Optimization of cutoff values for (1→3)-β-d-glucan and galactomannan assays in cerebrospinal fluid for the diagnosis of non-cryptococcal fungal infections of the central nervous system

Abstract

Fungal infections of the central nervous system (FI-CNS) pose substantial diagnostic challenges, owing to their diverse clinical presentations and the limited sensitivity of conventional diagnostic tests. Although serum (1→3)-β-d-glucan (BDG) and galactomannan (GM) assays are FDA-approved for the diagnosis of invasive fungal infections (IFIs), their effectiveness in cerebrospinal fluid (CSF) remains underexplored, and optimal cutoff values in CSF are not well established. This study aimed to assess the utility of BDG and GM assays in CSF for diagnosing non-cryptococcal FI-CNS. We conducted a prospective observational study at the National Institute of Mental Health and Neuro Sciences in India from January 2022 to December 2023, including CSF samples from patients suspected of fungal meningitis. The cases were categorized as proven, probable, or possible FI-CNS based on the revised EORTC/MSGERC criteria. Among 61 suspected cases, 2 were proven, 48 were probable, and 11 were possible FI-CNS. The control group included 23 patients without FI-CNS suspicion. BDG and GM testing in CSF followed manufacturers’ guidelines for serum. At the manufacturer’s recommended cutoff of 80 pg/ml, sensitivity of BDG was 94% and specificity was 78.3%. For GM, using the manufacturer’s recommended cutoff of 0.5 optical density index (ODI), sensitivity was 42% and specificity was 100%. Receiver operating characteristic curve analysis indicated optimal cutoffs of 72 pg/ml for BDG (sensitivity 96%, specificity 78.3%) and 0.47 ODI for GM (sensitivity 44%, specificity 100%). Combining both biomarkers increased sensitivity to 97.8%, suggesting that combined BDG and GM testing in CSF could significantly enhance the diagnostic accuracy and management of FI-CNS.

Introduction

Fungal infections of the central nervous system (FI-CNS) present considerable diagnostic challenges due to their diverse clinical presentations, non-specific imaging findings, and the limited sensitivity of conventional diagnostic tests.¹ An accurate and prompt diagnosis is essential for effective treatment and improved patient outcomes. While cryptococcal meningitis is the most common FI-CNS worldwide,^2,3 the incidence of non-cryptococcal CNS infections, such as hematogenous Candida meningoencephalitis (HCME), cerebral aspergillosis, and other mould infections, has increased over the past decade.^4,5Candida meningoencephalitis can occur spontaneously as a complication of disseminated candidiasis, particularly in premature neonates,⁶ or as a result of direct inoculation following head trauma or neurosurgical procedures, such as placement of ventriculostomy drains, shunts, stimulators, or prosthetic devices.^7,8 CNS candidiasis may manifest as microabscesses, vascular complications, or localized CNS candidiasis with macroabscesses and meningitis complicated with intracranial hypertension.^9,10 While older autopsy studies indicated that Candida species were responsible for 49% of cerebral mycoses cases¹¹ and that CNS involvement occurred in 48% of invasive candidiasis cases,¹² CNS dissemination is now rare, except in preterm infants due to their immature blood–brain barrier.¹³ Studies have shown that 5%–9% of candidemic infants develop Candida meningitis, with a mortality rate approaching 41%.⁶ In pediatric patients, HCME often presents with seizures, intraventricular hemorrhage, cortical blindness, neurocognitive deficits, and loss of developmental milestones. Apart from infants, CNS candidiasis has been reported in patients with hematological malignancies, organ transplants, intravenous drug use, diabetes, human immunodeficiency virus (HIV) infection, and primary immunodeficiencies like caspase recruitment domain-containing protein 9 (CARD9) deficiency. CNS embolic complications are reported in 12%–22% of Candida infective endocarditis cases.¹⁰

While moulds such as Aspergillus spp., Fusarium spp., Mucorales (such as Rhizopus, Mucor, and Lichtheimia), Cladophialophora bantiana, or Exserohilum rostratum are typically rare causes of FI-CNS, sporadic cases and iatrogenic outbreaks can have devastating consequences.^14–17 These infections are often linked to immune compromise and are usually acquired through the inhalation of conidia or direct inoculation after trauma or surgery, with subsequent hematogenous or contiguous spread.² Additionally, dimorphic fungi, such as Histoplasma capsulatum, Blastomyces dermatitidis, and Coccidioides spp., are important causes of FI-CNS in endemic regions, with their geographic distribution expanding due to several factors.¹⁸ These infections often result in severe outcomes, including meningitis, brain abscesses, spinal cord lesions, and meningoencephalitis, and can be fatal, especially in immunocompromised individuals.⁴

