Mine Durusu Tanriover, MD, Department of Internal Medicine, Section of General Internal Medicine, Faculty of Medicine, Hacettepe University, Sihhiye, Ankara 06100, Turkey. Tel.: +90 312 305 30 29. Fax: +90 312 305 22 94. E-mail: email@example.com
Invasive aspergillosis (IA) presents a diagnostic and therapeutic dilemma for the physicians who take care of the patients with severe underlying diseases and immunosuppression. This study aimed to evaluate the usefulness of serum galactomannan (GM) measurements in the routine practice and surveillance of IA along with possible caveats in diagnosis and treatment. Adult patients with high-risk haematological malignancies admitted to the Internal Medicine wards during the 2-year study period were followed up by daily visits for vital signs, existing or newly developing signs and symptoms, clinical and laboratory findings. Blood samples were analysed for GM levels by the ELISA method at the end of the study period. Data of 58 hospitalisation episodes in 45 patients were analysed. Proven IA was diagnosed in one patient, probable IA was diagnosed in four patients. The sensitivity was 60% and the specificity was 21% when the index cut-off for positivity was accepted as 0.5. The yield of GM testing may be influenced by many variables and each centre should evaluate the usefulness of this test in its own conditions.
Fungal infections, particularly invasive aspergillosis (IA), still present a diagnostic and therapeutic dilemma for the physicians who take care of the patients with severe underlying diseases and immunosuppression. Because the severity of the underlying disease, critical illness and acute conditions preclude the diagnosis most of the time, empirical antifungal treatment has been the mainstay of management of such patients until recently. Empirical approach has its own disadvantages including unnecessary exposure to toxic effects and drug interactions as well as increased cost. However, the search for an ideal diagnostic marker, which can guide pre-emptive therapy, has been inconclusive so far.1
The accuracy of the microbiological methods in diagnosing IA depends on the type of the specimen obtained. Tissue biopsies are the best as culture specimens, because histopathological confirmation can be done simultaneously. However, the critical illness of the patients usually does not allow an invasive procedure.2
Imaging modalities such as high resolution computed tomography (CT) are non-invasive options for diagnosing Aspergillus infections.3–5 Serial tomograms starting on the early days of the febrile neutropenic period are required to detect the halo sign that suggests IA in the appropriate host and setting.6,7
Galactomannan (GM), which is a polysaccharide cell-wall component of Aspergillus, is a promising molecule to search for the clues of Aspergillus infection and tissue invasion.8 Methods like enzyme immunoassay, radioimmunoassay and latex agglutination have been used to identify GM in different specimens.9,10 Commercial kits (Platelia®Aspergillus; Bio-Rad Laboratories, Marnes-la-Coquette, France) that use the monoclonal anti-GM antibody EB-A2 as both capture and peroxidase-linked antibodies in sandwich enzyme-linked immunosorbent assay (ELISA) are available.10,11 While the specificity of the test is quite high, reported sensitivities in different studies display wide variations.9,12–20 The dispute about the ideal cut-off point was a subject of matter as well as the reproducibility of the test. Recently, an index cut-off of 0.5 was accepted in Europe after the study by Maertens and colleagues.12,14,21–26
In this study, we aimed to evaluate the diagnostic accuracy of serial GM measurements in our high-risk patients along with the possible caveats in diagnosing and treating IA in our centre, and focused on the possible ways to use the method more effectively in our routine clinical practice in the future.
Patients and methods
This prospective cohort study was carried out in Hacettepe University Hospital for Adults. The study was approved by the ethics committee of the Faculty of Medicine (Approval date 12 July 2001, HEK 01/30-4). All patients (>16 years of age) with high-risk haematological malignancies admitted to the Internal Medicine wards during the 2-year study period were interviewed and those who consented were recruited for follow-up. High-risk haematological malignancies included acute leukaemias, chronic myelocytic leukaemia with blastic transformation, myelodysplastic syndromes that required intensive chemotherapy and high-grade non-Hodgkin’s lymphomas.
