- Design
- Multicentre target-trial emulation using hospital records
- Population
- 24,643 adults admitted with alcohol withdrawal at 754 US hospitals
- Primary outcome
- Length of stay; invasive mechanical ventilation
- Effect
- 4.1 vs 4.3 days (difference -0.21, 95% CI -0.42 to 0.00); ventilation 3.1% vs 2.7%
A comparative effectiveness study in JAMA Internal Medicine, published on 12 October 2026, emulated a trial in records from 754 US hospitals. It compared 24,643 adults admitted with alcohol withdrawal who started phenobarbital (11,919) or diazepam (12,724) on the day of admission. Median age was 45 and three-quarters were men.
There was no difference in length of stay (4.1 against 4.3 days; difference -0.21 days, 95% CI -0.42 to 0.00) or in invasive mechanical ventilation (3.1% against 2.7%; difference 0.4%, 95% CI -0.7 to 1.4). Escalation and complications of severe withdrawal did not differ. Patients started on phenobarbital left against advice less often (9.5% against 12.6%).
Phenobarbital has spread on the strength of comparisons with lorazepam. Against diazepam, a long-acting, self-tapering benzodiazepine, it looks broadly equivalent. That makes the choice one of local protocol, nursing familiarity and monitoring, rather than evidence of superiority. Observational data cannot remove all confounding, so a randomised trial would still be useful.
- Use whichever of phenobarbital or diazepam your unit has a protocol and monitoring for; neither showed an advantage.
- Prefer a long-acting agent with a defined dosing protocol over ad hoc short-acting doses.
- Monitor sedation and respiratory status closely with either drug, especially with other sedatives on board.
- Address the risk of self-discharge early with a withdrawal plan the patient understands.
Why it matters
It weakens the case for switching wards to phenobarbital as a better drug, rather than one of two acceptable choices.
Don't overread it
Observational target-trial emulation, not a randomised trial — residual confounding is possible.
The statistics, in plain English
The length-of-stay interval runs from -0.42 to 0.00 days, so any true difference is at most half a day. The ventilation interval crosses zero. Together these point to near-equivalence for the main outcomes.
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 internal medicine, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free