Alzheimer's Psychosis Drug Narrowly Misses, but Nursing Homes Still Need a Safer Option

A p-value of 0.0603 is the kind of result that makes a development team wince. Acadia Pharmaceuticals' remlifanserin came within a fraction of a point of hitting its primary endpoint in Alzheimer's disease psychosis, then cleared its key secondary endpoint and produced a safety profile that looked like placebo. Acadia is pressing on into Phase 3. The more interesting question is who is waiting on the other side. Psychosis in dementia is managed largely in nursing homes, with drugs that were never approved for it and that federal regulators have spent more than a decade trying to curb. For that market, a drug that does no harm and helps modestly could be worth more than its effect size suggests.

A Near Miss With a Clean Safety Profile

Acadia reported topline RADIANT results on September 24. The randomized, double-blind, placebo-controlled study tested once-daily remlifanserin at 60 mg and 30 mg. At six weeks, the 60 mg dose reduced the hallucinations and delusions score (SAPS-H+D) by 12.6 points from baseline, against 10.4 points on placebo, a standardized effect size of 0.26 and a p-value of 0.0603. On the key secondary measure, clinician-rated severity of Alzheimer's psychosis, the 60 mg dose improved 1.3 points against 0.9 for placebo, an effect size of 0.37 that was nominally significant at p=0.0077. Acadia will drop the 30 mg arm from its two ongoing Phase 3 studies and present full data at the CTAD conference in November.

The placebo response is the story inside the story. A 10.4-point improvement without active drug is large, and it is typical of neuropsychiatric trials in dementia, where caregiver attention, study-visit structure and natural fluctuation in symptoms all move scores. Designing Phase 3 to separate signal from that noise is now Acadia's central task. Chief executive Catherine Owen Adams said the results provide valuable information on the efficacy, safety and tolerability of the once-daily drug.

The safety readout is where the program looks strongest. Rates of adverse events, serious adverse events and discontinuations were similar to placebo across both doses. Acadia reported no signal of QT prolongation, no negative impact on motor symptoms or cognition, and no deaths in the remlifanserin arms. The company estimates that about 30% of people with Alzheimer's experience psychosis, against more than 7 million Americans living with the disease, and it notes that there is no FDA-approved drug for the condition.

Why Safety Is the Selling Point in Nursing Homes

In practice, dementia-related psychosis is often treated off-label with atypical antipsychotics, and much of that prescribing happens in long-term care. CMS has made reducing it a quality priority. According to the agency's National Partnership to Improve Dementia Care data report, 23.9% of long-stay nursing home residents were receiving an antipsychotic in late 2011. By the second quarter of 2025 that had fallen 40.6%, to 14.2%. The measure excludes residents with schizophrenia, Huntington's disease or Tourette's syndrome, and CMS says it does not expect the rate to reach zero because clinical indications do exist.

The decline may have been flattered by the data. In a 2025 memo to state survey agencies, CMS described inappropriate antipsychotic use as very dangerous for residents, acting as a chemical restraint or causing death. It cited a 2021 Inspector General finding that the Minimum Data Set might not accurately reflect how many residents are prescribed these drugs, and rebuilt the measure to add Medicare and Medicaid claims and Medicare Advantage encounter data. Under the old method, 14.64% of long-stay residents were receiving an antipsychotic. Under the new one, the figure is 16.98%. The updated measure went live on Nursing Home Care Compare in January 2026 and feeds the Five-Star quality rating.

That puts every facility's prescribing pattern on a public scorecard, which changes the commercial calculus for a psychosis drug. A therapy that works modestly but avoids sedation, motor decline and cardiac risk speaks directly to the clinical and regulatory worries that make nursing homes cautious about the drugs they use today.

The Falls Data Acadia Is Already Building

Acadia has run this playbook before with pimavanserin, sold as Nuplazid for Parkinson's disease psychosis. A retrospective analysis of 100% Medicare claims, reported by Psychiatric Times on September 18, matched three cohorts of 1,385 long-term care and nursing home residents each. Within six months, 19.86% of pimavanserin patients fell, against 25.27% on other atypical antipsychotics and 25.27% on quetiapine. Recurrent falls ran 7.44% against 12.85% and 12.13%. The adjusted relative risk of recurrent falls was 0.54 compared with other atypicals and 0.60 compared with quetiapine.

The study was led by researchers from Acadia and Anlitiks, so it should be read as sponsor-generated real-world evidence rather than an independent verdict, and claims data cannot fully adjust for why one resident got one drug over another. Still, it shows the shape of the evidence package Acadia is likely to build for remlifanserin: outcomes that matter to facilities, such as falls, fractures and hospital transfers, measured in the settings where the drug will actually be used. The study authors noted that falls among these patients are common and can lead to injuries and hospitalizations.

What Life Sciences Leaders Should Do Now

• Design Phase 3 around the placebo problem: Neuropsychiatric endpoints in dementia carry heavy placebo response, so enrichment strategies, rater training and caregiver-informed measures deserve as much attention as dose selection.
• Treat safety as a primary value claim: In long-term care, the absence of sedation, motor decline and QT risk is a reason to prescribe, so collect safety data at a depth that supports that message.
• Build falls and transfer outcomes into the evidence plan: Facilities and payers will ask about falls, fractures and hospitalizations, so plan prospective real-world studies alongside the pivotal trials.
• Engage the nursing home data ecosystem early: Electronic health records, MDS assessments and claims are where real-world outcomes will be measured, and partnerships set up before launch shorten the path to credible evidence.
• Plan for the behavioral health workforce gap: A new drug will reach residents only if someone is available to assess psychosis and monitor response, which makes clinician access part of the launch strategy.

Remlifanserin has not proven it works, and November's full data will matter. But the field it is entering has spent years defined by drugs that calm residents at the cost of their balance, alertness and safety. If Phase 3 delivers even a modest benefit on top of a placebo-like safety profile, the most persuasive data point may be the one CMS already publishes for every nursing home in the country.