A multicenter prospective study of 515 febrile neutropenia episodes in Argentina during a 5-year period

For better management of patients with febrile neutropenia, our study investigated the epidemiologic, microbiologic, and clinical characteristics of adult inpatients with febrile neutropenia and their mortality-associated factors. To this end, we carried out a prospective, observational, multicenter study in 28 Argentinian hospitals between 2007 and 2012. We included 515 episodes of febrile neutropenia from 346 patients, median age 49 years. Neutropenia followed chemotherapy in 77% of cases, half of the cases due to hematological malignancies. Most episodes were classified as high-risk according to MASCC criteria, and 53.6% of patients were already hospitalized at the onset of febrile neutropenia. Bloodstream infections were detected in 14% episodes; whereas an infectious source of fever was identified in 80% of cases. Mortality rate achieved to 14.95%. The binary regression analysis showed that persistence of fever at day 7, or neutropenia at day 14, dehydration and tachycardia at the onset of febrile neutropenia as well as prior infections were significantly associated with mortality. In addition to expanding our current knowledge on the features of adult patients with febrile neutropenia, present findings provide useful information for better management of them in Argentina, given the appropriate representativeness of centers participating in the study.


Introduction
Febrile neutropenia constitutes a differentiated clinical entity that is becoming increasingly frequent in daily hospital practice. The leading cause of neutropenia results from chemotherapy in patients with malignancies, making them more susceptible to bacterial infections because of the reduction of this critical component from innate defense mechanisms [1,2]. In turn, chemotherapy damages the integrity of the gastrointestinal mucosa, favoring the invasion a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 of endogenous germs [1,[3][4][5][6]. Since the ability to develop an adequate inflammatory response is impaired, the spectrum of clinical manifestations allowing to suspect an infectious process is also compromised, with fever being frequently the sole indicator of that [1,5,7,8]. The progression of infection in neutropenic patients may be fast, for which the early administration of empirical treatment is crucial, even before confirming any infectious source [1,5,8].
Current advances in terms of disease physiopathology, new imaging-based diagnosis, as well as microbiological and serological tests, resulted in better control measures and decreased morbimortality [5]. Nevertheless, case numbers continue to be significant, making the issue still relevant in terms of a health problem [1,9,10]. On the other hand, changes in the profile of the etiological agents together with the growing antimicrobial resistance renders the situation even more complex [5,6,8,10].
It follows that the management of febrile neutropenic patients remains challenging [1,5,10]. International and regional guidelines developed for the management of these patients, also establish the need for a proper epidemiological knowledge from different regional scenarios [5,6,8,10].
Given this background, we sought to identify the epidemiological, microbiological, clinical, and therapeutic characteristics of patients with febrile neutropenia from centers mostly located in the central area of Argentina. In a second step, we also searched for factors likely to be associated with fatal outcome.

Participants
All participants provided written and informed consent for the study, which was approved by the Ethic Committee from the School of Medical Sciencies National University of Rosario on July 2007 (resolution # 326/2007) in accordance with the World Medical Association's Declaration of Helsinki.
The study consisted of a multicenter, observational and prospective study of febrile neutropenia episodes in adult inpatients from 01-Aug-2007 to 02-Mar-2012, in 28 Hospital Departments of Internal Medicine from Argentina, led by investigators from "Hospital Provincial del Centenario" University Hospital. All participating centers habitually treat patients with febrile neutropenia episodes (FNEs) according to periodically revised local protocols, following national and international guidelines and recommendations. Inclusion criteria comprised age �15 years, along with the presence of neutropenia and fever according to the following definitions: for neutropenia a neutrophil count below 500 / mm 3 ; or less than 1,000 / mm 3 and likely to decrease to less than 500 / mm 3 in the next 48 hours; in the case of fever, presence of a single record of axillary temperature > 38.3˚C, or � 38˚C for one hour or more.
Day 0 corresponded to the time when the patient fulfilled inclusion criteria and was included in the study. The existence of an infective source was established provided the clinician made the diagnosis of a clinically documented infection based on clinical and imaging data. A microbiological infection was confirmed, whenever a microorganism was isolated in cultured samples from a clinically suspected source.
Criteria for study completion comprised the remission of fever and neutropenia. Information on whether the patient remained in the hospital for other reason or had been released from it was recorded at that time point. Death of the patient or the referral to another care center was also regarded as end of study.
For the remission of fever, we considered values lower than 38˚C for the latest 48 hours; whereas values higher than 500 neutrophils / mm 3 for at least during the two preceding days were regarded as neutrophil recovery.

