Clinical trial recruitment video and IRB review
The IRB reviews the finished tape, not the script, so every graphic is a claim. A script-lock workflow and the cure-certainty framings FDA guidance rules out.
FDA's guidance on recruiting study subjects contains one sentence that changes how you produce a recruitment video, and most production teams never read it. When an advertisement is taped for broadcast, the IRB should review the final audio or video tape. The guidance adds that an IRB may approve the wording in advance to avoid a re-tape, and that its review covers the mode of communication, not just the information. Delivery can push an approved script out of bounds.
Two things follow. The reviewer signs off on a rendered file — graphics, music, and the final shot — not a document. Iterate before you submit: a change afterwards costs another review cycle in a recruitment window that is usually already behind.
What the IRB is actually reviewing
FDA treats direct advertising for study subjects as the beginning of the informed consent and subject selection process, not as marketing that happens to sit next to it. The IRB reviews both the information in the ad and the mode of communication. The standard is not whether a claim is literally false. It is whether the ad states or implies a certainty of favorable outcome or benefits beyond what the protocol and consent document support.
"Implies" is the operative word, and it is where video differs from a print flyer. A flyer implies things with adjectives. A video implies things with music cues, casting, lighting, and the shot you chose to end on.
The sheet is also specific about scope. Limit a recruitment ad to what a prospective subject needs to determine eligibility and interest: who is running the study and where, the condition under study, a summary of eligibility criteria, what participation requires in time and commitment, and how to make contact. Anything outside that list is something you will be asked to defend.
Framings that fail a recruitment review
The FDA information sheet names several of these. The rest apply the same rule to picture and music: do not state or imply a certainty of favorable outcome.
| In the draft | Why it fails | What to do |
|---|---|---|
| "A new treatment for [condition]" | The sheet flags "new treatment," "new medication," and "new drug" used without explaining the article is investigational | "An investigational drug being studied for..." |
| "Feel better without the side effects" | States or implies a favorable outcome | Describe what participation involves, not what it produces |
| "Free treatment" | Named where the intent is only that subjects are not charged for participating | "There is no cost to you for study visits or the study drug" |
| "$1,200 COMPENSATION" as the headline | The sheet warns against emphasizing payment through large or bold type | State it plainly, in body copy weight, not as the hook |
| "This could be the cure" | Certainty of cure, in the most direct form | Cut |
| "Breakthrough," "cutting-edge," "the future of" | Common IRB flags for implied efficacy; not a named list in the sheet | Cut |
| Final shot of a recovered patient embracing family | An implied benefit claim made with picture | Neutral imagery, or end on the contact card |
Generated video makes the last row easy to miss. A script can pass a compliance read while a swelling string bed under a golden-hour hug still says "you will get better." Review with the sound off, then with the picture off. If either pass implies an outcome the script does not state, the picture is making a claim.
The script-lock workflow
The whole method is to make everything cheap and reversible up to the moment of submission, and frozen after it.
- Source every line from the approved consent document and protocol. Keep a two-column working file: script line on the left, the consent-document sentence it derives from on the right. An empty right column means the line is cut or escalated to the PI. This one habit removes most of what would otherwise come back.
- Write the graphics list at the same time as the script. Every super, lower third, statistic card, logo, and end frame, as text. Because the IRB reviews the finished tape, an unlisted visual is an unreviewed claim, and the graphics list is what turns "the video" into something a reviewer can mark up.
- Storyboard the shots before generating any of them. Each frame annotated with what it is doing and which part of the eligibility summary it supports. The AI storyboarding guide covers the mechanics; the discipline here is that a shot with no stated job is a shot that will imply something.
- Run the forbidden-framings pass on script, graphics, and storyboard together. Three separate documents, one checklist, in one sitting.
- Get internal sign-off on all three before you generate. PI, regulatory, and the site coordinator who will field the calls. The coordinator catches eligibility phrasing the other two miss, because they are the one who has to explain it forty times.
- Build to preview quality first. The explainer video generator handles the structured build, and the video editor assembles it on a re-renderable timeline. The 480p preview pass is free and carries a short per-user cooldown, which is enough to lock pacing, voiceover timing, and on-screen text before anything is charged. That split is laid out in previews and final export.
- Circulate the preview, not the script. This is the step that saves the resubmission, because reviewers form an opinion about tone from a moving picture, never from a document. A client review loop built on previews and share links is the pattern; content approval workflows that don't stall keeps it from becoming six rounds.
- Render the final and submit the file. The export is charged once regardless of how many clips sit on the timeline. Submit the rendered video with its burned-in text and music, plus the script and graphics list as supporting documents.
- Freeze, and log what was frozen. Keep the approved render, its file hash, and the EDL that produced it, so four months later you can show which file was signed off.
Variants are where approvals quietly break
A 45-second spot is rarely one file. It is 16:9 for the site, 9:16 for vertical placements, 1:1 for feed, a captioned version, and often a Spanish version. Each of those is a different finished tape.
Ask your IRB in writing how they want the set handled before you build it. Some review the family as one submission with variants attached. Some treat a reframe as new material, because the crop changes what is visible and where text sits. Assuming the permissive answer is how a site ends up running an unapproved cut on its highest-impression placement.
The same question applies to three things people forget:
- Site-specific contact cards. Many IRBs will approve a template with a defined variable field so that twelve sites share one approved body. Many will not. Find out before you build twelve videos.
- Captions. Burned-in captions are part of the picture and part of what is reviewed; a sidecar subtitle file may be treated differently. Burned-in captions vs SRT and VTT sidecars covers the tradeoff, but the deciding vote is your IRB's, not the platform's. Either way, set the text deliberately with the caption generator rather than letting a model render it.
- Language versions. A translated recruitment video is generally its own submission and usually needs a certified translation rather than a generated one. Build the picture once so the language versions differ only in voice and text. AI content localization strategy covers structuring for that; the certification requirement is a regulatory question for your team.
Casting and imagery
Two rules keep recruitment video out of trouble.
Do not depict study procedures the protocol does not describe. If the trial involves four infusion visits, do not show a patient at home with a pill organizer because it looked calmer. The picture describes the protocol whether you meant it to or not.
Do not generate patient likenesses that read as real study participants. Generated people are fine as neutral illustrative imagery and become a problem the moment they are framed as someone in this trial. A recruitment ad has no testimonials, because there are no results yet.
Prefer environment and materials imagery over faces: a clinic corridor, a study calendar, a phone with the site number on it. It is easier to approve, and it does not imply that a person got better.
FAQ
Can we A/B test recruitment creative?
Only across materials the IRB has approved. Testing two approved hooks against each other is ordinary media practice. Generating a third because the first two underperformed and pushing it live is running unapproved recruitment material. Plan the variants into the original submission, which costs one review cycle instead of three.
What can we say about compensation?
State it plainly, in the same weight as everything else, and describe it as compensation for time and travel rather than payment for participation. The guidance's concern is emphasis, so the failure mode is design as much as wording. A figure set in the largest type on screen is emphasized no matter how the sentence reads.
How long does IRB review of a video take?
That depends on your board, whether the study uses a central or local IRB, and where you land in their meeting calendar. Plan production backwards from the review calendar rather than the enrollment target, and get the script approved as text early if your board offers that, since the guidance specifically contemplates approving wording in advance to avoid a re-tape.
Does this apply to social posts and search ads too?
The guidance is about direct advertising to recruit subjects, and format does not change that. A 15-second vertical clip is recruitment material, and so is the thumbnail and ad copy above it. Submit the whole unit rather than the video file alone, because the reviewer's question is what a prospective subject sees.