- Design
- Systematic review and random-effects meta-analysis of randomised and quasi-randomised trials, GRADE assessed
- Population
- Children aged 2–59 months: 7,699 across four trials for fast-breathing pneumonia, 4,864 across three for chest indrawing without danger signs
- Primary outcome
- Treatment failure at day 6 and day 14; mortality; serious adverse events
- Effect
- Fast breathing: treatment failure RR 0.84 (95% CI 0.75–0.94, I² 0%). Chest indrawing: day-14 failure RR 0.94 (0.79–1.13); mortality RR 0.28 (0.09–0.86)
Two questions have sat unresolved behind the WHO pneumonia algorithm: does oral amoxicillin beat no antibiotic for fast-breathing pneumonia, and can outpatient oral treatment replace inpatient injectables for chest indrawing without danger signs? This review pooled seven randomised or quasi-randomised trials to answer both.
For fast-breathing pneumonia, four trials in 7,699 children found oral amoxicillin reduced treatment failure at day 14 by 16% (relative risk 0.84, 95% CI 0.75–0.94, I² = 0%, moderate certainty). The heterogeneity of zero is worth noting — the trials agree.
For chest-indrawing pneumonia without danger signs, three trials in 4,864 children found treatment failure similar at day 6 (RR 0.96, 0.83–1.11) and day 14 (0.94, 0.79–1.13), and mortality 72% lower with oral amoxicillin (RR 0.28, 0.09–0.86). Serious adverse events did not differ. The mortality finding deserves a careful reading rather than a headline: the authors state the absolute risk difference was small, which with so wide an interval means it rests on few deaths. What the data support is that outpatient oral treatment is not inferior — which in a setting where admission is difficult, expensive or unsafe is the finding that matters.
- Classify carefully: this applies to chest indrawing without danger signs, and the danger-sign assessment is what makes it safe.
- Give oral amoxicillin for fast-breathing pneumonia rather than withholding antibiotics — the trials agree on this.
- Arrange explicit follow-up at 48 hours for any child managed as an outpatient, and tell the caregiver which signs mean return immediately.
- Check the child can take and keep down oral medication before deciding against admission.
- This is directly usable in Indian practice, where distance, cost and bed availability often make admission the harder option — but only where the danger-sign assessment has actually been done.
Why it matters
The case for admitting a child with chest indrawing has rested on an assumption these trials do not support.
Don't overread it
The 72% mortality reduction rests on a small number of deaths and an interval too wide to quote as an effect size.
The statistics, in plain English
The treatment-failure comparisons are the reliable ones: intervals of 0.83–1.11 and 0.79–1.13 straddle 1.0 tightly, which is what non-inferiority looks like. The mortality relative risk of 0.28 with an interval from 0.09 to 0.86 is a different kind of number — the width tells you it comes from very few deaths, and an interval that wide is compatible with anything from a dramatic benefit to a modest one. Treat it as supporting the non-inferiority conclusion, not as a claim that oral treatment saves lives that injectables do not.
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 paediatrics, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free