A systematic review compared phenobarbital-based with benzodiazepine-based pathways for alcohol withdrawal in critically ill adults, drawing on sixteen non-randomised studies with twelve in the meta-analysis, and assessing bias with ROBINS-I.
Intubation did not differ significantly: an odds ratio of 0.62 with an interval from 0.20 to 1.87 across eleven studies, and I² of 73%. Hospital length of stay did not differ either, at -1.75 days with an interval from -5.07 to 1.56. ICU length of stay was shorter with phenobarbital, at -0.60 days with an interval from -0.79 to -0.41 and, notably, I² of 0%.
That last figure deserves attention. Zero heterogeneity across eight studies means they agreed closely, which is unusual and makes the ICU stay result the most solid finding here — while also being the smallest, at fourteen hours.
Exploratory subgroups suggested that front-loaded, protocolised phenobarbital-first pathways reduced intubation and hospital stay. The authors rate the certainty of that as very low, and it is worth saying why: these are non-randomised comparisons, so units adopting a phenobarbital protocol differ from those that have not in staffing, monitoring and case mix, and no adjustment recovers that.
- Entirely non-randomised evidence; confounding by unit rather than by patient
- ICU stay 0.6 days shorter, with unusually consistent agreement across studies
- Intubation and hospital stay showed no significant difference
- The front-loaded protocol subgroup is exploratory and rated very low certainty
- A pathway is being compared, not a drug — the protocol may matter more than the agent
The statistics, in plain English
The contrast between I² of 73% for intubation and 0% for ICU stay is the most informative thing here. High heterogeneity means the studies genuinely disagree and the pooled odds ratio is a weak summary; zero means they agree closely, so that estimate is trustworthy within the limits of the design. Non-randomised throughout, so the direction is suggestive rather than causal.
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