Rapid pan-microbial metagenomics for pathogen detection and personalised therapy in the intensive care unit: a single-centre prospective observational study.
Summary
A same-day pan-kingdom metagenomic service in ICU patients achieved high sensitivity (bacteria 97%, fungi 89%, viruses 89%) and identified additional pathogens in 30% of samples. Results altered antimicrobial therapy in 28% and contributed to immunomodulation in 20% of patients, demonstrating both patient-level and public health value.
Key Findings
- 94% of 114 samples passed QC; 94% of QC-passed samples yielded same-day preliminary results.
- Sensitivity in lower respiratory tract samples after 24 h: bacteria 97% (95% CI 87–100), fungi 89% (65–99), viruses 89% (71–98), with only one bacterial false positive.
- Metagenomics identified 42 additional pathogens in 32 (30%) of 107 samples beyond routine diagnostics.
- Antimicrobial therapy changed in 28% of samples (21% de-escalated, 7% escalated); immunomodulation initiated in 20% of patients.
- Clinically significant pathogens for infection control/public health were found in 14% of patients.
Clinical Implications
Incorporating rapid metagenomics into ICU diagnostics can enable earlier de-escalation/escalation of antimicrobials, support initiation of immunomodulators, and enhance infection control and surveillance. Multicentre evaluation should address outcomes, stewardship, and cost-effectiveness.
Why It Matters
This study operationalizes rapid pan-microbial metagenomics with demonstrated clinical decision impact, a key step beyond analytic validity alone. It sets a benchmark for integrating metagenomics into ICU sepsis workflows.
Limitations
- Single-centre design with modest sample size and potential selection bias.
- No randomized assessment of patient outcomes; exclusion of containment level 3 organisms.
- Resource and cost requirements not assessed.
Future Directions
Conduct multicentre trials to evaluate patient outcomes, antimicrobial stewardship metrics, and cost-effectiveness; standardize reporting thresholds; integrate with host-response biomarkers.
Study Information
- Study Type
- Cohort
- Research Domain
- Diagnosis
- Evidence Level
- III - Prospective observational cohort without randomization
- Study Design
- OTHER