Diagnostic performance of cerebrospinal fluid examination in cases with suspected acute meningitis: a systematic review and meta-analysis

Bacterial meningitis is a severe disease with high morbidity and mortality, in which urgent diagnosis and treatment can improve the prognosis. The diagnosis relies heavily on cerebrospinal fluid (CSF) examination, but the value of different CSF parameters has not been adequately studied. In preparation for the World Health Organization guidelines for diagnosis, treatment and care we performed a systematic review and meta-analysis to determine the diagnostic accuracy of commonly used CSF parameters (Gram stain, leukocyte count and differential, glucose, total protein, and lactate) for the diagnosis of bacterial meningitis. The outcomes were sensitivity, specificity, positive and negative predictive value, likelihood ratios and area-under the receiver operating characteristics curve (AUC). A total of 27 studies with 5,673 patients were included. CSF Gram stain had an overall sensitivity of 85% (95% confidence interval [CI] 55–96%) with a 99% specificity. Per pathogen the sensitivity varied with 93% for Streptococcus pneumoniae , 30–89% for Neisseria meningitidis , 10–25% for Listeria monocytogenes , and 25–65% for Haemophilus influenzae . CSF leukocyte count had a sensitivity of 77% (95% CI 74–81%) and specificity of 83% (95% CI 75–92%), CSF neutrophil count 82% (95% CI 70–94%) and 84% (95% CI 77–90%), CSF glucose 66% (95% CI 52–79%) and 85% (95% CI 72–98%), CSF to blood glucose ratio 88% (95% CI 83–93%) and 78% (95% CI 52–100%), CSF total protein 86% (95% CI 80–92%) and 79% (95% CI 70–88%) and CSF lactate 94% (95% CI 91–98%) and 86% (95% CI 74–98%), respectively. CSF Gram stain has a high specificity, with varying but mostly moderate sensitivity per pathogen. An elevated CSF leukocyte count, specifically neutrophils, CSF protein concentration and CSF lactate are predictive of bacterial meningitis. Furthermore, a decreased CSF glucose concentration and low CSF to blood glucose ratio are also predictive of bacterial meningitis. Further studies are needed to assess combinations of these tests and novel biomarkers to improve diagnostic accuracy.