Clinical Infectious Diseases

MAJOR ARTICLE

Levofloxacin Prophylaxis During Induction Therapy for

Pediatric Acute Lymphoblastic Leukemia
Joshua Wolf,1,2,3 Li Tang,4 Patricia M. Flynn,1,2 Ching-Hon Pui,2,5,6 Aditya H. Gaur,1,2 Yilun Sun,4 Hiroto Inaba,2,5 Tracy Stewart,1 Randall T. Hayden,6
Hana Hakim,1 and Sima Jeha2,5
Departments of 1Infectious Diseases, 4Biostatistics, 5Oncology, and 6Pathology, St. Jude Children’s Research Hospital, and 2Department of Pediatrics, University of Tennessee Health Science
Center, Memphis; 3Department of Paediatrics, University of Melbourne, Parkville, Victoria, Australia

Background. Infection is the most important cause of treatment-related morbidity and mortality in pediatric patients treated
for acute lymphoblastic leukemia (ALL). Although routine in adults with leukemia, antibacterial prophylaxis is controversial in
pediatrics because of insufficient evidence for its efficacy or antibiotic choice and concerns about promoting antibiotic resistance
and Clostridium difficile infection.
Methods. This was a single-center, observational cohort study of patients with newly diagnosed ALL, comparing prospectively
collected infection-related outcomes in patients who received no prophylaxis, levofloxacin prophylaxis, or other prophylaxis during
induction therapy on the total XVI study. A propensity score–weighted logistic regression model was used to adjust for confounders.
Results. Of 344 included patients, 173 received no prophylaxis, 69 received levofloxacin prophylaxis, and 102 received other
prophylaxis regimens. Patients receiving prophylaxis had longer duration of neutropenia. Prophylaxis reduced the odds of febrile
neutropenia, likely bacterial infection, and bloodstream infection by ≥70%. Levofloxacin prophylaxis alone reduced these infections,
but it also reduced cephalosporin, aminoglycoside, and vancomycin exposure and reduced the odds of C. difficile infection by >95%.
No increase in breakthrough infections with antibiotic-resistant organisms was seen, but this cannot be excluded.
Conclusions. This is the largest study to date of antibacterial prophylaxis during induction therapy for pediatric ALL and the
first to include a broad-spectrum fluoroquinolone. Prophylaxis prevented febrile neutropenia and systemic infection. Levofloxacin
prophylaxis also minimized the use of treatment antibiotics and drastically reduced C. difficile infection. Although long-term anti-
biotic-resistance monitoring is needed, these data support using targeted prophylaxis with levofloxacin in children undergoing
induction chemotherapy for ALL.
Clinical Trials Registration. NCT00549848
Keywords. prophylaxis; leukemia; child; levofloxacin; Clostridium difficile.

Infections remain the most frequent cause of serious treat- patients with acute leukemia. In addition, findings of animal
ment-related morbidity and mortality in children and adoles- studies suggesting that colonization with Clostridium difficile
cents with acute lymphoblastic leukemia (ALL); currently, up to might be reduced by replacing β-lactam antibiotics with fluoro-
4% of children with ALL die of infection [1–4]. Even nonfatal quinolones [7] are intriguing because hospital-acquired C. dif-
infections can result in permanent end-organ damage, contrib- ficile infection is associated with a 6.7-fold increase in the odds
ute to chemotherapy delay or modification, and increase anti- of mortality in hospitalized children [8]. However, no clinical
biotic exposure [5]. Most serious infections occur during the studies have shown that fluoroquinolone prophylaxis can pre-
relatively short induction phase of chemotherapy [2]. vent C. difficile infection in high-risk patients.
Primary antibacterial prophylaxis during chemotherapy-re- In the pediatric ALL population, primary antibacterial
lated neutropenia in adults reduces clinically documented prophylaxis is controversial, because data supporting its efficacy
infection, microbiologically documented infection, and infec- and safety are sparse. There are 2 published studies of antibac-
tion-related mortality [6]. The US National Comprehensive terial prophylaxis with fluoroquinolones for pediatric ALL. An
Cancer Network recommends antibacterial prophylaxis with unadjusted retrospective analysis, from Saudi Arabia, reported
levofloxacin, a broad-spectrum fluoroquinolone, for adult reduced bloodstream infection (BSI) and other infections with
ciprofloxacin prophylaxis during delayed intensification [9].
Received 11 May 2017; editorial decision 7 July 2017; accepted 24 July 2017; published online However, a randomized controlled trial of ciprofloxacin proph-
September 14, 2017.
ylaxis during induction therapy in Indonesia found a marked
Correspondence: J. Wolf, Department of Infectious Diseases, St Jude Children’s Research
Hospital, 262 Danny Thomas Pl, MS 320, Memphis, TN 38105 ([email protected]). increase in the risk of fever, BSI, and death [10]. No published
Clinical Infectious Diseases®  2017;65(11):1790–8 studies to our knowledge conducted in the pediatric population
© The Author 2017. Published by Oxford University Press for the Infectious Diseases Society have evaluated the use of broad-spectrum fluoroquinolones
of America. All rights reserved. For permissions, e-mail: [email protected].
DOI: 10.1093/cid/cix644 (which may have greater efficacy than narrow-spectrum agents

1790 • CID 2017:65 (1 December) • Wolf et al

such as ciprofloxacin), compared fluoroquinolones with other typically consisted of cefepime, ciprofloxacin, or vancomycin
prophylaxis regimens, or assessed the impact of antibacterial plus cefepime or ciprofloxacin and was initiated after the first
prophylaxis on C. difficile infection. The current study aimed documented neutropenia after chemotherapy. From August
to identify the effects of antibacterial prophylaxis during induc- 2014, institutional guidelines recommended levofloxacin
tion therapy for pediatric ALL on antibiotic exposure, infection, prophylaxis during all episodes of neutropenia expected to last
and febrile neutropenia, and on C. difficile infection. ≥7 days.

