Role of Molecular Biomarkers in the Diagnosis of Invasive Fungal Diseases in Children

Summary

This review summarizes the published literature and contemporary strategies for using the biomarkers galactomannan, (1→3)-β-d-glucan, Candida mannan antigen, and anti-mannan antibody and fungal polymerase chain reaction for diagnosing invasive fungal disease in children.

Abstract

Invasive fungal diseases are important clinical problems that are often complicated by severe illness and therefore the inability to use invasive measures to definitively diagnose the disease. Tests for a range of fungal biomarkers that do not require an invasive sample-collection procedure have been incorporated into adult clinical practice, but pediatric data and pediatric-specific recommendations for some of these diagnostic tools are lacking. In this review, we summarize the published literature and contemporary strategies for using the biomarkers galactomannan, (1→3)-β-d-glucan, Candida mannan antigen and anti-mannan antibody, and fungal polymerase chain reaction for diagnosing invasive fungal disease in children. Data on biomarker use in neonates and children with cancer, history of hematopoietic stem cell transplant, or primary immunodeficiency are included. Fungal biomarker tests performed on blood, other body fluids, or tissue specimens represent promising adjuncts to the diagnostic armamentarium in populations with a high prevalence of invasive fungal disease, but substantial gaps exist in the correct use and interpretation of these diagnostic tools in pediatric patients.

Invasive fungal diseases (IFDs) are significant causes of morbidity and death in pediatric patients with a compromised immune system. Definitively diagnosing infection with yeast or a filamentous fungus is difficult and often requires an invasive procedure to obtain a standard culture. Most clinical studies use a composite IFD outcome inclusive of “proven” or “probable” IFD, separating them from “possible” IFD. According to consensus international guidelines, the distinguishing factor between a possible or probable IFD designation is the presence of either direct (eg, microbiologic recovery) or indirect (eg, positive biomarker result) mycologic criteria [1]. Although species identification and the availability of susceptibility testing on culture-positive specimens is a distinct diagnostic advantage for the treatment team, the opportunity for microbiologic recovery is not always clinically feasible or timely.

Biomarkers, measured in blood or another clinical specimen, are promising adjuncts to traditional pathologic and microbiolic tools for diagnosing IFD. Currently, assays to measure fungal biomarkers, including galactomannan (GM), (1→3)-β-d-glucan (BDG), Candida mannan antigen and anti-mannan antibody, and fungal polymerase chain reaction (PCR) are available to aid in the diagnosis of IFD. A large number of cohort studies have documented the utility of these diagnostic tools in adult patients [2–6]. However, with exception of the GM assay, relatively few data on the use of these molecular tools in children exist. It is important to note that the available adult biomarker data cannot be extrapolated seamlessly to pediatric patients for a variety of reasons, including, but not limited to, differences in the pathophysiology of IFDs, the incidence of IFDs, and cutoff values for positive results in children versus adults. Two English-language guidelines have been published [7, 8], and they provided specific guidance for using fungal biomarkers for the diagnosis of IFDs in children; however, these guidelines address only children with cancer or pediatric hematopoietic stem cell transplant (HSCT) recipients.

To interpret the studies included in this review, it is important to understand the methodologic challenges inherent in biomarker studies. All studies on fungal biomarkers are limited by the lack of a true gold standard for the declaration of an IFD, which makes it difficult to accurately classify study subjects as having or not having an IFD. As a surrogate for a gold standard, the European Organization for Research and Treatment of Cancer/Invasive Fungal Infections Cooperative Group and the National Institute of Allergy and Infectious Diseases Mycoses Study Group (EORTC/MSG) published consensus definitions for IFD in 2002 and revised them in 2008 [1, 9]. Although the EORTC/MSG criteria have established a systematic approach to defining IFDs, limitations still exist. A positive GM assay result is one of the microbiologic criteria that can contribute to the assignment of probable IFD. Reports of studies that evaluate the performance characteristics of the GM assay often do not state how the authors handle the intrinsic bias of using GM assay results to define a case of IFD. Furthermore, defining the exact time of onset for an IFD presents difficulties, which leads to ambiguity when associating a positive fungal biomarker with an IFD diagnosis. Some studies have interpreted positive biomarker test results before a patient has met established EORTC/MSG criteria for IFD as a true-positive result and applauded the finding as an improvement on the time to diagnosis of IFD. However, some positive measurements of biomarker levels in advance of the patient meeting EORTC/MSG criteria can represent a false-positive result.

In this review, we highlight current knowledge of the uses and limitations of fungal biomarkers in the care of neonates and pediatric patients with cancer, history of HSCT, or primary immunodeficiency, who are at risk for IFD.

GM ASSAY

The Platelia Aspergillus enzyme-linked GM immunoassay (Bio-Rad, Hercules, CA) was designed specifically for the early diagnosis of invasive aspergillosis (IA) (Table 1). This assay is the oldest of the major fungal biomarker tests and has been available in Europe since 1997. It was approved by the US Food and Drug Administration in 2003 for use in adult patients and in 2006 for use in pediatric patients. The assay uses a sandwich enzyme-linked immunosorbent assay technique with a rat anti-GM monoclonal antibody (EB-A2) that recognizes galactofuranose epitopes of the GM molecule as both a capture and detector antibody. The recommended threshold for a positive result from a serum specimen is an optical density index of 0.5 or higher. This threshold is supported by the current international guidelines for defining IFDs [1].

Table 1.

