A nonprofit division of Sevadar Foundation Inc.

For clinicians  ·  For facilities  ·  For families  ·  For staff

Antipsychotics and mortality in dementia

The evidence base

The mortality signal for antipsychotics in dementia is real, it is randomized, and it is smaller than most people assume. Getting the size right matters, because both overstatement and dismissal lead somewhere bad.

The regulatory history

On April 11, 2005 the FDA issued a public health advisory on deaths with antipsychotics in elderly patients with behavioral disturbances, covering the atypical agents. Of seventeen placebo-controlled trials, fifteen showed numerical increases in mortality in the drug-treated group — approximately a 1.6 to 1.7-fold increase. The agency asked manufacturers to add a boxed warning.

On June 16, 2008 the FDA required manufacturers of the conventional antipsychotics to add the same boxed warning, having concluded that “elderly patients with dementia-related psychosis treated with conventional or atypical antipsychotic drugs are at an increased risk of death.”

The warning is unusually candid about its own evidence. Current FDA-approved labeling for a conventional agent states that the conventional-drug signal rests on observational studies and that “the extent to which the findings of increased mortality in observational studies may be attributed to the antipsychotic drug as opposed to some characteristic(s) of the patients is not clear.”

The size of the risk

The pivotal meta-analysis pooled 15 randomized placebo-controlled trials — 16 drug-placebo contrasts, generally 10 to 12 weeks, covering aripiprazole, olanzapine, quetiapine and risperidone — with 3,353 participants randomized to drug and 1,757 to placebo. Death occurred in 118 (3.5%) on drug and 40 (2.3%) on placebo: odds ratio 1.54 (95% CI 1.06 to 2.23), risk difference 0.01 (95% CI 0.004 to 0.02). That is roughly one additional death per hundred people treated over ten to twelve weeks.

Observational work suggests the absolute effect may be larger over longer periods. A retrospective case-control study of 90,786 patients aged 65 and over with dementia in the Veterans Health Administration reported, over 180 days versus matched non-users, absolute mortality risk increases of 3.8% for haloperidol (number needed to harm 26), 3.7% for risperidone (27), 2.5% for olanzapine (40) and 2.0% for quetiapine (50), with a dose-response among the atypicals. That study is observational and the authors say so.

Switching to an older drug is not a solution

A retrospective cohort of 22,890 Pennsylvania patients aged 65 and over initiating an antipsychotic found that conventional agents carried a higher adjusted risk of death than atypicals at every interval — relative risk 1.37 (95% CI 1.27 to 1.49) within 180 days, and 1.56 (95% CI 1.37 to 1.78) within 40 days. The finding held with and without dementia, inside and outside nursing homes, and survived propensity-score and instrumental-variable analysis. It is observational, and confounding by indication cannot be excluded.

What the drugs actually do for symptoms

CATIE-AD randomized 421 outpatients with Alzheimer disease and psychosis, aggression or agitation across 42 sites for up to 36 weeks. Its primary outcome — time to discontinuation for any reason — showed no significant difference from placebo (p = 0.52). Time to discontinuation for lack of efficacy favored olanzapine and risperidone, so there is a real efficacy signal. Time to discontinuation for intolerability favored placebo sharply: 24%, 16% and 18% for the three drugs against 5% on placebo. Global improvement did not differ significantly (32%, 26%, 29% versus 21%, p = 0.22). The authors’ conclusion: “Adverse effects offset advantages in the efficacy of atypical antipsychotic drugs.”

A 2021 Cochrane review of 24 randomized trials covering 6,090 participants found atypical antipsychotics probably reduce agitation slightly (standardized mean difference −0.21, 95% CI −0.30 to −0.12; 7 studies, 1,971 participants; moderate certainty) and probably have a negligible effect on psychosis (−0.11, 95% CI −0.18 to −0.03; 12 studies, 3,364 participants; moderate certainty), against somnolence (relative risk 1.93, high certainty), extrapyramidal symptoms (1.39) and serious adverse events (1.32).