METHODS Management of Febrile Neutropenia and Suspected Infection

The recommended febrile neutropenia treatment was cefepime
This was an observational cohort study of patients receiving
(or ceftazidime plus vancomycin if cefepime was unavailable),
induction therapy for newly diagnosed ALL. Infection-related
with addition of vancomycin or an aminoglycoside for speci-
adverse event data were collected prospectively by the study team
fied indications [12, 13] and substitution of meropenem for
and evaluated by cross-referencing with the electronic medical
cefepime in patients who had recently received cefepime or had
record and institutional microbiology and pharmacy databases.
suspected intra-abdominal or Bacillus cereus infection [13].
Data describing prophylaxis prescribing and antibiotic expo-
These guidelines did not change during the study period.
sure were collected retrospectively. The study was approved by
the St Jude Children’s Research Hospital Institutional Review Definitions Related to Infection
Board (XPD04-070). The Common Terminology Criteria for Adverse Events, ver-
sion 3.0 [14], were used to capture all infection-related adverse
Inclusion and Exclusion Criteria
events in the study database, and standard definitions were used
Eligible for inclusion were all patients enrolled in the total XVI
to further categorize episodes (Table 1). Antibiotic exposure
study for newly diagnosed ALL (ClinicalTrials.gov identifier:
was quantified as “antibiotic days,” “specific antibiotic days,”
NCT00549848) at St Jude Children’s Research Hospital from the
and “cumulative antibiotic exposure.” Antibiotic days were cal-
opening of the trial (29 October 2007) who completed induc-
culated for each patient as the simple proportion of induction
tion therapy before 6 January 2016. Similarly to a previous
days on which ≥1 systemic antibacterial antibiotic was admin-
study, total XV [11], induction therapy was administered over
istered (excluding Pneumocystis pneumonia prophylaxis), and
42 days and comprised 4 weeks of prednisone, 4 doses of weekly
specific antibiotic days were calculated as the proportion of
vincristine, 2 doses of weekly daunorubicin, and 1 or 2 doses
induction days on which a specific antibiotic was administered
of PEG-asparaginase, followed by 2 weeks of cyclophospha-
(cefepime and ceftazidime were combined). Cumulative antibi-
mide, cytarabine, and thioguanine, plus intrathecal chemother-
otic exposure was calculated as the sum of all specific antibiotic
apy according to risk category (Supplementary Table S1). The
days divided by the number of induction days, accounting for
induction phase could be modified or prolonged in response to
the potential additive effect of coadministered antibiotics.
infection or other complications. Participants with clinically or
microbiologically documented infection before induction ther-
apy initiation or who had fever before induction therapy requir- Statistical Methods
ing prolonged antibiotic therapy (>4 days) were excluded from Fisher’s exact test was used to compare proportions, and
this analysis to avoid misclassification of antibacterial treatment Kruskal-Wallis or Wilcoxon-Mann-Whitney tests were used to
as prophylaxis. Similarly, patients who developed febrile neutro- compare medians. Associations between primary antibacterial
penia or suspected infection during the first 7 days of induction prophylaxis and clinical outcomes were assessed using multiple
or after <2 days of neutropenia were excluded because there was logistic regression, and Firth’s penalized likelihood approach
insufficient opportunity to initiate primary prophylaxis. was used to address the problem of complete separation [20].
Propensity to receive prophylaxis was estimated with a multi-
Definition of Primary Prophylaxis nomial logistic regression model including age, sex, race, Down
Primary prophylaxis was defined as the administration of sys- syndrome, and leukemia type as predictors. Then, a logistic
temic antibacterial agents during induction therapy, before the regression model was used to model prophylaxis groups and
first episode of suspected or proven infection or febrile neutro- duration of profound neutropenia as predictors, with inverse
penia, with the intent to prevent infection. Appropriate peri- probability weighting based on the propensity score. Kaplan-
operative or Pneumocystis pneumonia prophylaxis and empiric Meier estimates of disease-free survival were used to graph the
short courses of antibiotic given for fever at presentation or time to first event for infection events in prophylaxis groups
for clinical events other than infection or febrile neutropenia, and were compared using the log-rank test. Statistical analysis
such as vomiting, headache, or vertigo, were not considered was performed with SAS (version 9.4; SAS Institute), and differ-
prophylaxis. From 2007 until July 2014, antibacterial prophy- ences were considered significant at P < .05 (2 tailed). No power
laxis was provided at the discretion of the treating clinician. It calculation was performed.

