Tanya Ramey has spent much of her career pushing psychiatric drug development to think past the diagnosis and toward the underlying mechanism, first at NIH and NIDA, where she helped build the phenotyping infrastructure the field is only beginning to use, and, until last month, as CMO of Usona Institute. Getting her to put that thinking into writing for Cascades Analytics is a genuine privilege, in this co-authored piece built around Tanya's own clinical and regulatory experience.
Psychiatric medications are typically prescribed based on a DSM-5 diagnosis, for instance, PTSD, MDD, GAD. Clinical trials use the same diagnosis-based approach to assess efficacy, and the same strategy was applied to psychedelics when they entered drug development.
The difference is that psychedelics appear to work on neurofunctional domains independently of DSM diagnosis. There are defensible mechanistic hypotheses, such as Gul Dölen’s critical window reopening, and the Carhart-Harris’s entropic brain/neuroplasticity window framing – these are nonspecific for depression, PTSD, nor addiction. The diagnosis-agnostic window of opportunity, represented by plasticity, reward-learning, arguably, positions the therapeutic content as supplied by what you pair with that open window. If the drug is a process, then the indication comes second.
The closest that anyone has come to doing the transdiagnostic trial properly was the kappa-antagonist (Eli Lilly compound LY-2456302, aticaprant). The trial was called Fast-fail Trial in Mood and Anxiety Spectrum Disorders (FAST-MAS), as a nod to an idea that such a trial design could fail the compound early on. This was the first attempt to use Research Domain Criteria (RDoC), linking anhedonia with ventral striatum reward anticipation activation on fMRI, across mood and anxiety continuum, dimensionally. In late-stage development this was abandoned, and the early mechanistic signal from phase 2 wasn’t carried into phase 3. Aticaprant development was discontinued in 2025 due to lack of efficacy in phase 3. However, its role in FAST-MAS was significant as one of the first examples of applying the dimensional RDoC methodology to test a specific neurobiological target (KOR antagonism) for anhedonia in a transdiagnostic psychiatric context. The lesson the whole industry took from that is: dimensional mechanism signal is real in Phase 2 but does not reliably convert into an approvable Phase 3, and the regulatory framework will not give you a dimensional label anyway.
Psilocybin has shown reductions in anhedonia in treatment-resistant depression (Shukuroglou et al., 2023) and decreases in demoralization in cancer-related psychological distress (Ross et al., 2016), but not necessarily uniformly across every individual depressive symptom that define an MDD diagnosis in DSM. Sleep disturbance, appetite changes, and somatic complaints are all core MDD criteria that psilocybin does not appear to primarily engage. That distinction matters enormously for how we design clinical trials with psychedelics. If only a subset of MDD patients carry the specific domain deficit psilocybin targets, then the average effect measured across the full diagnosed population is watered down. The signal is real, but the total trial population under study is diluting it.
This is an important problem for neuroscience drug development. DSM diagnoses, based on clinical presentation, touch domains heterogeneously. An MDD population contains patients who respond especially well and might not need readministrations of a drug to continue response and remission, because psilocybin hits their specific neurofunctional deficit, and patients who do not respond, or partially respond, because they don’t have enough of the target symptomatology. A DSM-level diagnosis can fully encompass a domain deficit, partially overlap with it, or barely touch it at all. Averaging across responsive, partially responsive, and non-responsive patients obscures what the drug does and for whom.
The logical next step would be to enroll patient populations in clinical trials based on, or enriched, for their domain profile. Assessing, using, for example, deep phenotyping batteries to ascertain the specific neurofunctional profile deficits at screening, and tracking how those deficits evolve with treatment. Which patients show the anhedonia signal before dosing? Which patients show the strongest cognitive rigidity? Enroll by that profile, and you can ascertain whether there is a real therapeutic signal. This could be the next frontier for precision psychiatry, matching the drug not to DSM diagnosis, but to the deficit the drug actually targets. At NIDA, T.R. initiated and led the deep phenotyping battery for Substance Use Disorders (SUD), Phenotyping Assessment Battery (PhAB), through to successful validation (Keyser-Marcus et al., 2021). Given the burden on the subjects when using such batteries, T.R. kept it operationally feasible in terms of length of administration, and investigators later created a brief version from it, PhAB-B (Parlier-Ahmad et al., 2023).
To utilize a dimensional approach in clinical trials would require redesigning clinical trial infrastructure that has been built around diagnostic categories for decades. Deep phenotyping batteries use for predictive symptomatic profiles enrichment, validated domain-specific instruments, adaptive enrollment strategies: none of this is standard practice yet. The FDA approves by DSM-based indication (except that TRD is nested within MDD, and is not DSM-defined), and accepting domain-level endpoints as clinically meaningful would require, for example, a predictive biomarker to undergo a CDER-involved Drug Development Tool (DDT) qualification process, which is slow, expensive, and offers no assurance of a successful outcome.
The safety question is the hardest one. When you enroll by diagnosis, comorbidities are relatively known. Domain-based enrollment opens the population in ways that are harder to map. For psychedelics, where adverse event profiles include psychological destabilization, mania triggering, and rare perceptual disturbances, that is closer to a black box, and a risky one. A sharper efficacy signal may come at the cost of a safety profile no one has ever mapped.
