Published on 18 September 2026
Quetiapine or haloperidol for delirium: are 215 patients enough to choose?
The essentials
This meta-analysis pools three randomized trials and 215 participants, 105 on quetiapine and 110 on haloperidol, with a mean age of 57.25 years. On delirium severity, the primary outcome, the mean difference is −0.80 (95% CI: −2.05 to 0.44; I² = 10%). On mortality, the risk ratio is 0.60 (95% CI: 0.29 to 1.27; I² = 0%). On sleep time, the mean difference is 1.59 (95% CI: −0.45 to 3.63; I² = 77%). On response rate, the risk ratio is 0.89 (95% CI: 0.51 to 1.56; I² = 85%). None of these results reaches statistical significance. The authors report “no significant difference in overall effect of quetiapine and haloperidol in delirium management” and call for further research, citing the small number of trials, the limited sample size and high heterogeneity. This absence of difference is not a demonstration of equivalence. With 215 patients, the analysis lacks the power to rule out either a clinically important advantage or a clinically important disadvantage. A later meta-analysis on the same question, covering four trials and 292 patients, ran a trial sequential analysis: the required information size was not reached and the data were judged inconclusive, with very low GRADE certainty for every outcome.
Context
Start with the regulatory frame, because it shapes the whole debate. The two drugs compared here do not have the same status in this indication. In the United Kingdom, the public assessment report of the Medicines and Healthcare products Regulatory Agency (MHRA) describes the authorized indication of haloperidol as the acute treatment of delirium in adults after non-pharmacological treatments have failed, and notes that newer antipsychotics are not authorized for acute delirium. The same indication appears in other European labels, including the French one. Authorization status varies from country to country and should be checked locally, but the pattern described by the MHRA is common: haloperidol has an indication that covers delirium, quetiapine does not.
This has a direct consequence. The everyday discussion in consultation-liaison psychiatry, geriatrics and intensive care takes place between a drug used within its authorization and a drug used off-label. The indication itself is also conditional: it only applies once non-drug measures have failed, which puts the pharmacological question in second place by design.
The question addressed by this meta-analysis is secondary in another sense too. What has been established for several years is that no antipsychotic has been shown to shorten delirium or reduce mortality compared with placebo. The systematic review by Nikooie and colleagues, published in Annals of Internal Medicine in 2019, concludes that the available evidence does not support routine use of haloperidol or second-generation antipsychotics to treat delirium in hospitalized adults. The MIND-USA trial, 566 intensive care patients, found no effect of haloperidol or ziprasidone on the number of days alive without delirium or coma. The Society of Critical Care Medicine guidelines suggested in 2018 not routinely using haloperidol, an atypical antipsychotic or a statin to treat delirium, a conditional recommendation based on low-quality evidence. Their 2025 focused update does not repeat that wording: it states that no recommendation can be made, for or against, on the use of antipsychotics versus usual care for treating delirium in critically ill adults, with low certainty. The absence of a recommendation is not a green light: it records that the data do not allow a decision.
First-line management is etiological and environmental. The UK NICE guideline asks clinicians to identify and treat the underlying cause or combination of causes, then, if the person is distressed, to use verbal and non-verbal de-escalation techniques first, with short-term haloperidol, usually for one week or less, coming only afterwards. National guidance elsewhere says the same thing. The French Haute Autorité de santé (HAS), in its May 2009 recommendation on acute confusion in older people, states that management requires above all treating the precipitating factors, and that non-drug measures alone are sufficient in a large number of cases.
Comparing two drugs head to head when neither has shown benefit over placebo amounts to ranking two tools whose usefulness is not established. The result still matters, because these prescriptions are written every day, but it does not change the order of priorities.
