The outcome of tuberculous meningitis is influenced by the stage of disease at the start of treatment. Initiation of antituberculous therapy is often delayed because of the inadequacy of presently available laboratory tests. Management of patients with possible tuberculous meningitis would thus be advanced by the development of an accurate, reliable, and rapid diagnostic test, particularly if it could be applied in settings with poor recourses.

Adenosine deaminase (ADA), an enzyme involved in purine catabolism, exists in at least three forms. ADA₁ is a monomeric protein with a molecular mass of approximately 35 kDa and two ADA₁molecules joined via a connecting protein form the dimeric ADA_1+CP. The third isoenzyme ADA₂ seems to be produced only by monocytes.1 Total CSF ADA has been suggested as a marker for tuberculous meningitis2 3; however, considerable variability and overlap, particularly with acute bacterial meningitis, has led some authors to question its clinical usefulness.4

An increased proportion of ADA₂ has been suggested to be a more specific means of diagnosing tuberculous effusions.2However, the use of CSF ADA₂ in the diagnosis of tuberculous meningitis has not been described.

Study subjects (all adults) were from a prospective cohort of patients undergoing a diagnostic lumbar puncture for suspected meningitis. Patient characteristics and investigations have previously been described in detail.5 In addition, CSF specimens from 10 patients at the Johannesburg Hospital were included (four tuberculous meningitis, four cryptococcal meningitis, two acute bacterial meningitis). Serum and CSF ADA analysis was performed on 11 specimens from patients with tuberculous meningitis (nine established by culture, two probable), nine with cryptococcal meningitis, 13 with acute bacterial meningitis (nine Neisseria meningitidis; two Streptococcus pneumoniae), nine with aseptic meningitis, and 19 with a normal lumbar puncture. All patients with tuberculous meningitis, cryptococcal meningitis, or aseptic meningitis, six of 13 with acute bacterial meningitis, and 10 of 19 of the normal lumbar puncture groups were HIV positive. Ethical approval was obtained from the committee for research on human subjects of the University of the Witwatersrand.

Tuberculous meningitis was confirmed if the CSF culture yieldedM tuberculosis. Probable disease was diagnosed in the presence of a lymphocytic pleocytosis (>20 cells/mm³), high CSF protein (>0.8 g/l), low CSF glucose (<60% of matched plasma glucose), and evidence of pulmonary tuberculosis. Cryptococcal meningitis) was diagnosed if either the indian ink stain, CSF fungal culture, or CSF cryptococcal antigen were positive. Acute bacterial meningitis was diagnosed in patients with acute onset of symptoms, pyrexia, a CSF neutrophilia with high protein, and low CSF glucose. Aseptic meningitis was diagnosed if there was a predominantly lymphocytic pleocytosis, a normal or moderately raised CSF protein (>0.7 g/l), and negative serology and bacterial, fungal, and mycobacterial culture.

Analysis of CSF was considered normal when there were < 5 leucocytes/mm³, protein <0.45 g/l, and negative culture and serology.

Matched CSF and serum samples were frozen within 6 hours of collection, stored at−20°C (the enzymes are stable for at least 4 weeks) and analysed within 1 week of collection. Tests were performed with laboratory staff unaware of the diagnosis. Total ADA activity was measured by an enzymatic spectrophotometric method on a Cobas Mira autoanalyser (Roche Diagnostics, Switzerland). Erythro-9-(2-hydroxyl-3-nonyl)-adenine, a selective ADA₁and ADA_1+CP inhibitor, was added to the reaction mixture at a concentration of 200 μM, allowing for the measurement of ADA₂ activity (in the same enzymatic system). The between batch coefficient of variation of the test at a dilution of 6 U/l was 8%. The proportion of total ADA which comprised ADA₂ could only be reliably estimated when total ADA was >2 U/l. ADA₂isoenzyme analysis is therefore only reported in the groups with a median total CSF ADA of >2 U/l.

Data were analysed using EpiInfo 6.04 (CDC, Atlanta) and PRISM 2.01 (GraphPad Software, USA). Continuous variables were compared using analysis of variance (ANOVA) and a 5% level of significance was used.

Comparison of total CSF and serum ADA and CSF ADA₂ in the diagnostic categories is shown in the table. Total CSF ADA was highest in patients with tuberculous meningitis. Using a cut off of ⩾6 U/l, the test was 90.9% sensitive in detecting tuberculous meningitis (10 of 11). The specificity was 94% (47 of 50) in all patients and 77.3% (17 of 22) compared with those with cryptococcal meningitis or acute bacterial meningitis. There were no significant differences between those with tuberculous meningitis established by culture and probable disease. Similarly, there were no significant differences in the CSF ADA concentrations in HIV positive and negative patients in the acute bacterial meningitis (mean 4.88 U/l vs 3.71 U/l; p=0.49) and normal lumbar puncture groups (mean 0.74 U/l v 0.12 U/l; p=0.14). Serum ADA concentrations were highest in patients with tuberculous meningitis or cryptococcal meningitis and significantly lower in those with acute bacterial meningitis.

There were significant differences in the mean proportion of total ADA that comprised ADA₂ in patients with tuberculous meningitis, cryptococcal meningitis, or acute bacterial meningitis (p<0.001). Using a cut off of 80% for the proportion of CSF ADA₂, the test was 100% sensitive and 86.4% specific in detecting tuberculous meningitis (positive predictive value 78.6%, negative predictive value 100%). An ADA₂ of >80% was found in three of nine of those with cryptococcal meningitis and none with acute bacterial meningitis. A cut off of 90% changed the sensitivity and specificity to 36.4% (four of 11) and 95.5% (21of 22) respectively.

The diagnosis of tuberculous meningitis remains a challenge. Acid fast bacilli are typically identified by microscopy in less than a quarter of patients and mycobacterial culture, the present “gold standard” for diagnosis is positive in 45%–90% of cases and may take up to 8 weeks to yield results. Recent advances in rapid automated liquid culture may reduce culture times. The use of the polymerase chain reaction (PCR) to detect Mycobacterium tuberculosis specific DNA may be of potential value; however, problems with its specificity have been encountered.

Sensitivities of 100% and specificities of 99%3 have been described for total CSF ADA. However, others have reported lower sensitivities and specificities and have suggested that CSF ADA measurement has “no advantages over conventional diagnostic criteria”.4 Possible reasons for the varying usefulness in different studies include different disease profiles, times to presentation, and ages, as children with tuberculous meningitis have lower ADA values.4

We were able to differentiate patients with tuberculous meningitis from those with aseptic meningitis or a normal lumbar puncture on the basis of the total CSF ADA. However, there was overlap between patients with tuberculous meningitis and those with cryptococcal meningitis or acute bacterial meningitis. A proportion of ADA₂ isoenzyme of >80% seems to be a reliable marker of tuberculous meningitis, yielding a sensitivity of 100% and specificity of 86.4%. The only other diagnostic category with patients who had >80% ADA₂was cryptococcal meningitis, which is easily diagnosed on indian ink staining and serology. Serum ADA concentrations were not useful in differentiating the cause of meningitis.

The laboratory technique for measuring ADA₂ is inexpensive (about £1 per test), relatively simple to perform, and can be adapted to an autoanalyser. It may thus be used in laboratories with limited resources. Measurement of ADA₂ produces results rapidly, thus potentially decreasing delays before therapy for tuberculous meningitis is initiated. These results seem promising and may make a valuable contribution to the early and accurate diagnosis of tuberculous meningitis. The use of CSF ADA₂ in the diagnosis of tuberculous meningitis should be further evaluated in larger series, including patients with other lymphocytic meningitides and different settings.