Someone Has to Watch: Observation and Documentation Under FDA's Final Psychedelics Guidance

The federal green light meets a hard problem

For years, psychedelic drug development sat at the edge of the pipeline. That is no longer where it lives. In a single week this July, the FDA finalized its guidance for clinical investigations of psychedelic drugs and scheduled a September public hearing on their future therapeutic use. HHS and the VA entered a five-year agreement to advance these therapies for veterans, and a wider wave of federal action followed, from a second FDA and VA scientific-collaboration agreement to new NIDA and ARPA-H funding for ibogaine research and an HRSA request for information on safe delivery. Days later, one of the largest pharmaceutical companies in the world committed billions to acquire a psychedelic developer.

The direction is set. What the momentum does not resolve is a design problem that sits at the center of every one of these programs, and it is not the molecule. It is observation.

Why the psychedelic session is so hard to observe

Most therapeutics are evaluated through brief, periodic clinic visits. Psychedelic drugs break that model. A single dosing session can involve several hours of altered consciousness, and subjects can remain vulnerable and suggestible for hours after. The clinically important activity does not sit between visits where a scheduled scale can sample it. It unfolds continuously, inside a long window that is genuinely difficult to capture by hand.

The final guidance is specific about what that window has to yield. Sponsors are advised to document the incidence, duration, and severity of all central nervous system effects across the session, including expected effects like euphoria, perceptual changes, and shifts in mood and cognition, and to characterize the onset and resolution of each. Because of abuse potential, expected psychoactive effects are to be recorded as adverse events even when the subject does not consider them adverse. Sessions are expected to be observed by two trained monitors for their full duration.

Read those expectations together and the guidance is asking for a complete, time-resolved, reviewable record of everything that happens across hours of altered consciousness, captured consistently enough that a third party can assess it later. That is a high bar, and a monitor writing notes in real time while also caring for a patient was never going to clear it.

The documentation burden is really a data-capture problem

When observation depends on a person watching and writing, three things degrade. Detail is lost, because no one can log the onset and resolution of every effect across a multi-hour session. Consistency erodes, because two monitors at two sites document differently. And nothing is re-reviewable, because a handwritten note cannot be handed back for independent assessment.

Reframe the requirement as a data problem and the answer changes shape. If the full session is captured as structured, timestamped video and audio, the record becomes complete and objective rather than reconstructed. Onset and resolution of effects tie to the exact moment they occur. Signals and alerts can be flagged and routed to the clinician who needs to respond. A reviewer can jump straight to the highlighted moment in the recording instead of relying on memory. And because the same capture protocol runs at every site, the resulting data is standardized and available for blinded central raters to evaluate after the fact.

That last point connects to a separate but related exposure in these trials: functional unblinding. Objective measures scored from a recording depend far less on whether the blind held, and can be re-checked against the source. Videra Health’s video-based behavioral models have been validated in peer-reviewed work, including detection of tardive dyskinesia at 0.89 AUC, matching or exceeding trained raters. The relevance here is straightforward: the same capture that documents a session to the guidance’s standard is also the substrate for objective measurement.

Observation does not stop when the session ends

The guidance extends the observation problem past the acute experience. It encourages sponsors to evaluate durability of response with a double-blind assessment at 12 weeks, to follow subjects for symptom recurrence, and, in the most informative designs, to build in blinded long-term follow-up out to roughly 12 months with prespecified retreatment criteria. It also flags that a REMS, along with additional postmarketing safety work, may be needed as larger populations are exposed.

Federal activity pushes the same way. The HHS and VA agreement calls for collecting and sharing real-world data on safety, effectiveness, and cost, and for building provider education. HRSA is asking how these therapies can be delivered safely in community settings. All of it depends on structured observation that reaches beyond the clinic and across time.

Remote, protocol-based assessment on a patient’s own device closes that gap. Between-visit video and audio check-ins let sponsors track durability and recurrence, capture safety signals as they emerge, and generate the longitudinal, real-world evidence that regulators, payers, and health systems are now explicitly requesting. The capability that documents the session is the same one that supports the follow-up window and the postmarketing commitments that come with approval.

What this means for clinical and regulatory teams

For a team designing a psychedelic program against this guidance, structured observation changes the evidence package in three concrete ways:

1. It produces a complete, timestamped session record that documents CNS effects to one standard, rather than reconstructing them from monitor notes.
2. It makes that record auditable and re-reviewable, so a signal can be verified against the source and, where appropriate, assessed by blinded central raters.
3. It carries observation into the durability and postmarketing window, where the durable-effect claim and the REMS obligations actually live.

None of this removes the clinician or replaces a validated primary measure. It adds a defensible observational layer beneath clinical judgment, aligned with the closer scrutiny these trials now receive.

The path forward

The federal support is real and the pipeline is filling. As trials scale, the sponsors who treat observation as core infrastructure rather than an afterthought will move through development faster, defend their data more confidently, and be readier for the postmarketing and REMS obligations that follow. Throughout, the goal is objective, defensible documentation of what happens to patients, with no stake in how any therapy is ultimately used.

To see how video-based capture performed inside a live study, read Videra Health’s case study on how an AI-powered eCOA revitalized a stalled Phase IV trial across all 45 sites.