Indian Pediatrics 1998; 35:841-850 

Diagnosis of Tuberculous Meningitis by Detection of Antigen and Antibodies in CSF and Sera

TUBERCULOSIS is a major public health problem in all developing countries including India. World wide, it kills more people than any other single infectious disease(1). Out of all forms, tuberculous meningitis (TBM) is the most devastating with mortality ranging from 13% to 71.9%(1,2). Early diagnosis and treatment is essential to reduce mortality and long term morbidity associated with TBM. The 'Gold Standard' for the diagnosis of any infection is bacterial isolation, and demonstration of tubercle bacilli would be the best way to confirm the tubercular etiology in meningitis, but unfortunately, the yield of tubercle bacilli on AFB staining is very low. Further, a very large number of bacilli, at least 10,000/ml is required for the bacilli to be demonstrated in smears(3) which may not be the case with every specimen. Culture takes a long period of 6 to 8 weeks and one cannot wait for culture report to start the therapy. There is therefore an urgent need for finding alternative method of diagnosis of tuberculosis which are reliable, time saving, cost effective and easy to perform.

Attempts have been made to improve the sensitivity and speed of detection of tubercle bacilli or their components by techniques such as radiometric determination of bacterial growth, gas chromatography, DNA hybridization and Polymerase chain reaction, but the involved high cost and the required high level of technical expertise are important barriers in their use(4). Enzyme linked immunosorbent assay (ELISA) has the advantage of being a cheaper and time saving test requiring relatively less technical expertise in performing the test. The present study was undertaken to determine the diagnostic potential of detection of H₃₇R_V antigen of M. tuberculosis in CSF by dot ELISA, and detection of antibodies (IgG) against H₃₇R_V antigen in CSF and serum by plate ELISA in diagnosis of tuberculous meningitis in children.

Subjects and Methods

Subject Selection

In this case control study, conducted in Children's Ward of Gandhi Memorial and Associated Hospitals, Lucknow, India, between May 1994 to May 1995, eligibility criteria for initial enrollment were defined as: children of either sex, below 12 years of age, presenting with features of CNS infection (fever, and altered sensorium, and convulsions, with or without headache, vomiting and neurological deficit), who had not been treated with antitubercular drugs in last 14 days, or antibiotics in last 7 days prior to admission. Detailed history and findings of clinical examination at the time of admission were noted. Written consent was obtained from the parents or the nearest relatives of each patient. Lumbar puncture was done and CSF obtained was subjected to biochemical and cytological examination, Gram's and Ziehl Neelson's staining, culture for mycobacterium in LJ media for 6-8 weeks, and bacterial and fungal culture. Part of CSF and serum from the blood drawn at the time of admission was stored in autoclaved eppendroff vials at -20^o C for further use. Cranial computed tomography was done on each patient.

Patients were carefully followed-up for 6 months and the details of therapy given and response to therapy were noted. Patients were labeled to be suffering from tuberculous meningitis if the following criteria were met: (a) CSF findings; pleocytosis of more than 20 cells per cumm with more than 60% being lymphocytes, and CSF proteins more than 100 mg/ dl and CSF sugar less than 60% of concomitant blood sugar levels (all 3 mandatory) with or without positive smear/culture for tubercle bacilli, and (b) clinical course and response to exclusive antitubercular treatment (gold standard for diagnosis adopted in the present study) and (c) CT scan findings (2 out of 4 following features)- exudates in basal cistern or sylvian fissure, hydrocephalus, infarct, Gyral enhancement. All three (a, b and c) were required to make a diagnosis of tuberculous meningitis. Those who received both antitubercular and antipyogenic treatment and also those who were given antimalarials during the first seven days of therapy were excluded from the study group. Patients who expired during their hospital stay, those who absconded and those who left against medical advice were also excluded from the study group. This was done purposely because response to therapy was the 'Gold standard' adopted in the study and lack of response could have resulted in error in classifying cases and controls. Fifty patients who qualified for these. criteria were de- fined as "cases". Staging of the disease was done using British Medical. Council system(5). Out of fifty patients, 2 (4%) patients were in stage I, 17 (34%) in stage II, and 31 (62%) patients were in stage III. Forty eight patients of nontubercular CNS disorders. 31 patients of pyogenic meningitis (diagnosed by CSF findings and positive cultures and response to exclusive anti-pyogenic meningitis treatment); 10 patients of encephalitis; 2 patients of CT scan proved brain tumor (cerebellar astrocytoma and medulloblastoma) and 5 patients of generalized tonic clonic seizures in whom no etiology could be found despite extensive diagnostic workup were included as "controls".

