Galactomannan detection for invasive aspergillosis in immunocompromised patients
Abstract
Background
Invasive aspergillosis is the most common life‐threatening opportunistic invasive mycosis in immunocompromised patients. A test for invasive aspergillosis should neither be too invasive nor too great a burden for the already weakened patient. The serum galactomannan enzyme‐linked immunosorbent assay (ELISA) seems to have the potential to meet both requirements.
Objectives
To obtain summary estimates of the diagnostic accuracy of galactomannan detection in serum for the diagnosis of invasive aspergillosis.
Search methods
We searched MEDLINE, EMBASE and Web of Science with both MeSH terms and text words for both aspergillosis and the sandwich ELISA. We checked the reference lists of included studies and review articles for additional studies. We conducted the searches in February 2014.
Selection criteria
We included cross‐sectional studies, case‐control designs and consecutive series of patients assessing the diagnostic accuracy of galactomannan detection for the diagnosis of invasive aspergillosis in patients with neutropenia or patients whose neutrophils are functionally compromised. The reference standard was composed of the criteria given by the European Organization for Research and Treatment of Cancer (EORTC) and the Mycoses Study Group (MSG).
Data collection and analysis
Two review authors independently assessed quality and extracted data. We carried out meta‐analysis using the bivariate method. We investigated sources of heterogeneity by adding potential sources of heterogeneity to the model as covariates.
Main results
We included 54 studies in the review (50 in the meta‐analyses), containing 5660 patients, of whom 586 had proven or probable invasive aspergillosis. When using an optical density index (ODI) of 0.5 as a cut‐off value, the sensitivity of the test was 78% (70% to 85%) and the specificity was 85% (78% to 91%). At a cut‐off value of 1.0 ODI, the sensitivity was 71% (63% to 78%) and the specificity was 90% (86% to 93%). At a cut‐off value of 1.5 ODI, the sensitivity was 63% (49% to 78%) and the specificity was 93% (89% to 97%). None of the potential sources of heterogeneity had a statistically significant effect on either sensitivity or specificity.
Authors' conclusions
If we used the test at a cut‐off value of 0.5 ODI in a population of 100 patients with a disease prevalence of 11% (overall median prevalence), two patients who have invasive aspergillosis would be missed (sensitivity 78%, 22% false negatives), and 13 patients would be treated unnecessarily or referred unnecessarily for further testing (specificity 85%, 15% false negatives). If we used the test at a cut‐off value of 1.5 in the same population, that would mean that four invasive aspergillosis patients would be missed (sensitivity 61%, 39% false negatives), and six patients would be treated or referred for further testing unnecessarily (specificity 93%, 7% false negatives). These numbers should, however, be interpreted with caution because the results were very heterogeneous.
Plain language summary
Measurement of serum galactomannan to detect invasive aspergillosis in immunocompromised patients
When the immune system of a patient is unable to fight infections (for example because of prolonged corticosteroid therapy, immunosuppressive drugs, haematological malignancies or HIV/AIDS) invasive or systemic aspergillosis can be a life‐threatening mycotic (fungal) infection. Establishing a diagnosis of invasive aspergillosis at an early stage of infection allows early antifungal treatment, but a definitive diagnosis can only be established after death. To enable early diagnosis in a way that is not burdensome for the already weakened patient, galactomannan testing may be promising. Galactomannan is a cell wall component of Aspergillus that is excreted by the fungus.
Study design
The authors of this systematic review found 54 studies that looked at the error rates of this galactomannan test. These studies compared the results of the galactomannan test with the results of a more elaborate diagnostic workup, so that the percentages of false positive results (patients without invasive aspergillosis, according to the elaborate testing, but with a positive galactomannan test) and false negative results (patients with invasive aspergillosis, according to the elaborate testing, but with a negative galactomannan test) could be calculated. The galactomannan test does not result in a yes/no answer, but in a so‐called 'optical density index' (ODI). The authors of the different studies defined the galactomannan test as positive when the ODI was above 0.5, 1.0 or 1.5. Four studies used a different ODI and these were not included in the meta‐analysis.
Studies and results
When an ODI of 0.5 or higher was said to be positive, the galactomannan test missed 22 out of every 100 patients with invasive aspergillosis and it resulted in a false positive test in 15 out of every 100 patients without invasive aspergillosis.
When an ODI of 1.0 or higher was said to be positive, the galactomannan test missed 29 out of every 100 patients with invasive aspergillosis and it resulted in a false positive test in 10 out of every 100 patients without invasive aspergillosis.
When an ODI of 1.5 or higher was said to be positive, the galactomannan test missed 37 out of every 100 patients with invasive aspergillosis and it resulted in a false positive test in only 7 out of every 100 patients without invasive aspergillosis.
Limitations
The studies showed variable results and had small numbers of patients with invasive aspergillosis.
Authors' conclusions
Summary of findings
| What is the diagnostic accuracy of the galactomannan ELISA for invasive aspergillosis for different cut‐off values? Patients/population: immunocompromised patients, mostly haematology patients Prior testing: varied, mostly underlying disease or symptoms (fever, neutropenia) Setting: mainly haematology or cancer departments, mainly inpatients Index test: a sandwich ELISA for galactomannan, an Aspergillus antigen Importance: depends on the time‐gain the test may give Reference standard: gold standard is autopsy, but that is almost never done; in most studies therefore the reference standard is composed of clinical and microbiological criteria Studies: patient series or case‐control studies, not using an in‐house test and not excluding possibly infected patients. Studies had to report cut‐off values that were used (n = 29). Each study can be present in more than one subgroup | |||||
| Subgroup | Effect (95% CI) | No. of participants (studies) | Prevalence (median, range) | Comments | What do these results mean? |
| Cut‐off 0.5 | Sensitivity 0.78 (0.70 to 0.85) Specificity 0.85 (0.78 to 0.91) | 394 proven or probable 3549 possible or no IA (27) | Median 11% (IQR 6.5% to 16%) | — | With a prevalence of 11%*, 11 out of 100 patients will develop IA Of these, 2 will be missed by the Platelia test (22% of 11), but will be tested again Of the 89 patients without IA, 13 will be unnecessarily referred for CT scanning |
| Cut‐off 1.0 | Sensitivity 0.71 (0.63 to 0.78) Specificity 0.90 (0.86 to 0.93) | 145 proven or probable 1246 possible or no IA (8) | Median 13% (IQR 4.2% to 31%) | — | With a prevalence of 11%*, 11 out of 100 patients will develop IA Of these, 3 will be missed by the Platelia test (29% of 11), but will be tested again Of the 89 patients without IA, 9 will be unnecessarily referred for CT scanning |
| Cut‐off 1.5 | Sensitivity 0.63 (0.49 to 0.77) Specificity 0.93 (0.89 to 0.97) | 209 proven or probable 2412 possible or no IA (15) | Median 7.4% (IQR 4.3% to 16%) | — | With a prevalence of 11%*, 11 out of 100 patients will develop IA Of these, 4 will be missed by the Platelia test (36% of 11), but will be tested again Of the 89 patients without IA, only 6 will be unnecessarily referred for CT scanning |
| Children | Sensitivity 0.84 (0.66 to 0.93) Specificity 0.88 (0.60 to 0.97) | 47 proven or probable 308 possible or no IA (in 6 studies) | Median 16% (IQR 10% to 16%) | 5 studies had a cut‐off of 0.5 and one had a cut‐off of 1.5 | Of the 100 children, 16 had IA |
| * Median prevalence over all studies was 11% (range 0.8% to 56%). CI: confidence interval; CT: computerised tomography; ELISA: enzyme‐linked immunosorbent assay; IA: invasive aspergillosis; IQR: interquartile range | |||||
Background
Target condition being diagnosed
Invasive aspergillosis is the most common life‐threatening opportunistic invasive mycosis in immunocompromised patients (Kontoyiannis 2002). Mortality in patients diagnosed with this condition ranges from 70% to 90% at one year (Upton 2007). Invasive aspergillosis is caused by ubiquitous Aspergillus species that invade (most often) from the lungs into the adjacent organs if the immune system is not able to fight the infection. Its incidence is still increasing, mainly because of the increasing number of patients undergoing intensified chemotherapy or receiving prolonged corticosteroid therapy, and due to the increasing number of transplant recipients (Denning 1998; Marr 2002).