The study at a glance
| Question (PICO) | |
|---|---|
| Population | |
| Hospitalized adults with delirium. 3 randomized trials, 215 participants, mean age 57.25 years (SD 13.09), 57.67% men. The care setting is not specified in the published abstract, which only says “hospitalized adults” | |
| Intervention | |
| Quetiapine monotherapy. Doses not reported in the abstract | |
| Comparator | |
| Haloperidol monotherapy. No placebo arm. Doses not reported in the abstract | |
| Outcomes | |
| Primary: delirium severity. Secondary: mortality, sleep time, response rate. The severity scale is not named in the abstract | |
| Design | |
| Systematic review with meta-analysis of randomized trials, conducted according to PRISMA. Searches in PubMed, Cochrane, Embase and Google Scholar, search period not reported in the abstract. Mean follow-up 5.75 days (SD 1.89). Random-effects model, RevMan software · CEBM 1a · Allocation: 105 quetiapine, 110 haloperidol | |
| Not reported in the abstract | |
| Protocol registration, risk of bias tool, GRADE assessment, names of the included trials, doses and treatment duration, funding, conflicts of interest |
Quality control
| Criterion | Status |
|---|---|
| Conduct framework | Solid |
| FindingReported as conducted according to PRISMA, four sources searched, random-effects model consistent with the expected heterogeneity, heterogeneity reported for each outcome | |
| Protocol registration | Caveat |
| FindingNo registration number appears in the abstract. Not mentioning a registration does not mean there was none, but readers cannot check that the outcomes were set before the analysis | |
| Risk of bias and grading | Caveat |
| FindingNeither a risk of bias tool nor a GRADE assessment is mentioned in the abstract | |
| Size of the evidence base | Critical |
| FindingThree trials, 215 participants. That is the sample size of a single medium-sized trial. Power is insufficient to detect even a sizeable difference | |
| Heterogeneity of secondary outcomes | Critical |
| FindingI² = 77% for sleep and 85% for response rate. At this level, pooling produces an average figure that describes none of the studies combined | |
| Publication bias | Caveat |
| FindingWith three studies, no test of funnel asymmetry can be interpreted. The risk is not ruled out, it simply cannot be assessed | |
| Interpretability of the primary outcome | Caveat |
| FindingThe abstract does not name the severity scale. A mean difference of 0.80 points cannot be interpreted without knowing which scale it was measured on | |
| Independence | Caveat |
| FindingNeither funding nor a conflict of interest statement appears in the abstract consulted. No funder is indexed, which does not mean there was none | |
| Authorship | Caveat |
| FindingTwenty-five authors sign a synthesis of three trials. The ratio raises questions about how the work was actually shared, although no conclusion about the quality of the content can be drawn from it | |
Results
| Outcome | Published result |
|---|---|
| Delirium severity | Mean difference −0.80 (95% CI: −2.05 to 0.44); I² = 10% |
| PEB readingNo difference detected, scale not specified heterogeneity is low, which is the strong point of this result. The interval covers both a 2.05-point advantage for quetiapine and a 0.44-point disadvantage, on a scale the abstract does not name | |
| Mortality | Risk ratio 0.60 (95% CI: 0.29 to 1.27); I² = 0% |
| PEB readingUnusable interval the point estimate suggests a 40% reduction, but the interval runs from a 71% reduction to a 27% increase. With 215 patients, the number of events is necessarily very small. This result supports no conclusion on safety | |
| Sleep time | Mean difference 1.59 (95% CI: −0.45 to 3.63); I² = 77% |
| PEB readingMajor heterogeneity the unit of measurement is not specified in the abstract; the later meta-analysis by de Oliveira reports the same estimate expressed in hours. With an I² of 77%, the pooled value represents none of the included studies | |
| Response rate | Risk ratio 0.89 (95% CI: 0.51 to 1.56); I² = 85% |
| PEB readingMajor heterogeneity heterogeneity of 85% makes pooling hard to defend. The definition of response is not specified in the abstract | |
A note on the order of magnitude of the primary outcome, with a caveat that has to come first. The abstract names neither the scale nor the included trials, but it does give the exact size of each arm. Three published head-to-head trials add up to exactly these numbers: Maneeton 2013 in Thailand, 24 versus 28 patients; Grover 2016 in India, 31 versus 32; Zakhary 2024 in Egypt, 50 versus 50. That makes 105 and 110. This is not published information. No trial name appears in the abstract: this arithmetic correspondence is a reading hypothesis, not a verified identification, and it is not unambiguous. Other randomized trials compare these two drugs, including Garg 2022, conducted in intensive care with a third, placebo arm, and the meta-analysis published later on the same question retained four trials totaling 292 patients. What follows therefore only holds if the hypothesis is correct, and readers should not treat it as data from the source. These three trials use the Delirium Rating Scale Revised 98, a scale on which both arms of the Maneeton trial improved by about 22 points in seven days. A between-treatment gap of 0.80 points would then sit far below what either drug produces on the score itself.
