Antipsychotics for delirium in palliative care: new RCT suggests non-drug alternatives are needed


Delirium is a common and serious medical condition (Oliver, 2016) that often occurs near the end of life, such as in palliative care settings. Research shows that 13-42% of people admitted to palliative care, and as much as 88% of people in their last days/hours experience delirium (Hosker & Bennett, 2016).

Current guidance focusses on prevention, eliminating reversible causes of delirium such as causative illnesses or drug side-effects. It suggests the use of antipsychotics such as olanzapine or haloperidol to treat the symptoms of delirium, but this is based on studies of delirium outside of palliative care (NICE, 2010).

A recent meta-analysis of antipsychotics for delirium (Kishi et al, 2016) included trials comparing antipsychotics to usual care and placebo amongst a variety of settings, not specifically palliative care. There was similar size in effectiveness of antipsychotics compared to both placebo and usual care for response rates at study endpoint (with a very good Number Needed to Treat of 2) but only 1 placebo controlled study. There was no difference between antipsychotics and placebo on severity of delirium in the 1 study that measured this; there was a large effect size advantage for antipsychotics compared to usual care, but the placebo comparison is more relevant. There was a non-significant advantage for haloperidol compared to newer antipsychotics for rate of response, but a significant advantage for newer antipsychotics to reduce time to response. Antipsychotics were associated with a higher rate of side-effects.

This paucity of evidence is interesting and highlights one of the pitfalls of evidence-based health care. The clinical experience of many doctors is that antipsychotics can be helpful in cases of delirium that they see on the wards. Yet there is an absence of good quality evidence for this common problem, possibly because of the difficulties of carrying out a high quality study in a group of patients with multiple underlying medical causes that can make creating comparative patient groups difficult.

The paper that is the subject of this blog (Agar et al, 2016) reports on a high quality double blind randomised controlled trial comparing haloperidol vs risperidone vs placebo for delirium in patients with life-limiting illness (88.3% cancer) in a hospice or hospital palliative care unit in Australia. By focussing on a particular patient group in a particular setting, the researchers are able to carry out a high-quality trial, although we must accept the downside that the results can only be applied to similar patients.

There is a paucity of evidence about the use of antipsychotics for delirium in patients at the end of life.

There is a paucity of evidence about the use of antipsychotics for delirium in palliative care patients.


The multi-site trial took almost 6 years to complete. Reasonable inclusion criteria were used:

  • Being able to speak English and swallow fluids
  • The presence of advanced incurable illness requiring inpatient care from palliative care team
  • DSM-IV Text Revision diagnosis of delirium
  • Memorial Delirium Assessment Scale (MDAS) score of 7 or more
  • The presence of the 3 target symptoms of delirium assumed to be associated with distress : inappropriate communication, inappropriate behaviour and illusions/ hallucination (these are thought to be the symptoms most responsive to antipsychotic treatment) measured using the Nursing Delirium Screening Scale (NUDesc scale)

Reasonable exclusion criteria were also used such as delirium not caused by substance withdrawal, history of adverse reactions to antipsychotics, expected lifespan less than 7 days.

Procedures were put in place to ensure randomisation and blinding of raters to treatment; dosage was based on previous trials; dosage was increased if active symptoms on the NUDesc were noted. Standardised procedures to use midazolam (a benzodiazepine) if NuDesc scores were high and benzotropine if (usually antipsychotic-induced) extrapyramidal side effects were used.

The primary outcome was the delirium symptom score on day 3 taking into account the baseline score. Given the absence of established response criteria for delirium response the study authors chose a 1-point difference on the NUDesc as a minimally clinically important change.

Secondary outcomes were survival, daily MDAS score, lowest delirium symptom score, daily use of midazolam, extrapyramidal symptoms and sedation. These were analysed on an intent-to-treat basis. Secondary mixed-model analysis was performed controlling for important variables such as medical comorbidity and prior cognitive impairment.

The study includes a CONSORT diagram illustrating what happened to potential participants: 1,819 were referred to study, 306 were assessed, 82 randomised to risperidone/81 randomised to haloperidol/86 randomised to placebo. These randomised patients were included in appropriate statistical tests. The study recruited according to plan.


The three groups (risperidone, haloperidol and placebo) were comparable on demographic and clinical variables.

For the primary outcome:

  • Placebo patients had significantly fewer NUDesc symptom scores (delirium distress symptoms) than either risperidone or haloperidol. This was confirmed on secondary mixed-model analysis taking into demographic and clinical variables.
  • Placebo group also required fewer doses of midazolam than the antipsychotic groups.