Establishing a diagnosis of invasive aspergillosis at an early stage of infection and subsequent early treatment improves the chances of survival (Upton 2007). However, clinical signs and symptoms are non‐specific and characteristic lesions on chest radiographs are frequently absent. The only definite reference standard to confirm invasive aspergillosis is autopsy, combined with culture from autopsy specimens. As a clinical reference standard, the demonstration of hyphen invasion in tissue specimens obtained by invasive procedures, in combination with a positive culture for Aspergillus species from the same specimens, establishes a diagnosis of invasive aspergillosis (Hope 2005; Singh 2005). The problem is that the patient's status often prohibits the use of invasive techniques. In addition, culturing of the causative agent can result in false negative or false positive results.
In 2001, a committee consisting of the Invasive Fungal Infections Cooperative Group of the European Organization for Research and Treatment of Cancer (EORTC), the Mycoses Study Group (MSG) and the National Institute of Allergy and Infectious Diseases proposed to grade the diagnosis of invasive aspergillosis at three levels of probability (Ascioglu 2002): proven, probable and possible invasive aspergillosis (seeTable 1). Unfortunately, these levels are only useful in research settings, because in clinical practice a large number of patients will be classified as 'possible', which may lead to overexposure to antifungal therapy if all 'possible' patients are treated (Subira 2003).
| Proven IA | Histopathologic or cytopathologic examination showing hyphae from needle aspiration or biopsy specimen with evidence of associated tissue damage; or positive culture result for a sample obtained by sterile procedure from a normally sterile and clinically or radiologically abnormal site consistent with infection |
| Probable IA | At least 1 host factor criterion; and 1 microbiological criterion; and 1 major (or 2 minor) clinical criteria from abnormal site consistent with infection |
| Possible IA | At least 1 host factor criterion; and 1 microbiological or 1 major (or 2 minor) clinical criteria from abnormal site consistent with infection. This category is not recommended for use in clinical trials of antifungal agents |
Host factor criteria are, for example, neutropenia, persistent fever, predisposing conditions, prolonged use of corticosteroids.
Microbiological criteria are positive culture from sputum, bronchoalveolar lavage fluid (BAL) samples or from sinus aspirate specimen; positive result for Aspergillus antigen in specimens of BAL, cerebrospinal fluid or two or more blood samples.
Major clinical criteria are, for example, new infiltrates on CT imaging (e.g. halo sign), suggestive radiological findings.
Minor clinical criteria are suggestive symptoms and signs.
The exact definitions of the EORTC/MSG criteria and their host factor, microbiological or clinical criteria can be found here (Ascioglu 2002).
CT: computerised tomography; EORTC: European Organization for Research and Treatment of Cancer; MSG: Mycoses Study Group
The main issues in the diagnosis of invasive aspergillosis are the following: a test needs to be sensitive in the early phase of the infection in order to start treatment early, but should not pose a large burden on the already weakened patient. Screening immunocompromised patients for invasive aspergillosis weekly or twice a week with such a test may lead to earlier treatment and better outcomes. When screening results are positive, patients may be referred for further confirmation of the diagnosis or they may be treated immediately.
Imaging techniques are neither invasive nor too great a burden for most patients. The presence of the so‐called 'halo sign' or the 'air crescent sign' on radiographs or computed tomography is indicative of invasive aspergillosis. These signs are, however, not long‐lasting: approximately a week after infection these signs disappear. The costs and the rapid accumulation of radiation associated with computed tomography (CT) scanning prevent its use as a screening tool for invasive aspergillosis. Furthermore, imaging techniques only give a clinical diagnosis, not a microbiological diagnosis. Microbiological diagnosis can be achieved through culturing of the fungus from normally sterile tissues or through histology of those tissues. These techniques, however, are time‐consuming and often too invasive for the patient.
An alternative is the use of laboratory tests. These include the detection of antigens (beta‐glucan or galactomannan), measurement of antibodies or nucleic acid detection techniques. Of these tests, the detection of galactomannan is currently the one that is most often used in practice. Galactomannan is a cell wall component of Aspergillus spp. and Penicillium spp. (Latgé 1994). It is excreted by the fungus during the growth phase and it has been suggested that the level of galactomannan is proportional to the fungal load in tissue and that the level of galactomannan has a prognostic value.
Index test(s)
There are currently two commercially available assays for the detection of galactomannan: the Pastorex® latex agglutination test and the Platelia® sandwich enzyme‐linked immunosorbent assay (ELISA) test. Of these two, the Pastorex® kit is only rarely used nowadays. The ELISA is mostly used for the detection of antigen in serum and in fluid that is obtained via bronchoalveolar lavage (BAL). Other specimens in which the test can also be used are cerebrospinal fluid (CSF) or urine (Ascioglu 2001; Hope 2005). We focused on the ELISA test in serum, because obtaining serum is less of a burden for the patient than collecting BAL fluid. Results of the ELISA are given as an optical density index (ODI), which is the ratio of the optical density of (usually) 1 ng/mL galactomannan versus the optical density of the sample. In order to enhance the sensitivity of the test, in the USA the manufacturer changed the recommended cut‐off for positivity in from 1.5 to 0.5 ODI.