Critical appraisal
| Domain | Judgment |
|---|---|
| Absence of difference versus equivalence | Critical |
| FindingThis is the main reading error on this topic. Finding no difference in 215 patients does not show that the two drugs are interchangeable. The meta-analysis published later by de Oliveira and colleagues in General Hospital Psychiatry, four trials and 292 patients, ran a trial sequential analysis: the required information size was not reached, GRADE certainty was very low for every outcome, and the authors concluded that the data remain inconclusive because of their scarcity and imprecision | |
| Population and generalizability | Critical |
| FindingMean age is 57.25 years (SD 13.09). This is neither a geriatric population nor a palliative care population. The care setting is not described in the abstract. Yet management differs in intensive care, acute geriatrics and end-of-life care, and the benefit-risk balance of antipsychotics worsens markedly with age and underlying dementia | |
| No placebo arm | Caveat |
| FindingThe comparison pits two treatments against each other, neither of which has shown, against placebo, a reduction in delirium duration or mortality. No difference between two unvalidated options validates either of them | |
| Doses and duration | Caveat |
| FindingDoses are not reported in the abstract, which does give a mean follow-up of 5.75 days (SD 1.89). In the published head-to-head trials, haloperidol doses are very low, in the range of 0.25 to 2 mg per day, and quetiapine doses range from 12.5 to 100 mg per day, over six to seven days. A result obtained at these doses and over this duration does not transfer to higher doses or to prolonged treatment | |
| Heterogeneity of secondary outcomes | Critical |
| FindingI² of 77% and 85%. The low heterogeneity of the primary outcome, 10%, contrasts with these values. It is also easier to obtain: three trials sharing the same scale and similar durations | |
| What the abstract does not say | Caveat |
| FindingProtocol registration, risk of bias, certainty grading, sensitivity analyses, subgroups, funding, conflicts of interest are not reported in the abstract | |
| Consistency with observational data | Solid |
| FindingThe largest cohort available on the subject, 17,115 older adults after major surgery, found in-hospital mortality of 3.7% on haloperidol, 3.3% on risperidone, 2.8% on olanzapine and 2.6% on quetiapine, with no significant difference, and none either for arrhythmias, pneumonia or stroke. The no-difference message is therefore consistent, at low to moderate oral doses | |
Three levels need to be kept apart. What is demonstrated: no antipsychotic has reduced delirium duration or mortality compared with placebo, and first-line management is etiological and non-pharmacological. What is suggested: at low doses and over a few days, the available trials show no signal of a difference between quetiapine and haloperidol on severity, which is not the same thing as a demonstrated similar effect. What is expert opinion: the choice between the two at the bedside, which today rests on the adverse effect profile, the patient’s background and regulatory status, for lack of an adequately powered comparison.
