Management principles for patients with neutropenia

Bacterial pathogens

When an infectious pathogen is identified as the source of FN, bacteria are the most common causes. Although bacteria as a group have remained as the most common identified etiology of pediatric FN, the epidemiology of causative pathogens has evolved. ^{, ,} The first reports on the epidemiology of bacterial infections during neutropenia most commonly implicated gram-negative pathogens, specifically Escherichia coli, Pseudomonas aeruginosa, and Klebsiella species. The transition from a gram-negative to a gram-positive bacterial predominance occurred in the latter 2 decades of the last century. This shift in pathogen type is likely multifactorial, but is often assumed to be related to increased reliance on central venous catheters and chemotherapy regimens that cause mucositis, resulting in an increase in pathogens such as viridans group streptococci. It is anticipated that gram-positive organisms will continue to predominate into the future as more centers will likely use prophylactic antibiotic regimens that have broader gram-negative activity in high-risk patient groups.

Despite the general predominance of gram-positive bacteria, significant variation in the epidemiology of bacteria during FN exists between centers. Table 8.1 displays identified bacterial pathogens across various international pediatric oncology studies between 1982 and 2018. ^{, , , ,} The variation by geographic location likely results from practice variation, such as approach to chemotherapy protocols, diagnostic testing practices, and prophylaxis regimens. Most recently, a pediatric randomized controlled trial of levofloxacin during periods of prolonged neutropenia in children with acute leukemia and those undergoing HCT was completed. Although levofloxacin was found to be effective, the rate of breakthrough infection was still 22% in the leukemia group and 11% in the HSCT group. Gram-positive organisms, most frequently viridans group streptococci, accounted for more than 77% of the breakthrough events.

Fungal pathogens

Invasive fungal diseases (IFDs) are rarely the source of the initial onset of fever during a neutropenic period. More typically, the concern for IFDs increases after a prolonged period of FN despite broad-spectrum antibacterial therapy. There are published consensus criteria for defining proven and probable IFDs that have been helpful to standardize the definition of IFDs across research studies and to provide some diagnostic criteria for clinicians. However, diagnosing IFDs by these published criteria can be difficult because invasive procedures are often needed to identify a fungal pathogen and patients with prolonged neutropenia cannot always tolerate such procedures. Therefore many published reports of IFD incidence as a source of FN may underestimate actual infection rates. Understanding these limitations, prospective multicenter data have documented a proven or probable IFD rate ranging from 3% to 5% of children hospitalized with fever and neutropenia. The rates of IFDs when considering prolonged neutropenia regardless of fever have been reported to be much higher. This highlights the fact that fever is not always present as a sign of IFD.

Candida species are the most common fungal pathogens identified during periods of FN. This is likely because Candida species commonly colonize the skin and intestinal tract and may become more dominant in the setting of prolonged exposure to broad-spectrum antibiotics. The skin and mucosal barriers are often compromised by the presence of central venous catheters and/or chemotherapy exposures that can allow for invasive of Candida isolates. Specific mortality data regarding invasive candidiasis in pediatric oncology patients and HSCT recipients are limited, but the attributable mortality of invasive candidiasis in all pediatric patients has been estimated to be 10%.

Episodes of invasive mold disease are less common but are much more challenging to treat and have significantly higher rates of case fatality. In contemporary pediatric cases series, less than two-thirds of patients with an invasive mold disease IMD responded to therapy in the first 12 weeks and 30% of patients died within the same time period. ^, Among the mold pathogens, Aspergillus species are most common, followed by organisms of the Mucorales order.

Viral pathogens

The advancement in viral diagnostic methodologies has resulted in better estimates of viral infections during periods of FN. Much of the interest in testing for a viral pathogen is the possibility that finding an explanation for fever may reduce the need for further diagnostic testing. The yield of viral testing in patients with FN has been reported in multiple studies ( Table 8.2 ). The frequency of laboratory-confirmed viral respiratory infection ranged from 8% to 59%. Of note, the study reporting an 8% incidence of viral respiratory infection obtained viral respiratory specimens via mouth swabs and thus likely underestimated the true rate. The range of infection rates for the remaining studies was 37% to 59%.

