The study was approved by the Research Ethics Committee of the University of Cape Town, and performed at GF Jooste Hospital, a public sector adult referral hospital serving a population of 1.3 million. Forty-three percent of patients admitted to the medical wards have proven or clinically suspected HIV [4], and TB incidence in the referral area exceeds 1000/100,000 person-years [5].
Participants and procedures
Patients undergoing lumbar-punctures (LPs) between 1st January 2006 and 31st December 2008 were studied. Results of all LPs performed for any indication were recorded, the majority being for suspected meningitis in acute admissions, with demographic information from laboratory request forms and clinical notes. Cerebrospinal-fluid (CSF) samples underwent macroscopic examination, protein and glucose quantification, cell counts using a Neubauer counting chamber following crystal violet staining, Gram-staining of centrifuged sediment, and bacterial culture on blood and chocolate agar for 72 hours. India-ink staining of centrifuged CSF, cryptococcal antigen (CrAg) testing (Meridian Cryptococcal Latex Agglutination System, Meridian Bioscience) if India-ink negative, and fungal culture of all samples on potato dextrose agar (PDA) slopes for 14 days were carried out. TB microscopy (Auramine-flourescent stain of the sediment when sufficient sample), liquid culture in mycobacterial growth indicator tubes (MGIT, Becton-Dickenson), Lowenstein-Jensen agar slopes if sufficient sample, and TB-PCR (Genotype MTBDRplus, Hain Lifesciences) on positive culture samples from MGIT were performed in cases of suspected TB meningitis (TBM) at the clinician's request, or when CSF findings were suggestive of TBM. In a limited number of cases TB-PCR was performed directly on CSF samples in an effort to rapidly obtain drug susceptibility data. VDRL and TPHA testing were carried out at the clinician's request.
Evaluation and outcomes
When a patient had more than one LP, a separate clinical episode was defined when there was ≥1 month between LPs, except in cases of TBM, where any repeat LP within 6 months was considered part of the same episode. These definitions were based on the best available evidence, and aimed to reflect the total burden of meningitis due to differing aetiologies, including relapse episodes. Median duration of admission for cryptococcal meningitis at our hospital is 15 days, with an inter-quartile range (IQR) of 13 to 20 days (unpublished data). For TB meningitis a cut-of off 6 months was chosen, as patients are treated with 6-9 months of therapy, and the TB programme regards a re-presentation with TBM during treatment as a deterioration of the initial episode, rather than a "recurrence" or "relapse".
Cases were classified by microbiological diagnosis, or in the absence of definitive microbiology as:
1) normal CSF (neutrophils ≤ 1 × 106/L, lymphocytes ≤ 5 × 106/L, protein ≤ 0.5 g/dL, and glucose ≥1.5 mmol/L with no organism isolated),
2) minor abnormalities (neutrophils 2-5 × 106/L, lymphocytes 6-20 × 106/L, protein 0.51-1.0 g/dL, or glucose 1.0-1.49 mmol/L and no organism isolated) or
3) markedly abnormal (neutrophils>5 × 106/L, lymphocytes>20 × 106/L, protein>1.0 g/dL, or glucose<1.0 mmol/L and no organism isolated).
Markedly abnormal cases were further classified into lymphocytic (lymphocyte predominance, LØ:NØ ratio ≥3:1), mixed (LØ:NØ ratio between 3 and 0.33), and pyogenic (neutrophil predominance, NØ:LØ ratio ≥3:1).
HIV status was determined for all patients with microbiologically confirmed disease during the last 2 years of the study from laboratory records and clinical notes.