Articles
Rapid pan-microbial metagenomics for pathogen detection and personalised therapy in the intensive care unit: a single-centre prospective observational study
Adela Alcolea-Medina, Luke B Snell, Gul Humayun, et al
Lancet Microbe 2025; 6: 101174
https://doi.org/10.1016/j.lanmic.2025.101174
Summary
Background
Most clinical metagenomic studies do not provide rapid results, detect pathogens from all microbial kingdoms, or measure clinical impacts. We aimed to evaluate the feasibility, performance, and clinical impacts of a rapid pan-microbial respiratory metagenomic service for patients admitted to intensive care units (ICUs).
Methods
This was a single-centre observational study of a rapid metagenomics service that tests respiratory samples from ICU patients at Guy’s and St Thomas’ hospitals, London, UK, between Dec 5, 2023, and April 12, 2024. Testing used a previously published pan-microbial metagenomics workflow, which simultaneously detects bacteria, fungi, and DNA and RNA viruses; provides same-day preliminary results after 2 h; and provides final results after 24 h. Patients were included if they were aged 18 years or older, admitted to the ICU, had confirmed respiratory failure requiring supplemental oxygen or advanced airway support, and had at least one of the following: (1) clinical suspicion of lower respiratory tract infection based on clinical, biochemical, or radiological findings, (2) sepsis of unknown origin, and (3) concern from an intensive care physician regarding inflammatory pathology. Patients with a suspected or confirmed containment level three organism were excluded. The outcome was performance characteristics of the metagenomic test compared with routine diagnostic testing, detection of additional pathogens by metagenomics, change in antimicrobial prescribing within 24 h of testing, and initiation of immunomodulation.
Findings
We processed 114 samples (1–5 per day) from 74 patients (39 [53%] female and 35 [47%] male). 107 (94%) of 114 samples passed quality control, of which 101 (94%) provided same-day preliminary results. Bacteria were detected in 45 (43%) of 104 tested specimens, fungal organisms in 17 (16%) of 104 tested specimens, and viruses in 28 (34%) of 83 tested specimens. Sensitivity in lower respiratory tract samples after 24 h was 97% (95% CI 87–100) for bacteria, 89% (65–99) for fungi, and 89% (71–98) for viruses, with only one false positive for bacteria. Metagenomics identified 42 pathogens not detected by other tests in 32 (30%) of 107 samples. Antimicrobial therapy was changed after metagenomic results from 30 (28%) of 107 samples: 22 (21%) were de-escalated and eight (7%) were escalated. Metagenomics contributed to the initiation of immunomodulation in 15 (20%) of 74 patients for a range of inflammatory conditions. Pathogens with clinical significance to local infection control or national public health were found in ten (14%) of 74 patients, including three invasive Group A streptococci, two parvovirus B19, and one each of HIV-1, measles virus, Mycobacterium tuberculosis, Neisseria meningitidis, and Mycoplasma pneumoniae.
Table 1. Demographic characteristics, clinical categorisation, and treatment outcomes
| Empty Cell | Value |
|---|---|
| Patients (n=74) | |
| Age, years | 50 (38–63) |
| Sex | |
| Female | 39 (53%) |
| Male | 35 (47%) |
| Admission category | |
| Severe respiratory failure | 48 (65%) |
| Medical | 16 (22%) |
| Surgical | 7 (9%) |
| Cardiac arrest | 3 (4%) |
| Level of respiratory support | |
| ECMO | 27 (36%) |
| Invasive ventilation | 40 (54%) |
| Other | 7 (9%) |
| Samples (n=114) | |
| Length of stay before sampling | |
| ≤48 h | 26 (23%) |
| >48 h | 88 (77%) |
| Suspected focus of infection | |
| CA-LRTI | 41 (36%) |
| HA-LRTI | 31 (27%) |
| Diagnostic uncertainty | 25 (22%) |
| Other focus | 13 (11%) |
| Suspected non-infectious | 4 (5%) |
| Broad spectrum antibiotics before sampling | 86 (75%) |
| Sample processing | |
| Pass | 107 (94%) |
| Fail | 7 (6%) |
| Outcome (n=107) | |
| Antimicrobial change | |
| De-escalate | 22 (21%) |
| Escalate | 8 (7%) |
| No change | 77 (72%) |
| Immunomodulation initiation | 15 (14%) |
Table 2. Performance characteristics of all samples and lower respiratory tract samples after 24-h results on a per-sample basis and separated by microbial kingdom
| Empty Cell | Number of samples | True positive | True negative | False positive | False negative | Sensitivity | Specificity | Negative predictive value | Positive predictive value |
|---|---|---|---|---|---|---|---|---|---|
| All samples | |||||||||
| Bacteria | 104 | 45 | 57 | 1 | 1 | 98% (88–100) | 98% (91–100) | 98% (89–100) | 98% (87–100) |
| Viruses | 83 | 28 | 52 | 0 | 3 | 90% (74–98) | 100% (93–100) | 95% (86–98) | 100% (88–100) |
| Fungi | 104 | 17 | 85 | 0 | 2 | 89% (67–99) | 100% (96–100) | 98% (92–99) | 100% (80–100) |
| Lower respiratory tract samples | |||||||||
| Bacteria | 89 | 38 | 49 | 1 | 1 | 97% (87–100) | 98% (89–100) | 98% (89–100) | 97% (85–100) |
| Viruses | 74 | 24 | 47 | 0 | 3 | 89% (71–98) | 100% (92–100) | 94% (84–98) | 100% (86–100) |
| Fungi | 89 | 16 | 71 | 0 | 2 | 89% (65–99) | 100% (95–100) | 97% (91–99) | 100% (79–100) |
Interpretation
Respiratory metagenomics for ICU patients showed good performance and turnaround time, and diverse clinical and public health benefits. This ability to inform both personalised patient therapy and infectious disease surveillance needs evaluation in multicentre studies.
Funding
None.
