Biomarkers and Fever During Neutropenia in Pediatric Cancer
Biomarkers and Fever During Neutropenia in Pediatric Cancer
This review updates our understanding of the quality of evidence available for studies investigating the predictive values of serum markers in the assessment of FN in children and young adults. It also examines the evidence, albeit limited, of their role at reassessment. The earlier review, which included studies published before February 2009, was unable to conclude that any one serum marker was more reliable due to limitations in the design and reporting of the included studies. A further 13 studies have since been published with CRP, PCT, IL-6 and IL-8 again most commonly examined. However, despite a total of now 37 studies, involving over 4689 episodes of FN, the challenges in deriving clinical conclusions are largely unchanged.
Marked heterogeneity limited direct comparisons between studies in this review. Four different FN definitions were used in the studies included in this review and 20 difference definitions in the studies included in the earlier review. Outcomes measures also varied between the studies and ranged from type of infection (bacteremia, bacterial infection, documented infection) to complication of the infection (ICU admission, SIRS, severe sepsis). Definitions of bacteremia were infrequently provided with only one study defining bacteremia secondary to coagulase-negative staphylococci, a frequent skin contaminant. Of the 3 studies examining the outcomes bacterial infection or clinically or microbiologically documented infection, only one used a standardized definition. Reassuringly, however, recognized definitions of SIRS and severe sepsis were used in the 2 studies that examined these outcomes. The variability of outcomes measures used between studies as well as a lack of standardized definitions used for bacteremia and other microbiologically or clinically documented infection further limit direct comparison of results.
In addition to the heterogeneity that exists between studies, children with FN are varied with regard to underlying immunity and risk of infection. Although the majority of studies included children with both hematological and solid malignancies, no study stratified results according to underlying diagnosis or chemotherapy pathway. Limited data exist about the kinetics of serum biomarkers in patients without acute infection undergoing different chemotherapy regimens. Of note, it has been suggested that high dose cytarabine, used in the treatment of acute lymphoblastic leukemia or non-Hodgkin lymphoma and well known to induce fever, may also induce a moderate release of both CRP and PCT.
In this and the earlier biomarker review, only 4 studies provided a direct comparison of the discriminatory power of biomarkers, with PCT and CRP most frequently compared. Although aforementioned differences between studies limit direct comparison of the ROC curve, PCT tended to have a higher sensitivity and specificity than CRP. This is in keeping with a review of the performance of biomarkers in immunocompetent children with sepsis. The prognostic ability of serial PCT levels, which peak at 6 hours after an infective stimulus (half-life 25–30 hours), may also be more pronounced than CRP that peaks at 36–50 hours (half-life 4–7 hours). Similar to results of the study by Hatzistilianou et al, multiple studies in immunocompetent children with sepsis show that PCT levels fall rapidly as appropriate antimicrobial therapy is initiated. Although further research in the kinetics of biomarkers over time is required, these data suggest that biomarkers such as PCT may have an important role in guiding duration of antibiotics as well as the need for escalation to cover potential resistant pathogens in the setting of ongoing fever beyond 24 to 48 hours.
In order to be clinically relevant, a diagnostic biomarker must be validated, readily available at point of care, and provide additional predictive ability over and above the information obtained on standard clinical assessment. Only a handful of studies have included biomarkers in the derivation of clinical decisions rules, with an elevated CRP and PCT included in the clinical decision rule criteria in 3 studies and 1 study, respectively. In reality, a serum biomarker will seldom be used alone to differentiate between low and high risk FN episodes. Future studies should focus on the additive sensitivity and specificity of biomarkers in predicting the risk of serious infection.
A recent systematic review by Lin et al investigated the role of PCT and CRP in predicting bacterial sepsis in children with FN. Using a bivariate meta-analysis model to construct ROC curves, the authors concluded that PCT had comparable discrimination to CRP. Our review, as well as previously published meta-analyses emphasize how the standard bivariate approach may be misleading and lead to clinically inappropriate conclusions. For CRP and PCT, we have shown that marked heterogeneity exists between published studies which is highlighted by the variations in sensitivity and specificity estimates (Figs. 2 and) and the negative likelihood ratios in the multivariate analysis (Table 2). We therefore caution interpretation of the review by Lin et al and suggest that more sophisticated meta-analysis techniques do not reliably indicate that CRP and PCT have comparable discriminative power.
There remains a paucity of robust and reproducible data on the use of biomarkers in prediction of serious infection in children with FN. Many of the identified limitations in published studies could be resolved by an individual participant data meta-analysis, similar to the large FN meta-analysis currently being conducted by the Predicting Infectious Complications of Neutropenic sepsis In Children with Cancer collaboration. Inconsistencies in outcome definition and reporting across all areas of FN research will also improve with pediatric consensus recommendations regarding study design, analysis and reporting. Until such time, further research on the use of biomarkers is warranted and should focus on how, and if, they can be incorporated into clinical decision rules to optimize the management of FN at initial presentation and in subsequent assessment.
