Last week's H1 review walked through Europe's access-first wins, including Switzerland's real-world psilocybin numbers, which I led with in Issue 1. It did not weigh those wins against two controlled academic readouts from the same half-year that pulled in the opposite direction. Both predate this newsletter; neither has appeared here, and together they locate the soft spot in the optimistic case. So here they are.

Start with treatment-resistant depression (TRD), the indication every commercial programme is chasing.

EPIsoDE missed. The German trial, run across Charité Berlin and the Central Institute of Mental Health in Mannheim, was triple-blind and used an active placebo (nicotinamide) to break the unblinding that flatters this field. It enrolled 144 TRD patients. The week-six response rate on a single 25mg dose was 17%, against 10.6% on placebo, a gap that was not statistically significant (OR 1.73, 95% CI 0.53-6.23, P = .19). Primary endpoint not met. It was published in JAMA Psychiatry in March and presented at ICPR in Haarlem on 5 June, inside a session on "negative clinical trials in psychedelic research." (JAMA Psychiatry)

Set that against the Zurich real-world cohort I led with in Issue 1, where clinician-rated scores fell well into the clinically meaningful range. Same disease, roughly the same drug. The open-label route looked convincing; the blinded trial could not separate psilocybin from its placebo. That does not prove expectancy explains the difference, since the two studies differ in more than their blinding, but it puts the burden of proof on the trial that was built to settle the question and didn't.

Karolinska adds the nuance. PSIPET, the first randomised psilocybin-for-depression trial in Sweden, did hit under active-placebo control, with rapid relief that held past three months. But its patients had ordinary recurrent depression, not the resistant kind. (Karolinska) Put the three readouts together and the pattern is uncomfortable for anyone selling TRD: psilocybin's signal holds up in open-label TRD and in controlled non-resistant depression, and softens exactly where the commercial value sits, in controlled trials in the resistant patients.

EPIsoDE's defenders point to its longer arc, an average 7.5-point HAMD-17 drop by week 12. The problem is structural. By that point every arm had crossed over to 25mg, so there is no untreated group left to compare against. (Research in Germany) Durability with no control group is a description, not a finding. To their credit, the EPIsoDE team presented these results at ICPR under that "negative trials" banner rather than dressing them up. The researchers are being straighter about the outcome than the field's boosters are.

This is why active placebos matter. When the comparator is inert, patients work out within minutes whether they got the drug, and expectation starts doing the work the molecule is supposed to do. Within TRD, the looser the design, the larger the effect looked: EPIsoDE controlled hard and found nothing, the Swiss real-world route had no control at all and looked strong. Keep that in mind every time a sponsor reports an open-label number.

Our position, unchanged from how I closed the H1 review: the access programmes are producing encouraging real-world signals, and a missed primary endpoint in a blinded trial is still a missed primary endpoint. Both things are true, and the second one is the one the field keeps underplaying.

17% vs 10.6%. EPIsoDE's week-six response rate for 25mg psilocybin against its active placebo in TRD, a gap that did not reach significance. A triple-blind trial with a credible placebo could not separate the drug from it at the primary endpoint, in the exact indication the field is built on. (JAMA Psychiatry)

BPL-003 (AtaiBeckley): the clock I flagged in H1 is now two weeks out. The H1 review listed first patients dosed in AtaiBeckley's ReConnection Phase 3 programme as a thing to watch. Here is the near-term test. After a successful FDA End-of-Phase 2 meeting in March, the company has reaffirmed Phase 3 initiation "in Q2 2026," most recently in its 12 May Q1 results. Q2 closes on 30 June, and no first-patient-dosed release has appeared. (AtaiBeckley IR)

One falsifiable prediction per issue. Scored in the next one.

AtaiBeckley will not announce a first BPL-003 Phase 3 patient dosed before 30 June. Expect initiation language (sites activated, screening open) at or near the wire, with confirmed dosing slipping into Q3. If a dosed-patient release lands this month, I mark myself wrong in Issue 3.

If something here is wrong, tell me, and I will correct it openly.

Thank you so much for reading.

Dhruv

Sources