Traditionally, FI-CNS are diagnosed by culture or direct microscopic demonstration of the fungus in cerebrospinal fluid (CSF) or brain biopsy samples. While culture remains the gold standard for diagnosis of invasive fungal infections (IFIs), its clinical utility is limited by low sensitivity and delayed results.^14,19 Histopathological examination of brain tissue from image-guided stereotactic biopsy is confirmatory but is invasive, resource-intensive and often not feasible.^4,20 Non-culture-based tests, such as biomarker assays, have the potential to enhance fungal diagnostics in terms of speed, accuracy, and cost-effectiveness, while obviating the need for extensive diagnostic procedures.²¹ The detection of cryptococcal capsular polysaccharide antigen in CSF has revolutionized the detection and therapeutic monitoring of cryptococcal meningitis.²² However, there is a paucity of specific biomarkers for diagnosing non-cryptococcal FI-CNS. Recently, biomarkers like (1→3)-β-d-glucan (BDG) and galactomannan (GM) are increasingly being evaluated as adjuncts to traditional diagnostic approaches for FI-CNS.²³

BDG is a polysaccharide found in fungal cell walls, particularly in Candida spp. and several ascomycetous moulds, including Aspergillus spp., Fusarium spp., Scedosporium spp., and E. rostratum.²⁴ BDG is typically undetectable in the CSF above a certain analytical threshold (30 pg/ml) when the blood–brain barrier is intact.²⁵ However, in isolated CNS infections, BDG becomes compartmentalized, leading to higher concentrations in the CSF. This, coupled with minimal spillover into the venous blood and a slower rate of clearance from the CSF, results in elevated BDG levels in FI-CNS.^26,27 These features make BDG a useful adjunct for the early detection of FI-CNS, even before culture results are available.

GM is a polysaccharide antigen present in the cell walls of Aspergillus spp., Fusarium spp., Paecilomyces spp., Penicillium spp., and H. capsulatum. Traditionally, detecting GM in serum and bronchoalveolar lavage fluid (BALF) has been a valuable method for diagnosing invasive pulmonary aspergillosis.²⁸ However, its utility also extends to CSF testing, offering significant diagnostic insights in cases of cerebral aspergillosis. In patients with chronic meningitis, detecting GM in CSF helps distinguish CNS mould infections from HCME. Viscoli et al.²⁹ postulated that the CSF albumin/serum albumin quotient could estimate the amount of GM crossing the blood–brain barrier. By comparing expected and actual CSF GM indexes, they found that 99% of GM in the CSF of patients with cerebral aspergillosis was produced intrathecally, reflecting blood–brain barrier damage and confirming that high GM levels in CSF are indicative of localized CNS infection rather than translocation from the bloodstream. Furthermore, the presence of GM in CSF is included as a microbiological criterion in the revised European Organization for Research and Treatment of Cancer/Mycoses Study Group Education and Research Consortium (EORTC/MSGERC) definitions for diagnosing invasive mould diseases of the CNS.³⁰

Integrating BDG and GM assays into the diagnostic algorithm for FI-CNS could facilitate early detection and improve patient outcomes, particularly when traditional methods like fungal cultures or imaging provide inconclusive or delayed results. Currently, none of these assays are approved for testing of CSF samples. The available data on the performance of these biomarkers in CSF are mainly derived from retrospective reviews and studies involving specific cohorts, showing variable sensitivities and specificities across different cutoffs.^23,31,32 Unlike serum and BALF, the positive and negative predictive values (PPV and NPV) and optimal thresholds of fungal biomarkers in CSF samples have not been established. To date, only a few studies have evaluated their performance for the diagnosis of non-cryptococcal FI-CNS. Considering their promising role in the diagnosis of FI-CNS, we evaluated the performance characteristics and clinical utility of BDG and GM assays in CSF at different thresholds for test positivity.