Summary of Fungi Detected by Fungal Biomarkers

Fungal Biomarker	Fungal Species Detected	Fungal Species Not Detected
Galactomannan	Aspergillus, Penicillium, Paracoccidioides, Histoplasma, Fonsecaea, Cryptococcus	Saccharomyces, Candida, Pichia
(1→3)-β-d-glucan	Aspergillus spp, Candida spp, Fusarium spp, Trichosporon spp, Saccharomyces cerevisiae, Acremonium spp, Coccidioides immitis, Histoplasma capsulatum, Sporothrix schenckii, Pneumocystis jirovecii	Cryptococcus spp, Blastomyces dermatitidis (in yeast form), Absidia spp, Mucor spp, Rhizopus spp; potentially lower detection of C parapsilosis and C famata
Candida mannan	C albicans, C glabrata, C tropicalis	C parapsilosis, other fungi

Adult Data

Before and since its approval, numerous studies that examined the diagnostic utility of the GM assay in individual adult subpopulations at high risk for IA across various clinical scenarios have been published. A meta-analysis of studies in which serial surveillance GM testing was used in high-risk populations found the GM assay to have an overall sensitivity of 71% and specificity of 89% [2]. Sensitivity was highest in patients with a hematologic malignancy at risk for neutropenia (70%) and for patients undergoing HSCT (82%), although data on neutrophil counts were not included. The GM assay had poor sensitivity (22%) in solid organ transplant recipients [10], likely related to the variance in pathogenesis in these patients, in whom neutropenia predisposing to fungal angioinvasion is less common [11]. The operating characteristics of the assay are also poor in the setting of mold-active antifungal therapy [12–14]. As with any diagnostic tool, the prevalence of IA in the studied population will affect the positive predictive value (PPV) and negative predictive value (NPV) of the GM assay. In the meta-analysis described above, in the setting of a baseline prevalence (pretest probability) of 5%, the GM assay would have a PPV (posttest probability) of 31% [2]. The PPV increased to 69% with a baseline prevalence of 20%. These findings underscore the importance of understanding the baseline risk for IA in the patient to be tested.

Pediatric Data

To date, 23 studies with the GM assay have been performed in children, including 18 prospective [15–32] (Table 2) and 5 retrospective [33–37] studies. Many of these studies have been reviewed [38–40]. The authors of systematic reviews concluded that the cutoff values and assay operating characteristics in children and adults seem similar [38]. In general, specificity of the GM assay has been good in prospective pediatric studies; most studies reported a specificity of >87% [39]. Sensitivity was as high as 100% for proven and/or probable IA, but only in studies with multiple cases of IA [39], which indicates that GM assay use should be limited to pediatric populations with a high likelihood of IA. The GM assay has been used as a screening and a diagnostic tool in studies focused on pediatric patients with cancer or HSCT recipients; sensitivity and specificity were highly variable, but the NPVs were consistently high [40]. Important to note is that it seems that age and weight are not specific factors in GM assay performance in children; however, false positivity in premature infants has been reported (discussed in detail later) [41].

Table 2.

Prospective Pediatric GM Assay Studies

Authors	Subjects	Sample Years	No. of Children	No. of GM Samples	ODI Cutoff	IFD Definition	Indication for Testing	Proven/ Probable IA	Sensitivity (%)	Specificity (%)	False-Positive Rate (%)a
Rohrlich et al, 1996 [15]	Children	NR	37	425	≥0.93	Guiot et al (1994) [115]	Surveillance (neutropenia)	10	100	92.6	5.7
Sulahian et al, 2001 [16]	Adults and children	1995–1998	347	2376	≥1.5	Locally defined	Surveillance (neutropenia)	9	100	89.9	10.1
Herbrecht et al, 2002 [17]	Adults and children	1997–2001	42	3294	≥1.5	EORTC/MSG 2002	Diagnosis (F&N or suspected pulmonary infection) or surveillance (HSCT)	2b	NR	47.6	44 (F&N), 75 (HSCT)c
Challier et al, 2004 [18]	Adults and children	1999–2001	20	207d	≥1.0	EORTC/MSG 2002	Clinically diagnosed IA	12b	NR	NR	NR
El-Mahallawy et al, 2006 [19]	Children	2003–2004	91	NR	≥0.5	EORTC/MSG 2002	Diagnosis (F&N)	28	79	61	NR
Hovi et al, 2007 [20]	Children	2000–2002	98	932	≥0.5	EORTC/MSG 2002	Surveillance (induction chemotherapy for ALL, chemotherapy for AML, HSCT with neutropenia or immunosuppression for GVHD)	1	NR	93	5.7
Steinbach et al, 2007 [21]	Children	2003–2004	64	826	≥0.5	EORTC/MSG 2002	Surveillance (HSCT with neutropenia or GVHD)	1	0	98.4e	1.6
Hayden et al, 2008 [22]	Children	NR	56	990	≥0.5	EORTC/MSG 2002	Surveillance (neutropenia or immunosuppressive risk)	17	65.7	87.2	1.0
Armenian et al, 2009 [23]	Children	2006–2007	68	1086	≥0.5	EORTC/MSG 2002	Surveillance (neutropenia, steroid use or HSCT)	3	NR	98.7	0.1
Zhang et al, 2009 [24]	Children	NR	88	NR	≥0.5	NR	NR	14b	71.4	91.9	NR
Fisher et al, 2012 [25]	Children	2004–2007	198	1865	≥0.5	EORTC/MSG 2002	Surveillance (neutropenia or HSCT)	1	0	95	5.2
Badiee et al, 2012 [26]	Children	2008–2009	62	230	≥0.5	EORTC/MSG 2008	Surveillance (hematology disorders)	10	90	92	NR
Choi et al, 2013 [27]	Children	2007–2010	83	640	≥0.5	EORTC/MSG 2008	Surveillance (F&N, GVHD, corticosteroids or HSCT) and diagnosis	23	91.3	81.7	18.3
Jha et al, 2013 [29]	Children	2010–2011	78	NR	≥0.5	EORTC/MSG 2002	Diagnosis (fever)	25b	84	38	NR
Dinand et al, 2016 [28]	Children	2007–2011	145	405	≥0.7	EORTC/MSG 2008	Diagnosis (F&N)	45b	82.2	82.5	13.7
Gefen et al, 2015 [30]	Children	2010–2011	46	510	≥0.5	NR	Surveillance (HSCT or high-risk leukemia)	NR	80	66	NR
Gupta et al, 2017 [31]	Children	2013–2014	125	125	≥0.5	EORTC/MSG 2008	Diagnosis (fever or F&N)	62	81.9	49	NR
Loeffler et al, 2017 [32]	Children	2012–2015	39	543	≥0.5	EORTC/MSG 2008	Surveillance (HSCT)	4	100	NR	2.1

Abbreviations: ALL, acute lymphoblastic leukemia; AML, acute myelogenous leukemia; EORTC/MSG, European Organization for Research and Treatment of Cancer/Invasive Fungal Infections Cooperative Group and the National Institute of Allergy and Infectious Diseases Mycoses Study Group; F&N, fever and neutropenia; GM, galactomannan; GVHD, graft-versus-host disease; HSCT, hematopoietic stem cell transplant; IA, invasive aspergillosis; IFD, invasive fungal disease; NR, not reported; ODI, optical density index.

