Prueba (1→3)‐β‐D‐glucano para la detección de infecciones fúngicas invasivas en pacientes inmunodeprimidos o en estado crítico
Resumen
Antecedentes
Las infecciones fúngicas invasivas (IFI) son infecciones oportunistas que ponen en peligro la vida y que se producen en pacientes inmunodeprimidos o gravemente enfermos. La detección y el tratamiento tempranos de las IFI son esenciales para reducir la morbilidad y la mortalidad en esas poblaciones. El (1→3)‐β‐D‐glucano (BDG) es un componente de la pared celular fúngica que se puede detectar en el suero de los pacientes infectados. La prueba BDG en suero es una forma de detectar rápidamente estas infecciones e iniciar el tratamiento antes de que se conviertan en una amenaza para la vida. Hay cinco versiones diferentes de la prueba BDG disponibles en el mercado: Fungitell, Glucatell, Wako, Fungitec‐G y Dynamiker Fungus.
Objetivos
Comparar la exactitud diagnóstica de las pruebas BDG en suero disponibles en el mercado para detectar determinadas infecciones fúngicas invasivas (IFI) entre pacientes inmunodeprimidos o gravemente enfermos.
Métodos de búsqueda
Se hicieron búsquedas en MEDLINE (vía Ovid) y en Embase (vía Ovid) hasta el 26 de junio de 2019. Se utilizó SCOPUS para realizar una búsqueda de citas hacia adelante y hacia atrás de artículos relevantes. No se impusieron restricciones por idioma o diseño del estudio.
Criterios de selección
Se incluyeron todas las referencias publicadas en o después de 1995, que es cuando las primeras pruebas comerciales de BDG estuvieron disponibles. Se consideraron los estudios publicados y revisados por pares sobre la exactitud de las pruebas diagnósticas de BDG para el diagnóstico de infecciones fúngicas en pacientes inmunodeprimidos o en cuidados intensivos que utilizaron los criterios de la European Organization for Research and Treatment of Cancer (EORTC) o su equivalente como estándar de referencia. Se consideraron todos los diseños de estudio (casos y controles, cohortes consecutivas prospectivas, y estudios de cohortes retrospectivos). Se excluyeron los estudios de casos y los estudios con menos de diez participantes. También se excluyeron los estudios en animales y de laboratorio. Se excluyeron los resúmenes de congresos porque no proporcionaron información suficiente.
Obtención y análisis de los datos
Se siguieron los procedimientos estándar descritos en el Manual Cochrane para Revisiones de Exactitud de Pruebas Diagnósticas (Cochrane Handbook for Diagnostic Test Accuracy Reviews). Dos autores de la revisión, de forma independiente, revisaron los estudios, extrajeron los datos y realizaron una evaluación de la calidad de cada estudio. Para cada estudio se creó una matriz de 2 × 2 y se calculó la sensibilidad y la especificidad, así como un intervalo de confianza (IC) del 95%. La calidad de los estudios incluidos se evaluó mediante la herramienta Quality Assessment of Studies of Diagnostic Accuracy‐Revised (QUADAS‐2). No fue posible realizar un metanálisis debido a la considerable variación entre los estudios, con la excepción de Candida, por lo que se han proporcionado estadísticas descriptivas como la curva de rendimiento diagnóstico y los diagramas de bosque (forest plot) por marca de la prueba para mostrar la variación en los resultados de los estudios.
Resultados principales
En esta revisión se incluyeron 49 estudios con un total de 6244 participantes. Cerca de la mitad de estos estudios (24/49; 49%) se realizaron con pacientes que presentaban cáncer o neoplasias hematológicas. La mayoría de los estudios (36/49; 73%) se centraron en la prueba BDG de Fungitell. Le siguieron Glucatell (cinco estudios; 10%), Wako (tres estudios; 6%), Fungitec‐G (tres estudios; 6%) y Dynamiker (dos estudios; 4%). Aproximadamente tres cuartas partes de los estudios (79%) utilizaron un diseño de estudio consecutivo prospectivo o retrospectivo; el resto utilizó un diseño de casos y controles.
Según los niveles de corte recomendados por el fabricante para la prueba de Fungitell, la sensibilidad osciló entre el 27% y el 100%, y la especificidad entre el 0% y el 100%. Para la prueba de Glucatell, la sensibilidad osciló entre el 50% y el 92%, y la especificidad, entre el 41% y el 94%. En muy pocos estudios se han utilizado las pruebas Dynamiker, Wako y Fungitec‐G, pero las sensibilidades y especificidades individuales oscilaron entre el 50% y el 88%, y entre el 60% y el 100%, respectivamente. Los resultados muestran diferencias considerables entre los estudios, incluso por fabricante, lo que impidió un metanálisis formal. La mayoría de los estudios (32/49; 65%) no informaron de un alto riesgo de sesgo en los dominios de QUADAS‐2. Los dominios de QUADAS‐2 con un mayor riesgo de sesgo incluyeron la selección y el flujo de participantes y el momento.
Conclusiones de los autores
Se observó una considerable heterogeneidad entre los estudios, y esas diferencias impidieron que se realizara un metanálisis formal. Debido a la amplia variación de los resultados, no es posible determinar la exactitud diagnóstica de la prueba BDG en contextos específicos. Los futuros estudios que determinen la exactitud de las pruebas BDG se deberían vincular a la forma en que se utiliza la prueba en la práctica clínica y deberían describir claramente el protocolo de muestreo y la relación entre el momento de la prueba y el momento del diagnóstico.
Resumen en términos sencillos
Medición de β‐D‐glucanos para detectar la infección fúngica invasiva en pacientes inmunodeprimidos
¿Por qué es importante mejorar el diagnóstico de las infecciones fúngicas invasivas?
Las infecciones fúngicas se producen en pacientes que no pueden luchar contra la infección, y estas infecciones pueden poner en peligro la vida de este grupo de pacientes. Estas infecciones suelen ser difíciles de diagnosticar. El hecho de no reconocer una infección fúngica cuando está presente (un resultado de prueba falso‐negativo) provoca un retraso en el tratamiento y desenlaces más deficientes. Un diagnóstico incorrecto de la infección (un resultado falso positivo) puede dar lugar a que se desperdicien recursos y a una investigación y tratamiento innecesarios.
¿Cuál es el objetivo de esta revisión?
El objetivo de esta revisión es determinar cuán exacto es un análisis de sangre para el diagnóstico de las infecciones fúngicas en pacientes que no pueden combatir la infección. Los autores de la revisión incluyeron 49 estudios para responder a esta pregunta.
¿Qué se estudió en esta revisión?
Se compararon cinco tipos de análisis de sangre. Todas estas pruebas utilizan métodos bioquímicos similares para detectar la presencia de una molécula de azúcar (β‐D‐glucano) que es un componente de la pared celular fúngica. Esta molécula no suele aparecer en la sangre, por lo que su detección indica que hay hongos presentes. Las pruebas requieren una muestra de sangre, que luego se envía a un laboratorio para su análisis. El diagnóstico de las infecciones fúngicas es difícil, y a menudo se realiza sólo después de que la enfermedad ha avanzado. Los análisis de sangre pueden proporcionar un diagnóstico más temprano, por lo que ofrecerían una ventaja sobre los métodos actuales.
¿Cuáles son los principales resultados de la revisión?
Esta revisión incluyó estudios de 6244 pacientes que estaban en riesgo de contraer infecciones fúngicas. Los resultados de los estudios muestran que la exactitud varió ampliamente entre los estudios. La variación fue tan grande que no fue posible obtener una estimación fiable de la exactitud de las diversas pruebas.
¿En qué medida son fiables los resultados de los estudios de esta revisión?
