Methods for rapid diagnosis of meningitis etiology in adults

Infectious meningitis is caused by numerous pathogens and may be life threatening [1–3]. The classic symptoms of meningitis are fever, neck stiffness and headaches, although these features are not present uniformly [2–4]. Broadly, the most common causes of infectious meningitis are viral, bacterial, mycobacterial and fungal. Parasitic and non-infectious causes of meningitis also occur but are not the focus of this review.

In the USA, meningitis accounts for more than 72,000 hospitalizations with a total expenditure of 1.2 billion dollars annually [5]. Many types of infectious meningitis carry high rates of mortality and long-term complications including neurologic deficits and cognitive impairment [3,6,7]. Mortality due to bacterial meningitis ranges from 10 to 20% in high resource settings and as high as 50% in lower resource settings where bacterial meningitis is estimated as the fourth leading cause of disability [3,6–9]. In comparison to bacterial meningitis, aseptic (typically viral) meningitis typically caries a good prognosis (4.5% mortality rate) [3]. Tuberculous meningitis has a higher mortality rate, up to 50% in HIV-infected subjects and of individuals who survive tuberculosis meningitis, approximately 50% suffer from neurologic disability [6,10]. Like tuberculous meningitis, meningitis due to fungi carries significant mortality. Cryptococcal meningitis carries an acute in-hospital mortality of 30–50% [11]. World-wide, meningitis due to Cryptococcus accounts for 15% of AIDS-related deaths [12]. Other forms of fungal meningitis are rarer, but also deadly. Antinori et al. reported a case fatality rate of 63.5% due to Aspergillus meningitis for immunocompetent and 83% for immunocompromised patients [13]. Meningitis cause by Coccidioides has mortality of 90% at 1 year and 100% at 2 years if untreated [14]. CNS infections due to Histoplasma have a 39% case fatality rate [15].

It is not always possible to determine the etiology of meningitis. A study by Sulaiman et al. only identified the etiology in about 32% of cases, while other studies found similar ranges [16–18]. This is likely related to the lack of sensitivity of CSF cultures for non-bacterial pathogens as well as the underutilization of viral molecular and serologic testing [16]. Of course the etiology in any individual case may be influenced by many factors including test availability, geographic region, host, and many others – the main point is that the exact etiology is not always uncovered. Due to the high mortality and morbidity of many types of meningitis, it is critical to obtain a diagnosis or initiate empiric treatment rapidly as soon as possible [1]. Clinician diagnoses must be informed by historical information such as duration of symptoms, travel and country of origin, host immune capabilities, vaccination status, as well as an understanding of the appropriate diagnostic testing based on the probable epidemiology [1]. This review will focus on the diagnosis of bacterial, mycobacterial, fungal meningitis due to their worse prognosis and need for accurate diagnosis as well as the various diagnostic tests used in these conditions. Given viral meningitis typically is generally associated with relatively good outcomes, PCR testing of viral etiologies will not be a major focus of this review.

This is an extensive narrative review. We searched pubmed for ‘meningitis and diagnosis’, ‘epidemiology and meningitis’, ‘risk factors and meningitis’, ‘stem cell transplant and meningitis’, ‘organ transplant and meningitis’, ‘bacterial meningitis’, ‘tuberculosis meningitis’, ‘fungal meningitis’, ‘histoplasma meningitis’, ‘coccidioides meningitis’, ‘cryptococcal meningitis’, ‘blastomyces meningitis’ and ‘Aspergillus meningitis’ to gather data in addition to using material cited in some of the sources found via these searches.

Epidemiology

Although an estimated 16 million cases of bacterial meningitis occurred worldwide in 2013, only 4100 cases per year occur in the USA [2,19]. In Western countries the incidence of bacterial meningitis over the past 10–20 years has declined by approximately 3–4% per year and currently is approximately 0.8 cases per 100,000 per year [20]. The rates of bacterial meningitis are significantly higher in many African countries with an incidence of 10–40 per 100,000 persons per year [20]. The most common etiologies in order of frequency are Streptococcus pneumoniae, Neisseria meningitidis, and Listeria monocytogenes [4,20]. Meningitis due to H. influenzae type B and N. meningitidis has decreased over the past 10–20 years, due to vaccination [3,20,21].

Aseptic meningitis is relatively common with an annual incidence of 7.6 per 100,000 adults in the USA [6]. Enteroviruses, herpes simplex virus, varicella zoster virus and West Nile virus are the most common causes of infectious aseptic meningitis, other causes include cytomegalovirus and human immunodeficiency virus (HIV), among others [18,22,23]. Aseptic meningitis can also be caused by non-infectious etiologies such as adverse reaction to medications, chemotherapy, vaccinations or inflammatory diseases [22,23]. In 30–65% of cases of aseptic meningitis, the etiology is not definitively identified [22].

Tuberculosis (TB) meningitis incidence is not known, but likely occurs in about 1–5% of TB cases worldwide [24]. In 2017 WHO estimated 10 million incident cases of tuberculosis, which would correlate to 100,000–500,000 cases per year of TB meningitis [25].

