Phenobarbital has been gaining ground over benzodiazepines for alcohol withdrawal in critical care, largely on the basis of single-centre experience. This systematic review assessed what the accumulated evidence actually supports. Sixteen non-randomised comparative studies of critically ill adults with alcohol withdrawal were reviewed and 12 pooled, comparing phenobarbital-based with benzodiazepine-based pathways.
The headline claims did not hold. Intubation was not significantly different (odds ratio 0.62, 95% CI 0.20 to 1.87, I-squared 73%), nor was hospital length of stay (mean difference -1.75 days, 95% CI -5.07 to 1.56, I-squared 63%). What did hold was ICU length of stay, shorter by 0.60 days with phenobarbital (95% CI -0.79 to -0.41), and notably with I-squared of 0% — the studies agreed on this one.
The more interesting finding is about implementation rather than pharmacology. In exploratory subgroup analysis, front-loaded protocolised phenobarbital-first pathways looked better on both intubation and hospital stay, while adjunctive or mixed phenobarbital strategies did not. That distinction — a designed pathway versus phenobarbital added as rescue when benzodiazepines are failing — may explain much of the disagreement in the literature. It is also, the authors are clear, very low certainty and hypothesis-generating.
What to do with it. If your unit is considering a change, the evidence supports designing a front-loaded phenobarbital-first protocol rather than adding phenobarbital to an existing benzodiazepine pathway as a rescue agent — the two are different interventions and only one shows promise. But the whole evidence base is non-randomised, so patients were assigned to pathways by clinicians who knew how sick they were, and confounding by indication runs in both directions. Half a day of ICU is a modest return on which to rebuild a protocol; the case for doing so rests more on coherence than on the size of the effect.
- If changing practice, design a front-loaded phenobarbital-first protocol rather than using it as rescue.
- Expect roughly half a day less ICU stay; do not promise less intubation, which was not shown.
- All 16 studies were non-randomised, so confounding by indication cannot be excluded.
- Heterogeneity was high for intubation and hospital stay but zero for ICU stay — trust that result most.
- The subgroup finding favouring protocolised pathways is exploratory and very low certainty.
The statistics, in plain English
The I-squared values are what separate the reliable finding from the unreliable ones here. I-squared of 73% for intubation means the studies disagreed substantially, so the pooled odds ratio of 0.62 is an average across studies pointing different ways and its wide interval of 0.20 to 1.87 reflects that. I-squared of 0% for ICU length of stay means the studies agreed closely, which is why a difference as small as 0.60 days comes with a tight interval. Across all of it, these are non-randomised comparisons: sicker patients get different pathways, and no statistical adjustment removes that entirely, which is why the authors rate certainty as very low.
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