The review’s most useful sentence for anyone running a facility: “The apparent effectiveness of the drugs seen in daily practice may be explained by a favourable natural course of the symptoms, as observed in the placebo groups.”

What withdrawal does

DART-AD randomized 165 UK care-home residents with Alzheimer disease already established on an antipsychotic to continue or switch to placebo; 128 started treatment. Survival was reduced in those who continued: 12-month cumulative survival 70% (95% CI 58 to 80) against 77% (64 to 85), with divergence widening to 46% against 71% at 24 months and 30% against 59% at 36 months. Its companion paper found no significant difference in cognition or neuropsychiatric symptoms at 6 months, with some evidence that residents whose baseline symptoms were more severe benefited from continuing.

That is a small trial and the mortality finding has not been replicated. The 2018 Cochrane review of antipsychotic withdrawal — 10 studies, 632 participants, all on an antipsychotic for at least three months — found low-certainty evidence that the drugs can be successfully discontinued with little or no important effect on behavioral and psychological symptoms, and stated of mortality that “we are uncertain whether discontinuation of antipsychotics leads to a decrease in mortality.”

The subgroup finding is the clinically important one. Two trials enrolled only patients who had responded to antipsychotic treatment for psychosis, agitation or aggression; in those, stopping was associated with a higher risk of leaving the study early due to symptomatic relapse, or a shorter time to relapse. People with more severe baseline symptoms may also worsen. Cochrane’s own summary: those groups “may benefit behaviourally from continuation.”

Which is why the answer is a monitored attempt rather than a policy. Most people can come off. The ones who cannot are identifiable, and they are precisely the ones a target-driven reduction program would harm.

Other harms

  • Stroke. A self-controlled case series of 6,790 UK patients with incident stroke found rate ratios during exposed periods of 1.73 (95% CI 1.60 to 1.87) for any antipsychotic — and 3.50 (95% CI 2.97 to 4.12) in patients with dementia against 1.41 without.
  • Pneumonia. A nested case-control study in adults 65 and over found odds ratios versus past use of 2.61 (95% CI 1.48 to 4.61) for atypicals and 1.76 (1.22 to 2.53) for conventionals, dose-dependent. A separate cohort found risk highest in the first week after initiation (odds ratio 4.5, 95% CI 2.8 to 7.3).
  • Fracture. In a nationwide cohort of 1,540,915 Danes aged 65 and over, incidence rate ratios in the first 30 days after initiation ranged from 1.97 for risperidone to 2.98 for haloperidol. That cohort is adults 65 and over, not restricted to dementia.

All four are observational. Note how consistently the risk concentrates at initiation — that is the practical finding.

When they are the right answer

The American Psychiatric Association’s 2016 practice guideline is specific, and it is not a prohibition. Its recommendations, in summary: outside imminent-danger situations, use antipsychotics for agitation or psychosis in dementia only when symptoms are severe, dangerous, or causing significant distress; review the response to non-pharmacological interventions first; start low and titrate to the minimum effective dose; if there is no clinically significant response after a four-week trial at an adequate dose, taper and withdraw; if there is a response, attempt a taper within four months unless prior attempts caused recurrence; assess at least monthly during the taper and for at least four months after; and in the absence of delirium, do not use haloperidol first-line.

The quality measure, and its 2026 change

The CMS National Partnership to Improve Dementia Care in Nursing Homes was launched in 2012. On CMS’s own reporting, the share of long-stay nursing home residents receiving an antipsychotic fell from 23.9% in the fourth quarter of 2011 to 14.2% in the second quarter of 2025, a relative decrease of 40.6%. CMS states plainly that circumstances exist where clinical indications are present and that it “does not expect that the national prevalence of antipsychotic medication use will decrease to zero.”

Two things about that measure are worth knowing. First, in January 2023 CMS announced audits of schizophrenia coding, on the concern that “some nursing homes have erroneously coded residents as having schizophrenia, which can mask the facilities’ true rate of antipsychotic medication use,” noting that in a pilot there was “an absence of comprehensive psychiatric evaluations and behavior documentation.” Second, the measure itself was respecified effective January 1, 2026: the diagnosis exclusions now require corroborating claims data rather than the assessment item alone, and the numerator draws on pharmacy claims as well. Figures produced under the old and new specifications are not comparable, and this site does not mix them.