Prophylaxis in Pediatric ALL • CID 2017:65 (1 December) • 1791

Table 1. Definitions Related to Infections

Outcome Definition

Neutropenia Absolute neutrophil count ≤500/μL

Profound neutropenia Absolute neutrophil count ≤100/μL
Febrile neutropenia Core body temperature ≥38.3°C or ≥38.0°C for ≥1 h in context of neutropenia [15, 16]
Microbiologically documented Diagnosed bacterial, viral, fungal, or parasitic infection, with supportive microbiological evidence, such as a positive culture,
infection antigen or PCR test results, or characteristic histopathological findings [15]
Clinically documented infection Infection diagnosed by the treating clinician for which a specific microbial cause could not be demonstrated [15]
Likely bacterial infection Any infection with a microbiologically documented bacterial cause or that was clinically documented in categories typically
attributed to bacterial infection, including pneumonia, skin and soft-tissue infection, osteomyelitis or myositis, enterocolitis,
otitis media or externa, sinusitis, epididymo-orchitis, CVC pocket or tunnel infection, pharyngitis, perianal abscess or celluli-
tis, peritonitis, lymphadenitis, or culture-negative sepsis
Bloodstream infection Any infection caused by a recognized pathogen that was isolated from ≥1 blood culture in the context of a compatible clinical
illness; common commensal bacteria were included if identified from multiple culture sets, or if a single blood culture set
was collected before start of antibiotic therapy and the result deemed clinically significant by the treating clinician [17, 18]
Severe sepsis Infection in conjunction with severe dysfunction of cardiovascular or respiratory systems or ≥2 other organ systems [19];
fever or elevated WBC count not required
Urgent intervention Infection requiring admission to the intensive care unit, fluid resuscitation, or supplemental oxygen or associated with clini-
cian diagnosis of septic shock
Clostridium difficile infection Identification of C. difficile toxin in stool by enzyme immunoassay (from 2007 to April 2010) or toxin gene PCR (from April
2010 onward) in the presence of diarrhea
Enterocolitis Syndrome of abdominal pain in conjunction with radiologically documented bowel-wall thickening or microbiologically docu-
mented C. difficile infection

Abbreviations: CVC, central venous catheter; PCR, polymerase chain reaction; WBC, white blood cell count.

RESULTS documented infection, enterocolitis, C. difficile infection, likely

Of 505 patients assessed for eligibility, 161 were excluded because bacterial infection, or BSI (Table 4) [12]. Reductions in C. difficile
of the onset of suspected infection before induction therapy infection and enterocolitis occurred only in patients receiving
(n = 64), within the first 7 days of induction (n = 41), or after <2 fluoroquinolone prophylaxis; patients receiving cefepime-based
days of neutropenia (n = 19), or because of fever at presentation prophylaxis had a higher risk of both (Supplementary Figure S2
requiring prolonged treatment (n = 37). Of the remaining 344 and Supplementary Table S3). Multiple logistic regression analy-
patients, 173 received no primary prophylaxis, 69 received levo- sis (Supplementary Table S4) and Kaplan-Meier analysis showed
floxacin prophylaxis, and 102 received other prophylaxis (Table similar results (Figure 2 and Supplementary Figure S3).
2 and Supplementary Figure S1). Other prophylaxis regimens Levofloxacin prophylaxis alone was associated with a sim-
comprised cefepime, ciprofloxacin, or vancomycin plus cefepime ilar, statistically significant reduction in febrile neutropenia,
or ciprofloxacin (Supplementary Table S2). The median dura- microbiologically documented infection, and likely bacterial
tions of neutropenia and profound neutropenia were longer in infection. (Table 5) There was also significantly less enterocolitis
patients receiving prophylaxis (Table 2), so duration of profound and C. difficile infection (Table 5). Multiple logistic regression
neutropenia was included with propensity score in multivariate analysis (Supplementary Table S4) and Kaplan-Meier analysis
analyses. Analyses related to C. difficile infection also included showed similar results (Figure 2 and Supplementary Figure S4).
duration of meropenem exposure, because this drug was tar- Compared with patients receiving other prophylaxis reg-
geted by antimicrobial stewardship interventions. imens, patients receiving levofloxacin prophylaxis had a simi-
Total antibiotic days and cumulative antibiotic exposure were lar risk of febrile neutropenia, clinically documented infection,
significantly increased in patients who received prophylaxis microbiologically documented infection, likely bacterial infec-
(P < .001) (Figure 1). The pattern of antibiotic use in patients tion, and BSI (Table 5). However, levofloxacin prophylaxis was
receiving levofloxacin differed from that in patients receiving associated with a significantly lower rate of C. difficile infec-
no prophylaxis; levofloxacin exposure increased, but there was tion (adjusted odds ratio, 0.04; 95% confidence interval, <.01 to
a concomitant reduction in exposure to cefepime/ceftazidime, .36; P < .001) (Table 5). There was some heterogeneity among
vancomycin, meropenem (P < .001 for all comparisons), and alternative regimens: Pooled ciprofloxacin-based regimens,
aminoglycosides (P = .002) (Figure 1). predominantly ciprofloxacin plus vancomycin, seemed to be
Table 3 shows the incidence of infection-related adverse events superior for some outcomes, but no individual regimen was
in this population. After adjustment for other covariates, patients significantly superior to levofloxacin (Supplementary Figure S2
receiving any antibacterial prophylaxis were significantly less and Supplementary Table S3). Multiple logistic regression analy-
likely than those receiving no prophylaxis to have febrile neu- sis (Supplementary Table S4) and Kaplan-Meier analysis (Figure
tropenia, clinically documented infection, microbiologically 2 and Supplementary Figure S4) again showed similar results.

1792 • CID 2017:65 (1 December) • Wolf et al

Table 2. Characteristics of Included Patients With or Without Prophylaxis

Patients, No. (%)a

Levofloxacin Prophylaxis
Characteristic No Prophylaxis (n = 173) (n = 69) Other Antibiotic Prophylaxis (n = 102) P Valueb

Age, median (IQR), y 5.8 (3 –11.9) 6.8 (3.9–11.1) 7 (3.6 –11.9) .67
Age group .75
≥10 y 50 (29) 18 (26) 32 (31)
<10 y 123 (71) 51 (74) 70 (69)
Sex .59
Male 103 (60) 43 (62) 56 (55)
Female 70 (40) 26 (38) 46 (45)
Race .86
White 134 (77) 56 (81) 80 (78)
Others 3 (23) 13 (19) 22 (22)
Down syndrome 4 (2) 1 (1) 2 (2) >.99
ALL type .57
T 29 (17) 15 (22) 21 (21)
B 144 (83) 54 (78) 81 (79)
ALL risk category .14
Low 89 (51) 37 (54) 51 (50)
Standard 81 (47) 28 (41) 43 (42)
High 3 (2) 4 (6) 8 (8)
Duration of neutropenia, 17 (11–24) 18 (12–23) 20 (17–25) .002
median (IQR), d
Duration of profound 6 (2–13) 7 (4–12) 11 (5–16) .001
neutropenia, median
(range), d

Abbreviations. ALL, acute lymphoblastic leukemia; IQR, interquartile range.

a
Data represent No. (%) of patients except where otherwise specified.
b
Fisher exact test was used for categorical and Kruskal-Wallis test for continuous variables.