Discussion
This review updates our understanding of the quality of evidence available for studies investigating the predictive values of serum markers in the assessment of FN in children and young adults. It also examines the evidence, albeit limited, of their role at reassessment. The earlier review, which included studies published before February 2009, was unable to conclude that any one serum marker was more reliable due to limitations in the design and reporting of the included studies. A further 13 studies have since been published with CRP, PCT, IL-6 and IL-8 again most commonly examined. However, despite a total of now 37 studies, involving over 4689 episodes of FN, the challenges in deriving clinical conclusions are largely unchanged.
Marked heterogeneity limited direct comparisons between studies in this review. Four different FN definitions were used in the studies included in this review and 20 difference definitions in the studies included in the earlier review. Outcomes measures also varied between the studies and ranged from type of infection (bacteremia, bacterial infection, documented infection) to complication of the infection (ICU admission, SIRS, severe sepsis). Definitions of bacteremia were infrequently provided with only one study defining bacteremia secondary to coagulase-negative staphylococci, a frequent skin contaminant. Of the 3 studies examining the outcomes bacterial infection or clinically or microbiologically documented infection, only one used a standardized definition. Reassuringly, however, recognized definitions of SIRS and severe sepsis were used in the 2 studies that examined these outcomes. The variability of outcomes measures used between studies as well as a lack of standardized definitions used for bacteremia and other microbiologically or clinically documented infection further limit direct comparison of results.
In addition to the heterogeneity that exists between studies, children with FN are varied with regard to underlying immunity and risk of infection. Although the majority of studies included children with both hematological and solid malignancies, no study stratified results according to underlying diagnosis or chemotherapy pathway. Limited data exist about the kinetics of serum biomarkers in patients without acute infection undergoing different chemotherapy regimens. Of note, it has been suggested that high dose cytarabine, used in the treatment of acute lymphoblastic leukemia or non-Hodgkin lymphoma and well known to induce fever, may also induce a moderate release of both CRP and PCT.
In this and the earlier biomarker review, only 4 studies provided a direct comparison of the discriminatory power of biomarkers, with PCT and CRP most frequently compared. Although aforementioned differences between studies limit direct comparison of the ROC curve, PCT tended to have a higher sensitivity and specificity than CRP. This is in keeping with a review of the performance of biomarkers in immunocompetent children with sepsis. The prognostic ability of serial PCT levels, which peak at 6 hours after an infective stimulus (half-life 25–30 hours), may also be more pronounced than CRP that peaks at 36–50 hours (half-life 4–7 hours). Similar to results of the study by Hatzistilianou et al, multiple studies in immunocompetent children with sepsis show that PCT levels fall rapidly as appropriate antimicrobial therapy is initiated. Although further research in the kinetics of biomarkers over time is required, these data suggest that biomarkers such as PCT may have an important role in guiding duration of antibiotics as well as the need for escalation to cover potential resistant pathogens in the setting of ongoing fever beyond 24 to 48 hours.
In order to be clinically relevant, a diagnostic biomarker must be validated, readily available at point of care, and provide additional predictive ability over and above the information obtained on standard clinical assessment. Only a handful of studies have included biomarkers in the derivation of clinical decisions rules, with an elevated CRP and PCT included in the clinical decision rule criteria in 3 studies and 1 study, respectively. In reality, a serum biomarker will seldom be used alone to differentiate between low and high risk FN episodes. Future studies should focus on the additive sensitivity and specificity of biomarkers in predicting the risk of serious infection.
A recent systematic review by Lin et al investigated the role of PCT and CRP in predicting bacterial sepsis in children with FN. Using a bivariate meta-analysis model to construct ROC curves, the authors concluded that PCT had comparable discrimination to CRP. Our review, as well as previously published meta-analyses emphasize how the standard bivariate approach may be misleading and lead to clinically inappropriate conclusions. For CRP and PCT, we have shown that marked heterogeneity exists between published studies which is highlighted by the variations in sensitivity and specificity estimates (Figs. 2 and) and the negative likelihood ratios in the multivariate analysis (Table 2). We therefore caution interpretation of the review by Lin et al and suggest that more sophisticated meta-analysis techniques do not reliably indicate that CRP and PCT have comparable discriminative power.
There remains a paucity of robust and reproducible data on the use of biomarkers in prediction of serious infection in children with FN. Many of the identified limitations in published studies could be resolved by an individual participant data meta-analysis, similar to the large FN meta-analysis currently being conducted by the Predicting Infectious Complications of Neutropenic sepsis In Children with Cancer collaboration. Inconsistencies in outcome definition and reporting across all areas of FN research will also improve with pediatric consensus recommendations regarding study design, analysis and reporting. Until such time, further research on the use of biomarkers is warranted and should focus on how, and if, they can be incorporated into clinical decision rules to optimize the management of FN at initial presentation and in subsequent assessment.
Source...