There is a broader caution here that belongs on a page about medication burden. A quality measure creates pressure toward the measured number. The evidence above says most residents can come off an antipsychotic safely, and that a specific identifiable minority relapse when they do. A program that pursues the number rather than the individual will find that minority the hard way.

Common questions

How much does an antipsychotic increase the risk of death in dementia?

In the randomized trials behind the FDA warning, roughly one additional death per hundred people treated over ten to twelve weeks — odds ratio 1.54, risk difference 0.01. Observational studies over longer periods suggest a larger absolute effect, with numbers needed to harm between 26 and 50 over 180 days depending on the drug. Antipsychotics in dementia.

Can an antipsychotic ever be the right choice in dementia?

Yes, and the guidelines say so. The American Psychiatric Association reserves non-emergency use for symptoms that are severe, dangerous, or causing significant distress, after the response to non-pharmacological approaches has been reviewed, with a four-week efficacy checkpoint and a taper attempt within four months. That is a narrow indication, not an absent one. Non-pharmacological interventions.

Is it safe to stop an antipsychotic someone has been on for years?

For most people, yes, on low-certainty randomized evidence — withdrawal usually succeeds without symptom rebound. The exceptions are identifiable: people who responded well to the drug for psychosis, aggression or agitation, and people with more severe baseline symptoms, are more likely to relapse. That is a reason for a monitored attempt with a plan, and it is why nobody should change a prescription without their own prescriber. Tapering principles.

Back to the evidence base.

What the evidence does and does not show

Sedating medications impair gait, balance and reaction time, and their labels say so. That part is not in doubt. What has not been shown is that a deprescribing program reduces fall rates — the trials that tested it were small, short, and mostly measured a medication count rather than a fracture. Those are two different questions, and absence of evidence from underpowered trials is not evidence that the drugs do not cause falls. What the evidence actually shows about deprescribing and falls.

