From www.bloodjournal.org by guest on January 20, 2015. For personal use only.

CLINICAL TRIALS AND OBSERVATIONS

Microbiologically documented infections and infection-related mortality in

children with acute myeloid leukemia
Lillian Sung,1 Beverly J. Lange,2 Robert B. Gerbing,3 Todd A. Alonzo,3,4 and James Feusner5
1Division

of Haematology/Oncology, Hospital for Sick Children, Toronto, ON; 2Division of Oncology, Childrens Hospital of Philadelphia, PA; 3Childrens Oncology
Group, Arcadia, CA; 4Division of Biostatistics, University of Southern California, Los Angeles, CA; and 5Division of Oncology, Childrens Hospital of Oakland, CA

The primary objective was to describe the

prevalence and characteristics of microbiologically defined infections and
infection-related mortality (IRM) in 492
children with acute myeloid leukemia enrolled on CCG 2961. Secondary objectives were to determine the relationship
between demographic, disease-related,
and therapeutic variables, and infections
and IRM. Institutions documented infections prospectively. Age, ethnicity, body
mass index, leukemia karyotype, treatment, and institutional size were exam-

ined for association with infection outcomes. More than 60% of children
experienced such infections in each of 3
phases of chemotherapy. There were 58
infectious deaths; cumulative incidence
of IRM was 11% plus or minus 2%. Thirtyone percent of infectious deaths were
associated with Aspergillus, 25.9% with
Candida, and 15.5% with alpha hemolytic
streptococci. Age older than 16 years
(hazard ratio [HR], 3.32; 95% confidence
interval [CI], 1.87-5.89; P < .001), nonwhite ethnicity (HR, 1.85; 95% CI, 1.10-

3.09; P .02), and underweight status

(HR, 3.06; 95% CI, 1.51-6.22; P .002)
were associated with IRM, while size of
the treating institution was not. Thus,
age, ethnicity, and BMI were important
contributors to IRM. Fungi and Grampositive cocci were the most common
organisms associated with IRM and, in
particular, Aspergillus species was the
largest contributor to infectious deaths.
(Blood. 2007;110:3532-3539)
2007 by The American Society of Hematology

Introduction
Infections are an important cause of morbidity and mortality for
children and adults with acute myeloid leukemia (AML). These
patients are at particularly high risk of infection, likely related to
the intensity of their therapy resulting in repeated episodes of
prolonged and profound neutropenia. Infections not only contribute
to mortality but also prolong hospitalization, compromise chemotherapy delivery, affect quality of life, and increase health care
utilization. Furthermore, protracted empiric and therapeutic use of
broad-spectrum antibiotics may contribute to evolution of resistant
microbiologic flora.
Although it is known that invasive infections are common in
children with AML, little is known about predictors of infections in
this population. A recent report from the Childrens Cancer Group
(CCG) found that body mass index (BMI) was associated with
treatment-related mortality.1 More specifically, the authors found
that in children enrolled onto the phase 3 trial, CCG 2961,
overweight patients were less likely to survive (HR, 1.88; 95% CI,
1.25-2.83; P .002) and more likely to experience treatmentrelated mortality (HR, 3.49; 95% CI, 1.99-6.10; P .001) compared with middleweight patients. However, the analysis did not
focus on infection-related mortality.
Given this finding, we wished to specifically explore infections
and infection-related mortality in children with AML and to
determine how they are influenced by BMI and other potential
predictors of outcome, which included age, ethnicity, leukemia
cytogenetics, and size of the treating institution. Our primary
objective was to describe the prevalence and characteristics of
infections and infectious deaths in children with AML enrolled

onto CCG 2961. The secondary objectives were to determine the

relationship between demographic, disease-related, and therapeutic
variables, and infections and infection-related mortality during
AML chemotherapy in children.

Patients and methods

Patients
This study was approved by each participating institutions institutional
review board (Document S1, available on the Blood website; see the
Supplemental Materials link at the top of the online article). Informed
consent was obtained in accordance with the Declaration of Helsinki.
The patients included in this report were those enrolled onto CCG 2961.
Children between the ages of one month to 21 years with newly diagnosed
AML or myelodysplastic syndrome were eligible. Patients with acute
promyelocytic leukemia, Down syndrome, or AML as a second malignancy
were excluded. In addition, for this specific analysis, we also excluded
patients with myelodysplastic syndrome, AML patients with isolated
granulocytic sarcoma, and infectious complications during phase 3 allogeneic stem cell transplantation (SCT).
Therapy consisted of 4 phases: (1) induction (idarubicin, daunomycin,
cytarabine, thioguanine, etoposide, and dexamethasone [IdaDCTER]);
(2) consolidation (randomization between a second IdaDCTER versus
fludarabine, cytosine arabinoside, and idarubicin); (3) intensification (cytosine arabinoside and L-asparaginase or SCT for those patients with a
matched related donor); and (4) immune modulation (randomization to
interleukin-2 or none in patients without donors after phase 3). In phase 2,
IdaDCTER contained 5 mg/m2 per day of idarubicin for 4 days and 200
mg/m2 per day of cytarabine on days 0 to 3 and days 10 to 13 as continuous

Submitted May 24, 2007; accepted July 25, 2007. Prepublished online as
Blood First Edition paper, July 27, 2007; DOI 10.1182/blood-2007-05-091942.