Clinical pathway
There is substantial variation in the way the galactomannan ELISA is currently used in the clinic. Some clinicians do not use it at all, while others use the galactomannan ELISA as a screening tool, to monitor whether patients at risk develop invasive aspergillosis or not. In those cases, serum is tested for invasive aspergillosis once or twice every week. Sometimes the galactomannan ELISA is used to test for invasive aspergillosis in BAL fluid when it is already suspected and in those situations the test is only used in serum when there is no BAL fluid. In most situations, the galactomannan ELISA is used as a triage test: if the ELISA is positive, patients will be referred for further diagnostic testing or they will be referred for antifungal therapy (Segal 2006). Further diagnostic testing may involve either laboratory testing of BAL fluid, CT scanning or radiography, or a combination of tests. Patients may also be referred for further diagnostic work‐up on the basis of clinical signs and symptoms.
Objectives
Our primary objective was to assess the diagnostic accuracy of galactomannan detection in serum for the diagnosis of invasive aspergillosis in immunocompromised patients, at different cut‐off values for test positivity.
Secondary objectives
We aimed to study several possible sources of heterogeneity: subgroups of patients, different interpretations of the EORTC/MSG criteria as the reference standard and study design features.
Methods
Criteria for considering studies for this review
Types of studies
Studies that assessed the diagnostic accuracy of galactomannan detection by the Platelia© sandwich ELISA test, with either prospective or retrospective data collection, were eligible. The galactomannan ELISA could be assessed alone or in comparison to other tests.
Participants
Studies had to include patients with neutropenia or patients whose neutrophils are functionally compromised. We included studies with the following patient groups:
-
patients with haematological malignancies, receiving haematopoietic stem cell transplants, chemotherapeutics or immunosuppressive drugs;
-
solid organ transplant recipients and other patients who are receiving immunosuppressive drugs for a prolonged time;
-
patients with cancer who are receiving chemotherapeutics;
-
patients with a medical condition compromising the immune system, such as HIV/AIDS and chronic granulomatous disease (CGD, an inherited abnormality of the neutrophils).
Index tests
A commercially available galactomannan sandwich ELISA (Platelia©) was the test under evaluation. We only included studies concerning galactomannan detection in serum. We excluded studies addressing detection in BAL fluid, a number of other body fluids, such as CSF or peritoneal fluid, and tissue. We also excluded studies evaluating in‐house serum galactomannan tests.
Target conditions
The target condition of this review was invasive aspergillosis, also called invasive pulmonary aspergillosis or systemic aspergillosis.
Reference standards
The following reference standards can be used to define the target condition:
-
autopsy;
-
the criteria of the EORTC/MSG (Ascioglu 2002; De Pauw 2008); or
-
the demonstration of hyphal invasion in biopsies, combined with a positive culture for Aspergillus species from the same specimens.
The gold standard for this diagnosis is autopsy, combined with a positive culture of Aspergillus species from the autopsy specimens, or with histopathological evidence of Aspergillus. Autopsy is rarely reported, therefore we decided to take the criteria of the EORTC/MSG as the reference standard. These criteria divide the patient population into four categories: patients with proven invasive aspergillosis, patients who probably have invasive aspergillosis, patients who possibly have invasive aspergillosis and patients without invasive aspergillosis (seeTable 1). This division is based on host factor criteria, microbiological criteria and clinical criteria. Clinical studies have shown that these criteria do not match autopsy results perfectly. This especially true for the possible category. For clinical trials investigating the effect of treatment, for example, it is recommended that only the proven and probable categories are used (Borlenghi 2007; Subira 2003).
The exclusion of patients with 'possible' invasive aspergillosis, which can be regarded as group of 'difficult or atypical' patients, is likely to affect the observed diagnostic accuracy of a test. Also, the exclusion of any other of the reference standard groups may affect the accuracy of the index test. We therefore excluded studies explicitly excluding one of the four categories of patients from the review, as well as studies in which it is not clear how many patients with proven, probable, possible or no invasive aspergillosis had positive or negative index test results.
Search methods for identification of studies
Electronic searches
We searched MEDLINE (through PubMed), EMBASE (through Ovid) and ISI Web of Knowledge for relevant articles. We updated the search on 24 June 2011 and again on 17 February 2014 by searching the complete databases again with revised search terms. We compared the results of the updated search with the results of the previous search and removed duplicates. The revised search strategies can be found in Appendix 1.
Searching other resources
To identify additional published, unpublished and ongoing studies, we:
-
entered relevant studies identified from the above sources into PubMed and then use the 'Related Articles' feature;
-
searched the Science Citation Index to identify articles that cite the relevant articles;
-
checked the reference lists of all relevant studies.
In the protocol, we stated that we would also contact authors and industry, but due to time constraints we were not able to do this.
Data collection and analysis
Selection of studies
After we removed all articles on animal studies, plant studies, mycotoxin studies and studies of allergic aspergillosis from the set of retrieved articles, two authors (ML, YD) selected potentially relevant articles based on title and abstract. Afterwards, we obtained the full paper of each potentially eligible article located in the search. Three review authors (ML, CV, YD) independently assessed eligible articles for inclusion. We resolved disagreements by discussion. We included all articles on which disagreement could not be resolved.
Data extraction and management
We extracted the following:
-
author, year of publication and journal;
-
study design;
-
study population;
-
reference standard and performance of the reference standard;
-
performance of the index test;
-
QUADAS‐2 items;
-
data for two‐by‐two tables.
The data extraction form was accompanied by a background document that stated how each item on the form should be interpreted. We standardised the form and piloted it on two primary diagnostic studies, including the quality assessment. Six review authors in total extracted data and assessed quality. Two review authors independently assessed each article. One author had a methodological background and the other a microbiological background. The articles were randomly allocated to a pair of assessors. We resolved disagreements by discussion.
Assessment of methodological quality
Study quality was assessed using QUADAS‐2 (Whiting 2003; Whiting 2011). The items of the QUADAS‐2 tool and their interpretation are described in Appendix 2. Results are presented in the text and in a graph. We did not calculate a summary score estimating the overall quality of a article since the interpretation of such summary scores is problematic and potentially misleading (Juni 1999; Whiting 2005).
Statistical analysis and data synthesis
Our reference standard was the set of EORTC/MSG criteria that can be used to classify patients to one of four groups: proven, probable, possible and no invasive aspergillosis. This resulted in a two‐by‐four table: a positive or negative galactomannan test result in each one of the four reference groups. To calculate test accuracy and to reflect the categories that are used in clinical practice to guide further management, we made the post hoc decision to define the proven and probable patients as having invasive aspergillosis and we defined the possible and no invasive aspergillosis patients as not having invasive aspergillosis, in order to construct two‐by‐two tables. We assessed other divisions between patients having and not having invasive aspergillosis in a subgroup analysis, but because of the limited clinical value and statistical limitations (not enough proven invasive aspergillosis patients, for example) our focus was on the ability of the galactomannan ELISA to discriminate between patients that were either classified as proven or probable and patients who were classified as possible or no invasive aspergillosis. We excluded studies reporting insufficient data for the construction of a two‐by‐two table from the final analyses.