En los estudios incluidos, el diagnóstico de la infección fúngica invasiva se realizó mediante los criterios desarrollados por la European Organization for Research and Treatment of Cancer (EORTC)*. Los criterios de la EORTC se consideran fiables y los estudios se realizaron en general correctamente, por lo que es probable que los diagnósticos de referencia fueran exactos. La exactitud de los análisis de sangre para detectar infecciones fúngicas invasivas varió mucho. Algunos estudios encontraron que el análisis de sangre era exacto, pero otros encontraron que no lo era mucho. No se entiende el motivo de esta variación.
*Los criterios de la EORTC proporcionan el diagnóstico de referencia. Los resultados del análisis de sangre se comparan con el diagnóstico de referencia.
¿Para quiénes son relevantes los resultados de esta revisión?
La mayoría de los estudios incluidos se realizaron en centros médicos académicos u hospitales públicos en los Estados Unidos, Alemania e Italia. Las afecciones subyacentes más comunes fueron el cáncer (47%) y el ingreso en cuidados intensivos (33%). La mayoría de los participantes eran adultos. La prevalencia general de la infección fúngica invasiva fue del 28%.
¿Cuáles son las implicaciones de esta revisión?
La exactitud diagnóstica varió ampliamente entre los estudios. No está claro si las pruebas pueden detectar con exactitud las infecciones fúngicas invasivas. Las pruebas detectaron con exactitud la enfermedad en algunos estudios, pero en otros no. No se comprenden los motivos de la variación en la exactitud.
¿Cuál es el grado de actualización de esta revisión?
Los autores de la revisión buscaron y revisaron los estudios publicados hasta junio de 2019.
Authors' conclusions
Summary of findings
| Participants/Populations: immunocompromised people at risk for invasive fungal infections | |||||
| Prior testing: none | |||||
| Settings: hospital setting | |||||
| Index test: commercially available serum BDG test | |||||
| Importance: test needed to accurately detect fungal infections in susceptible people at an early enough stage to facilitate antifungal treatment | |||||
| Reference standard: EORTC/MSG criteria, or by microscopy or autopsy | |||||
| Studies: 49 studies with 6244 participants | |||||
| 1. Test assay | |||||
| Test/Subgroup | No. of participants (studies) | Overall sensitivity (range) | Overall specificity (range) | Implications | Quality and comments |
| Fungitell | 4316 (36) | 27% to 100% | 0 to 100% | Wide variation in sensitivity | |
| Glucatell | 957 (5) | 50% to 92% | 41% to 94% | Wide variation in sensitivity | |
| Wako | 420 (3) | 50% to 86% | 89% to 100% | Insufficient number of studies for meta‐analysis | |
| Fungitec‐G | 353 (3) | 67% to 88% | 60% to 85% | Too few studies for meta‐analysis | |
| Dynamiker Fungus | 198 (2) | 64% to 81% | 78% to 80% | Too few studies for meta‐analysis | |
| 2. Fungal organism | |||||
| Test/Subgroup | No. of participants (studies) | Sensitivity estimate (95% CI) | Specificity estimate (95% CI) | Implications | Quality and comments |
| Candida | 1185 (10) | 81% (75% to 86%) | 64% (56% to 72%) | Results are more homogeneous for Candida testing than for all fungi | |
| BDG: beta‐D‐glucan test; CI: confidence interval; EORTC/MSG: European Organization for Research and Treatment of Cancer Mycoses Study Group. | |||||
Background
Target condition being diagnosed
Invasive fungal infection (IFI) is a major cause of morbidity and mortality in immunosuppressed and critically ill people (Lemonovich 2018; Person 2010). Prompt diagnosis is important because early initiation of appropriate antifungal therapy improves patient outcomes (Chamilos 2008; Garey 2006; Morrell 2005; von Eiff 1995). Diagnosis of IFI is challenging because the standard methods of clinical diagnosis (e.g. clinical signs and symptoms, host risk assessment, physical examination, radiography) are not specific to IFI. In addition, traditional microbiological methods often have limited clinical utility because cultures are frequently negative or become positive only in advanced stages of infection (Clancy 2013). Histopathologic examination of infected tissue has been the historic gold standard, but invasive testing may not be feasible in unstable participants or in those with underlying coagulopathy. Although composite definitions for IFI have been developed by the European Organization for Research and Treatment of Cancer Mycoses Study Group (EORTC/MSG), these definitions are best suited for research purposes (De Pauw 2008). The EORTC/MSG diagnostic categories of IFI include proven, probable, and possible disease.
Current strategies for prevention and management of IFI include antifungal prophylaxis, pre‐emptive therapy, empiric treatment, and treatment of established infection (Leroux 2013). Universal prophylaxis is effective and logistically easy, but the medications can have toxic effects, can potentially promote antimicrobial resistance, and are expensive. Empiric therapy based on symptoms or treatment of established IFI potentially delays initiation of potentially life‐saving therapy. In contrast, pre‐emptive therapy is a more selective approach in which people are sequentially monitored and treatment is based on detection of laboratory biomarkers in blood, often before clinical signs or symptoms of an IFI are apparent. Pre‐emptive approaches are designed to identify the highest‐risk people who are most likely to benefit from early antifungal therapy. Examples of fungal biomarkers include circulating fungal DNA and cell wall components such as galactomannan (GM), glucuronoxylomannan, mannan, and (1→3)‐beta‐D‐glucan (BDG). Tests designed to detect these markers may be deployed as part of a pre‐emptive treatment strategy or may be used to facilitate selection of empiric treatment for symptomatic at‐risk people.
Index test(s)
Non‐invasive, non‐culture‐based methods for diagnosing invasive fungal disease have the potential for significant clinical utility (Powers‐Fletcher 2016). BDG is a cell wall polysaccharide found in a wide variety of medically important fungi including Candida species (spp) (Aspergillus spp and Pneumocystis jirovecii; important exceptions are Mucorales, Cryptococcus spp, and the yeast form of Blastomyces (Wright 2011)). Assays designed to detect BDG in human serum have been used both as an adjunct for diagnosis of IFI and for serial surveillance during periods of risk. Commercially available assays include the Fungitell and Glucatell assays (Associates of Cape Code, Falmouth, MA, USA), which are used in America and in Europe, as well as the Fungitec‐G assay (Seikagaku Kogyo Corporation, Tokyo, Japan) and the Wako test (Wako Pure Chemical Industries, Osaka, Japan), both of which are used in Japan. The Dynamiker Fungus assay (Dynamiker Biotechnology Ltd, Tianjin, China) is a new test that was recently developed in China.
These assays are based on the ability of the BDG molecule to induce clot formation in the hemolymph of horseshoe crabs. BDG activates Factor G, which is a serine protease in the horseshoe crab coagulation cascade. Activated Factor G then converts an inactive proclotting enzyme to its active form, which, in turn, cleaves an artificial substrate that can be detected. The assays differ in the substrate used for detection. The Fungitell and Glucatell assays use a chemiluminescent method. The Glucatell test differs from the Fungitell test in that the Glucatell reagent is processed to eliminate Factor C. This makes the Glucatell test more specific for BDG linkages. The Glucatell reagent does not react to other polysaccharides including beta‐glucans with other glycosidic linkages. For the other assays, Dynamiker Fungus uses a spectrophotometric method, the Wako assay is a turbidometric method, and Fungitec‐G is a colorimetric method. Each of these tests uses a different interpretive cut‐off value. In the Fungitell and Glucatell assays, a value of 60 pg/mL or less is negative, a value of 60 to 80 pg/mL is equivocal, and a value of 80 pg/mL or more is positive. For the Fungitec‐G assay, a value greater than 20 pg/mL is considered positive, whereas for Wako, it is 11 pg/mL. The Dynamiker Fungus test considers values above 95 pg/mL as positive. These differences may be due to the fact that the reagents are obtained from different genera of horseshoe crabs (Fungitell reagents are extracted from Limulus polyphemus, whereas Fungitec and Wako reagents are extracted from Tachypleus tridentatus).