Materials and methods

Study specimens

This prospective observational study included residual CSF samples from patients who had their samples submitted to the Neuromicrobiology Laboratory of NIMHANS for evaluation of fungal meningitis between January 2022 and December 2023. Patient data, including demographics, predisposing factors, underlying disease, HIV status, and any history of organ transplantation or immunosuppression, were recorded. Diagnostic data included bacterial and fungal culture results from blood, CSF, and respiratory specimens (e.g., sputum or BALF), as well as serological data for toxoplasmosis, syphilis, brucellosis, leptospirosis, scrub typhus, Lyme borreliosis, neurocysticercosis, and salmonellosis. Pulmonary and CNS imaging findings were also documented, along with histopathological results from biopsies of the CNS and other sites, when available. The CSF samples were prospectively collected and screened for cryptococcal capsular polysaccharide antigen using CrAg lateral flow assay (IMMY Inc., Norman, OK, USA). The samples that tested negative for cryptococcal meningitis were stored frozen and tested in batches for BDG and GM as a part of the study protocol. A single CSF sample was collected from each patient by lumbar puncture (LP) and tested only once. Serum samples were not included for evaluation, as previous studies demonstrated a poor correlation between serum BDG and GM levels and fungal invasion of the CNS.^26,27,29

Case definitions

The diagnosis of meningitis required the presence of compatible symptoms and/or signs accompanied by CSF pleocytosis (CSF white blood cell count, >5 cells/μl), hypoglycorrhachia (CSF glucose, <40 mg/dl), or elevated protein (CSF protein, >50 mg/dl), and brain imaging abnormalities (micro- or macroabscesses, multiple hemorrhages/infarction or ischemia, meningeal enhancement, and space-occupying or ring-enhancing lesions).¹ FI-CNS was defined according to the revised EORTC/MSGERC criteria for IFIs³⁰ with adjustments to accommodate for FI-CNS detailed below. In order to avoid major bias from the inclusion of the assays under validation, CSF GM was removed as a microbiological criterion to define probable FI-CNS.

Proven FI-CNS was defined by compatible brain imaging findings, along with evidence of fungal elements in brain histopathology or a positive culture from CSF or a brain biopsy specimen. The diagnosis of probable FI-CNS required the presence of at least one host factor, compatible brain imaging findings on CT or MRI scans, a proven or probable IFI at a non-CNS site, and the lack of an alternative diagnosis for brain lesions. The medical records were examined to confirm that no alternative diagnoses were present for 2 years following the initial diagnosis. Possible FI-CNS was defined by the presence of a host factor and compatible brain imaging findings, but without mycological evidence of an IFI (i.e., negative or absent culture, microscopy or histopathology) and no other alternative diagnosis for the brain lesions.

Patients were classified as immunosuppressed if they met at least one of the following criteria: neutropenia (<500 neutrophils/µl) for >10 days, HIV seropositive with a CD4 lymphocyte count of ≤200 cells/µl, daily use of >10 mg of systemic corticosteroids for a minimum of 4 weeks, anti-tumor necrosis factor therapy within the last 6 months, solid organ or hematopoietic stem cell transplant recipients on long-term immunosuppressive therapy, malignancy requiring cytotoxic chemotherapy, or receiving systemic immunomodulators for other medical conditions. For immunocompromised patients, CSF with a normal cell count, normal to mild protein elevation, and normal to slightly low glucose levels were still considered for analysis, as they often exhibit less pronounced inflammatory responses in their CSF compared to immunocompetent individuals with meningitis.

Patients with no evidence of immunosuppression but with one or more direct inoculation risks, such as trauma or neurosurgical procedures (e.g., placement of ventriculostomy drains, shunts, stimulators, and prosthetic reconstructive devices), were classified as immunocompetent.

The control group only included patients with a definitive alternative diagnosis for brain lesions, no evidence of IFI, and no brain imaging abnormalities consistent with FI-CNS. This included individuals diagnosed with Parkinson’s disease, stroke, demyelinating disorders, autoimmune encephalitis, Wernicke’s encephalopathy, viral encephalitis, cerebral venous thrombosis, carcinomatous meningitis, tuberculous meningitis, and spinocerebellar ataxia.

CSF analysis and fungal biomarker assays

The CSF samples were subjected to cytological and biochemical analyses (measurement of protein, glucose and lactate levels). For culture, 500 µl of CSF was inoculated onto Sabouraud dextrose agar and incubated at 37°C for 4 weeks.