Worldwide, fungal meningitis causes substantial mortality, particularly among immunocompromised persons [5]. The primary pathogens implicated are Cryptococcus, Coccidioides, Histoplasma, and Candida. In 2014, Rajasingham et al. estimated 223,100 incident cases of cryptococcal meningitis with 181,100 annual fatalities [12]. Sub-Saharan Africa accounted for 73% of the 223,1000 cases [12]. Meningitis due to Cryptococcus neoformans is most commonly seen in immunocompromised individuals, especially those with HIV [5,26]. However in high-income countries, 30% of cryptococcal meningitis occurs in apparently immunocompetent individuals, particularly if due to C. gattii [5]. Meningitis due to Coccidioides, Histoplasmosis or Candida typically occurs as a results of dissemination and so is relatively uncommon in hosts with intact immune systems. Meningeal involvement occurs in 33–50% of disseminated coccidioidomycosis and 10–20% of disseminated histoplasmosis [5].

Regional variation in frequency and etiology of meningitis occurs as well. For instance, the ‘meningitis belt’ (Senegal to Ethiopia in sub-Saharan Africa) experiences high numbers of bacterial meningitis, prior to the introduction of the conjugate meningococcal vaccine for serotype A [20,21]. In areas with high rates of HIV such as Malawi, South Africa, Uganda, Zambia and Zimbabwe, Cryptococcus and TB are the first and second most common causes of meningitis [26]. In Iran, H. influenza type B meningitis is relatively common due to a lack of routine vaccination [9]. Coccidioides meningitis is a frequent causes of meningitis in Arizona and California whereas Histoplasma meningitis is more common in the Mississippi and Ohio river valleys of the United States and much of Latin America [27,28].

Host factors

The likelihood of a particular etiology of meningitis is also affected by the host. In general, bacterial meningitis is more commonly seen in older individuals, possibly due to immune system decline over time [3,4,7]. Adriani et al. found that almost half of patients with bacterial meningitis had a predisposing condition and a third had an immunodeficiency (diabetes mellitus, alcoholism, cancer, history of organ transplant, HIV, asplenia, complement deficiency or anatomical defects) [7]. Common bacterial etiologies of N. meningitidis, H. influenzae and S. pneumoniae – all are encapsulated organisms that require protection via B lymphocyte production of immunoglobulins, immunoglobulin activating the complement cascade, and appropriate splenic function to remove circulating bacteria [3,7,29]. Disruptions, quantity or function of B cells, immunoglobulin, complement and asplenia increase risk of meningitis due to encapsulated organisms [29]. Candida meningitis is more common in patients with ventriculoperitoneal shunts, and cochlear implants have been associated with a higher incidence of pneumococcal meningitis [5,7,29]. Head trauma and neurosurgery are associated with higher rates of aerobic Gram-negative bacilli and Staphylococcus aureus [2,3].

Fungal (most notably cryptococcal) meningitis is more common in patients with CD4 T cell dysfunction, most notably persons living with HIV/AIDS [5]. Compared with those without HIV co-infection, persons with TB and HIV are more likely to develop extrapulmonary disease, including tuberculosis meningitis [30]. Finally, HIV-infected individuals are at higher risk than the general population for meningitis due to typical bacterial pathogens such as S. pneumoniae as well as less common bacteria such as Salmonella and Listeria species [7,29]. Figure 1 shows a diagnostic approach to subacute meningitis in persons with advanced HIV, Figure 2 shows a diagnostic approach to subacute meningitis in persons a presumed normal immune system. The figures are based on the literature cited within this review, but also our clinical experience. Of note, Figure 2 refers to the CSF to blood glucose ratio. Glucose is transported across the blood–brain barrier, therefore CSF glucose levels are directly rated to plasma levels [31]. Elevations in the plasma glucose will cause the CSF glucose levels to be elevated but CSF levels are decreased with certain CSF infections. When interpreting CSF glucose levels they should be compared with plasma levels [31]. CSF/blood glucose level is typically low in bacterial and tuberculous meningitis, but can also be seen with other causes as well [3,27,32,33].

TNF-α inhibitors are commonly used for a number of auto-immune conditions and have been linked to numerous infections including histoplasmosis and tuberculosis, including meningitis at a higher rate than the general population [5,34–36]. Moreover, infliximab in particular causes increased incidence of Listeria meningitis although other TNF-α inhibitors have been linked as well [34,37,38]. Among solid organ transplant recipients, the annual incidence of bacterial meningitis is 5.4-times higher than the general population, Listeria meningitis is of particular concern (110-fold increased risk) [7,39]. Cryptococcal meningitis risk is also high among solid organ transplant recipients and those with cirrhosis [40,41]. Patient who have undergone hematopoietic stem cell transplant (HSCT) are also at increased risk for infectious meningitis, particularly those having undergone allogenic HSCT [42]. S. pneumoniae, N. meningitidis, and Aspergillus spp. are of particular concern in the HSCT population [27,42–45].