Studies vary in the criteria used for BDG positivity. For example, a single positive BDG result may be sufficient to classify a person as "BDG positive" in some studies, whereas other studies may use more stringent criteria such as two consecutive positive tests, or two positive tests within a specified time period. Similarly, studies use different sampling plans, which may affect test performance. Some studies may use a single sample, whereas others may use a prolonged sampling regimen (e.g. twice‐a‐week sampling for several weeks).
Clinical pathway
Presentation
The fungi capable of causing invasive disease in humans are a diverse group of eukaryotic microorganisms including yeasts, molds, and dimorphic fungi. Candida and Aspergillus are the pathogens most commonly diagnosed after solid organ transplantation or critical care (Pappas 2010), and Aspergillus and other filamentous fungi predominate after hematopoietic stem cell transplantation or as a complication of cytotoxic chemotherapy for hematologic malignancy (Kontoyiannis 2010; Neofytos 2009). In addition, Pneumocystisjirovecii remains an important opportunistic pathogen that affects people with AIDS and those receiving cytotoxic or immunosuppressive therapy. Clinical signs and symptoms of IFI vary widely. The clinical presentation of IFI varies widely according to the infecting pathogen, the overall net state of immunosuppression (i.e. the host), and the site and severity of infection. Invasive candidiasis comprises a spectrum of diseases including bloodstream infection and deep‐seated infection (e.g. intra‐abdominal abscess), which may occur independently or concurrently. The filamentous fungi typically present with pulmonary or sino‐cerebral disease. Pneumonia is the most common manifestation of Pneumocystis.
Standard diagnostic practice
In general, the current approach to IFI diagnosis combines a variety of complementary testing modalities. Diagnostic imaging helps clinicians to identify potential sites of infection. Cultures of blood, body fluids, and/or tissue are performed in combination with molecular tests and serum fungal biomarkers in an attempt to detect and identify fungi. Use of targeted imaging may help to guide biopsy sampling of infected tissue for histopathology.
Alternative test(s)
Classical methods of diagnosis include direct stains for fungi (i.e. calcofluor white, cytology, or histopathology) and fungal culture. Despite availability of a variety of test modalities, the clinical utility of this routine testing is often limited. For example, cultures are slow and relatively insensitive. Positive results, however, are useful for definitive organism identification and antifungal susceptibility testing. Cytology and calcofluor white stains applied to body fluid also lack sensitivity. Furthermore, deciphering colonization from invasive disease can be extremely difficult when samples are obtained from non‐sterile body sites such as the respiratory tract. Visualizing fungal elements in tissue remains the diagnostic gold standard for IFI, but invasive testing may not be feasible for critically ill or coagulopathic people. Additionally, biopsy results may be affected by sampling error, and current staining techniques are neither genus‐ nor species‐specific. This level of organism discrimination, however, is essential for selection of optimal antifungal therapy.
Detection of fungal biomarkers including nucleic acid and cell wall components helps support the diagnosis of IFI. Rapid polymerase chain reaction (PCR) techniques targeting fungi have been widely applied in clinical practice (Arvanitis 2014; Avni 2011; Fan 2013; Lu 2011a; Mengoli 2009; Sun 2011). Unfortunately, laboratory‐developed PCR tests lack standardization, and commercial assays are not widely available. The Candida T2 assay (T2 Biosystems, Lexington, MA, USA) is a rapid and accurate test for the detection of Candida DNA directly in whole blood (Tang 2019). Unfortunately, this test targets only the five most common Candida species and requires expensive instrumentation/reagents. Detecting mannan antigen and anti‐mannan antibodies also has potential utility for the diagnosis of invasive candidiasis, but commercial assays are mainly limited to European markets (Mikulska 2010). Last, lateral flow assays for Aspergillus GM have been developed for use with serum and bronchoalveolar lavage samples (Mercier 2019; Verdaguer 2007). A potential benefit of antigens like GM and BDG is that these polysaccharides can be detected non‐invasively in blood at an early stage of infection, whereas release of fungal DNA may be negligible in initial phases of the disease (Monique 2006). Alternatively, limitations of the Aspergillus GM test include limited sensitivity in non‐neutropenic patient populations and potential cross‐reactivity with closely related fungi or other antigenically similar substances (Demiraslan 2017; Verweij 2006; Viscoli 2004).
Rationale
Here we perform an updated review of the BDG literature with a focus on immunocompromised or critically ill people. BDG suffers from many of the same limitations as the Aspergillus GM test. Sensitivity may vary by population and organism type, and false positives are thought to result from cross‐reacting substances in certain medications or materials, or possibly in bacteria (Marty 2006; Tran 2016; Wright 2011). Thus, it is important to understand the diagnostic performance of BDG across a variety of at‐risk populations and testing strategies. Our objective was to provide summary estimates of the diagnostic performance of BDG that could be used to inform future guideline updates and serve as a benchmark for emerging diagnostics tests such as PCR and the Candida T2 assay. Both BDG and the Aspergillus GM test results have been incorporated into the revised EORTC/MSG criteria for probable IFI.
Objectives
Primary objective
To compare the diagnostic accuracy of commercially available tests for serum BDG to detect selected invasive fungal infections (IFIs) among immunocompromised or critically ill people.
Secondary objectives
To assess possible sources of heterogeneity that could affect sensitivity and specificity estimates in this study (see Investigations of heterogeneity).
Methods
Criteria for considering studies for this review
Types of studies
Published peer‐reviewed studies that compared the results of BDG tests against a clearly defined reference standard (EORTC criteria or equivalent) for diagnosis of IFI were included in the analysis.
We included the following types of studies.
-
Retrospective studies in which BDG samples were collected from consecutive people at risk.
-
Prospective studies in which BDG samples were collected from consecutive people at risk.
-
Case‐control studies in which controls were people at risk.
We excluded the following types of studies.
-
Case reports or case series.
-
Studies reported only as meeting abstracts.
-
Case‐control studies using healthy controls, due to the high risk of spectrum bias.
-
Animal studies.
Participants
Study participants included the following categories of immunocompromised people, with results for both the index test and the reference test.
-
Those with cancer, specifically:
-
patients with hematologic malignancies; those receiving stem cell transplants, chemotherapeutics, or other immunosuppressive drugs; and
-
patients receiving chemotherapy.
-
-
Those receiving prolonged immunosuppressive therapy for:
-
solid organ transplant; or
-
connective tissue diseases.
-
-
Individuals with congenital or acquired immune disorders, including:
-
human immunodeficiency virus (HIV) or acquired immunodeficiency syndrome (AIDS); or
-
inherited immune disorders.
-
-
People receiving treatment in the intensive care unit (ICU).
There was no restriction on age or comorbidities.
Index tests
We included studies that used any commercially available BDG tests that were approved for clinical use and followed the manufacturer's recommended cut‐off values.
-
Fungitell (cut‐off: 80 pg/mL).
-
Glucatell (cut‐off: 80 pg/mL).
-
Wako (cut‐off: 11 pg/mL).
-
Fungitec‐G (cut‐off: 20 pg/mL).
-
Dynamiker Fungus (cut‐off: 95 pg/mL).
Target conditions
The target condition included proven or probable IFI due to Aspergillus or Candida, or other IFIs as defined by EORTC/MSG criteria (De Pauw 2008). It should be noted that EORTC/MSG criteria were developed for people with malignancy and for hematopoietic stem cell transplant recipients; these criteria are not easily generalizable to all risk groups and/or fungal diseases. Therefore, Pneumocystis jirovecii pneumonia (PJP) and Candida studies outside of the cancer population were also included if proven infection was determined by microscopy (Pneumocystis) or by sterile site culture (Candida). People with colonized Candida were considered as non‐cases.