The publication costs of this article were defrayed in part by page charge
payment. Therefore, and solely to indicate this fact, this article is hereby
marked advertisement in accordance with 18 USC section 1734.

The online version of this article contains a data supplement.

2007 by The American Society of Hematology

3532

BLOOD, 15 NOVEMBER 2007 VOLUME 110, NUMBER 10

From www.bloodjournal.org by guest on January 20, 2015. For personal use only.
BLOOD, 15 NOVEMBER 2007 VOLUME 110, NUMBER 10

infusions. The fludarabine-based regimen contained 12 mg/m2 per day of

idarubicin for 3 days and 2400 mg/m2 per day of cytarabine for 3 days as a
continuous infusion (B.J.L., F.O. Smith, J.F., D. Barnard, P. Dinndorf, S.
Feig, N. Heerema, C. Arndt, R. Arceci, N. Seibel, M. Weiman, K.
Dusenbery, K. Shannon, S. Luna-Fineman, R.B.G., and T.A.A., Outcomes
in CCG-2961, a Childrens Oncology Group phase 3 trial for untreated
pediatric acute myeloid leukemia (AML); a report from the Childrens
Oncology Group, manuscript submitted April 2007).
For this study, drug dosing was based on weight in kilograms up to age
3 years and by body surface area for children older than 3 years. Drug doses
were modified for hyperbilirubinemia only for this group of patients, and
there were no dose modifications for underweight or overweight patients.
There were uniform guidelines for the use of empiric antibiotic and
antifungal therapy as well as granulocyte colony-stimulating factor (GCSF). In patients with an absolute neutrophil count of .5 109/L (500/L)
or less and oral temperature higher than 38C twice in 12 hours or higher
than 38.5C once, empiric systemic antibiotics were mandated. Broadspectrum antibiotics such as ceftazidime or triple antibiotic combinations of
an aminoglycoside, extended spectrum penicillin, and an antibiotic with
activity against alpha hemolytic streptococci and staphylococci were
specified. In terms of empiric antifungal coverage, the protocol specified
that in the setting of persistence of fever after 7 days of broad-spectrum
antibiotic coverage or the emergence of a new fever in neutropenic patients
with negative blood cultures, intravenous amphotericin B at 0.5 mg/kg per
day was to be initiated. G-CSF was specified beginning 2 days after
completion of induction and consolidation until the absolute neutrophil
count was higher than 1.5 109/L (1500/L).
This trial was open to patient accrual on August 30, 1996. The study was
suspended in October 1999 because of concern regarding increased
treatment-related mortality compared with the previous CCG AML trial.
The study was reopened in May 2000 and met accrual target December 4,
2002. The analysis presented in this report is based on children that were
enrolled in the presuspension period because of differences in how
infections were reported in the 2 time periods.
Outcomes
Infections were prospectively collected by the institutional clinical research
associates using a standardized data collection form. Microbiologically
documented infections with the same organisms within the same phase of
therapy were counted as different infections if they occurred more than
7 days apart. An infectious episode was defined as a microbiologically or a
clinically documented infection. In the case of common contaminants such
as coagulase-negative staphylococci and Gram-positive bacilli, insufficient
clinical information was available to distinguish between likely contaminants versus invasive infections and in this study; positive cultures were
included as infectious episodes. Clinically documented infections such as
pneumonia or typhlitis without microbiologic documentation were documented, but are not included in this report apart from the description of
infectious episodes.
Potential predictors related to demographics, disease, and
treatment
Potential predictors of infection outcomes that were examined were age
(0- 2, 2-16 and 16 years), ethnicity (white versus nonwhite), BMI
(overweight, middleweight, and underweight BMI status), leukemia cytogenetics, time to an absolute neutrophil count of at least 1.0 109/L
(1000/L; more than and fewer than the median number of days) and
institution size ( or median of 6 patients per year annual accrual).
Body mass index at diagnosis was calculated as weight in kilograms
divided by the square of the height in meters.2 For patients older than
2 years, underweight was defined as BMI of 10th percentile or lower;
overweight, as BMI of the 95th percentile or higher; and middleweight, as
BMI higher than the 10th to lower than the 95th percentile. For patients ages
1 to 2 years, 95th percentile or higher and 10th percentile or lower of weight
for length were used to define overweight and underweight, respectively.2
Those younger than 1 year were not included in the analysis by BMI

INFECTIONS IN PEDIATRIC AML

3533

because of lack of consensus definitions for overweight and underweight in

this age group.
Statistical analysis
Data were analyzed through to June 9, 2006. Observed differences in
proportions were tested using the chi-squared test and Fisher exact test
when data were sparse. To account for each patient having different lengths
of time at risk related to relapse, death, or SCT, the number of infections
was expressed as the rate of infections per day. The time period at risk was
during on-study chemotherapy administration and did not include time
following removal from study for any reason, relapse or SCT. The number
of infections was modeled using Poisson regression and treatment effects
were expressed as the incidence rate ratio (IRR). The IRR expresses the
increase in risk of an outcome for a one-unit change in the covariate and can
be considered analogous to a relative risk.
Infection-related mortality was analyzed using a Cox proportional
hazards model where the primary event was death attributed to infection.3
Patients who relapsed, had a noninfection-related death, completed protocol
therapy, or underwent protocol SCT were censored at the time of the event.
Patients who withdrew from protocol therapy before completion were
censored 30 days from the time they were taken off study. The 5-year
estimate for infection-related mortality was calculated by cumulative
incidence where patients who died due to infection were the primary events.
Patients who either relapsed or died not due to infection were competing
events, and patients who underwent protocol SCT were censored. For all
analyses, patients lost to follow-up were censored at their date of last known
contact or at a cutoff 6 months prior to June 9, 2006, to compensate for the
tendency of deaths and relapses being reported sooner than ongoing
follow-up. All statistical analysis was performed using the SAS statistical
program (SAS-PC, Version 9.1; SAS Institute, Cary, NC). All tests of
significance were 2-sided and statistical significance was defined as P value
less than .05.