The CSF samples were tested for BDG using the Fungitell® Assay (Associates of Cape Cod, Inc., East Falmouth, MA, USA) according to the manufacturer’s recommendations for serum. The absorbance was measured spectrophotometrically using an ELx808 Absorbance Microplate Reader (Bio-Tek Instruments Inc., Winooski, VT, USA). The kinetic curves were inspected at 405 nm, 490 nm, and delta OD (405–490 nm) for signs of interference and were evaluated for a smooth increase in optical density, comparable to that of the standards, as per the manufacturer’s instructions. Kinetic curves that showed evidence of optical interference and/or unusual kinetic patterns were excluded from analysis. Data analysis was performed using the Gen5 Secure v. 3.11 software (Bio-Tek Instruments Inc., Winooski, VT, USA). The glucan concentration was determined from the average of the MeanV replicate values derived from the kinetic standard curve. The results were considered valid if at least two sample replicates had a coefficient of variation (CV) <30% for BDG values ≥60 pg/ml (indeterminate or positive using the manufacturer’s serum cutoffs), or if all sample replicates had BDG values ≥80 pg/ml (serum category positive) or 60–79 pg/ml (serum category indeterminate) or <60 pg/ml (serum category negative). The CSF samples that failed to meet these criteria were retested in the next assay, and the results were considered valid if they met the acceptance criteria. The reportable range of the Fungitell® BDG assay for serum samples is 31 pg/ml to 500 pg/ml. In this study, the diagnostic performance of BDG in CSF samples was determined by setting the upper limit of BDG concentration at 523.438 pg/ml and the lower limit at 7.812 pg/ml.

For all cases, GM was evaluated using the Platelia® Aspergillus Ag enzyme immunoassay (Bio-Rad, Marnes-la-Coquette, Paris, France), according to the manufacturer’s recommendations for serum, but using 300 µl of CSF instead of serum. CSF GM was tested in duplicate with a positivity threshold of 0.5 optical density index (ODI). In cases of discordant duplicates, strategies including repeat testing, additional diagnostic tools (such as culture or PCR), and clinical correlation were employed to determine whether the discordance was due to assay performance, technical errors, or other factors like treatment response or comorbidities.

Data analysis

Descriptive statistics were used to analyse the demographic and clinical characteristics of the four patient groups: proven, probable, and possible FI-CNS and control. For categorical data, frequency and proportion were used with confidence intervals (CIs) and expressed as median with interquartile range (IQR) or mean ± standard deviation (SD). Patients in the control group were compared with proven, probable and possible categories using Pearson’s χ² test/Fisher’s exact test (for categorical variables) and the Kruskal–Wallis H test with Dunn’s pairwise comparison (for continuous variables). Logistic regression models were created including both BDG and GM as continuous covariates to assess whether combining these markers would improve predictive performance. In order to determine the optimal BDG and GM cutoff values for CSF samples, the receiver operating characteristic (ROC) curves were created by plotting the sensitivities of BDG and GM assays at several numerical cutoff values vs. 1 − specificity at those levels. The area under the curve (AUC) was estimated with an associated 95% CI using the trapezoidal rule for BDG and GM, individually. The sensitivity, specificity, PPV, and NPV were estimated as binomial proportions over a range of previously reported threshold values for both BDG and GM. The optimal cutoff values for BDG and GM were calculated based on Youden’s J index. A two-tailed P-value of < .05 was considered significant. Statistical analyses were performed using SAS version 9.3 (SAS Institute, Inc., Cary, NC) and GraphPad Prism version 10.2.3 for Windows (GraphPad Software, Boston, MA, USA).