Duration of symptoms

Meningitis can be broadly categorized as acute (<5 days), subacute (5–30 days), and chronic (>30 days). Bacterial meningitis typically presents in less than 24 h and though viral meningitis may have an acute presentation, generally, the onset of symptoms is not as rapid (median 2 days) nor as severe [3,6,16,46]. However, it is important to note that patients with certain risk factors, such as those with a solid organ transplant, may have a more protracted clinical course compared with the general population [39]. Another exception is Listeria meningitis with approximately one third of patients having symptoms for more than four days [1,2]. Subacute or chronic meningitis is typically due to viruses, fungi or mycobacteria (M. tuberculosis); 24 weeks of symptoms is most typical of latter two categories [1,16,27,47]. Yet, TB meningitis can occur with as little as 1 week or as long as 1 year of symptoms [33]. Moreover patients with Histoplasma meningitis typically have symptoms for at least 2 months and Blastomyces meningitis is typically diagnosed after an average of 3 months of symptoms [15,48,49]. Table 1 summarizes typical durations of symptoms by etiology.

General diagnostic tests

Cerebrospinal fluid (CSF) analysis is key to determining the etiology of meningitis [2,46,50]. Nonspecific, standard CSF testing includes white blood cell count with differential, glucose, and protein [1,46]. Initial results return within hours (ideally faster) and can help one differentiate the more likely etiologies. Yet, none of these tests have high specificity and are to be interpreted in the setting of other relevant clinical information. Table 2 shows the typical patterns of basic CSF tests for by etiology.

Testing the CSF for inflammatory markers, such as C-reactive protein (CRP), procalcitonin, lactic acid and ferritin may aid in the differentiation of bacterial from viral meningitis but are not commonly used and have been more frequently studied as blood tests [4,51–53]. It should be noted that CSF lactate, in particular is used in some settings. However, a CSF lactic acid concentration of 4.2 mmol/l or greater has a sensitivity of 96% and specificity of 100% for differentiating bacterial from viral meningitis although elevations also occur in cryptococcal meningitis, TB meningitis and a number of noninfectious conditions such as seizure, ischemia or hemorrhage [52,53]. Other CSF biomarkers to differentiate bacterial from viral meningitis include procalcitonin (cut-off 0.235 pg/ml, sensitivity 96.4%, specificity 80%), ferritin (cut-off 4.6 ng/ml, sensitivity 92.9%, specificity 68%), and CRP (cut-off 1.3 mg/l, sensitivity 92%, specificity 84%) [51]. Finally, CT and magnetic resonance imaging (MRI) of the head can provide additional information in diagnosis of suspected meningitis but in most cases are not helpful in determining the causative agent [27].

Table 3 summarizes the major, specific tests by etiology of meningitis.

Bacterial meningitis except tuberculous meningitis

Culture & Gram stain

CSF culture is considered the gold standard to diagnosis bacterial meningitis, yet is only positive in 70–85% of persons with bacterial meningitis who have not received antimicrobial therapy prior to lumbar puncture [19,51,52]. Final results of the culture are often not available for 48 h or more [19,52,54]. Gram stain is more rapid and has good specificity but sensitivity is poor (10–93% depending on the organism and whether or not antibiotics were given prior to CSF collection) [19,52]. Gram stain is most useful for S. pneumoniae.

Latex agglutination

Latex agglutination is relatively inexpensive and rapid (∼10 min) [2,6,55–58]. Sensitivity varies by microorganism and manufacturer – 78–100% for H. influenzae, 59–100% for S. pneumoniae, 69–100% for S. agalactiae and 50–93% for N. meningitis [4,52]. Kits to detect multiple common bacterial pathogens, have also been developed [8,55]. Latex agglutination may be particularly useful in patients with a negative Gram stain when the LP is performed after antibiotic administration or among bacteria with fastidious culture growth [6,55,59]. Yet, overall use is limited by inadequate negative predictive value (e.g., inability to rule out bacterial meningitis), and so latex agglutination is only sporadically used.

Limulus lysate assay to detect endotoxin

The limulus lysate assay utilizes lysate prepared from the amebocyte of the horseshoe crab, Limulus polyphemus, to identify endotoxins [52,60,61]. The test was originally used for detection of lipopolysaccharide endotoxin in blood, urine and pharmaceutical products, but has been adapted to CSF [60,61]. Results are typically available in 20–30 min and a positive test suggests Gram-negative bacterial meningitis [52,60]. As this assay detects only endotoxin, its inability to detect Gram-positive bacteria is a major limitation [52,60]. In suspected bacterial meningitis without an etiology determined by culture, this assay can help identify if etiology is from a Gram-negative organism.

Nucleic acid amplification tests

Nucleic acid amplification tests (NAATs) such as polymerase chain reaction (PCR) testing can provide rapid and accurate diagnoses for bacterial meningitis [2,62]. Depending on the method used, results can be available in as little as 15 min, and no longer than 3 h for a single sample [63–68]. Of course, if the PCR test is not performed on site, shipping time increases the time to results significantly.

