Evaluation of six risk factors for the development of bacteremia in children with cancer and febrile neutropenia

Febrile neutropenia is a well-known entity in children with cancer, being responsible for the high risk for infection that characterizes this population. For this reason, cancer patients are hospitalized so that they can receive prophylactic care. Risk factors have been used to classify patients at a high risk for developing bacteremia. The present study evaluates whether those risk factors (C-reactive protein, hypotension, leukemia as the cancer type, thrombocytopenia, recent chemotherapy, and acute malnutrition) apply to patients at the Unidad Nacional de Oncología Pediátrica. We evaluated 102 episodes in 88 patients, in whom risk factors and blood cultures were tested. We observed no statistical relationship between the six risk factors and bacteremia. There was also no relationship between bacteremia and the simultaneous presence of two, three, or more risk factors. A significant relationship of C-reactive protein and platelet count with other outcome factors was observed.


INTRODUCTION
Children with cancer who develop febrile neutropenia are at high risk for severe infections such as bacteremia and sepsis.Children with cancer account for 12.8%-17.4% of severe sepsis in children, with an associated mortality rate of 5.5% (excluding patients receiving stem-cell transplantation) 1 .
The relationship between fever, neutropenia, bacteremia, and sepsis has been widely known for more than 40 years.Certain characteristics of the neutropenia episode, such as severity, cause, duration, and time of resolution, determine vulnerability of patients and outcome when they develop an acute infection 2,3 .Additionally, in patients with immune deficiencies, signs of infection are often blurred; many times, the first and only sign is fever 2 .Fever can be the result of many causes, but approximately 60% of the time, it is caused by bacterial infection with or without bacteremia 4 .The appearance of clinical signs in any kind of infection is closely associated with absolute neutrophil count: a level below 100 cells/mm 3 is strongly associated with a positive blood culture 1 .
In spite of numerous studies, knowledge of this subject is far from complete.Even the exact incidence of febrile neutropenia, especially after chemotherapy, is unknown.Studies of antibiotic prophylaxis during neutropenia suggest that this complication is not as frequent as was thought: 47% in children with tumours or leukemia, and 68% in children with acute leukemia 5 .In the population of the Unidad Nacional de Oncología Pediátrica (unop), we know that the incidence of neutropenia in children with acute lymphoblastic leukemia treated with 6-mercaptopurine and methotrexate as maintenance therapy is 3.5%, similar to the 4% rate seen in the international literature 6 .Because of high rates of bacterial infections, the proven good results with prophylactic antibiotics, and difficulty in assessing who is more at risk of severe infection, all patients with febrile neutropenia used to receive prompt in-hospital evaluation and broad-spectrum intravenous antibiotics until the resolution of the fever and maintenance of immune integrity 7 .This strategy has been very successful in reducing mortality and other sequelae of infectious complications.However, this aggressive approach poses risks such as side effects from antibiotics, resistant bacterial strains, fungal infections, and substantial economic and psychological effects for both the family and the unop.Recent evidence has also shown risk heterogeneity in this population, with half the population potentially benefiting from gentler management: earlier hospital discharge, shorter antibiotic course and use of oral medication, or ideally, outpatient treatment 4,7 .
In a review of 27 studies involving 5208 episodes of febrile neutropenia, 17 risk factors for the development of bacteremia were documented.However, no risk factor appeared to be specific to the pediatric Evaluation of six risk factors for the development of bacteremia in children with cancer and febrile neutropenia E.J. Asturias MD,* J.E. Corral MD, † and J. Quezada MD † population, and all had a high failure percentage for identifying low-risk patients: 7% in adults, and 12% in children 5 .Finally, after evaluating 447 episodes in children only, Santolaya et al. 4 identified 5 independent risk factors (in order of significance): presence of C-reactive protein (crp) 90 mg/L or more, hypotension, leukemia relapse as the cancer type, platelet count below 50,000/mm 3 , and interval of 7 days or fewer since the last chemotherapy.There was also evidence that other previously described risk factors such as fever severity and absolute monocyte count were not statistically relevant 3 .A year later, a prediction model for low or high risk of invasive bacterial infection was developed with good results: 92% sensitivity, 76% specificity, and 82% positive and 90% negative predictive value 7 .In addition to those 5 risk factors, any kind of malnutrition in a patient was identified (in a study conducted by Velásquez at unop) as being associated with a 1.25 probability of that patient experiencing at least 1 episode of neutropenia 6 .The Velásquez finding, and the fact that normal nutrition status is a determinant of a good immune response, suggested the inclusion of malnutrition as a 6th risk factor.
In the present study, our aim was to evaluate the association of the 5 previously described risk factors-and the less recognized malnutritionwith bacteremia in children with cancer and febrile neutropenia, so as to support selective treatment in those patients and to improve, to a small degree, their quality of life.

