Emergency_medicine · Acute infection and sepsis assessment
TriVerity
Inflammatix, Inc.
TriVerity reads gene expression rather than looking for a pathogen: 29 messenger RNAs of the immune response itself, amplified from one blood tube in about half an hour. In its 1,222-patient pivotal study across 22 emergency departments the bacterial and viral scores out-performed C-reactive protein, procalcitonin and white-cell count in the 729 patients with a consensus infection adjudication, and the severity score out-classified qSOFA for who would need critical care within a week. What no study has yet shown is that those better numbers change what happens to the patient — the outcomes study is enrolling now.
Performance Metrics
Clinical Evidence
The evidence base is anchored by the pivotal SEPSIS-SHIELD study (NCT04094818), run for FDA clearance and published in Nature Medicine in 2025: a prospective enrolment of 1,222 adults with suspected acute infection or sepsis across 22 emergency departments in the United States and Europe. The headline diagnostic figures rest on the 729 patients whose infection status reached consensus adjudication, the paper's primary diagnostic endpoint: there the bacterial score reached an area under the curve of 0.83 and the viral score 0.91 — higher than C-reactive protein, procalcitonin or white-blood-cell count. On the full population of 1,222 under forced adjudication, which groups the probable and unlikely cases with the certain ones, the figures fall to 0.76 for bacterial and 0.83 for viral infection. The severity score reached 0.78 in the 1,120 patients evaluable for the prognostic endpoint, predicting the all-cause need for critical-care interventions within seven days and reclassifying risk beyond clinical assessment with qSOFA alone. Each of the three scores held rule-in specificity above 92 per cent and rule-out sensitivity above 95 per cent at its outer interpretation bands, and the diagnostic scores report in five bands with rising likelihood ratios rather than as a bare positive-or-negative. A separate analytical validation in the Journal of Clinical Microbiology (2025) covers precision, reproducibility, interference and ease of use of the cartridge-and-instrument system, with sample-to-answer in roughly 30 minutes and under a minute of hands-on time. The intellectual lineage runs back through a 2020 Nature Communications paper describing the 29-mRNA neural-network classifier from which the test's algorithms descend. The honest caveats: the pivotal and analytical papers are manufacturer studies, the severity readout is prognostic rather than diagnostic, and no published trial yet shows the scores changing antibiotic decisions, admissions or outcomes — that question belongs to the interventional TIMED study (NCT06637904), registered as a non-randomised pre/post design with an estimated 300 participants and a start date of 8 November 2024, at Johns Hopkins and OSF Saint Francis, with endpoints of SEP-1 bundle compliance and time to emergency-department disposition. The company's launch announcement cited 600 patients; the registry figure is recorded here.
| Study | Design | n | Sensitivity | Specificity | AUC | Published |
|---|---|---|---|---|---|---|
| Liesenfeld O, Arora S, et al. | ProspectiveProspective | 1,222 | Rule-out >95% (each score, outer band) | Rule-in >92% (each score, outer band) | 0.83 bacterial / 0.91 viral (n=729 consensus-adjudicated); 0.76 / 0.83 by forced adjudication on all 1,222; 0.78 severity (n=1,120) | Nature Medicine, 2025 (PMID 41028541; DOI 10.1038/s41591-025-03933-y); SEPSIS-SHIELD pivotal study (NCT04094818), conducted for FDA clearance — prospective enrolment of 1,222 adults with suspected acute infection or sepsis across 22 US and European emergency departments. Primary diagnostic endpoint on the 729 consensus-adjudicated patients: bacterial AUROC 0.83, viral 0.91, out-performing C-reactive protein, procalcitonin and white-blood-cell count; forced adjudication on all 1,222: 0.76 and 0.83. Severity score AUROC 0.78 in 1,120 evaluable patients for all-cause critical-care interventions within 7 days, reclassifying risk versus qSOFA alone |
Inside the algorithm
Editorial featureHow TriVerity reads a patient.
Five stages — from raw input to verdict — drawn from manufacturer documentation and the public regulatory record.
- INGEST
- NORMALISE
- DETECT
- LOCALISE
- VERDICT
Stage 01 · INGEST
One tube of blood, stabilised for RNA.
The input is a single draw of whole blood collected into a PAXgene Blood RNA tube, which stabilises the messenger RNA of circulating immune cells at the moment of draw. That choice defines the device: TriVerity never looks for a pathogen. It reads the transcriptional state of the host immune system — the signal that is present from the first hours of infection, whether or not a culture will ever grow.
Input
Whole blood (PAXgene)
Inference
Per-sample, on demand
Inside the Auris+ Listing
Five more sections complete this device’s Auris+ Listing.
Decision Ledger
ProPro unlocks a private, cross-vendor log of every case you read on this device — what the AI called, what you concluded, and a one-line reason if you overrode it. Ready for the EU AI Act's deployer logging obligations when they land in 2028.
Clinical Evidence Deep Dive
ProPro unlocks the structured clinical-evidence summary — study count, target patient population, and a tabular accuracy-metrics view drawn from peer-reviewed sources.
Peer-Reviewed Publications
ProPro unlocks the curated peer-reviewed publication list with PubMed cross-links — the citation backbone of every editorial verdict.
Post-Market & Regulatory Conditions
ProPro unlocks the post-market surveillance summary, recall record, and the conditions of approval that bound real-world use.
AI Algorithm Version History
ProPro unlocks the chronological record of algorithm version changes — what changed when, drawn from manufacturer changelogs and regulatory filings.
Regulatory Approvals
Safety Record
No recalls, FDA safety communications or published adverse-event reports attributable to the TriVerity Test System or the Myrna instrument were identified in public sources as of August 2026 — read as none found rather than a completed audit, since the FDA MAUDE and recall databases could not be queried directly from this environment. The cautions that matter are structural. The test does not identify an organism: a high bacterial score says the immune system looks bacterially provoked, not which bacterium or which antibiotic, so culture and susceptibility work continue alongside it. The severity score is all-cause prognosis, not a sepsis diagnosis. And the roughly 30-minute turnaround, fast for a molecular test, still means the result lands after the first triage decisions — the intended slot is the disposition decision, not the door.
Intended Use & Indications
The TriVerity Test System is an automated, semi-quantitative in-vitro diagnostic that measures the relative expression of host-response genes in RNA isolated from whole blood collected in a PAXgene Blood RNA tube. A single-use TriVerity cartridge runs on the Myrna instrument, which amplifies 29 host-response messenger RNAs plus three housekeeping genes by reverse transcription loop-mediated isothermal amplification and interprets the result with machine-learning-derived algorithms. The system reports three independent readouts for adult patients presenting to the emergency department with suspected acute infection or suspected sepsis: a score for the likelihood of bacterial infection, a score for the likelihood of viral infection, and a score for the risk of severe illness, the latter grounded in the all-cause need for critical-care interventions within seven days of presentation. The scores are adjuncts to clinical assessment: they do not identify a pathogen, do not replace culture or clinical judgment, and the treating clinician remains responsible for diagnosis and management.