Patients who gave informed consent were included in the study starting from the day they were admitted to the wards and followed up until death, discharge or withdrawal of consent, whichever occurred earlier. Death or discharge within 10 days of hospitalisation, less than 10 days of neutropenia or major difficulty in obtaining blood samples were the exclusion criteria. Demographic characteristics, underlying diseases and risk factors for invasive fungal infections (IFI), such as administration of chemotherapy, corticosteroids, antimicrobials, total parenteral nutrition within 30 days and stem-cell transplantation within 1 year, were noted. Patients were followed up by daily visits for vital signs, existing or newly developing signs and symptoms, clinical and laboratory findings. Colony stimulating factors, chemotherapeutic and antimicrobial agents administered were recorded during each visit. Culture growths and the results of the imaging studies were also noted.
The study protocol required that blood be drawn twice a week during the follow-up of the patients, however because of the problems in venous access and reluctance of the patients, regular sampling could not be performed all the time. Blood samples were then transported to the laboratory and preserved at −70 °C until all the specimens were analysed by the ELISA method at the end of the study period.
All patients with haematological malignancies who developed fever were consulted with the infectious diseases team as a routine part of patient care at our centre. GM levels were tested subsequently; therefore the primary physician and the infectious diseases consultant were not aware of the results during patient care. No antifungal prophylaxis was used in this cohort of patients. Patients were treated with amphotericin B formulations during inpatient periods and discharged on oral itraconazole when indicated for IA.
Invasive fungal infections were defined according to the European Organization for Research and Treatment of Cancer – Mycoses Study Group (EORTC-MSG) consensus case definitions.27 As this study aimed to evaluate the accuracy of GM in diagnosis, GM positivity was not used as a microbiological criterion for classifying IA.
Galactomannan levels were studied by sandwich ELISA commercial kit (Platelia®Aspergillus; Bio-Rad Laboratories) in accordance with the manufacturer’s instructions. Results are checked with positive and negative controls. The GM index was expressed as the ratio of the optical density of the sample relative to the optical density of the threshold control.
The total number of hospitalisation episodes required for the final analysis was calculated as 57 when the risk of IFI was predicted to be 15% in high-risk haematological malignancies in our hospital. Data analysis was performed with the softwares spss version 10.0 (SPSS Inc., Chicago, IL, USA) and stata version 9.0 (StataCorp LP, College Station, TX, USA). In addition to the cut-off point of 1.5 that was originally recommended by the manufacturer of the GM Platelia kit, 1.0, 0.7 and 0.5 cut-off points were also used to calculate sensitivity, specificity, negative and positive predictive values. Calculations were made separately for single positive values and at least two consecutive positive results (within 1 week) as well as classifying the data as proven plus probable cases or proven plus probable plus possible cases.
A total of 83 hospitalisation episodes were included in the study; however, 25 episodes were excluded from analysis because of the death of the patients soon after their inclusion in the study (n = 8), neutropenia <10 days (n = 7), absence of neutropenia (n = 6), problems with the venous access route (n = 1) and short period of hospitalisation (n = 3). Fifty-eight hospitalisation episodes in 45 patients were eligible for final analysis (Table 1). The underlying haematological malignancy was acute myeloblastic leukaemia in 35 patients, acute lymphoblastic leukaemia in six patients, chronic myelocytic leukaemia-blastic transformation in two patients, biphenotypical leukaemia in one patient and high-grade non-Hodgkin lymphoma in one patient.
Table 1. Characteristics of the episodes with regard to the IA class in which serum GM were evaluated
According to the EORTC-MSG case definitions, one patient had proven IA (sinopulmonary aspergillosis). The diagnosis was confirmed by the demonstration of invading hyphae in the necrotic specimen taken from the lateral wall of the nose. Probable IA was diagnosed in four and possible IA was diagnosed in 20 episodes. Thirty-three episodes were defined as not having IA.
Dyspnoea and cough were the leading complaints in proven and probable IA cases (Table 2). Bacteraemia was present in 21.2%, 30% and 60% of the episodes without IA, with possible IA and with probable/proven IA, respectively. One case of candidaemia and one case of disseminated fusariosis were identified, both of which did not have IA according to EORTC-MSG criteria. Aspergillus flavus was cultured from either blood, sputum or bronchoalveolar lavage in three episodes of three different patients, while Aspergillus fumigatus was cultured from bronchoalveolar lavage in two episodes of probable IA. Bronchoalveolar lavage could only be performed in nine episodes overall.