Procedures
Evaluation of patients consisted of a complete physical examination, biochemistry and hematology studies, sets of blood cultures, or additional samples from infected sites if needed. Recorded data included demographic and epidemiologic variables, cause of neutropenia, i.e., cancer chemotherapy, underlying disease, and comorbidities. Information on patient antecedents before day 0 included: ambulatory or hospitalized condition at the onset of FNEs, prior use of antibiotics, presence of previous fungal infection, or neutropenia episodes. Data from physical examination, routine laboratory tests on admission, blood and urine cultures, use of a central venous catheter (CVC), empiric antibiotic treatment, administration of granulocyte colony-stimulating factor (G-CSF), detection of clinical or microbiological infection, evolution and outcome (death, recovery or referral) were also recorded. The clinical source of fever was registered by the medical staff according to its location: lung, skin, gastrointestinal tract, urinary, and central nervous system, among others. The biochemistry and microbiological studies were carried out at each center according to standard procedures.
The MASCC risk index (Multinational Association for Supportive Care in Cancer) [11] was estimated on admission, as did the ECOG (Eastern Cooperative Oncology Group) performance score [12,13]. The MASCC risk index developed and validated to identify febrile neutropenic patients with different risk of complications and death [8,[14][15][16], served to separate individuals into low or high-risk groups (values � or < 21 points, respectively) [11].
The ECOG scale, which is widely used to assess the functional status of cancer patients, mainly their daily life ability and autonomy, ranged from 0 (patients with autonomy and normal activity) to 4 (bed-ridden patients at high risk of complications during febrile neutropenia) [13,17].
As stated, all patients or their legal representatives provided written informed consent before enrolment in the study.

Statistical analysis
Quantitative results were expressed as median and 25-75 percentiles when required. Univariate analysis was performed by using the Chi-square test or Fisher's exact test for categorical variables, whereas the Mann-Whitney test was employed for continuous variables (survivors versus non-survivors comparisons). Variables yielding significant differences in univariate analysis were entered into a binary logistic regression model for the multivariate approach. The level of statistical significance was set at p <0.05. Data were analyzed by employing the SPSS package.