Further, commercial NAATs have been developed to detect multiple bacterial pathogens (some include viral and fungal pathogens as well) and while sensitivities and specificities are often >90% for each pathogen, test performance varies by pathogen and assay [4,52,62,63,65–70]. Table 4 summarizes three of the commercially available PCR assays for bacterial meningitis that detect multiple pathogens and their gene targets. For example, the BioFire Film Array meningitis/encephalitis panel (FilmArray ME) is a PCR panel that rapidly (1 h) detects 14 pathogens including six bacteria, seven bacteria and one fungus directly from CSF [19,71]. Leber et al, found that the FilmArray ME panel had an overall percentage of agreement with comparative testing of 99.8% and a specificity of 99.2% [19]. However, this panel does not perform equally for all pathogens and is particularly weak for Cryptococcus and herpes simplex 1 and 2 compared with standard diagnostics for these conditions [19,71]. Further, the FilmArray ME panel and all NAATs only detect the pathogens they were designed to target and so will miss pathogens that they are not designed to target.

Further limitations of NAATs include cost, infrastructure requirements, easy contamination of samples, and the need for additional laboratory personnel training [63,66–68,70]. However, NAATs may be particularly useful for the diagnosis of bacterial meningitis when patients have received antibiotics prior to LP. Numerous NAATs technologies are currently being utilized in high-income countries [2]. Future developments in point of care testing (e.g., cartridge based) may facilitate more wide use of NAATs in smaller hospitals and low-resource settings with less training [70].

Loop-mediated isothermal amplification (LAMP) is a NAAT that can be used to detect bacterial DNA [72,73]. LAMP uses DNA primers and polymerases that bind to target DNA which is subsequently amplified [74]. Importantly, LAMP is meant to be isothermal and so require much less laboratory infrastructure. Further, its positivity can be interpreted by the naked eye. Compared with PCR, LAMP is less expensive and requires less extensive laboratory resources [73]. LAMP assays have been shown to have a sensitivity of up to 90% and specificity of up to 100% with capability of detecting pathogens in 20–60 min and at >100-fold lower detection limits than traditional PCR [72,73,75]. However, until recently multiplexing via LAMP has not been possible, although a recent LAMP assay was developed to detect three bacterial pathogens via LAMP integration into a chip [73,75]. This method has potential applicability as a point of care diagnostic test for bacterial meningitis in low-resource settings.

Recently, new molecular techniques have been developed that can detect a pathogen’s nucleic acid by harnessing clustered regularly interspaced short palindromic repeats (CRISPER) and CRISPR associated sequences (Cas) related technology [76–79]. Gootenberg et al. utilized the Cas13a enzyme to develop a platform called Specific High-Sensitivity Enzymatic Reporter Unlocking (SHERLOCK) that can rapidly detect selected Gram-negative pathogens and distinguish them from other bacteria [77,78,80,81]. Whether these technologies have clinical utility is uncertain – to our knowledge they have not been tested on CSF.

Next-generation sequencing (NGS) has the capability to detect a wide variety of pathogens. Two recent studies used NGS to detect bacterial (S. pneumoniae, H. influenzae, Enterococcus faecium and S. aureus) and viral pathogens in CSF [82,83]. These reports, combined with literature regarding CSF NGS in subacute meningitis lend themselves to optimism regarding a potential role for NGS in pathogen detection in difficult to diagnose meningitis, yet, this field is young and role of NGS in meningitis is unclear [84,85].

Matrix-assisted laser desorption ionization time of flight mass spectrometry

Matrix-assisted laser desorption ionization time of flight mass spectrometry (MALDI-TOF MS) is routinely being used in many centers in high-income countries to quickly identify bacteria from blood cultures and recently has been adapted for use on CSF [2,86,87]. MALDI-TOF MS identifies organisms by comparing the mass spectrum of the test isolate to the database of reference spectra [86,88]. Hartmeyer et al. described using MALDI-TOF MS to identify S. pneumoniae in CSF within 1.5 h from receiving the sample [86,89]. The CSF culture subsequently grew S. pneumoniae, confirming the diagnosis [86]. Similarly, Segawa et al. described rapid identification of Klebsiella pneumoniae from CSF by MALDI-TOF MS [89]. The cost of identifying bacteria via MALD-TOF MS is approximately 17–32% of traditional identification methods although the up-front cost of purchasing the equipment is significant and performance varies by pathogen [87]. MALDI-TOF MS is limited by the amount of organism present and can be improved by sample centrifugation [87]. MALDI-TOF MS use for meningitis is experimental at this point but may hold some promise for future use.

Tuberculosis meningitis

Multiple clinical diagnostic criteria have been developed to identify patients likely to have TBM, however none of these have proven broadly generalizable, or adequate as the sole tool for diagnosis of TBM [33,90].

Culture & AFB smear

CSF acid-fast bacillus (AFB) smear microscopy and culture are the typical methods of diagnosis of TB meningitis (TBM) but neither is adequate [91–96]. AFB smear is a microscopic technique that uses Ziehl–Neelsen stain to visualize organisms under direct microscopy [95]. AFB smear is rapid, affordable and widely available but has a low sensitivity (10–20%) [27,30,91,93,97,98]. AFB smear can be improved by examination of at least 6 ml of CSF, examining the sample for at least 30 min or modification of the Ziehl–Neelsen stain using cytospin slides with Triton processing [30,95,97,99]. Lowenstein–Jensen (LJ) and mycobacterial growth inhibitor tube (MGIT) are positive in 45–60% of TBM cases but results are too slow for clinical decision making (6 weeks for LJ, 2–4 weeks for MGIT) [27,91,93,96,98]. Despite sensitivity limitations and biosafety requirements, culture is essential for drug susceptibility testing [98]. In low-resource settings, biosafety requirements are generally only available at national reference laboratories, causing diagnostic delay [98].