There were 28 episodes of bacterial BSI in 28 patients dur- of antibacterial prophylaxis in predominantly adult patients
ing the study period (Supplementary Table S5). Although BSI showed that prophylaxis reduced febrile episodes, documented
was less common in patients receiving prophylaxis, the caus- infection, infection-related mortality, and all-cause mortal-
ative organisms were similar (Supplementary Table S5), and ity [6]. Few studies of fluoroquinolone prophylaxis in children
the proportions of BSI episodes complicated by severe sepsis with leukemia have been published; most showed a reduc-
or requirement for urgent intervention did not differ between tion in infection-related adverse events [9, 21–24], but some
patients receiving prophylaxis and those receiving none (2 of 9 showed a paradoxical increase in these complications [10, 25].
[22.2%] vs 4 of 19 [21.1%], respectively; P > .99). One patient Trimethoprim-sulfamethoxazole prophylaxis can also reduce
receiving levofloxacin prophylaxis developed bacteremia caused bacterial infection [26] but was routine in all groups in this study.
by an extended-spectrum β-lactamase–producing Escherichia The current study provides new information by comparing
coli strain that was resistant to levofloxacin and cefepime. prophylaxis with levofloxacin, a broad-spectrum fluoroquinolone,
to alternative prophylaxis regimens and by excluding patients who
DISCUSSION
had no opportunity to receive prophylaxis, carefully controlling
In this largest study to date, primary antibacterial prophylaxis for potential confounders and examining C. difficile infection.
during induction therapy for ALL was associated with markedly Our study has several strengths related to the institutional setting
reduced rates of febrile neutropenia, clinical or microbiologi- and study design. All patients were treated according to a single
cally documented infection, likely bacterial infection, and BSI. protocol at a single institution and were provided with high-qual-
Although prophylaxis increased overall antibiotic exposure, ity supportive care [2, 27, 28]. We used prespecified standard
patients receiving levofloxacin prophylaxis had less exposure definitions of febrile neutropenia, clinically or microbiologically
to cefepime/ceftazidime, vancomycin, and aminoglycosides. documented infection (which includes nonbacterial infections),
Unexpectedly, prophylaxis with levofloxacin drastically reduced and BSI [15]. We added the category of likely bacterial infection,
the risk of C. difficile infection and all-cause enterocolitis. comprising infections typically attributed to bacteria.
The observed reduction in infection and febrile neutropenia Patients with early infection or febrile neutropenia were
extends data obtained from adult populations. A meta-analysis excluded from analysis because they might falsely raise the

Prophylaxis in Pediatric ALL • CID 2017:65 (1 December) • 1793

Figure 1. Antibiotic exposure during induction for each antibiotic prophylaxis group. Data are shown in a box plot; each box plot illustrates the upper and lower quartile
(box), median (line inside box), adjacent values (whiskers), and outliers (open circles). Antibiotic exposure and cumulative antibiotic exposure were significantly greater in
patients receiving levofloxacin or other prophylaxis than in those receiving no prophylaxis (P < .001 for all comparisons). However, patients receiving levofloxacin prophylaxis
had less exposure to cefepime/ceftazidime, vancomycin, meropenem, or aminoglycosides (data not shown) when compared with those receiving no prophylaxis (P < .001,
P < .001, P < .001, and P = .002, respectively) or other prophylaxis (P < . 001, P < .001, P < .001, and P = .04, respectively).

apparent rate of infection in the no-prophylaxis group. Similarly, antipseudomonal cephalosporin and β-lactam antibiotics, espe-
excluding patients who had an infection before receiving cially cefepime, was the most important contributor to the risk
chemotherapy or who received prolonged treatment for fever of C. difficile infection, but this study did not assess the effects
at presentation ensured that only potentially preventable out- of quinolone exposure [30].
comes were analyzed. Not all patients with fever at presentation In an animal model, fluoroquinolone antibiotics did not
were excluded, because fever occurs in about 59% of patients impair resistance to colonization with C. difficile, whereas
and is rarely related to infection [29]. Detailed chart review β-lactam antibiotics did [7]. Therefore, using fluoroquinolone
ensured appropriate classification of prophylaxis regimens. prophylaxis to reduce exposure to other antibiotics might pre-
Although observational studies of this type can be subject to vent C. difficile infection, but this is the first study to show such
selection bias, we aimed to ameliorate such bias by applying a a benefit. Previous studies of quinolone prophylaxis in adults
propensity score–weighted logistic regression model to account showed no such effect on C. difficile infection [6], and some
for differences between groups. studies have even identified fluoroquinolone exposure as a risk
The observed reduction in enterocolitis and C. difficile infec- factor for C. difficile infection [35]. The difference in the cur-
tion associated with levofloxacin prophylaxis is important rent study may be related to the marked reduction in the use
because these conditions can cause serious illness, malnutri- of broad-spectrum antibiotic therapy or to predominance of
tion, chemotherapy delay, or poor absorption of medications quinolone-susceptible C. difficile [2, 35].
for cancer treatment, all of which can reduce the chance of Other potentially confounding variables that affect the risk of
cure [8, 30–33]. Furthermore, 2 studies found that C. difficile C. difficile infection in children with leukemia include shorter
infection in hospitalized children markedly increased the risk hospital stay and granulocyte colony-stimulating factor admin-
of all-cause mortality [8, 34]. Patients receiving levofloxa- istration [36], but there was no change in discharge guidelines
cin prophylaxis were protected from this infection, despite an during the study period, and granulocyte colony-stimulating
incongruous increase in their total antibiotic exposure [30, 32]. factor was not used in this protocol. Importantly, the institu-
A recent study in children with ALL found that exposure to tional rate of C. difficile infection did not change during the