Although some authors have suggested these rates of viral detection support routine comprehensive viral testing at the time of presentation for FN, ^, the utility of routine viral testing is not clear. First, ideally the identification of a viral pathogen should inform clinical management decisions. However, identification of a virus does not necessarily exclude the possibility of a concomitant bacterial infection. The percentage of patients with both a viral and bacterial infection ranged from 13% to 33%. ^, Some clinicians are comfortable stopping antibiotics during the FN episode in the setting of a viral syndrome but only for low-risk patients with FN. Based on the possibility of both bacterial and viral infection, empiric antibiotics are often continued in high-risk FN episodes. Second, the sensitivity of viral polymerase chain reaction testing results in detection of virus well after clinical resolution, and thus viral detection by polymerase chain reaction may not always confirm the source of fever in a neutropenic patient. Finally, there are limited effective antiviral therapeutic agents available, and thus detection of some viruses will not inform targeted antiviral therapy. Considering these reasons collectively, viral testing should be limited to patients in whom positive results would allow for de-escalation of antibiotic therapy (e.g., in a low-risk FN episode) or initiation of an appropriate antiviral therapy (e.g., neuraminidase inhibitor for influenza). Of note, the hospital’s infection prevention and control division may desire testing for symptomatic patients to inform appropriate isolation precautions that could limit hospital transmission of viral pathogens.

Initial risk stratification

Early comparative studies highlighted the effectiveness of early initiation and continuation of empirical antibiotic and subsequently empirical antifungal therapy to reduce the morbidity and mortality associated with FN. ^, These studies served as the foundation for standards of care for FN management that have been applied for decades across all episodes of FN. However, use of empirical broad-spectrum antibiotics and antifungal agents for prolonged periods in all patients is not ideal as it presents risks for medication toxicities, prolongs hospital stays, and potentiates evolution of resistance. As not all FN episodes carry the same risk for infection, it is important to stratify each FN episode into risk groups for true infection. Such risk stratification can inform evidence-based decisions for more discriminant use of anti-infective agents and other health care resources.

Identifying which children are at a lower risk of complications can allow for a reduction in the intensity of anti-infective therapy and monitoring. Conversely, identifying children at higher risk of complications can allow for prophylactic approaches, rapid escalation of therapy, or closer observation. Fortunately, there have been substantial research efforts to identify criteria for stratifying FN episodes into low and high risk. Often, these studies leverage a composite of factors to derive risk prediction models or rules. More than 25 such risk prediction studies have been conducted in pediatric cancer. ^, These studies have been heterogeneous and have included different pediatric cancer populations and different clinical endpoints (such as bacteremia, serious infection, death, and intensive care unit admission), thereby reducing the ability to combine the individual study data into a composite analysis. However, review of the individual studies can be informative.

There have been six prediction models derived from pediatric cohorts that (1) focused on identifying patients at low risk for infection using data elements evident on a single FN assessment and (2) have been validated. Selection of a single schema that can be applied across all clinical scenarios has not been possible, potentially because of heterogeneity in clinical settings and resources. Therefore clinicians should review each of the validated low-risk stratification schemas, choose which schema matches their clinical setting, and determine if the application of that schema is feasible for their center. The choice of strategy should be determined by an institution’s ability to implement more complex rules and the timeliness of receipt of required components of the rule, such as C-reactive protein. Whichever schema is chosen, centers should establish a quality improvement infrastructure to routinely monitor their process for identification of low-risk FN episodes and outcomes of these episodes to ensure the chosen prediction model is safe and continues to have local applicability. Of note, these prediction models were derived in cohorts of children with cancer and chemotherapy-induced FN and thus their applicability to FN episodes in the post-HSCT period is not known.