Reference standards
We included studies that used the following reference standards for invasive fungal disease.
-
Autopsy.
-
EORTC/MSG criteria from either 2002 or 2008 guidelines (Ascioglu 2002; De Pauw 2008).
-
Microscopy or sterile site culture for proven PJP or Candida infection, respectively.
The criteria for proven IFI are listed below.
Microscopic analysis of sterile material
-
Molds: histopathologic, cytopathologic, or direct microscopic examination of a specimen obtained by needle aspiration or biopsy in which hyphae or melanized yeast‐like forms are seen accompanied by evidence of associated tissue damage
-
Yeast: histopathologic, cytopathologic, or direct microscopic examination of a specimen obtained by needle aspiration or biopsy from a normally sterile site
Culture of sterile material
-
Molds: recovery of a mold or black yeast by culture obtained by sterile procedure from a normally sterile and clinically or radiologically abnormal site consistent with an infectious disease process, excluding bronchoalveolar lavage fluid, a cranial sinus cavity specimen, and urine
-
Yeast: recovery of a yeast by culture of a sample obtained by a sterile procedure (including a freshly placed drain < 24 hours) from a normally sterile site showing a clinical or radiological abnormality consistent with an infectious disease process
Blood culture
-
Molds: blood culture that yields a mold in the context of a compatible infectious disease process
-
Yeast: blood culture that yields yeast or yeast‐like fungi
The criteria for probable IFI include host factors (e.g. receipt of allogeneic stem cell transplant), clinical criteria, and mycologic criteria. As of 2008, the EORTC/MSG mycologic criteria now include biomarker tests such as BDG or GM. This creates a possible source of incorporation bias because the index test (BDG) is sometimes used as part of the reference standard for possible IFI. Therefore, we excluded studies that used BDG as part of the reference standard.
Search methods for identification of studies
Electronic searches
An initial search to identify articles related to the diagnostic accuracy of BDG using the search strategies described in Appendix 1 and Appendix 2 was completed in April 2017. The last update was performed on 26 June 2019.
-
MEDLINE (R) via Ovid (1946 to June week 3, 2019).
-
Embase via Ovid (1980 to week 25, 2019).
Because the commercial BDG test was not implemented until 1995, the search was restricted to articles published in 1995 or later. The search was not restricted with respect to language or study design.
We performed an additional electronic search based on the set of potentially relevant studies identified in June 2017 from MEDLINE and Embase. This search was a forward and backward citation search to identify all studies cited by or citing the set of potentially relevant studies. This citation search was performed using SCOPUS on June 6, 2017.
Data collection and analysis
Selection of studies
Two review authors (RLS, SKW) screened the titles and abstracts of all articles to identify potentially relevant studies. Disagreements were resolved by discussion.
Each study in the set of potentially relevant studies was given a full‐text review. An initial abstract form (see Appendix 3) was used to retrieve preliminary information that was used to determine whether the article met the inclusion criteria. Full‐text review was performed independently by two review authors (RLS, SKW). This included information on study design, participant population, sample type, and IFI category (proven, probable, or possible), and whether EORTC/MSG, autopsy, or another method was used as the reference standard. These were reviewed together, and any discrepancies were resolved by discussion between the two review authors. Foreign language articles were assessed by a native speaker with scientific training (but not screened in duplicate) or were translated using Google Translate and reviewed by two review authors (RLS, SKW). The review authors who determined relevance were not blinded to trial authors, publishing journal, or results.
Data extraction and management
Two review authors (SKW, BSW) extracted additional information from the selected studies on the condition (cancer, ICU, organ transplant, etc.), study design (prospective consecutive, retrospective consecutive, or case‐control), sample type (serum, urine, bronchoalveolar lavage fluid, or cerebrospinal fluid), fungal organism (mixed IFI, Candida, Aspergillus, or Pneumocystis jirovecii), and reference standard used (EORTC or study‐specific), using the data abstract form provided in Appendix 4. True‐positive, false‐positive, false‐negative, and true‐negative values were obtained to calculate sensitivity and specificity estimates. Additional information extracted during the full‐text review included use of antifungal agents, sampling protocol, the assay used and the cut‐off value, the number of positive samples needed to constitute a positive test result, and age of the population. All data were recorded, and discrepancies were resolved through discussion or by a third review author (RLS).
Assessment of methodological quality
We assessed study quality using the Quality Assessment of Studies of Diagnostic Accuracy‐Revised (QUADAS‐2) tool (Whiting 2011). Bias was assessed in four domains: participant selection, index test, reference standard, and flow/timing, and applicability was assessed in the first three domains only (participant selection, index test, and reference standard). Both were independently graded as low, high, or unclear quality by two review authors (SKW, BSW), using the interpretations listed in Appendix 5. Discrepancies were then resolved by discussion or were moderated by a third review author (RLS).
Statistical analysis and data synthesis
We transferred data into 2 × 2 matrices to calculate sensitivity and specificity for each study. We used reported values of true positives/negatives and false positives/negatives to calculate sensitivity and specificity. If these values were not reported, we back‐extrapolated using reported sensitivity and specificity values.
Individual study data were presented graphically as forest plots by assay type. Studies were also plotted in receiver operating characteristic (ROC) space. We used the bivariate random‐effects model for meta‐analysis of the pairs of sensitivity and specificity (Reitsma 2005; van Houwelingen 1993). We restricted the analysis to standard cut‐off values recommended by test manufacturers. All statistical analyses were completed using Stata v.14.2 (Stata Corporation, College Station, TX, USA). However, with the exception of studies involving Candida, we were unable to perform a formal meta‐analysis for fungal groups because of high heterogeneity within the data, which prevented estimations of summary accuracy. This diversion from the protocol is explained in the Differences between protocol and review section.
Investigations of heterogeneity
When heterogeneity is present, subgroup analysis can be performed to determine the source. Heterogeneity between studies was supposed to be assessed by meta‐regression performed on pre‐selected covariates. We planned to investigate whether the following covariates or patterns of covariates had contributed to this.
-
Variation across participant subgroups (people with cancer or in the ICU compared to other groups; pediatric versus adult studies).
-
Variability in the number of positive results used to define a positive test (single positive result versus two consecutive positive samples).
-
Differences due to sampling strategies (single sample taken versus multiple samples collected over the length of stay).
-
Study design factors, including prospective versus retrospective and consecutive versus case‐control.
-
Test interference (antifungal prophylaxis, pre‐emptive therapy, etc.).
-
Definition of IFI: using proven and probable IFI (as defined above) as the definition of the target condition, and comparing it only to proven IFI when compared to all other categories and using proven, probable, and possible IFI compared to no IFI.
Because we were unable to perform a formal meta‐analysis, we used ROC plots to visually investigate these potential sources of heterogeneity. This diversion from the protocol is explained in the Differences between protocol and review section.
Sensitivity analyses
We planned to compare pooled sensitivity and specificity estimates for studies that had low overall risk of bias versus those with at least one high risk of bias. However, we were unable to do this because a formal meta‐analysis was not performed. This diversion from the protocol is explained in the Differences between protocol and review section.
Results
Results of the search
Through the literature search in MEDLINE and Embase, we identified 10,354 references. Duplicate references were identified and removed (N = 1671), resulting in 8683 articles. The initial review of titles and abstracts yielded 211 potentially relevant articles (Figure 1).
Study flow diagram.
The citation search in SCOPUS (forward and backward search based on potentially relevant references) identified 6726 references. Of these, 747 were duplicates. Also, 124 references had extensive missing information and could not be retrieved, yielding a final number of 5855 references.
