REDUCE Haloperidol

Effect of Haloperidol on Survival Among Critically Ill Adults With a High Risk of Delirium

van den Boogaard. JAMA 2018; 319:680–690 doi:10.1001/jama.2018.0160

Clinical Question

  • In critically ill adults, does haloperidol 1 mg or haloperidol 2 mg given three times a day compared to placebo reduce mortality?


  • Delirium, which is acute confusion, inattention and fluctuating consciousness, represents neurological dysfunction and is common in critically ill patients, with up to 80% of ventilated patients suffering from delirium
  • Although several risk factors have been identified, and validated assessment tools have been produced, there are no pharmacological agents with clear benefit in preventing or treating delirium
  • Haloperidol is widely used to treat delirium, but lacks conclusive evidence to support this
    • The HOPE-ICU trial did not demonstrate a difference in coma-free and delirium-free days in critically ill adults with delirium


  • Randomised, controlled trial
  • Multi-centre study
  • Random allocation in 1:1:1 ratio across three groups
  • Only the trial pharmacist and safety management board were aware of allocation sequence, so that patients and clinical staff were blinded to group allocation
  • Consent was sought immediately on admission to critical care
    • Deferred consent was obtained from those unable to prospectively give consent
  • Sample size calculation based upon before-after observational study
    • Assumed effect size would produce Hazard Ratio of 0.85 (15% relative reduction in death over time)
    • Defined Power 0.8 and α error 0.05
    • Required 715 patients per group
  • Analysis was modified intention-to-treat (those that withdrew consent were not included)
    • Subgroup and per-protocol sensitivity analyses also conducted
  • Independent data and safety management board reviewed results at four points with defined thresholds for stopping – either for futility or superiority
    • The 1 mg Haloperidol group was stopped at the fourth review on grounds of futility – the board were blinded to allocation and unaware that this was 1 mg Haloperidol group


  • 21 centres across the Netherlands, including university and non-teaching hospitals
  • July 2013 to December 2016


  • Inclusion: Age over 18 years without delirium and admitted to trial Intensive Care Unit with anticipated stay of at least 2 days
  • Exclusion: Delirium prior to inclusion, dementia, Parkinson disease, alcohol abuse, acute neurological disease, use of antipsychotics or history of psychiatric disease, prolonged QTc or “clinically relevant ventricular arrhythmia”, death expected within 2 days, pregnancy, intolerance to haloperidol, inability to consent
  • 15,882 screened, 1,796 were randomised and 1,789 were included in the primary analysis
  • Baseline characteristics between groups were similar (1 mg vs 2 mg vs placebo):
    • Mean age: 66 vs 67 vs 67 years
    • Surgical admission: 47% vs 46% vs 46%
    • Emergency admission: 81% vs 82% vs 81%
    • Mechanical ventilation: 71% vs 68% vs 65%
    • APACHE 2 score: 20 vs 19 vs 19
    • Sepsis diagnosis: 30% vs 37% vs 33%
    • PRE-DELIRIC score: 18 vs 19 vs 19


  • Haloperidol 1 mg
    • 1 mg Haloperidol administered three times daily
    • Reduced to 0.5 mg for patients over 80 years, or less than 50 kg, or had liver failure (bilirubin > 50 µmol/L)
  • Haloperidol 2 mg
    • 2 mg Haloperidol administered three times daily
    • Reduced to 1 mg for patients over 80 years, or less than 50 kg, or had liver failure (bilirubin > 50 µmol/L)