1794 • CID 2017:65 (1 December) • Wolf et al

Table 3. Unadjusted Incidence of Infection-Related Adverse Events by Prophylaxis Group

Patients, No. (%)

Prophylaxis Regimen

Outcome No Prophylaxis (n = 173) Any Prophylaxis (n = 171) Levofloxacin (n = 69) Other Antibiotic (n = 102)

Febrile neutropenia 106 (61.3) 74 (43.3) 28 (40.6) 46 (45.1)

Febrile neutropenia without 60 (34.7) 49 (28.7) 19 (27.5) 30 (29.4)
CDI or MDI
Any documented infectiona 97 (56.1) 71 (41.5) 27 (39.1) 44 (43.1)
CDI 52 (30.1) 37 (21.6) 13 (18.8) 24 (23.5)
MDI 69 (39.9) 43 (25.1) 18 (26.1) 25 (24.5)
BSI 19 (11) 9 (5.3) 4 (5.8) 5 (4.9)
Clostridium difficile infection 17 (9.8) 9 (5.3) 0 (0) 9 (8.8)
Any enterocolitis 25 (14.5) 15 (8.8) 3 (4.3) 12 (11.8)
Likely bacterial infection 64 (37) 33 (19.3) 11 (15.9) 22 (21.6)

Abbreviations: BSI, bloodstream infection; CDI, clinically documented infection; MDI, microbiologically documented infection.
a
“Any documented infection” included CDI or MDI, with or without neutropenia.

study period (Supplementary Figure S5). The effect of prophy- Other than neutropenia duration, risk factors for infection
laxis during induction on C. difficile infection in postinduction did not differ significantly between groups (Table 2); known
chemotherapy cannot be determined from this study. variables were addressed by a propensity score–weighted mul-
The study has some limitations. Grouping alternative tivariate logistic regression analysis (Tables 4 and 5). Another
prophylaxis regimens provides a robust comparison group potential confounder is the time period, because routine lev-
but can mask differences between heterogeneous regimens ofloxacin was introduced in 2014 and the other groups were
(Supplementary Figure S2 and Supplementary Table S3). The treated in 2008–2014. There is no evidence that this explains
study was not powered to detect a difference in outcomes the differences between groups, because there were no trends
between different fluoroquinolones, and the reduction in C. in infection rates in study patients who received no proph-
difficile infection was seen in both levofloxacin and ciproflox- ylaxis before 2014 (Supplementary Figure S6) and no sig-
acin-based prophylaxis regimens (Supplementary Table S3). nificant decrease in institutional rates of hospital-acquired
Until 2014, antimicrobial prophylaxis was provided at the infection or C. difficile infection during the entire study period
discretion of the treating physician, so the potential for con- (Supplementary Figure S5).
founding is an important issue, especially in the context of a Antimicrobial stewardship interventions introduced dur-
single-center study, which can amplify the effects of confound- ing the study period included prospective audit with feedback
ers. Clinicians may have preferentially prescribed prophylaxis for meropenem and linezolid use; however, no intervention
to patients at higher risk of infection, but this confounding by aimed at reducing the use of vancomycin, cefepime/ceftazi-
indication would be expected to mask, rather than inflate, the dime, or aminoglycosides. Meropenem exposure was included
benefits of prophylaxis. as a covariate in the analysis of C. difficile infections to reduce

Table 4. Propensity Score–weighted Multivariate Logistic Regression Analysis for Effectiveness of Primary Prophylaxis

Outcome Crude OR (95% CI) P Value Adjusted OR (95% CI)a P Value

b
Febrile neutropenia 0.48 (.31–.74) <.001 0.23 (.14–.40) <.001b
b
Febrile neutropenia with CDI 0.44 (.22–.89) .02 0.30 (.14–.65) .002b
b
Febrile neutropenia with MDI 0.40 (.23–.70) .001 0.25 (.14–.48) <.001b
CDI 0.64 (.39–1.05) .08 0.54 (.32–.90) .02b
MDI 0.51 (.32–.80) .004b 0.40 (.24–.65) <.001b
BSI 0.45 (.20–1.03) .06 0.30 (.13–.73) .008b
c
Clostridium difficile infection 0.51 (.22–1.18) .11 0.38 (.16–.93) .04b
b
Likely bacterial infection 0.41 (.25–.66) <.001 0.26 (.15–.45) <.001b
Any enterocolitis 0.57 (.29–1.12) .10 0.44 (.21–.91) .03b

Abbreviations: BSI, bloodstream infection; CDI, clinically documented infection; CI, confidence interval; MDI, microbiologically documented infection; OR, odds ratio.
a
The propensity score of prophylaxis groups was estimated using a multinomial logistic regression model with age, sex, race, Down syndrome, and leukemia type as predictors; a logistic
regression model was then used to model prophylaxis groups and exposure to profound neutropenia during induction as predictors of clinical outcomes, with inverse probability weighting
using the propensity score.
b
Significant difference (prophylaxis versus no prophylaxis; P < .05).
c
Also adjusted for exposure to meropenem.

