Haloperidol and Ziprasidone for Treatment of Delirium in Critical Illness

Girard et al. NEJM 2018. Published on line first Oct 22nd DOI:10.1056/NEJMoa1808217

Clinical Question

  • In delirious adult ICU patients, does the administration of haloperidol or ziprasidone reduce the duration of delirium when compared with placebo?


  • Delirium is the most common manifestation of acute brain dysfunction during critical illness, affecting 50 to 75% of patients who receive mechanical ventilation in an intensive care unit (ICU)
  • It is estimated that around 85% of delirious ICU patients have hypoactive delirium
  • Patients with delirium have higher mortality, longer periods of mechanical ventilation and hospital stays, higher risk of long-term cognitive impairment. Delirium also interferes with medical care
  • Delirium is a difficult condition to treat (and prevent) and the evidence base is mixed in regard to the use of antipsychotic agents. Options include typical (haloperidol) and atypical (olanzapine, quetiapine, risperidone) antipsychotics
  • Trials using haloperidol and atypical antipsychotic medications have failed to show a clear benefit with their use in the management of delirium. In addition, haloperidol has not been shown to be useful in the prevention of delirium in ICU patients HOPE-ICU study
  • Haloperidol and ziprasidone are dopamine D2 antagonists. Both drugs have activity at other receptors, including antagonism at 5-hydroxytryptamine2 (5-HT2) receptors
  • Haloperidol remains the most commonly used medication for delirium management in the UK despite a limited evidence of benefit


  • A randomised, double-blind placebo-controlled phase 3 trial
  • Delirium was determined by the use of the Confusion Assessment Method for the ICU (CAM-ICU), which rates four features of delirium to determine a binary outcome of whether delirium is present or absent
  • Patients were assessed for the development of delirium for five days, or until death or ICU discharge
  • Patients were randomised to placebo, haloperidol or ziprasidone in a 1:1:1 ratio by using blocked randomisation with patients stratified by trial site
  • The study was powered at 80% to detect a 2-day difference between groups in days alive without delirium or coma


  • 16 intensive care units in the USA
  • September 2010-March 2018


  • Inclusion: Adult ICU patients (over 18 years of age), who had received either invasive or non-invasive ventilation, vasopressors or mechanical cardiac support
  • Exclusion: Patients with severe cognitive impairment at baseline. Patients at high-risk for medication side effects because of pregnancy, breast-feeding, a history of torsades de pointes, QT prolongation, a history of neuroleptic malignant syndrome, or allergy to the study medication
  • 20,914 patients screened. 16,306 met one or more exclusion criteria, leaving 4608 patients
    • 3,425 (74%) out of the remaining 4608 patients refused consent
    • 1183 (26%) consented to be assessed further
    • 566 (48%) patients had delirium and met the criteria for randomisation
  • Baseline characteristics were similar between the three groups
    • APACHE II scores in the three groups were between 28 and 30
    • 528 patients (93.2) received invasive mechanical ventilation prior to randomisation
    • 505 patients (89.2%) had hypoactive delirium at randomisation


  • Haloperidol administered intravenously
    • 2.5mg in patients younger than 70 years of age
    • 1.25mg in patients older than 70
    • maximum of 10mg per administration and 20 mg per day


  • Ziprasidone administered intravenously
    • 5mg in patients younger than 70 years of age
    • 2.5mg in patients older than 70
    • maximum of 20mg per administration and 40 mg per day


  • An intravenous placebo was administered of the same volume as the intervention drug

Management common to both groups

  • All trial medications were colourless and were administered in identical bags
  • All medications were administered every 12 hours, at 10:00 and 22:00 hours
  • Dosages were doubled if the patient still had delirium
  • The volume and dose of a trial drug was halved if the patient did not have delirium on two consecutive CAM-ICU assessments


  • Primary outcome: No statistical difference in the adjusted median number of days alive and without coma was:
    • 8.5 (95% CI 5.6 to 9.9) in the placebo group
    • 7.9 (95% CI 4.4 to 9.6) in the haloperidol group
    • 8.7 (95% CI 5.9 to 10.0) in the ziprasidone group
    • p value across trial groups was 0.26
  • Secondary outcome: No statistical difference in:
    • 30-day and 90-day mortality
    • time free from ventilation
    • ICU discharge
    • ICU readmission
    • hospital discharge
    • death at 30 days: 27% in the placebo group, 26% in the haloperidol group and 28% in the ziprasidone group
  • Safety: Prolongation of the corrected QT interval was more common in the ziprasidone group than in the haloperidol group or placebo group. Torsades de pointes developed in two patients in the haloperidol group during the intervention period but this was several days after the trial drug was stopped

Authors’ Conclusions

  • In this large, double-blind, randomised, placebo-controlled trial, we found no evidence that the use of haloperidol or ziprasidone had an effect on the duration of delirium among patients with acute respiratory or shock in the ICU


  • The double-blind study design, with matched placebo, minimised bias
  • The approach to ensuring that the placebo process was carried out appeared to be meticulous
  • An intention to treat analysis was used
  • Adjustment was made for potentials confounders
  • A power calculation was performed to detect a two-day difference in days free from delirium
  • This was a multi-centre study, incorporating 16 ICUs, improving external validity
  • Two antipsychotic medications were studied, one typical (haloperidol) and a newer agent (ziprasidone)
  • Wide inclusion criteria, and the use of a validated assessment tool to screen patients, improved the generalisability of the study


  • A large number of patients, or their representatives, refused consent. As a result, selection bias was likely to be an issue in the randomised cohort
  • The incidence of delirium, at 48%, was lower than predicted in the study design. This may have meant that the study was underpowered to detect a 2-day difference between groups in days alive without delirium
  • The vast majority of patients had hypoactive, rather than hyperactive, delirium. Using antipsychotics routinely to treat hypoactive delirium is not standard practice in most ICUs
  • Generalisability to a European or Australasian setting is limited in view of the differences in ICU practices between the USA and elsewhere
  • Twice daily CAM-ICU assessments is unlikely to represent usual care in most units. Despite this screening process, it took seven years to recruit the pre-specified number of patients
  • The primary outcome of days alive and without delirium is a composite quantitative variable, with binary components (dead or alive, coma or no coma) and a continuous component (days without delirium)

The Bottom Line

  • The administration of intravenous dopamine antagonists failed to improve the duration of delirium in this study
  • I rarely treated hypoactive delirium with antipsychotic medication and my practice will therefore not be influenced by this paper
  • I predominantly use quetiapine in the management of hyperactive delirium, which has a different mechanism of action to haloperidol (with an alpha agonist component), and I will continue with this practice until further evidence becomes available
  • The underlying pathophysiology of disordered mentation in the ICU is complex and probably involves multiple neurotransmitters in addition to alterations to the structural integrity and neural pathways of the brain

External Links


Summary author: Fraser Magee
Summary date: 16/11/18
Peer-review editor: Steve Mathieu

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.