^aAccording to sample, not patient, unless noted.

^bIncludes possible cases that constitute the majority of cases.

^cFalse-positive rate according to patient, not sample.

^dPooled samples from adults and children.

^eExcluded patients treated with piperacillin-tazobactam.

As noted previously for adults, the GM assay might be less sensitive in certain pediatric patients who do not have neutropenia as their primary risk factor for IA. A study that included 16 children with IA and a primary immunodeficiency (chronic granulomatous disease [CGD] [n = 10] or hyperimmunoglobulin E syndrome [Job syndrome] [n = 6]) found that GM was detected in the serum of only 25% of the patients with IA and CGD or hyperimmunoglobulin E syndrome versus 80% (24 of 30) in children with IA and neutropenia (P = .0004) [42]. In addition, the authors of recent reviews highlighted the high false-negativity rate of the GM assay in the setting of patients with CGD and an IFD in general and in those with Aspergillus nidulans infection in particular [43, 44].

In 2 published guidelines, pediatric data were considered, and recommendations for GM testing in children were formulated. The first publication is the pediatric-specific European guidelines for the diagnosis, prevention, and treatment of IFDs in children with cancer or those undergoing HSCT [7]. The authors noted that the GM assay results in operating characteristics in children similar to those in adult patients and that the use of GM for monitoring in children is warranted and reasonable [7]. Monitoring was defined as serial screening of serum GM level twice per week in children at high risk for IFD. A specific note was made that performance of the GM assay with serum from pediatric patients under nonsurveillance conditions (ie, evaluation of a pulmonary infiltrate) is unclear. Weaker support exists for use of the GM assay on bronchoalveolar lavage (BAL) specimens or cerebrospinal fluid (CSF). The second publication is the pediatric-specific international fever and neutropenia guideline, originally published in 2012 and updated in 2017 [8, 45]. The original guideline supported GM surveillance testing in hospitalized neutropenic patients at high risk (weak recommendation, moderate-quality evidence) [45]. The 2017 update no longer contains specific recommendations for GM surveillance testing. Regarding the use of fungal biomarker testing to guide empirical antifungal management in patients with febrile neutropenia that persists despite antibacterial therapy, the guidelines recommend consideration of not routinely testing serum GM levels (weak recommendation, moderate quality evidence) [8]. The rationale for this recommendation includes concerns for the low PPV in many studies and the fact that the biomarker is limited to detection of Aspergillus species and thus would not identify other fungal pathogens that could present in children with prolonged fever and neutropenia. In addition, the consistently high NPV for Aspergillus might not be as clinically informative in this setting when other mold pathogens could be the cause of the prolonged fever. Authors of the guidelines expressed concern about the utility of GM testing in children who are receiving mold-active antifungal prophylaxis [7] or did not comment because of the lack of data [8]. Mold-active antifungal prophylaxis is not recommended universally for children at risk for IA, but some guidelines support prophylaxis in centers with a high baseline incidence of invasive mold infection [7, 46]. Centers that use such prophylaxis need to consider whether surveillance GM testing is less effective and might opt for symptom-directed testing, as recommended in a recent adult-cohort publication [13]. The prevailing data indicate that mold-active antifungal prophylaxis nullifies the signal of serum GM such that surveillance is not recommended (strong recommendation, high-quality evidence) by the current Infectious Diseases Society of America aspergillosis treatment guidelines for patients who are receiving mold-active antifungal prophylaxis [47]. However, it might still be useful to apply the GM assay to bronchoscopy specimens from patients on antimold prophylaxis who are undergoing bronchoscopy for an infectious concern. Future results from pediatric-specific comparative effectiveness studies on the utility of mold-active antifungal prophylaxis and GM surveillance testing in this setting might help clarify ideal approaches for pediatric clinicians.

Most studies to date have focused on assessing the GM assay on serum specimens; however, the assay has been performed on other specimens, including BAL fluid, urine, and CSF. Two studies retrospectively examined GM levels in BAL specimens from pediatric patients [35, 36]. Each study found excellent correlation between serum and BAL GM levels, validating findings in adult patients. The authors of 1 study suggested an optimal pediatric optical density index cutoff of 0.87 in BAL fluid [35], similar to the 1.0 cutoff proposed for adult patients [48] and the 0.85 cutoff suggested by receiver operator curve analyses in adult patients [49]. A recent prospective pediatric case-control study used a cutoff of 0.5 for GM in BAL fluid, similar to that used for serum, resulting in a sensitivity of 87.5% and a PPV of 93.33% [50]. Several adult studies found that the GM assay on BAL fluid is more sensitive than the serum GM assay [49, 51], and the results of 3 pediatric studies suggested that the combined use of both modalities led to improved overall detection. Another pediatric study evaluated surveillance urine and blood GM testing in patients with prolonged neutropenia. Sensitivity of the GM assay on urine specimens was high, but a large number of false-positive results occurred when standard thresholds were used [25]. The measurement of GM levels in CSF was reported in several case reports and case series, including 3 pediatric case reports [52–54].