Host-derived tests

CSF adenosine deaminase has shown sensitivities ranging from 50 to 100% and specificity from 63 to 99%, further, adenosine deaminase has difficulty distinguishing TBM from bacterial meningitis [47,97,100–102]. CSF adenosine deaminase cannot be used alone to rule out TBM [97,102,103]. Variable performance, cost and infrastructure requirements have limited use worldwide [103]. TB IFN-γ release assays measure host T lymphocyte release of IFN-γ when stimulated by M. tuberculosis specific antigens to diagnose latent and active tuberculosis [91,92]. CSF IFN-γ release assays show sensitivity of 75–92% and specificity of 90–100% [92,94,99]. IFN-γ release assays are limited by frequent indeterminate results in those co-infected with HIV, high CSF volume requirement (>2 ml), laboratory infrastructure and lab personnel training [91,92,94,97,99,103].

Delta-like 1 ligand, fetuin and vitamin D binding protein are other immune biomarkers that have been considered for TBM, yet have shown limited utility in an African HIV-infected population [104]. Heat shock proteins found at higher levels in CSF in patients with TBM compared with non-TBM patients and so have been theorized to be a biomarker for TBM – their role as a potential diagnostic test is not clear [100]. Though no immune markers have proven to be adequate diagnostic tests for TBM thus far, given the inadequate of performance of any one test to detect TB bacilli to date in CSF, they may play a role as adjunctive tests in the future [103].

Nucleic acid amplification tests

GeneXpert MTB/Rif (Xpert) is an automated, rapid, cartridge-based PCR assay designed to detect M. tuberculosis DNA (and rifampin resistance) in clinical samples [26,27,30,47,91,94,105,106]. Xpert has shown variable performance on CSF in different study populations, but in general has sensitivities similar to culture (50–60%) with a much more rapid turnaround (∼2 h) [26,27,30,47,94,105,106]. Specificity has generally been excellent, those studies with possible false-positive results may have been required to label them as such due to the inadequacy of TBM reference standards – the likelihood of false detection of M tuberculosis DNA in the CSF in the setting of chronic meningitis in those without an alternative diagnosis is quite low. The sensitivity of Xpert is higher in HIV-infected individuals, and when higher volumes (at least 5 ml) of CSF are centrifuged [94,105,107]. This cartridge-based assay is easy to use, carries low risk for sample cross-contamination, and does not require a specific biological safety environment as does TB culture [106,108]. However, sensitivity with Xpert is still inadequate, rifampin resistance performance unreliable and despite lower prices in low-resource settings, the prices of the modules and the cartridges are still relatively high. Inadequate negative predictive value makes Xpert unsuitable as a single ‘rule-out’ test for TBM [105,106,109]. Still, this technology has widely adopted in TB endemic countries [94,97].

Subsequently GeneXpert MTB/Rif Ultra (Xpert Ultra) was developed as a new assay using the Xpert platform to improve sensitivity and performance of rifampin resistance detection and deliver semiquantitative results in 90 min [10,103,110,111]. Xpert showed sensitivity of 95% for microbiologically proven, definitive TBM versus 45% for Xpert or culture and 70% for probable or definite TBM versus 43% for Xpert or culture [10]. Xpert Ultra has 93–99% negative predictive value versus (90%) for Xpert in TBM depending on the standard used [10,111]. Since March 2017, the WHO has recommended the use of Xpert Ultra as a replacement for Xpert, however addition studies are needed to confirm these findings [110].

Amplicor MTB (Cobas, Amplicor) and Amplified M. tuberculosis Direct Test (MTD) are among the commercially available PCR platforms [112–116]. Amplicor PCR has a sensitivity of ∼40% and specificity of ∼90–100% for TBM, similar to Xpert [112,114]. Yet, false-positive CSF results with Amplicor have been documents in persons with herpes encephalitis, staphylococcal spondylitis and cerebral symptoms secondary to drug interactions [113]. Further, these platforms are generally only located in reference laboratories, require trained laboratory personal, are not fully contained, and are expensive limiting their broad utility [105,115]. The Gen-Probe amplified M. tuberculosis Direct test (AMTD) is a transcription-medicated amplification procedure that detects M. tuberculosis complex rNA directly from specimens [117]. While data on extrapulmonary specimens in general in promising, studies on its use on CSF specimens are limited [117,118].

