Psilocybin’s best result in alcohol use disorder is four years old, and it ends at week 36.
That boundary has not moved. The randomized trial that produced it, run by Michael Bogenschutz at NYU and published in JAMA Psychiatry in 2022, measured drinking through week 36. No controlled trial has published drinking results beyond that point, in a search of PubMed and ClinicalTrials.gov on October 25. A paper published October 15 in Drug and Alcohol Dependence is the program’s first published look past the line, and it is uncontrolled: among patients still drinking heavily, drinking fell after one more psilocybin session at month one and then returned to where it started.
That is not evidence that the drug fails. Everyone in the extension knew what they were taking, there was no comparison group, and the people in it had already been through the 36-week trial. What it shows is how little the field can yet say about the property that would make a few-session treatment valuable: whether its effect lasts.
What the 2022 trial established, and where it stopped
The trial randomized 95 adults with alcohol dependence and dosed 93 of them, 48 with psilocybin and 45 with diphenhydramine, an antihistamine used as an active placebo. Patients received two dosing sessions, the first at 25 mg/70 kg and the second at 25 to 40 mg/70 kg, within 12 sessions of motivational enhancement and cognitive behavioral therapy. They were outpatients who were still drinking when they enrolled.
Over weeks 5 to 36, heavy drinking days made up 9.71 percent of days on psilocybin against 23.57 percent on diphenhydramine, a difference of 13.86 percentage points with a confidence interval of 3.00 to 24.72, a moderate effect. That is the number the field cites, and it holds up.
The qualifications around it are cited less often. Complete abstinence across weeks 5 to 36 was 22.9 percent against 8.9 percent, which did not reach significance (p=.06); for abstinence, the significant separation came only in the final four weeks, in an analysis with no correction for multiple comparisons. Enrollment stopped at 95 patients when the pandemic suspended recruitment, against a design of up to 180, so the trial is smaller than planned.
The 36-week endpoint was the design, not a hidden limitation. But it means the longest controlled observation of psilocybin’s effect on drinking is roughly eight months.
The follow-up, and what it found
The new paper reports an open-label extension written into the original protocol: patients who completed the blinded phase were offered one more session. Fifty-nine received a single open-label dose of 25 to 40 mg/70 kg with four hours of manualized therapy; 29 had received psilocybin in the blinded phase and 30 diphenhydramine. Follow-up ran four months.
Psilocybin was well tolerated, and no serious adverse events occurred. On drinking, the published result is narrow. Among participants with higher drinking intensity when they entered the extension, heavy drinking days, drinks per day and drinking days fell at month one and, in the authors’ words, “returned to entry levels thereafter.” The authors also describe “maintained remission among most participants who entered abstinent.” The published abstract does not say how many of the 59 were in either group, so neither finding can be sized.
The authors do not claim to know why the reductions did not hold. Whether the limited durability reflects “the lower intensity of extension treatment, the relative treatment resistance of residual drinking, differences in motivation, or other design-related factors cannot be determined from these data,” they write. That sentence names the confound that runs through this literature.
The design could not separate those explanations. There was no control group and no blinding, and the patients chose to enter after completing a demanding trial. A drop at month one followed by a return to entry levels fits a real but short drug effect, expectancy that wore off, or a mixture. In a group defined by heavy drinking at entry, some of the month-one drop could also be regression toward more typical drinking, which would make the short-lived benefit smaller still. Maintained remission in people who entered abstinent says little on its own, because nothing shows how they would have done without the extra dose.
Two more controlled trials, pointing different ways
Outside the NYU program, the controlled record consists of two small trials published last year, and they disagree.
In Zurich, Rieser and colleagues randomized patients who had completed alcohol withdrawal treatment in the prior six weeks to a single 25 mg psilocybin dose or a mannitol placebo, with five sessions of psychotherapy. Among the 37 patients analyzed, there was no significant difference in abstinence at four weeks (16.80 days against 13.80, p=0.55) or in alcohol use per day, and the effect sizes were smaller still at six months. The authors concluded that a single dose with five psychotherapy sessions may not be enough for severely affected patients after withdrawal.