Table 2. Signs and symptoms of patients with regard to IA episodes
Total, n = 58 (%)
Proven/ probable IA, n = 5 (%)
Possible IA, n = 20 (%)
No IA, n = 33 (%)
IA, invasive aspergillosis; ALT, alanine aminotransferase; n, total number of episodes.
Papular skin lesion
Necrotic skin lesions
At least one thoracic CT was performed in 36 episodes. CT was ordered by the ward staff when the patient had prolonged fever without a focus, pulmonary signs and symptoms or pathological findings on plain radiograms. Among the 22 episodes in which no thoracic CT was performed, 12 had prolonged fever and neutropenia despite broad-spectrum antimicrobial therapy. Although indicated theoretically, CT was not ordered in these episodes at the discretion of the ward staff. The mean time from the first day of the febrile neutropenic episode to the first thoracic CT – if it was ever performed – was 15.7 ± 12.1 days. The most common finding was a nodule (53.4%). Halo sign and air-crescent sign were rather rare (6.9% and 2.7%, respectively).
We evaluated the concordance of the clinical diagnosis of IA made by the infectious diseases consultant, the initiation of antifungal therapy and the consensus definitions of EORTC-MSG. The consultant doctor was aware of the results of the microbiological and radiological studies, however, not of the GM assays. There was 100% agreement between the diagnosis of the consultant doctor and EORTC-MSG case definitions in patients with proven and probable IA. On the other hand, 85% of the patients with possible IA and 9.1% of those without IA according to EORTC definitions were considered to have IA clinically by the consultant. Moreover, 95% of the patients with possible IA and 30.3% of the patients with no IA received amphotericin B either with a clinical suspicion of IA or empirically for prolonged fever of unknown origin. The mean duration of amphotericin B use was 31 days for episodes with proven and probable IA, 26.5 days for episodes with possible IA and 6.2 days for episodes without IA.
A total of 545 serum samples were analysed by ELISA for GM levels. Regular sampling could not be carried out in all cases (in 22 of 58 episodes, more than 7 days elapsed in between two sampling dates at least once). During the course of the only proven IA, all of the serum GM levels were above 1.5 cut-off point (Fig. 1). The GM levels of the patient were positive at the beginning of the follow-up and soon rose to >10.0. Thoracic CT obtained 1 week later revealed a cavitary lesion in the lung and amphotericin B was started. Necrotic tissue in the nose and destruction of the bone on CT were noted and biopsied. Septate hyphae were demonstrated in the histopathological samples of the necrotic nasal tissue. The patient died 80 days after her admission because of uncontrolled malignancy.
None of the four probable episodes demonstrated consecutive GM positivities when the cut-off point was accepted as 1.5. One of them was positive consecutively when the cut-off was lowered to 1.0. All the probable cases had at least one GM level equal to or above 0.7. The case of fusariosis had a GM level of 1.8 after 5 days of growth of the fungus in the blood, necrotic nasal mucosa and skin specimen cultures. Candidaemia was detected in a patient with no IA in a period when GM values of 4.3 and 2.5 were measured.
The timing of GM positivity with respect to CT findings and culture growths could not be evaluated in all of the episodes. Lack of regular and timely CT imaging and high rate of false positivity and negativity were the obstacles to make this evaluation. However, the data of the only proven IA (Fig. 1) and the four probable IA episodes (Fig. 2) were summarised regarding the time elapsed between CT findings, culture growths and GM positivities.
Six episodes out of 24 episodes of probable and possible IA failed to demonstrate consecutive GM positivities when the cut-off of 0.7 was accepted (Table 3). When the cut-off was lowered to 0.5, four episodes had negative results on consecutive samples. On the other hand, 20 episodes out of 33 with no IA had positive GM results with a cut-off of 0.7 (Table 4). Four more episodes were rendered false positive when the cut-off was lowered to 0.5. Characteristics of patients with false positive GM results and factors coinciding with the period of false positivities were summarised in Table 4. Patients received beta lactam antibiotics in all episodes but one. Piperacillin-tazobactam and/or amoxicillin-clavulanate were used in 19 episodes out of 58. In particular cases with false positive results, piperacillin-tazobactam was used in four of 20 episodes and amoxicillin in one episode (Table 4).
