Seven-year performance of a clinical metagenomic next-generation sequencing test for diagnosis of central nervous system infections

Abstract

Metagenomic next-generation sequencing (mNGS) of cerebrospinal fluid (CSF) is an agnostic method for broad-based diagnosis of central nervous system (CNS) infections. Here we analyzed the 7-year performance of clinical CSF mNGS testing of 4,828 samples from June 2016 to April 2023 performed by the University of California, San Francisco (UCSF) clinical microbiology laboratory. Overall, mNGS testing detected 797 organisms from 697 (14.4%) of 4,828 samples, consisting of 363 (45.5%) DNA viruses, 211 (26.4%) RNA viruses, 132 (16.6%) bacteria, 68 (8.5%) fungi and 23 (2.9%) parasites. We also extracted clinical and laboratory metadata from a subset of the samples (n = 1,164) from 1,053 UCSF patients. Among the 220 infectious diagnoses in this subset, 48 (21.8%) were identified by mNGS alone. The sensitivity, specificity and accuracy of mNGS testing for CNS infections were 63.1%, 99.6% and 92.9%, respectively. mNGS testing exhibited higher sensitivity (63.1%) than indirect serologic testing (28.8%) and direct detection testing from both CSF (45.9%) and non-CSF (15.0%) samples (P < 0.001 for all three comparisons). When only considering diagnoses made by CSF direct detection testing, the sensitivity of mNGS testing increased to 86%. These results justify the routine use of diagnostic mNGS testing for hospitalized patients with suspected CNS infection.

Fig. 1. Distribution of tests ordered by year and geographic location.

a,b, Distribution of tests ordered by state (a) and internationally (b). A total of 4,075 mNGS tests were performed from CSF samples collected from the United States, California being the most frequent state of origin (n = 2,420 samples). Reference laboratories such as Associated Regional and University Pathologists, Inc., Labcorp and Mayo Clinic (n = 722) receive tests from multiple states, so the location of individual samples cannot be tracked and thus are excluded from the figure. 14.8% (n = 715) of samples were sent from pediatric hospitals. c, Number of tests performed by year and number of positive results, excluding results that were reported possible or likely contaminants. *Data shown are samples analyzed up to April 2023. Source data

Fig. 2. Summary of positive results by mNGS testing (n = 4,828).

a, Number and types of organisms detected by mNGS testing. b, Detected bacterial species (total and unique), including those that are typical (nonfastidious) and atypical (uncommon, fastidious and/or difficult to diagnose). c, Detected DNA viruses, RNA viruses (including arboviruses and enteroviruses), fungi and parasites. ^aOther DNA viruses detected included human parvovirus 4 (n = 1), human parvovirus B19 (n = 4) and human herpesvirus 8 (n = 1). ^bOther RNA viruses detected included lymphocytic choriomeningitis virus (n = 3), astrovirus (n = 2), calicivirus (n = 2), coronavirus 229E (n = 2), SARS-CoV-2 (n = 1), human T cell lymphotropic virus (n = 1), human parechovirus (n = 1) and measles virus (n = 1). ^cOrthobunyaviruses detected included Cache Valley virus (n = 3), Jamestown Canyon virus (n = −1), La Crosse virus (n = 1) and Potosi virus (n = 1). Potosi virus is a novel species in this genus, not reported before as causing disease in humans. ^dOther fungi detected included Alternaria sp. (n = 2), Mucorales sp. (n = 2), Epicoccum sp. (n = 1) and Sporothrix schenkii (n = 1). BKPyV, BK polyomavirus or human polyomavirus 1; CMV, cytomegalovirus; COX, coxsackievirus; CTFV, Colorado tick fever virus; DENV, dengue virus; EBV, Epstein-Barr virus; EV, enterovirus; HBV, hepatitis B virus; HCV, hepatitis C virus; HEV, hepatitis E virus; HHV-6, human herpesvirus 6; HHV-7, human herpesvirus 7; HSV, herpes simplex virus; HIV, human immunodeficiency virus; HTLV-2, human T cell lymphotropic virus 2; JCPyV, JC polyomavirus or human polyomavirus 2; NAAT, nucleic acid amplification testing; POWV, Powassan virus; SLEV, St. Louis encephalitis virus; VZV, varicella-zoster virus; WNV, West Nile virus; YFV, yellow fever virus; ZIKV, Zika virus. Source data

Fig. 3. Evaluation of true positive mNGS test results and comparison to other microbiologic tests.

a, The proportional Venn diagram displays the overlap among four modalities (mNGS, CSF direct detection, non-CSF direct detection and serologic testing) in diagnosis of CNS infections. b, Pathogens detected by mNGS testing only (n = 48) or mNGS testing first (n = 19). Among the 19 pathogens that were detected by mNGS testing first, 11 pathogens were detected by another microbiologic test run in parallel, whereas 8 were detected by another test, but only for orthogonal confirmation of the initial mNGS positive result. c, Number and types of pathogens detected by each of the 4 diagnostic modalities. d, 2 × 2 contingency tables showing the comparative performance of mNGS testing compared to other diagnostic modalities. The P value is based on comparison between mNGS and a conventional testing modality using the two-sided McNemar’s test. CMV, cytomegalovirus; EBV, Epstein-Barr virus; HHV-6, human herpesvirus 6; HHV-7, human herpesvirus 7; LCMV, lymphocytic choriomeningitis virus; Neg, negative; NPV, negative predictive value; Pos, positive; PPV, positive predective value; VZV, varicella-zoster virus. Source data

Fig. 4. Evaluation of false-negative mNGS test results.

a, 26 samples were positive by direct detection CSF testing and negative by mNGS testing. Of these false-negative results, 10 (38.5%) were attributed to high DNA host background, a known limitation of mNGS approaches, and 4 (15.4%) to persistent antigen positivity from fungal infection after onset of treatment. b, 43 samples were positive by serology and negative by mNGS testing. Most of these false-negative results can be explained by the presumed absence of nucleic acid in the samples at the time of CSF collection, a known limitation of direct detection methods. c, 24 samples were positive by non-CSF direct detection testing and negative by mNGS testing. In these cases, the causative pathogen is presumed absent in CSF and only detectable from infected tissue or abscesses. ^aHSV-1, HSV-2, CMV, EBV and VZV. ^bListeria monocytogenes, Sporothrix shenckii., and EBV. ^cRhizopus sp., Candida albicans, Balamuthia mandrillaris, enterovirus and VZV. ^dMycobacterium chelonae, Pseudomonas aeruginosa, Staphylococcus aureus (n = 3), Enterococcus faecalis, Citrobacteri koseri, Streptococcus mitis, Cutibacterium acnes, Aggregatibacter sp., Bacillus cereus, Parvimonas micra and Streptococcus intermedius. CMV, cytomegalovirus; CSF DD, CSF direct detection testing; EBV, Epstein-Barr virus; HSV, herpes simplex virus; non-CSF DD, direct detection testing of samples other than CSF; VZV, varicella-zoster virus; WNV, West Nile virus. Source data