Results
A total of 492 children were included in this analysis. Table 1
illustrates the demographics of these children. The median age was
9.6 years (range, 0.005-21.0 years). Of the 492 children, 48 (10%)
were overweight and 150 (30%) were underweight. Following
phase 1, 389 (80%) children were in complete remission, 35 (7%)
were in partial remission, and 27 (6%) had persistent disease. Six
withdrew from the study, and 2 were not evaluable. During phase 1,
33 (7%) died.
The most common sites of microbiologically documented
infections were blood, pulmonary, and gastrointestinal tract (Table
2). The majority of subjects had at least one microbiologically
documented infection during each phase of therapy (Table 3).
Infections with Gram-positive cocci predominated as the etiology
of microbiologically documented infections. More specifically,
coagulase-negative Staphylococcus and alpha hemolytic Streptococcus were the most common cause of at least one infection during
each phase of therapy, with 18%, 17%, and 19% of patients having
at least one positive culture with coagulase-negative Staphylococcus, and 10%, 27%, and 20% of patients having at least one
positive culture with alpha hemolytic Streptococcus, in phases 1, 2,
and 3, respectively. Pseudomonas species, Escherichia coli, and
Klebsiella species were the most common Gram-negative organisms. Fungi also were a prominent cause of such infections,
occurring in 18%, 21%, and 14% of subjects in phases 1, 2, and 3,
respectively. Four percent to 10% of patients had at least one
infection with Candida species. However in phase 2 almost twice
as many patients had at least one Aspergillus species compared
with Candida species infections.

From www.bloodjournal.org by guest on January 20, 2015. For personal use only.
3534

BLOOD, 15 NOVEMBER 2007 VOLUME 110, NUMBER 10

SUNG et al

Table 1. Demographic variables and treatment response of 492

children with de novo acute myeloid leukemia enrolled in CCG 2961
Characteristic

No.

Sex
Male

263

53

Female

229

47

0 to younger than 2

107

22

2 to 16

318

65

67

14

White

316

64

Black

42

Hispanic

99

20

Other

33

NA

Underweight, 10th percentile or lower

150

30

Normal weight, higher than 10th and lower

294

60

48

10

Age, y

Older than 16
Race

Unknown
BMI percentile at diagnosis

than 95th percentile

Overweight, 95th percentile or higher
FAB classification
M0

29

M1

83

17

M2

136

28

M4

122

25

M5

79

16

M6

10

M7

24

AML NOS
Cytogenetics
Favorable: t(8;21) or abnormal 16
Intermediate: not favorable or unfavorable
Unfavorable: 5/7 abnormalities
Unknown

78

26

209

70

13

192

NA

N 492.
BMI indicates body mass index; FAB, French-American-British; NOS, not
otherwise specified; and NA, not applicable.

To examine whether alpha hemolytic streptococci were associated with severe infection, we compared patients with and without
at least one episode of alpha hemolytic streptococcal bacteremia
with respect to reporting of any grade 3 or 4 toxicity and,
specifically, grade 3 or 4 pulmonary, renal, or cardiac toxicity.
Those with alpha hemolytic streptococci were more likely to have
any reported severe toxicity (80% versus 63% [P .018] in phase

1 and 69% versus 49% [P .018] in phase 3). In addition, those

with alpha hemolytic streptococcal bacteremia were more likely to
have grade 3 or 4 clinical pulmonary toxicity (dyspnea at rest or
with normal activity) (10% versus 3% [P .046] for phase 1 and
6% versus 2% [P .019] for phase 2) and grade 3 or 4 functional
pulmonary toxicity (required oxygen or assisted ventilation) (27%
versus 14% [P .016] for phase 1 and 18% versus 10% [P .048]
for phase 2).
Recurrence of infection was common between phases 1 and 2
but less so for all 3 phases. For example, in the 51 children with
alpha hemolytic streptococci isolated in phase 1 (23 of whom
completed all 3 phases), 14 also had a positive culture for the same
organism in phase 2, while only 1 child had a positive culture in all
3 phases. Similarly, for the 7 children with respiratory syncytial
virus in phase 1 (3 of whom completed all 3 phases), 3 also had
a positive culture for this virus in phase 2, while only 1 child had
a positive culture in all 3 phases.
To account for the possibility of subjects having more than one
infection and differing periods at risk, Table 4 illustrates the results
of the Poisson regression. Table 4 illustrates that the rates of
microbiologically documented bacterial infections per day in each
of the 3 phases of therapy were similar. Table 4 also illustrates that
BMI impacted the rate of bacterial infections only during phase 2,
with a significantly higher rate of infections in overweight children.
Conversely, age, ethnicity, duration of neutropenia, and institution
size did not significantly impact the rate of bacterial infections. In
terms of institution size, we chose to compare institutions with
more than 6 versus 6 or fewer accruals per year, as that was the
median accrual rate for all institutions participating in the trial. In
addition, we also examined cutoffs of 2, 3, 4, and 5 patients per
year; none of them was significantly associated with the rate of
bacterial infections (data not shown). In phase 2, BMI also was
predictive of the number of all microbiologically documented
infections (P .007), the number of infectious episodes (P .003),
and the number of bacteremias (P .038), with overweight
children having more infections compared with normal weight
children (data not shown).
The cumulative incidence of infection-related mortality was
11% plus or minus 2% during the period at risk that excluded SCT,
relative to the cumulative incidence of death from any cause as a
first event of 18% plus or minus 4%. Table 5 illustrates the
microorganisms associated with infection-related mortality. When
both polymicrobial and single-pathogen infection-related mortality