Baseline characteristics of the sample population
A total of 346 patients, experiencing 515 FNEs were enrolled. Age at the FNE episode was 49 (34-60) years, distributed among 227 women and 288 men. Nearly half of episodes (46%) occurred in the range age of 40 to 64 years. Data about patient characteristics on study admission (day 0) are depicted in Table 1. As shown, 53.6% of patients were already hospitalized at the onset of the episode, whereas one-third of them had the antecedent of prior FNEs or were undergoing antibiotic treatment before developing neutropenia. Among those receiving prophylaxis (n = 63, 12.2%), regimens included cotrimoxazole (n = 37, 7.2%), quinolones (n = 19, 3.7)%, or other compounds (n = 7, 1.3%).
The main comorbidities were HIV infection, treatment with immunosuppressants or corticosteroids, and diabetes. The MASCC risk index corresponded to the high-risk group (<21) in 95.3% of cases, while 47.4% of patients presented a �2 ECOG score. The median neutrophil counts by total episodes was 100 (0-348) cells/mm 3 , with 242 FNEs (47%) showing profound neutropenia (�100 neutrophils / mm3). Tachycardia and tachypnea, followed by hypotension, dehydration, and oliguria were the most frequent alterations seen on physical examination at the onset of FNEs (Table 1).
Most FNEs (395) were secondary to chemotherapy, 291 of them due to oncohematological malignancies, i.e., non-Hodgkin's lymphoma (NHL), acute myeloid leukemia (AML), acute lymphocytic leukemia (ALL), and Hodgkin lymphoma. Another 104 episodes (20.2%) resulting from chemotherapy, had a diagnosis of solid tumors, such as breast, lung, and colorectal cancer, in decreasing order. In 120 episodes (23.3%), neutropenia was unrelated to chemotherapy, and due to the presence of primary hematological disorders, infections, drugs or autoimmune diseases. In 14 cases, neutropenia causes remained unidentified ( Table 2). Blood cultures performed on day 0 were positive in 72 of FNEs (14%), showing a slight predominance of Gram-negative bacilli (52.8%). The main isolated microorganism was Escherichia coli, followed by methicillin-sensitive Staphylococcus aureus, Klebsiella pneumoniae, and pseudomonas species (Table 3).
When analyzing the association between antimicrobial prophylaxis and the type of bacteria recovered from blood cultures, we found an insignificant trend for Gram-positive predominance among patients undergoing prophylaxis (5/8 Gram-positive bacteremias from patients under prophylaxis, versus 29/64 from those with no prophylaxis, p = 0.463). The same trend was recorded when analyzing prophylaxis according to cotrimoxazole and quinolones treatments (data not shown).
Throughout the evolution of FNEs, one to three additional blood cultures were taken in 312 cases, 33 of them yielding positive results (10.6%). Again, there was a small predominance of Gram-negative bacilli, with K. pneumoniae being the most frequently isolated one (Table 3). Urine cultures performed in all FNEs yielded positive results in 49 cases (9.5%); as did another 13 cases during their evolution. In both occasions, E. coli predominated (Table 3).
Mycological blood cultures performed at day 0 (n = 153) were all negatives, whereas in another 90 samples collected during evolution, the presence of Candida albicans was found in 2 out of them (2.2%).
Stoll cultures were carried out as required. Among 103 stoll cultures performed at the onset of FNE, 3 of them led to pathogen isolation, which may be incriminated as being possibly involved in diahrrea (Table 3).

Confirmed infection
In overall, an infectious cause was identified in 80% of FNEs. More than 68% of FNEs (n = 355) had a diagnosis of an infectious process, based on clinical data (n = 235, 46%) or clinical and microbiological evidence (n = 120, 23%). There were another 58 cases (11%) with a positive microbiological result but no clinical focus. In the remaining 102 episodes, clinical or microbiological evidence of infection was lacking.

Treatment
Two third of FNEs were given G-CSF (n = 341). Information dealing with empiric antibiotic therapy on day 0 is provided in Table 4. The prevailing antibiotic schedule consisted of the combination of ceftazidime plus amikacin, followed by piperacillin/tazobactam and Table 3. Results from cultures performed at day 0 and during the evolution of febrile neutropenia episodes.

Blood cultures Isolated microbe (n) n
Blood cultures at the onset (n = 515)

Total 33
Urine culture Isolated microbe (n) Urine cultures at the onset (n = 515)

Total 13
Stoll cultures at the onset (n = 103)

None
Microorganisms were indicated according to decreasing order of presentation
On crude analysis, MASCC risk index category was associated with mortality (Table 5). Other variables associated with mortality were: hospitalization at the onset of FNE, presence of CVC at presentation, treated infection before the onset FNEs, hypotension/tachycardia/ tachypnea/dehydration at the onset of FNE, initial use of vancomycin, unremitted fever at day 7 or unremitted neutropenia at day 14, as well as initial positive blood cultures, an ECOG score � 3,or MASCC risk index <15, and neutrophils <50 cells/mm 3 on day 0. Female gender, chemotherapy for a solid tumor cause of FNEs, and initial empirical treatment with ceftazidime plus amikacin showed a protective relationship with mortality on crude analysis. Variables that continued to show mortality-associated statistical significance, upon multivariate analysis, were the presence of dehydration and tachycardia, the existence of a treated infection before the onset FNEs, along with unremitted fever or neutropenia at days 7 and 14, respectively (Table 6). After removing non-cancer patients, results did not deviate from the ones recorded in the whole sample, except for tachycardia, that lost statistical significance upon multivariate analysis (S1 Table).

