- Design
- Systematic review and meta-analysis of 9 studies, random effects
- Population
- 10,785 adults with non-variceal upper GI bleeding (2,638 ultra-early, 8,147 early)
- Primary outcome
- Mortality and rebleeding
- Effect
- Mortality OR 0.88 (95% CI 0.58–1.34); rebleeding OR 1.23 (0.90–1.67); ICU admission OR 1.41 (1.20–1.66)
This systematic review and meta-analysis pooled nine studies of 10,785 adults with non-variceal upper gastrointestinal bleeding, comparing endoscopy within six hours of presentation (2,638 patients) with endoscopy after six hours (8,147), most within the 24 hours guidelines recommend.
Ultra-early endoscopy did not reduce mortality (OR 0.88, 95% CI 0.58–1.34) or rebleeding (OR 1.23, 0.90–1.67), and did not change transfusion, surgery, repeat endoscopy or length of stay. ICU admission was more frequent in the ultra-early group (OR 1.41, 1.20–1.66). Results held in high-risk patients, most often defined as Glasgow-Blatchford score above 12.
The pooled studies are mostly observational, so sicker patients are likely to have been scoped sooner — which may explain the higher ICU rate and blunt any benefit. But the direction matches the one randomised trial in high-risk patients, which also found no mortality gain from endoscopy within six hours. The practical message for the medical registrar at 2 a.m. is that resuscitation comes first and a stable patient can wait for a planned list.
- Resuscitate first: restrictive transfusion (usually at haemoglobin below 7–8 g/dL), IV proton pump inhibitor, correct coagulopathy.
- Aim for endoscopy within 24 hours in non-variceal bleeding; rushing to within six hours showed no benefit.
- Calculate the Glasgow-Blatchford score at presentation; a score of 0–1 can often be managed as an outpatient.
- Escalate urgently for haemodynamic instability despite resuscitation — that is the exception, not a high score alone.
- Suspected variceal bleeding is a different pathway; this analysis does not apply.
Why it matters
It removes the pressure for out-of-hours endoscopy in stable patients, which carries its own risks and costs.
Don't overread it
Most included studies were observational, so sicker patients were probably scoped sooner — the higher ICU rate is likely confounding, not harm.
The statistics, in plain English
The mortality odds ratio of 0.88 has a confidence interval from 0.58 to 1.34 — a wide range consistent with either benefit or harm, which means the studies cannot show a difference. Pooling observational studies does not remove their biases.
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 internal medicine, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free