We used the data from the two‐by‐two tables to calculate sensitivity and specificity for each study. We present individual study results graphically by plotting the estimates of sensitivity and specificity (and their 95% confidence intervals) in both forest plots and the receiver operating characteristic (ROC) space. We used a bivariate random‐effects approach for the meta‐analysis of the pairs of sensitivity and specificity and for the construction of a summary ROC curve (Reitsma 2005). We incorporated covariates in the bivariate model to examine the effect of potential sources of bias and variation across subgroups of studies. Due to the bivariate nature of the model, effects of covariates on sensitivity and specificity can be modelled separately.
If more than one threshold was reported in a study, we selected one of those thresholds to incorporate in the meta‐analysis. In that case, we chose the threshold of 0.5, if reported, because this is the positivity threshold currently recommended by the manufacturer. Meta‐analyses were restricted to those studies that reported one of the most often used cut‐off values (0.5, 1.0 or 1.5).
Investigations of heterogeneity
We investigated heterogeneity in the first instance through visual examination of forest plots of sensitivities and specificities and through visual examination of the ROC plot of the raw data. We addressed the following three sources of heterogeneity: effect of cut‐off value, effect of the reference standard and existence of clinical subgroups.
a. Effect of cut‐off value
A main source of heterogeneity in diagnostic test accuracy reviews is differences in the applied cut‐off value between studies. We expected studies to report three different cut‐off values: 1.5 ODI (the value previously prescribed by the manufacturer), 0.5 ODI (the value prescribed by the manufacturer nowadays) and 1.0 ODI (an intermediate value). We therefore first investigated what was the influence of these cut‐off values on sensitivity and specificity by including the cut‐off value as a covariate in the meta‐regression model.
Some studies defined a positive test result as one single sample that exceeded the cut‐off value, while others defined a test result as positive when at least two subsequent samples (taken within a week) exceeded the cut‐off value. The latter was only reported in studies that used the galactomannan ELISA to monitor whether the patients developed invasive aspergillosis. The single sample definition was both used in these screening studies and in studies that only tested for galactomannan when there was suspicion of invasive aspergillosis (e.g. fever not responsive to antibacterial medication). We examined the impact of single sample versus subsequent sample by adding subsequent testing as a covariate to the previous analysis.
b. Effect of the definition of invasive aspergillosis
Our reference standard consists of the criteria of the EORTC/MSG, as published by Ascioglu 2002 or De Pauw 2008. This reference standard classified patients into four groups. We studied what the effect was of our definition of 'diseased' patients (i.e. either proven or probable invasive aspergillosis) and 'non‐diseased' patients (either possible or no invasive aspergillosis) versus other definitions of diseased (only proven patients or all patients except no invasive aspergillosis) and non‐diseased (all patients except the proven invasive aspergillosis patients or only no invasive aspergillosis patients).
c. Clinical subgroups
We explored the possible influence of clinical subgroups by stratified analyses and by including additional covariates in the regression analyses. We carried out these additional analyses by adding these variables as covariates to the analyses. We used the following variables as covariates in the meta‐analyses:
-
children versus adults;
-
distinctive groups of patients (e.g. patients at high risk versus patients at low risk of invasive aspergillosis; solid organ transplants versus haematological patients);
-
use of antifungal prophylaxis (yes versus no);
-
use of antifungal therapy (yes versus no).
Sensitivity analyses
To assess whether methodological quality influenced the results we found, we compared the results of only studies fulfilling certain quality criteria with the overall results. We also did a sensitivity analysis for Chinese studies only and for our choice to regard proven and probable as being diseased versus possible and no invasive aspergillosis as being non‐diseased.
Results
Results of the search
Our updated search in 2011 resulted in 4377 unique titles (Figure 1). After removing totally irrelevant titles (allergic aspergillosis, plants, animals, studies published before 1995), we checked 972 titles and abstracts in duplicate for eligibility. After assessment of the selected 158 full‐text articles, 87 articles seemed to be eligible for inclusion. From these 87 articles, we excluded 49 from the review for various reasons. We found no extra studies through additional searches or reference checking. The updated search in 2011 revealed 6 extra studies compared to the original version of the review. The updated search in 2014 resulted in 1605 additional unique titles; of these, we checked the full texts of 168 papers and included 38 studies. During data extraction, we excluded a further 22 studies and thus included another 16 studies in the review. In total, this review includes 54 relevant articles (of which 6 plus 16 were not in the original version). As four studies did not report the results of a cut‐off value of 0.5 ODI, 1.0 ODI or 1.5 ODI, we used 50 studies in the meta‐analyses.
Study flow diagram.
Included studies
The Characteristics of included studies table lists the characteristics of the 54 included studies, containing a total of 197 participants with proven invasive aspergillosis, 573 participants with probable invasive aspergillosis, 43 participants that were classified in one group as proven or probable invasive aspergillosis, 980 participants with possible invasive aspergillosis, 5284 participants with no invasive aspergillosis and 1219 participants that were classified in one group as possible or no invasive aspergillosis. One study included a group of nine so‐called 'suspected' patients, in between proven and no invasive aspergillosis (Williamson 2000). We considered this group as possibles.
Most studies reported diagnostic accuracy based on the results in individual patients, whereas seven studies reported test results for treatment, neutropenic or disease episodes, without exactly stating how many episodes there were per patient. As most patients will have only one or two episodes, we did not expect diagnostic test accuracy to change by the inclusion of these studies and therefore we have included them in the analyses.
The reference standard was formed by the EORTC/MSG criteria that defined the proven, probable, possible or no invasive aspergillosis categories. We also included studies that used criteria that were similar to the EORTC criteria (thus, defining groups of patients with ordinal certainty of invasive aspergillosis). Eleven studies were done in China and used the criteria from the Editorial Board of the Chinese Journal of Internal Medicine. As these criteria also divide the patients into proven, probable, possible and no invasive aspergillosis categories, we assumed that they were sufficiently similar to include these studies. However, we did carry out a sensitivity analysis to assess their effect on the results.
Twenty studies presented the results for a cut‐off value of a single test above 0.5 ODI; 17 presented the results for a cut‐off value of two subsequent tests above 0.5. Ten studies reported the results for a cut‐off value of a single test above 1.0, while seven presented the results for a cut‐off value of two subsequent tests above 1.0. Fifteen studies reported the results for a cut‐off value of a single test above 1.5, while 10 presented the results for a cut‐off value of two subsequent tests above 1.5. A few of these studies also reported the results for other cut‐off values, such as 0.38 (one study) or 0.80 (two studies). One study did not report a cut‐off value at all and three studies reported a mixed cut‐off value of either subsequently above 0.5 or a single value above 0.8.
Excluded studies
We excluded 67 articles (seeCharacteristics of excluded studies table). Possible invasive aspergillosis patients were excluded by 24 studies, which we excluded from the review for that reason. Four assessed another test than the commercially available galactomannan ELISA or an obsolete version, and 14 studies did not provide sufficient details to make two‐by‐two tables. One study was published in duplicate, one was a letter to the editor and one provided two‐by‐two tables based on samples rather than individual patients. Five studies needed to be translated, but we could not find a translator.