LAMP has also been used to diagnose of TBM [94,98,119]. LAMP amplifies DNA in single tube at a constant temperature in only 60 min, potentially making it more suitable for resource-limited settings [94,98,119]. LAMP based assays have been compared with conventional PCR and with sensitivities of 88–96% compared with 53% with conventional in-house PCR [94,99,119]. Despite its advantages, further study is required before widespread adoption of LAMP. Lastly, as recently noted by Michael Wilson and colleagues, NGS may play a role in detection of M. tuberculosis in CSF [85].

Other TB meningitis testing

Detection of TB lipoarabinomannan (LAM) in the CSF has been explored as a potential diagnostic technique for TBM [91,94,98,100,120]. Patel et al. tested CSF for LAM antigen and found a sensitivity of 64% and specificity of 69% for TBM [120]. Cox, et al. found 75% sensitivity and 80% specificity for definite TBM using large volume postmortem CSF samples, tested with the LAM lateral flow assay (LFA, Alere, MA, USA) [121] Others found zero positive results among 67 samples (12 with definite TBM) tested with the same LAM LFA on CSF obtained from LP [122]. Immunocytochemical staining is used to identify antigens within the cytoplasm of macrophages and one study found a sensitivity of 73% and specificity of up to 100% for TBM in CSF [123]. Although immunocytochemical staining has been developed into an assay for wider use, it still requires expertise and so its role TBM diagnosis is uncertain [123].

Diagnostic techniques for multiple types fungal meningitis

1,3-Beta-D-glucan

1,3-Beta-D-glucan (BDG) measurement in CSF has been used to detect a variety of fungi including all fungi discussed below, yet specificity for any individual fungi is inadequate [14,124,125]. Although it was previously thought that Cryptococcus did not release sufficient BDG for detection, this was disproven by Rhein and colleagues who found that among HIV-infected patients with cryptococcal meningitis, BDG showed a sensitivity of 89% and specificity of 85% compared with cryptococcal antigen detection with higher BDG levels being associated with mortality [11,125,126]. There remains a FDA-required black-box label incorrectly stating that Cryptococcus does not produce BDG. CSF BDG has also been studied for Histoplasma meningitis with a sensitivity of 53% and specificity of 87% [124]. Thus, while BDG may have utility in steering clinicians towards fungal meningitis – BDG is not able to reliably diagnose any particular fungal pathogen. A positive BDG result should prompt further focused fungal diagnostics and strong consideration of antifungal therapy.

Next-generation sequencing

NGS is an emerging diagnostic testing platform for subacute and chronic meningitis, including fungal meningitis. Initial results are promising, however further study is required to fully elucidate the role of NGS in the diagnosis of fungal meningitis [84,85,127].

Histoplasma meningitis

CSF culture is the standard for diagnosis of Histoplasma meningitis but is only positive in 19–65% of cases and growth can take up to 4 weeks [15,27,124,128–131]. Culture can be improved by using at least three samples of at least 10 ml of CSF – impossible in most cases [15,128,131,132].

Antibodies/antigen

Immunodiffusion, complement fixation and enzyme immunoassay (EIA) can be used to detect Histoplasma antibodies in CSF [15,28,124,128–130,133,134]. Moreover, a new IgM and IgG EIA showed sensitivity of 82% (vs 51% for immunodiffusion or complement fixation) and specificity of 93% to diagnose Histoplasma meningitis [128]. Yet, antibodies are typically not able to be detected early in infection, and may persist for years after infection [15,135,136]. False-positive CSF results can occur in patients with high levels of serum antibodies due to passive diffusion when blood–brain barrier is impaired due to inflammation or due to cross-reaction with other fungi [129,137]. Antibody response may be impaired in immunosuppressed individuals [132].

Histoplasma antigen can be detected in CSF via EIA [15,124,128–130,134]. Multiple generations of EIA have been designed with the most recent allowing quantification [136]. Antigen sensitivity is as high as 85% when ethylene diamine tetraacetic acid (EDTA) pre-treatment is used [128]. As with antibody testing, false-positive results can occur due to cross-reactivity with other fungi, passive diffusion across the blood–brain barrier or traumatic LP [27,128,129,132,136]. Testing for the organism in non-CNS specimens can also be useful when Histoplasma meningitis is secondary to disseminated histoplasmosis [15,27,132,133]. Finally, combined, CSF antigen and antibody testing have shown synergistic performance in one study [128].

Coccidioides meningitis

Wet mount and cytospin preparation can be used to identify Coccidioides spherules in CSF to confirm diagnosis; however, the spherules are seen in less than 10% of cases [138–140]. CSF culture is only positive in 15–30% of the cases of coccidioidal meningitis although using large volumes of CSF can improve yield [27,138,139,141,142]. Eosinophilic predominance in CSF can be an indicator of coccidioidal meningitis, but is not specific [139,142,143]. Meningeal biopsy can be used but is invasive and so is rarely used.

CSF anti-coccidioidal antibodies may be detected via immunodiffusion, complement fixation or EIA [27,138–140,142,144]. Complement fixation has a sensitivity of 59–94% while EIA has sensitivity as high as 85%, both have excellent specificities [27,138]. Latex agglutination is less sensitive with more false positives [27,145]. Antibodies less helpful due to limited production early in infection or in immunocompromised persons, and prior (or current) coccidioidomycosis exposure may lead to antibody leakage from the serum into the CSF leading to false-positive CSF testing [14,27,138,139,141,142,144,145]. Cross-reaction with may occur with endemic fungi or Cryptococcus [137].

