Aspergillus Galactomannan Antigen in the Cerebrospinal Fluid of Bone Marrow Transplant Recipients with Probable Cerebral Aspergillosis

Abstract

The Aspergillus galactomannan test was performed on cerebrospinal fluid and serum samples from 5 patients with probable cerebral aspergillosis and from 16 control patients. Cerebrospinal fluid galactomannan levels were significantly higher in aspergillosis patients, and most galactomannan was produced intrathecally. Comparison of serum galactomannan values in pulmonary and cerebral aspergillosis patients showed significant overlapping. Detection of Aspergillus galactomannan in cerebrospinal fluid may be diagnostic of cerebral aspergillosis.

Bone marrow transplant recipients are at high risk of developing Aspergillus infection, due to the combined effects of profound and prolonged neutropenia and the use of high-dose steroid therapy for management of graft-versus-host disease.

Although pulmonary infection is the most frequent manifestation of aspergillosis, cerebral disease is not uncommon, being detected in 10 to 20% of all cases of invasive aspergillosis. Most patients die within a few days after diagnosis, with a mortality rate as high as 95% (10). Cerebral involvement can occur even without any pulmonary disease (2). In recent years, a sandwich enzyme-linked immunosorbent assay (ELISA), “Platelia Aspergillus” (Bio-Rad Laboratories, Marnes La Coquette, France), for diagnosis of Aspergillus infection by the detection of galactomannan (GM) antigens, has aroused interest (7, 9, 12, 13). The sensitivity and specificity of this test are good, and it sometimes gives positive results before the onset of symptoms or radiological abnormalities (9, 11, 13, 17). There is some evidence that the test is reliable even when used on bronchoalveolar lavage or cerebrospinal fluids (CSF) (8, 16, 18). In this study we have investigated the value of the Platelia Aspergillus ELISA with CSF specimens for the diagnosis of cerebral aspergillosis. All serum and CSF samples were processed according to the manufacturer's instructions, and ELISA results were tabulated as the ratio of the optical density of the sample tested to the optical density of the threshold positive control (GM index). All comparisons were performed using the Mann-Whitney U test for unpaired data sets. Data were expressed as box plots indicating median values with interquartile ranges to graphically represent the GM index in patient and control groups (StatView 4.2 program; Abacus Concepts, Berkeley, Calif.).

From 1995 to 2000, at the Bone Marrow Transplant Unit of the Hematology Department of San Martino General Hospital in Genoa, Italy, five patients with probable aspergillosis, as defined by the EORTC-MSG (European Organization for Research and Treatment of Cancer-Mycoses Study Group) consensus group (1), were observed. For all patients, albumin, immunoglobulin G (IgG), and glucose concentrations were calculated in parallel, in CSF and serum samples. In addition, the IgG index [calculated as (CSF IgG level/serum IgG level)/(CSF albumin level/serum albumin level)] was also determined as an estimate of intrathecal IgG production (15). The main clinical characteristics are shown in Table 1. In computed tomographic scans all patients had hypodense lesions, with ring enhancement. In no case was a cerebral biopsy performed, because of the patients' underlying conditions and severe thrombocytopenia. All patients died after a median of 40 days after diagnosis. Autopsies were performed for three of five patients and showed diffuse cerebral infiltration with hyaline septate, acute branching hyphae, diagnostic of cerebral mold infection. For these patients, premortem CSF cultures were negative and postmortem cultures were not performed. The two patients who did not undergo autopsy had at least one nasal swab positive for Aspergillus spp. within the 30 days preceding the onset of neurological symptoms. The median CSF GM index for the five patients with probable cerebral aspergillosis (eight samples) was 10.52 (mean, 81.931; standard deviation [SD], 200.958; range, 0.549 to 578.5). By contrast, the median CSF GM index from the 16 control patients (33 samples) was 0.287 (mean, 0.281; SD, 0.13; range, 0.09 to 0.749). As shown in Fig. 1, there was no overlap between the two groups of patients.

TABLE 1.

Clinical characteristics of five bone marrow transplantation patients with probable cerebral aspergillosis

Patient no.	Underlying diseasea	Time (days) from BMTb to:		Clinical presentation	Histology at autopsy	Aspergillus colonization
		Neurological symptoms	1st CSF sample
1	NHL	72	74	Fever, hemiparesis, and coma	Diffuse cerebral infiltration with hyaline septate, acute branching hyphae	No
2	ALL	5	44	Fever; neurological deficit at the right arm and at the left leg; rapid neurological deterioration; seizures; coma	Diffuse meningeal and parenchymal infiltration with hyaline septate, acute branching hyphae	No
3	SAA	113	118	Fever, seizures, sensory loss	Not performed	Yes
4	ALL	54	61	Fever, sinusitis, seizures, sensory loss, hemiparesis, and coma	Not performed	Yes
5	AML	185	190	Anisocoria, hemiparesis	Diffuse cerebral infiltration with hyaline septate, acute branching hyphae	No

^aNHL, non-Hodgkin lymphoma; ALL, acute lymphoblastic leukemia; SAA, severe acute aplasia; AML, acute myeloid leukemia.

^bBMT, bone marrow transplantation.

FIG. 1.

Box plot analysis of CSF GM index for patients with or without cerebral aspergillosis, demonstrating the specificity of the test when performed in CSF.