Methodological quality of included studies
Figure 2 shows the overall quality of the 54 included studies.
Methodological quality graph: Review authors' judgements about each methodological quality item presented as percentages across all included studies.
Slightly more than half of the studies had included a representative patient spectrum. Two studies reported the results of a case‐control study, but in these studies the controls were randomly selected from the previously tested population. In eight studies it was not clear whether they used a case‐control design or not. Eleven studies were not clear about how they interpreted the EORTC/MSG criteria or whether they used other criteria as reference standard, five of which were of Chinese origin. The time between the galactomannan ELISA and the actual diagnosis was reported in 10 studies and eight of these reported an acceptable time gap. Partial and differential verification was not a problem. Most studies (n = 29) reported explicitly that they did not include the galactomannan ELISA in the EORTC/MSG criteria, but there were 12 studies that explicitly reported that galactomannan testing was part of the reference standard. Blinding of both the results of the reference standard and the results of the index test was reported variably. Most studies reported no details at all about any uninterpretable or indeterminate index test results.
The quality assessment results for the individual studies can be found in Figure 3.
Methodological quality summary: Review authors' judgements about each methodological quality item for each included study.
Findings
The sensitivity of the 54 studies varied from 0% to 100% and the specificity from 21% to 100% (see Figure 4 and Figure 5). The wide range of the sensitivity was largely due to chance variation, because of small numbers of patients with the target condition (proven or probable) in the various studies, ranging from 1 to 98 (median 12). For instance, if there is only one patient with proven or probable invasive aspergillosis in a study and this patient had a positive test, the sensitivity would be 100%, but if he or she had a negative test result, the sensitivity would be 0%. Small numbers of patients were not an issue in the possible or no invasive aspergillosis groups (median 92, range 8 to 773).
Forest plots of sensitivity and specificity. The squares represent the sensitivity and specificity of one study, the black line its confidence interval. Studies are grouped by reported cut‐off value. If a study reported accuracy data for more than one cut‐off, its results are included in more than one subgroup. TP = True Positive; FP = False Positive; FN = False Negative; TN = True Negative.
Forest plot of the included studies. TP = True Positive; FP = False Positive; FN = False Negative; TN = True Negative. Between brackets the 95% confidence intervals (CI) of sensitivity and specificity. The figure shows the estimated sensitivity and specificity of the study (blue square) and its 95% confidence interval (black horizontal line).
Plot of sensitivity versus specificity for all 54 studies, with different symbols for the different cut‐off values. The width of the symbols is proportional to the inverse standard error of the specificity in every study and the height of the symbols is proportional to the inverse standard error of the sensitivity.
The median prevalence of invasive aspergillosis patients was 11% (range 0.8% to 56%; interquartile range 4.5% to 21%). This prevalence is based on the proportion of proven and probable patients in the studies that included consecutive series of patients with a comparable risk of developing invasive aspergillosis (in contrast to case‐control studies, where the numbers of cases and controls, and thus the prevalence, is determined by the researchers).
We carried out meta‐analyses on the 50 studies that reported a cut‐off value of 0.5, 1.0 or 1.5, either as a single test result or subsequently.
After selecting one two‐by‐two table from each study, we had 27 two‐by‐two tables for a cut‐off value of 0.5, eight two‐by‐two tables for 1.0 and 15 two‐by‐two tables for 1.5. As expected, applying a higher cut‐off value led to lower sensitivities and higher specificities; these effects of cut‐off value were statistically significant for specificity only (P value = 0.07 for sensitivity and P value = 0.03 for specificity) (Figure 6).
Summary ROC plots for galactomannan test at three different cut‐off values. The graph shows the point estimates of sensitivity and specificity (solid dots) and the 95% confidence regions (dotted lines) around it. Data for individual studies are not shown in this plot.
A participant could be defined as test‐positive in two ways: a single sample above the cut‐off value; or two subsequent samples above the cut‐off value. In those situations, a participant was only defined as having a positive enzyme‐linked immunosorbent assay (ELISA) when two subsequent test results were both above the cut‐off value. This increases specificity, as the number of false positive results will decrease, and it decreases sensitivity. This effect was not statistically significant (P value = 0.73 for sensitivity and P value = 0.12 for specificity).
The results for both analyses are presented in Table 2.
| Cut‐off | Analysis | Studies (n) | Sensitivity (95% CI) | Specificity (95% CI) |
| 0.5 | 27 | 0.78 (0.70 to 0.85) | 0.85 (0.78 to 0.91) | |
| Single sample | 13 | 0.79 (0.69 to 0.88) | 0.80 (0.71 to 0.90) | |
| Subsequent samples | 14 | 0.77 (0.67 to 0.87) | 0.88 (0.81 to 0.94) | |
| 1.0 | 8 | 0.71 (0.63 to 0.78) | 0.90 (0.86 to 0.93) | |
| Single sample | 4 | 0.72 (0.62 to 0.82) | 0.87 (0.81 to 0.93) | |
| Subsequent samples | 4 | 0.70 (0.59 to 0.80) | 0.92 (0.88 to 0.96) | |
| 1.5 | 15 | 0.63 (0.49 to 0.77) | 0.93 (0.89 to 0.97) | |
| Single sample | 8 | 0.64 (0.48 to 0.80) | 0.92 (0.86 to 0.97) | |
| Subsequent samples | 7 | 0.61 (0.45 to 0.78) | 0.95 (0.91 to 0.98) |
CI: confidence interval
Effect of different definitions of 'diseased' patients
Table 3 shows the results of the analyses based on different definitions of 'diseased' patients. In the analyses above, sensitivity was calculated based on both the proven and the probable patients. If we calculate sensitivity with the results of only the proven patients, sensitivity improves from 78% (70% to 85%) to 89% (79% to 99%) for subgroups with a cut‐off value of 0.5 optical density index (ODI). In these analyses, the probable patients were considered to have no invasive aspergillosis and are used (with the possibles and the no invasive aspergillosis group) to calculate specificity, which decreased from 85% (78% to 91%) to 72% (62% to 82%). These effects, increasing sensitivity and decreasing specificity, are the same for all cut‐off values.