  • Placebo
    • 0.9% sodium chloride administered three times daily

Management common to both groups

  • Non-pharmacological methods to prevent delirium were part of daily ICU care, and implementation was nearly 90%
    • Early mobilisation
    • Efforts to improve circadian rhythm
    • Noise reduction
    • Reduced sedation and avoidance of benzodiazepines
    • Hearing and visual aids
  • Doses for all three groups appeared identical and were given 1 mL volume (or 0.5 mL for reduced doses as described above)
  • Initial dose given as soon as possible and at least within 24 hours
  • Study drug given until day 28, or until ICU discharge, or until delirium occurred
  • CAM-ICU was used to identify patients suffering from delirium
  • Upon delirium occurring, the study medication was stopped and open-label haloperidol was recommended
  • A standardised regime of increasing-decreasing haloperidol was given for hyperactive (agitated) delirium
    • Up to 5 mg every 8 hours
    • Reduced by 50% on each day after resolution of delirium and stopped on day three if they remained free from delirium
    • Rescue medication could include midazolam, clonidine, propofol or dexmedetomidine
    • Haloperidol could be administered intravenous, intramuscular or oral
  • Study medication was not restarted if delirium resolved or if the patient was readmitted to the ICU


  • Only the 2 mg Haloperidol group was compared against Placebo group, as the 1 mg Haloperidol group had been stopped early on grounds of futility
  • Primary outcome: There was no difference in survival rates between the two groups
    • Median time until death censored at 28 days:
      • 2 mg Haloperidol: 28 days
      • Placebo: 28 days
      • Hazard Ratio: 1.003 (95% CI 0.78 to 1.30)
    • 28-day mortality:
      • 2 mg Haloperidol: 16.67%
      • Placebo: 17.26%
      • Absolute Risk Reduction: 0.59% (95% CI -3.29% to 4.47%; P-value 0.7790)
  • Secondary outcome:
    • Survival to 90 days did not differ between groups
    • Incidence of delirium was not different
      • 2 mg Haloperidol: 33.3%
      • Placebo: 33.0%
    • Duration of open-label haloperidol to treat delirium was not different
    • Dose of open-label haloperidol required to treat delirium was not different
    • No differences were identified in other outcome measures including duration of ventilation, unintended extubation, readmission rates or lengths of stay
    • No differences became apparent in per-protocol analysis
    • Adverse events did not differ between groups
    • No interactions with any predefined subgroups were identified

Authors’ Conclusions

  • Prophylactic haloperidol therapy did not improve survival in critically ill patients at high risk of developing delirium


  • Appropriate concealment of allocation sequence, randomisation strategy and blinding of group allocation, which strengthen the validity of these results
  • Multi-centre design improves validity and generalisability to the wider population
  • Appropriate safety and futility checks put in place with early termination of 1 mg Haloperidol group
    • It is unethical to continue research on humans in the knowledge that either the data will never be sufficient to answer the relevant hypothesis or that the intervention is harmful
  • Appropriate intention-to-treat and sensitivity analysis of data
  • Time-to-event analysis offers greater statistical power to detect a significant difference compared to an analysis of incidence of events at a fixed time point


  • I doubt the biological plausibility that a small dose of prophylactic haloperidol could modify survival at 28 days
  • The power calculation assumed an effect size of 15% relative difference, which would also seem a biologically implausible effect size, even though it was based on a suitable observational study
  • Median survival at 28 days is not a useful statistical description in this trial
    • The median describes the outcome of the middle patient, if they were lined up in numerical order of their outcome
    • Censoring means that, in this case, if patients lived to 28 days and possibly beyond, their survival outcome was recorded as “28”
    • Since more than 50% lived at least 28 days, the middle patient’s result was therefore “28”, as described by the median survival outcome in this trial
    • The survival rate was always likely to be more than 50% in this trial (corresponding to a mortality rate of less than 50%), so the median survival to day 28 was always going to be “28”
    • However, the Hazard Ratio describes the approximate rate of death over time and is an appropriate and useful description
  • The study sample only included Dutch hospitals, which limits the extrapolation to non-European populations
  • No differentiation was made between hypo-active and hyper-active delirium, so conclusions about these subgroups cannot be made
  • Data on delirium outcomes were not collected from 7 centres due to lack of research staff

The Bottom Line

  • This was a methodologically sound and well conducted trial, which was based upon a hypothesis that lacked biological plausibility
  • The trial results confirm that Haloperidol 1 mg or 2 mg three times a day given prophylactically in critically ill adults will not prevent delirium, reduce delirium duration or modify mortality risk

External Links


Summary author: Duncan Chambler
Summary date: 20 June 2019
Peer-review editor: Adrian Wong

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.