Discussion
According to study purposes, we herein report on the epidemiological, clinical, and microbiological features of patients with FNEs, their therapeutic management, and disease outcome. To our knowledge it constitutes the first report in terms of its multicentric and prospective nature in the region, enrolling a substantial number of FNEs.
Many published studies were based on retrospective data from patients enrolled in cancer clinical trials [4,6,9,10,18]. Some prospective studies in adults include only a few centers, with a smaller number of patients [19][20][21][22][23]. Other researchers, exclusively concentrate on postchemotherapy neutropenia [13,24,25], only evaluate the first episode of febrile neutropenia [17], the ones occurring in the context of a new indication of chemotherapy [26], or they do not report further episodes occurring in the same patient [11].
Our approach of including all FNEs of any etiology, in 346 relatively young patients is closer to the real clinical scenario. In line with other studies [1,8,11,19,21,27], most episodes (77%) were secondary to chemotherapy, although 106 cases of neutropenia of different etiologies were recorded. Also, more than half of patients were already hospitalized at the onset of FNEs, a situation known to confer a higher risk of complications and poor prognosis [11,17,25,28], as did in the present series, although differences became insignificant upon multivariate analysis. Nearly 20% of cases were already receiving anti-infectious treatment at the onset of FNE, with this variable behaving as an independent mortality predictor.
According to the MASCC risk index [11], nearly all FNEs (95.3%) fell within the high-risk category, which is greater than values found in the studies from by Jin et al. (23.6%) [20] or Rabagliati et al. (41%) [22]. Reduced neutrophil counts at day 0, was found to be compatible with a poor prognosis of patients [5]. In the present series, a neutrophil count lower than 50 cells/mm 3 was also related to a bad prognosis on crude analysis, although differences became statistical insignificant upon multivariate analysis.
While in only 20% of episodes fever had an unknown origin, in the remaining ones, an infectious process was detected based on clinical (46%), microbiological (11%) or combined (23%) criteria. Such amount of recognized infections is higher than findings from other series, although more recent studies show values above figures recorded in earlier studies [8, 21, 27,  29]. The increased presence of infections may be due to the availability of better diagnostic methods, although it cannot be ruled out that single clinical criteria may have led to an overestimation of it. Overall, the distribution of affected sites agrees with findings from other studies of FNEs, that is pneumonia, the involvement of the gastrointestinal tract, skin and soft tissues, and the oral cavity [1,5,8,21]. As well as in other reports, bacteremia was present between 10 and 37% of blood samples [5,8,24], 15% of present FNEs had positive blood cultures. In a study carried out in Buenos Aires [19], 36.9% of blood cultures yielded positive results, like two Chilean studies reporting 30.5% and 31.4% positivity [21,22]. Our lower bacteremia may be due to that more than 30% of these patients were under antimicrobial treatment before the onset of FNEs, either for prophylactic reasons or the existence of a prior infection. We only found an insignificant trend for Gram-positive bacteriemia in cases under antibiotic prophylaxis, probably due to its limited use in our patient series. Following the Gram-negative to Gram-positive shift in pathogen predominance seen worldwide during the 1980s, Gram-positive bacteria became predominant in most industrialized regions, with less developed countries still showing a Gram-negative bacteria predominance, possibly related to lower antibiotic prophylaxis [30][31][32][33][34][35]. However, in the last decade, most centers report similar Gram-positive and Gram-negative bacteremia rates in patients with FNEs [6,36]. In agreement with studies from the region [21] or Southeast Asia [20], our series also revealed a slight predominance of Gram-negative bacteria (52.8%). As seen in former reports [1, 5, 20-22, 27, 37, 38] we also found a predominance of E. coli, K. pneumoniae and pseudomonas species. Within Gram-positive bacteria, methicillin-susceptible Staphylococcus aureus and methicillin-resistant Staphylococcus aureus followed by coagulase-negative staphylococci were the prevailing forms. The latter one was also found to predominate among Gram-positive cocci [20][21][22]37].
The fact that G-CSF was indicated in 66% of the FNEs suggests some overuse, as current guidelines mostly recommend its administration before the establishment of FNEs in patients at high risk [8,36,[39][40][41]. However, recent guideline recommendations indicate that G-CSF should be considered in patients with fever and neutropenia who are at high risk for infectionassociated complications or have prognostic factors predictive of poor clinical outcomes [40].