Table 2. Sites of microbiologically documented infections

Phase 2
Ida-DCTER
regimen, N 342

Phase 1, N 635
Site

No.

Blood

No.

Fludarabine
regimen, N 313

Np.

Phase 3, N 318
No.

357

56

191

56

190

61

219

69

Pulmonary

67

11

48

14

54

17

35

11

Gastrointestinal tract

90

14

30

23

11

Skin

33

20

11

15

32

20

Central nervous system

Urinary tract
Central venous catheter

0.3

0.3

13

3
2

Liver

10

Sinus, upper respiratory tract

17

14

10

Other

21

12

IdaDCTER regimen indicates idarubicin, daunomycin, cytarabine, thioguanine, etoposide, and dexamethasone; and fludarabine-based regimen, fludarabine, cytosine
arabinoside, and idarubicin.

From www.bloodjournal.org by guest on January 20, 2015. For personal use only.
BLOOD, 15 NOVEMBER 2007 VOLUME 110, NUMBER 10

INFECTIONS IN PEDIATRIC AML

3535

Table 3. Percentage of patients with at least one microorganism during each phase of therapy
Phase 2
Phase 1, N 492

Total, N 407

IdaDCTER, N 205

Fludarabine, N 202

Phase 3, N 248

Microorganism characteristic

No.

No.

No.

No.

No.

At least 1 organism

297

60

300

74

157

77

143

71

176

71

1 organism

158

32

150

37

79

39

71

35

98

40

2 organisms

80

16

73

18

41

20

32

16

46

19

3 organisms

39

53

13

28

14

25

12

24

10

4 organisms

11

16

13

191

39

205

50

100

49

105

52

111

45

91

18

70

17

33

16

37

18

47

19

5 or more organisms
Gram-positive cocci
Coagulase-negative staphylococci
Staphylococcus aureus

14

Alpha hemolytic streptococci

51

10

109

27

45

22

64

32

49

20

Other streptococci

17

Enterococcus species

27

26

18

Other Gram-positive cocci

28

27

16

11

18

Gram-negative bacilli and cocci

87

18

106

26

62

30

44

22

69

28

Pseudomonas species

21

29

14

15

11

Escherichiae coli

18

15

10

14

Klebsiella species

15

44

11

31

15

13

19

Other enterobacteriaceae

13

16

15

Acinetobacter species

Serratia species

16

14

Other Gram-negative bacilli

Gram-negative anaerobe

Gram-negative cocci, coccobacilli

2
14

Fungus

88

18

86

21

48

23

38

19

34

Candida albicans

21

Other Candida species

30

20

10

10

Aspergillus species

20

42

10

23

11

19

Fusarium species

Mucor

22

23

15

14

Other fungus
Viruses

36

31

15

16

17

Varicella-zoster virus

Herpes simplex virus

24

18

10

Cytomegalovirus

Respiratory syncytial virus

Influenza virus

Other

Sum of columns and rows do not match, as some patients had multiple organisms.
IdaDCTER regimen indicates idarubicin, daunomycin, cytarabine, thioguanine, etoposide, and dexamethasone; and fludarabine-based regimen, fludarabine, cytosine
arabinoside, and idarubicin.

were considered, coagulase-negative staphylococci, alpha hemolytic streptococci, Candida species, and Aspergillus species were
the most common etiologic agents related to infectious deaths. In
particular, in phase 2, Aspergillus species accounted for 46% of
infection-related deaths on protocol therapy. The following lists the
proportion of infections with specific pathogens that resulted in
fatal infections: coagulase-negative staphylococci 14 (6.7%) of
208, alpha hemolytic streptococci 9 (4.3%) of 209, Candida
species 15 (17.6%) of 85, and Aspergillus species 18 (26.1%) of 69.
Pseudomonas species contributed to 6 deaths, Klebsiella species to
7, and other Gram-negative bacilli to 15. Of the 58 infection-related
deaths, the proportion attributed to specific organisms were as
follows: coagulase-negative staphylococci 14 (24.1%) of 58, alpha
hemolytic streptococci 9 (15.5%) of 58, Candida species 15
(25.9%) of 58, and Aspergillus species 18 (31.0%) of 58. In other
words, Aspergillus species was the largest contributor to infectionrelated mortality in these children. It is important to note that this
study could not distinguish which coagulase-negative staphylococcal infections were contaminants. Thus it is likely that some of
these infections were not the cause of death. In addition, it is