Regulatory status
The table below uses the French summaries of product characteristics (SmPC) as a documented example. Wording, indications and dates differ between countries: check the label that applies where you practice.
| Drug | Indication covering delirium | Warning carried by the summary of product characteristics (French SmPC) |
|---|---|---|
| Haloperidol (HALDOL) | Yes Acute treatment of delirium when non-pharmacological treatments have failed, in adults. The injectable form carries the same indication, with the intramuscular route recommended, an electrocardiogram before injection and continuous ECG monitoring if the intravenous route is used. The authorization therefore builds the prior failure of non-drug measures into its very wording. The MHRA describes the same indication in the United Kingdom | Section 4.4, under the heading “Increased mortality in elderly people with dementia”. Electrocardiogram recommended at initiation, monitoring for QTc prolongation and ventricular arrhythmias, discontinuation if QTc exceeds 500 ms. Text updated April 8, 2026 |
| Quetiapine (XEROQUEL LP, prolonged release) | No indications are limited to schizophrenia, manic and depressive episodes of bipolar disorder, prevention of recurrence in patients who have already responded, and adjunctive treatment of major depressive episodes. Neither delirium nor agitation in older adults | Quetiapine is not approved for the treatment of dementia-related psychosis. Increased mortality reported in ten-week trials in patients with dementia, 5.5% versus 3.2% on placebo, and an increased risk of cerebrovascular events. QT prolongation reported at therapeutic doses. Somnolence and sedation are common. Text updated April 22, 2025 |
A marketing authorization is neither a prescribing recommendation nor proof of efficacy. The MHRA public assessment report of December 2021, devoted specifically to haloperidol in older patients with acute delirium, finds that efficacy data in this population are limited, and restates that haloperidol should be reserved for cases where non-drug methods have failed, at the lowest dose and for the shortest possible duration. The same document notes that newer antipsychotics are not authorized for acute delirium.
Two points clarify the scope of these texts. The first concerns the excess mortality warning. It does not target delirium: it targets older adults with dementia treated with an antipsychotic. The haloperidol SmPC states that elderly patients with dementia-related psychosis treated with antipsychotics are at increased risk of death, while haloperidol keeps, in the French label, an indication for persistent aggression and psychotic symptoms in moderate to severe Alzheimer’s dementia or vascular dementia when non-pharmacological treatments have failed. The quetiapine SmPC states that the drug is not approved for the treatment of dementia-related psychosis. The warning therefore becomes relevant as soon as the confused patient also has dementia, a frequent situation in hospital, but it should not be read as a warning specific to the delirium indication.
The second concerns the framework for off-label prescribing. Prescribing quetiapine for delirium is off-label. Most jurisdictions accept off-label use under conditions, and these conditions usually include the absence of a suitable authorized alternative, a justification based on current scientific evidence, informing the patient and documenting the reasoning in the medical record. In France, for example, article L. 5121-12-1 of the Public Health Code only allows it when no appropriate authorized alternative exists and the prescriber considers it indispensable in light of established scientific data, with patient information, the mention “off-label prescription” on the prescription, information on reimbursement conditions and a documented rationale in the medical record. Since haloperidol has an indication that covers precisely this situation, the no-alternative condition is not met by default: it has to be argued case by case, on the basis of the patient’s profile, and not on the result of this meta-analysis, which demonstrates nothing in favor of quetiapine.
Level of evidence
PEB assessment: low confidence in the absence of a difference in severity, very low confidence for mortality, sleep and response rate, no confidence in the idea that the two drugs are equivalent. The work is honest in its conclusion, since the authors themselves point to the scarcity of trials, the small samples and the heterogeneity. What limits its reach is not the synthesis method but the material available: three trials, 215 patients, very low doses, a middle-aged population, an undescribed care setting. In practice, the regulatory question and the question of whether antipsychotics are indicated in delirium at all weigh more than the result of this comparison.
The colleague test
What an experienced colleague would say if you presented this study in two minutes, between two consultation-liaison calls.
“Three trials and 215 patients settle nothing, and above all they don’t say the two are the same. What matters first is finding the cause, not picking the drug. If I really have to prescribe, haloperidol has the indication and quetiapine doesn’t, at a low dose and for a few days. And in an 85-year-old with dementia, what both labels say is excess mortality, for dementia and not for delirium, but the patient has both.”
What this means in practice: when facing two treatments that no placebo-controlled trial has validated, a head-to-head comparison only moves the question elsewhere. The first step is still the search for a cause.
