Compass Pathways filed an 8-K on July 7 disclosing 26-week data from COMP006, its second pivotal Phase 3 trial of COMP360 psilocybin in treatment-resistant depression. The company’s framing is unambiguous: rapid onset and remarkable durability through at least six months. The underlying data support a real and clinically meaningful signal. They also come with four qualifications, printed on Compass’s own investor slides, that the word “remarkable” does not convey. Reading exactly what was measured, and how, is where this dataset earns its place as the most important psychedelic-industry filing of the summer.
What COMP006 is
COMP006 dosed 581 participants with treatment-resistant depression, randomized to 25 mg COMP360 (N=296), 10 mg (N=142), or a 1 mg control (N=143), two fixed doses three weeks apart. The population is genuinely severe: participants averaged more than three years in their current depressive episode and more than six lifetime episodes. The trial runs in three parts, blinded through Week 9, blinded through Week 26, then open-label to Week 52. The Week 6 primary endpoint was already public in February, a statistically significant 3.8-point MADRS separation between the 25 mg and 1 mg arms, p<0.001. This filing adds the six-month picture.
The headline number, read correctly
Compass states that 39 percent of the 25 mg arm achieved a clinically meaningful MADRS reduction of at least 25 percent at Week 6 and maintained that response on average through Week 26, up from 25 percent in COMP005’s single-dose design, a real improvement the company attributes to the second dose. A separate slide defining the Week 6 responder categories gives, out of an evaluable population of 269, 48 remitters, 68 responders or partial responders, and 153 non-responders. Add remitters and responders together and you get 116, which against that 269 denominator is 43 percent, not 39.
The two numbers are not in conflict once the denominators are made explicit. The 39 percent figure uses the full randomized 25 mg arm, all 296 patients, as its base; 116 divided by 296 is 39.2 percent. The 43 percent figure uses only the 269 patients captured in the Week 6 evaluable analysis, which appears to exclude roughly 27 patients who left the trial before that assessment. Counting those 27 as non-responders, rather than excluding them from the denominator entirely, is the more conservative and more standard convention in a chronic, hard-to-treat population where dropout itself is informative. Compass’s headline number, in other words, is the stricter of the two readings, not a favorable rounding. That is a genuinely reassuring finding about how the company chose to report the topline. It is also exactly the kind of detail a press release headline cannot convey and a slide deck can bury.
What “maintained through Week 26” is actually built on
Four qualifications appear directly on Compass’s own slides, and each narrows what the durability claim can support.
The Week 26 durability analysis is explicitly labeled a post hoc analysis, meaning it was not the trial’s pre-specified statistical test, and no p-value is presented at that timepoint. The Week 6 result is a powered, statistically significant finding. The six-month persistence of that result is a descriptive characterization, not an inferential one.
The Part B results, covering Week 9 through Week 26, are based on observed data only, meaning the change-from-baseline figures reflect the patients still in the trial and being measured at each point, with no imputation for those who dropped out. That is the standard completers-based approach many trials use for open-ended follow-up, and it is not improper, but it means the “maintained response” line describes the patients who stayed, not a full intention-to-treat accounting through six months.
Most consequentially, 58 percent of the 25 mg arm received retreatment after Week 9, either another dose of study drug or, per the protocol, a permitted antidepressant. So the population whose response was “maintained through Week 26” is, for a majority of patients, not a population that received one or two psilocybin doses and was simply followed. It is a population where most participants received an additional therapeutic intervention along the way. That does not invalidate the finding, retreatment-driven durability is still durability, and a meaningful share of initial responders who had not fully remitted by Week 6 went on to remission after retreatment in Part B. But it changes what is being claimed. The six-month result describes an ongoing treatment protocol allowing supplemental dosing, not the persistence of a single intervention’s effect, which is the claim the “one or two doses” framing in Compass’s own messaging can imply to an unfamiliar reader.
One reassuring methodological note sits alongside these caveats: Compass states that separation between the 25 mg and 1 mg arms was maintained through the randomized, blinded Part B period out to Week 26, meaning the comparison during that stretch was still conducted under blinding, a real point in its favor given how often this desk has flagged unblinding as the central threat to subjective psychedelic-trial endpoints.
Safety, reported plainly
Through 26 weeks, any treatment-emergent adverse event occurred in 94.6 percent of the 25 mg arm, versus 93.0 percent at 10 mg and 86.7 percent at 1 mg, consistent with a drug whose acute perceptual effects, nausea, headache, and dizziness among them, are common and expected on dosing days. Serious adverse events were low and similar across arms, 5.7 percent at 25 mg versus 6.3 percent at the 1 mg control and 3.5 percent at 10 mg. Within the 25 mg arm, serious adverse events included suicidal ideation in 4 of 296 participants (1.4 percent), a rate not higher than the 1 mg control arm’s 2.8 percent, and one suspected completed suicide (0.3 percent), which the investigator determined, based on the circumstances and timing, was not related to treatment. The 1 mg control arm recorded one instance of suicidal behavior and one suicide attempt. These are small counts in a population selected for treatment-resistant depression, a condition independently associated with elevated suicide risk, and Compass reports its independent Data Safety Monitoring Board found no meaningful imbalance in suicidality between arms. The numbers should be read in that context, as safety monitoring in a high-risk population, rather than as evidence in either direction beyond what such small samples can support.
Where this sits in the pipeline
Compass says its rolling NDA submission and initial FDA review are underway, with final submission targeted for the fourth quarter of 2026, backed by a national priority review voucher, and commercial launch, subject to approval and DEA rescheduling, anticipated in the first half of 2027. That timeline places COMP006’s Week 26 data as very likely the last major efficacy readout before the application is complete.
This is also the dataset the desk’s recurring duration-and-economics thread has been waiting for. The entire case for psilocybin therapy rests on a small number of supervised sessions producing benefit that lasts, since the session itself, hours long, staffed, and monitored, is the expensive part. COMP006 is evidence that benefit can persist to six months. It is not yet evidence, at the statistical standard the word “remarkable” implies, that it persists independent of the retreatment many patients also received. That distinction will matter directly to Compass’s label, to how payers eventually price a psilocybin course, and, by comparison, to how the field reads Definium’s competing lysergide program, where the underlying molecule’s much longer acute duration makes the same durability-versus-session-cost tradeoff sharper still.
The frame
Nothing here should be read as a rebuttal of Compass’s data. A 39 percent response rate maintained on average through six months, in a chronically treatment-resistant population, with a safety profile the company’s own DSMB found unremarkable relative to prior studies, is a genuinely positive result and the strongest evidence yet behind the most advanced psilocybin drug candidate. But durability, in the sense that matters for how this drug will actually be used and paid for, is supported by this filing and not yet established to the standard “remarkable” suggests. Post hoc, observed data only, and a majority-retreated population are not disqualifying details. They are the specific, printed-on-the-slide reasons the honest claim is narrower than the press release. The NDA is coming this year regardless. What Week 26 actually shows, read past the topline, is the more useful story than what the topline says it shows.