Assessing psychedelics for their domain-targeting capabilities would give the field something it does not currently have: a precise account of what these drugs target, and in whom. Today, the model is largely borrowed from classical antidepressant development: the same endpoints, the same populations, the same regulatory logic. That framework was not designed for compounds that appear to work transdiagnostically, on functional dimensions that do not respect the boundaries of the DSM.
“Assessing psychedelics for their domain-targeting capabilities would give the field something it does not currently have: a precise account of what these drugs target, and in whom.”
The registrational psychedelic programs are deliberately the opposite of transdiagnostic. They are DSM-based, single-indication, symptom-scale primary endpoints: COMP360 in TRD on MADRS, Usona in MDD on MADRS also, Resilient (formely Lykos) in PTSD on CAPS-5. Where the field is innovating on design, it is on the blinding axis, not the dimensional one. The EPIsoDE trial, for instance, ran a triple-blind, active-placebo-controlled Phase 2b in TRD, and there is now a head-to-head testing psychedelic therapy against antidepressants under equal unblinding conditions. That is the methodological frontier the field has chosen: fixation on blinding, not on capturing the transdiagnostic dimension.
The transdiagnostic insight is expressed at a portfolio strategy level, rather than a trial design. Compass Pathways run the same molecule across TRD, anorexia, OCD, PTSD, as separate programs. Definium’s DT120 (formely MindMed’s MM120) runs across GAD and MDD. That is to note that many baskets are being run sequentially and independently, which captures the breadth commercially through label expansion while demoting both the statistical power of borrowing across baskets and the mechanistic true dimensional readout. The transdiagnostic nature of psychedelics is reflected as serial simultaneous single-indication development programs, which one can also see as a tacit admission that the real dimensional approach is not buildable under the current regulatory framework in neuroscience.
At the molecule level, the non-hallucinogenic plasticity-promoting analogs are a bet that the dimension is neuroplasticity, decoupled from subjective experience. If that pans out clinically, it reframes the entire class as plasticity therapeutics, the most dimensional framing imaginable. But this framing is pre-registrational. The moment the molecule reaches Phase 3, it goes straight back into the same indication-labeling wall as everything else. That is the platform-versus-indication tension: the regulator buys diseases, not platforms, even when the underlying science is about platforms.
Reliance on DSM is asynchronous with construct-centric transdiagnostic neuroscience. What psychedelics do dimensionally, the regulatory system cannot reward. However, dimensional endpoints can be positioned inside a DSM indication. The pragmatic approach is to run phase 2A with a primary endpoint based on a neurofunctional construct, then run the next trial – POC with dimensional enrichment within an indication. For example, if there is a signal on criterion D (Negative Alterations in Cognition and Mood) on CAPS- 5 in phase 2A, you can launch two POCs: PTSD enriched for criterion D on CAPS-5 and MDD using MADRS, so the population is dimensionally enriched for 2 indications, even though the label stays DSM diagnostic. Two programs could run simultaneously. This is the most realistic way to bias toward the mechanism without abandoning the indication. Have transdiagnostic secondary or exploratory endpoints, and a mechanistic mediator, such as behavioral tasks, plus the putative mediator (a plasticity biomarker, or the acute subjective surrogate). You satisfy the FDA on the DSM-centric label while accumulating the dimensional dataset. Dimensional signals stay exploratory and underpowered, which means they can never carry the weight of a claim. The FDA still approves by an indication.
Psychedelics may not mark the end of DSM-centered psychiatry. But they may be the strongest argument yet for building a parallel framework, one that asks not what disorder a patient has, but what neurofunctional deficit they carry, and whether this drug can address it.
On July 8 at noon ET, Cascades Analytics Live welcomes Tanya Ramey, M.D., Ph.D., former CMO of Usona Institute and NIDA program director, for a conversation on the future of psychedelic drug development. Free. RSVP on LinkedIn.
References
- Shukuroglou M, Roseman L, Wall M, Nutt D, Kaelen M, Carhart-Harris R. Changes in music-evoked emotion and ventral striatal functional connectivity after psilocybin therapy for depression. Journal of Psychopharmacology. 2023;37(1):70–79. doi:10.1177/02698811221125354. PMID: 36433778.
- Ross S, Bossis A, Guss J, Agin-Liebes G, Malone T, Cohen B, Mennenga SE, Belser A, Kalliontzi K, Babb J, Su Z, Corby P, Schmidt BL. Rapid and sustained symptom reduction following psilocybin treatment for anxiety and depression in patients with life-threatening cancer: a randomized controlled trial. Journal of Psychopharmacology. 2016;30(12):1165–1180. doi:10.1177/0269881116675512. PMID: 27909164.
- Keyser-Marcus LA, Ramey T, Bjork JM, Adams A, Moeller FG. Development and Feasibility Study of an Addiction-Focused Phenotyping Assessment Battery. The American Journal on Addictions. 2021;30(4):398–405. doi:10.1111/ajad.13170. PMID: 33908104.
- Parlier-Ahmad AB, Eglovitch M, Legge C, Keyser-Marcus LA, Bjork JM, Adams A, Ramey T, Moeller FG, Martin CE. Development and clinical feasibility study of a brief version of an addiction-focused phenotyping battery in females receiving buprenorphine for opioid use disorder. Brain and Behavior. 2023;13(7):e3128. doi:10.1002/brb3.3128. PMCID: PMC10454248.