What you can do with this on Monday morning. The result does not change the prescription, but it gives concrete arguments for the conversation with the team and with the family.
- Go back to the etiological work-up before discussing the drug. Untreated pain, urinary retention, fecal impaction, hypoxia, sepsis, hyponatremia, alcohol or benzodiazepine withdrawal, anticholinergic drug burden, sensory deprivation. The NICE guideline makes identifying and treating the cause its first recommendation, and the French HAS writes that non-drug measures alone are sufficient in a large number of cases.
- Say clearly what the treatment will not do. No antipsychotic has been shown to reduce delirium duration or mortality compared with placebo. The aim is to control dangerous agitation or distress, nothing more. Telling the team this avoids dose escalation when the picture does not resolve.
- Choose knowing the framework. Haloperidol carries the indication for acute treatment of delirium when non-pharmacological treatments have failed, in the United Kingdom as in France. Quetiapine does not: prescribing it here is off-label, which regulatory frameworks generally accept only when no suitable authorized alternative exists, with patient information and a documented rationale (in France, also a mention on the prescription). The available comparative data do not establish that either drug is superior, and so do not remove the need for that justification.
- Set the duration and dose from the start. For older adults, the French HAS recommends starting at a quarter or half of the adult dose, preferring the oral or orodispersible route over injection, and limiting treatment to 24 to 48 hours. NICE speaks of one week or less. Writing the stop date in the initial prescription is the simplest way to avoid automatic renewal.
- Monitor what can be monitored. Electrocardiogram at initiation for haloperidol, with QTc and ventricular arrhythmia monitoring according to the SmPC, which requires stopping treatment above 500 ms. For quetiapine, QT prolongation reported at therapeutic doses, and sedation and somnolence usual in the first three days. Actively look for extrapyramidal signs and akathisia in both cases.
- Two situations where you do not prescribe the same way. Parkinson’s disease and dementia with Lewy bodies: NICE recommends particular caution, and the HAS warns that even a low dose of a neuroleptic can cause excessive sedation, or even coma. Older adults with dementia: both SmPCs carry an excess mortality warning, which targets this population and not delirium as such, and the quetiapine label adds an explicit statement that it is not approved for dementia-related psychosis.
- Take the reading grid with you. A meta-analysis of three trials and 215 participants mostly tells you what is missing. For any non-significant result, the useful question is what the confidence interval rules out, not what the point estimate suggests.
Frequently asked questions
Does this meta-analysis show that quetiapine and haloperidol are equivalent?
No. It shows that no difference was detected in 215 patients across three trials. That is a different piece of information. An equivalence study is designed in advance, with a predefined equivalence margin and a sample size calculated to exclude it. Nothing of the kind here. The meta-analysis published later on the same question, with a trial sequential analysis, concludes that the required information size has not been reached and that the data remain inconclusive.
Do these results apply to my 88-year-old patient in acute geriatrics, or to a patient in palliative care?
Nothing supports that. The mean age of participants is 57.25 years and the care setting is not described in the abstract. These are precisely the situations where the benefit-risk balance of antipsychotics is most debated. In older adults with dementia, both summaries of product characteristics carry an excess mortality warning. At the end of life, these texts carry no specific warning, but no efficacy data from this meta-analysis can be transferred to that setting either.
Is it permitted to prescribe either drug in this indication?
Haloperidol has an indication covering the acute treatment of delirium after non-pharmacological treatments have failed, in both oral and injectable forms, in the United Kingdom according to the MHRA and in France, among other countries. Quetiapine has no indication in this situation: prescribing it there is off-label. Authorization status varies between countries and should be checked locally. Off-label use is generally acceptable only when no suitable authorized alternative exists, a condition that is not a given here since haloperidol is indicated, and it then requires informing the patient and documenting the rationale in the record. In France, for example, the Public Health Code also requires the mention “off-label prescription” on the prescription. It engages the prescriber’s responsibility and may not be reimbursed on the same terms.
Should an antipsychotic be prescribed for delirium?
