Diagnostic accuracy of a commercial multiplex PCR for the diagnosis of meningitis and encephalitis in an Italian general hospital.

Infectious meningitis and encephalitis are potentially life-threatening conditions caused mostly by bacterial and viral agents. Rapid diagnosis and prompt treatment are associated with a more favorable outcome. In recent years nucleic acid amplification tests have been developed to speed detection and identification of pathogens directly from cerebrospinal fluid (CSF). The aim of this study was to compare the diagnostic accuracy of a commercially available multiplex PCR assay for etiological diagnosis of infectious meningitis directly from CSF samples with culture. A secondary endpoint was to look for a possible screening threshold based on main CSF indices and urgent blood test results, to define CSF samples with low pre-test probability of PCR and/or culture-positive result. We performed a secondary analysis of results of CSF samples already processed as part of routine clinical care from February 2016 to December 2018. In all, 109 CSF samples were included in the study and a total of 14 bacteria were identified by either PCR, culture or both methods, along with nine samples positive for viruses. The comparison of PCR results with culture showed no significant difference: 7/109 (6.4%) vs 13/109 (11.9%) respectively, p=0.07. After exclusion of the isolates not detectable by the multiplex PCR panel, the diagnostic accuracy was: 100% (95% confidence interval (CI): 54.1% to 100%) sensitivity; 98.9% (95% CI: 93.5% to 99.9%) specificity; 85.7% (95% CI: 42% to 99.2%) positive predictive value; 100% (95% CI: 95.1% to 100%) negative predictive value; 96 (95% CI: 13.6 to 674.6) LR+; Zero LR-; Cohen's kappa: 0.918, p<0.0001. CSF protein value ≤ 28 mg/dl and CSF glucose/blood glucose ratio ≥0.78 were associated with both PCR-negative result for bacteria or viruses and culture-negative result. The multiplex PCR evaluated in this study showed a very good diagnostic performance compared to culture, and the thresholds found can be a useful tool to best choose which samples to test.