In Nîmes, Luquiens and colleagues ran a feasibility trial in a French inpatient addiction program, randomizing 30 recently detoxified patients with depressive symptoms two to one to two sessions of 25 mg or 1 mg, three weeks apart, on top of intensive relapse prevention. Feasibility was the primary outcome. On secondary measures at 12 weeks, 11 of 20 patients on 25 mg were abstinent against 1 of 9 on 1 mg, with one control lost to follow-up (p=0.043). Relapse rates were 35 percent and 50 percent, a difference that was not statistically significant.
Set beside NYU, a tempting pattern appears. The two controlled trials with positive results used two doses inside substantial treatment programs. The single-dose Zurich trial found no difference, and the single-session extension saw only a brief change.
It should not be called a pattern yet. Dose count, therapy intensity, setting and population all move together across these studies, so nothing here separates them, and three small controlled trials, the largest with 93 dosed patients, cannot support a dose-response or intensity-response claim. One alternative reading is that psilocybin does more for people who are still drinking than as relapse prevention after detox. The strongest result enrolled people who were still drinking, while Zurich enrolled patients after withdrawal treatment and the extension came after a completed course of treatment. Nîmes cuts against that reading, since its patients had already detoxified, though it is a 30-patient feasibility trial.
Almost everyone knew what they had taken
All three controlled trials measured blinding, and in all three almost everyone guessed correctly. In the 2022 trial, participants guessed their assignment correctly after 93.6 percent of first sessions and 94.7 percent of second sessions, and therapists after 92.4 and 97.4 percent. In Zurich, 92 percent of participants guessed correctly. In Nîmes, 93.3 percent did, and four control participants refused their second session after working out which arm they were in; one of them took MDMA on their own.
A 25 mg psilocybin dose is hard to mistake, and active placebos have not solved that. It means every durability figure in this literature is entangled with expectancy, and an open-label extension sits at the far end of that problem and does not escape it.
What the pooled evidence says, and how far out it reaches
A systematic review published October 8 in the Journal of Psychopharmacology gathered six trials totaling 202 patients, three of them randomized. Against controls, psilocybin showed moderate short-term advantages on drinks per day, drinking days and heavy drinking days. At what the authors call mid- to long-term follow-up, only drinking days remained significantly better, and the odds of maintaining abstinence were higher.
The review defines that mid- to long-term window as 4 to 28 weeks. Its longest analysis, in other words, ends earlier than the 36 weeks the anchoring trial reached.
The next trials stop their comparisons sooner
The two largest trials now enrolling will shape the next several years, and neither is designed to extend the controlled record.
ERPPAD, the first Phase 3 trial in this indication, is sponsored by the university hospital in Nîmes and publicly funded through France’s national hospital clinical research program. It dosed its first patient on July 1 and plans 172 patients at eight French hospitals. It keeps the two-dose structure, comparing two 25 mg sessions three weeks apart with two 3 mg sessions, in patients with severe alcohol use disorder and depressive symptoms whose last drink was 10 to 60 days before enrollment. Its primary endpoints, relapse and time to relapse, are measured at month 6. After that, according to the hospital, every participant is offered an optional third 25 mg dose, with follow-up extended to 12 months. Once participants in either arm can choose a full dose, the later months are no longer a randomized comparison, the same limit the NYU extension ran into. Primary completion is currently estimated for June 2030.
NYU’s own follow-on, a 200-patient trial, is mechanistic. It tests a single 30 mg dose against placebo with four supportive sessions, and about half of its 23 primary outcomes are brain-imaging measures taken within days or weeks of dosing. Drinking is measured as the percentage of days without heavy drinking through week 24. Primary completion is estimated for May 2030.
Month 6 is about week 26. Both comparisons end before week 36. Two smaller registered trials list longer drinking outcomes. A 90-patient placebo-controlled trial in Sydney names a 52-week primary outcome, but its registry record has not been updated since June 2024 and lists no results. A 36-patient dose-comparison study at Brigham and Women’s Hospital in Boston plans relapse measures through week 48 and has not begun enrolling. Unless one of them reports first, the longest controlled look at psilocybin’s effect on drinking will still date from 2022 when the Phase 3 trial is scheduled to finish.
That is the gap the extension exposes without being able to fill it. The controlled case for psilocybin in alcohol use disorder rests on one trial of 93 dosed patients, with drinking measured to eight months, in a population that mostly knew what it had taken. The largest trials under way stop their comparisons sooner, and the one published look beyond the line had no control. How long the effect of a few sessions lasts has not yet been tested under conditions that could answer it.