Prophylaxis in Pediatric ALL • CID 2017:65 (1 December) • 1795

Figure 2. Kaplan-Meier analysis of time to infectious complications during induction for each antibiotic prophylaxis group. Patients receiving levofloxacin prophylaxis dur-
ing induction therapy for pediatric acute lymphoblastic leukemia had a lower cumulative incidence of C. difficile infection (A) than those receiving no prophylaxis (P = .008) or
other prophylaxis regimens (P = .01) and a lower cumulative incidence of enterocolitis (B) than those receiving no prophylaxis (P = .03). Patients receiving any prophylaxis had
a lower cumulative incidence of febrile neutropenia (C) (P < .001) and likely bacterial infection (D) (P < .001), but there was no significant difference in the cumulative incidence
of febrile neutropenia or likely bacterial infection between patients receiving levofloxacin and those receiving other prophylaxis regimens (P = .52 and P = .36, respectively).

the risk of confounding by these stewardship interventions. In the possibility of neutropenia as an adverse effect of fluoro-
2015, a 2-month nationwide shortage of cefepime necessitated quinolones [10]. A related consideration is that infection-re-
a temporary modification of institutional guidelines to include lated mortality during leukemia therapy is more common
ceftazidime plus vancomycin as first-line therapy for febrile in low-income countries and malnourished patients, so our
neutropenia. Accordingly, ceftazidime and cefepime exposure findings may not be generalizable to those settings [10].
were merged in the final analysis to avoid a spurious reduction The implementation of antibacterial prophylaxis has
in cefepime exposure during the shortage. Despite these efforts been limited by concerns regarding possible adverse conse-
to address and ameliorate the effects of identified variables, quences, especially the development of antibiotic resistance
confounding of the study by unidentified variables remains [37]. Although antibiotic-resistant infections did not signifi-
possible. cantly increase in this study, the sample size and time period
The prolonged neutropenia in patients receiving proph- were inadequate to measure the medium- or long-term risk.
ylaxis might be related to the indication for prophylaxis Studies of antibacterial resistance after levofloxacin proph-
and was, therefore, included in the multivariate analysis. ylaxis have yielded mixed results. A meta-analysis found
However, a lower neutrophil nadir was seen in an Indonesian no significant increase in quinolone-resistant infection
study of children receiving ciprofloxacin prophylaxis, raising (relative risk, 1.2; 95% confidence interval, .8–1.7), but the

1796 • CID 2017:65 (1 December) • Wolf et al

Table 5. Propensity Score–Weighted Multivariate Logistic Regression Analysis for Effectiveness of Levofloxacin Prophylaxis

Outcome Crude OR (95% CI) P Value Adjusted OR (95% CI)a P Value

Levofloxacin vs no prophylaxis
Febrile neutropenia 0.43 (.24–.76) .004a 0.28 (.15–.52) <.001b
b
Febrile neutropenia with CDI 0.33 (.11–.99) .048 0.25 (.09–.69) .007b
Febrile neutropenia with MDI 0.44 (.21–.93) .03b 0.34 (.17–.71) .004b
CDI 0.54 (.27–1.07) .08 0.48 (.26–.91) .02b
MDI 0.53 (.29–.99) .045b 0.49 (.28–.88) .02b
BSI 0.50 (.16–1.52) .22 0.42 (.15–1.16) .09
Clostridium difficile infectionc 0.06 (<.01 to .48) .003b 0.03 (<.01 to .24) <.001b
Likely bacterial infection 0.32 (.16–.66) .002b 0.24 (.12–.48) <.001b
b
Any enterocolitis 0.27 (.08–.92) .04 0.23 (.08–.67) .008b
Levofloxacin vs other prophylaxis
Febrile neutropenia 0.83 (.45–1.54) .56 1.17 (.64–2.14) .60
Febrile neutropenia with CDI 0.64 (.19–2.15) .47 0.74 (.25–2.19) .59
Febrile neutropenia with MDI 1.16 (.48–2.82) .74 1.68 (.74–3.84) .21
CDI 0.75 (.35–1.61) .47 0.85 (.44–1.65) .63
MDI 1.09 (.54–2.19) .82 1.41 (.76–2.63) .27
BSI 1.19 (.31–4.61) .80 1.85 (.54–6.35) .33
C. difficile infectionc 0.07 (<.01 to .57) .008b 0.04 (<.01 to .36) <.001b
Likely bacterial infection 0.69 (.31–1.53) .36 0.85 (.41–1.74) .65
Any enterocolitis 0.34 (.09–1.26) .11 0.38 (.12–1.14) .08

Abbreviations: BSI, bloodstream infection; CDI, clinically documented infection; CI, confidence interval; MDI, microbiologically documented infection; OR, odds ratio.
a
The propensity score of prophylaxis groups was estimated using a multinomial logistic regression model with age, sex, race, Down syndrome, and leukemia type as predictors; then a
logistic regression model was used to model prophylaxis groups and exposure to profound neutropenia during induction as predictors of clinical outcomes, with inverse probability weighting
using the propensity score.
b
Significant difference (P < .05).
c
Also adjusted for exposure to meropenem.

