrow Transplant 1990; 5:387–93.
8. Engelhard D, Elishoov A, Strauss N, et al. Nosocomial coagulase-
27. Engelhard D, Elishoov H, Or R, et al. Cytosine arabinoside as a major
negative staphylococcal infections in bone marrow transplantation
risk factor for Streptococcus viridans septicemia following bone marrow
recipients with central vein catheter. Transplantation 1996; 61:
transplantation: a 5-year prospective study. Bone Marrow Transplant
430–4.
1995; 16:565–70.
9. Viscoli C, Castagnola E, Giacchino M, et al. Bloodstream infections in
28. Engels EA, Barza M. Efficacy of quinolone prophylaxis in neutropenic
children with cancer: a multicentre surveillance study of the Italian
cancer patients: a meta-analysis. J Clin Oncol 1998; 16:1179–87.
Association of Paediatric Haematology and Oncology. Supportive
29. Claessen DC, Burke JP, Ford CD, et al. Streptococcus mitis sepsis in
Therapy Group–Infectious Diseases Section. Eur J Cancer 1999; 35:
bone marrow transplant patients receiving oral antimicrobial prophy-
770–4.
laxis. Am J Med 1990; 89:441–6.
10. Madani T. Clinical infections and bloodstream isolates associated with
fever in patients undergoing chemotherapy for acute myeloid leukemia. 30. Viscoli C, Paesmens M, Sanz M, et al. Association between antifungal
Infection 2000; 28:367–73. prophylaxis and rate of documented bacteremia in febrile neutropenic
11. Ninin E, Milpied N, Moreau P, et al. Longitudinal study of bacterial, cancer patients. Clin Infect Dis 2001; 32:1532–7.
viral, and fungal infections in adult recipients of bone marrow trans- 31. Gosse C, Faure E, Ardouin P, Andremont A, Tancrede C. Experimental
plants. Clin Infect Dis 2001; 33:41–7. intestinal microflora in mice with a changeable population size of E.
12. Reed EC. Infectious complications during autotransplantation. He- coli. Prog Clin Biol Res 1985; 181:175–7.
matol Oncol Clin North Am 1993; 7:717–35. 32. Palmblad J, Lonnqvist B, Carlsson B, et al. Oral ketoconazole pro-
13. Bochud PY, Eggiman T, Calandra T, Van Melle G, Saghafi L, Francioli phylaxis for Candida infections during induction therapy for acute
P. Bacteremia due to viridans Streptococcus in neutropenic patients with leukemia in adults: more bacteremias. J Intern Med 1992; 231:363–70.
cancer: clinical spectrum and risk factors. Clin Infect Dis 1994; 18: 33. Schaffner A, Schaffner M. Effect of prophylactic fluconazole on the
25–31. frequency of fungal infections, amphotericin B use, and health care
14. Hughes WT, Armstrong DA, Bodey GP, et al. Guidelines for the use costs in patients undergoing intensive chemotherapy for hematologic
of antimicrobial agents in neutropenic patients with unexplained fever. neoplasia. J Infect Dis 1995; 172:1035–41.
J Infect Dis 1990; 161:381–96. 34. Kern W, Behre G, Rudolf T, et al. Failure of fluconazole prophylaxis
15. Soussy C, Cluzel R, Courvalin P. Definition and determination of in to reduce mortality or the requirement of systemic amphotericin B
vitro antibiotic susceptibility breakpoints for bacteria in France. Eur J therapy during treatment for refractory acute myeloid leukemia. Can-
Clin Microbiol Infect Dis 1994; 13:238–46. cer 1998; 83:291–301.

Gram-Positive Infections and Neutropenia • CID 2003:36 (15 January) • 157

35. Menichetti F, Del Favero A, Martino P, et al. Itraconazole oral solution 36. Martino P, Micozzi A, Gentile G, Raccah R, Girmenia C, Mandelli F.
as prophylaxis for fungal infections in neutropenic patients with he- Piperacillin plus amikacin vs. piperacillin plus amikacin plus teico-
matologic malignancies: a randomized, placebo-controlled, double- planin for empirical treatment of febrile episodes in neutropenic pa-
blind multicenter trial. Clin Infect Dis 1999; 28:250–5. tients receiving quinolone prophylaxis. Clin Infect Dis 1992; 15:290–4.

Downloaded from http://cid.oxfordjournals.org/ by guest on April 7, 2016

158 • CID 2003:36 (15 January) • Cordonnier et al.