important to emphasize that many infectious deaths were polymicrobial, and one infection or treatment of one infection may have
led to subsequent lethal infections.
If infection-related mortality is classified by association with a
single or multiple pathogens, then 22 (37.9%) were associated with
only one microorganism, most often Aspergillus species (7 deaths).
The next most common causes of single-pathogenassociated
infection-related mortality were Escherichia coli and Pseudomonas aeruginosa in 3 children for both. In other words, polymicrobial infection-related mortality was more common than singlepathogen infection-related mortality. In particular, only 2 of
14 fatal infections with coagulase-negative Staphylococcus had this
organisms as the sole isolate.
Table 6 illustrates potential predictors of infectious mortality
in univariate analysis and demonstrates that BMI was significantly associated with infection-related mortality, with underweight children having a 3-fold increase in mortality and
overweight children having a 1.5-fold increase in mortality. In
addition, nonwhite children were more likely to die of infection
with a hazard ratio of 1.85. In addition, children older than

From www.bloodjournal.org by guest on January 20, 2015. For personal use only.
3536

BLOOD, 15 NOVEMBER 2007 VOLUME 110, NUMBER 10

SUNG et al

Table 4. Univariate regression predicting the risk of microbiologically documented bacterial infection (any site) by phase of treatment
Phase 1
Characteristic
Rate of infections per day

IRR (95% CI)

Phase 2
P

0.0220

IRR (95% CI)

Phase 3
P

0.0226

IRR (95% CI)

0.0222

Age, y
0 to younger than 2
2 to 16
Older than 16

1.22 (0.97-1.55)

.087

1.15 (0.90-1.46)

.255

1.22 (0.89-1.68)

Ref

NA

Ref

NA

Ref

.215
NA

1.09 (0.82-1.46)

.563

1.10 (0.83-1.45)

.512

1.49(1.01-2.20)

.045

Ethnicity
White
Not white

Ref

NA

Ref

NA

Ref

NA

0.99 (0.81-1.22)

.954

1.03 (0.84-1.25)

.785

1.01(0.78-1.32)

.928

0.95 (0.65-1.38)

.773

1.05 (0.72-1.53)

.807

0.89 (0.44-1.81)

.751

Ref

NA

Ref

NA

Ref

NA

1.09 (0.79-1.49)

.603

1.54 (1.19-2.00)

.001

0.701(0.45-1.08)

.108
.309

BMI percentile at diagnosis*

Underweight
Normal weight
Overweight
Cytogenetics
0.81 (0.59-1.11)

.198

0.85 (0.65-1.10)

.214

0.84 (0.59-1.18)

Intermediate

Favorable

Ref

NA

Ref

NA

Ref

NA

Unfavorable

0.88 (0.57-1.36)

.558

0.89 (0.46-1.73)

.728

2.07 (0.97-4.42)

.062

Unknown

0.96 (0.77-1.20)

.738

0.86 (0.69-1.06)

.149

0.64 (0.47-0.86)

.003

1.11 (0.88-1.39)

.380

1.18 (0.95-1.46)

.137

0.87 (0.66-1.14)

.314

Ref

NA

Ref

NA

Ref

NA

1.04 (0.82-1.34)

.729

1.01 (0.80-1.29)

.921

1.07 (0.77-1.48)

.694

Ref

NA

Ref

NA

Ref

NA

Time to ANC higher than 1000/L

Median no. days or fewer
More than the median no. of days
Institution size
6 or fewer
More than 6

IRR indicates incidence rate ratio; BMI, body mass index; ANC, absolute neutrophil count; Ref, reference group; and NA, not applicable.
*Underweight ( 10th percentile); normal weight (between 10th and 95th percentile); and overweight ( 95 percentile).
Institution size categorized by median number of enrollments.

16 years were more likely to die of infections compared with

children 2 to 16 years of age. Neither cytogenetics nor
institution size was associated with infection-related mortality.
We also examined whether the cumulative risk of death as a first
event or infection-related mortality was significantly different
between those enrolled prior to and after March 16, 1998
(n 290 in each group). These were not different, with 5-year
cumulative incidences of death as a first event of 17.0% plus or
minus 5% in the earlier group and 18.0% plus or minus 5.0% in
the latter group (P .639 by Gray test) and infection-related
mortality of 11.0% plus or minus 3.0% in both groups (P .804
by Gray test).
A multiple regression analysis then was conducted to examine
whether the effects of BMI, ethnicity, and age were independently
associated with infection-related mortality. Because only children
older than one year were included in the BMI analysis, the multiple
regression model included age as a dichotomous variable by older
than 16 years of age versus younger children. In this model, age
older than 16 years (HR, 3.46; 95% CI, 1.94 to 6.17; P .001),
nonwhite ethnicity (HR, 2.05; 95% CI, 1.17 to 3.58; P .012), and
underweight status (HR, 3.59; 95% CI, 1.74 to 7.39; P .001) all
remained independent predictors of infection-related mortality.
There was one randomized intervention relevant to our analysis,
which was the randomization during phase 2 between IdaDCTER
versus the fludarabine-based regimen. During phase 2, there were
24 treatment-related deaths and there was significantly more
treatment-related mortality associated with the fludarabine-based
regimen, with 7 deaths with IdaDCTER and 17 deaths with
fludarabine (P .01). Infection-related mortality also was higher
with the fludarabine regimen (HR, 1.92; 95% CI, 0.97 to 3.81;
P .063). There was no difference in the rate of bacterial
infections between the 2 randomized regimens, with an IRR of 1.09
(95% CI, 0.90 to 1.32; P .357). However, the rate of alpha
hemolytic streptococcal bacteremia was higher in the fludarabine-