Table 3. Characteristics of patients with false negative GM results and possible factors coinciding with the period of false negativities (false negativity is accepted as failure to demonstrate at least two GM ≥0.7 on two consecutive samples in patients with probable or possible IA)
Herpes labialis and diarrhoea episodes, amphotericin B (10th day of admission)
Staphylococcus aureus (BAL), A. flavus (BAL),
Infiltration, centrilobular densities, effusion
Amphotericin B (20th day of admission)
BL, hepatitis B
Nodule, ground glass
On itraconazole for previous IA, switched to amphotericin B upon admission
Pseudomonas aeruginosa (BAL)
Nodule, infiltration, lobar pneumonia
Cefepime, CAM, PIP-TAZ
On itraconazole for previous IA, switched to amphotericin B 1 week after admission, neutropenic enterocolitis
Amphotericin B (28th day of admission), discharged with itraconazole
Table 4. Characteristics of patients with false positive GM results and possible factors coinciding with the period of false positivities (false positivity is accepted as ≥0.7 on two consecutive samples in patients with no IA).
Values in parentheses show results for single positivity above the cut-off.
In recent years, monitoring of serum GM levels by ELISA has become popular for the early diagnosis of IA because of its standardisation and the applicability in routine practice. In this study, we evaluated the way we handle high-risk patients for IA and the applicability of serum GM measurements in our routine practice and surveillance.
The reported sensitivity and the specificity of the serum GM measurements by Aspergillus Platelia® kit vary widely in the literature, mostly because of heterogeneity among the studies.20 Sensitivities as high as 100% are reported, whereas some studies report no positive results in proven cases or sensitivity as low as 17%.14,16,28–31 A recent meta-analysis revealed an overall sensitivity of 61% and specificity of 93% for proven and probable cases.20 Although the sensitivity of GM assay differs among patient groups and may be very low, its specificity is quite good.20 This variation in the performance of the test is thought to be related to the inconsistency of the patient populations and the specimens used, the uncontrolled variables during the specimen transport or processing, and the different disease definitions and cut-off points.25
In this study, with only five episodes of IA (proven and probable), we found 60% sensitivity and a very low specificity (20.8%) for GM assay with the use of the generally accepted 0.5 cut-off value. The very low positive predictive values in our study can also be explained by the low number of IA in our patient population. The predictive values of a test in clinical practice depend critically on the prevalence of the abnormality in the patients being tested; the rarer the abnormality the lower will be the positive predictive value.
Several factors may explain the very low sensitivity. First of all, as the kinetics of GM is variable, we might have missed the positive values in the patients in whom regular twice weekly sampling could not be carried out because of the reluctance of the patients or the problems with the venous access. Mean number of serum samples per episode was 9.4 in this study, whereas it was 36.5 (in proven IA episodes) and 30.6 (in all episodes) in the series of Maertens, where 100% sensitivity was reported.12,32 The requirement of two consecutive results for positivity further decreased the sensitivity, given the fact that in 38% of the episodes, more than 7 days have elapsed in between two samples. In ideal study conditions, however, these patients would be excluded from the analysis.33 Second, the lack of invasive diagnostic interventions and autopsy probably had a great impact.12 Many of the possible and probable cases could have been upgraded to a higher level if microbiological criteria had been obtained. The performance of GM assay has been much worse in studies evaluating routine practices in which an ideal study setting could not be constructed.34 Another factor may be the high rate of empirical antifungal drug use in episodes with a prior or current episode of suspected IA, or with cavitary lesions in the lungs. This might have led to negative GM results because of the decreased release of the molecule into the bloodstream and especially in patients with mild infection.14 There might have been problems also during the transport and the storage of the specimens that we could not have controlled, which might have led to false negative results. Moreover, patients who encountered Aspergillus before their follow-up episodes might have developed anti-Aspergillus antibodies, which is a reported cause of GM false negativity.25
The GM-ELISA assay has been shown to demonstrate specificity of above 90% in most reports, contrary to this study, which documented a very poor specificity. The high false positivity of the method might be related to several factors in this study. First of all, concomitant beta lactam use such as piperacillin-tazobactam and amoxicillin-clavulanate might have contributed to some extent as reported previously.35–39 It seems as if cefepime is a reason for false positivity with regard to data in Table 4, however, it is very hard to conclude that cefepime cross reacts with GM. The empirical treatment for febrile neutropenia in our hospital included cefepime during the study period, so nearly all the patients were treated with this beta lactam antibiotic including those with false positive GM results. Moreover, the frequency of the GM testing is not adequate to conclude that cefepime is a causative agent for false positivity.
