Phenobarbital has been displacing benzodiazepines in intensive care alcohol withdrawal pathways largely on enthusiasm and single-centre experience. This systematic review assembled what evidence exists: 16 non-randomised comparative studies of critically ill adults, 12 of which entered the meta-analysis. There are no randomised trials.
Intubation was not significantly different between phenobarbital-based and benzodiazepine-based pathways (11 studies, odds ratio 0.62, 95% CI 0.20 to 1.87), with substantial heterogeneity at I-squared 73%. Hospital length of stay was also no different (5 studies, mean difference -1.75 days, 95% CI -5.07 to 1.56, I-squared 63%). Intensive care length of stay was shorter with phenobarbital by 0.60 days (95% CI -0.79 to -0.41), and here heterogeneity was zero.
The more interesting observation is structural. In exploratory subgroup analysis, front-loaded protocolised phenobarbital-first pathways were associated with lower intubation and shorter hospital stay, whereas adjunctive or mixed use was not. The authors are explicit that certainty is very low and that this is hypothesis-generating.
That distinction is worth carrying regardless of the drug. A protocol that front-loads treatment, uses a defined escalation, and does not rely on repeated bedside judgement is a different intervention from the same drug given ad hoc. Much of what gets attributed to phenobarbital may be attributable to having a protocol at all.
Practically, this does not justify switching a working benzodiazepine pathway. It does justify examining whether your existing pathway is protocolised and front-loaded, which is the modifiable part.
- Do not switch an established benzodiazepine pathway to phenobarbital on this evidence — all 16 studies were non-randomised
- The consistent finding is a 0.6 day shorter intensive care stay, which is real but small
- Examine whether your withdrawal pathway is protocolised and front-loaded; that structure may matter more than the agent
- Treat the subgroup finding on phenobarbital-first pathways as hypothesis-generating, as the authors do
- Watch for the patient in whom withdrawal is not the only diagnosis — sepsis, head injury and Wernicke's all mimic it
The statistics, in plain English
Two numbers show why the intensive care stay result is the only one worth acting on. The intubation odds ratio has an interval from 0.20 to 1.87 and I-squared of 73%, meaning the included studies disagreed with each other substantially and the pooled estimate is close to meaningless. The intensive care stay result has I-squared of 0%, meaning the studies agreed, and a tight interval. I-squared measures how much of the variation between studies exceeds what chance would produce; above about 50% the pooled figure should be treated as a description of disagreement rather than an estimate. All of this rests on non-randomised data assessed with ROBINS-I, so confounding by indication remains the dominant concern.
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