False-Positive GM Assay Results

Early reports on pediatric GM testing suggested an unacceptably high false-positive rate for children compared to that for adults. One study included surveillance testing results in children with a hematologic malignancy and adult patients who were undergoing allogeneic HSCT [16]. The reported false-positive rates were 10.1% (34 of 338) in children and only 2.5% (10 of 406) in an adult cohort. The medical records of 25 pediatric patients were reviewed, and the investigators concluded that the presence of concurrent mucositis, which might the favor passage of GM from the digestive tract, or early antifungal therapy, which might diminish the detectability of IFD, could have led to false-positive reactions. In another large study of adult and pediatric patients with fever of unknown origin [17], false-positive results again were observed in children significantly more frequently (44.0%) than in adults (0.9%) (P < .0001). In addition, 3 of the 11 children with false-positive results had repeated false-positive results, whereas the adult patients had only 1 false-positive result. A more recent retrospective study of 141 children with acute lymphoblastic leukemia who were screened twice per week during periods of neutropenia reported that more than half (52.1%) of the 179 positive serum samples were false positives and that one-fourth (26.1%) were of undetermined significance [37].

False-positive results are of particular concern in premature neonates. In 1 study, 6 premature infants without documented IA had a positive GM assay result in surveillance testing. In repeat testing, 5 of these premature infants were found to test positive persistently [41]. The etiology of these false-positive results is hypothesized to result from cross-reactivity of a membrane-associated molecule of Bifidobacterium bifidum subsp pennsylvanicum, which was found to mimic the epitope recognized by the EB-A2 antibody used in the GM-detection kit [55]. Bifidobacterium spp comprise 91% and 75% of the total microflora in breastfed and formula-fed neonates, respectively. Various brands of milk formula have been reported to yield positive GM assay results, with speculated translocation into serum samples after ingestion, but 3 samples of human milk that were tested in that study had a negative result [56].

In early reports, some batches of piperacillin-tazobactam were found to yield false-positive reactions in the GM assay [57]. Several older studies excluded patients who were receiving piperacillin-tazobactam or automatically classified GM results for such patients as false-positive. It is important to note that in recent years, manufacturing changes for piperacillin-tazobactam have eliminated cross-reactivity with the GM assay [58].

BDG ASSAY

BDG is an integral cell wall glucose polysaccharide found in the majority of fungi. It is notable that this polysaccharide is also the target of echinocandin antifungal agents. The BDG assay can detect Aspergillus spp, Candida spp, Fusarium spp, Trichosporon spp, Saccharomyces cerevisiae, Acremonium spp, Coccidioides immitis, Histoplasma capsulatum, Sporothrix schenckii, and Pneumocystis jirovecii (Table 1). Also notable is that several clinically important fungi, including Cryptococcus spp, Blastomyces dermatitidis (in yeast form), Absidia spp, Mucor spp, and Rhizopus spp, are not detected by the BDG assay because of their low or absent BDG production. Factor G is a coagulation factor of the horseshoe crab that serves as a highly sensitive natural detector of BDG via a modified Limulus endotoxin assay. Factor G is prepared from amebocytes of either the North American horseshoe crab (Limulus polyphemus) or the Japanese horseshoe crab (Tachypleus tridentatus), and assay properties differ in chromogenic reactivities and cutoff values depending on the source of crab species. Each of the 5 commercially available assays relies on lysate from a specific horseshoe crab species. Fungitell (Associates of Cape Cod, East Falmouth, Massachusetts), which uses the North American horseshoe crab, is approved and available in the United States and Europe. Another assay that uses the North American horseshoe crab is available for research use only (Glucatell [Associates of Cape Cod]). Fungitec G (Seikagaku, Tokyo), Wako (Wako, Japan), and Maruha (Maruha-Nichiro, Tokyo) are available only in Japan and use the Japanese horseshoe crab as the source for the assay lysate. The Fungitell BDG assay was approved by the US Food and Drug Administration in 2004.

Adult Data

Numerous adult trials that investigated the utility of the BDG assay have been completed. A meta-analysis of studies that investigated the utility of BDG testing for detecting IFDs in adult patients with hematologic or oncologic disease determined that the sensitivity, specificity, PPV, and NPV of 2 consecutive tests were 49.6%, 98.9%, 83.5%, and 94.6%, respectively [3]. An IFD prevalence of 10% was used to estimate the PPV and NPV. The majority of the 6 studies included in the meta-analysis used a screening methodology, whereas 1 study measured BDG for suspected IFD. BDG has also been investigated in the context of Pneumocystis jirovecii pneumonia (PCP) diagnosis. In 2 meta-analyses, which mostly included adult human immunodeficiency virus–positive patients, serum BDG levels had a sensitivity of 94.8% to 96% for PCP diagnosed by the presence of the pathogen in sputum or BAL fluid [59, 60].

Pediatric Data

There have been few publications on the utility of BDG testing to detect IFDs in children at risk. The studies that have been published have been limited primarily to case reports or small series [61–69]. Five prospective studies with the BDG assay were performed in children with a hematologic malignancy or pediatric HSCT recipients [26, 31, 70–72] (Table 3). Sensitivities of the BDG assay ranged from 50% to 82%, and specificities ranged from 46% to 82%. In 1 recent prospective study of pediatric patients (between birth and 21 years of age) with fungemia or microbiologically proven invasive candidiasis, the specificity increased from 67% to 76% and the NPV from 91.7% to 92.6% with an increase in the cutoff from 80 to 100 pg/mL [72]. Although specific numbers were not provided, 4 of the 5 studies noted high rates of false-positive results. The studies also were limited in some cases by use of a noncommercial assay [70] or definitions of IFDs that were not always consistent with EORTC/MSG criteria [71].

Table 3.