For the secondary outcomes:

  • 34 participants died during the study period (9 in the placebo group, 9 in the haloperidol group, and 16 in the risperidone group):
    • Median survival was longer for placebo than antipsychotic groups
    • In a post hoc analysis, those receiving an antipsychotic drug were approximately 1.5 times more likely to die than those receiving placebo (95% confidence interval 0.18 to 2.01, p=0.01, i.e. statistically significant but the confidence intervals included no effect)
  • Placebo had significantly lower MDAS delirium score than risperidone. Placebo had lower MDAS scores but not significantly so than haloperidol. No significant differences for lowest delirium score
  • Significantly greater number of extrapyramidal side-effects in antipsychotic groups than placebo
  • No significant difference in sedation between risperidone and placebo, but haloperidol significantly more sedative than placebo
This evidence suggests that people in palliative inpatient settings with life-limiting illness and delirium should be managed by other means other than antipsychotics.

This evidence suggests that people in palliative inpatient settings with life-limiting illness and delirium should be managed by means other than antipsychotics.

Strengths and limitations


  • Double blind randomised controlled trial with clear protocols
  • Use of standardised outcome measures (even if there is a lack of agreement as to their clinical significance)
  • Use of statistical methods to control for effects of important clinical variables that can effect outcomes


  • Only used oral meds so limited participants
  • 6 year multi-site study so hard to replicate
  • Results can only be said to apply to limited-life patients in palliative care settings

Implications for practice

The results seem conclusive: placebo had better delirium outcomes and reduced side-effects than antipsychotics for delirium encountered in palliative care inpatient settings. Survival seemed better on placebo but the post-hoc analysis confidence intervals included no effect.

Currently this means people in palliative inpatient settings with life-limiting illness and delirium should be managed by other means such as reversing identifiable causes and supportive interventions, though these are resource-intensive and may not be provided (Maust and Kales, 2016).

It also highlights the importance of carrying out trials of antipsychotics for delirium in general medical/surgical wards. This will clearly be difficult due to the complicating effects of a wide range of comorbid conditions and premorbid health states, but delirium is common in hospitals so this type of information is vital.

The evidence of doctors’ experience is that antipsychotics are helpful in delirium, but it is important to test if this experience has led to an accurate conclusion or a mirage that is at best exposing people to side-effects or at worst worsening their survival prospects.

What training and resources are required to deliver environmental or biopsychosocial interventions in place of a prescription?

What training and resources are required to deliver environmental or biopsychosocial interventions in place of a prescription?


Primary paper

Agar MR, Lawlor PG, Quinn S, Draper B, Caplan GA, Rowett D, Sanderson C, Hardy J, Le B, Eckermann S, McCaffrey N, Devilee L, Fazekas B, Hill M, Currow DC. (2016) Efficacy of Oral Risperidone, Haloperidol, or Placebo for Symptoms of Delirium Among Patients in Palliative Care: A Randomized Clinical TrialJAMA Intern Med. Published online December 05, 2016. doi:10.1001/jamainternmed.2016.7491

Other references

Maust DT, Kales HC. (2016) Medicating Distress. JAMA Intern Med. 2017;177(1):42-43. doi:10.1001/jamainternmed.2016.7528

Hosker CMG, Bennett MI. (2016) Delirium and agitation at the end of life.

Kishi T, Hirota T, Matsunaga S, Iwata N. (2016) Antipsychotic medications for the treatment of delirium: a systematic review and meta-analysis of randomised controlled trials. J Neurol Neurosurg Psychiatry. 2016 Jul;87(7):767-74. doi: 10.1136/jnnp-2015-311049. Epub 2015 Sep 4. [Abstract]

National Institute for Health and Care Excellence. Delirium: prevention, diagnosis and management. NICE guideline 103. (2010)

Oliver D. (2016) David Oliver: Delirium matters.

Photo credits

Share on Facebook Tweet this on Twitter Share on LinkedIn Share on Google+
Mark as read
Create a personal elf note about this blog
Profile photo of Samei Huda

Samei Huda

Dr Samei Huda has been working as a Consultant Psychiatrist for over 20years including a year in Australia. He also completed a Diploma in Cognitive Therapy (though he doesn’t see patients for therapy) and a Master of Science in Psychiatry. He has been working with an Early Intervention in Psychosis team since 2007. Over his career he has directly assessed, managed and treated hundreds of patients with psychosis of different types. He recommends people with an interest in psychosis to read (he is not affiliated with them). He has written a book evaluating the evidence for diagnosis and treatment in psychiatry called “The Medical Model in Mental Health” published by Oxford University Press. He has recently completed an MA in Philosophy of Mental Health from UCLAN.

More posts - Website

Follow me here –