Antigen detection via EIA exhibits 89–93% sensitivity and 100% specificity and as opposed to antibody detection, may have a role in differentiating immune reconstitution inflammatory syndrome (IRIS) from relapse [138,141]. Limitations include cross-reactivity with other endemic fungi, such as histoplasma [138,141]. Coccidioides antigen detection is currently available only via reference laboratory, limiting its speed [138]. PCR assays for Coccidioides have been used but are not routinely available [27]. Testing by more than one modality and specimen type (e.g., urine, serum, CSF, etc) may increase diagnostic yield [138,145]. For example, one study showed combined antigen and antibody testing with 98% sensitivity and specificity [138].

Cryptococcal meningitis

The most accurate method to diagnose cryptococcal meningitis is cryptococcal antigen detection although culture is important in some situations [11]. India ink staining of the CSF for direct identification of Cryptococcus is a readily available technique however it can be lead to missed diagnosis, sensitivity is as low as 42% when fungal burden is <1000 colony-forming units/ml, and the best case is 85% sensitivity [11,27,146–148]. Fungal culture is considered the standard for diagnosis of cryptococcal meningitis, growth can take up to 10 days and false-negative results can occur with a low fungal burden although using higher CSF volume (100 vs 10 μl) can improve sensitivity [11,147]. Despite the limitations of quantitative fungal culture, quantitative culture remains highly clinically useful for monitoring response to treatment and critical for differentiating relapse of cryptococcal meningitis from IRIS, where cryptococcal antigen detection falls short [26]. The FilmArray ME PCR panel has the capability of detecting Cryptococcus in CSF but is inadequate compared with antigen testing and it’s exact role in the diagnosis of cryptococcal meningitis is not well defined [11,149,150]. FilmArray ME has distinct diagnostic utility for distinguishing probable culture-positive relapse from culture-negative paradoxical IRIS. NGS can detect Cryptococcus in CSF samples, but its role is not yet clear [127].

Cryptococcus capsular polysaccharide antigen

Detection of Cryptococcus capsular polysaccharide antigen (CrAg) in CSF, serum, plasma or whole blood is key to rapid diagnosis of cryptococcal meningitis [11]. This test was previously accomplished via latex agglutination or EIA; however in 2011 a CrAg lateral flow assay (LFA) was developed [11,146,148,151–153]. CrAg LFA is a point of care test with results available in 10 min, stability at room temperature (crucial for use in many locations where the infrastructure for cold-chain storage of tests is not practical), and relatively low cost at US$2 per test [11,26,148,151,152]. CrAg LFA in CSF has sensitivity and specificity as high as 99% with whole blood, serum and plasma, being nearly as accurate for diagnosis of meningitis [11,27,40,146–148,152,154,155]. This combination of accuracy, rapidity, cost and heat stability has led to widespread adoption of the CrAg LFA. Yet, lower sensitivity (91%) has been noted with high fungal burdens due to the pro-zone or hook effect (high cryptococcal load interfering with antigen–antibody complex of the assay resulting in a false negative) – this issue resolves with dilution (sensitivity 100%) [11,146,156]. In addition, false-positive results have been reported in lower incidence settings at low titers [157]. Importantly, although a quantitative titer can provide prognostic information, CrAg does not decay at a reliable rate from the CSF (can remain for years) and so cannot be used to monitor treatment response or differentiate IRIS from fungal relapse [11,27,146,152].

A recent study identified a group of HIV infected patients with symptomatic meningitis, positive serum CrAg, but negative CSF testing [158]. Thus, blood CrAg should be obtained in any immunocompromised patient in which Cryptococcal meningitis is suspected as very low antigen in CSF may be too low to detect while higher levels are present in the blood while true cryptococcal meningitis exists – this is pathophysiologically plausible given that Cryptococcus first enters the lungs, then disseminates and eventually seeds the CNS [158]. Because serum CrAg can be detected in blood 3 weeks prior to the onset of symptoms of meningitis (with typically a median of another 2 weeks until meningitis diagnosis), targeted serum CrAg screening, can allow detection and treatment of early, subclinical cryptococcosis and prevention of meningitis [11,151,158,159]. Thus, the World Health Organization (WHO) and many national HIV guidelines strongly recommend CrAg screening in HIV patients with a CD4 <100 cells/μl and to consider CrAg screening among CD4 counts 100–200 cells/μl [146,151,158,159].

Blastomyces meningitis

Diagnosis of Blastomyces meningitis is difficult and often delayed [160,161]. CSF culture is the standard diagnostic test and takes at least 5 days, but may take up to 4 weeks and is only 64% sensitive [27,49,160,162–165]. CSF culture is more sensitive using fourth ventricle CSF, but this is not practical in most cases [160,162–164]. EIA detection of Blastomyces antigen in CSF has been reported but diagnostic performance is unclear [160,161]. Given that Blastomyces antigen testing in other body fluids frequently cross reacts with other fungi (particularly Histoplasma), cross-reaction in the CSF would be a concern as well [27,160,161]. Serological assays are commercially available for Blastomyces and can be used on CSF, however, they are of limited utility for diagnosis of meningitis due to low sensitivity and specificity [27,48,163]. Tissue biopsy or diagnosis of blastomycosis outside of the CNS are other diagnostic options [49,160,162,163,166].

