Management of Non-neutropenic Fever in Children with Cancer

INTRODUCTION—FEVER IN THE NEUTROPENIC CHILD Infections are a major complication in children undergoing treatment for cancer. Profound neutropenia, commonly defined as an absolute neutrophil count (ANC) <500/µL, has been known for a long time as the major risk factor for severe infection. As fever (mostly defined as a temperature higher than 38.3 °C or 38.5 °C mostly defined as a temperature higher C) in the neutropenic host may indicate an infection, early institution of empirical antibiotic therapy has become the standard of care.1 While this strategy has decreased mortality to less than 3%,2 it leads to overtreatment of a proportion of children.3 Over the last 2 decades, empirical antibiotic treatment in this patient population has been refined to decrease the time of hospitalization and antibiotic use. This is important, as hospitalization decreases the quality of life of patients and their families, limits the number of available hospital beds, and may contribute to higher rates of resistance. In this regard, the dogma of stopping antibiotic therapy only with hematological recovery has been modified in low-risk patients, where antibiotics may now be discontinued in the absence of hematological recovery if blood cultures are negative and the patient is stable and afebrile for at least 24 hours.4 Non-neutropenic Fever—Epidemiology Unfortunately, the approach in children with cancer who develop fever during a period of non-neutropenia is less clear, and not surprisingly, current strategies vary widely.5 A recent survey among 227 Children's Oncology Group (COG) institutions revealed that in well-appearing children with non-neutropenic fever, more than 80% of the centers routinely institute intravenous empirical antibiotics, whereas about 20% of the institutions withhold antibiotics based on either clinical appearance or on a risk stratification tool. Similarly, although the most common agent administered in this situation was ceftriaxone, about 10% of the institutions relied on agents covering a broader spectrum (eg, piperacillin-tazobactam). Importantly, despite the pivotal role of neutropenia as a risk factor for bacterial infection, several other arms of host defenses may be affected in non-neutropenic children with cancer.6 For example, indwelling central catheters, as well as mucositis, may facilitate the invasion of bacterial and fungal pathogens, and the impairment of lymphocyte subsets such as T- and B-cells may decrease the host defense against viruses. Therefore, non-neutropenic children with cancer are at lower risk for severe infection compared with neutropenic patients, but again, these patients still are immunologically different compared with healthy children. The occurrence of fever in non-neutropenic cancer patients seems to be a relatively common event. A recent study performed in two major pediatric cancer centers in Germany analyzed children with acute lymphoblastic leukemia, acute myeloid leukemia, or non-Hodgkin lymphoma (NHL).7 The 210 analyzed patients developed a total of 776 episodes of fever during intensive chemotherapy, with 205 episodes occurring in the absence of neutropenia (26.4%). Similar results were reported in 108 children, mostly diagnosed with leukemia or lymphoma,8 in a mixed pediatric and adult population with non-M3 acute myeloid leukemia,9 and in adults treated for acute leukemia,10 with 25%, 10.6% and 33% of febrile episodes without neutropenia, respectively. Causes of Non-neutropenic Fever The causes of fever during episodes without neutropenia were evaluated in 1 study with 371 patients (79 and 292 patients with leukemia/lymphoma or a solid tumor, respectively).11 Most of the 356 non-neutropenic febrile episodes (67%) were due to infection, whereas noninfectious causes included the underlying malignancy, concurrent medication, or status postsurgery. With a focus on infections, a pooled analysis of 15 studies including 4106 episodes of non-neutropenic fever in pediatric cancer patients showed that upper respiratory infections were frequent and occurred in up to 60%, whereas the pooled average rate of bacteremia was only 8.2%.12 Importantly, the percentage of episodes of bacteremia seems higher in neutropenic than in non-neutropenic patients,7,13 and Gram-positive bacteria are more often isolated than Gram-negative pathogens. Infections due to coagulase-negative staphylococci were most observed, followed by Enterococcus spp, Staphylococcus aureus and Streptococcus pneumonia, whereas Enterobacterales, Pseudomonas spp and Acinetobacter spp were the most common Gram-negative pathogens.12 Of note, bloodstream infections due to fungi were diagnosed in only 7 patients (Candida spp in 6 patients, Trichosporium spp in 1 patient). Importantly, the presence of a central venous catheter (CVC) was associated with bacteremia, with a significantly higher risk for Broviac/Hickman type or peripherally