- Design
- Systematic review with prespecified dual frequentist and Bayesian random-effects meta-analysis of randomised trials
- Population
- Seven randomised trials, 2,408 adults; all eligible trials enrolled patients with septic shock
- Primary outcome
- 28-day mortality
- Effect
- Risk ratio 0.87 (95% CI 0.76–1.01, P = 0.06, I² 0%, moderate certainty); Bayesian posterior probability of RR < 1 = 97.2%
Seven randomised trials in 2,408 adults, all with septic shock, compared peripheral perfusion-guided resuscitation — titrating to capillary refill time — against standard care or lactate-guided resuscitation. The frequentist analysis gave a 28-day mortality risk ratio of 0.87 (95% CI 0.76–1.01, P = 0.06), moderate certainty, with no heterogeneity at all (I² = 0%). The two largest and least biased trials were individually non-significant.
The prespecified Bayesian analysis put the posterior probability that the risk ratio is below 1 at 97.2%, and that held across different prior assumptions. The authors present both without pretending the second settles what the first did not.
This is the clearest recent example of the two frameworks answering different questions. The frequentist result says the data are compatible with no effect. The Bayesian result says that, given these data, benefit is far more likely than harm. Both are true. What neither establishes is the size of any benefit, and that is what would decide whether to rebuild a resuscitation protocol around capillary refill. The honest position is that this is a good reason to keep assessing peripheral perfusion at the bedside and a poor one to abandon lactate.
- Keep checking capillary refill time at the bedside — it is free, repeatable, and this analysis gives no reason to stop.
- Do not replace lactate-guided targets with a capillary refill protocol on this evidence.
- Standardise how capillary refill is measured in your unit if you use it — pressure, duration and site all vary between clinicians.
- Remember the population: every eligible trial enrolled septic shock, so this says nothing about other vasodilatory shock states.
- In resource-limited settings where serial lactate is not available, peripheral perfusion assessment remains the practical alternative and this supports rather than undermines it.
Why it matters
A result reported as non-significant and a result reported as 97% probable are the same result, and which one reaches a protocol committee is arbitrary.
Don't overread it
A high Bayesian posterior probability of any benefit is not evidence that the benefit is clinically meaningful.
The statistics, in plain English
P = 0.06 and a 97.2% posterior probability of benefit are not in conflict — they answer different questions. The confidence interval (0.76 to 1.01) asks whether these data would be surprising if the treatment did nothing, and the answer is only just. The posterior probability asks how plausible benefit is given the data, and gets a high number largely because the point estimate favours treatment and I² = 0% means the trials agree. Neither tells you the effect is large enough to matter; the interval's lower bound of 0.76 and upper bound of 1.01 span exactly that uncertainty.
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 emergency & critical care, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free