An Emergent Change in Epidemiologic and Microbiological Characteristics of Bloodstream Infections in Adults With Febrile Neutropenia Resulting From Chemotherapy for Acute Leukemia and Lymphoma at Reference Centers in Chile, Ecuador, and Peru

Abstract Background Febrile neutropenia is a life-threatening condition commonly observed in patients with hematologic malignancies. The aim of this article is to provide updated knowledge about bloodstream infections in febrile neutropenia episodes within the Andean region of Latin America. Method This retrospective study was based in 6 hospitals in Chile, Ecuador, and Peru and included adult patients with acute leukemia or lymphoma and febrile neutropenia between January 2019 and December 2020. Results Of the 416 febrile neutropenia episodes, 38.7% had a bloodstream infection, 86% of which were caused by gram-negative rods, with Klebsiella pneumoniae, Escherichia coli, and Pseudomonas aeruginosa being the most frequently identified bacteria. K pneumoniae isolates were more frequently resistant than E coli to cefotaxime (65% vs 39.6%), piperacillin-tazobactam (56.7% vs 27.1%), and imipenem (35% vs 2.1%) and were more frequently multidrug resistant (61.7% vs 12.5%). Among P aeruginosa, 26.7% were resistant to ceftazidime, piperacillin-tazobactam, and imipenem, and 23.3% were multidrug resistant. Overall 30-day mortality was 19.8%, being higher with vs without a bloodstream infection (26.7% vs 15.3%, P = .005). Fever duration was also significantly longer, as well as periods of neutropenia and length of hospital stay for patients with bloodstream infection. Additionally, the 30-day mortality rate was higher for episodes with inappropriate vs appropriate empirical antibiotic therapy (41.2% vs 26.6%, P = .139). Conclusions Considering the high rates of bacteria-resistant infection and 30-day mortality, it is imperative to establish strategies that reduce the frequency of bloodstream infections, increasing early identification of patients at higher risks of multidrug bacteria resistance, and updating existing empirical antibiotic recommendations.

Febrile neutropenia (FN) is a medical emergency frequently observed among patients with acute leukemia (AL) and lymphoma (L) after high-intensity myeloablative chemotherapy.It is associated with high morbidity and mortality rates, increasing costs, and the use of health care resources [1][2][3][4][5].
During FN episodes, bloodstream infections (BSIs) have been documented in 30% to 50% of cases, either by gram-negative rods (GNRs) or gram-positive cocci (GPC), with a predominance of one over the other changing over time [6].Patients with FN episodes require urgent empirical antibiotic therapy, based on international, national, or local guidelines [7,8].Patients with AL and L frequently receive antimicrobial prophylaxis, which increases selective pressure on them, leading to antibiotic resistance [9].
Antimicrobial resistance is a global health threat [10].In particular, GNRs have emerged as an important issue in different clinical settings [11], including hematologic cases [12].Several resistance mechanisms have been identified, such as extendedspectrum β-lactamases (ESBLs), amp-C β-lactamases, carbapenemases, among others, and very few antimicrobial therapy Bacteremia in Febrile Neutropenia • OFID • 1 Open Forum Infectious Diseases M A J O R A R T I C L E options are available [13].Additionally, greater bacterial resistance was accelerated during the COVID-19 pandemic, as were BSIs due to multidrug-resistant (MDR) bacteria at many health care centers, especially in the intensive care units [14,15].
Considering the aforementioned scenario, it is highly likely that bacterial resistance has increased among patients with AL and L in health care centers within the Andean region, as has been described for other countries in Asia, Europe, and North America and in a few Latin American countries [16][17][18].Consequently, some of the empirical antibiotic therapies recommended in current guidelines could no longer be appropriate [19], at least for a proportion of the neutropenic cases, putting patients at increased risks of missing days of adequate treatment with the threat of severe sepsis and death.There is a scarcity of published epidemiologic data from the Andean region of Latin America, and existing reports are out of date [20][21][22].Hence, there is an urgency to update corresponding epidemiologic information by center, country, and region.
The aim of this research was to offer updated knowledge about the epidemiology and clinical and microbiological characteristics of BSI in FN episodes observed in patients with AL and L in 6 selected centers within the Andean region.

Study Design
This retrospective study was conducted at the following 6 reference hospitals in main cities from 3 countries within the Andean region of Latin America: The study population included adults with AL or L (≥18 years old) who were receiving chemotherapy and experienced FN episodes between 1 January 2019 and 31 December 2020.Patients undergoing hematopoietic cell transplantation (HCT) were excluded.