Study Population
Our study included children younger than 18 years of age who were hospitalized with fever and neutropenia at unop from April 8 to October 15, 2008.Children hospitalized for fewer than 48 hours, those who had received antibiotics 7 days before admission, and those who had undergone bone marrow transplantation were excluded.Informed consent was obtained from the parents.The study was approved by the unop ethics committee and the Universidad Francisco Marroquín research committee.

Study Design
In this prospective observational study, eligible children were evaluated by unop pediatricians at admission and daily during hospitalization.One of the investigators reviewed hospital admissions every 2 or 3 days and recruited patients who had fulfilled the criteria of fever (axillary temperature ≥ 38.5°C or 38°C in two different measurements more than 1 hour apart) and neutropenia (absolute neutrophil count ≤ 1000/mm 3 ).All children with febrile neutropenia underwent blood culture and crp evaluation at admission.Predefined risk factors for severe infection and sepsis were evaluated during hospitalization and at discharge, if available.Patients were considered bacteremic if their blood culture was positive for any pathogen except coagulase-negative staphylococci; two positive blood cultures were considered true bacteremia.Monitoring, antibiotic selection, and daily management were managed by unop medical personnel and staff.Antimicrobial agents were selected empirically based on the Infectious Diseases Society of America 2002 guidelines for the use of antimicrobial agents in neutropenic patients with cancer 8 .

Risk Factor Definitions
Serum crp was considered high at 96 mg/L or more.Hypotension was defined as a systolic or diastolic blood pressure below the fifth percentile for age and height.Any type of leukemia was considered a risk factor.Thrombocytopenia was considered a risk factor at a platelet count of 50,000/mm 3 or lower.Recent chemotherapy was considered a risk factor when any kind of chemotherapeutic agent had been administered 7 days or fewer before consultation.Malnutrition was defined as a weight-for-height Z score of -2.0 or lower.

Statistical Analysis
The 6 risk factors were evaluated individually in univariate analysis with bacteremia as the key outcome.All the analyses were performed using EpiInfo (version 3.5.1,2008: Centers for Disease Control and Prevention, Atlanta, GA, U.S.A.) and the Fisher exact test.Additional analyses were performed using that same test or a comparison of means by Student t-test or Mann-Whitney and Wilcoxon tests (depending on the assumption of homogeneous variances, using a Bartlett test for inequality of population variances).

RESULTS
During the 7-month period, 128 eligible patients with febrile neutropenia presented for consultation, 98 were approached for the study, and the parents of 88 consented to participation.Those 88 patients experienced 102 episodes of febrile neutropenia: 74 children (84%) had 1 episode, 11 (12%) had 2 episodes, and 2 (2%) had 3 episodes.During the analysis, 2 episodes were excluded because of errors in the inclusion criteria, and 4 others because of a lack of blood culture results, for a total of 96 evaluable bacteremia episodes.Of the patients involved in these episodes, 45% were male; ages ranged from 8 months to 18 years (mean age: 6.5 ± 4.4 years).
Table i shows the cancer types in the study population.The predominant type was acute lymphocytic leukemia, followed by acute myeloid leukemia.Nineteen patients had relapsing acute lymphocytic leukemia.In 4 patients, a typified cancer diagnosis was not present on admission.One patient had been diagnosed 5 years and 10 months earlier (mean time from diagnosis: 7.9 ± 12.1 months).None of the children had undergone bone marrow transplantation.
At admission, 22 patients (23%) showed clinical evidence of infection, the most common symptoms being diarrhea (27%), celulitis (18%), and otitis (14%).For initial antimicrobial treatment, the most frequently used agent was a cephalosporin, mainly ceftazidime, used in 89 episodes (in 67 episodes as monotherapy and in 15 episodes in combination with a second drug, mainly clindamycin or gentamicin).In 5 episodes, more than two drugs were required on admission.In the rest of the episodes, ciprofloxacin or clindamycin was used as the main antimicrobial therapy.Vancomycin was indicated in 15 episodes, but in only 3 episodes after admission.The mean antibiotic course was 12.7 ± 11.9 days (range: 3-64 days).
A bacteria was isolated in 11 episodes.Table ii shows the bacterial species found.
The risk factor of hypotension was present in none of the children.No specific risk factor was statistically predictive of bacteremia.Nor was the association of 2, 3, or more risk factors predictive of the presence of a positive blood culture at admission.In 8 patients, no risk factors were present, and in 2 patients, 5 risk factors were present.No child presented with all the risk factors (mean: 2.01 ± 1.07).Table iii summarizes the presence of the risk factors previously associated with bacteremia.Intensive care was required for 6 patients, and 5 patients died during the course of the study-2 from septic shock, and 1 each from herpetic encephalopathy, hepatic failure, and stroke.The mean crp level in the patients who died was 130.25 ± 123.05 mg/L; it was 73.41 ± 83.19 mg/L in the patients who survived (92 episodes).When the child who died of hepatic failure, a non-infectious diagnosis, was excluded from the analysis, a correlation was observed between crp level and mortality (t = 1.988, p < 0.05).
Children who experienced more days with fever during hospitalization had a higher crp level on admission (115.04 mg/L) than did those with fewer fever days (71.99 mg/L).
Children who were thrombocytopenic had a higher mean of hospitalization days and more time with fever during hospitalization than did those who presented without thrombocytopenia (p = 0.0018 and p = 0.026 respectively).Patients who had thrombocytopenia on admission (61 episodes) were hospitalized for a mean of 20 ± 24.6 days and experienced a mean of 6 ± 7.9 days with fever.Patients who did not have with thrombocytopenia on admission (29 episodes) were hospitalized for a mean of 9 ± 6.8 days and experienced fever for a mean of 3 ± 3.8 days.
We also observed that the mean platelet count differed between patients with or without bacteremia.The mean platelet count in bacteremic patients was 49,000 ± 65,000/mm 3 ; in non-bacteremic patients, it was 104,000 ± 134,000/mm 3 .This difference was not statistically significant (t = 1.336, p > 0.05).
Children with a positive blood culture did not have a lower absolute neutrophil count (p = 0.54).Severe neutropenia (an absolute neutrophil count of 500/ mm 3 or fewer) was not a risk factor for bacteremia; it was not a risk factor even at lower values of 100/mm 3 or fewer (p = 0.38 and p = 0.61 respectively).

