# Lesson 9: MDMA-Assisted Therapy for PTSD: Evidence, Regulation and Research Integrity

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

---

**Sarah:** In the second phase three trial of MDMA-assisted therapy for PTSD, the people who took inactive capsules improved by almost fifteen points on the main symptom scale.

**Kiffer:** And the people who took MDMA improved by about twenty-four. Both groups had the same kind of two-therapist team, the same preparation and the same integration sessions. So the number that tells you what the drug added is closer to nine.

**Sarah:** That leaves two questions. Where did the fifteen points come from, and how far can we trust the nine when most participants could tell which capsules they'd taken?

**Kiffer:** Those two questions were at the heart of the FDA's review in 2024, and Sarah and I don't fully agree on what should follow from them.

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

**Kiffer:** And I'm Kiffer. The gap between twenty-four and fifteen sits at the centre of Lesson nine, MDMA-Assisted Therapy for PTSD: Evidence, Regulation and Research Integrity, so let's start with what the nine points measure.

**Sarah:** Question one. What does a difference of nine points on the CAPS-5 actually tell us?

**Kiffer:** Start with the design. In MAPP two, both arms received three preparatory sessions, three eight-hour medication sessions and nine integration sessions with a two-person team. The planned difference was the capsule. So the gap between arms, 8.9 points, estimates what MDMA added to an intensive psychotherapy. The full change in the MDMA arm, 23.7 points, mixes the drug with everything the two arms shared.

**Sarah:** And everything they shared is a lot. The placebo arm improved by 14.8 points.

**Kiffer:** Several things could produce that together: the therapy itself, many hours of attention from two therapists, the natural course of PTSD, regression to the mean, and expectation. The trial can't pull them apart, because every arm received all of them. So the accurate description of the benefit is the drug's addition to one particular therapy.

**Sarah:** Could a trial ever pull them apart? I'd want to know whether the drug helps without that therapy, or with a different one.

**Kiffer:** The FDA's final guidance from July 2026 notes that a factorial design may be useful for that. In such a trial, participants are randomized twice, once to drug or placebo and once to the therapy or no therapy, which gives four groups and lets you estimate each contribution. With MDMA, every group would still need safety monitoring during sessions, so in practice the comparison would be the psychotherapy compared with monitoring alone.

**Sarah:** Here's one for everyone listening. The CAPS-5 total adds up twenty symptom ratings, each from zero to four. Take a hypothetical person whose total falls by nine points. What might that look like in their symptoms? Take a few seconds, and pause the audio if you'd like more time.

*(Pause)*

**Kiffer:** There are many ways to make nine. Spread evenly, nine divided by twenty is 0.45, a little under half a step on every symptom. Or two symptoms could fall from four, extreme, all the way to zero, while one more drops a single step. Four plus four plus one is nine. For the person, those are very different changes. One is a general easing. The other might be nightmares and an exaggerated startle disappearing while everything else stays where it was.

**Sarah:** And the diagnosis could move differently again, because it depends on how many symptoms in each cluster are rated two or higher.

**Kiffer:** Right. A general easing could leave every cluster above its threshold, while losing the startle symptom could take the arousal cluster below the two symptoms it needs. It's also worth remembering that nine is a difference between two group averages, so it can't tell us how much any one person gained from the drug.

**Sarah:** Which brings me to a comparison that tripped me up. In a network meta-analysis of psychological treatments for PTSD, eye movement desensitization and reprocessing, or EMDR, had a standardized mean difference of about two compared with waitlist. The MAPP trials came in at 0.91 and 0.7. My first reaction was that EMDR works more than twice as well.

**Kiffer:** That's a common reading, and it runs into trouble once you look at what each trial compared, so let's build it with invented numbers. Imagine a hypothetical trial with three arms: a waitlist, the therapy with placebo, and the therapy with MDMA. The waitlist improves by four points, the therapy with placebo by fifteen and the therapy with MDMA by twenty-four. The standard deviation of the change is eleven points in every arm.

**Sarah:** So compared with the therapy plus placebo, the MDMA arm's effect size is nine divided by eleven, about 0.8. Here's the question for everyone. Using the same made-up numbers, what's the MDMA arm's effect size compared with the waitlist? Take a few seconds.

