SAN FRANCISCO (GenomeWeb) – Researchers from the University of California, San Francisco have found that a metagenomic next-generation sequencing test in cerebrospinal fluid for patients with an unknown neurological illness could diagnose more cases than conventional culture and PCR methods, but that serology-based testing was still important for yielding a diagnosis in cases where all other tests failed.
The prospective, multi-center study, published on Wednesday in the New England Journal of Medicine, found that almost 30 percent of around 200 patients evaluated for suspected meningitis, encephalitis, or myelitis could be diagnosed with an infection by one of the methods used, including metagenomic NGS, PCR, culture, and serology. Another 20 percent ultimately received other diagnoses, including of autoimmune disorders, while about half remained undiagnosed.
Metagenomic NGS diagnosed 13 more patients than culture or PCR-based methods, but also missed a diagnosis in 26 patients who were either diagnosed by serology or from a different sample than the cerebrospinal fluid on which mNGS testing was performed. In 19 cases, mNGS and conventional testing agreed on the diagnosis.
Over the last several years, metagenomic sequencing has been looked at as an attractive method for infectious disease diagnostics because of its unbiased nature — it will analyze all nucleic acid present, so physicians do not have to test for one pathogen at a time — and because it does not rely on culture, so they can identify unculturable bacteria and viruses.
However, it remains unclear where in the clinical testing paradigm mNGS would be best implemented and for what clinical indications, and whether it would be a cost-effective alternative to other infectious disease tests.
The study is "much needed" in that it "illustrates both the usefulness and limitations of metagenomic sequencing in a real-life clinical setting," said Anne Piantadosi, an infectious disease physician at Massachusetts General Hospital who previously presented on efforts her group has made in developing its own metagenomic sequencing test for central nervous system infections.
Earlier this year, infectious disease diagnostics firm Karius also published results of its own metagenomic sequencing assay, which included both a study on its utility in sepsis and an evaluation of its performance for the first 2,000 clinical samples. The Karius test differs in that it analyzes circulating cell-free DNA from patients' plasma instead of CSF and is not focused specifically on neurological illness.
In the NEJM study, the researchers evaluated the use of the mNGS test for uncovering the cause of neurological illness. The test analyzes both DNA and RNA from CSF and makes use of a custom bioinformatics pipeline, SURPI+.
Previously, the UCSF team published an analytical validation of the test in Genome Research, demonstrating 73 percent sensitivity and 96 percent specificity compared to conventional testing. It also had 81 percent positive predictive agreement and 99 percent negative predictive agreement after further evaluating the samples with discrepant test results and determining that for some of those, the mNGS test did make the correct call.
Charles Chiu, senior author of the NEJM study and a professor of laboratory medicine and medicine at UCSF and associate director of the UCSF Clinical Microbiology Laboratory, said that the clinical lab now runs the test one to two times per week, batching between 13 and 16 samples per run. The test is offered for patients with evidence of CSF inflammation, meningitis, encephalitis, or myelitis, or for patients who are highly immunocompromised, he said, and UCSF charges about $2,200 for it.
In the study, patients with neurological illness were enrolled at eight different participating hospitals: UCSF; the Zuckerberg San Francisco General Hospital; the University of California Davis; the University of California, Los Angeles; Children's Hospital Los Angeles; Children's Hospital Colorado; St. Jude's Children's Research Hospital, and the Children's National Medical Center. All samples were tested via mNGS at UCSF.
Overall, 204 patients who met the study criteria enrolled. The researchers analyzed their samples via metagenomic sequencing and conventional direct-detection infectious disease tests that included culture, PCR, and antigen testing, as well as indirect testing, such as serologic testing.
In total, 103 patients, or just over half, were diagnosed, 57 of them with an infection. Of the 58 infectious agents identified, mNGS and conventional testing agreed on 19, while 26 were diagnosed via conventional tests only and 13 by mNGS only.
Piantadosi said that it was "notable that the [mNGS test] detected 13 infections that were missed by routine diagnostics, highlighting the limitations of our existing testing methods, as well as the fallibility of clinicians' diagnostic choices."
In addition, for eight of those 13 patients, clinicians reported that the mNGS test "favorably affected their clinical reasoning," and for seven of the 13 patients, the results guided therapy.
For the 26 patients where mNGS testing did not identify a pathogen, but conventional testing did, the researchers further analyzed why the mNGS test missed the pathogen. Eleven cases were only diagnosed via serology while seven cases missed by mNGS ended up being diagnosed from a sample other than cerebrospinal fluid, such as a brain biopsy.
Piantadosi said that this finding illustrates the limitations of metagenomic sequencing, primarily that "nucleic acid from the responsible pathogen must be present in the sample being tested."
For instance, some of the samples only diagnosed by serologic methods were cases of West Nile virus, and, as other researchers have previously reported, symptoms of West Nile fever often persist after the virus has been cleared, but antibodies can still be detected by serologic methods.
In eight cases, pathogen levels were below the threshold for which the mNGS test would be called a positive, but in all of those cases, there were sequencing reads that mapped to the causative pathogen.
The authors also noted that one complicating factor was that although CSF samples were obtained at a median of three days after the patient's initial presentation, for 72 patients, samples were obtained at around eight days after presentation. Many of these patients had already been on antibiotics, which may have reduced the levels of pathogen nucleic acids in the CSF samples.
In addition, Piantadosi said that around half of the patients in the study never received a diagnosis from any method, which "emphasizes just how challenging these clinical syndromes are, and underscores the need for continued improvement in diagnostic methods for both infectious and non-infectious causes of meningitis and encephalitis."
Chiu noted that the researchers are now planning to conduct several follow-on studies, including one testing mNGS in plasma to diagnose fever and sepsis. In addition, he said, the researchers are working on making use of the sequencing data from human RNA in the CSF samples to see whether they can be used to "discriminate between bacterial infection, viral infection, and autoimmune disease on the basis of the host response," he said.
To increase turnaround time, the team is also looking to move the assay, which currently runs on an Illumina sequencing instrument, to Oxford Nanopore Technologies' nanopore sequencing platform. That would enable the team to reduce turnaround time to six hours from between 72 and 96 hours, Chiu said.
The NEJM study was one result of the Precision Diagnosis of Acute Infectious Diseases (PDAID) project, which was funded by the California Initiative to Advance Precision Medicine with support from the Sandler, William F. Bowes Jr., and Charles and Helen Schwab Foundations, as well as in-kind support from UCSF. Another expected outcome from the PDAID project is a cost-effectiveness analysis, which researchers from the University of California, Berkeley are conducting.