DISCUSSION AND CONCLUSIONS
Recent evidence has suggested that 5 simple clinical and laboratory parameters assessed at admission can predict the presence of invasive bacterial infection in children with cancer and febrile neutropenia 4,7 .However, in the present study, neither those factors nor malnutrition showed an association with the presence of bacteremia in our population of children with fever and neutropenia.
Apart from those discouraging results, two variables-platelet count and crp level-are related to important aspects of evolution and prognosis in children with cancer presenting with febrile neutropenia.Increasing crp values were directly associated with the number of days with fever during hospitalization and with mortality from infection.A difference in the mean crp on admission between bacteremic and non-bacteremic patients was also observed.However, a cut-off value of 96 mg/L may not be applicable in all populations; predictability may vary with the risk of infection and the availability of health care.
Another consideration in our analysis is the exclusion of 1 febrile episode.This patient met the inclusion and exclusion criteria, but showed an extremely low crp level (13 mg/L) compared with others in the group.This finding may be explained by the fact that the patient had hepatic failure secondary to hepatoblastoma, and crp is synthesized in the liver 9 .That episode reduced the significance of the crp test; its exclusion validated the association (p = 0.049).
No clear association of thrombocytopenia (≤50,000 platelets/mm 3 ) with bacteremia emerged, but we observed a difference in the mean platelet count between bacteremic and non-bacteremic patients.That finding invites an assessment of platelet count cut-off values in our population.
With regard to the recent chemotherapy and the development of bacteremia, it is important to emphasize that numerous factors determine the timing and severity of febrile neutropenia after chemotherapy.The type and dose of chemotherapy determines the degree of bone marrow depletion.Agents such as nitrosoureas and mitomycin produce longer depletion.Host factors such as bone marrow cellularity and ability to metabolize chemotherapeutics determine the severity of febrile neutropenia 10 .
This observational study found a predominance of gram-negative bacterial isolates (64%).Most of the literature shows a preponderance of gram-positive bacteria isolated in these patients (58%-80%) [11][12][13][14][15] ; other studies suggest that gram-negative bacteremia may be on the increase in pediatric cancer patients 16- 18 .In developed countries, these changes may be related to the use of more-intensive chemotherapy regimens.In developing countries, gram-negative bacteremia has classically prevailed because of less use of prophylactic antibiotics and central lines 19 .
An important consideration is that bacteremia only, and not invasive bacterial infection (a positive bacterial culture from blood, urine, cerebrospinal fluid, or catheters), was the main outcome variable in this study.Other non-bloodstream infections are closely related to the evaluated risk factors, and their inclusion may validate the findings of Santolaya et al. 4,7 for our population.
One of the biggest limitations of this study is the absence of patients who presented with hypotension on admission.Even children who developed septic shock during hospitalization did not present with hypotension at admission, probably because hypotension is a late and severe sign in septic shock 21 .Clinical trials of patients with more advanced hemodynamic compromise are necessary to assess the predictive value of that variable.However, the most important limitation in our study was the small number of episodes registered (96 vs. 447 in Santolaya et al. 4 ).We recognize that larger studies have more analytic power, and these 6 factors may still be valid in our population, pending larger studies.
Finally, our study reinforces the need for studies in other populations of children with cancer and neutropenia to evaluate the power of risk factors that could predict which patients are at risk of severe invasive bacterial infection.Designing a practical and reliable classification system will give patients the benefits of a selective approach and may improve their quality of life.

ACKNOWLEDGMENT
This study was completely funded by unop, and none of the authors has any financial associations with any commercial company related to the study topic.

table iii
Summary of contingency tables crp = C-reactive protein.

table i
Cancer type at admission, by frequency