Haloperidol in ICU delirium: probable benefit in both subtypes, possibly more in hyperactive

Design

Post hoc Bayesian analysis of a randomised placebo-controlled trial

Population

987 adult ICU patients with delirium (540 hypoactive, 447 hyperactive)

Primary outcome

Days alive and out of hospital at 90 days, by motor subtype

Effect

Benefit in both subtypes; 76.5% probability of larger effect in hyperactive (difference 2.9 days, CrI −4.8 to 10.6)

AID-ICU randomised adult intensive care patients with delirium to haloperidol or placebo. This post hoc Bayesian analysis of 987 participants looked at whether the effect differed between hypoactive (55%) and hyperactive (45%) delirium.

Haloperidol was associated with more days alive and out of hospital, and with lower mortality, in both subtypes. The probability of a larger effect in hyperactive delirium was 76.5% for days out of hospital (difference 2.9 days, 95% credible interval −4.8 to 10.6) and 73% for mortality. Serious adverse reactions and rescue medication use did not differ by subtype.

This analysis does not support withholding haloperidol simply because delirium is hypoactive, and does not prove a difference between subtypes either: the credible intervals are wide. It is a reason not to dismiss haloperidol in delirious ICU patients, alongside the non-drug measures that remain the foundation.

• Treat reversible causes of delirium first: hypoxia, hypoglycaemia, sepsis, retention, pain and drugs.
• Screen for hypoactive delirium actively; it is easily missed without a tool such as CAM-ICU.
• Consider haloperidol in ICU delirium of either subtype, with ECG monitoring of the QT interval.
• Keep non-drug measures — sleep, orientation, mobilisation, family presence — as the foundation.

Why it matters

It challenges the idea that haloperidol only helps the agitated patient.

Don't overread it

This is a post hoc subgroup analysis; the differences between subtypes are uncertain.

The statistics, in plain English

Bayesian analysis gives probabilities rather than p-values. A 76.5% probability of a larger effect in hyperactive delirium means it is more likely than not, but far from certain; the credible interval from −4.8 to 10.6 days includes no difference.

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.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.