| Cut‐off value and analysis | Proven and probable versus possible and no IA | Proven versus probable, possible and no IA | Proven, probable and possible versus no IA | ||||||
| n | Sensitivity (95% CI) | Specificity (95% CI) | n | Sensitivity (95% CI) | Specificity (95% CI) | n | Sensitivity (95% CI) | Specificity (95% CI) | |
| 0.5 ODI | 27 | 0.78 (0.70 to 0.85) | 0.85 (0.78 to 0.91) | 18 | 0.89 (0.79 to 0.99) | 0.72 (0.62 to 0.82) | 19 | 0.55 (0.41 to 0.69) | 0.87 (0.80 to 0.94) |
| 1.0 ODI | 8 | 0.71 (0.63 to 0.78) | 0.90 (0.86 to 0.93) | 8 | 0.79 (0.70 to 0.89) | 0.83 (0.78 to 0.88) | 8 | 0.54 (0.44 to 0.65) | 0.93 (0.90 to 0.96) |
| 1.5 ODI | 15 | 0.63 (0.49 to 0.77) | 0.93 (0.89 to 0.97) | 14 | 0.65 (0.48 to 0.83) | 0.91 (0.86 to 0.96) | 14 | 0.54 (0.36 to 0.71) | 0.97 (0.94 to 0.99) |
CI: confidence interval; IA: invasive aspergillosis; ODI: optical density index
When we calculated sensitivity based on all patients except the no invasive aspergillosis patients, then sensitivity decreased (from 78% to 55%). This is due to the addition of the possible patients, who will be more often falsely negative if we classify them as having invasive aspergillosis. Specificity in these analyses was only based on no invasive aspergillosis patients and increased from 85% (78% to 91%) to 87% (80% to 94%) for subgroups with a cut‐off value of 0.5 ODI. Again this effect is present for all cut‐off values.
Clinical subgroups
It is possible that the accuracy of the following clinical subgroups could differ and therefore they are a potential source of heterogeneity:
-
children versus adults;
-
distinctive groups of patients;
-
use of antifungal prophylaxis;
-
use of antifungal therapy.
Children versus adults
Within the set of 54 studies there were seven studies that reported data on children. Foy 2007 reported separate results for both adults and children. Badiee 2013, Bialek 2002, De Mol 2013, Hovi 2007, Jha 2013 and Zhang_X 2009 only reported results for children. Five studies reported a cut‐off of 0.5; one a cut‐off of 1.5 and one study did not report a cut‐off (Hovi 2007).
Whether the analyses were based only on children or not had no significant effect on either sensitivity (P value = 0.09) or specificity (P value = 0.69).
Effect of distinctive groups of patients
We were not able to investigate the effect of distinctive groups of patients either due to the absence of such patient groups in the included studies or because this information was not presented in the articles. Some studies reported the inclusion of high‐risk patients, but the definition of high‐risk was not always clear or the definition of high‐risk matched the inclusion criteria of studies that did not report that they included high‐risk patients. Also the type of underlying disease was not always clearly reported.
Therefore, we decided post hoc to analyse the effect of prevalence of invasive aspergillosis on the accuracy of the galactomannan test and the effect of the way the patients were selected for the study, as a proxy for disease severity. High prevalence of invasive aspergillosis may reflect a population that is at high risk of developing invasive aspergillosis. The effect of prevalence on sensitivity and specificity was not significant when it was in addition to cut‐off value as a covariate in the regression analysis (sensitivity, P value = 0.99; specificity, P value = 0.96).
Another post hoc analysis to investigate the effect of distinctive patient groups was the assessment of the effect of the selection of patients on the accuracy of the galactomannan test. We divided the studies into three groups: (1) studies that did not restrict the patients that would be included in the study and that used the galactomannan ELISA as a screening test in all patients (n = 15; median prevalence 7.6%, interquartile range 2.0% to 12%); (2) studies that included only patients who had fever for a certain number of days and whose fever was not responsive to antibiotic treatment (n = 17; median prevalence 12%, interquartile range 8.6% to 23.0%); (3) studies that used other selection methods, mostly based on underlying diseases, or that did not report clearly how they selected their patients, or that did use a combination of selection methods (n = 14; median prevalence 14%, interquartile range 6.7% to 34%). The difference in selection methods had no significant effect on either sensitivity (P value = 0.21 for fever; P value = 0.21 for unselected) or specificity (P value = 0.088 for fever; P value = 0.89 for unselected).
Use of antifungal prophylaxis or antifungal therapy
Eighteen studies used antifungal prophylaxis, 22 studies did not and 11 studies provided no details on the use of prophylaxis. The use of prophylaxis was not significantly associated with either sensitivity (P value = 0.16) or specificity (P value = 0.63). Thirty‐eight studies used a therapeutic antifungal intervention (mostly amphotericin B), three studies did not and 10 studies were not clear on whether they used therapy or not. Most studies that did use antifungal therapy kept monitoring galactomannan levels during therapy. Use of antifungal therapy had a significant effect on sensitivity (P value = 0.04), but not on specificity (P value = 0.32).
Sensitivity analysis
We explored the impact of risk of bias by doing a sensitivity analysis only for studies that had a low risk of bias for each of the QUADAS‐2 domains (Figure 7). We also analysed the effect of different reference standards, studies including haematology patients only and Chinese articles. For all the sensitivity analyses the confidence intervals overlap with the overall meta‐analysis.
Sensitivity analyses
Discussion
Summary of main results
We included 54 studies in the review, but the results of the meta‐analyses are based on the 50 studies that explicitly reported the use of the commercially available galactomannan enzyme‐linked immunosorbent assay (ELISA) and the cut‐off value(s), and that included results for all four categories of invasive aspergillosis patients: proven, probable, possible and no invasive aspergillosis. Quality features that were poorly reported were: the time between the galactomannan ELISA and the actual diagnosis, whether reference and index tests were performed in a blinded fashion and the source of funding.
The mean sensitivity of the galactomannan ELISA at a cut‐off of 0.5 optical density index (ODI) was 78% (70% to 85%) and the specificity was 85% (78% to 91%). At a cut‐off value of 1.0 ODI, sensitivity was 71% (63% to 78%) and specificity was 90% (86% to 93%). At a cut‐off value of 1.5 ODI, sensitivity was 63% (49% to 77%) and specificity was 93% (89% to 97%). Sensitivity in particular was very heterogeneous. Part of this heterogeneity can be explained by the inclusion of small studies and by the inclusion of studies with low prevalence. Seesummary of findings Table. When two subsequent positive test results were needed to indicate a patient as being 'positive', sensitivity decreased slightly and specificity increased considerably. There were no potential sources of heterogeneity that had a significant effect on either sensitivity or specificity, except that in studies in which the patients were curatively treated, the sensitivity was slightly higher than in studies that did not report curative treatment during the study period.