based regimen compared with IdaDCTER (IRR, 1.84; 95% CI,

1.25 to 2.72; P .002), and those allocated to the fludarabinebased regimen had a higher percentage of patients with at least one
alpha hemolytic streptococcal infection compared with IdaDCTER
(22.4% for IdaDCTER versus 33.2% for the fludarabine regimen,
P .016). The duration of neutropenia did not explain these
differences, as the median time to absolute neutrophil count
recovery was shorter for the fludarabine-based regimen (49 days
for IdaDCTER versus 40 days for the fludarabine-based regimen;
P .001). Similarly, prevalence of mucositis did not explain these
differences, and occurred in 32.2% with IdaDCTER versus 25.7%
with the fludarabine regimen (P .185).

Discussion
We found that most children with AML experience invasive
infections and that Gram-positive cocci are the most common
microbiologically documented etiology of these infections. The
cumulative incidence of infection-related mortality was 11% plus
or minus 2% during chemotherapy administration excluding SCT.
Polymicrobial infections are commonly associated with infectionrelated mortality in AML patients as a group, and in individual
patients, fungi are implicated in more than half of infectious deaths.
More specifically, Aspergillus species was the largest contributor to
infection-related mortality. Children who were underweight, nonwhite, and older than 16 years were at higher risk of deaths from
infection.
Our report is important for several reasons. First, to our
knowledge, we have the only large pediatric AML study in which
infections were collected prospectively and in which a specific
emphasis was placed on collecting infection data. This distinction
is important as it likely increased the quality and compliance of
infection reporting, thus making our study a more accurate

From www.bloodjournal.org by guest on January 20, 2015. For personal use only.
BLOOD, 15 NOVEMBER 2007 VOLUME 110, NUMBER 10

INFECTIONS IN PEDIATRIC AML

3537

Table 5. Microorganisms responsible for infection-related mortality on protocol therapy

Phase 2
Phase 1

Total

IdaDCTER

Fludarabine

Phase 3

Microorganism characteristic

Total no. of infection-related deaths

24

100

24

100

29

17

71

10

100

Gram-positive cocci

11

46

15

63

57

11

65

25

Coagulase-negative staphylococci

21

33

29

35

Staphylococcus aureus

Alpha hemolytic streptococci

17

17

14

18

Other streptococci

Enterococcus species

13

13

29

Other Gram-positive cocci

17

17

29

12

33

11

46

71

35

25

13

18

13
13

Gram-negative bacilli and cocci

Pseudomonas species
Escherichiae coli

29

Klebsiella species

17

43

Other enterobacteriaceae

17

Acinetobacter species

Serratia species

Other Gram-negative bacilli

13

Gram-negative anaerobe

Gram-negative cocci, coccobacilli

Fungus

11

46

15

63

57

11

65

44

Candida albicans

13

Other Candida species

25

17

24

Aspergillus species

13

11

46

57

41

25

Fusarium species

Mucor

Other fungus

Viruses

14

13

Varicella-zoster virus

Herpes simplex virus

Cytomegalovirus

Respiratory syncytial virus

Influenza virus

14

Other

Sum of columns and rows do not match, as some patients had multiple organisms as the cause of death.
IdaDCTER regimen indicates idarubicin, daunomycin, cytarabine, thioguanine, etoposide, and dexamethasone; and fludarabine-based regimen, fludarabine, cytosine
arabinoside, and idarubicin.

Table 6. Potential predictors of infection-related mortality

Characteristic

Hazard ratio (95% CI)

0.81 (0.37-1.76)

.597

Age, y
0 younger than 2
2 to 16
Older than 16

Ref

NA

3.32 (1.87-5.89)

.001

Ethnicity
White
Not white

Ref

NA

1.85 (1.10-3.09)

.020

3.06 (1.51-6.22)

.002

BMI percentile at diagnosis*

Underweight
Normal weight
Overweight

Ref

NA

1.58 (0.76-3.30)

.224

Cytogenetics
Favorable

Ref

NA

Intermediate

0.95 (0.43-2.08)

.889

Unfavorable

2.32 (0.70-7.64)

.168

Institution size
6 or fewer
More than 6

Ref

NA

0.84 (0.45-1.55)

.571

BMI indicates body mass index; NA, not applicable; and Ref, reference group.
*Underweight ( 10th percentile; normal weight [between 10th and 95th
percentile]); and overweight ( 95th percentile).
Institution size categorized by median number of enrollments per year.

reflection of the real incidence of microbiologically documented

infections compared with studies in which these data were collected retrospectively. Second, we have described one of the largest
cohorts of children with AML in which infection-related mortality
is described. This point is important as infection-related mortality
is the most clinically relevant infection end point, and our study is
one of a few pediatric AML studies sufficiently powered to model
infectious deaths. Finally, our study is the only one to track
individual patients over several courses of chemotherapy and, thus,
we have been able to describe the recurrence risks with specific
microorganisms.
Infection-related mortality in this study was higher than that
demonstrated in other pediatric AML studies. An infectionrelated mortality of 25 (7.3%) of 341 during chemotherapy was
reported for children treated according to a United Kingdom
protocol, Medical Research Council 10 trial (MRC-10).4
A retrospective analysis of a German study AML-BFM 93 noted
855 infectious complications occurring in 304 children with
AML, with 20 (6.6%) dying of infection (5.4% in nonDown
syndrome children).5 Similarly, a single institution study found
that 5 (6.4%) of 78 of their children with AML died of
infection.6 The reason for higher infection-related mortality on
CCG 2961 may be related to the intensely myelosuppressive
nature of this chemotherapy regimen. Alternatively, in intensive