Prospective Pediatric BDG Assay Studies

Authors	Subjects	Sample Years	No. of Children	No. of BDG Samples	BDG Cutoff (pg/mL)	IFD Definition	Indication for Testing	Proven/ Probable IA	Sensitivity (%)	Specificity (%)	False-Positive Rate
Zhao et al, 2009 [70]	Children	2007–2008	130	NR	≥10a	NR	Diagnosis (persistent fever)	22	81.8	82.4	NR
Badiee et al, 2012 [26]	Children	2008–2009	62	230	≥80	EORTC/MSG 2008	Surveillance (hematologic disorders)	10	50	46	NR
Koltze et al, 2015 [71]	Children	2012–2014	34b	702	≥80	EORTC/MSG 2008c	Surveillance (HSCT)	6	80	82	NR
Gupta et al, 2016 [31]	Children	2013–2014	125	125	≥80	EORTC/MSG 2008	Diagnosis (fever or F&N)	62	80.2	47.2	NR
Salvatore et al, 2016 [72]	Children	2012–2015	127	NR	≥80 ≥100	EORTC/MSG 2008	Surveillance (all positive cultures and control population)	27	77 78	67 76	NR

Abbreviations: BDG, (1→3)-β-d-glucan; EORTC/MSG, European Organization for Research and Treatment of Cancer/Invasive Fungal Infections Cooperative Group and the National Institute of Allergy and Infectious Diseases Mycoses Study Group; F&N, fever and neutropenia; HSCT, hematopoietic stem cell transplant; IA, invasive aspergillosis; IFD, invasive fungal disease; NR, not reported.

^aNoncommercial assay.

^bSystemic mold prophylaxis routinely administered.

^cInconsistent application of criteria.

A recent retrospective study of 118 children, which included 790 serum samples, examined the BDG assay with a cutoff of ≥80 pg/mL using the Fungitell assay as surveillance for IFD in patients with neutropenia [73]. Using EORTC/MSG criteria to diagnose IFD, 9 unique proven or probable cases were found, and the sensitivity and specificity of the BDG assay were 75% and 56%, respectively. Specificity increased to 90% when 2 positive test results were required, but then the sensitivity decreased to 50%. Exclusion of patients on antifungal therapy also improved the performance of assays for BDG [73]. A case-control study using a novel BDG-detection assay, the GKT assay (Gold Mountain Tech Development Co., Ltd, Beijing) included 56 patients with documented candidemia and 210 patients without candidemia [74]. The patients with candidemia included 38 who were born prematurely, 10 with a gastrointestinal abnormality, 6 with congenital heart disease, and 3 with a neurologic abnormality. The sensitivity and specificity for candidemia were 68% and 91%, respectively; it was noted, however, that the BDG levels were low or negative in patients with Candida parapsilosis or Candida famata candidemia, which might be a result of smaller amounts of BDG in the cell walls of these species.

The authors of 1 pediatric case series reported detectable CSF BDG levels in 9 patients with probable or proven central nervous system Candida or Aspergillus infection [75]. BDG results reverted to negative in every patient who completed treatment. A recent detailed study of successful treatment of Aspergillus ventriculitis in a pediatric patient revealed the diagnostic value of CSF BDG testing and the pivotal role for monitoring therapeutic response [54]. CSF GM testing lacked the analytical dynamic range to be used for monitoring therapeutic response in this case. Very few data on the use of BDG testing for the diagnosis of PCP in pediatric patients are available. One small case series found BDG levels of >500 pg/dL in 3 pediatric patients with confirmed or presumed PCP [76].

It is important to note that the appropriate cutoff for a positive BDG assay result in children has not been established, and data suggest that the adult cutoff of 80 pg/mL with the Fungitell or Glucatell assay might not be appropriate for children. The differences were first noted in the initial pediatric BDG study that retrospectively examined BDG levels in the serum of children without an IFD. BDG baseline values in this setting were approximately one-third higher in children than in adults [77]. Similar concerns about increased baseline values were raised in a retrospective study of 61 neonates with or without invasive candidiasis. The optimal cutoff for distinguishing candidiasis with the Fungitell assay was 125 pg/mL [78]. Additional research studies with large pediatric cohorts of children at risk for IFDs are needed to better define the optimal threshold for positivity in children. Because of the limited data on this assay in children, there are no definitive recommendations in any pediatric or neonatal guidelines for its routine use. The European pediatric-specific guideline for the diagnosis, prevention, and treatment of IFDs in children with cancer or those undergoing HSCT notes insufficient data to base clinical decisions on BDG assay results [7]. The pediatric-specific international fever and neutropenia guideline recommends not using the BDG assay to guide empiric antifungal management in patients with fever and neutropenia that persist while the patient is undergoing antibacterial therapy (strong recommendations, low-quality evidence) [8].

Potential causes of false-positive results with the assay include blood products, hemodialysis, surgical gauze, piperacillin-tazobactam, ampicillin-clavulanate, mucositis, alcohol swabs, replacement immunoglobulin, and others (reviewed in reference 79). Persistently positive BDG levels also were reported after treatment of candidemia in a pediatric HSCT recipient, and exhaustive studies revealed no persistent focus of infection [67]. A recent study in pediatric intensive care unit patients found no difference in BDG levels in children with and those without bacteremia [80]. False-negative BDG assay results can occur in patients who are already undergoing antifungal therapy. High concentrations of bilirubin and triglycerides also reportedly inhibit the performance of the assay and cause false-negative BDG values [81].

FUNGAL PCR

Multiple publications have reported on the utility of nucleic acid–based detection of fungal DNA through PCR. These studies have focused most commonly on Aspergillus spp or Candida spp as the pathogens of interest. Quantitative real-time PCR testing of BAL fluid or induced sputum is widely used to diagnose PCP. Various PCR techniques were used in the published studies, including pathogen-specific PCR assays (for the detection of a single genus or species) or “panfungal” PCR approaches that rely on conserved gene sequences located in ribosomal RNA genes or the spacer DNA between genes (internal transcribed spacer [ITS]). Multiple gene targets and primers have been used for both the pathogen-specific and panfungal PCRs. In addition to the variability in the gene targets, there are many variations in protocols for sample preparation, specimen type, and reaction format. The specimen type can be plasma, serum, or whole blood. Studies have used reaction formats of standard PCR, nested PCR, semi-nested PCR, multiplex PCR, and real-time PCR techniques, each of which carries different advantages and challenges. The European Aspergillus PCR Initiative (EAPCRI) was established to address these issues and optimize protocols and accelerate routine clinical use for laboratory diagnosis of aspergillosis [82].