*(Pause)*

**Kiffer:** Twenty-four minus four is twenty, and twenty divided by eleven is about 1.8. The therapy with placebo, compared with the waitlist, comes out at eleven divided by eleven, exactly one. So the same invented treatment gets an effect size of 0.8 or 1.8, depending only on what it's compared with.

**Sarah:** And many psychotherapy trials use a waitlist, while the MAPP trials used an active treatment as the comparison. The populations differ as well. The MAPP trials stopped all psychiatric medications and excluded people with many medical and psychiatric conditions.

**Kiffer:** Yes, and ICER noted that inactive comparators may inflate effect sizes. It found no head-to-head trials and rated the evidence comparing MDMA-assisted therapy with trauma-focused psychotherapies as insufficient, and the MAPP two investigators themselves called for head-to-head studies of their therapy compared with cognitive behavioural therapy and prolonged exposure.

**Sarah:** Question two. In MAPP two's end-of-trial survey, forty-nine of the fifty-two people who answered in the MDMA arm believed they'd had MDMA. In the placebo arm, thirty-three of forty-four correctly believed they'd had placebo, nine thought they'd had MDMA, and two couldn't tell. How much of the nine points could expectation explain?

**Kiffer:** The FDA's position in 2024 was that the size of the bias "cannot be quantified," which is fair, because expectancy was never measured. We can answer a narrower question, though. How large would an expectancy effect have to be to explain the whole difference? Here's a deliberately simple, made-up model. Suppose the drug did nothing at all. Suppose that anyone who believes they got MDMA improves by some extra amount, call it E, compared with anyone who doesn't believe it.

**Sarah:** I'd like to try this one. Everyone else, here's the question: how big would E need to be to produce a difference of 8.9 points? Take a few seconds.

*(Pause)*

**Sarah:** Okay. In the MAPP two survey, ninety-four percent of the MDMA arm guessed correctly, and seventy-five percent of the placebo arm guessed correctly. The difference is nineteen percentage points. So E times 0.19 has to equal 8.9, and E is 8.9 divided by 0.19, which is about forty-seven points.

**Kiffer:** Before you go on, compare that number with the scale itself. Does forty-seven make sense?

**Sarah:** No. The average baseline score in MAPP two was thirty-nine. An expectancy effect of forty-seven would wipe out more than the average person's entire score. I see what I did. In the placebo arm, guessing correctly means believing you got placebo. I subtracted two shares that measure opposite beliefs.

**Kiffer:** Right. What the model needs is the share in each arm who believe they got MDMA.

**Sarah:** In the MDMA arm that's forty-nine of fifty-two, about ninety-four percent. In the placebo arm, two people were positive they'd had MDMA and seven thought so, which is nine of forty-four, about twenty percent. The difference is about seventy-four percentage points. 8.9 divided by 0.74 is about twelve. So believing you got MDMA would have to be worth about twelve CAPS-5 points.

**Kiffer:** Now compare twelve with the 14.8 points the whole placebo arm improved, with the therapy, the attention and the time all included. It comes close, although twelve is a gap between believing and not believing, so it could be shared between a boost for people who think they got MDMA and disappointment for people who think they didn't, which the FDA raised as a possibility. My own reading is that a purely expectancy explanation is a demanding one, because it needs a belief gap bigger than the ten points the trials counted as a response, and there's still plenty of room for expectancy to have inflated the estimate.

**Sarah:** And many participants knew what MDMA feels like. ICER reported that about forty percent of participants across the trials had used it before. There's also a trap in the model itself, isn't there? The survey was taken at the end of the trial.

**Kiffer:** Yes, and it's a big one. People who felt much better may have concluded they'd had MDMA, so belief at the end may be partly a result of improvement as well as a cause. That's one reason to measure expectancy early, and the FDA's final guidance on psychedelic trials, from July 2026, suggests an expectancy questionnaire before randomization as well as at the end of treatment, along with blinding questionnaires, and says alternatives to an inert placebo may be considered.

**Sarah:** To be fair to the investigators, they made points of their own. The arms didn't separate after the first session, the placebo-arm dropouts didn't cluster after it, and the raters scoring the CAPS-5 were blind.

**Kiffer:** Those observations make a large expectancy effect less likely, and they leave room for a moderate one. The calculation can't settle which, because it only tells us how large the effect would have to be to explain the whole difference.