Not as a first-line measure. Guidelines agree on etiology and environmental measures. The Society of Critical Care Medicine guidelines suggested in 2018 not routinely using haloperidol or an atypical antipsychotic to treat delirium, a conditional recommendation based on low-quality evidence, and their 2025 update no longer makes a recommendation, for or against, for lack of sufficient evidence. Drug treatment is considered in the face of distress or risk, for the shortest possible duration.
Is the mortality result, a risk ratio of 0.60, reassuring for quetiapine?
No, and it should not be read that way. The confidence interval runs from 0.29 to 1.27, that is, from a 71% reduction to a 27% increase. With 215 patients, the number of deaths is necessarily small. This result allows neither a conclusion of benefit nor the exclusion of excess risk.
What should be taken from the reported heterogeneity?
It is low on the primary outcome, 10%, which makes the severity estimate relatively consistent across trials. It is major for sleep, 77%, and for response rate, 85%. At these levels, the pooled value properly describes none of the studies combined and should not be quoted on its own.
Annotated bibliography
Zainab S, Fatima E, Khalid MS, Tariq A, Zaheer L, Abbas K, Ahmad E, Sheikh AR, Zakir E, Arshad MZ, Sabir M, Zahra N, Kanwal A, Khan MR, Ali F, Razia IT, Khilji SQ, Ihtasham A, Khan AA, Imran H, Hussain D, Nasir A, Arain MS, Jawed H, Najam S (2025). Effect of quetiapine versus haloperidol on delirium severity in hospitalized adults: A systematic review and meta-analysis. Asian Journal of Psychiatry, 116, 104808. DOI 10.1016/j.ajp.2025.104808 · PMID 41483728. The source study analyzed here, published online on December 30, 2025. No funder indexed and no conflict of interest statement in the abstract consulted, which says nothing about whether they exist.
de Oliveira LGAM, Squizzato FAT, da Silva Matos HS, Maia AS, Lima-Araújo I, Cintra ACG, Amaral-Marques E, da Silva FPR, de Almeida ABB, Fagundes TP (2026). Quetiapine Versus Haloperidol for the Treatment of Delirium in Hospitalized Adults: A Meta-Analysis of Randomized Controlled Trials with Trial Sequential Analysis. General Hospital Psychiatry, 101, 159-165. DOI 10.1016/j.genhosppsych.2026.06.005 · PMID 42365709. Later meta-analysis on the same question, four trials and 292 patients. No significant difference in severity (mean difference −1.99; 95% CI: −4.70 to 0.72; I² = 73.6%), in hospital and intensive care length of stay, or in mortality, with a non-significant trend favoring quetiapine on extrapyramidal symptoms (risk ratio 0.14; 95% CI: 0.02 to 1.06). The trial sequential analysis shows that the required information size has not been reached and GRADE certainty is very low for all outcomes: the authors conclude that the data are inconclusive because of their scarcity and imprecision. The key reference for avoiding overinterpretation.
Maneeton B, Maneeton N, Srisurapanont M, Chittawatanarat K (2013). Quetiapine versus haloperidol in the treatment of delirium: a double-blind, randomized, controlled trial. Drug Design, Development and Therapy, 7, 657-667. DOI 10.2147/DDDT.S45575 · PMID 23926422. Double-blind trial over seven days, 52 medical inpatients, 24 on quetiapine 25 to 100 mg per day and 28 on haloperidol 0.5 to 2 mg per day. Reductions of about 22.9 and 21.7 points on the Delirium Rating Scale Revised 98, p = 0.59. This trial gives the order of magnitude of within-group changes against which to compare the between-group difference of the meta-analysis.
Grover S, Mahajan S, Chakrabarti S, Avasthi A (2016). Comparative effectiveness of quetiapine and haloperidol in delirium: A single blind randomized controlled study. World Journal of Psychiatry, 6(3), 365-371. DOI 10.5498/wjp.v6.i3.365 · PMID 27679777. Single-blind trial, 63 patients, 31 on quetiapine 12.5 to 75 mg per day and 32 on haloperidol 0.25 to 1.25 mg per day, over six days, Delirium Rating Scale Revised 98. Comparable improvement in both arms. Single blinding is the main methodological limitation of this trial.
