TB Meningitis in an HIV-Endemic Setting

INTRODUCTION: Tuberculous meningitis (TBM) is a clinical challenge because early treatment is essential yet early diagnosis is difficult. Furthermore, TBM is often associated with HIV co-infection. This diagnostic study had three main objectives: 1) To test the performance of existing diagnostic rules in an HIV-endemic setting; 2) To determine which admissions variables are most predictive of TBM; 3) To compare the clinical presentation of TBM in HIV-positive vs. HIV-negative patients. METHODS: This prospective, observational cohort consists of 96 patients from Tygerberg Hospital in Stellenbosch, South Africa. Criteria for entry were age > 15 years, clinical features of meningitis, and an abnormal lumbar puncture (high CSF protein or CSF pleocytosis). Data include the standard clinical and laboratory values: age, gender, duration of symptoms, neurological exam findings, CSF values, blood, microscopy, culture, imaging, etc. 1) Two diagnostic algorithms described by Thwaites (Lancet, 2002) are tested: LRM (logistic regression) and CART (classification tree). We determine each patient's TBM status according to Thwaites' criteria, and then calculate the sensitivities and specificities of LRM and CART. 2) Distributions are compared between Definite TBM and Not TBM groups, first in HIV-positive and then in HIV-negative patients, using the appropriate tests for continuous and categorical variables. 3) The same methods are used to compare distributions between HIV-positive and HIV-negative patients with Definite TBM. RESULTS: In HIV-negative patients, LRM had a sensitivity of 100% and a specificity of 37.5%, while CART had a sensitivity of 91% and a specificity of 50%. In HIV-positive patients, this fell to sensitivity = 100%, specificity = 12.5% for LRM, and sensitivity = 94%, specificity = 12.5% for CART. In HIV-negative patients, factors including CSF white cell count, WBC count, blood sodium, age, duration of symptoms, and presence of systemic symptoms were statistically significant (p < 0.05). In HIV-positive patients, only two variables were statistically significant: CSF glucose and positive chest x-ray. In patients with Definite TBM, clinical presentation was generally similar, regardless of HIV status- with the exceptions of age and WBC count. CONCLUSION: The data suggest that in HIV-positive patients, there is considerable overlap in presentation between TBM and other forms of meningitis, making a challenging diagnostic problem even harder. Current algorithms are in urgent need of revision, especially in HIV-positive patients.