**Sarah:** Before question three, a short content note, because this part discusses sexual misconduct by a therapist in a clinical trial. 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. And for sexual violence in British Columbia, VictimLinkBC is at one eight hundred, five six three, zero eight zero eight. So, question three. MAPS says that at a Canadian trial site, every session had two therapists and was recorded, and the study had monitoring and supervision. Why did none of that protect the participant?

**Kiffer:** I'll stick to what MAPS and CBC News have reported. According to the statement MAPS published in May 2019, at a small phase two study in Vancouver, one of the two co-therapists began a sexual relationship with a participant around the end of 2015 or early 2016, after her active treatment and while she was still enrolled in follow-up. MAPS said the other therapist reportedly learned of it, tried unsuccessfully to stop it, and didn't report it. MAPS says it learned of it in October 2018, when the participant forwarded complaints she had filed with British Columbia's regulatory colleges earlier that year.

**Sarah:** And the session recordings existed the whole time.

**Kiffer:** They did. MAPS says that in November 2021, a journalist who had viewed the recordings interviewed its staff, and it opened a compliance review two days later. In March 2022 it reported a provisional finding that both therapists had substantially deviated from the treatment manual during her treatment. CBC News reported that MAPS staff hadn't viewed the recordings until November 2021, about six years after they were made. MAPS itself listed the safeguards in place: two providers in every session, video recording, study monitoring and clinical supervision. It said none had been sufficient.

**Sarah:** So if I go through the layers, by MAPS's provisional finding the second person in the room was also involved in the departures from the manual. The recordings weren't watched in time to deter or detect anything. And the relationship began outside the study sessions, where session-based safeguards don't reach.

**Kiffer:** And one layer was missing altogether. CBC reported that one of the therapists wasn't a licensed psychologist in British Columbia, and because the province didn't regulate therapists or counsellors in 2022, no regulatory body could investigate that therapist. The route that worked was the participant's own complaint, years later. She has spoken publicly, and CBC reported her view that if these drugs are legalized for use in psychotherapy, there needs to be a strong system in place to keep patients safe.

**Sarah:** The case reached the scientific record as well. In August 2024, the journal Psychopharmacology retracted three phase two papers, because the authors had known of the unethical conduct at that site without disclosing it and hadn't fully declared a potential competing interest.

**Kiffer:** And the two phase three reports haven't been retracted. So let's test what a better system would look like. Here's an invented excerpt from a hypothetical trial protocol. One, participants may continue to see their study therapists privately after the trial if both agree. Two, concerns about the study should be raised with the site's lead therapist. Three, sessions will be recorded and kept for training. Four, all team members will sign the sponsor's code of ethics.

**Sarah:** So for everyone: if you sat on the research ethics board reviewing that excerpt, what would you ask the team to change? Take a few seconds.

*(Pause)*

**Sarah:** I'd start with the first sentence. MAPS reported that the participant had moved near the therapists and kept seeing them for therapy, and the relationship began during follow-up. So I'd want clear rules on contact during follow-up and after the trial, with any continuing care declared to someone outside the team. And under the second sentence, a concern about the lead therapist would go straight to the lead therapist. A participant needs at least one route that doesn't pass through the people the concern is about, explained at consent and repeated at follow-up.

**Kiffer:** Good. In Canada those routes include the sponsor, the research ethics board, Health Canada, the regulatory college of any licensed team member and, for criminal conduct, the police, and a service like VictimLinkBC can help a person decide what to do. On the third sentence, Health Canada's 2022 expectations say a protocol that records sessions should state the purpose, how the recordings will be used, how often they'll be reviewed and how long they'll be kept, and the participant can withdraw consent. I'd add a reviewer outside the team who watches a sample of each team's sessions soon after they happen.

**Sarah:** The excerpt also says nothing about touch.

**Kiffer:** A board should catch that. Health Canada expects consent to any touch before the drug is given and again in the moment, contact that's non-intimate and non-sexual, and touch that's optional except to prevent imminent harm. That first agreement has to be reached while the participant is sober, and sexual or romantic contact with a participant is never acceptable.

**Sarah:** And the code of ethics in the fourth sentence sounds reassuring on its own.

**Kiffer:** It's a good start, and MAPS reported in 2019 that it had published a formal code of ethics. I'd add a written duty for every team member to report a concern about a colleague to someone outside the team, and I'd ask that unlicensed members work under the direct supervision of a licensed therapist, as Health Canada expects. Each change puts a check where it doesn't depend on the goodwill of the therapists in the room.