Results

A total of 75 CSF samples were received for evaluation of fungal meningitis during the study period. Five samples were from patients who did not meet the case criteria, while nine had inconclusive BDG results after repeat testing (CV > 30%) and hence were excluded from analysis. Finally, 61 CSF samples from suspected cases of FI-CNS were considered for evaluation. The control group included 23 CSF samples from patients for whom FI-CNS was not suspected. Among 61 suspected cases that met the inclusion criteria, 2 (3%) were classified as proven, 48 (79%) as probable, and 11 (18%) as possible FI-CNS. The mean age of FI-CNS suspected groups was 38.2 ± 15.7 years, and 31 (51%) were males. The control group consisted of 23 patients with a mean age of 37.5 ± 15.6 years. Table 1 outlines the host factors and clinical, radiological, and laboratory findings that were used to categorize the patients into proven, probable, and possible FI-CNS and control groups. In FI-CNS suspected cases, the most common clinical presentations were headache (74%), altered mental status (49%), fever (43%), limb weakness (34%), focal neurological deficit (33%), seizures (30%), and confusion (28%). The median duration of symptoms was 60 days (range, 21–730 days). A total of 46 (75.4%) FI-CNS suspected cases exhibited CSF pleocytosis (>5 cells/μl), with a median CSF cell count of 42 cells/µl (range, 6–900 cells/µl). Lymphocytic pleocytosis was observed in 65.6% of the cases, while 9 (14.8%) patients had no CSF abnormalities. Elevated CSF protein (>50 mg/dl) and lactate (>25 mg/dl) levels and hypoglycorrhachia (<40 mg/dl) were present in 68%, 49%, and 15% of the patients, respectively (Table 1).

Meningeal enhancement was the most common radiological abnormality observed in 44% of FI-CNS suspected cases, followed by multiple infarcts (38%), T2/FLAIR hyperintensity (28%), ring-enhancing lesions (18%), space-occupying lesions (16%), features of raised ICP (13%), and nodular enhancement (6.5%) (Table 1). An initial diagnosis of CNS tuberculosis was made in 12 (19.7%) patients, including 1 proven, 8 probable, and 3 possible cases of FI-CNS, based on clinical and radiological findings.

Using the manufacturer’s recommended serum cutoff value of 80 pg/ml, CSF BDG was positive in 48 (79%) of 61 suspected cases of FI-CNS. The results were positive in both cases of proven FI-CNS with a mean BDG concentration of 408 pg/ml and 45 (94%) probable cases with a mean BDG concentration of 263.4 pg/ml and a median concentration of 226 pg/ml (range, 24–523 pg/ml). In the “possible” category, the mean and median BDG concentrations were 48.9 pg/ml and 43 pg/ml (range, 13–83 pg/ml), respectively. Eighteen (78%) of 23 patients in the control group had negative CSF BDG results with a median concentration of 38 pg/ml (range, 8–340 pg/ml) (Kruskal–Wallis H test, χ² = 43.06; P < .001). Five patients in the control group (Parkinson’s disease [n = 2], Wernicke’s encephalopathy [n = 1], stroke [n = 1], and tuberculous meningitis [n = 1]) had positive CSF BDG with concentrations ranging from 95 to 340 pg/ml (Table 2 and Fig. 1A). A non-fungal definitive diagnosis was retained in these patients and could therefore be considered as false-positive results.

At the manufacturer’s recommended serum cutoff of 0.5 ODI, CSF GM was positive in 21 (34%) suspected FI-CNS patients, indicating CNS mould disease. This includes both proven cases (mean GM ODI, 2.04) and 19 of 48 (39.6%) probable cases (mean GM ODI, 0.69 ± 0.71). The patients in the “possible” (mean GM ODI, 0.22 ± 0.12) and “control” (mean GM ODI, 0.22 ± 0.11) groups had negative CSF GM results (Kruskal–Wallis H test, χ² = 18.99; P < .001) (Table 2 and Fig. 1B). The correlations between BDG and GM levels for proven, probable, and possible FI-CNS and control groups are shown in Fig. 2A.

Since the final diagnoses in possible FI-CNS cases were uncertain, they were excluded from ROC curve analyses and performance evaluations. The results of BDG and GM assays in CSF at different cutoff values are shown in Table 3. The ROC analysis for discrimination between FI-CNS cases and controls yielded an area under the ROC curve (AUC) of 0.894 (95% CI, 0.806–0.982) for BDG (Fig. 2B) and 0.760 (95% CI, 0.648–0.872) for GM (Fig. 2C). The optimal cutoff with the highest Youden’s J index was 72 pg/ml for BDG (sensitivity 96%, specificity 78.3%, PPV 90.6%, and NPV 90%), and 0.47 ODI for GM (sensitivity 44%, specificity 100%, PPV 100%, and NPV 45.1%). None of the possible cases had a positive GM test.

