# Lesson 10: Psychedelics and Addiction: Alcohol, Tobacco, Opioids and Community Models

*Companion-podcast transcript • Sarah & Kiffer*  
*Office Hours episode to listen to after working through the lesson*

---

**Sarah:** Two trials gave psilocybin to people with alcohol use disorder. In the first, heavy drinking days over thirty-two weeks were about ten percent with psilocybin and about twenty-four percent with an active placebo. In the second, a single dose made no difference at four weeks or at six months.

**Kiffer:** People usually ask which trial is right. My answer is that both probably are. They gave different amounts of the drug and of therapy, to people at different points in their drinking, so they answered different questions.

**Sarah:** That's a comfortable answer. It gets less comfortable when somebody has to decide which result to build on.

**Kiffer:** Agreed, and the way through is to line the trials up and ask what each one actually tested.

**Sarah:** Welcome back to Office Hours. I'm Sarah.

**Kiffer:** And I'm Kiffer. Trials that seem to disagree come up all through Lesson ten, Psychedelics and Addiction: Alcohol, Tobacco, Opioids and Community Models, so let's start with those two.

**Sarah:** Question one. Why do the psilocybin trials in alcohol use disorder conflict, and what should we make of a trial that finds nothing?

**Kiffer:** Line them up and the differences are large. Bogenschutz and colleagues gave two psilocybin sessions inside twelve weeks of motivational and cognitive behavioural therapy, to people who were still drinking, and compared them with diphenhydramine, an antihistamine used as an active placebo. The Swiss trial reported by Rieser and colleagues gave one twenty-five milligram dose with five sessions of brief psychotherapy, to people who had just finished withdrawal treatment, and its placebo was mannitol, an inactive substance. One trial asked whether psilocybin helps people cut down. The other asked whether one dose helps people who have stopped drinking stay stopped.

**Sarah:** The placebo detail is the one that caught my attention. With an inactive placebo, people should find it even easier to tell which group they're in. So if expectancy were doing most of the work, you'd predict that the Swiss trial would find a difference, and it didn't.

**Kiffer:** That's a useful inference. Broken blinding makes room for expectancy effects, and if participants in the Swiss trial could tell which group they were in, its result suggests that this alone doesn't guarantee a positive finding. I'd add one caution. The Swiss trial analyzed thirty-seven people, so a modest benefit could have gone undetected. A null result from a small trial is evidence about one regimen in one group of people, and it argues for testing the features that differ, such as the number of doses and the amount of therapy.

**Sarah:** And in the larger trial, the control group improved a great deal too. Heavy drinking days in the diphenhydramine group fell from about forty-nine percent at screening to about twenty-four percent over follow-up, with no psilocybin at all.

**Kiffer:** That reflects therapy, trial attention and natural change. Nearly every trial we've discussed gives the drug inside a program of psychotherapy, and the only one that varied the drug and the therapy separately was the English ketamine trial. So when we say psilocybin worked in the larger trial, we mean it added something on top of a lot of therapy.

**Sarah:** Let's put the headline number to work. The difference was 13.86 percentage points over a window of two hundred twenty-four days, which works out to about thirty-one fewer heavy drinking days per person. The confidence interval for that difference ran from 3.0 to 24.7 points. What range of heavy drinking days does the interval correspond to? Take a few seconds, and pause the audio if you'd like more time.

*(Pause)*

**Kiffer:** Three divided by one hundred, times two hundred twenty-four, is about 6.7 days. And 24.7 divided by one hundred, times two hundred twenty-four, is about 55.3 days. So the data fit anything from about a week to nearly two months fewer heavy drinking days per person over the period.

**Sarah:** And the top end looks odd to me. An upper limit of 24.7 points is a little more than the control group's own figure of 23.6.

**Kiffer:** That's worth noticing, and there's a straightforward explanation. The 23.6 is an estimate too, with its own uncertainty, and the interval is for the difference between the two true figures. A true figure of twenty-nine percent in the control group and four percent with psilocybin, both values the data allow, would give a difference of twenty-five points. So the top of the interval describes a very large effect that the data can't rule out. The practical reading is that one trial of ninety-five people, stopped at about half its planned size when COVID-19 halted recruitment, leaves a very wide range.

