How (and why) to get MDMA-assisted therapy approved by the FDA

Recommendations for Lykos Therapeutics (formerly MAPS), and arguments for the Food and Drug Administration.

The FDA rejected MDMA-assisted psychotherapy as a valid treatment for severe PTSD earlier this year, on my 30th birthday, citing various concerns with the large-scale Phase 3 clinical trials sponsored by Lykos Therapeutics (the non-profit formerly known as the Multidisciplinary Association for Psychedelic Studies).

But first, I’d like to begin by sharing this relevant paper published in the British Journal of Medicine in 2002 titled ”Parachute Use to Prevent Death and Major Trauma Related to Gravitational Challenge: Systematic Review of Randomized Controlled Trials”, which I think perfectly captures why the FDA's rejection of MDMA-assisted psychotherapy was the wrong decision: if it were up to the FDA, parachutes wouldn’t be an approved treatment for falling out of airplanes, either. There have been "some problems with the data collection" in the those studies.

---

Below you will find brief summaries of the various reasons the FDA rejected this bid for approval, followed by my recommendations for and responses to each of those concerns. Some of my responses are directed more at the FDA; others are directed more at Lykos Therapeutics (for how to revise the clinical trials next time in order to get MDMA-assisted therapy approved as a safe, clinicially valid form of treatment, or for points they should make in their formal appeal of the rejection).

1. Prior MDMA Experience in Participants

FDA Concern: "The FDA noted that a significant proportion of trial participants had prior experience with MDMA—approximately 40%—which could introduce expectancy bias, potentially skewing the results."

My Response/Recommendation:

Around 40% of the actual intended patient population (people suffering from severe PTSD) do have prior experience with MDMA, according to my best educated guess (I'm sure decent data exists out there for this though), which means that trying to only run the study on participants who don't have MDMA experience would actually make the study _less_representative and clinically valid, not more.

2. Double-Blind Design Challenges

FDA Concern: "The FDA noted that the psychoactive effects of MDMA made it challenging to maintain the study's double-blind design, as many participants could likely discern whether they received the active drug or a placebo."

My Response/Recommendation:

Lykos Therapeutics should propose using a similar but already FDA-approved medication as the control, rather than a truly "empty" placebo. I believe niacin has been used in the past, but this is simply not a good placebo when the alternative that participants will be comparing it to is literal ecstasy. A much better candidate for the placebo would be something like a moderately to moderately-high dose of Adderall or Dexedrine (i.e., amphetamine), which are already medications approved by the FDA (and designated as sufficiently safe) yet still produce a significant psychotropic effect. (And if it turns out that amphetamine-assisted therapy works just as well for PTSD as MDMA does, well then that would also be a great discovery to make!)

3. Cardiovascular Concerns

FDA Concern: "The FDA expressed apprehension regarding the cardiovascular effects of MDMA, such as increased blood pressure and heart rate, and the potential for abuse."

My Response/Recommendation:

The increased blood pressure and heart rate are short-term, and there have not been significant long-term cardiovascular risks shown from occasional MDMA use in the literature. If the FDA is concerned about the risks of even _short-term_elevated blood pressure and heart rate, then they should probably also ban exercise, sex, and scary movies.

4. Potential for Abuse

FDA Concern: "The FDA raised concerns about the potential for MDMA to be abused."

My Response/Recommendation:

I'm not exactly sure what they mean by "abuse" here specifically—like do they think people will receive MDMA-assisted therapy and then go out and start doing molly every weekend? Or keep coming back for more MDMA-assisted therapy even if they don't need it or it doesn't work? Either way, these are testable hypotheses that can be evaluated alongside the rest of the metrics in this study. All you need to do is survey the study participants over some period of time after their MDMA/placebo experience and see if they tend to feel more of a desire to use MDMA recreationally (or to seek out more MDMA-assisted therapy sessions) or not, after they have the officially sanctioned experiences.

5. Sexual Misconduct and Standardization of Psychotherapy Protocols

FDA Concern: "The FDA highlighted concerns about reports of adverse events, including a notable incident of alleged sexual misconduct by a therapist during the trial, and the standardization of psychotherapy protocols across different trial sites."

My Response/Recommendation:

This is literally a textbook, perfect case in which having AI facilitate the psychotherapy — rather than a human — would kill two birds with one stone: AI cannot sexually abuse patients, and you can perfectly standardize it by using the exact same AI model/LLM with the exact same system prompt. And by standardizing the therapy and eliminating physical risks from human therapists, you would probably increase the quality as well while you're at it.

I'm sure many folks at Lykos will find the idea of replacing humans with AI in this sort of therapy rather unpalatable, but I would say a few things to any skeptics of this idea:

1. Would you rather complain about getting chocolate instead of vanilla and get to eat zero ice cream at all, or shut up and enjoy some chocolate ice cream?

2. Humans will still be involved in the therapy, but it can be registered nurses rather than psychotherapists, present in the room with the patient for safety and emotional support – not actually providing the psychotherapeutic services or guiding them through the experience.

3. AI speech-to-text and text-to-speech are now sufficiently good that the entire therapy session could really feel to the patient like it was happening with a real human therapist. They would still speak their answers using their voice and hear responses out loud, like a normal human-to-human conversation.

4. This would dramatically reduce the cost of running these sorts of experiments, as I assume that one of the main drivers of the high cost of MDMA-assisted psychotherapy is paying a human psychotherapist for 8 hours of work at $200/hr. The comparable cloud compute costs of running AI models to provide similar services would be negligible.

5. Most of the benefits of MDMA-assisted psychotherapy don't come from brilliant insights provided by the human psychotherapists—they come from the introspective insights of the patients themselves, under the effects of a drug that enables clear, safe introspection. Given this widely accepted view of why MDMA-assisted psychotherapy works, I think it would work at least as well, if not better, with large language models and AI voices as the therapists.

6. Adverse Events and Long-Term Durability

FDA Concern: "The FDA raised concerns about the adequacy of data collection on adverse events and the long-term durability of the treatment's effects."

My Response/Recommendation:

The FDA should consider that the alternative to MDMA-assisted psychotherapy isn't just a bunch of veterans and sexual assault victims continuing on with their normal lives, free from any other interventions or coping mechanisms. They are instead coping with their pain in other ways, which often involve severe addiction and alcoholism, suicide, and extreme psychological suffering. Concerned about potential adverse events and long-term durability? How's the adverse events profile of tens of thousands of suicides every single year? And what's the long-term durability of getting drunk as a treatment for depression?

You need to compare these things to their actual, realistic alternatives, rather than comparing MDMA-assisted therapy to some hypothetical alternative that doesn't reflect the empirical realities of people living with severe PTSD.