Zakhary T, Ahmed I, Luttfi I, Montasser M (2024). Quetiapine Versus Haloperidol in the Management of Hyperactive Delirium: Randomized Controlled Trial. Neurocritical Care, 41(2), 550-557. DOI 10.1007/s12028-024-01948-w · PMID 38561588. Conducted in intensive care, unlike the two previous trials, 100 patients with a mean age of 68 years and hyperactive delirium, quetiapine 25 to 50 mg per day versus haloperidol 1 to 2 mg per day. Overall response rate 92%, with no difference between arms, p = 0.609. Shorter intensive care stay on quetiapine, 10.1 versus 11.7 days, p = 0.018, and more hours of sleep on day seven, 6 versus 3.5, p < 0.001, with no mortality benefit. The combined sample sizes of these three trials match the 215 participants of the meta-analysis, but this match is a reading hypothesis: the included trials are not named in the abstract.
Garg R, Singh VK, Pratishruti, Singh GP (2022). Comparison of Haloperidol and Quetiapine for Treatment of Delirium in Critical Illness: A Prospective Randomised Double-blind Placebo-controlled Trial. Journal of Clinical and Diagnostic Research, 16(7), UC31-UC33. Double-blind placebo-controlled trial in intensive care, 45 patients split into three arms of 15. Primary outcome: days alive without delirium over fourteen days, 9.45 on haloperidol, 8.64 on quetiapine and 8.57 on placebo, with no significant difference. Cited here because it shows that head-to-head trials are not limited to the three above, and because it adds the placebo arm that the analyzed meta-analysis lacks.
Nikooie R, Neufeld KJ, Oh ES, Wilson LM, Zhang A, Robinson KA, Needham DM (2019). Antipsychotics for Treating Delirium in Hospitalized Adults: A Systematic Review. Annals of Internal Medicine, 171(7), 485-495. DOI 10.7326/M19-1860 · PMID 31476770. Sixteen randomized trials and ten observational studies. No difference in delirium duration, length of stay or mortality, for haloperidol or for second-generation antipsychotics. Cardiac effects more frequent. The authors’ conclusion: the data do not support routine use of these drugs in this indication. This is the reference that frames the debate.
Girard TD, Exline MC, Carson SS, et al. (2018). Haloperidol and Ziprasidone for Treatment of Delirium in Critical Illness. New England Journal of Medicine, 379(26), 2506-2516. DOI 10.1056/NEJMoa1808217 · PMID 30346242. MIND-USA trial, 566 intensive care patients with acute respiratory failure or shock and hypoactive or hyperactive delirium. Median days alive without delirium or coma of 8.5 on placebo, 7.9 on haloperidol and 8.7 on ziprasidone, p = 0.26. The placebo-controlled trial that the comparison analyzed here lacks.
Kim DH, Lee SB, Park CM, Levin R, Metzger E, Bateman BT, Ely EW, Pandharipande PP, Pisani MA, Jones RN, Marcantonio ER, Inouye SK (2023). Comparative Safety Analysis of Oral Antipsychotics for In-Hospital Adverse Clinical Events in Older Adults After Major Surgery: A Nationwide Cohort Study. Annals of Internal Medicine. DOI 10.7326/M22-3021 · PMID 37665998. Nationwide cohort of 17,115 older adults after surgery. In-hospital mortality of 3.7% on haloperidol, 3.3% on risperidone, 2.8% on olanzapine and 2.6% on quetiapine, with no significant difference, and likewise for arrhythmias, pneumonia and stroke. Contrary to a widespread belief, these observational data show no excess mortality risk specific to low- to moderate-dose oral haloperidol. Observational study: confounding by indication cannot be eliminated despite adjustment.
