BMS-986368

BMS-986368 (irafamdastat; also previously designated CC-97489 and ABX-1772 [11]) is Bristol-Myers Squibb's oral dual FAAH/MAGL inhibitor, now in two Phase 2 trials: one in agitation associated with Alzheimer's disease (n=120) and one in multiple sclerosis spasticity (n=200) [1][2]. The compound blocks both enzymes that break down the brain's own cannabinoids, aiming to dial up endogenous endocannabinoid tone enough to calm agitation without the sedation and movement-disorder problems that limit antipsychotic use in dementia. The AD agitation indication is the commercially interesting one: brexpiprazole (Rexulti) won this label in May 2023 as the only FDA-approved option, leaving room for a cleaner second mover [3]. Note that brexpiprazole is a D2/D3 and 5-HT1A partial agonist atypical antipsychotic, so it is a competitor in the same indication rather than a competitor in the same drug class. There are no approved FAAH or MAGL inhibitors. For BMS, sitting on roughly $48B in annual revenue but facing loss of exclusivity on Eliquis and Revlimid, this is a small early-stage shot in a CNS indication that has burned a long line of dementia drugs [4].

This is a novel compound, never approved anywhere. Both Phase 2 programs are actively recruiting: the Alzheimer's agitation trial (NCT06808984) targets 120 patients and the MS spasticity trial (NCT06782490) targets 200 [1][2]. The Phase 1 package is essentially complete: a multi-cohort safety and PK study in healthy young, elderly, and Japanese volunteers (NCT06411730), a drug-drug interaction and food-effect study using itraconazole as a CYP3A4 probe (NCT06170723), and a radiolabeled mass-balance study to characterize absorption, metabolism, and excretion (NCT06227975) [5][6][7]. Detailed Phase 1 PK findings, including half-life, Tmax, dose-limiting observations, and any CNS biomarker data that informed Phase 2 dose selection, have not been publicly disclosed at the level of a peer-reviewed publication or AAIC abstract that we have located. No FDA breakthrough therapy, fast track, orphan drug, or RMAT designation has been disclosed for either indication. The sponsor of record on ClinicalTrials.gov is still listed as Celgene, the BMS subsidiary acquired in 2019, which is BMS's standard convention for assets that came through that pipeline. Based on primary completion timing on the active Phase 2 trials, readouts most plausibly land in late 2026 to 2027, assuming enrollment holds in the AD population where competition for trial-eligible patients is heavy.

FAAH (fatty acid amide hydrolase) and MAGL (monoacylglycerol lipase) are the two enzymes the brain uses to break down its own cannabinoids: FAAH chews up anandamide, and MAGL chews up 2-arachidonoylglycerol [8]. Block both and you let those natural cannabinoids accumulate at their receptors, turning up the volume on a system that normally helps modulate pain, anxiety, mood, and arousal. The pitch versus THC is that you only get an effect where endocannabinoids are actually being released, so in theory no global intoxication, less abuse potential, and a cleaner side effect profile than direct CB1 agonism. CNS penetration is the make-or-break PK question for this mechanism, and human exposure data has not been publicly disclosed. Pre-clinical descriptions from chemistry vendors and database entries characterize irafamdastat as a CNS-penetrant small-molecule inhibitor [11], but quantitative brain or CSF exposure in humans, particularly relative to plasma, is not in the public domain. For the MS spasticity indication, spinal cord exposure is required for activity, and that data gap is currently the largest unknown for both Phase 2 programs. The single-target version of this idea has a difficult history. The most infamous case was Bial's BIA 10-2474 in 2016, where a French Phase 1 trial killed one healthy volunteer and caused severe neurologic injury in five others; the toxicity was later traced to off-target effects rather than pure FAAH inhibition [9]. Pfizer's PF-04457845, a clean FAAH-only inhibitor, failed in osteoarthritis pain and PTSD-related sleep. The dual FAAH/MAGL bet is partly a response to those failures: hitting both nodes is thought to give broader endocannabinoid elevation without the CB1 receptor desensitization seen with chronic MAGL inhibition alone. Validation remains mostly preclinical and mechanistic; no FAAH or MAGL inhibitor has ever been approved.