A total of 98 serum and CSF samples from 50 cases with clinical diagnosis of tuberculous meningitis and 48 controls with CNS disorders of nontubercular etiology were subjected to immunodiagnostic tests under evaluation, namely, detection of H₃₇R_V M. tuberculosis antigen in CSF by Dot ELISA, and detection of antibodies (IgG) against H₃₇R_V strain of M. tuberculosis in CSF and serum by plate ELISA.

Laboratory Methods

Antigen Extraction

H₃₇R_V strain of Mycobacterium tuberculosis was cultured in Sauton's medium at 37^oC for four weeks. The growth was subjected to heat treatment for 15 minutes in boiling water bath to kill the cells followed by centrifugation at 8000 rpm for 10 minutes. The cells were suspended in Tris 0.01 M(pH 7.0) and O.1M Phenyl Methyl Sulphonyl Fluoride (PMSF) and then were sonicated for a total of 45 minutes in cold at 20 Kilocycles (MSC Soni Prep.) Sonication was followed by centrifugation at 12000 rpm for 90 minutes and the supernatant with antigen was collected and stored at -20°C. The protein concentration as estimated by Lowry's method was 1.8 mg/ml(5).

Antibody Detection by Plate ELISA

H₃₇R_V antigen in carbonate buffer (0.06 M, pH 9.6) (200 µl) was coated in microtitre plates (Tarsons) and incubated overnight at 4^oC. Following morning, plates were washed with phosphate buffer saline (pH 7.3) containihg 0.05% Tween-20 (PBS-T) and non-specific sites were blocked with 30/0 Bovine Serum Albumin (BSA, Sigma Chemical Co, USA). After incubation for 90 minutes at 3TC, 200 µI of test sample diluted in PBS (serum 1:20 dilution, CSF 1:2 dilution) were added to each well. Each sample was tested in duplicate. After 90 minutes of incubation at 37°C plates were washed with (PBS-T) and 200 µI of 1:1000 diluted anti-human IgG-horse. radish peroxidase conjugate (Sigma chemical Co., USA) . was added, followed by. incubation for 90 minutes at 3TC, and, after another wash, substrate (200 Ill) of Ortho Phenylene Oiamine (OPO, 40 mg/50 ml) in phosphate citrate buffer (pH 6.0) containing hydrogen peroxide was added and reaction was allowed to proceed at room temperature in dark for 15-20 minutes. Reaction was terminated by addition of 5 N sulphuric acid to each well and readings were taken by ELISA reader at 492 nm optical density.

Antigen Detection by Dot ELISA

This was performed using nitrocellulose paper (Schtcher & Schnell, Germany) on which CSF samples in 1:2 dilution were vacuum blotted in duplicate (Bioblot, Bionad). After washing with Tris buffer saline containing 0.5% Tween.20 (TBS-T) and blocking-of non-specific sites by 3% Bovine serum albumin for 90 min at 37^oC, rabbit antibody to H₃₇R_V at 1:800 dilution was added and again incubated at 37^oC for 90 minute. After washing with TBS-T, nitrocellulose paper was incubated again with 1:1000 diluted horse radish peroxidase conjugated antirabbit IgG (Sigma Chemicals Co, USA) followed by washing and addition of substrate, 3-3 diaminobenzidine tetrahydrochloride (Sigma Chemicals Co, USA) with 0.05% hydrogen peroxide. After 10-15 minutes of incubation at room temperature, the strips were washed thrice with distilled water, and were allowed to dry. Results were recorded visually as positive or negative.

Statistical Analysis

Results of dot ELISA for detection of H37Rv antigen in CSF were recorded as dichotomous variable, namely, positive or negative. Results of detection of antibodies against H37Rv antigen in CSF and serum by plate ELISA were recorded as optical densities ,(continuous variable) compared between various groups using ANOVA, and Mann Whitney test, as and where applicable. Results of ELISA test were taken as positive when the value of mean optical density obtained was more than 2 standard deviations above mean for that test in controls. Sensitivity, specificity, positive/negative predictive values were computed as indices of applicability of diagnostic test under evaluation.