We compared results of the citation search (SCOPUS) with results of the initial search (MEDLINE and Embase) and identified 1162 references that had already been included in the initial search. We reviewed titles and abstracts of the remaining 13,376 references. In total, we identified 233 potentially relevant studies (211 from the MEDLINE/Embase searches, 20 from the citation search, and 2 additional articles through handsearching) that were initially reviewed for inclusion using the abstract form in Appendix 3.
After reviewing the 233 potentially relevant studies, we identified 100 studies that met study criteria in which we conducted a full‐text review (Appendix 4). We contacted two study authors to receive clarification on possible IFI results and study design, which we received. At the conclusion of the full‐text review, we identified 49 studies to be included in the systematic review (Table 1; Characteristics of included studies). A flow diagram of the selection process is shown in Figure 1.
| Study name | Study design | Underlying condition | Fungal type | Test brand | Samples taken | Reference standard |
| Prospective consecutive | ICU | Mixed | Fungitell | Single | EORTC 2008 | |
| Prospective consecutive | Organ transplant | Mixed | Fungitell | Multiple | EORTC 2008 | |
| Retrospective consecutive | Cancer | Mixed | Fungitell | Single | EORTC 2008 | |
| Prospective consecutive | Cancer | Aspergillosis | Glucatell | Multiple | EORTC 2008 | |
| Prospective consecutive | Cancer | Mixed | Fungitell | Single | EORTC 2008 | |
| Prospective consecutive | Mixed at‐risk | Mixed | Fungitell | Multiple | EORTC 2008 | |
| Retrospective case‐control | ICU | Candida | Fungitell | Single | Culture from blood or sterile site | |
| Retrospective consecutive | Mixed at‐risk | PJP | Fungitell | Single | Microscopy | |
| Retrospective case‐control | ICU | Aspergillosis | Fungitell | Unknown | EORTC 2008 | |
| Prospective consecutive | ICU | Candida | Fungitell | Single | EORTC 2008 | |
| Retrospective case‐control | Mixed at‐risk | PJP | Wako | Single | PCR | |
| Retrospective case‐control | Cancer | Mixed | Fungitell | Multiple | EORTC 2008 | |
| Prospective consecutive | Cancer | Aspergillosis | Fungitell | Multiple | EORTC 2008 | |
| Retrospective consecutive | ICU | Candida | Fungitell | Single | Culture from blood or sterile site | |
| Prospective consecutive | Cancer | Mixed | Fungitell | Single | EORTC 2008 | |
| Prospective consecutive | Cancer | Aspergillosis | Fungitell | Multiple | EORTC 2002 | |
| Retrospective consecutive | Cancer | Mixed | Glucatell | Multiple | EORTC 2008 | |
| Prospective consecutive | Cancer | Mixed | Glucatell | Multiple | EORTC 2008 | |
| Prospective consecutive | ICU | Mixed | Fungitell | Multiple | EORTC 2008 | |
| Prospective consecutive | Cancer | Aspergillosis | Fungitec | Unknown | EORTC 2002 | |
| Prospective consecutive | Cancer | Aspergillosis | Glucatell | Multiple | EORTC 2002 | |
| Unknown | Cancer | Aspergillosis | Fungitec | Multiple | Study‐specific, comparable to EORTC | |
| Prospective consecutive | Cancer | Aspergillosis | Wako | Multiple | EORTC 2002 | |
| Prospective consecutive | Cancer | Mixed | Fungitell | Multiple | EORTC 2008 | |
| Retrospective consecutive | Cancer | Mixed | Fungitell | Multiple | EORTC 2008 | |
| Retrospective consecutive | ICU | Aspergillosis | Fungitell | Unknown | EORTC 2008 | |
| Retrospective consecutive | ICU | Aspergillosis | Fungitell | Single | EORTC 2008 | |
| Prospective consecutive | ICU | Candida | Fungitell | Multiple | Culture from blood or sterile site | |
| Retrospective consecutive | Cancer | Mixed | Fungitec | Single | EORTC 2008 | |
| Retrospective consecutive | ICU | Candida | Fungitell | Single | EORTC | |
| Prospective consecutive | ICU | Mixed | Fungitell | Single | Study‐specific, comparable to EORTC | |
| Prospective consecutive | ICU | Candida | Fungitell | Multiple | Culture | |
| Retrospective case‐control | Cancer | Aspergillosis | Fungitell | Multiple | EORTC 2008 | |
| Retrospective consecutive | Cancer | Aspergillosis | Fungitell | Multiple | EORTC 2008 | |
| Prospective consecutive | ICU | Candida | Fungitell | Multiple | Culture | |
| Unknown | Cancer | Mixed | Glucatell | Multiple | EORTC 2002 | |
| Retrospective case‐control | Cancer | Mixed | Fungitell | Single | EORTC 2002 | |
| Retrospective case‐control | Mixed | Candida | Fungitell | Single | Culture from blood or sterile site or pathology | |
| Prospective consecutive | ICU | Mixed | Fungitell | Multiple | EORTC 2008 | |
| Prospective consecutive | Cancer | Mixed | Fungitell | Multiple | EORTC 2002 | |
| Retrospective consecutive | Mixed at‐risk | Mixed | Fungitell | Single | EORTC 2008 | |
| Retrospective consecutive | HIV | PJP | Fungitell | Single | Microscopy/PCR | |
| Prospective consecutive | Cancer | Mixed | Wako | Multiple | EORTC 2002 | |
| Prospective consecutive | ICU | Candida | Dynamiker | Single | Culture from blood or sterile site | |
| Prospective consecutive | Organ transplant | Mixed | Fungitell | Multiple | EORTC 2008 | |
| Prospective consecutive | ICU | Mixed | Fungitell | Multiple | EORTC 2008 | |
| Prospective consecutive | Mixed at‐risk | Mixed | Fungitell | Single | EORTC 2008 | |
| Retrospective case‐control | Cancer | Candida | Fungitell | Multiple | Culture from blood or sterile site | |
| Retrospective case‐control | Mixed at‐risk | Mixed | Dynamiker | Single | EORTC 2008 |
EORTC: European Organization for Research and Treatment of Cancer; ICU: intensive care unit; PCR: polymerase chain reaction; PJP: Pneumocystis jirovecii pneumonia.
Basic features of included studies
Details of the included studies are presented in Table 2, 'Overall characteristics of included studies'. We included 49 studies with a total of 6244 participants.
| Characteristic | n | Percentage |
| Underlying condition | ||
| Cancer HIV/AIDS ICU Mixed at‐risk Organ transplant | 23 1 16 7 2 | 46.9% 2.0% 32.7% 14.3% 4.1% |
| Age of patients | ||
| Adult Neonate Pediatric Mixed Unknown | 26 3 3 5 12 | 53.1% 6.1% 6.1% 10.2% 24.5% |
| Study design | ||
| Prospective consecutive Retrospective consecutive Retrospective case‐control | 26 12 11 | 53.1% 24.5% 22.4% |
| Fungal type | ||
| Aspergillus only Candida only PJP only Mixed fungal types | 12 10 3 24 | 24.5% 20.4% 6.1% 49.0% |
| Test brand | ||
| Fungitell Glucatell Fungitec‐G Wako Dynamiker Fungus | 36 5 3 3 2 | 73.5% 10.2% 6.1% 6.1% 4.1% |
| Sampling strategy | ||
| Single sample Multiple samples Unknown | 20 26 3 | 40.8% 53.1% 6.1% |
| Reference standard used | ||
| EORTC Proven Candida PJP microscopy/PCR Study‐specific | 36 8 3 2 | 73.5% 16.3% 6.1% 4.1% |
| Low risk of bias | ||
| Participant selection Index test Reference standard Flow and timing | 35 49 41 41 | 71.4% 100.0% 83.7% 83.7% |
EORTC: European Organization for Research and Treatment of Cancer; ICU: intensive care unit; PCR: polymerase chain reaction; PJP: Pneumocystis jirovecii pneumonia.