From www.bloodjournal.org by guest on January 20, 2015. For personal use only.
3538

SUNG et al

chemotherapy protocols, treatment-related mortality may decline as treating centers become more familiar with managing
these complex children.7 More specifically, children enrolled on
MRC-10 had a decline in treatment-related mortality from
17.8% in the earlier period to 9.6% in the later period.7 Given
that we are presenting infection-related mortality for the earliest
children enrolled on CCG 2961, our higher rates may reflect an
early experience with a new treatment protocol rather than
fundamentally different rates of infection-related mortality.
It is not known why underweight children with AML have
higher infection-related mortality and overweight children have
more infections depending on the specific chemotherapy administered. One possible explanation is that since drug doses were not
modified depending on BMI, that drug exposure was significantly
different in underweight and overweight children. However, this
hypothesis is not supported by the duration of neutropenia and time
to initiation of the next phase of chemotherapy, which were shorter
in overweight children treated on this protocol.1 Another possibility
is that underweight and overweight children have important
comorbidities or nutritional compromise that could explain differences in infection outcomes.
The factors predictive of microbiologic documented infections
did not precisely match those predictive of infection-related
mortality. For example, while overweight status was predictive of
microbiologically documented infections in phase 2, only underweight children were at significantly high risk of infection-related
mortality. Similarly, while we did not find that ethnicity was
a significant predictor of microbiologically documented infections,
we found that nonwhite children had a higher risk of infectionrelated mortality. This divergence could be hypothesized to be
related to several potential issues. First, infection-related mortality
includes mortality related to clinically documented infections as
well as microbiologically documented infections, and factors
predictive of clinically documented infections may differ from
those predictive of microbiologically documented infections. Second, the analysis of the number of infections was conducted
separately for each phase, whereas infection-related mortality
reflected a summary of events occurring throughout the course of
therapy. Third, the inability to distinguish common contaminants
from invasive infection may have diluted our ability to predict
invasive microbiologically documented infections. Finally, it is
possible that the lack of significantly increased infection-related
mortality in the overweight group is related to sample size, given
that an analysis of the entire CCG 2961 patient cohort did
demonstrate increased infection-related mortality in the overweight
group.1
As CCG 2961 contained one randomization in phase 2, this
design provided an opportunity to more precisely estimate the
contribution of different drugs to infection outcomes. It is
interesting that the fludarabine-based regimen was associated
with more treatment-related mortality and a higher rate of alpha
hemolytic streptococcal infection despite also being associated
with a shorter duration of neutropenia. There were 3 major
differences between the 2 regimens that might explain these
findings, namely, the presence of fludarabine, higher daily dose
of idarubicin, and higher daily dose of cytarabine associated
with the fludarabine-based regimen (although the cumulative
doses of idarubicin and cytarabine were similar). Given previous reports that have found that high-dose cytarabine is a risk
factor for invasive viridans group streptococcal infections,8-13 it
seems that the latter may be the more likely etiologic factor

BLOOD, 15 NOVEMBER 2007 VOLUME 110, NUMBER 10

behind worse infection outcomes associated with the fludarabine

regimen.
A limitation of our study is that insufficient clinical information
was available to distinguish between likely contaminants versus
invasive infections. Therefore, the estimate of the number of
Gram-positive infections may be an overestimate. Similarly, we did
not know whether urine specimens were catheter, midstream, or
bagged specimens and thus, we likely have overestimated the rate
of this infection as well. Conversely, given that our classification of
infection relied on microbiology, we likely underestimated the risk
of Aspergillus infections, an important point when considering the
large apparent contribution of Aspergillus to infection-related
mortality even when only microbiology was considered. Finally,
we chose to censor patients at the completion of protocol therapy or
30 days following removal from protocol to exclude those who
died of infection related to a relapse or salvage protocol. However,
it is possible that some patients may have experienced late
infection-related mortality (for example, from Aspergillus) following completion of protocol therapy or removal from protocol
therapy, and thus we may have underestimated both the risk of
infection-related mortality as well as the contribution of Aspergillus to infectious deaths.
For those caring for children with AML, knowledge of the
polymicrobial nature of infection-related mortality emphasizes the
need to continue broad-spectrum coverage in these patients during
neutropenic infectious episodes, even if investigations reveal a
specific microorganism. In addition, given the large contribution of
invasive fungal infection and in particular Aspergillus to infectious
deaths, careful consideration of antifungal coverage and thorough
evaluation for fungus are warranted in the appropriate setting.
In summary, microbiologically documented infections and
infection-related mortality are common in pediatric AML. Aspergillus species was the most common contributor to infectious deaths.
Age, ethnicity, and BMI were predictors of infection-related
mortality. Further understanding of the etiology behind this relationship will help develop future AML protocols.