Results

Maximum children (38%) were in the age group of 1-3 years. The youngest and the oldest patients in which TBM was diagnosed were 8 months and 9 years old, respectively. Male to female ratio was 1.77:1. History of tubercular contact was present in 38% (n=19) children. Fever was the commonest presenting feature in 94% (n=47), followed by convulsions in 64% (n=37). Acid fast bacilli could be demonstrated by direct smear examination and culture in only 6% (n=3) CSF samples of patients of tuberculous meningitis. Twenty five patients (50%) with tuberculous meningitis and 64.58% (n=31) patients of other diseases were vaccinated with BCG. On tuberculin testing induration of more than 10 mm was noted in 52% (n=26) patients with tuberculous meningitis while none of the children in control group were tuberculin positive.

Results of dot ELISA for detection of H₃₇R_V antigen in CSF were positive in 90% (n=45) cases of tuberculous meningitis, and 4.17% (n = 2) patients of CNS disorders of nontubercular etiology [3.22% (n=1) patients of pyogenic meningitis (due to N. meningitidis) and 50% (n=1) patient with brain tumor (cerebellar astrocytoma)] (Table I). Sensitivity of the detection of H37Rv antigen in CSF was 90% with specificity of 95.83% (Table II).

The mean ±SD (range) of optical densities obtained on plate ELISA for detection of antibodies against H₃₇R_V antigen in CSF and serum, and cutoff values for considering results of these two test to be positive, are shown in Table I (Figs. 1 & 2). Optical densities for both the tests were significantly higher in patients of clinically diagnosed tuberculous meningitis as compared to patients of CNS disorders of non tubercular etiology (p <0.0001). Further, there was no statistically significant difference in optical densities among various sub-groups of controls (p > 0.05). Results of detection of antitubercular antibodies against H₃₇R_V antigen in CSF were positive in 74% (n=37) cases of IBM, and 10.42% (n=5) controls [in 6.45% (n=2) patients of pyogenic meningitis, 10% (n=1) patient of encephalitis and 100% (both) patients of brain tumor) (Table I and Fig. 1). Sensitivity of the test was 74% and specificity was 89.58% (Table II).

Results of detection of antibodies against H₃₇R_V antigen in serum was positive in 50% of cases and 8.33% of controls [6.45% (n = 2) patients of pyogenic meningitis and 20% (n = 2) patients of encephalitis] thus making it 50% sensitive and 91.67% specific for diagnosis of TBM (Tables I & II, Fig. 2).

 

Fig 1. Detection of Antibodies in CSF in Cases and Controls. Each dot in horizontal plane represents the absorbance value recorded for individual patient. I-I Each line shows the mean optical density in respective groups. - - - The broken line represents the cutoff value (0.531) [250 above mean value in control group (all groups taken together) above which the test was considered to be positive].

Fig. 2. Detection of Antibodies in Serum in Cases and Controls. Each dot in horizontal plane represents the absorbance value recorded for individual patient. I-I Each line shows the mean optical density in respective groups. - - - The broken line" represents the cutoff value (0.466) [2SD above mean value in control group (all groups taken together) above which the test was considered to be positive].

Detection of antigen in CSF was found to have maximum sensitivity and specificity (90% and 95.83%, respectively), positive as well negative predictive values (95.74% and 90.19%, respectively), followed by detection of antibody in CSF with positive predictive value of 88.10% and negative predictive value of 76.78%. Detection of antitubercular antibodies in serum was found to have the least positive as well negative predictive value (86.21% and 63.76, respectively) (Table II).

Discussion

We adopted the response to antitubercular therapy as the alternative gold standard for diagnosis of tuberculous meningitis in the study. Bacteriological confirmation could be obtained in only 6% (n=3) clinically diagnosed cases of tuberculous meningitis who otherwise satisfied the criteria adopted in the study and responded to exclusive antitubercular therapy. It might appear from the results that diagnosis of tuberculous meningitis can be made with certainty using the criteria adopted in the study but it should be realized that this highly selected studied sample does not truly represent the spectrum seen in hospital practice where a clinician often comes across situations where making a diagnosis is a real difficult task. The aim of the study was to evaluate the diagnostic potential of indirect, simple and inexpensive methods of diagnosis which may be applied in such cases posing diagnostic difficulties and should be interpreted considering this point. Alternative criteria, namely, CSF and CT scan findings, response to therapy, follow-up often have to substitute for 'Gold Standard' especially in tuberculous meningitis and have been used by other workers in making diagnosis of tuberculous meningitis while evaluating a diagnostic test(7,8).

A large number of studies have been conducted with wide variation in the methodology employed by various workers in immunodiagnosis of tuberculous meningitis and a wide variety of antigens are available for immunoassay in tuberculosis(9-36). We used sonicated extract of H₃₇R_V antigen for immunodiagnosis of tubercular etiology in our study. H₃₇R_V strain of M. tuberculosis has been used by many workers for immunodiagnosis of tuberculosis in different settings using varied methodology and yielded good results(14,20 ,22,31-35).