We investigated the performance of a combination of BDG and GM in cases of FI-CNS. Defining a positive diagnosis as either test positive and a negative diagnosis as both tests negative, the sensitivity and specificity of the combined assay were 96.5% and 78.3% at the manufacturers’ recommended BDG/GM serum cutoffs of 80/0.5, respectively, while they were 97.8% and 78.3% at the optimal cutoffs of 72/0.47 identified in the present study. In a logistic regression model, excluding the possible FI-CNS cases, when both BDG and GM were included as continuous covariates, the effect of BDG was statistically significant (P = .0006), while GM was not (P = .069). Combining BDG with GM resulted in an improvement in the AUC (0.918 [95% CI, 0.842–0.994]) compared to BDG alone (0.894 [95% CI, 0.806–0.982]) (Fig. 2D).

Discussion

We investigated the comparative performances of two biomarkers, BDG and GM, in CSF for diagnosing FI-CNS. Using manufacturers’ recommended serum cutoff values and optimized CSF thresholds, the study assessed the diagnostic accuracy of these biomarkers alone and in combination for distinguishing between proven, probable, and possible FI-CNS cases, as well as control conditions.

As previously reported,^33,34 the majority (49/61, 80.3%) of FI-CNS in our study occurred in immunocompromised patients with IFI, indicating that these cases are likely a result of hematogenous spread from a primary site. In immunocompetent patients, direct inoculation following neurosurgical procedures was the primary risk factor. However, a systematic review of 235 proven cases of CNS aspergillosis found that only 43.5% of the patients were immunocompromised³⁵ suggesting that such infections can occur even in the absence of classical risk factors. As noted earlier,²⁵ the CSF cell count and biochemical findings in our survey were highly variable, making them unreliable for diagnosing FI-CNS. Nearly 15% of the patients had no CSF abnormality, and a presumptive diagnosis of FI-CNS was based on clinical presentations. Although radiological features are non-specific, unique imaging findings in the form of leptomeningitis, multiple infarcts, T2/FLAIR signal abnormalities, and ring-enhancing lesions were observed in our patients, consistent with previous studies.^1,36,37 The fact that 19.7% of patients initially diagnosed with CNS tuberculosis had clinical and radiological features mimicking FI-CNS underscores the diagnostic challenge posed by these infections, particularly in countries where tuberculosis is endemic. Differentiating between tuberculosis and fungal infection is crucial, as treatment strategies differ significantly.

The utility of CSF BDG for the diagnosis of FI-CNS was first described by Petraitiene et al.²⁷ They found that BDG levels in CSF correlated directly with the brain fungal burden, as demonstrated in a preclinical rabbit model of HCME. Further studies also highlighted the usefulness of CSF BDG in monitoring the response to antifungal therapy.^38–40 In addition to HCME, CSF BDG has also been evaluated for diverse causes of fungal meningitis, including infections caused by Exserohilum,^39–41Coccidioides,³¹Cryptococcus,⁴²Histoplasma,⁴³ and Aspergillus spp.^38,41,44 However, the CSF BDG cutoffs used in these studies were not uniform, resulting in variable performance characteristics across different thresholds. Our study shows that at the manufacturer’s recommended serum cutoff of 80 pg/ml, the CSF BDG assay demonstrates high sensitivity (94%) and moderate specificity (78.3%) for diagnosing FI-CNS. Notably, using an optimal CSF threshold of 72 pg/ml, determined by the Youden method⁴⁵ the sensitivity and NPV of the assay increased to 96% and 90%, respectively, with no change in specificity. So far, only two studies have evaluated the performance of BDG in CSF samples using optimal cutoff values around 70 pg/ml.^34,41 At a cutoff of 72 pg/ml, we observed a higher sensitivity compared to Bigot et al. (96% vs. 73% at a cutoff of 73 pg/ml)³⁴ and Malani et al. (96% vs. 91% at a cutoff of 66 pg/ml).⁴¹ However, our specificity was lower than the values reported in both studies (78.3% vs. 83.5%³⁴ and 92%⁴¹). The analysis by Malani et al.⁴¹ was conducted on a retrospective cohort of E. rostratum meningitis resulting from contaminated methylprednisolone injections in the context of a known exposure and high pre-test probability. Therefore, those findings cannot be generalized to other fungal causes of meningitis. On the other hand, the specificity of CSF BDG may have been underestimated in our study. Notably, two probable FI-CNS patients, one with Candida and the other with Aspergillus infection, had negative CSF BDG results. These patients had received 7 days of empiric amphotericin B treatment prior to LP, which could have contributed to the false-negative BDG results. Additionally, nine patients with possible FI-CNS had negative CSF BDG and GM results, and no clear explanation for the abnormal imaging could be identified. It is possible that a low fungal load, leading to minimal BDG release, or infection with less GM-producing moulds, could explain these false-negative results. Furthermore, five patients in the control group with alternative diagnoses for their brain lesions tested positive for CSF BDG, suggesting the possibility of either an undiagnosed fungal infection or a compromised blood–brain barrier allowing BDG translocation from the blood into the CSF. However, given the well-documented cross-reactivity of BDG with intravenous immunoglobulin, blood products, gauze, and certain antibiotics,^46,47 the possibility of diagnostic false-positive results in these five control cases cannot be refuted. Contrary to the findings of Bigot et al.³⁴ our analysis shows a high PPV for CSF BDG (90.6% vs. 20.5%), which could be attributed to differences in patient populations studied and the varying prevalence of fungal meningitis across different geographical regions. Notably, Bigot and colleagues³⁴ used fungal biomarkers other than BDG (GM, mannan, and Aspergillus fumigatus qPCR) to define highly probable and probable cases, a strategy we chose not to adopt in our study to avoid potential bias from including GM, which was evaluated in parallel. As previously reported,^26,40,41 we observed low CSF BDG levels in patients without evidence of FI-CNS, while those with proven meningitis had higher mean BDG concentrations (408 pg/ml) than probable cases (263.4 pg/ml). This supports the role of BDG in differentiating between varying levels of infection severity. Our findings reiterate those of Forster et al.²³ indicating that CSF BDG may be more valuable in patients with a high pre-test probability.