To rule out the possibility that the presence of GM in CSF was merely due to passive transfer, the percentage of intrathecally produced GM was calculated for each patient. The chemicophysical characteristics of the GM molecule have not been completely characterized yet. According to the information available (6), the fraction of GM reacting with the EB-A2 monoclonal antibody used in this test is part of a more complex molecule whose molecular weight is intermediate between those of albumin and IgG. Therefore, we postulated that the CSF albumin/serum albumin quotient (QAlb), which indicates the amount of albumin crossing the blood-brain barrier, could represent a maximal estimate of the amount of GM crossing the barrier. As shown by the increased QAlb values (Table 2), all patients had some degree of blood-brain barrier damage. Based on the QAlb, we calculated the expected CSF GM index, i.e., the index we should have found if the presence of GM in CSF were due only to passive transfer. Finally, the actual CSF GM index was estimated and, for each patient, the difference between the measured and expected CSF GM indexes (which indicates intrathecally produced GM) was calculated and divided by the measured GM index. The final quotient represented the percentage of intrathecally produced GM. Overall, the median expected CSF GM index was 0.053 (SD, 0.196), while the median measured CSF GM index was 26.144 (SD, 252.900). The difference was highly statistically significant (P < 0.001). The percentage of the total CSF galactomannan that was intrathecally produced ranged from 95.6 to 99.9%.

TABLE 2.

CSF and serum parameters, expected and measured CSF GM index, and percentage of total GM content that was intrathecally produced

Patient no.	QAlb (mg/dl)a	IgG indexb	Serum GM index	CSF GM index		Intrathecally produced GMc
				Measured	Expected
1	23.886	0.671	3.099	1.715	0.074	95.684
2	18.605	0.674	25.709	578.499	0.478	99.917
3d	8.125	0.699	2.626	26.378	0.021	99.919
4e	17.778	0.647	3.520	26.144	0.063	99.761
5	11.950	0.988	1.307	0.549	0.016	97.155

^aNormal values, 1.92 to 7.3 mg/dl.

^bNormal values, 0.236 to 0.669.

^cExpressed as a percentage of total GM.

^dIn one additional CSF sample, the GM index was 18.417.

^eIn two additional CSF samples, the GM index values were 2.586 and 1.161.

To determine whether patients with cerebral or pulmonary aspergillosis might be distinguished just on the basis of the serum GM index (which might render the CSF test unnecessary), the test was also performed on a second control group of five patients with documented pulmonary aspergillosis and without cerebral aspergillosis (45 samples), and these results were compared with those obtained for our original group of patients with cerebral aspergillosis (five patients, 56 samples). As shown in Fig. 2, the median serum GM index was higher in patients with cerebral aspergillosis (median, 3.05; mean, 6.007; SD, 5.91; range, 1 to 25.709) than in patients with pulmonary aspergillosis (median, 2.118; mean, 37.154; SD, 212.082; range, 0 to 1,424.14) (P < 0.01). However, a box plot analysis showed considerable overlap between the two groups (95% confidence interval, 4.46 to 7.55 for cerebral aspergillosis versus −18.39 to 92.7 for pulmonary aspergillosis). In fact, 78% of the GM index values in patients with cerebral aspergillosis were contained within the 25th and 75th percentiles of those obtained from patients with pulmonary aspergillosis.

FIG. 2.

Box plot analysis of serum GM index for patients with pulmonary or cerebral aspergillosis, demonstrating the poor discriminatory ability of the test when performed in serum.

We recognize that none of our patients with cerebral disease might be defined as having documented aspergillosis, since in no case was confirmation obtained by culture from infected tissues. Three of these patients had autopsy-documented mold infections, and two had compatible clinical disease with evidence of nasal colonization. However, all of them had several positive results by serum GM antigen detection, a test which has been shown to be sensitive and specific for the diagnosis of aspergillosis. According to some authorities, positive Aspergillus GM antigenemia combined with appropriate clinical signs in patients at high risk for aspergillosis constitutes evidence which is almost equivalent to histological documentation of hyphae, even in the absence of confirmation by culture (3). In addition, Aspergillus accounts for most of the cases of brain abscess in bone marrow transplant recipients (2), and cross-reactivity for the EB-A2 monoclonal antibody has been demonstrated only with filamentous fungi, which usually are not involved in invasive infections (5, 9, 14). A possible exception is Fusarium, which, however, rarely causes isolated cerebral lesions. Moreover, a recent and specific report seems to exclude any cross-reaction (14). For all these reasons, we considered another etiology extremely unlikely. Other studies have already suggested the potential value of the GM antigen detection test for diagnosis of cerebral aspergillosis. Verweij et al. (16) studied 26 CSF samples from a single patient with documented Aspergillus meningitis and found that the test result became positive 45 days before the positive CSF culture. Comparison with control CSF specimens from patients without aspergillosis showed good specificity. These authors also suggested that the cutoff for CSF positivity might be lower than the cutoff for serum positivity. Kami et al. (4) studied 5 patients with documented cerebral aspergillosis and 11 controls. They found that the test was 100% sensitive and specific for the diagnosis of cerebral aspergillosis.

In conclusion, in the present study we showed a statistically significant difference in the CSF GM indexes between patients with cerebral aspergillosis and a control group of patients with other neurological diseases. The great majority of the GM in CSF (95.68 to 99.92%) was apparently produced intrathecally, as shown by the difference between expected and measured CSF GM indexes in cerebral aspergillosis patients. Finally, serum GM indexes seem to be unable to discriminate between patients with pulmonary and cerebral disease, as shown by the remarkable overlapping of values between the two groups. Our study suggests that the Aspergillus CSF GM index might be diagnostic for cerebral aspergillosis in patients at high risk for aspergillosis and with a compatible neurological disease. Further studies on larger patient populations might be necessary in order to fully evaluate the role of this test in the diagnosis of this dreadful disease.