In the immunological spectrum of human tuberculosis, patients range from localized tuberculosis with well developed to those having poorly developed immunity and assay for detection of antigen and antibodies will depend on the position of the patient in the spectrum(37). The relative value of an ELISA system will be deter- mined by the preponderance of antigen, antibody or immune complexes in the CSF depending on the stage of the disease(36).

Results for detection of H37Rvantigen in CSF by dot ELISA were negative in 10% (n = 5) cases who met the criteria for diagnosis of TBM adopted in the study. Levels of antigen below the detectable limits may be possible explanation for these cases or it .could be that the manifestation of TBM were predominantly due to an Ag-Ab immune complex mediated reaction(21). Two positive results were obtained in 1 case of pyogenic meningitis in whom, N. meninigitidis was isolated on CSF culture and case of cerebellar astrocytoma. Detection of H₃₇R_V antigen in CSF was found to be 90% sensitive and 95.83% specific for diagnosis of tuberculous meningitis.

The CSF antibody test was 74% sensitive and 89.61% specific for diagnosis of TBM. Varying results have been demonstrated by other workers. Bal et al. detected antibodies to the same antigen as ours (H37Ry) in all the 9 patients of TBM with one false positive result in pyogenic meningitis(20). Sipsova et al. (Unpublished Abstract, 1990) reported sensitivity and specificity of detection of antibodies in CSF as 76.9% and 45% respectively. Samuel et al. reported sensitivity of 66-74% and specificity of 100% when using radioimmunoassay technique with this antigen(31).

False positivity of detection of both antigen and antibodies in CSF of patients with pyogenic meningitis needs some explanation for its obvious importance. Cross reactions have been reported between polysaccharides of mycobacteria and other bacterial genera(38,39). A 65 Kba Ag of Mycobacteria known as 'heatshock protein' has been shown to be present in E. coli also, and appears to be present in a wide variety of other bacteria also(40). Among the 2 patients with brain tumor, both antigen and antibody were detected in CSF in the patients with cerebellar astrocytoma, whereas the other patient with medulloblastoma demonstrated presence of antigen only in CSF. We have no possible explanation for this. However, sharing of Ag between tumor cells and M. tuberculosis antigen has been observed by other workers also(41,42).

The serum antibody test was found to be 50% sensitive and 91.67% specific for diagnosis of TBM. Krambovitis et al. reported a sensitivity of 84.5% and specificity of 84% for diagnosis of extra pulmonary tuberculosis(32). The false negativity in serum has also been explained by them on the basis of formation of immune complexes(32). Further, antibodies to H₃₇R_V were detected in serum in 74% CSF samples from patients of tuberculous meningitis, whereas same antibody was detectable in serum in only 50% of these patients. This phenomenon of lower sensitivity of detection of serum anti- body as compared to detection of anti-bodies in body fluids has also been observed and explained by other workers according to whom specific antitubercular antibodies may be present in CSF in active disease but may not be found in serum(43).

No test can be expected to match the diagnostic accuracy of demonstration of tubercle bacilli in diagnosis of tubercular etiology of the disease but considering its low sensitivity and the difficulties encountered in clinical practice, alternate or indirect methods when employed with the complete clinical profile may be useful for therapeutic decision making process. The usefulness of immunodiagnosis in different settings can be assessed by calculating predictive . value which vary with the prevalence of the disease under consideration in the geographical area where such studies are conducted. High prevalence rates lead to more true positive predictive accuracy. Similarly, lower prevalence rates lead to fewer false positive results and hence better negative predictive accuracy. In areas of high prevalence like India, a negative test may be helpful to exclude mycobacterial disease from further consideration. How- ever, false negative tests results would be unacceptable in such situations and a very high sensitivity would be required.

Detection of antigen in CSF with the highest sensitivity (90%), specificity (95.83%), positive predictive. value (95.74%), as well as negative predictive value (90.19%) appears to be the most promising test for the diagnosis of tuberculous meningitis, followed by detection of antibodies in CSF, with slightly lower sensitivity (74%), specificity (89.58%) and positive and negative predictive values of 88.10% and 76.78%, respectively. Detection of antibodies in serum with lowest sensitivity and lowest positive and negative predictive values does not appear to be much useful for this purpose, atleast in highly endemic areas like India. However, more studies involving larger number of patients need to be conducted to support and confirm these findings and before their routine use in initiating therapy is adopted or advocated.