While the BDG assay does not identify or detect the presence of a particular fungal pathogen, GM is fairly specific for Aspergillus spp. However, cross-reactivity of GM has been observed with other moulds and dimorphic fungi⁴⁸ and recent studies have also reported false positives with respiratory tract Candida in non-hematological patients.⁴⁹ While a positive GM test in the context of Scedosporium infection is possible, it is uncommon. Kauffmann-Lacroix et al.⁵⁰ reported positive GM in a case of mycetoma caused by Scedosporium apiospermum, which was thought to result from cross-reactivity with Aspergillus GM or a dual infection where Aspergillus could not be isolated. Similarly, Ledoux et al.⁵¹ described positive GM in a case of pulmonary co-infection with Rhizopus oryzae and S. apiospermum, potentially due to an undetected Aspergillus co-infection. Therefore, the positive GM result in our confirmed FI-CNS case involving Scedosporium boydii may be explained by cross-reactivity or a dual infection with Aspergillus, which could not be cultured.

While there is extensive data on the performance of the GM assay in serum and BALF for invasive aspergillosis, its diagnostic accuracy in CSF samples remains largely unknown. Most available information on CSF GM is derived from anecdotal case reports, small retrospective cohorts, and systematic reviews of CNS aspergillosis.^52–54 There is limited data on its utility in non-Aspergillus FI-CNS. Komorowski et al.⁵⁴ conducted the first systematic review and meta-analysis on the diagnostic accuracy of CSF GM in CNS aspergillosis and found a mean sensitivity, specificity, PPV and NPV of 69%, 94%, 52.5%, and 96.3%, respectively, at an ODI cutoff of 0.5. An excellent diagnostic performance was also reported by Chong et al.⁵³ and Mercier et al.²⁸ using an ODI cutoff of 0.5 to 2.0. In this study, we observed that at an optimal cutoff of 0.47, CSF GM exhibited high specificity (100%), indicating its excellent ability to rule out CNS mould infections when the result is negative. However, its low sensitivity (44%) suggests that many FI-CNS cases may be missed, especially those caused by Candida or less GM-producing non-Aspergillus moulds. The reduced sensitivity may also be attributed to early-stage infections, where GM concentrations are too low to be detectable. This could explain the false negative CSF GM results in 28 probable FI-CNS cases. A previous study from our center found cerebral aspergillosis as the most common CNS mould infection in India, with Aspergillus flavus being the predominant pathogen, followed by A. fumigatus.⁵⁵ Therefore, the positive CSF GM results in 20 probable FI-CNS cases in this study likely indicate Aspergillus infection.