Adult Data

A meta-analysis summarized 25 studies that evaluated blood-based Aspergillus PCR in patients with a hematologic disease who were at high risk and reported a pooled sensitivity of 84% and a specificity of 76%. When the definition of true positive required 2 positive PCR results, the specificity increased to 94% [4]. For PCR diagnosis of invasive candidiasis, a separate meta-analysis reviewed 54 studies (unspecified age in 22 studies, adults alone in 17, children alone in 8, and both children and adults in 7) that included 4694 patients. The study designs varied in testing methodology and patient inclusions; both neutropenic and nonneutropenic patients were included, and disease was categorized as candidemia or invasive candidiasis (proven, probable, or possible) according to EORTC/MSG 2008 criteria. When used in the context of suspected invasive candidiasis, PCR had a sensitivity of 95% and specificity of 92% for diagnosing candidemia. When it was used to study patients with proven or probable invasive candidiasis versus true-negative patients, sensitivity and specificity were similar (93% and 95%, respectively) [5].

Recent guidelines for the diagnosis of PCP in patients with a hematologic malignancy and HSCT recipients recommended real-time PCR for the routine diagnosis of PCP [83]. BAL fluid is the preferred specimen, because a negative PCR result from a specimen collected noninvasively is not considered adequate to rule out PCP.

After the development of DNA-extraction methods, primers, and probes for quantitative PCR (qPCR) assays of mucormycosis in BAL fluid and plasma [84], several studies revealed the value of qPCR in the laboratory diagnosis of pulmonary and disseminated mucormycosis in predominantly adult populations of immunocompromised patients and patients with burn wounds [85–87]. Circulating DNA from Mucorales spp was detected in 9 of 10 patients between 68 and 3 days before the diagnosis of mucormycosis was confirmed by positive culture or histopathologic examination. However, the qPCR assay was negative in a patient with disseminated mucormycosis caused by a Lichtheimia sp [85].

A recent advance in PCR technology for the diagnosis of candidemia is the T2Candida assay (T2Biosystems, Inc, Lexington, Massachusetts) [88]. After a DNA-amplification step, T2-weighted magnetic resonance is used to detect the amplified product within 3 to 5 hours. The assay is performed on whole blood and can detect 5 distinct Candida species: Candida albicans, Candida tropicalis, Candida parapsilosis, Candida krusei, and Candida glabrata. In a study that compared the T2Candida assay to standard blood cultures, the sensitivity and specificity for detection of the 5 species from seeded blood samples were 100% and 97.8%, respectively [88]. A second study used patient-derived blood culture specimens manually inoculated with Candida to evaluate the T2Candida assay. The overall sensitivity and specificity for 250 contrived samples were 91.1% and 99.4%, respectively [89].

Pediatric Data

Eleven prospective studies of Aspergillus or pan-fungal serum PCR in children have been performed [18, 19, 23, 26, 31, 32, 90–94] (Table 4), and 2 retrospective studies of Aspergillus or panfungal PCR have been performed in children [95, 96]. Results in children have been mixed, with sensitivities ranging from 63% to 100% depending on the study, specific patient population, and selection of PCR assay. A recent retrospective study examined the combined utility of a multifungal DNA microarray and Aspergillus-specific PCR assay for the detection of IFD in biopsy specimens from 18 proven and 29 probable cases of IFD, 36 of which represented proven or probable IA [96]. The combined assays had a sensitivity of 79% and specificity of 69% for any IFD, including IA. When evaluated on blood samples from the same patients, the sensitivity of the PCRs was reduced to 33%, and specificity decreased 75%. Three studies were performed to examine Candida PCR in children [62, 65, 97]. The largest study included 54 neonates and children with a central venous catheter and persistent fever [97]. Blood cultures from 8 patients were positive for Candida spp. Candida DNA via multiplex nested PCR targeting the ITS region was detected in 13 patients, including the 8 children with a positive blood culture result, and an additional 5 patients who tested positive on the basis of PCR results only. PCR results were negative in all 28 control group patients who had no evidence of any type of bloodstream infection. The time to results was only 24 hours, which is an improvement on standard culture methods. In addition to potential delayed time to identification, blood cultures can also “miss” 50% of invasive candidiasis cases [79]. A second study found positive results for Candida PCR in 5 cases of “candidiasis” among pediatric surgery patients [62]. A limitation of that study was using a definition of candidiasis based on positive fungal biomarkers in 3 of the 5 cases, including 1 that was positive only by Candida PCR. The third study focused on patients colonized with Candida spp and found that all Candida serum sample PCR results were negative in 20 colonized patients [65]. One small pediatric study evaluated the T2Candida assay with whole blood from 15 children with candidemia and 9 children without candidemia [98]. The concordance of these results with blood culture results was 100% despite low sample volumes that required reducing T2Candida assay testing volumes from the standard of 4 to 2 mL.

Table 4.