inserted central catheter compared with ports.12 In addition, the presence of clinical signs such as hypotension, chills or rigor were significantly associated with bacteremia. Importantly, the clinical significance of bacteremia in non-neutropenic patients may be different from that in neutropenic children, which is suggested by the data of a prospective single-center study in febrile non-neutropenic pediatric cancer patients with a CVC.14 Among a total of 937 febrile episodes, 834 episodes were classified as low risk (predicted risk for bacteremia <10%), and in 603 episodes, antibiotics were withheld within the 7 days of presentation. True bacteremia was detected in 7 out of these children not receiving antibiotics, of whom all were in good clinical condition and all but one were afebrile at reassessment.14 Nevertheless, complicated courses of infections with intensive care unit (ICU) admissions and even lethal events have been described in non-neutropenic cancer patients presenting with fever, but the rates of these events are significantly lower compared to that in the neutropenic host. Whereas an analysis of 15 studies evaluating 4,106 non-neutropenic febrile episodes reported infrequent ICU admissions and no deaths,12 one recent two-center study observed in 205 episodes of fever during a period of non-neutropenia that three children had to be transferred to the ICU (one patient each with metapneumovirus pneumonia, pulmonary aspergillosis, and sepsis syndrome).7 All three patients died due to the infectious complication, resulting in an infection-related mortality of 1.4% in this study. To this end, bacteremia may not be the only clinically relevant endpoint of studies evaluating risk prediction models in non-neutropenic febrile children, and other severe infectious complications have to be considered. Risk stratification To evaluate risk factors for a complicated clinical course of non-neutropenic fever, which may potentially be used for developing clinical practice guidelines, a retrospective two-center study focused on children treated for acute lymphoblastic leukemia, the most common cancer in childhood.15 The study population consisted of 524 children in whom 1,591 febrile episodes without severe neutropenia occurred. Upper respiratory infection was the most common infectious complication, accounting for 381 (67%) episodes, whereas documented bloodstream infections were observed in only 30 (1.9%) cases; almost all patients had an implantable port and not a tunneled catheter. No infection was seen in patients who had recent exposure to cytarabine, a cytotoxic agent whose administration is frequently associated with fever, and only one child with high-risk leukemia died from an infection (disseminated varicella during intensive treatment). Multivariate analysis revealed that the presence of hypotension, chills and rigors, higher temperature, and leukemia with infant phenotype were independently associated with bacteremia. Both centers differed regarding their standard strategy in febrile non-neutropenic fever: one center administered empirical anti-pseudomonal antibiotics to all febrile children when a CVC was present, whereas the other center withheld antibiotics in those who appeared well and met the predefined low-risk criteria. Therefore, the data clearly suggest that a subset of patients may not require antibiotic treatment. To define the risk for bacterial infection, most studies rely on clinical parameters, whereas the use of biomarkers such as procalcitonin, interleukin (IL) 6 or 8 or the C-reactive protein was not included. In this respect, a recent meta-analysis of procalcitonin in adult cancer patients revealed that the pooled sensitivity and specificity of the biomarker were moderate to low with 60% and 78%, respectively, and sensitivity and specificity did not differ from those of the C-reactive protein.16 Similar results were described for IL-8.17 Therefore, these parameters do not seem to have a meaningful impact on clinical decision-making. Clinical Decision Making One group has extensively investigated the setting of non-neutropenic fever in children with cancer, with a significant impact on this field.14,18,19 In a first step, they performed a retrospective analysis of 932 febrile episodes in 463 non-neutropenic pediatric cancer patients.18 Using logistic regression modeling, they evaluated a number of candidate predictors, including age, cancer diagnosis, presence of CVC, in- and outpatient presentation, body temperature, chills, hypotension, ANC or absolute monocyte count. Based on these results, they developed a risk prediction model (Esbenshade/Vanderbilt, EsVan), which subsequently underwent some minor modification. These models were externally validated at 6 sites, and a reasonable discrimination regarding the risk of bloodstream infections was obtained.19 The model EsVan2b, which outperformed the other models, was able to predict bloodstream infections