References

  1. U.S. Food and Drug Administration. Public Health Advisory: Deaths with Antipsychotics in Elderly Patients with Behavioral Disturbances. April 11, 2005. fda.gov. 2005. https://wayback.archive-it.org/7993/20170406045738/https://www.fda.gov/Drugs/DrugSafety/PostmarketDrugSafetyInformationforPatientsandProviders/ucm053171.htm.
  2. U.S. Food and Drug Administration. Information for Healthcare Professionals: Conventional Antipsychotics. June 16, 2008. fda.gov. 2008. https://wayback.archive-it.org/7993/20170112031213/http://www.fda.gov/Drugs/DrugSafety/PostmarketDrugSafetyInformationforPatientsandProviders/ucm124830.htm.
  3. Schneider LS, Dagerman KS, Insel P. Risk of death with atypical antipsychotic drug treatment for dementia: meta-analysis of randomized placebo-controlled trials. JAMA. 2005;294(15):1934–1943. PMID 16234500 · DOI 10.1001/jama.294.15.1934.
  4. Wang PS, Schneeweiss S, Avorn J, et al. Risk of death in elderly users of conventional vs. atypical antipsychotic medications. The New England Journal of Medicine. 2005;353(22):2335–2341. PMID 16319382 · DOI 10.1056/NEJMoa052827.
  5. Maust DT, Kim HM, Seyfried LS, et al. Antipsychotics, other psychotropics, and the risk of death in patients with dementia: number needed to harm. JAMA Psychiatry. 2015;72(5):438–445. PMID 25786075 · DOI 10.1001/jamapsychiatry.2014.3018.
  6. Schneider LS, Tariot PN, Dagerman KS, et al. Effectiveness of atypical antipsychotic drugs in patients with Alzheimer’s disease. The New England Journal of Medicine. 2006;355(15):1525–1538. PMID 17035647 · DOI 10.1056/NEJMoa061240.
  7. Mühlbauer V, Möhler R, Dichter MN, Zuidema SU, Köpke S, Luijendijk HJ. Antipsychotics for agitation and psychosis in people with Alzheimer’s disease and vascular dementia. Cochrane Database of Systematic Reviews. 2021;12(12):CD013304. PMID 34918337 · DOI 10.1002/14651858.CD013304.pub2.
  8. Ballard C, Hanney ML, Theodoulou M, et al. The dementia antipsychotic withdrawal trial (DART-AD): long-term follow-up of a randomised placebo-controlled trial. The Lancet Neurology. 2009;8(2):151–157. PMID 19138567 · DOI 10.1016/S1474-4422(08)70295-3.
  9. Ballard C, Lana MM, Theodoulou M, et al. A randomised, blinded, placebo-controlled trial in dementia patients continuing or stopping neuroleptics (the DART-AD trial). PLoS Medicine. 2008;5(4):e76. PMID 18384230 · DOI 10.1371/journal.pmed.0050076.
  10. Van Leeuwen E, Petrovic M, van Driel ML, et al. Withdrawal versus continuation of long-term antipsychotic drug use for behavioural and psychological symptoms in older people with dementia. Cochrane Database of Systematic Reviews. 2018;3(3):CD007726. PMID 29605970 · DOI 10.1002/14651858.CD007726.pub3.
  11. Douglas IJ, Smeeth L. Exposure to antipsychotics and risk of stroke: self controlled case series study. BMJ. 2008;337:a1227. PMID 18755769 · DOI 10.1136/bmj.a1227.
  12. Trifirò G, Gambassi G, Sen EF, et al. Association of community-acquired pneumonia with antipsychotic drug use in elderly patients: a nested case-control study. Annals of Internal Medicine. 2010;152(7):418–425. PMID 20368647 · DOI 10.7326/0003-4819-152-7-201004060-00006.
  13. Knol W, van Marum RJ, Jansen PAF, Souverein PC, Schobben AFAM, Egberts ACG. Antipsychotic drug use and risk of pneumonia in elderly people. Journal of the American Geriatrics Society. 2008;56(4):661–666. PMID 18266664 · DOI 10.1111/j.1532-5415.2007.01625.x.
  14. Torstensson M, Leth-Møller K, Andersson C, Torp-Pedersen C, Gislason GH, Holm EA. Danish register-based study on the association between specific antipsychotic drugs and fractures in elderly individuals. Age and Ageing. 2017;46(2):258–264. PMID 27932365 · DOI 10.1093/ageing/afw209.
  15. Reus VI, Fochtmann LJ, Eyler AE, et al. The American Psychiatric Association Practice Guideline on the Use of Antipsychotics to Treat Agitation or Psychosis in Patients With Dementia. The American Journal of Psychiatry. 2016;173(5):543–546. PMID 27133416 · DOI 10.1176/appi.ajp.2015.173501.
  16. Yohanna D, Cifu AS. Antipsychotics to Treat Agitation or Psychosis in Patients With Dementia. JAMA. 2017;318(11):1057–1058. PMID 28975291 · DOI 10.1001/jama.2017.11112.
  17. Centers for Medicare & Medicaid Services. National Partnership to Improve Dementia Care in Nursing Homes: Antipsychotic Medication Use Data Report. cms.gov. 2026. https://www.cms.gov/files/document/data-report-national-partnership-improve-dementia-care-nursing-homes-antipsychotic-medication-use.pdf.
  18. Centers for Medicare & Medicaid Services. Adjusting Quality Measure Ratings Based on Erroneous Schizophrenia Coding, and Posting Citations Under Dispute. Memorandum QSO-23-05-NH, January 18, 2023. cms.gov. 2023. https://www.cms.gov/files/document/qso-23-05-nh-adjusting-quality-measure-ratings-based-erroneous-schizophrenia-coding-and-posting.pdf.

Bibliographic records on this page were retrieved from PubMed. Every reference links to its DOI or PubMed record.

Written and medically reviewed by Dr. Gurpreet Singh Padda, MD, MBA, MHP, founder and medical director of DWARAA. Last reviewed . This page is educational. DWARAA does not prescribe, deprescribe, diagnose or treat.