Prospective Pediatric Aspergillus or Panfungal PCR Studies

Authors	Subjects	Sample Years	No. of Children	Type of PCR	No. of PCR Samples	IFD Definition	Indication for Testing	Proven/ Probable IA	Sensitivity (%)	Specificity (%)	False-Positive Rate (%)
Bialek et al, 2002 [92]	Children	NR	17	Aspergillus	71	EORTC/MSG 2002	Surveillance (before and after HSCT)	3a	NR	NR	NR
Challier et al, 2004 [18]	Adults and children	1999–2001	20	Aspergillus 28S rRNA	207b	EORTC/MSG 2002	Clinically diagnosed IA	12a	NR	NR	NR
El-Mahallawy et al, 2006 [19]	Children	2003–2004	91	Panfungal 18S rRNA	NR	EORTC/MSG 2002	Diagnosis (F&N)	28	75	92	8.3
Cesaro et al, 2008 [93]	Children	2004–2005	62	Aspergillus 18S rRNA	536	EORTC/MSG 2002	Diagnosis (F&N, HSCT with fever, or suspected IA)	8	63	81	19c
Armenian et al, 2009 [23]	Children	2006–2007	52	Aspergillus 28S rRNA	554	EORTC/MSG 2002	Surveillance (neutropenia, steroid use, or HSCT)	3	NR	NR	87.5
Hummel et al, 2009 [90]	Children	2000–2007	71	Aspergillus 18S rRNA	291d	EORTC/MSG 2002	Diagnosis	5	80	81	NR
Mandhaniya et al, 2012 [94]	Children	2008–2009	29	Aspergillus and Candida 28S rRNA	29	EORTC/MSG 2002	Diagnosis (fever while on mold prophylaxis)	0e	0	36	NR
Badiee et al, 2012 [26]	Children	2008–2009	62	Aspergillus	230	EORTC/MSG 2008	Surveillance (hematologic disorders)	10	80	96.2	NR
Reinwald et al, 2014 [91]	Children	1997–2000	95	Aspergillus	967	EORTC/MSG 2008	Diagnosis (fever)	0	NR	NR	10
Gupta et al, 2016 [31]	Children	2013–2014	125	Panfungal 28S rRNA	125	EORTC/MSG 2008	Diagnosis (fever, F&N)	62	82.7	54	NR
Loeffler et al, 2017 [32]	Children	2012–2015	39	Aspergillus ITS1- 5.8S rRNA	543	EORTC/MSG 2008	Surveillance (HSCT)	4	100	NR	3.9

Abbreviations: EORTC/MSG, European Organization for Research and Treatment of Cancer/Invasive Fungal Infections Cooperative Group and the National Institute of Allergy and Infectious Diseases Mycoses Study Group; F&N, fever and neutropenia; HSCT, hematopoietic stem cell transplantation; IA, invasive aspergillosis; NR, not reported; PCR, polymerase chain reaction; rRNA, ribosomal RNA.

^aIncludes possible cases.

^bPooled samples from adults and children.

^cFalse-positive rate according to patient, not sample.

^dNonblood samples included.

^ePatient with mucormycosis not included.

Two pediatric studies of PCP PCR testing of upper and lower respiratory tract specimens in children at risk for PCP have been performed [99, 100]. PCR results were positive in 100% of the cases with positive direct immunofluorescence for the cystic forms of P. jirovecii. Samuel et al [99] found that nasopharyngeal aspirates in children with PCP had similar rates of PCR reactivity when compared to BAL fluid for the detection of PCP. However, it was noted that PCR results were 7-fold more likely to be positive than those of immunofluorescence, and the specificity of a positive PCR test for true disease could not be determined. A later study from the same center also found that PCR was approximately 2.5 times more sensitive in the detection of PCP in BAL fluid than other methods [100].

There is a paucity of reported pediatric cases with mucormycosis diagnosed with qPCR [101, 102]. In 1 case, a 3-year-old child who underwent allogeneic HSCT and had disseminated mucormycosis caused by Rhizomucor pusillus was serially monitored with serum qPCR [101].

PCR technology for the diagnosis of IFD potentially has several advantages, including theoretically higher sensitivity than culture- or antigen-based methodologies. However, the lack of a standardized methodology limits widespread uptake of PCR diagnostics for detecting Candida and Aspergillus infections. False-positive results caused by nonspecific primers or the presence of clinically irrelevant fungal DNA are also a concern. Until we have standardized methods and acceptable operating characteristics for a standardized approach, use of PCR for the detection of IFD in children cannot be routinely recommended. The pediatric-specific European guidelines for the diagnosis, prevention, and treatment of IFDs in children with cancer or those undergoing HSCT make no general recommendation for fungal PCR use [7]. The pediatric-specific international fever and neutropenia guideline recommends against the routine use of fungal PCR testing of blood to guide empiric antifungal management in patients with fever and neutropenia that persist while the patient is undergoing antibacterial therapy (strong recommendations, moderate-quality evidence) [8].

CANDIDA MANNAN ANTIGEN AND ANTI-MANNAN ANTIBODY

Cell wall mannan comprises approximately 7% of Candida cell dry weight and is a major circulating antigen during infection. The assay was derived originally against C albicans mannan using the EB-CA1 monoclonal antibody [103, 104]. The Candida mannan antigen and anti-mannan antibody assays are available currently in Europe (Bio-Rad Platelia Candida antigen and Platelia Candida antibody) but are not commercially available in the United States.

Adult Data

A review of 14 studies on adult patients with invasive candidiasis examined the performance of Candida mannan antigen and anti-mannan antibody in patients with a hematologic disease and in ICU patients. The tests performed best in combination; used together, they had a sensitivity of 83% (increased from approximately 60% for the individual assays) and a specificity of 86% [6]. Sensitivity seems to be best for C albicans, followed by C glabrata and C tropicalis, most likely because of the similarity between the mannan antigens of these species. In a subset of studies on patients with a hematologic malignancy, sensitivities ranged from 71% to 100% and specificities from 53% to 92% [6]. A recent prospective study of 67 adults with a hematologic malignancy found a high false-positive rate for anti-mannan antibody and a combined specificity for mannan antigen and anti-mannan antibody of 74.8% [105]. In an adult nonneutropenic ICU population with invasive candidiasis, the two tests combined had a sensitivity and specificity of 54.8% and 59.9%, respectively [106]. In a head-to-head comparison, the sensitivity of mannan antigen and anti-mannan antibody combined were comparable to that of the BDG assay but with a lower specificity [103].