Our results compared with other reports
Several reviews have been published in recent years about the (lack of) usefulness of the galactomannan ELISA for the diagnosis of invasive aspergillosis (Mennink‐Kersten 2004; Pfeiffer 2006; Segal 2006; Verdaguer 2007). Most of these reviews, however, are based on non‐systematic methods. Pfeiffer and colleagues undertook a systematic approach to summarise all available studies until 2005 (Pfeiffer 2006). Although this meta‐analysis has methodological limitations (sensitivity and specificity were summarised separately, for example), their results for the different cut‐off value subgroups did not differ much from ours (Leeflang 2006). As a change in cut‐off value will always lead to an opposite change in sensitivity and specificity across studies, we studied the effect of other potential factors by including them as covariates additional to the cut‐off value. This gives a more realistic estimation of the sensitivity and specificity belonging to a certain group of studies. Pfeiffer 2006 also recommended that a higher rather than a lower cut‐off value improves diagnostic test accuracy. They only looked, however, at the diagnostic odds ratio (DOR) for this conclusion. Using the DOR to guide clinical decisions regarding the use of a diagnostic test has some serious limitations. It does not take into account the relative importance of false negative or false positive results. A test with a sensitivity of 70% and a specificity of 90% has the same DOR as a test with a sensitivity of 90% and a specificity of 70%, but the clinical consequences of missing a diseased patient (false negative) are not identical to those of given unnecessary treatment to a non‐diseased patient (false positive).
Strengths and weaknesses of the review
We reviewed the diagnostic accuracy of a commercially available galactomannan ELISA to diagnose invasive aspergillosis according to the most recent insights and methods for diagnostic meta‐analyses. The results can, however, be biased by the use and implementation of the reference standard in a way that we have not been able to detect. We only included studies that used the European Organization for Research and Treatment of Cancer (EORTC)/Mycoses Study Group (MSG) criteria or a similar reference standard, but we can imagine that these criteria may still be interpreted subjectively, especially regarding the host factor criteria. Differences in interpretation of the reference standard may have been the reason for the large differences we found in the distribution of patients with proven, probable, possible and no invasive aspergillosis. A relatively large proportion of proven and probable patients may suggest that the reference standard is interpreted in a liberal way, which would then lead to more patients with proven/probable invasive aspergillosis that in reality might not have it. In that case, the estimated sensitivity will be lower than the true sensitivity.
Another indication that the reference standard may not be interpreted in the same way for each study is the variation in prevalence that we found among the studies. Prevalence of proven and probable invasive aspergillosis combined varied from 0.8% to 44%, with an overall median of 9.3%. Variation in prevalence can be caused by several different mechanisms, among which are differences in reference standard interpretation, differences in underlying population, differences in selection of participants and differences in referral pattern (the way through which the participants have been referred to the study location). These mechanisms may also cause differences in diagnostic test accuracy, but their effect on sensitivity and specificity may vary. For example, patient groups with a higher prevalence may include more severe cases of disease (Mulherin 2002). In that situation, one would expect that patient groups with a higher prevalence would also show a higher sensitivity, because more severe cases will result in more true positives and fewer false negatives. We found, however, no significant effect of prevalence on diagnostic accuracy, either when we used prevalence as a continuous covariate (results not shown), or when we used prevalence as dichotomous covariate. This may be because the patients were selected for being 'at high risk' before they developed invasive aspergillosis, so that this only changes the proportion of patients with proven or probable invasive aspergillosis in this group, but not the severity of this disease in the high‐risk group. Another explanation may be that there is no relationship between the severity of invasive aspergillosis and the serum galactomannan titre.
Another factor that we could not control was the time between the index test and the reference standard. Our reference standard was a composite reference standard, therefore the final diagnosis could have been made at several time points and at different time intervals from the index test. If the time between the index test and the reference standard is too long, the true disease status of the patient may have changed by the time the reference standard is assessed.
We defined the proven and probable patients as having invasive aspergillosis and we defined the possible and no invasive aspergillosis patients as not having invasive aspergillosis, in order to construct two‐by‐two tables. Whether this would have influenced our results depends on the association between the galactomannan test results and the true underlying invasive aspergillosis status in the 'probables' and in the 'possibles'.
Applicability of findings to the review question
We reviewed the diagnostic accuracy of only one test, but it would have been worthwhile to investigate the relative value of the galactomannan ELISA in addition to all other tests that can be performed. However, the galactomannan test has the advantage that it is not an invasive test and hence can be assessed in very ill patients. In some patients, it may therefore be the only available test. In that case, this review gives a valuable overview of the possibilities and weaknesses of the test. Furthermore, the current use of the galactomannan ELISA and its place in the clinic differs from place to place. It would therefore have been very difficult to make comparisons that would have been relevant for a broader public.
In some clinics, the galactomannan test is used in addition to the clinical presentation of the patient and chest radiographs, as a tool to monitor whether the immunocompromised patient develops invasive aspergillosis. If a patient has fever and pulmonary symptoms that do not respond to antimicrobial therapy, he or she will be referred for high‐resolution computed tomography (HRCT). If the galactomannan test is positive, the patient will also be referred for HRCT; it is generally believed that the galactomannan test becomes positive before clinical signs of aspergillosis develop. Hence, the use of this test will lead to earlier referral for HRCT, before clear symptoms develop, and to earlier treatment, if the test is positive. This, in turn, may lead to a higher treatment success rate.
This supposed advantage of the galactomannan test, however, leans on three assumptions: (1) the Platelia test is indeed positive before the patient shows signs and symptoms; (2) the HRCT also shows signs of invasive aspergillosis at that moment; and (3) earlier treatment results in a higher success rate. Of the 42 studies that we included in our review, 24 did not report any useful information about point in time at which the galactomannan test was positive. Five studies reported that the test was never positive before either CT, diagnosis or clinical signs. The other studies that reported the time between a positive galactomannan test and other tests or clinical signs reported time periods varying from around 60 days before to around 50 days after any other evidence (either CT, radiology, clinical signs, fever, diagnosis) for aspergillosis. It was not possible to calculate a mean or median time span, or even a probability of the galactomannan test being positive earlier than other diagnostic evidence. So we could not evaluate the probability that the first two assumptions are true.
Methodological quality graph: Review authors' judgements about each methodological quality item presented as percentages across all included studies.
Methodological quality summary: Review authors' judgements about each methodological quality item for each included study.
Forest plots of sensitivity and specificity. The squares represent the sensitivity and specificity of one study, the black line its confidence interval. Studies are grouped by reported cut‐off value. If a study reported accuracy data for more than one cut‐off, its results are included in more than one subgroup. TP = True Positive; FP = False Positive; FN = False Negative; TN = True Negative.
Forest plot of the included studies. TP = True Positive; FP = False Positive; FN = False Negative; TN = True Negative. Between brackets the 95% confidence intervals (CI) of sensitivity and specificity. The figure shows the estimated sensitivity and specificity of the study (blue square) and its 95% confidence interval (black horizontal line).
Plot of sensitivity versus specificity for all 54 studies, with different symbols for the different cut‐off values. The width of the symbols is proportional to the inverse standard error of the specificity in every study and the height of the symbols is proportional to the inverse standard error of the sensitivity.
