Clinical utility of CSF metagenomic next-generation sequencing in suspected CNS infection: performance against a composite reference standard and read-count stratification.

BACKGROUND: Metagenomic next-generation sequencing (mNGS) of cerebrospinal fluid (CSF) is increasingly used to identify pathogens in suspected central nervous system (CNS) infections. However, integrating these results into real-world clinical decision-making remains problematic, particularly given the lack of standardized quantitative metrics beyond raw read counts. METHODS: We retrospectively analyzed 46 patients with suspected encephalitis, meningitis, or meningoencephalitis who underwent CSF mNGS testing. Etiologic certainty was classified using a composite clinical reference standard as Definite, Probable, or Unlikely. We assessed concordance between mNGS findings and the Likely etiology category (Definite or Probable), calculated diagnostic performance metrics, characterized the detected pathogens, and explored a tiered interpretation framework based on maximum read counts per patient (<10, 10-49, and >=50). Trends across read-count strata were evaluated using the Cochran-Armitage test, and exact binomial 95% confidence intervals (CIs) were calculated. RESULTS: CSF mNGS detected pathogens in 13 of 46 patients (28.3%). Positivity increased with greater adjudicated diagnostic certainty, from 0% in Unlikely cases to 11.8% in Probable cases and 73.3% in Definite cases. Within the composite reference framework, mNGS showed 40.6% sensitivity, 100% specificity, 100% positive predictive value, and 42.4% negative predictive value, indicating stronger rule-in than rule-out performance. Viral detections predominated, with herpes simplex virus type 1 and varicella-zoster virus as the most frequent pathogens; all findings should be interpreted in the context of DNA-only testing. Among mNGS-positive patients with Likely etiologies, the proportion classified as Definite increased across higher max-read strata, but these tier-specific estimates were imprecise and should be viewed as exploratory. CONCLUSION: In this real-world cohort, positive CSF mNGS results supported an infectious etiology more strongly than negative results excluded it. Max-read-based stratification may have exploratory interpretive value for positive findings, but it should not be considered a validated clinical decision rule and requires confirmation in larger multicenter studies with standardized reference standards.