**Sarah:** So it fits that a 2026 meta-analysis called the evidence mixed.

**Kiffer:** It does. It pooled six trials with two hundred two participants in all, only three of which were randomized, and found moderate short-term advantages for psilocybin, with only one drinking measure still significantly better at longer follow-up. That echoes the LSD trials from 1966 to 1970, where the benefit was present at six months and no longer significant by twelve. How long any benefit lasts is the question the field hasn't yet answered.

**Sarah:** Question two. How much should a large odds ratio from a small trial move us? I'm thinking of the smoking trial, where seventeen of forty-two people quit with psilocybin and four of forty with the nicotine patch.

**Kiffer:** That's an odds ratio of 6.12, with a confidence interval from about two to about twenty-three. Here's a way to feel how fragile it is. Suppose, hypothetically, that four more people in the patch group had quit, so eight of forty. What would the odds ratio be? Sarah's going to try it aloud, and everyone else can try it first. Take a few seconds.

*(Pause)*

**Sarah:** Okay. Seventeen out of forty-two is 0.405, and eight out of forty is 0.20. Then 0.405 divided by 0.20 is about 2.02. So the odds ratio would drop from about six to about two.

**Kiffer:** Compare that with the real trial. With four quitters in the patch group, the odds ratio was 6.12 and the risk ratio was 4.05. Which of those two did you just calculate?

**Sarah:** The risk ratio. I divided the proportions. For odds, I need the people who quit over the people who didn't. Seventeen over twenty-five is 0.68, and eight over thirty-two is 0.25. Then 0.68 divided by 0.25 is 2.72. So the odds ratio would be about 2.7, and the risk ratio about two.

**Kiffer:** That's right, and it passes the check from the lesson. Whenever the treated group does better, the odds ratio sits above the risk ratio, and further above when the outcome is common, so 2.7 above two is what you'd expect. Four extra quitters in a group of forty would cut the odds ratio by more than half. With Woolf's approximation, the method the lesson's calculator uses, the interval would run from about one to about seven, right at the edge of no difference.

**Sarah:** And the number needed to treat would go from about three to about five, because the risk difference shrinks from about thirty points to about twenty.

**Kiffer:** Which is why the authors called it a pilot. A pilot's job is to estimate an effect for a larger trial, and when one arm has four successes, a handful of people either way changes the estimate a lot. So a large odds ratio from a small trial supports running a bigger trial, and it gives a weak basis for headlines about quitting.

**Sarah:** Before question three, a short content note, because this part discusses deaths after ibogaine use and deaths from opioid overdose. If you'd like support, you can call or text nine eight eight, the Suicide Crisis Helpline, at any time. The Hope for Wellness Help Line for Indigenous people is at one eight five five, two four two, three three one zero. So, question three. For ibogaine and opioid use disorder, is the bigger danger the heart, or what happens after treatment?

**Kiffer:** Both are real, and they work on different time scales. The cardiac risk is concentrated in the hours and days after a dose. The risk after treatment is the one Canada's national guideline describes for withdrawal management alone. Tolerance falls, and a return to use then carries a higher risk of overdose death.

**Sarah:** On the cardiac side, ibogaine and its long-lasting metabolite, noribogaine, block hERG potassium channels in heart muscle, and the QT interval lengthens. A forensic review covered nineteen known deaths outside West Central Africa between 1990 and 2008, from one and a half to seventy-six hours after ingestion. And ibogaine is usually offered as a single session that ends withdrawal, which is the situation the guideline warns about. In the Mexican cohort, only half of the participants reported no opioid use in the month after treatment.

**Kiffer:** Let's put rough sizes on the two risks, with a hypothetical group built from two real figures. A preprint posted in June 2026, which hasn't been peer reviewed, pooled records from eleven international clinics and reported six deaths within seventy-two hours among ten thousand three hundred eighty-two people treated for opioid use disorder. And a 2017 review that pooled cohort studies found all-cause mortality of 11.3 per thousand person-years during methadone treatment and 36.1 after people left it.