through multiple performance sites, in particular regarding the prediction of high-risk bacteremia (Gram-negative or S. aureus infection). In the next step, the group applied this risk prediction, the publicly available Web-based Esbenshade Vanderbilt models, in a single institution, estimated the risk for bacteremia and used these results for clinical decision-making in well-appearing pediatric cancer patients with non-neutropenic fever.14 Notably, the treating provider decided whether the patient was ill or was well-appearing, but no reproducible criteria were predefined. For patients with a low risk of bacteremia (defined as a risk of less than 10%), discharge home without antibiotics was recommended.14 Those patients with an intermediate risk of bacteremia (10%–39.9%) received an antibiotic before discharge, and those with high risk of bacteremia (higher than 40%) received broad-spectrum antibiotics, respectively.14 Among 937 episodes of fever without neutropenia, the prediction model identified 88.9%, 8.6% and 2.3% of low-, intermediate- and high-risk episodes, with an incidence of bacteremia of 1.9%, 13.6% and 54.5%, respectively. Importantly, in the low-risk group, only 21.1% of children received empirical antibiotics, whereas 72.3% of those children did not require intravenous antibiotics within 7 days of presentation. In addition, no deaths or clinical decompensations attributable to antibiotic delay were observed. This study demonstrated that using the prediction model, antibiotic usage could safely be reduced in febrile, non-neutropenic cancer patients. The most recent study by this group validated the model externally in 18 academic medical centers, most of them located in the United States.20 In 2565 evaluable episodes, the rate of bloodstream infection was 4.7%, and 27 (1.1%) of the patients needed ICU care. Five patients (0.2%) died, but there was only one potential infection-related death; all 5 patients had received initial empirical antibiotics. In total, initial empirical antibiotic therapy was withheld in only 14.9% of episodes, and in this population, no ICU admission or death attributable to not receiving antibiotics was observed. This work is impressive but needs several comments. First, the results of these studies need to be validated in other countries and health-care settings (eg, low- and middle-income countries, different settings of bacterial epidemiology and resistance, different use or not of antibacterial prophylaxis, policies of patient follow-up (ie, local/ remote and homecare/ no homecare). Second, the impact of targeted therapies, which may be associated with severe infections during non-neutropenic episodes, is unclear.21 Third, patients with fever during a non-neutropenic episode may develop serious clinical problems or even die due to infectious complications that are not caused by bloodstream infections, such as severe viral or fungal infections. Therefore, studies need to implement these endpoints in the validation of risk prediction models. Last, the legal aspects of a publicly available web-based prediction model, which is a medical device, have to be clarified. For example, it has to be addressed who is legally responsible if a patient is misclassified by the prediction model and develops serious medical problems. In addition, it is important to note that medical devices manufactured anywhere in the world that are then marketed in the European Union require Conformité Européenne marking. CONCLUSIONS AND PERSPECTIVES Fever during periods without neutropenia is a common event in children with cancer, but the clinical strategies in this setting still vary widely. Risk factors for bloodstream infections in non-neutropenic febrile cancer patients have been well described and validated, and data are promising to successfully use these risk factors for clinical decision making. However, in addition to bacteremia, there are other infectious complications such as viral or fungal infections, which have also been reported to cause a severe clinical course in this setting. In our opinion, careful patient monitoring and follow-up are key to the ultimate success of this strategy. An additional observation time in the hospital, which had been included in the Australian-UK-Swiss (AUS) rule for children with febrile neutropenia, may also be of additional value in the setting of non-neutropenic fever.22 At the moment, we are planning to include the current knowledge in a common conservative risk stratification and management strategy for evaluation in a multinational multicenter observational study (Fig. 1). By doing so, we will hopefully be able to further optimize the strategies to safely and evidence-based reduce antibiotic usage in children with cancer who present with non-neutropenic fever.FIGURE 1.: Potential multinational multicenter observational study for evaluating a common risk stratification and management for febrile non-neutropenic children with cancer.