The commonest antibiotic schedule employed in this patient series consisted of ceftazidime plus amikacin, followed by piperacillin/tazobactam, or carbapenems and cefepime as monotherapy. The latest guidelines from the Argentine Society of Infectious Diseases [5] and the Infectious Diseases Society of America [8] and meta-analytical approaches [42] regard ceftazidime plus amikacin as a second choice in favor of the abovementioned alternatives. Most cases from the present studies were included before the availability of these guidelines [5,8] when ceftazidime plus amikacin were recommended as the initial empirical treatment [3,43,44]. Nevertheless, since evidence is not conclusive in this regard [45], these therapeutic schedules continue to be adequate for the management of FNEs, with other recent guidelines still including ceftazidime as first-line therapy [36,46,47]. In our hands, although initial treatment with ceftazidime plus amikacin was associated with lower mortality -probably because such treatment was not of first choice for severe patients-, further multivariate analysis revealed no differences in mortality depending on the initial antimicrobial therapy.
The fact that vancomycin was indicated in more than half of FNEs raises the possibility of some overuse. Vancomycin has precise indications in this kind of patients [5,8], for which institutional surveillance is currently being implemented to establish more accurate handling of it. In the same sense, the use of antifungal therapy (35.1% of episodes) seems high, given the lower amount of cases with confirmed invasive mycosis. The fact that antifungal therapy does not require mycological confirmation [5,8,48,49], and patients presented prolonged and severe neutropenia mostly linked to oncohematological malignancies, may have accounted for such indication.
In line with other studies [3,41], the mean recovery time of neutrophils situated between 6 and 8 days. Results also revealed that the longer the time for neutrophil recovering (over 14 days in our series) the poorer the prognosis, as seen in former reports [37,43].
On admission, patients fit well with the high-risk category (95.3%), showing a mortality rate (14.9%) like the one recorded in a study of comparable risk [22]. Mortality rate increase with the MASCC prognostic index, in agreement with other studies showing mortality as low as 3% if the MASCC score was >21, further reaching to 36% provided the MASCC score was <15 [10]. As stated, patients were at high risk of complications, i.e., increased MASCC risk index and ECOG scores, low neutrophil counts, prolonged FNEs, many already hospitalized patients with previous FNEs, Gram-negative bacteremia, as well as a high occurrence of hematological neoplasms, and lung involvement. According to multivariate analysis, mortality-associated variables were: unremitted fever at day 7, unremitted neutropenia at day 14, tachycardia, dehydration, and previous infections. The lost of MASCC statistical significance following multivariate analysis may be related to the fact that some of the adjusting variables were also employed for the score construction. As stated, analysis among oncologic patients (the MASCC score was originally developed for them), revealed a quite similar pattern of mortality association. Except for tachycardia which was only associated with mortality in the multivariate analysis from the whole patient sample. The lower number of FNEs among oncologic patients may account for such lost of statistical association.
The study by Gómez Roca et al. [19] identified the presence of comorbidities, tachypnea, systolic hypotension, and a clinical infectious process as factors of poor prognosis.
This type of observational study has limitations. For instance, there existed variations in causes of febrile neutropenia, and features of centers in charge of patients. Management was decided at each center according to its usual practice, likely contributing to some therapeutic dissimilarities. In the same sense, surveillance for multidrug-resistant bacteria was not usually carried out in many centers, for which data regarding this issue is lacking, as did information about Clostridium difficile toxin. Also, there was no information on lifestyle parameters, primary healthcare data, or socioeconomic data, emerging as potential confounding factors.
Against such constraints, our study did collect data about underlying diseases which may influence study endpoints and revealed a consistent and relatively high association, likely to reduce the risk of unknown confounders. These facts, along with the proper representativeness of centers allowing to get an informative number of cases strengthen study validity serving to expand present knowledge on the clinical and microbiological features of patients with febrile neutropenia in Argentina. Present kind of observational data also help to identify associations between exposures and outcomes, and the strength of such associations, providing useful information for clinical decisions and better management of FNEs in the region.
Supporting information S1 Table. Variables associated with mortality during febrile neutropenia episodes excluding non-cancer patients. (DOCX) S1 File. Five hundred and fifteen febrile neutropenia episodes in Argentina during a 5-year period database. (SAV)