**Sarah:** Those are two different kinds of number, one per treatment and one per year of follow-up.

**Kiffer:** They are, and that's the point of the exercise. Imagine a hypothetical thousand people on methadone who each have one ibogaine treatment and then leave methadone. If the clinic figure and the cohort death rates applied to them, how would deaths around the treatment compare with the extra deaths over the following year from being out of agonist treatment? Take a few seconds.

*(Pause)*

**Sarah:** Six divided by ten thousand three hundred eighty-two is about 0.6 deaths per thousand treatments, with an exact upper limit of about 1.1 per thousand from the lesson's calculator. For the year afterward, my first instinct was to set 0.6 beside 36.1, but some of those thirty-six would have died even in treatment. A thousand people followed for a year gives roughly a thousand person-years, so I'd expect about eleven deaths in treatment and about thirty-six out of it. The excess is about twenty-five, more than twenty times the figure for deaths around treatment, even at its upper limit.

**Kiffer:** That's the arithmetic, and you caught the units. For people on buprenorphine, the rates were 4.3 and 9.5, an excess of about five per thousand, still several times that figure, though it comes from only three cohorts. The example also has traps. The clinic figure is a floor, because the clinics took part voluntarily and the deaths weren't independently adjudicated, and its upper limit allows only for chance. The 36.1 comes from other cohorts of people who left methadone for any reason, and part of the gap may reflect who leaves. If ibogaine kept people abstinent, their risk might be lower, and if many returned to use with lost tolerance, it might be higher. Nobody has measured it.

**Sarah:** So the calculation leaves the safety question open, and it shows which outcome a trial can't afford to skip.

**Kiffer:** That's how I'd read it. Public funding for trials, like the fifty million dollars Texas committed in June 2025, to be matched by private investment, seems reasonable to me if the trials are built around that outcome: a comparison group, cardiac screening and monitoring, a planned link to agonist treatment for anyone who wants it, and overdoses counted after treatment. A trial that follows people for a month would repeat a gap in some of the observational studies. HSCI 402 develops the agonist treatments that would sit beside it.

**Sarah:** Before question four, another short content note, because this part touches on concerns about the appropriation of Indigenous medicines. You can call or text nine eight eight at any time, and the Hope for Wellness Help Line for Indigenous people is at one eight five five, two four two, three three one zero. Question four. What can community and group models of psychedelic care claim from the evidence so far?

**Kiffer:** Let's take the two British Columbia examples in turn, starting with the older one. In 2013, Gerald Thomas and colleagues reported a preliminary observational study of the Working with Addiction and Stress retreats, delivered in a rural First Nations community. The retreats combined four days of group counselling with two expert-led ayahuasca ceremonies.

**Sarah:** And the results differed across measures. Among twelve participants assessed before and six months after, hopefulness, empowerment, mindfulness and the meaning and outlook parts of a quality of life measure improved. Self-reported alcohol, tobacco and cocaine use declined, and the fall in problematic cocaine use was statistically significant, while cannabis and opiate use did not decline. The authors called the study preliminary, said this form of therapy appears to be associated with improvements in factors related to problematic substance use, and asked for more rigorous research.

**Kiffer:** That careful reporting is worth noticing. With twelve people, no comparison group and a program that combined counselling, ceremony and community, the study can't credit any single element. Its contribution is documenting a program delivered in a First Nations community, with outcomes that include hope, empowerment and meaning. Questions about who designs programs like this, who holds the data and how results are shared belong to Indigenous research ethics, which the series' graduate course develops.

**Sarah:** And those questions connect to the concerns from many Indigenous Nations that an Indigenous-led group described in 2023: cultural appropriation, a lack of recognition of the sacred place these medicines hold, exclusion from research and practice, and the patenting of traditional medicines. The group's principle of Respect calls for relationships with Indigenous knowledge holders, adherence to regional protocols, and free, prior and informed consent.

**Kiffer:** Then there's Roots to Thrive in Nanaimo, which embeds three ketamine sessions in a twelve-week therapy group it calls a community of practice. Its outcome evaluations concern health care workers and firefighters with post-traumatic stress, depression or anxiety, and a 2026 qualitative report describes psilocybin groups for people with terminal illness. None of its published evaluations reports outcomes for substance use disorders.