Ethical Approvals
Before data collection began, the institutional review board at Pontificia Universidad Católica de Chile, the coordinating center, as well as all local institutional review boards from the participating centers approved the protocol and waived consent for enrollment considering the retrospective design.

Patient Consent
Our study did not include factors necessitating patient consent.

Procedures
Based on a survey conducted at the beginning of the study, it was determined that microbiological diagnostic capabilities, clinical approaches, and antibiotic availability for empiric therapy were similar among the 6 centers.However, differences were identified in the strategies of antibiotic-resistant bacteria colonization surveillance and antibiotic prophylaxis use.
The BSI diagnostic approach at each health care center included at least 2 sets of aerobic blood cultures.The susceptibility test was carried out via automated or manual methods.Regarding the automated approach, 3 hospitals used Vitek 2 (bioMérieux) and 1 used Phoenix (BD Diagnostics).With respect to the manual methods, 1 hospital used disks diffusion, and another used agar dilution.All the laboratories used techniques for ESBL detection, as well as for carbapenemases, such as Carba-NP, Blue Carba, and the modified carbapenem inactivation method.Immunochromatography tests were performed to identify carbapenemases.
Documenting clinical sources of infections was based on published definitions [23].In general, antibiotic management recommendations by FN international guidelines [7] were used as references at each center.The following were recorded and collected via a standardized form and the REDCap platform: demographic and clinical characteristics, main comorbidities, AL/L disease, bacterial colonization, antibiotic prophylaxis, BSI, antibiotic susceptibilities, clinical focus, therapy, length of hospital stay, fever and neutropenia duration, and mortality.

Febrile Neutropenia
FN was defined as an oral temperature >38.3 °C or 2 consecutive readings >38.0 °C for 2 hours and an absolute neutrophil count <500 cells/mm 3 or one expected to fall to <500 cells/ mm 3 during the next 48 hours [7,24].

Bloodstream Infection
BSI was defined as bacteria identification from blood cultures while the patient remained neutropenic in the same FN episode.Contamination was excluded, which was defined as common skin flora species recovered from a single blood culture bottle (eg, coagulase-negative Staphylococcus).Early BSI was defined when the event occurred from the first to fourth days of FN and late BSI when it occurred from the fifth day onward.

Bacterial Colonization
Bacterial colonization was defined as bacterial identification obtained from rectal swabs before the FN episode.

Multidrug Resistance
Multidrug resistance was defined as bacteria resistant to >1 antibiotic in >3 categories [25].

Extensive Drug Resistance
Extensive drug resistance was defined if the bacteria was checked for susceptibility to all relevant antibiotics and found nonsusceptible to ≥1 agent in all but <2 categories [25].

Inappropriate Antimicrobial Therapy.
Inappropriate antimicrobial therapy involved the absence of antimicrobial prescription to which the microorganism was susceptible or the use of an agent to which the organism was resistant [26].

Outcomes
The main clinical outcomes were days of fever, neutropenia, length of hospital stay, and 30-day mortality.

Statistical Analysis
Patient demographics and characteristics of the FN episodes are presented by count and percentage for categorical data and mean and SD or median and range for continuous data.
The chi-square or Fisher exact test was used for dichotomous variables and the t test for continuous variables.Statistical analysis was performed with Stata SE 15.0 software for Windows (StataCorp).P ≤ .05 was considered statistically significant.

FN Episodes
In total, 416 FN episodes were registered.As shown in Table 1, female gender and acute lymphatic leukemia were more frequent.The highest number of episodes was identified in a health center from Ecuador, followed by 1 in Peru and 1 in Chile.Some degree of mucositis was registered in 217 (52.4%)FN episodes, and patients had a central venous catheter in place in 136 (32.8%) episodes.Eighty-eight (21.1%) episodes had a known bacterial colonization: 67 (16.1%)GNR, 14 (3.4%)GPC, and 7 (1.9%)GNR and GPC.