Acknowledgments
L.S. is supported by a Career Development Award with the
Canadian Child Health Clinician Scientist Training Program, a
strategic training program with the Canadian Institutes of Health
Research. B.J.L. holds the Yetta Dietch Novotny Chair in Clinical
Oncology. We also acknowledge the support of Cancer Control and
Acute Myeloid Leukemia committees at the Childrens Oncology
Group.

Authorship
Contribution: R.B.G. and T.A.A. analyzed the data; L.S. wrote the
paper. All authors designed and performed the research, revised the
paper for critical content, and approved the final version of the
paper.
Conflict-of-interest disclosure: The authors declare no competing financial interests.
Correspondence: Lillian Sung, Division of Haematology/
Oncology, Hospital for Sick Children, 555 University Avenue,
Toronto, ON Canada M5G 1X8; e-mail: lillian.sung@sickkids.ca.

From www.bloodjournal.org by guest on January 20, 2015. For personal use only.
BLOOD, 15 NOVEMBER 2007 VOLUME 110, NUMBER 10

INFECTIONS IN PEDIATRIC AML

3539

References
1. Lange BJ, RBG, Feusner J, et al. Excess mortality in morbidly obese and underweight pediatric
patients with acute myeloid leukemia. JAMA.
2004;293:203-211.
2. Ogden CL, Kuczmarksi RJ, Flegal KM, et al. Centers for Disease Control and Prevention 2000
growth charts for the United States: improvements to the 1977 National Center for Health Statistics version. Pediatrics. 2002;109:45-80.
3. Cox DR. Regression models and life-tables (with
discussion). J Royal Stat Soc B. 1972;34:187-220.
4. Stevens RF, Hann IM, Wheatley K, Gray RG.
Marked improvements in outcome with chemotherapy alone in paediatric acute myeloid leukemia: results of the United Kingdom Medical Research Councils 10th AML trial: MRC Childhood
Leukaemia Working Party. Br J Haematol. 1998;
101:130-140.
5. Lehrnbecher T, Varwig D, Kaiser J, Reinhardt D,
Klingebiel T, Creutzig U. Infectious complications
in pediatric acute myeloid leukemia: analysis of

the prospective multi-institutional clinical trial

AML-BFM 93. Leukemia. 2004;18:72-77.
6. Brunet AS, Ploton C, Galambrun C, et al. Low
incidence of sepsis due to viridans streptococci in
a ten-year retrospective study of pediatric acute
myeloid leukemia. Pediatr Blood Cancer. 2006;
47:765-772.
7. Riley LC, Hann IM, Wheatley K, Stevens RF.
Treatment-related deaths during induction and
first remission of acute myeloid leukaemia in children treated on the Tenth Medical Research
Council acute myeloid leukaemia trial (MRC
AML10): The MCR Childhood Leukaemia Working Party. Br J Haematol. 1999;106:436-444.
8. Engelhard D, Elishoov H, Or R, et al. Cytosine
arabinoside as a major risk factor for Streptococcus viridans septicemia following bone marrow
transplantation: a 5-year prospective study. Bone
Marrow Transplant. 1995;16:565-570.
9. Gamis AS, Howells WB, DeSwarte-Wallace J,
Feusner JH, Buckley JD, Woods WG. Alpha he-

molytic streptococcal infection during intensive

treatment for acute myeloid leukemia: a report
from the Childrens cancer group study
CCG-2891. J Clin Oncol. 2000;18:1845-1855.
10. Kern W, Kurrle E, Schmeiser T. Streptococcal
bacteremia in adult patients with leukemia undergoing aggressive chemotherapy: a review of 55
cases. Infection. 1990;18:138-145.
11. Kern W, Kurrle E, Vanek E. High risk of strep-tococcal septicemia after high dose cytosine arabinoside treatment for acute myelogenous leukemia. Klin Wochenschr. 1987;65:773-780.
12. Rossetti F, Cesaro S, Putti MC, Zanesco L. Highdose cytosine arabinoside and viridans streptococcus sepsis in children with leukemia. Pediatr
Hematol Oncol. 1995;12:387-392.
13. Woolley I, Curtis D, Szer J, et al. High dose cytosine arabinoside is a major risk factor for the development of hepatosplenic candidiasis in patients with leukemia. Leuk Lymphoma. 1997;27:
469-474.

From www.bloodjournal.org by guest on January 20, 2015. For personal use only.

2007 110: 3532-3539

doi:10.1182/blood-2007-05-091942 originally published
online July 27, 2007

Microbiologically documented infections and infection-related mortality

in children with acute myeloid leukemia
Lillian Sung, Beverly J. Lange, Robert B. Gerbing, Todd A. Alonzo and James Feusner

Updated information and services can be found at:

http://www.bloodjournal.org/content/110/10/3532.full.html
Articles on similar topics can be found in the following Blood collections
Clinical Trials and Observations (4009 articles)
Neoplasia (4212 articles)
Information about reproducing this article in parts or in its entirety may be found online at:
http://www.bloodjournal.org/site/misc/rights.xhtml#repub_requests
Information about ordering reprints may be found online at:
http://www.bloodjournal.org/site/misc/rights.xhtml#reprints
Information about subscriptions and ASH membership may be found online at:
http://www.bloodjournal.org/site/subscriptions/index.xhtml

Blood (print ISSN 0006-4971, online ISSN 1528-0020), is published weekly by the American Society
of Hematology, 2021 L St, NW, Suite 900, Washington DC 20036.
Copyright 2011 by The American Society of Hematology; all rights reserved.