**Sarah:** Here's one for everyone. In a pilot comparison, fourteen health care practitioners did the group program alone and thirty-five did it with ketamine, and their improvement didn't differ significantly. Does that show ketamine adds nothing? Think of two reasons to be careful. Take a few seconds.

*(Pause)*

**Kiffer:** Two reasons stand out. The allocation wasn't random, so however people ended up in each group, they may have differed in ways that matter, such as how severe their symptoms were. And fourteen people is a very small group, so a real but moderate difference could easily go undetected. A third point is that both groups improved and nobody went without the program, so we can't tell how much of either group's improvement came from the program itself.

**Sarah:** So the fair summary is that the group may do much of the work, which fits the protocol's description of ketamine as an amplifier, and what ketamine adds is still an open question. And none of this concerns addiction.

**Kiffer:** Right. Both examples support studying these models in people with substance use disorders, ideally in randomized trials, with communities shaping any program that draws on their ceremonies. Neither gives a basis for advertising them as addiction treatment today.

**Sarah:** Okay. Now the part where we disagree. The smoking trial didn't try to blind anyone, and I think that was the right call.

**Kiffer:** Let me set out the authors' reasoning first. Everyone knew whether they received psilocybin or the patch, and the authors acknowledged that expectancy could have contributed. They argued that an unblinded comparison was reasonable given how poorly double-blind psychedelic trials keep the blind.

**Sarah:** And the record supports them. In the larger alcohol trial, people guessed their allocation after 93.6 percent of first sessions, even with an active placebo. A trial that is described as blinded when the blind has failed may mislead readers more than one that says openly what everyone knew. And psilocybin compared with the patch is a question that people who smoke and their clinicians actually face.

**Kiffer:** That's a fair argument for an open comparison with a real treatment. My worry is the comparison group. People who sign up for a psilocybin trial and get the patch may be disappointed and try less hard, which would push the patch result down and widen the gap.

**Sarah:** But the patch group did about as well as you'd expect. Ten percent prolonged abstinence is a little above the roughly eight percent the authors cite for nicotine replacement alone in a recent meta-analysis. That doesn't look like a group that gave up.

**Kiffer:** It's a reasonable point. Then again, that group also had thirteen weeks of cognitive behavioural therapy, so you might have expected a bit more than nicotine replacement alone achieves. The ten percent fits both readings, and that's what bothers me. In an open trial, the numbers can't tell us which reading is right.

**Sarah:** So where do you land?

**Kiffer:** I'd still try to blind where it can be done, for example with an active placebo, and I'd ask every participant before dosing what they expect, so that expectancy can be measured and taken into account. For a pilot like this one, an open comparison with biological checks seems reasonable to me. For the larger trial, I'd want those expectation measures in both groups.

**Sarah:** I'm more comfortable with open designs than Kiffer is, but we agree on the practical rule. When a trial's blind is weak or absent, look for an outcome that's checked biologically, a comparison treatment that participants could believe in, and some measure of what people expected going in.

**Kiffer:** Let's pull it together with three things to take away.

**Sarah:** First, trials that seem to conflict often tested different things. The psilocybin trials differed in doses, therapy and the stage of drinking, and one trial's interval, from about a week to nearly two months of heavy drinking days, shows how little a single small trial can settle.

**Kiffer:** Second, large ratios from small trials are fragile. Four more quitters in a patch group of forty would have cut the smoking trial's odds ratio from about six to under three, so read a pilot as an estimate for the next trial to test.

**Sarah:** And third, for opioid use disorder, safety extends past the session. The risk around the dose is concentrated in the days after it, the risk from lost tolerance follows, and in our hypothetical example the excess deaths from leaving agonist treatment were the larger number. Any program has to keep people connected to care and count what happens afterward.

**Kiffer:** If you'd like more practice, rework the smoking trial problem from this episode with six quitters in the patch group, and check that your odds ratio comes out above your risk ratio.

**Sarah:** Next time, it's Lesson eleven, Access, Regulation and Practice in Canada and the World, where we look at how a molecule becomes a medicine and at the routes by which people in Canada can lawfully reach these treatments.

**Kiffer:** Take care, everyone.

**Sarah:** See you in Lesson eleven.