Microbiological Identification
Regarding FN etiology, the cause of fever was identified in 330 (79.3%) episodes.A BSI was documented in 161 cases (38.7%), while a focus of infection was registered without a documented BSI in 152 episodes (36.5%).A bacterial non-BSI, viral, and/or fungal infection was identified in 17 (4.1%)episodes, and fever was of unknown origin in 86 (20.7%) cases.
A total of 198 bacteria were isolated from blood cultures, of which 108 (54.5%) were identified in early BSI and 90 (45.4%) in late BSI.As shown in Table 2, GNRs were more frequently identified than GPCs.Klebsiella pneumoniae, Escherichia coli, and Pseudomonas aeruginosa were the most frequent GNRs, while Staphylococcus epidermidis, Staphylococcus aureus, and Enterococcus faecium were the most frequent GPCs.
Figure 2 compares resistance rates of K pneumoniae, E coli, and P aeruginosa isolated in early vs late BSI, with a significant increase in E coli resistance to piperacillin-tazobactam and amikacin, as well as P aeruginosa resistance to piperacillintazobactam.A significant decrease in E coli ciprofloxacin resistance was also observed.
In relation to the 26 GPCs, 9 of 13 (69.2%)S epidermidis were methicillin resistant, and 2 of 4 (50%) E faecium were vancomycin resistant.On the contrary, all the S aureus were methicillin susceptible.

Outcomes
Clinical outcomes are presented in Table 3, showing statistically significant unfavorable results for BSI vs non-BSI episodes.In addition, it shows clinical outcomes in terms of the appropriateness of empirical antibiotic therapy for early BSI.Results were unfavorable for episodes involving inappropriate antimicrobial therapy, although without statistically significant differences.
We did not observe differences in the outcomes among those episodes receiving vs not receiving antibiotic prophylaxis.