Devlin JW, Skrobik Y, Gélinas C, et al. (2018). Clinical Practice Guidelines for the Prevention and Management of Pain, Agitation/Sedation, Delirium, Immobility, and Sleep Disruption in Adult Patients in the ICU. Critical Care Medicine, 46(9), e825-e873. DOI 10.1097/CCM.0000000000003299. PADIS guidelines. Suggest not routinely using haloperidol, an atypical antipsychotic or a statin to treat delirium: conditional recommendation, low-quality evidence.
Lewis K, Balas MC, Stollings JL, McNett M, Girard TD, et al. (2025). A Focused Update to the Clinical Practice Guidelines for the Prevention and Management of Pain, Anxiety, Agitation/Sedation, Delirium, Immobility, and Sleep Disruption in Adult Patients in the ICU. Critical Care Medicine, 53(3), e711-e727. DOI 10.1097/CCM.0000000000006574. Focused update of the previous guidelines. The panel states that it cannot make a recommendation, for or against, on the use of antipsychotics versus usual care for treating delirium in critically ill adults, with low certainty. This is the text to cite today: it withdraws the negative 2018 wording without endorsing these drugs.
National Institute for Health and Care Excellence. Delirium: prevention, diagnosis and management in hospital and long-term care, CG103, section 1.7 on treating delirium, last updated 2023. nice.org.uk. Identify and treat the cause or combination of causes first. For a person who is distressed, use verbal and non-verbal de-escalation techniques first, then consider short-term haloperidol, usually for one week or less, at the lowest effective dose with cautious titration, with particular attention to cardiac and neurological effects, especially in Parkinson’s disease and dementia with Lewy bodies.
Haute Autorité de santé (May 2009). Confusion aiguë chez la personne âgée : prise en charge initiale de l’agitation (Acute confusion in older people: initial management of agitation). has-sante.fr (in French). French national guidance, cited as a concrete example of a principle shared by guidelines: management requires above all treating the precipitating factors, with non-drug measures alone sufficient in a large number of cases. If treatment is needed: always start at a quarter or half of the adult dose, prefer the oral or orodispersible route over injection, limit to 24 to 48 hours. Warning about Lewy body disease, where even a low dose of neuroleptic can cause excessive sedation or even coma. An older recommendation, not updated to date, but the only specific French national reference.
Medicines and Healthcare products Regulatory Agency (December 2021). Review of the use of haloperidol in elderly patients with acute delirium, Public Assessment Report. gov.uk. Confirms that the authorized indication covers the acute treatment of delirium in adults after non-pharmacological treatments have failed, and that newer antipsychotics are not authorized in this situation. Finds that efficacy data in older adults are limited and restates the rule of the lowest dose for the shortest duration. The main regulatory anchor of this analysis; authorization status may differ in other countries.
Agence nationale de sécurité du médicament et des produits de santé (ANSM, France). HALDOL 5 mg, tablet, summary of product characteristics, updated April 8, 2026. French public medicines database (in French). Indication for acute treatment of delirium when non-pharmacological treatments have failed, in adults, also present in the summary of product characteristics of HALDOL 5 mg/ml, solution for injection, updated on the same date, which recommends the intramuscular route and continuous ECG monitoring for intravenous administration. Section 4.4 “Increased mortality in elderly people with dementia”. Electrocardiogram recommended at initiation, monitoring of QTc and ventricular arrhythmias, discontinuation above 500 ms. Cited as an example of a European label.
Agence nationale de sécurité du médicament et des produits de santé (ANSM, France). XEROQUEL LP 300 mg, prolonged-release tablet, summary of product characteristics, updated April 22, 2025. French public medicines database (in French). Indications limited to schizophrenia, manic and depressive episodes of bipolar disorder, prevention of recurrence and adjunctive treatment of major depressive episodes. Quetiapine is not approved for the treatment of patients with dementia-related psychosis, with mortality of 5.5% versus 3.2% on placebo in ten-week trials and an increased cerebrovascular risk. QT prolongation reported at therapeutic doses.