**Sarah:** Okay. Now the part where we disagree.

**Kiffer:** Whether people should have access before better trials exist.

**Sarah:** I think the FDA was right to decline approval in August 2024, and I'd be wary of opening even a narrow route like Australia's for now. We don't know how much of the nine to twelve points survives expectancy. Durability rests on a single follow-up visit that lost about a quarter of participants. Positive effects like euphoria weren't recorded as adverse events, and liver tests weren't done in phase three. And at one site, every protection failed, and the sponsor learned of it only through the participant's complaint years later. Access now asks patients to carry all of that uncertainty.

**Kiffer:** Those are real gaps, and I agree the evidence fell short of what's needed to approve a product for marketing. Where I see it differently is for people whose PTSD hasn't responded to established treatments. Two randomized trials agreed with each other, the raters were blind, and dropout from the MDMA arms was about five percent. Compare that with a large trial in United States veterans, where about fifty-six percent left prolonged exposure and about forty-seven percent left cognitive processing therapy. For that narrower group, I think a supervised route with strict conditions is defensible.

**Sarah:** That's an indirect comparison, though, the same kind we were careful about in question one. And ICER noted that intensive formats of trauma-focused therapy have also reached a pooled attrition of about five percent. The trials also didn't require that people had already tried other treatments, so they don't test your narrower group directly.

**Kiffer:** Those are fair points, and I'll take them. My argument rests more on the two trials agreeing, although they share the same blinding problem, and on how many people are left without a treatment that works for them. The FDA noted that response to the two medications approved for PTSD in the United States rarely exceeds sixty percent, and that fewer than twenty to thirty percent reach full remission. Australia has allowed authorized psychiatrists with ethics committee approval to prescribe MDMA for PTSD since the first of July 2023, and sixty-three had been authorized by the end of July 2026.

**Sarah:** And where does Canada stand?

**Kiffer:** As of October 2026, no MDMA product has market authorization in Canada. The Special Access Program takes case-by-case requests, now under the Controlled Substances Regulations that came into force on the first of October 2026.

**Sarah:** But Australia's route answers a different question. The TGA rescheduled MDMA so that authorized psychiatrists could prescribe it. It hasn't evaluated any MDMA product, and the figures it released in 2026 count psychiatrists with no patient numbers. If access expands without collecting outcomes, we learn very little from the people who are treated.

**Kiffer:** That's your strongest point, and I agree with it. A route that doesn't record outcomes can't tell anyone whether it's helping. So where I land is a narrow route for people whose PTSD hasn't responded to other treatments, a group the MAPP trials didn't specifically test, tied to a registry that would collect the first real evidence for them, while broad approval waits for a trial that measures expectancy before randomization and follows people, blinded, for about twelve months. The question is live in the United States, too. In August 2026, the renamed sponsor, Resilient Pharmaceuticals, was reported to have resubmitted its application without a new phase three trial.

**Sarah:** I'm more cautious than Kiffer, but we agree on the practical rule. Any access outside a trial should come with the evidence the trials lacked, meaning every drug effect recorded, expectancy measured and twelve months of follow-up, and with protections that don't depend on the therapists in the room.

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

**Sarah:** First, the MAPP trials estimate what MDMA added to an intensive therapy, about nine to twelve CAPS-5 points. The placebo arm's improvement, almost fifteen points in MAPP two, belongs to the therapy and everything around it, and effect sizes from waitlist-controlled trials can't be lined up with these.

**Kiffer:** Second, when a blind fails, ask how large the bias would have to be to explain the result, using the share of each arm that believes it received the drug. In MAPP two, belief would have to be worth about twelve points, and measuring expectancy before randomization would help answer the question.

**Sarah:** And third, safeguards protect participants when they don't all rest on the same people. That means agreement on touch while the participant is sober, recordings reviewed in time by someone outside the team, licensed or supervised team members, rules that reach past the sessions, and a way to raise concerns outside the team.

**Kiffer:** If you'd like more practice, rework the expectancy problem from this episode, counting the four people who couldn't tell as half-believers, and check whether the answer moves much.

**Sarah:** Next time, it's Lesson ten, Psychedelics and Addiction: Alcohol, Tobacco, Opioids and Community Models, where the evidence turns to alcohol, tobacco and opioid use, and to community programs in British Columbia.

**Kiffer:** Take care, everyone.

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