The AD agitation trial (NCT06808984) is a Phase 2, placebo-controlled study randomizing 120 patients with a primary endpoint of change from baseline in Cohen-Mansfield Agitation Inventory (CMAI) total score [1]. That is the same primary endpoint Otsuka and Lundbeck used to win brexpiprazole's 2023 FDA approval, so the regulatory path is well-trodden. The concern is power: brexpiprazole's two Phase 3 trials enrolled 345 and 406 patients respectively to detect roughly 5-point CMAI differences, and 120 is light by that standard [3]. Either BMS is treating this as a proof-of-concept that will need a larger Phase 3 to confirm, or they expect a much larger effect size than brexpiprazole's. The MS spasticity trial (NCT06782490) enrolls 200 patients and uses change in Total Numeric-transformed Modified Ashworth Scale, Most Affected Lower Limb (TNmAS-MALL) as its primary outcome [2]. Ashworth-family scales are notoriously rater-dependent, but the transformed numeric version reduces some of that variability. Both trials are placebo-controlled, which is appropriate given no FAAH/MAGL drug has prior human efficacy data. Neither trial uses a biomarker-selected subpopulation, which keeps recruitment broader but accepts more efficacy heterogeneity in the readout. The specific dose and dosing regimen used in each Phase 2 trial (single or multiple oral dose levels, once-daily vs. twice-daily, titration schedule) are not detailed in the public ClinicalTrials.gov record we reviewed, beyond the route being oral. For a CNS drug where target-engagement and safety margin are both dose-sensitive, this is information a careful reader would want before drawing conclusions from the readouts.

Our model estimates a 11% chance this drug is eventually approved. It starts from the historical base rate for Phase 2 drugs in this area (about 24%), then adjusts using ten facts about the trial and sponsor. What moves the number most: it is helped by more secondary endpoints than usual and the sponsor's strong record of getting drugs approved; it is held back by weak or limited earlier-phase results and heavier-than-usual blinding. The other facts land near average for this stage, so they leave the estimate roughly where the base rate put it.

Efficacy risk is the single biggest factor. CMAI in AD agitation runs notoriously high placebo response rates, often 20 to 30% improvement in placebo arms, and 120 patients gives limited margin to separate from that noise. The MS spasticity trial faces a similar problem with Ashworth-family scales, which have low test-retest reliability. Safety risk carries class history. The BIA 10-2474 fatality in 2016 was traced to off-target effects, not pure FAAH inhibition, but it set a regulatory expectation of careful CNS and hepatic monitoring for anything in this space [9]. Anandamide signaling overlaps with body-temperature regulation, appetite, and learning, so chronic dosing can pull on multiple systems at once. Liver enzyme elevation is the most likely on-target watch item; dose-limiting CNS effects are the off-target watch item. The absence of public CNS PK and dose-selection rationale (see mechanism and trial_design) makes it harder to pre-judge the safety margin from outside. Execution risk: both trials run through a single sponsor with no co-developer or partnership disclosed. If BMS deprioritizes the asset during portfolio review, there is no backup mechanism for development continuity. The Celgene-of-record naming may matter for internal politics at BMS, and we have not located a recent BMS R&D Day slide that explicitly ranks irafamdastat against other neuroscience pipeline assets, so its priority inside the BMS portfolio is not transparent to outside readers. Commercial risk: brexpiprazole (Rexulti) is already approved for AD agitation and is prescribed off-label heavily in nursing homes despite the dementia black box on antipsychotics. Payers will likely demand step therapy through brexpiprazole and generic atypicals before covering a novel agent. To justify branded pricing, BMS-986368 would need to show either superior efficacy or a cleaner safety profile, particularly on sedation, falls, and EPS (extrapyramidal symptoms, the movement-disorder side effects from dopamine receptor blockade). Composition-of-matter patent timing is also material to the commercial case but is not detailed here; the asset originated from an Abide Therapeutics / Celgene chemistry program, and the full prosecution and Orange Book picture once an NDA is filed will determine effective market exclusivity.

Worth watching, not yet signal. The asset is too early to move BMY stock and too dependent on a single Phase 2 readout in a brutal indication to bet on commercially. The data point to wait for is the AD agitation CMAI readout, plausibly tracking for late 2026 to 2027 based on the Phase 2 primary completion timing on NCT06808984 [1]. A CMAI separation of 5 points or more with no serious hepatic or neuropsychiatric signals would make this interesting; anything less, particularly given the small sample size, will read as inconclusive even if directionally positive. The MS spasticity readout has a partial validation tailwind: Sativex/nabiximols won European approval for MS spasticity in 2010, which establishes that CB1 receptor engagement in the spinal cord can reduce spasticity in this population. Important distinction though: Sativex is a direct CB1/CB2 partial agonist (plant-extracted THC:CBD), so it validates the receptor system and the indication but it does not validate the enzyme-inhibition approach. The specific open question for BMS-986368 is whether FAAH/MAGL inhibition raises local endocannabinoid tone enough to produce comparable spinal-cord receptor activation without the systemic exogenous cannabinoid exposure that limits Sativex. A positive MS spasticity signal would establish that irafamdastat engages the endocannabinoid axis meaningfully in humans, but it would not on its own validate the AD agitation thesis. For BMS, this is one of many neurology shots from the Celgene integration; the company's near-term story is dominated by Eliquis and Revlimid loss of exclusivity, not by Phase 2 endocannabinoid drugs [4]. Check back at the next BMS R&D update or AAIC presentation for any guidance shift on Phase 2 timing, where this asset falls in the disclosed neuroscience priority stack, and again when the first CMAI data drops.