-
Participants: nearly half of the studies involved people with hematologic and oncologic diseases (N = 23; 47%), followed by people in the ICU (N = 16) and mixed at‐risk cases (N = 7) (see Table 2). A majority (N = 26; 53%) were focused on adult populations, five had both adult and pediatric cases, and six focused solely on pediatrics (N = 3) or neonates (N = 3).
-
Study design: of the 49 studies, a little more than half (N = 26; 53%) were prospective in design. Of the 23 retrospective studies, 12 were consecutive, with the remaining employing a case‐control design.
-
Assay characteristics: a majority (N = 36) used the Fungitell assay, followed by Glucatell (N = 5), Fungitec‐G (N = 3), Wako (N = 3), and Dynamiker Fungus (N = 2).
-
Sampling: almost all studies reported only estimates based on a single positive test, although two studies did provide results based on two consecutive positive samples. Sampling design varied greatly between studies, ranging from a single sample collection (41%) to multiple samples collected over several weeks. Studies that collected multiple samples reported differing criteria for the classification of a positive BDG result, such as using the first sample collected or the highest BDG value recorded.
-
Organisms: studies that included all types of organisms were most common (N = 24), although several focused exclusively on Candida (N =10) or Aspergillus (N = 12). Per the selection criteria, all studies either used the EORTC/MSG criteria (N = 36) or followed the criteria used in the diagnosis of proven PJP or candidiasis.
-
Language: almost all studies (N = 47) were published in English, with the two remaining articles published in Chinese and Japanese.
Excluded studies
From our full‐text review, we identified 51 studies to be excluded (see Characteristics of excluded studies) for the following reasons.
-
Unable to determine 2 × 2 cell counts for overall sensitivity and specificity estimates (N = 17).
-
Used cut‐off values that did not follow the manufacturer's recommended level or utilized two cut‐off values to determine a positive test (N = 11).
-
Included BDG tests as part of the reference standard or did not follow EORTC/MSG guidelines (N = 7).
-
Did not meet inclusion criteria for the study population or inclusion criteria were unknown (N = 6).
-
Included probable (PJP or Candida) or possible IFI cases in the IFI definition, which could not be separated (N = 5).
-
Other reasons (N = 4).
Methodological quality of included studies
Thirty‐two studies had no concerns regarding risk of bias or applicability among the four QUADAS‐2 domains (Figure 2 and Figure 3). Details on bias for individual studies are provided in the Characteristics of included studies table. For studies that had high risk of bias or concerns regarding applicability, this was due mainly to (1) case‐control design (Cornu 2018; De Vlieger 2011; Dichtl 2018; Fontana 2012; Metan 2012; Persat 2008; Pini 2019; Verduyn Lunel 2009; White 2017), and (2) exclusion of possible IFI cases from the findings (Hammarstrom 2015; Hammarstrom 2018; Jin 2013; Theel 2013).
Risk of bias and applicability concerns summary: review authors' judgements about each domain for each included study.
Risk of bias and applicability concerns graph: review authors' judgements about each domain presented as percentages across included studies.
Thirty‐six of the studies failed to report the time interval between BDG testing and the reference standard. Only 13 studies provided a time frame; however, we still judged all other studies as low bias if other criteria were met (Figure 2).
Due to study design criteria, all studies pre‐specified cut‐off values or reported values that met the manufacturer's recommendations. BDG is an objective quantitative test that is generally performed without knowledge of the participant's true infection status. Therefore, failure to blind investigators to the reference test poses little risk of bias with respect to interpretation of the BDG test result. Thus, even if the study did not report blinding, we considered both the index test and reference standard domains to still be at low bias (Figure 2; Figure 3).
The reference standard was likely to classify IFIs correctly by using either EORTC/MSG criteria or confirmation by culture or microscopy. The EORTC/MSG criteria were revised in 2008. One of the important changes was that BDG was added as a criterion for IFI. Thus, to avoid incorporation bias, studies had to exclude BDG from the diagnostic criteria. Forty‐two studies reported that they did not incorporate BDG testing in the reference standard, and we excluded two studies that included BDG as part of the reference test. It is unclear in seven studies whether BDG testing had been excluded (Ceesay 2015; De Vlieger 2011; Fontana 2012; Gupta 2017; Lahmer 2016a; Lahmer 2016b; White 2017). We elected to include these studies. Most studies were careful not to incorporate BDG testing, and we assumed that these studies most likely would have done so as well.
Findings
The prevalence of IFI ranged from 4% to 59% among all studies (mean 23%, 95% confidence interval (CI) 18% to 28%). In addition, estimates of sensitivity and specificity varied widely. Due to the high degree of heterogeneity between studies, we did not perform a formal meta‐analysis, with the exception of Candida.
-
Fungitell (36 studies): sensitivity for individual studies ranged between 27% and 100% and specificity range between 0% and 100% (Figure 4; Figure 5). A large amount of uncertainty was noted in study estimates, as evidenced by wide confidence intervals in the forest plot (Figure 4). Because IFI is relatively rare, many studies had a small number of positive cases. Koo had the largest study, with a study population of 871 (Koo 2009).
-
Glucatell (5 studies): study estimates for sensitivity and specificity for Glucatell also ranged widely. Sensitivity ranged from 50% to 92%, and specificity ranged from 41% to 94%, among the 5 studies (Figure 4; Figure 5).
-
Wako (3 studies): only 3 studies used the Wako assay at the manufacturer's specified cut‐off level. Two studies reported lower sensitivities (55% and 50%, respectively) but higher specificities (98% and 89%) (Figure 4; Figure 5) (Kawazu 2004; Senn 2008). Dichtl 2018 reported fairly high sensitivity (86%) and specificity (100%) among a group of 98 people.
-
Fungitec‐G (3 studies): estimates for the 3 studies using Fungitec ranged from 67% to 88% for sensitivity and from 60% to 85% for specificity (Figure 4; Figure 5).
-
Dynamiker (2 studies): only 2 recent studies had published results regarding Dynamiker Fungus (Figure 4; Figure 5). White 2018 reported sensitivity of 81% and specificity of 78%, and Shabaan 2018 reported sensitivity and specificity of 64% and 80%, respectively.
Forest plot of tests: Fungitell, Glucatell, Wako, Fungitec, Dynamiker.
Summary ROC plot of tests: 1 Fungitell, 2 Glucatell, 3 Wako, 4 Fungitec, 5 Dynamiker.
We included 10 studies in the meta‐analysis, from which an estimate for Candida could be obtained. Estimated sensitivity and specificity for these studies was 81.3% (95% CI 75.3% to 86.0%) and 64.1% (95% CI 55.6% to 71.8%), respectively. Almost all (N = 9; 90%) used Fungitell, with 8 of the 10 studies involving people in ICU settings. Forty per cent utilized multiple samples, and the remainder relied on a single test.
Investigations of heterogeneity
Heterogeneity was assessed by ROC plots that examined differences in individual sensitivity and specificity estimates by participant population, fungal organism, reference standard, and single versus multiple testing.
Heterogeneity could not be explained by the participant population (Figure 6). We restricted this analysis to a single test platform to limit a potential source of variation. We selected the Fungitell assay because it was the most commonly used test platform (36 of 49 studies). In 13 studies involving participants with cancer, sensitivity ranged from 33% to 100% and specificity ranged from 0% to 100%. In 15 studies involving participants who had been admitted to the ICU, sensitivity ranged from 27% to 100% and specificity ranged from 20% to 94%. Finally, in 5 studies with a mixture of participants, sensitivity ranged from 40% to 100% and specificity ranged from 42% to 91%. All participant groups had a wide range of sensitivity and specificity. Considerable overlap could be seen in the ranges of sensitivity and specificity for each group. It was not possible to identify an underlying condition that was associated with higher or lower levels of sensitivity or specificity.