DISCUSSION
This retrospective study revealed a predominance of GNR BSI in FN episodes among patients with AL and L from 6 reference hospitals in Chile, Ecuador, and Peru, with an overall high frequency of antibiotic resistance and high mortality rates.
Results indicated that 38.9% of FN episodes exhibited BSI.This is within the range described for hematologic cases and >14% reported by Parodi et al, according to a multicentric prospective study in Argentina focusing on FN in patients with hematologic and solid organ tumors [27].This can be explained by the characteristics of the patients in the current study, among which AL was the most frequent hematologic disease, given that its FN episodes are known to be high-risk events typically leading to complications.The finding of the present study regarding GNR predominance has been reported in several studies from different health care centers but with different magnitudes [27][28][29], including catheter-related BSI [30].Higher GNR prevalence indicates a shift with respect to a Chilean study that described a GPC predominance [20].Some health care centers continue to report a predominance of GPC [28,31,32] or an equal distribution of GNR and GPR [33].
Among GNR BSI cases, the present study identified K pneumoniae, E coli, and P aeruginosa as the bacteria with the highest resistance rates to the antibiotics recommended by existing international guidelines [7].In fact, the resistance rates revealed here were higher than those reported by Cattaneo et al in an Italian multicenter prospective study of patients with AL, which showed incidences of 23.2% for ESBL and 9% for CRE, as well as 21% for multiresistant P aeruginosa [34].Similarly, in a multicenter study in the United States, including FN after chemotherapy and patients with HCT, Zimmer et al identified E coli, K pneumoniae, and P aeruginosa as the most frequent GNRs, with >85% Enterobacterales susceptibility to cefepime, piperacillin-tazobactam, and carbapenems and >90% Pseudomonas susceptibility to cefepime and piperacillintazobactam, as well as 86% susceptibility to carbapenems [33].Regarding MDR bacteria, in a retrospective study in a single center from Brazil that included chemotherapy and patients with HCT, the incidence of CRE was 16.6% [32].Conversely, a multicenter European study noted a CRE incidence of 8.4% [35].Furthermore, in a retrospective study from 1 health care center in China, Wang et al reported 89 BSIs in 348 FN episodes in patients who received HCT, with a higher frequency of GNRs and an overall 30.1% incidence of MDR and 12.3% of CRE.The latter study demonstrated that patients with high-risk diseases, prolonged neutropenia, and carbapenemresistant GNRs presented independent risk factors for BSIrelated mortality [36].All these data underscore the importance of better comprehending local epidemiologic data.
In view of the high mortality rates found in the present study -26.7% in BSI episodes and 45.2% in those with GNR BSI who received inappropriate antimicrobial therapy-it is crucial to apply different interventions to decrease BSI rates: (1) use of antibiotic prophylaxis, (2) prompt identification of higher-risk FN episodes for GNR BSI to adapt to our empirical approach, and (3) updated recommendations to timely and effectively treat antibiotic-resistant GNRs when confirmed.
Although this was not a study to evaluate the effectiveness of antibiotic prophylaxis, a tendency to reduce early BSI was observed according to our data.However, given the higher frequency of antibiotic-resistant K pneumoniae detected, this practice must be directly evaluated to better identify its advantages and risks.Publications on this topic raise the question regarding the benefits for the ecology of resistance in each center [37].In our opinion, it is mandatory to implement a program of antibiotic resistance surveillance in relation to prophylaxis use.
Regarding how to identify patients at higher risk for MDR infections, Herrera et al reported a clinical score to stratify risk for BSIs with CRE in patients with cancer and HCT.The scoring system incorporated key risk factors, each of which was assigned a corresponding point value, facilitating a comprehensive risk assessment: >10 days of hospitalization (2 points), prior treatment with antibiotics for >7 days (2 points), and current colonization with KPC carbapenem-producing Enterobacterales (5 points).Patients scoring ≥7 points demonstrated a remarkable specificity of 98.3%, a positive predictive value of 77.7%, and a negative predictive value of 90.9% for CRE bacteremia, signifying a substantial risk.In the absence of any of these factors, the probability of CRE was only 1.9% [38].For the design of our study, we could not confirm the impact of these factors, but we observed a very low frequency of KPC BSI among those not colonized.Based on our results and considering >20% CRE and P aeruginosa, it is strongly recommended to enhance MDR GNR screenings with rectal swabs [39,40] and customize the optimal therapy against colonizing GNR in local protocols [41].In a similar way, Garcia-Vidal et al assessed the risk of MDR GNR infection at FN onset in hematologic cases using machine learning at a health care center in Spain and identified other factors: age >45 years, prior antibiotic use, first-ever FN in a hospitalization period, previous admissions for FN, at least 15 prior hospital visits, high-risk hematologic conditions, and hospitalization in a room formerly occupied by patients with MDR GNR isolation [42].
With regard to updating empirical therapy, considerations such as ceftazidime-avibactam, accessible in all 3 countries of the study, could be suggested as empirical treatment for FN in patients colonized by KPC-or OXA-48-producing Enterobacterales [39,43].Nonetheless, the clinical outcomes of this approach require future evaluation.Moreover, ceftolozane/tazobactam could be considered an alternative due to its potent activity against P aeruginosa strains according to Latin American data [44], and its empirical application in FN has yielded safety data and improved clinical outcomes when compared with cefepime, piperacillin/tazobactam, and meropenem [45].
When interventions are being planned to improve care for patients with FN, epidemiologic data must be available to guide the empirical antibiotic approach.However, caution must be taken to avoid unnecessary broad-spectrum antibiotic prescriptions by using discriminatory tools to decide on deescalation algorithms upon BSI identification.In a recent study from Peru, implementation of a rapid polymerase chain reaction-based blood culture identification panel had an impact on antibiotic use in this patient setting [46].
The study reported here has some limitations, such as the retrospective design, the different sample sizes at different centers, the differences in the use of antimicrobial prophylaxis, and dissimilar colonization surveillance strategies.Another issue is that some resistance rates were calculated in <30 determinations, which makes it is necessary to evaluate these data with precaution.Finally, during the COVID-19 pandemic, during which this study was being carried out, hospitals were in demand with high occupancy rates, which could have affected patient care.Despite that, the results presented are an alert for physicians in charge of hematologic cases within the region, given that several health care centers could be experiencing similar realities.
In conclusion, it is mandatory to continuously monitor BSIs in FN episodes to better understand the real epidemiology at hand, including, for instance, the finding here regarding the high prevalence of GNR and MDR BSIs in patients with AL and L within the Andean region.It is important to review and adapt existing empirical antibiotic recommendations toward avoiding inappropriate antimicrobial therapy in these vulnerable patients.

Table 2 . Frequency of Bacteria Identified in BSI: Total and Separated by Early and Late BSI
Abbreviation: BSI, bloodstream infection; GNR, gram-negative rod; GPC, gram-positive cocci.aOne S epidermidis and 1 S haemolyticus were not included because they were considered contamination.

Table 3 . Clinical Outcomes of Febrile Neutropenia Episodes
Comparison includes those with and without BSI, early BSI episodes, and early BSI episodes with GNR identification by AAT and IAT.Bold indicates P ≤ .05.Abbreviations: AAT, appropriate antibiotic therapy; BSI, bloodstream infection; GNR, gram-negative rod; IAT, inappropriate antibiotic therapy.a The episode with Sphingomonas paucimobilis was excluded from analysis because the antibiogram was not available.