Summary ROC plots for galactomannan test at three different cut‐off values. The graph shows the point estimates of sensitivity and specificity (solid dots) and the 95% confidence regions (dotted lines) around it. Data for individual studies are not shown in this plot.
| What is the diagnostic accuracy of the galactomannan ELISA for invasive aspergillosis for different cut‐off values? Patients/population: immunocompromised patients, mostly haematology patients Prior testing: varied, mostly underlying disease or symptoms (fever, neutropenia) Setting: mainly haematology or cancer departments, mainly inpatients Index test: a sandwich ELISA for galactomannan, an Aspergillus antigen Importance: depends on the time‐gain the test may give Reference standard: gold standard is autopsy, but that is almost never done; in most studies therefore the reference standard is composed of clinical and microbiological criteria Studies: patient series or case‐control studies, not using an in‐house test and not excluding possibly infected patients. Studies had to report cut‐off values that were used (n = 29). Each study can be present in more than one subgroup | |||||
| Subgroup | Effect (95% CI) | No. of participants (studies) | Prevalence (median, range) | Comments | What do these results mean? |
| Cut‐off 0.5 | Sensitivity 0.78 (0.70 to 0.85) Specificity 0.85 (0.78 to 0.91) | 394 proven or probable 3549 possible or no IA (27) | Median 11% (IQR 6.5% to 16%) | — | With a prevalence of 11%*, 11 out of 100 patients will develop IA Of these, 2 will be missed by the Platelia test (22% of 11), but will be tested again Of the 89 patients without IA, 13 will be unnecessarily referred for CT scanning |
| Cut‐off 1.0 | Sensitivity 0.71 (0.63 to 0.78) Specificity 0.90 (0.86 to 0.93) | 145 proven or probable 1246 possible or no IA (8) | Median 13% (IQR 4.2% to 31%) | — | With a prevalence of 11%*, 11 out of 100 patients will develop IA Of these, 3 will be missed by the Platelia test (29% of 11), but will be tested again Of the 89 patients without IA, 9 will be unnecessarily referred for CT scanning |
| Cut‐off 1.5 | Sensitivity 0.63 (0.49 to 0.77) Specificity 0.93 (0.89 to 0.97) | 209 proven or probable 2412 possible or no IA (15) | Median 7.4% (IQR 4.3% to 16%) | — | With a prevalence of 11%*, 11 out of 100 patients will develop IA Of these, 4 will be missed by the Platelia test (36% of 11), but will be tested again Of the 89 patients without IA, only 6 will be unnecessarily referred for CT scanning |
| Children | Sensitivity 0.84 (0.66 to 0.93) Specificity 0.88 (0.60 to 0.97) | 47 proven or probable 308 possible or no IA (in 6 studies) | Median 16% (IQR 10% to 16%) | 5 studies had a cut‐off of 0.5 and one had a cut‐off of 1.5 | Of the 100 children, 16 had IA |
| * Median prevalence over all studies was 11% (range 0.8% to 56%). CI: confidence interval; CT: computerised tomography; ELISA: enzyme‐linked immunosorbent assay; IA: invasive aspergillosis; IQR: interquartile range | |||||
| Proven IA | Histopathologic or cytopathologic examination showing hyphae from needle aspiration or biopsy specimen with evidence of associated tissue damage; or positive culture result for a sample obtained by sterile procedure from a normally sterile and clinically or radiologically abnormal site consistent with infection |
| Probable IA | At least 1 host factor criterion; and 1 microbiological criterion; and 1 major (or 2 minor) clinical criteria from abnormal site consistent with infection |
| Possible IA | At least 1 host factor criterion; and 1 microbiological or 1 major (or 2 minor) clinical criteria from abnormal site consistent with infection. This category is not recommended for use in clinical trials of antifungal agents |
| Host factor criteria are, for example, neutropenia, persistent fever, predisposing conditions, prolonged use of corticosteroids. Microbiological criteria are positive culture from sputum, bronchoalveolar lavage fluid (BAL) samples or from sinus aspirate specimen; positive result for Aspergillus antigen in specimens of BAL, cerebrospinal fluid or two or more blood samples. Major clinical criteria are, for example, new infiltrates on CT imaging (e.g. halo sign), suggestive radiological findings. Minor clinical criteria are suggestive symptoms and signs. The exact definitions of the EORTC/MSG criteria and their host factor, microbiological or clinical criteria can be found here (Ascioglu 2002). CT: computerised tomography; EORTC: European Organization for Research and Treatment of Cancer; MSG: Mycoses Study Group | |
| Cut‐off | Analysis | Studies (n) | Sensitivity (95% CI) | Specificity (95% CI) |
| 0.5 | 27 | 0.78 (0.70 to 0.85) | 0.85 (0.78 to 0.91) | |
| Single sample | 13 | 0.79 (0.69 to 0.88) | 0.80 (0.71 to 0.90) | |
| Subsequent samples | 14 | 0.77 (0.67 to 0.87) | 0.88 (0.81 to 0.94) | |
| 1.0 | 8 | 0.71 (0.63 to 0.78) | 0.90 (0.86 to 0.93) | |
| Single sample | 4 | 0.72 (0.62 to 0.82) | 0.87 (0.81 to 0.93) | |
| Subsequent samples | 4 | 0.70 (0.59 to 0.80) | 0.92 (0.88 to 0.96) | |
| 1.5 | 15 | 0.63 (0.49 to 0.77) | 0.93 (0.89 to 0.97) | |
| Single sample | 8 | 0.64 (0.48 to 0.80) | 0.92 (0.86 to 0.97) | |
| Subsequent samples | 7 | 0.61 (0.45 to 0.78) | 0.95 (0.91 to 0.98) | |
| CI: confidence interval | ||||
| Cut‐off value and analysis | Proven and probable versus possible and no IA | Proven versus probable, possible and no IA | Proven, probable and possible versus no IA | ||||||
| n | Sensitivity (95% CI) | Specificity (95% CI) | n | Sensitivity (95% CI) | Specificity (95% CI) | n | Sensitivity (95% CI) | Specificity (95% CI) | |
| 0.5 ODI | 27 | 0.78 (0.70 to 0.85) | 0.85 (0.78 to 0.91) | 18 | 0.89 (0.79 to 0.99) | 0.72 (0.62 to 0.82) | 19 | 0.55 (0.41 to 0.69) | 0.87 (0.80 to 0.94) |
| 1.0 ODI | 8 | 0.71 (0.63 to 0.78) | 0.90 (0.86 to 0.93) | 8 | 0.79 (0.70 to 0.89) | 0.83 (0.78 to 0.88) | 8 | 0.54 (0.44 to 0.65) | 0.93 (0.90 to 0.96) |
| 1.5 ODI | 15 | 0.63 (0.49 to 0.77) | 0.93 (0.89 to 0.97) | 14 | 0.65 (0.48 to 0.83) | 0.91 (0.86 to 0.96) | 14 | 0.54 (0.36 to 0.71) | 0.97 (0.94 to 0.99) |
| CI: confidence interval; IA: invasive aspergillosis; ODI: optical density index | |||||||||
| Test | No. of studies | No. of participants |
| 1 Platelia ‐ all cut‐offs Show forest plot | 50 | 7955 |
| 2 Platelia in children Show forest plot | 7 | 472 |
