Meta-analysis of Cerebrospinal Fluid Cell Count and Biochemistry to Diagnose Meningitis in Infants Aged < 90 Days

Abstract

Objective Cerebrospinal fluid (CSF) white blood cell (WBC) count, protein, and glucose (cytochemistry) are performed to aid in the diagnosis of meningitis in young infants. However, studies have reported varying diagnostic accuracies. We assessed the diagnostic accuracy of CSF cytochemistry in infants below 90 days and determined the certainty of evidence. Study Design We searched PubMed, Embase, Cochrane Library, Ovid, Cumulative Index to Nursing and Allied Health Literature (CINAHL), and Scopus databases in August 2021. We included studies that evaluated the diagnostic accuracy of CSF cytochemistry compared with CSF culture, Gram stain, or polymerase chain reaction in neonates and young infants &lt;90 days with suspected meningitis. We pooled data using the hierarchical summary receiver operator characteristic (ROC) model. Results Of the 10,720 unique records, 16 studies were eligible for meta-analysis, with a cumulative sample size of 31,695 (15 studies) for WBC, 12,936 (11 studies) for protein, and 1,120 (4 studies) for glucose. The median (Q1, Q3) specificities of WBC, protein, and glucose were 87 (82, 91), 89 (81, 94), and 91% (76, 99), respectively. The pooled sensitivities (95% confidence interval [CI]) at median specificity of WBC count, protein, and glucose were 90 (88, 92), 92 (89, 94), and 71% (54, 85), respectively. The area (95% CI) under ROC curves were 0.89 (0.87, 0.90), 0.87 (0.85, 0.88), and 0.81 (0.74, 0.88) for WBC, protein, and glucose, respectively. There was an unclear/high risk of bias and applicability concern in most studies. Overall certainty of the evidence was moderate. A bivariate model-based analysis to estimate the diagnostic accuracy at specific thresholds could not be conducted due to a paucity of data. Conclusion CSF WBC and protein have good diagnostic accuracy for the diagnosis of meningitis in infants below 90 days of age. CSF glucose has good specificity but poor sensitivity. However, we could not identify enough studies to define an optimal threshold for the positivity of these tests. Key Points