- Design
- Multinational prospective cohort with stool metagenomics and machine learning
- Population
- 679 outpatients with cirrhosis in seven countries
- Primary outcome
- Non-elective hospitalisation within 90 days
- Effect
- 25% admitted; AUC clinical 0.79, microbiome 0.74, combined 0.84
A multinational metagenomic study in Gut (21 September) profiled stool from 679 outpatients with cirrhosis in seven countries and followed them for 90 days. A quarter needed non-elective hospitalisation.
Patients who were admitted had lower microbial diversity across all countries, despite very different baseline compositions. After adjusting for country, severity, aetiology and treatment, 19 species were independently associated with admission, including more Enterococcus faecium and loss of short-chain-fatty-acid-producing commensals, with more antibiotic resistance genes. A model combining clinical and microbiome features predicted admission better than clinical features alone (AUC 0.84 vs 0.79).
Stool metagenomics is not a clinical test, and the gain over clinical prediction is modest. But the consistent loss of diversity supports the clinical instinct to avoid needless antibiotics and PPIs, which both disrupt the microbiome in cirrhosis.
- Review the indication for long-term PPIs in patients with cirrhosis
- Avoid unnecessary antibiotics; use rifaximin and prophylaxis where indicated
- Watch closely those with advancing MELD or recent decompensation
- Do not order microbiome testing for prognosis outside research
Why it matters
It links a modifiable factor, the gut microbiome, to short-term outcomes across very different populations.
Don't overread it
The microbiome findings are associations and the prediction model has not been externally validated.
The statistics, in plain English
An AUC of 0.84 means the combined model correctly ranks a hospitalised and a non-hospitalised patient 84% of the time, against 79% for clinical data alone. The models were developed in this cohort; independent validation is needed.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for gastroenterology & hepatology, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free