Summary ROC plot of underlying medical conditions for Fungitell studies.
Heterogeneity could not be explained by the reference standard (Figure 7). This analysis was also restricted to studies performed with the Fungitell assay. In 26 tests using EORTC criteria, sensitivity ranged from 27% to 100% and specificity ranged from 0% to 100%. In seven studies testing for Candida, sensitivity ranged from 76% to 100% and specificity ranged from 41% to 81%.
Summary ROC plot of reference standard for Fungitell studies.
In some studies, the BDG test was performed once, and in other studies, BDG testing was performed multiple times (e.g. twice a week). Heterogeneity could not be explained by the number of tests (Figure 8). This analysis was restricted to studies performed with the Fungitell test. In 16 studies that used a single sample, sensitivity ranged from 27% to 100% and specificity ranged from 12% to 98%. In 18 studies that used multiple tests per person, sensitivity ranged from 33% to 100% and specificity ranged from 0% to 100%. Both groups had a wide range of sensitivity and specificity with substantial overlap. It was not possible to identify a sampling policy that was associated with higher or lower levels of sensitivity or specificity.
Summary ROC plot of single versus multiple sampling for Fungitell.
Studies that focused on Candida infection did appear to be more homogeneous than those focused on other fungal organisms (Figure 9); therefore, we obtained a summary estimate for these studies.
Summary ROC plot of fungal organisms for Fungitell studies.
Discussion
Summary of main results
Literature on the use of (1→3)‐β‐D‐glucan (BDG) for diagnosis of invasive fungal infection (IFI) shows wide variation in diagnostic accuracy. Sensitivity ranged from 27% to 100%, and specificity ranged from 0% to 100%. Because of this variation, we did not perform a formal meta‐analysis, with the exception of Candida studies.
There were many potential sources of heterogeneity. These include study design (case‐control retrospective, prospective), differences in populations (immunodeficient versus critically ill), sampling (single sample, multiple samples, monitoring with two samples per week), assays (Fungitell, Glucatell, etc.), target organisms (all IFI, Candida, Aspergillus, etc), and threshold for positivity (one positive BDG test, two consecutive positive BDG tests). Application of European Organization for Research and Treatment of Cancer (EORTC) criteria is another potential source of heterogeneity. The accuracy with which physicians perform this task may vary, and because the number of physicians in any study is low, differences in classification accuracy are unlikely to average out. We are not aware of any agreement of studies on EORTC criteria. This variation made it difficult to obtain meaningful estimates of sensitivity and specificity. Thus, it is not possible to predict how the BDG test will perform in a particular context.
Going forward, it would be helpful if studies limited variation in these factors. Prospective studies should be preferred over case‐control and retrospective studies. Prospective studies are more closely aligned with the clinical context and allow various sampling policies to be compared in a single study. For example, one could perform twice‐weekly sampling and compare the diagnostic accuracy of the first positive BDG result, two consecutive BDG‐positive results, positive BDG when a person is first symptomatic, etc. It is not clear whether studies on individual organisms are helpful. Several studies focused on infections in a single organism. Although such studies provide useful knowledge regarding test performance, they do not address a clinically relevant question. The clinical question that is addressed by BDG testing is whether a person has an IFI rather than whether a person is infected with a particular organism. It might be better to conduct instrument comparisons in laboratory studies rather than in clinical studies.
We found that the quality of studies was generally good. Risk of bias was generally low. We did exclude a number of case‐control studies that included healthy controls. These study designs produce inflated estimates of sensitivity and specificity due to spectrum bias (White 2019). There is some room for improvement in reporting. Studies should not include BDG as part of the reference test and should explicitly state this. Also, it would be helpful if studies reported results for all four EORTC categories (proven, probable, possible, none). One must aggregate categories to calculate sensitivity and specify; however, to facilitate meta‐analysis, results should be available as individual categories. Studies should report timing of the BDG test relative to the reference test, and whether the reference test was blinded to the BDG test result.
Comparison of our results with other meta‐analyses
Four meta‐analyses on the diagnostic accuracy of BDG have been previously published (He 2015; Karageorgopoulos 2011; Lu 2011b; White 2019). These meta‐analyses studied the use of BDG in similar populations of people (immunocompromised) and included between 13 and 28 studies. Our analysis summarized 49 studies, which reflects the large number of studies conducted in the past five years. Previous reviews have also reported high levels of heterogeneity.
Strengths and weaknesses of the review
This review represents the most up‐to‐date systematic assessment of BDG test performance. The high level of heterogeneity is a significant limitation. Current BDG diagnostic literature remains impacted by variability in study design, heterogenous populations, limited information on baseline use of antifungal therapy or other potential assay interferences, and lack of consistently robust adjudication of potential colonization versus invasive disease. In addition, microscopy for PJP and culture confirmation of Candida are imperfect reference standards that may miss true cases of invasive disease and may impact calculations of test specificity.
There was considerable variation in the prevalence of probable/proven IFI (range 4% to 59%). This could reflect differences in populations or differences in interpretation of the reference standard. Variation in interpretation of the reference standard is a potential source of heterogeneity. We used simple descriptive methods to investigate potential sources of heterogeneity; however, this was largely unsuccessful. Future work might benefit from the application of a latent‐class meta‐analysis, which could potentially address the issue of variable, imperfect reference standards across studies.
Applicability of findings to the review question
We summarized the diagnostic accuracy of several BDG tests. We found significant heterogeneity between study estimates. Given this variability, a summary estimate is unlikely to be applicable at any given location. We were unable to make a meaningful comparison between different commercial tests, and we were unable to determine factors that affect diagnostic accuracy (e.g. population, positivity criteria, sampling).
Study flow diagram.
Risk of bias and applicability concerns summary: review authors' judgements about each domain for each included study.
Risk of bias and applicability concerns graph: review authors' judgements about each domain presented as percentages across included studies.
Forest plot of tests: Fungitell, Glucatell, Wako, Fungitec, Dynamiker.
Summary ROC plot of tests: 1 Fungitell, 2 Glucatell, 3 Wako, 4 Fungitec, 5 Dynamiker.
Summary ROC plot of underlying medical conditions for Fungitell studies.
Summary ROC plot of reference standard for Fungitell studies.
Summary ROC plot of single versus multiple sampling for Fungitell.
Summary ROC plot of fungal organisms for Fungitell studies.