Pediatric Data

All of the pediatric Candida antigen reports to date have included, often exclusively, neonates [61, 69, 107–112]. The largest study examined 184 preterm infants and required a mannan level of ≥0.5 ng/mL in at least 2 serum samples to be considered positive [110]. Only 70 patients with at least 3 mannan measurements were included in the final analysis. Twelve neonates were found to have invasive candidiasis according to conventional testing; the sensitivity of the assay was 92% (11 of 12). Important to note is that in 8 of the 12 cases of invasive candidiasis, the mannan antigen result was positive before the blood culture by a median of 8.5 days (range, 4–18 days). A false-negative assay result occurred in a patient with C parapsilosis candidemia. This lack of detection for C parapsilosis is consistent with the difference in sensitivity for antigenic detection by EB-CA1 monoclonal antibody of the different Candida species mannose epitopes. Because the assay was derived against C albicans mannan, it is less likely to detect C parapsilosis mannan [103, 104]. The limited ability of the mannan assay to detection C parapsilosis was confirmed in other studies with a higher prevalence of C parapsilosis events [61]. All 10 neonates in that study had a positive BDG assay result, whereas only 70% had a positive result for mannan antigen. The patients who tested positive for mannan antigen were diagnosed with non-parapsilosis Candida infection. Only limited data on the utility of this assay for nonserum specimens exist. The detection of mannan antigen in BAL fluid from preterm neonates has been studied as an indication for empirical antifungal therapy in infants at high risk [111, 112], and 1 such report of CSF testing in a neonate and a child was published [107].

Nonneonatal pediatric data regarding the Candida mannan and/or anti-mannan assay are lacking. Only 5 nonneonatal patients with candidemia were included a pediatric study of Candida mannan [61]. The BDG assay result was positive for all 5 patients with a hematologic malignancy, whereas Candida mannan was detected in only 2 of these 5 candidemic patients (1 with C albicans and 1 with Candida lusitaniae infections). A pediatric intensive care unit study reported 100% sensitivity of the mannan antigen assay for candidemia and 60% sensitivity for the anti-mannan antibody assay [113]. A high false-positive rate was found (23%), but the authors speculated that many of these results represented undiagnosed invasive candidiasis as evidenced by clinical response to antifungal therapy. A study of pediatric patients with cancer colonized with Candida spp reported a low rate of positive test results for mannan antigen (10%) [65]. Even with additional data, the persistent inability of the mannan antigen assay to detect C parapsilosis is worrisome, because C parapsilosis is the causative pathogen in a significant proportion of all pediatric invasive candidiasis [114].

SUMMARY RECOMMENDATIONS FOR FUNGAL BIOMARKER USE IN PEDIATRIC PATIENTS

The GM assay has operating characteristics in nonneonatal children similar to those in adults, and the sensitivity and specificity are optimized for populations of patients with underlying neutropenia. However, the GM assay has a high rate of false positivity in neonatal patients and poor sensitivity in patients with a primary immunodeficiency or solid organ transplant recipients without neutropenia. Use of the BDG assay in neonates and older children has not been well described, and pediatric guidelines currently discourage reliance on it for diagnosis, because the optimal cutoff is not known and infections with several important fungal pathogens are not detected by the assay. Fungal PCR testing from blood currently suffers from a lack of standardization. The Candida mannan antigen/antibody approach has also undergone only limited testing in nonneonatal children, leading to an inability to offer clear recommendations on its use.

The ultimate use of biomarkers in diagnosing IFDs might require multiple assays in combination to overcome limits in sensitivity and specificity. Pediatric-specific studies of the biomarkers alone and combined are needed to define the limitations of each assay and to design the optimal multipronged diagnostic strategy. One recent study found improved sensitivity by combining biomarkers [31]. Furthermore, as with any diagnostic test, the utility of testing these biomarkers is founded on the pretest probability of infection in the child to be tested. Even if the sensitivity and specificity are optimal for a given biomarker or combination of biomarkers, the results might not be useful if the likelihood of IFD is low. Therefore, pediatric-specific guidelines on fungal biomarker testing must be individualized for the clinical scenario that incorporates the estimated IFD prevalence (ie, pretest probability) for that clinical scenario. For instance, biomarker testing might be useful when a child has a condition that predisposes him or her to IFD (eg, prolonged neutropenia) and presents with signs and symptoms of IFD, but testing for that same biomarker might not be useful for surveillance testing in the absence of signs and symptoms of IFD. In addition, an understanding of the epidemiology of pathogens that can cause IFD in specific patient populations is important when testing for specific fungal biomarkers. For instance, a negative result for a biomarker with specificity to a fungal pathogen (eg, the GM assay and Aspergillus) might provide a high NPV for that pathogen but not provide any guidance on the risk for other potential fungal pathogens. Extrapolating current pediatric biomarker data to other clinical scenarios is difficult, such as scenarios with differences in the rates of antimold prophylaxis use or specific immunosuppressive agents used. Continual study and updates to recommendations are necessary.

Notes

Financial support. This work was supported in part by the National Institutes of Health (grants 1R01 AI03315-A1 and 1R01 HD081044-01 to B.T.F and W.J.S.), the Henry Schueler Foundation and the Save Our Sick Children Foundation (to T.J.W.), and the Medical College of Wisconsin Department of Pediatrics and Children’s Hospital of Wisconsin Research Institute (to A.R.H.).

Supplement sponsorship. This article appears as part of the supplement “State of the Art Diagnosis of Pediatric Invasive Fungal Disease: Recommendations From the Joint European Organization for the Treatment of Cancer/Mycoses Study Group (EORTC/MSG) Pediatric Committee,” sponsored by Astellas.

Potential conflicts of interest. B.T.F. has received research funding from Ansun, Enzon, and Wyeth and currently receives research funding from Pfizer and Merck. T.L. is currently receiving a research grant from Gilead Sciences, is a consultant for Merck/MSD, Basilea, and Gilead Sciences, and is on the speaker’s bureau for Astellas, Gilead Sciences, Merck/MSD, and Pfizer. W.J.S. has received basic science research grants from Astellas and is on advisory boards for Merck, Astellas, and Gilead. T.J.W. receives research grants for experimental and clinical antimicrobial pharmacotherapeutics from Astellas, Cubist, Theravance, The Medicines Company, Allergan, Novartis, Merck, and Pfizer and has served as consultant to Astellas, Actavis, ContraFect, Drais, iCo, Novartis, Pfizer, Methylgene, SigmaTau, and Trius. A.R.H.: No reported conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.