Fungal infections other than IA may yield positive GM results. Histoplasma, Penicillium, Cryptococcus, and Blastomyces are among the fungi that have been reported to cause false positive results.40–43 There are controversial reports regarding Fusarium.43,44 The disseminated fusariosis case in this study yielded two positive results among 18 measurements, and no other fungal infection could be shown. Candida, a cause of frequent opportunistic infections in neutropenic patients, could not be shown to yield positive results in GM assay before.43 In this study, the case with candidaemia had false positive GM results; however, the concomitant use of piperacillin-tazobactam in that case was probably the reason for false positivity (Table 4). The disruption of the integrity of gastrointestinal mucosa might have led to false positivity, as 57% of the patients experienced at least one diarrhoea attack during the follow-up (Table 4).45
This study revealed the discrepancy between the diagnose made in routine clinical practice and EORTC-MSG case definitions, as reported previously.46 In all of the cases of proven and probable IA, the consultant started antifungal therapy with a diagnosis of IA. However, in 85% of patients classified as possible IA, the consultant suspected of IA, and 95% of them received antifungal therapy either on clinical grounds or empirically. More dramatically, 9.1% of patients in the class without IA were suspected to have IA clinically and antifungal therapy was administered in 30.3% at some time during their follow-up. These findings are in accordance with the suggestion that the EORTC-MSG definitions were developed to guide clinical trials and to provide homogeneity of case definitions, but not to guide antifungal therapy.27
Administration of antifungal therapy to patients with possible IA – 95% in our series – might be considered as unnecessary and over-treatment as the likelihood of IA is rather low in these patients.12 It was recently demonstrated that antifungal therapy could be reduced from 35 to 7.7% by implementing a diagnostic algorithm.32 Developing pre-emptive therapy strategies will not only prevent unnecessary antifungal treatment but also will help diagnosing the episodes early in the period of IA when signs and symptoms are lacking in the window period.32 The greyest zone in the correlation of clinical picture and the EORTC-MSG classification is the possible IA group. The blade is two-sided; non-specific signs may be related to a non-existing IA or subclinical infection might be overlooked without adequate microbiological evidence. However, we detected cavitating nodules or halo sign in CT scans in 40% of the possible IA episodes. In other words, at least some of these cases probably do represent a group of patients with IA with inadequate microbiological evaluation who could have been upgraded to a higher risk class if they had been evaluated with adequate and appropriate cultures and tissue samples. Nodules on thoracic CT, which represent the most common finding in this study, can be caused by a vast array of pathologies in neutropenic patients, including IA. In routine clinical practice, it is very difficult to exclude the diagnosis of IFI in these patients unless biopsies are performed. This might have been the rationale why so many patients without a clear evidence of IA received antifungal therapy. In conditions where all diagnostic tools cannot be utilised effectively, it is natural for the clinician to feel safer on empirical approach.
The limitations of the study include the low number of probable and proven cases in the cohort, which might have led to worse results than some other studies in the literature. However, it is a valuable experience to discuss as it may demonstrate the caveats of empirical approach as well as the difficulty of implementing a GM and CT based pre-emptive strategy in a true cohort, which we face every day in routine clinical practice.
In conclusion, GM testing has been a major advance in the medical care of the patients with haematological malignancies. However, each centre should evaluate the usefulness of this test in its own conditions. The specific characteristics of the environment such as renovations that might increase exposure of the patients to Aspergillus species and result in anti-Aspergillus antibodies, as well as certain therapeutic practices, i.e. use of piperacillin-tazobactam in febrile neutropenic patients, rate of utilisation of imaging techniques and other microbiological diagnostic procedures, and the non-ideal settings of real life may profoundly influence the yield of this important serological marker for early diagnosis.
The authors want to thank Infectious Diseases research nurse Nimet Simsek for her efforts in specimen collection and Muge Durusu for the preparation of figures and tables. This study was supported with a grant from the Scientific and Technical Research Council of Turkey, Health Sciences Research Grant Group.