| Participants/Populations: immunocompromised people at risk for invasive fungal infections | |||||
| Prior testing: none | |||||
| Settings: hospital setting | |||||
| Index test: commercially available serum BDG test | |||||
| Importance: test needed to accurately detect fungal infections in susceptible people at an early enough stage to facilitate antifungal treatment | |||||
| Reference standard: EORTC/MSG criteria, or by microscopy or autopsy | |||||
| Studies: 49 studies with 6244 participants | |||||
| 1. Test assay | |||||
| Test/Subgroup | No. of participants (studies) | Overall sensitivity (range) | Overall specificity (range) | Implications | Quality and comments |
| Fungitell | 4316 (36) | 27% to 100% | 0 to 100% | Wide variation in sensitivity | |
| Glucatell | 957 (5) | 50% to 92% | 41% to 94% | Wide variation in sensitivity | |
| Wako | 420 (3) | 50% to 86% | 89% to 100% | Insufficient number of studies for meta‐analysis | |
| Fungitec‐G | 353 (3) | 67% to 88% | 60% to 85% | Too few studies for meta‐analysis | |
| Dynamiker Fungus | 198 (2) | 64% to 81% | 78% to 80% | Too few studies for meta‐analysis | |
| 2. Fungal organism | |||||
| Test/Subgroup | No. of participants (studies) | Sensitivity estimate (95% CI) | Specificity estimate (95% CI) | Implications | Quality and comments |
| Candida | 1185 (10) | 81% (75% to 86%) | 64% (56% to 72%) | Results are more homogeneous for Candida testing than for all fungi | |
| BDG: beta‐D‐glucan test; CI: confidence interval; EORTC/MSG: European Organization for Research and Treatment of Cancer Mycoses Study Group. | |||||
| Study name | Study design | Underlying condition | Fungal type | Test brand | Samples taken | Reference standard |
| Prospective consecutive | ICU | Mixed | Fungitell | Single | EORTC 2008 | |
| Prospective consecutive | Organ transplant | Mixed | Fungitell | Multiple | EORTC 2008 | |
| Retrospective consecutive | Cancer | Mixed | Fungitell | Single | EORTC 2008 | |
| Prospective consecutive | Cancer | Aspergillosis | Glucatell | Multiple | EORTC 2008 | |
| Prospective consecutive | Cancer | Mixed | Fungitell | Single | EORTC 2008 | |
| Prospective consecutive | Mixed at‐risk | Mixed | Fungitell | Multiple | EORTC 2008 | |
| Retrospective case‐control | ICU | Candida | Fungitell | Single | Culture from blood or sterile site | |
| Retrospective consecutive | Mixed at‐risk | PJP | Fungitell | Single | Microscopy | |
| Retrospective case‐control | ICU | Aspergillosis | Fungitell | Unknown | EORTC 2008 | |
| Prospective consecutive | ICU | Candida | Fungitell | Single | EORTC 2008 | |
| Retrospective case‐control | Mixed at‐risk | PJP | Wako | Single | PCR | |
| Retrospective case‐control | Cancer | Mixed | Fungitell | Multiple | EORTC 2008 | |
| Prospective consecutive | Cancer | Aspergillosis | Fungitell | Multiple | EORTC 2008 | |
| Retrospective consecutive | ICU | Candida | Fungitell | Single | Culture from blood or sterile site | |
| Prospective consecutive | Cancer | Mixed | Fungitell | Single | EORTC 2008 | |
| Prospective consecutive | Cancer | Aspergillosis | Fungitell | Multiple | EORTC 2002 | |
| Retrospective consecutive | Cancer | Mixed | Glucatell | Multiple | EORTC 2008 | |
| Prospective consecutive | Cancer | Mixed | Glucatell | Multiple | EORTC 2008 | |
| Prospective consecutive | ICU | Mixed | Fungitell | Multiple | EORTC 2008 | |
| Prospective consecutive | Cancer | Aspergillosis | Fungitec | Unknown | EORTC 2002 | |
| Prospective consecutive | Cancer | Aspergillosis | Glucatell | Multiple | EORTC 2002 | |
| Unknown | Cancer | Aspergillosis | Fungitec | Multiple | Study‐specific, comparable to EORTC | |
| Prospective consecutive | Cancer | Aspergillosis | Wako | Multiple | EORTC 2002 | |
| Prospective consecutive | Cancer | Mixed | Fungitell | Multiple | EORTC 2008 | |
| Retrospective consecutive | Cancer | Mixed | Fungitell | Multiple | EORTC 2008 | |
| Retrospective consecutive | ICU | Aspergillosis | Fungitell | Unknown | EORTC 2008 | |
| Retrospective consecutive | ICU | Aspergillosis | Fungitell | Single | EORTC 2008 | |
| Prospective consecutive | ICU | Candida | Fungitell | Multiple | Culture from blood or sterile site | |
| Retrospective consecutive | Cancer | Mixed | Fungitec | Single | EORTC 2008 | |
| Retrospective consecutive | ICU | Candida | Fungitell | Single | EORTC | |
| Prospective consecutive | ICU | Mixed | Fungitell | Single | Study‐specific, comparable to EORTC | |
| Prospective consecutive | ICU | Candida | Fungitell | Multiple | Culture | |
| Retrospective case‐control | Cancer | Aspergillosis | Fungitell | Multiple | EORTC 2008 | |
| Retrospective consecutive | Cancer | Aspergillosis | Fungitell | Multiple | EORTC 2008 | |
| Prospective consecutive | ICU | Candida | Fungitell | Multiple | Culture | |
| Unknown | Cancer | Mixed | Glucatell | Multiple | EORTC 2002 | |
| Retrospective case‐control | Cancer | Mixed | Fungitell | Single | EORTC 2002 | |
| Retrospective case‐control | Mixed | Candida | Fungitell | Single | Culture from blood or sterile site or pathology | |
| Prospective consecutive | ICU | Mixed | Fungitell | Multiple | EORTC 2008 | |
| Prospective consecutive | Cancer | Mixed | Fungitell | Multiple | EORTC 2002 | |
| Retrospective consecutive | Mixed at‐risk | Mixed | Fungitell | Single | EORTC 2008 | |
| Retrospective consecutive | HIV | PJP | Fungitell | Single | Microscopy/PCR | |
| Prospective consecutive | Cancer | Mixed | Wako | Multiple | EORTC 2002 | |
| Prospective consecutive | ICU | Candida | Dynamiker | Single | Culture from blood or sterile site | |
| Prospective consecutive | Organ transplant | Mixed | Fungitell | Multiple | EORTC 2008 | |
| Prospective consecutive | ICU | Mixed | Fungitell | Multiple | EORTC 2008 | |
| Prospective consecutive | Mixed at‐risk | Mixed | Fungitell | Single | EORTC 2008 | |
| Retrospective case‐control | Cancer | Candida | Fungitell | Multiple | Culture from blood or sterile site | |
| Retrospective case‐control | Mixed at‐risk | Mixed | Dynamiker | Single | EORTC 2008 | |
| EORTC: European Organization for Research and Treatment of Cancer; ICU: intensive care unit; PCR: polymerase chain reaction; PJP: Pneumocystis jirovecii pneumonia. | ||||||
| Characteristic | n | Percentage |
| Underlying condition | ||
| Cancer HIV/AIDS ICU Mixed at‐risk Organ transplant | 23 1 16 7 2 | 46.9% 2.0% 32.7% 14.3% 4.1% |
| Age of patients | ||
| Adult Neonate Pediatric Mixed Unknown | 26 3 3 5 12 | 53.1% 6.1% 6.1% 10.2% 24.5% |
| Study design | ||
| Prospective consecutive Retrospective consecutive Retrospective case‐control | 26 12 11 | 53.1% 24.5% 22.4% |
| Fungal type | ||
| Aspergillus only Candida only PJP only Mixed fungal types | 12 10 3 24 | 24.5% 20.4% 6.1% 49.0% |
| Test brand | ||
| Fungitell Glucatell Fungitec‐G Wako Dynamiker Fungus | 36 5 3 3 2 | 73.5% 10.2% 6.1% 6.1% 4.1% |
| Sampling strategy | ||
| Single sample Multiple samples Unknown | 20 26 3 | 40.8% 53.1% 6.1% |
| Reference standard used | ||
| EORTC Proven Candida PJP microscopy/PCR Study‐specific | 36 8 3 2 | 73.5% 16.3% 6.1% 4.1% |
| Low risk of bias | ||
| Participant selection Index test Reference standard Flow and timing | 35 49 41 41 | 71.4% 100.0% 83.7% 83.7% |
| EORTC: European Organization for Research and Treatment of Cancer; ICU: intensive care unit; PCR: polymerase chain reaction; PJP: Pneumocystis jirovecii pneumonia. | ||
| Test | No. of studies | No. of participants |
| 1 Fungitell Show forest plot | 36 | 4316 |
| 2 Glucatell Show forest plot | 5 | 957 |
| 3 Wako Show forest plot | 3 | 420 |
| 4 Fungitec Show forest plot | 3 | 353 |
| 5 Dynamiker Show forest plot | 2 | 198 |