The ROC curve analysis and logistic regression modelling indicated that the BDG assay outperformed GM, highlighting the better overall performance of BDG in distinguishing between FI-CNS and non-fungal controls. This is expected, as the widespread presence of BDG across numerous fungal pathogens makes it a highly effective biomarker for detecting FI-CNS across a broad range of fungal species. Optimizing the cutoff values for both BDG and GM further improved the diagnostic metrics. The combined model, including both BDG and GM, showed an improvement in AUC, suggesting that both tests used in combination can enhance the diagnostic precision and provide a more comprehensive assessment of FI-CNS.

Our study involved a rigorous case selection process based on the revised EORTC/MSGERC criteria. Unlike previous retrospective studies, the prospective design of our study enhanced the robustness of our data and the interpretation of results. This was particularly crucial for assessing the utility of fungal biomarkers in rare conditions like FI-CNS. To ensure accuracy, all CSF samples were initially screened for potential bacterial causes of meningitis, preventing interference with the BDG and GM results. Additionally, the inclusion of the “possible” category added relevance to our findings, as studies indicate that a significant proportion of patients with possible IFI may actually have the disease, regardless of their immune status, thus warranting treatment by clinicians.^28,56,57

Our study has some limitations. First, the sample sizes in both the “proven” and “control” categories were relatively small due to the rarity of FI-CNS and the limited occurrence of non-infectious meningitis. Second, given that the sensitivities of traditional culture (14%) and PCR (29%) are relatively low,^41,58 our study does not have an ideal gold standard method for diagnosis, which may have resulted in some true cases being missed. This limitation prompted us to include an analysis based on host factors and clinical criteria. Third, our data may not accurately represent the disease prevalence in the at-risk population. The significant number of false positives in non-fungal conditions highlights a limitation of the BDG assay. It remains unclear if these false-positive results are due to biological variables in the CSF itself or to factors related to sample collection, such as the use of cotton gauze or alcohol swabs to clean the LP site. Moreover, BDG levels cannot determine the specific fungal species responsible for the infection. Thus, while the CSF BDG assay is a valuable tool for detecting FI-CNS, it should be interpreted in conjunction with relevant host factors, as well as clinical and radiological findings. We were unable to assess serum BDG levels in all suspected FI-CNS cases. Although serum BDG testing might seem like a potential alternative to LP for diagnosing FI-CNS, several studies have reported its low sensitivity and poor correlation with this condition.^26,27 CSF GM, on the other hand, while highly specific for CNS aspergillosis, lacks sensitivity, particularly in cases of FI-CNS caused by Candida or non-Aspergillus moulds.

Conclusion

We found that BDG, with an optimal cutoff of 72 pg/ml, and GM at 0.47 ODI in CSF samples, can effectively diagnose non-cryptococcal FI-CNS. These findings underscore the potential of BDG as a sensitive biomarker for detecting FI-CNS, though its specificity is compromised by false positives in non-fungal conditions. GM, while highly specific for Aspergillus infections, exhibits lower sensitivity, making it more appropriate as a confirmatory test for CNS aspergillosis rather than a screening tool. The combination of both biomarkers improves diagnostic accuracy, addressing both CNS candidiasis and mould infections, compared to using either test alone, though each has its limitations. Clinicians should interpret these findings alongside other diagnostic tools and clinical judgment to achieve a more precise diagnosis. Future research should aim to refine these assays and investigate additional markers to further improve the diagnostic accuracy and management of FI-CNS.

Acknowledgments

We acknowledge the support of Associates of Cape Cod, Inc. (Falmouth, MA, USA) and Anand Brothers (Karampura, New Delhi, India) for laboratory instrument set-up, staff training, logistics, and providing diagnostic kits and reagents.

Author contributions

Arghadip Samaddar (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Visualization, Writing—original draft), Gregory R. Kowald (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Resources, Software, Supervision, Validation, Writing—review & editing), Jenevi Margaret Mendonsa (Data curation, Investigation, Methodology, Software, Visualization, Writing—review & editing), Nagarathna S (Investigation, Supervision, Visualization, Writing—review & editing), and Veena Kumari H.B (Supervision, Software, Visualization, Writing—review & editing).

Funding

No funds were received for this study.

Conflict of interest

None of the authors has any commercial or other association that might pose a conflict of interest.

Data availability

The data supporting the findings of this study are available from the corresponding author upon reasonable request.

References