follow-up period in these studies was insufficient to address Supplementary Data

the question of long-term resistance [6]. Other groups have Supplementary materials are available at Clinical Infectious Diseases online.
Consisting of data provided by the authors to benefit the reader, the posted
reported increased resistance, including cross-resistance,
materials are not copyedited and are the sole responsibility of the authors,
after extended use of quinolones [38, 39], although this did so questions or comments should be addressed to the corresponding author.
not necessarily negate the benefit of prophylaxis [38]. In the
current study, exposure to antipseudomonal β-lactam anti- Notes
biotics, aminoglycosides, and vancomycin was reduced. This Acknowledgments. We thank Keith A. Laycock, PhD, ELS, Elaine
Tuomanen, MD, Jose Ferrolino MD, and Marnie Dorsey for support in the
suggests that fluoroquinolone prophylaxis might shift antibi- preparation of the manuscript.
otic use away from agents less crucial for treating infection, Financial support. This work was supported by the National Institutes
thereby balancing the overall increase in exposure. Large- of Health (grants CA21765, CA36401-26S1, and GM92666) and by ALSAC.
Potential conflicts of interest. All authors: No reported conflicts of
scale long-term investigations are needed to assess this pros-
interest. All authors have submitted the ICMJE Form for Disclosure of
pect [39]. Potential Conflicts of Interest. Conflicts that the editors consider relevant to
Ours is the first study to examine the effects of primary anti- the content of the manuscript have been disclosed.
bacterial prophylaxis with levofloxacin on serious infections References
and antibiotic exposure in children undergoing induction ther- 1. Christensen MS, Heyman M, Mottonen M, et al. Treatment-related death in
childhood acute lymphoblastic leukaemia in the Nordic countries: 1992–2001. Br
apy for ALL. Although all prophylaxis regimens, including lev- J Haematol 2005; 131:50–8.
ofloxacin, reduced the risk of febrile neutropenia and systemic 2. Inaba H, Pei D, Wolf J, et al. Infection-related complications during treatment for
childhood acute lymphoblastic leukemia. Ann Oncol 2017; 28:386–92.
infection, levofloxacin prophylaxis also shifted antibiotic use
3. O’Connor D, Bate J, Wade R, et al. Infection-related mortality in children with
away from agents typically used to treat infection and dramat- acute lymphoblastic leukemia: an analysis of infectious deaths on UKALL2003.
ically reduced the risk of enterocolitis and C. difficile infection. Blood 2014; 124:1056–61.
4. Slats AM, Egeler RM, van der Does-van den Berg A, et al. Causes of death–other
There was no documented increase in breakthrough infections than progressive leukemia–in childhood acute lymphoblastic (ALL) and myeloid
with resistant organisms, but this remains possible. These find- leukemia (AML): the Dutch Childhood Oncology Group experience. Leukemia
2005; 19:537–44.
ings support the targeted use of levofloxacin prophylaxis in chil- 5. Fiser RT, West NK, Bush AJ, Sillos EM, Schmidt JE, Tamburro RF. Outcome of
dren and adolescents with ALL who are undergoing induction severe sepsis in pediatric oncology patients. Pediatr Crit Care Med 2005; 6:531–6.
6. Gafter-Gvili A, Fraser A, Paul M, et al. Antibiotic prophylaxis for bacterial
therapy, with close long-term monitoring of antibiotic resist- infections in afebrile neutropenic patients following chemotherapy. Cochrane
ance patterns. Database Syst Rev 2012; 1:CD004386.

