When the FDA approved Bristol Myers Squibb’s Cobenfy in September 2024, it cleared the first genuinely new mechanism of action for schizophrenia in decades. Cobenfy, the combination of xanomeline and trospium chloride formerly known as KarXT, is a muscarinic agonist. It works through the M1 and M4 acetylcholine receptors rather than by blocking the dopamine D2 receptor, which is the route every prior antipsychotic took and the source of the side-effect profile that drives patients off their medication. As a monotherapy for adults with schizophrenia, Cobenfy is approved and selling.

That sentence is the entire good-news case, and it is worth stating plainly before the rest of the picture, because the rest of the picture is a sequence of misses. Every commercial bet built on top of the monotherapy approval has either failed or stalled, and the settings where the muscarinic class has failed are the settings that hold the volume.

The adjunctive miss

Most patients who take an antipsychotic are not treatment-naive. They are already on a drug that is not controlling their symptoms, and the prescriber is deciding what to add. That adjunctive market, a new agent layered on top of an existing atypical antipsychotic, is where the prescribing volume concentrates. It is also where Cobenfy missed.

In the ARISE trial, a roughly 400-patient study reported in April 2025, adjunctive Cobenfy produced a 2.0-point improvement on the Positive and Negative Syndrome Scale versus placebo added to an atypical antipsychotic at week 6, with a p-value of 0.11. That does not clear statistical significance. The safety and tolerability profile held, which is why the company said it would keep analyzing the data and talking to regulators, but the efficacy signal in the high-volume setting was not there. A drug approved for the clean monotherapy population had failed to show it adds measurable benefit in the population where most prescribing decisions actually happen.

The rival that failed outright

If Cobenfy were the only muscarinic in development, the adjunctive miss would read as a single-program problem. It is not. The other lead muscarinic candidate for schizophrenia, AbbVie’s emraclidine, failed both of its pivotal Phase 2 trials in November 2024. EMPOWER-1 and EMPOWER-2 each missed the primary endpoint of a statistically significant PANSS reduction versus placebo at week 6. AbbVie, which had acquired emraclidine through its purchase of Cerevel Therapeutics, took a $3.5 billion impairment charge tied to the program in January 2025. On the day of the readout, AbbVie shares fell more than 12 percent and Bristol Myers shares rose by a similar amount, the market correctly reading one failure as the other’s competitive relief.

The two drugs are not the same mechanism, and the difference is the most interesting part of the class story. Xanomeline, the active component of Cobenfy, is an agonist at both M1 and M4 receptors. Emraclidine is a positive allosteric modulator selective for M4 alone. Emraclidine’s failure, read against Cobenfy’s monotherapy success, points toward M1 agonism contributing to the antipsychotic effect, which would mean the broad “muscarinic” label covers mechanisms that do not behave alike. The class is not one bet. It is several, and so far only the M1-and-M4 agonist has cleared a pivotal trial, and only as monotherapy.

The Alzheimer’s program slipped a year

The largest expansion opportunity for Cobenfy is not in schizophrenia at all. It is in Alzheimer’s disease psychosis, which affects an estimated 40 to 60 percent of Alzheimer’s patients and has no FDA-approved treatment specific to it. The drugs used off-label are atypical antipsychotics carrying black-box warnings for increased mortality in elderly dementia patients. A muscarinic agent that worked here would own a category.

That readout has moved. In December 2025, Bristol Myers said it had identified irregularities in clinical-trial execution at a limited number of sites in the Phase 3 ADEPT-2 study, excluded the data from those sites, and would enroll additional patients. The trial, originally expected to read out by the end of 2025, is now expected by the end of 2026, with the parallel ADEPT-1 and ADEPT-4 studies on the same revised timeline. The company framed the exclusion as a commitment to data integrity, which is the correct framing, but the practical effect is that the single most valuable Cobenfy readout is a year later than planned and arrives with a question about site conduct attached.

The launch itself is modest

The monotherapy approval is real, and the monotherapy launch is small. In its first nine months on the market across 2025, Cobenfy recorded about $105 million in sales. That is a respectable start for a psychiatric launch and a long way from the numbers that would justify the mechanism’s billing as the future of the category. The drug needs the adjunctive indication or the Alzheimer’s indication, or both, to become the franchise the mechanism was supposed to support. One of those has failed and the other has slipped.

The Cobenfy program also sits inside a difficult year for its sponsor. Bristol Myers recorded Phase 3 setbacks in 2025 across Opdualag, Camzyos, Reblozyl, and the Johnson and Johnson-partnered milvexian, and its chief medical officer at the time, Samit Hirawat, stepped down amid the run of results, succeeded by AstraZeneca veteran Cristian Massacesi. Cobenfy is one of the products meant to offset patent losses on the company’s top sellers, which raises the stakes on the ADEPT readout rather than lowering them.

The ledger

Read as a class rather than a single drug, the muscarinic story in mid-2026 is a ledger with one entry on each side and two still open. Monotherapy schizophrenia: approved and selling, if modestly. Adjunctive schizophrenia: missed. A second, M4-selective molecule: failed and written down. Alzheimer’s psychosis: pending, delayed to the end of 2026, and carrying a site-conduct asterisk.

The reason adjunctive and acute-exacerbation trials are so hard to win is not specific to muscarinics, and it is the subject the desk takes up separately in its primer on reading a schizophrenia trial. The short version is that the placebo arms in these studies have been improving for years, which shrinks the gap any drug has to clear. A 2.0-point miss at p=0.11 is the kind of result that gap produces.

None of this unwinds the mechanism. A non-dopaminergic antipsychotic that patients tolerate is a real advance, and the monotherapy approval is the proof. But the market has been pricing the muscarinic class as a platform, and a platform needs to work in more than one setting. As of now it has worked in one. The next datapoint that changes that is the ADEPT readout at the end of 2026. Until it lands, the new era in schizophrenia is one drug, in one population, posting nine-figure sales in a category that needs ten.