Roughly a third to a half of people with Alzheimer's disease develop psychosis at some point, meaning hallucinations, delusions, or both. There is no medication approved in the United States to treat it.
The trial program most likely to change that just moved further away. Bristol Myers Squibb told investors on Thursday that topline results from its three ADEPT studies, testing the schizophrenia drug Cobenfy in Alzheimer's disease psychosis, will now begin arriving in early 2027 and spread across that year. As recently as April, the company still expected all three to report in 2026.
For families, the delay does not change anything about this week. What it does is extend the period in which the only pharmacologic options are drugs used off-label, most of which carry a boxed warning about increased mortality in exactly this population.
Sixteen Months of Slipping Timelines
Chief Executive Chris Boerner said the company now anticipates readouts "to begin in early 2027 and be spread across the year." He attributed the shift to slower enrollment in the ADEPT-2 and ADEPT-4 studies and slower accrual of relapse events in ADEPT-1.
This is the second public delay. In December 2025, BMS reopened enrollment in ADEPT-2 after identifying what it called irregularities in trial execution at a few study sites. The company excluded data from those sites, informed the Food and Drug Administration, and continued the study after an independent data monitoring committee reviewed it. At that point the readout moved to the end of 2026.
Not every element of the current delay is unfavorable. Chief Medical Officer Cristian Massacesi said slower relapse accrual in ADEPT-1 could indicate the drug is working, since fewer relapses means the trial takes longer to reach its required number of events. He also acknowledged that because the study remains blinded, the company cannot tell whether that reflects Cobenfy or a control arm performing better than historical experience.
BMS said it plans to release safety and efficacy data this year from the open-label lead-in portion of ADEPT-1 and from a rollover study of patients who completed the trials. An interim analysis of ADEPT-1 could come later in 2026. Open-label data, in which everyone knows what they are receiving, cannot establish efficacy.
The Boxed Warning at the Center of Current Practice
Understanding why this program matters requires understanding what clinicians reach for now.
In 2005, the FDA applied a boxed warning to all antipsychotics stating that elderly patients with dementia-related psychosis treated with these drugs face an increased risk of death. The warning came from pooled analyses of 17 placebo-controlled trials in this population, which found death rates roughly 1.6 to 1.7 times higher among treated patients.
No antipsychotic has been approved for psychosis in Alzheimer's disease. Two adjacent approvals exist. Pimavanserin was approved in 2016 for hallucinations and delusions associated with Parkinson's disease psychosis, and is explicitly not approved for dementia-related psychosis unrelated to Parkinson's. Brexpiprazole was approved in May 2023 for agitation associated with Alzheimer's dementia, a different symptom, and retains the class boxed warning.
The result is that psychosis in dementia is managed largely off-label, with second-generation antipsychotics, benzodiazepines, antidepressants, and mood stabilizers, all with modest evidence and meaningful side effects including sedation, falls, and stroke. Guidelines recommend non-drug approaches first and limit antipsychotic courses to short durations. An FDA public workshop convened with the Duke-Margolis Center has examined whether the boxed warning evidence should be reanalyzed.
BMS neuroscience development head Laura Gault described the condition in the company's earlier statement as an area of "tremendous unmet medical need."
Reading a Delay Without Overreading It
A pushed readout is not a failed trial, and it should not be reported as one.
What is known: enrollment has been slower than planned in two studies, relapse events have accrued more slowly than modeled in a third, and one study previously had site conduct problems serious enough to exclude data. What is not known: whether Cobenfy works for this indication. The trials remain blinded, and no efficacy result exists.
Cobenfy is already approved for schizophrenia in adults. Its mechanism, muscarinic receptor agonism, is different from the dopamine-blocking drugs that carry the class warning, which is the basis for hoping it might have a different safety profile in frail older patients. Hoping is the accurate verb. That has not been demonstrated in this population.
Other programs are moving. Acadia's remlifanserin received FDA fast-track designation for Alzheimer's disease psychosis, with phase 2 RADIANT topline results expected in September or October 2026 and phase 3 studies beginning. MapLight Therapeutics expects phase 2 results in this indication in the second half of 2027. None of these is close to an approval decision.
Options for Families Managing Hallucinations Today
Nothing in this reporting should prompt anyone to start, stop, or change a medication. These decisions belong with a clinician who knows the patient.
The first practical step when hallucinations or delusions appear is a medical evaluation for a reversible cause. Urinary tract infection, pneumonia, dehydration, uncontrolled pain, constipation, poor sleep, and newly added medications can all precipitate or worsen psychotic symptoms in dementia, and treating the underlying problem sometimes resolves them without adding a psychiatric drug.
Non-drug approaches are recommended first in guidelines, and they are more concrete than they sound. Identifying triggers, keeping routines and environments consistent, improving lighting to reduce misperception, correcting hearing and vision problems, and avoiding confrontation about the content of a delusion are all standard. Families often find that arguing with a delusion escalates it.
When medication is used, reasonable questions for the prescriber include what specific symptom is being targeted, what the plan is for reassessing and stopping, and what side effects should trigger a call. Sedation, new unsteadiness, or a fall warrants prompt contact.
Symptoms that require urgent evaluation rather than a scheduled visit include a sudden change in alertness or confusion over hours to days, fever, new weakness, or any situation where the person or a caregiver is unsafe.
The Alzheimer's Association operates a 24-hour helpline staffed by clinicians, and caregiver support programs are frequently covered under Medicare. What happens next is measurable: ADEPT-1 interim results and open-label data are expected later this year, remlifanserin phase 2 data in early fall, and ADEPT topline readouts beginning in early 2027. MedicalDaily will report each.
Frequently Asked Questions
What was announced? Bristol Myers Squibb said topline results from its three ADEPT trials of Cobenfy in Alzheimer's disease psychosis will begin arriving in early 2027 rather than during 2026.
Why was it delayed? Slower enrollment in ADEPT-2 and ADEPT-4, and slower accrual of relapse events in ADEPT-1. An earlier delay followed conduct irregularities at some study sites.
Is there an approved treatment for psychosis in Alzheimer's? No. No medication is FDA-approved for this indication. Brexpiprazole is approved for agitation in Alzheimer's dementia, which is a different symptom.
What is the boxed warning? Since 2005, all antipsychotics carry an FDA warning that elderly patients with dementia-related psychosis treated with them face an increased risk of death.
Does the delay mean the drug failed? No. The trials remain blinded, and no efficacy results exist. The company said slower relapse accrual could reflect benefit, but cannot determine that while blinded.
What should families do when hallucinations start? Seek a medical evaluation for reversible causes such as infection, pain, dehydration, or new medications, and ask about non-drug approaches before adding a psychiatric drug.
When is the next data expected? Open-label ADEPT-1 data and a possible interim analysis later in 2026, remlifanserin phase 2 results in September or October 2026, and ADEPT topline readouts starting in early 2027.