Aspergillus meningitis

Aspergillus meningitis diagnosis is also difficult and in many cases is not obtained until postmortem examination [13,45,167]. In a study by Antinori, et al., only 56% of patients were diagnosed prior to death [13]. CSF fungal culture is only 31% sensitive among all hosts, and 18% in immunocompromised hosts [13,43,44,168,169]. Multiple, large volume CSF cultures may improve sensitivity [13,167]. Galactomannn antigen is a cell wall polysaccharide released by Aspergillus that can be detected in body fluids including CSF via ELISA, radioimmunoassay, latex agglutination or western blot [44]. CSF galactomannan sensitivity is 70–90% and specificity 70–100% [13,27,44,169]. As opposed to cryptococcal antigen, CSF Aspergillus galactomannan antigen declines with treatment and so may be used to monitor treatment response [167,169]. Cross-reactivity to piperacillin–tazobactam has been reported [44]. Aspergillus antibody detection in CSF is possible but performance is unreliable and although PCR has detected Aspergillus in CSF, diagnostic performance is unclear [45,167]. Often the diagnosis of Aspergillus meningitis occurs only after aspergillosis is diagnosed elsewhere [27].

Viral meningitis

Viral meningitis has significantly less mortality compared with other types of meningitis, and treatment is limited to supportive measures in most cases [23]. NAATs are often used for diagnosis of specific pathogens. Some argue these tests are underutilized, as positive results may allow rapid discontinuation of empiric antibiotics and decrease length of hospitalization [6,23,52]. Others test only for those viral etiologies with the potential for treatment (e.g., herpes viruses) or those with the potential for more poor outcomes [23].

Barriers to adoption

The largest barrier to adoption of many of the technologies discussed remains cost, this is particularly true for newer molecular techniques such as those utilizing CRISPR or NGS. Excellent performance, minimal lab equipment and personnel training requirements, heat stability and ease of use are all crucial for adoption of new technologies as well. The CrAg LFA remains the standard to which other diagnostic tests for meningitis must strive. Performance and cost are optimal. The CrAg LFA test is easy to use and can be performed at the bedside. CrAg LFA does not require extensive training, a steady electrical supply or cold-chain distribution. For CSF diagnostic tests to be truly accessible and useful worldwide, all of these attributes must be met. Thus, noncryptococcal meningitis diagnostic tests continue to require improvements. While syndromic PCR panels are attractive in terms of requiring smaller volumes of CSF to detect multiple pathogens (e.g., Biofire FilmArray ME), their limitations must be considered. First, these are only able to detect pathogens that they are designed to detect, which is generally appropriate based on high-income country epidemiology of meningitis, yet multiplex panels may miss unexpected pathogens, thus a thorough history and expert physician diagnostic acumen is still required. This seems obvious but the lack of TB in the panel means this panel is less useful in most low and middle resource countries. Further, to truly use the panel correctly, one must understand its limitations and supplement the panel as needed (e.g., the FilmArray ME panel does not eliminate the need for CrAg testing).

Conclusion

Despite the existence of numerous diagnostic techniques, determining the etiology of infectious meningitis remains difficult and cumbersome in many cases. Delay in diagnosis is a significant contributor to mortality in bacterial, tuberculosis and fungal meningitis. When evaluating a patient for meningitis, physicians must incorporate many factors to determine the most appropriate tests to order. CSF volume from an LP should be conserved for use on tests of the most likely etiology, rather than ordering numerous tests from the start (a common practice in the era of electronic medical records and auto-populated order sets). Physicians must consider local epidemiology, duration of symptoms, current and recent medications, current immune status, country of origin, current living situation, social history, vaccination status and history of travel among other factors. Currently, excellent diagnostic techniques exist for cryptococcal meningitis along with many bacteria and viruses that cause meningitis. Yet significant room for improvement exists with other bacteria, viruses, fungi and TB. GeneXpert MTB/Rif Ultra seems to be a major improvement in the diagnosis of TBM, yet cases are still missed. Thus, additional improvements in TBM are urgently needed. Cutting-edge technologies such as NGS hold significant promise but require extensive bioinformatics and sample processing expertise – the role of NGS in meningitis diagnosis is not yet clear.

Future perspective

The field of diagnostic testing in infectious diseases is ever evolving. Impressive strides have been made over the last 5 years in NGS, and these technologies are starting to be used clinically. Newer molecular techniques that utilize the CRISPR-Cas9 system are in their infancy but may have potential with further investigation. It remains difficult to assess which new diagnostic techniques will become available in low income setting due to significant monetary and practical barriers for implementation. Yet, to truly improve diagnosis of meningitis worldwide, diagnostic test development must focus on these practical issues that affect test availability in many settings worldwide.