Prophylaxis in Pediatric ALL • CID 2017:65 (1 December) • 1797

 7. Buffie CG, Bucci V, Stein RR, et al. Precision microbiome reconstitution restores 23. Sulis ML, Blonquist TM, Athale UH, et al. Effectiveness of antibacterial prophy-
bile acid mediated resistance to Clostridium difficile. Nature 2015; 517:205–8. laxis during induction chemotherapy in children with acute lymphoblastic leu-
8. Sammons JS, Localio R, Xiao R, Coffin SE, Zaoutis T. Clostridium difficile infec- kemia. Blood 2015; 126: 249.
tion is associated with increased risk of death and prolonged hospitalization in 24. Yeh TC, Liu HC, Hou JY, et al. Severe infections in children with acute leukemia
children. Clin Infect Dis 2013; 57:1–8. undergoing intensive chemotherapy can successfully be prevented by ciprofloxa-
9. Yousef AA, Fryer CJ, Chedid FD, Abbas AA, Felimban SK, Khattab TM. A pilot cin, voriconazole, or micafungin prophylaxis. Cancer 2014; 120:1255–62.
study of prophylactic ciprofloxacin during delayed intensification in children 25. Felsenstein S, Orgel E, Rushing T, Fu C, Hoffman JA. Clinical and microbiologic
with acute lymphoblastic leukemia. Pediatr Blood Cancer 2004; 43:637–43. outcomes of quinolone prophylaxis in children with acute myeloid leukemia.
10. Widjajanto PH, Sumadiono S, Cloos J, Purwanto I, Sutaryo S, Veerman AJ. Pediatr Infect Dis J 2015; 34:e78–84.
Randomized double blind trial of ciprofloxacin prophylaxis during induction 26. Rungoe C, Malchau EL, Larsen LN, Schroeder H. Infections during induction
treatment in childhood acute lymphoblastic leukemia in the WK-ALL protocol therapy for children with acute lymphoblastic leukemia. the role of sulfameth-
in Indonesia. J Blood Med 2013; 4:1–9. oxazole-trimethoprim (SMX-TMP) prophylaxis. Pediatr Blood Cancer 2010;
11. Pui CH, Campana D, Pei D, et al. Treating childhood acute lymphoblastic leu- 55:304–8.
kemia without cranial irradiation. N Engl J Med 2009; 360:2730–41. 27. McCullers JA, Williams BF, Wu S, et al. Healthcare-associated infections at a chil-
12. Freifeld AG, Bow EJ, Sepkowitz KA, et al. Clinical practice guideline for the use dren’s cancer hospital, 1983–2008. J Pediatric Infect Dis Soc 2012; 1:26–34.
of antimicrobial agents in neutropenic patients with cancer: 2010 update by the 28. Pui CH, Evans WE. A 50-year journey to cure childhood acute lymphoblastic
Infectious Diseases Society of America. Clin Infect Dis 2011; 52:e56-93. leukemia. Semin Hematol 2013; 50:185–96.
13. Lehrnbecher T, Phillips R, Alexander S, et al; International Pediatric Fever and 29. Khurana M, Lee B, Feusner JH. Fever at diagnosis of pediatric acute lympho-
Neutropenia Guideline Panel. Guideline for the management of fever and neutro- blastic leukemia: are antibiotics really necessary? J Pediatr Hematol Oncol 2015;
penia in children with cancer and/or undergoing hematopoietic stem-cell trans- 37:498–501.
plantation. J Clin Oncol 2012; 30:4427–38. 30. Fisher BT, Sammons JS, Li Y, et al. Variation in risk of hospital-onset Clostridium
14. National Cancer Institute. Common terminology criteria for adverse events v3.0 difficile infection across β-lactam antibiotics in children with new-onset acute
(CTCAE). NCI, NIH, DHHS. August 9, 2006. lymphoblastic leukemia. J Pediatric Infect Dis Soc 2014; 3:329–35.
15. Haeusler GM, Phillips RS, Lehrnbecher T, Thursky KA, Sung L, Ammann RA. Core 31. Sandora TJ, Fung M, Flaherty K, et al. Epidemiology and risk factors for
outcomes and definitions for pediatric fever and neutropenia research: a consensus Clostridium difficile infection in children. Pediatr Infect Dis J 2011; 30:580–4.
statement from an international panel. Pediatr Blood Cancer 2015; 62:483–9. 32. Stevens V, Dumyati G, Fine LS, Fisher SG, van Wijngaarden E. Cumulative antibi-
16. Santolaya ME, Alvarez AM, Becker A, et al. Prospective, multicenter evaluation otic exposures over time and the risk of Clostridium difficile infection. Clin Infect
of risk factors associated with invasive bacterial infection in children with cancer, Dis 2011; 53:42–8.
neutropenia, and fever. J Clin Oncol 2001; 19:3415–21. 33. Tai E, Richardson LC, Townsend J, Howard E, Mcdonald LC. Clostridium difficile
17. Centers for Disease Control and Prevention. Bloodstream infection event (cen- infection among children with cancer. Pediatr Infect Dis J 2011; 30:610–2.
tral line-associated bloodstream infection and non-central line-associated 34. de Blank P, Zaoutis T, Fisher B, Troxel A, Kim J, Aplenc R. Trends in Clostridium
bloodstream infection). Available at: http://www.cdc.gov/nhsn/PDFs/pscManu- difficile infection and risk factors for hospital acquisition of Clostridium difficile
al/4PSC_CLABScurrent.pdf. Accessed 26 April 2017. among children with cancer. J Pediatr 2013; 163:699–705 e1.
18. Centers for Disease Control and Prevention. NHSN organism list (all organ- 35. Dingle KE, Didelot X, Quan TP, et al; Modernising Medical Microbiology
isms, top organisms, common commensals, MBI organisms, and UTI bacteria). Informatics Group. Effects of control interventions on Clostridium difficile infec-
Available at: http://www.cdc.gov/nhsn/XLS/master-organism-Com-Commen- tion in England: an observational study. Lancet Infect Dis 2017; 17:411–21.
sals-Lists.xlsx. Accessed 26 April 2017. 36. Sung L, Aplenc R, Alonzo TA, Gerbing RB, Lehrnbecher T, Gamis AS.
19. Goldstein B, Giroir B, Randolph A; International Consensus Conference on Effectiveness of supportive care measures to reduce infections in pediatric AML:
Pediatric Sepsis. International pediatric sepsis consensus conference: definitions a report from the Children’s Oncology Group. Blood 2013; 121:3573–7.
for sepsis and organ dysfunction in pediatrics. Pediatr Crit Care Med 2005; 6:2–8. 37. Haeusler GM, Slavin MA. Fluoroquinolone prophylaxis: worth the cost? Leuk
20. Heinze G, Schemper M. A solution to the problem of separation in logistic regres- Lymphoma 2013; 54:677–8.
sion. Stat Med 2002; 21:2409–19. 38. Kern WV, Klose K, Jellen-Ritter AS, et al. Fluoroquinolone resistance of
21. Cruciani M, Concia E, Navarra A, et al. Prophylactic co-trimoxazole versus norfloxa- Escherichia coli at a cancer center: epidemiologic evolution and effects of discon-
cin in neutropenic children–perspective randomized study. Infection 1989; 17:65–9. tinuing prophylactic fluoroquinolone use in neutropenic patients with leukemia.
22. Laoprasopwattana K, Khwanna T, Suwankeeree P, Sujjanunt T, Tunyapanit W, Eur J Clin Microbiol Infect Dis 2005; 24:111–8.
Chelae S. Ciprofloxacin reduces occurrence of fever in children with acute leu- 39. Macesic N, Morrissey CO, Cheng AC, Spencer A, Peleg AY. Changing microbial
kemia who develop neutropenia during chemotherapy. Pediatr Infect Dis J 2013; epidemiology in hematopoietic stem cell transplant recipients: increasing resist-
32:e94–8. ance over a 9-year period. Transpl Infect Dis 2014; 16:887–96.

1798 • CID 2017:65 (1 December) • Wolf et al