Regulation — pharma-specific
FDA Diversity Plans Guidance (April 2022) — representative-population parity for AI-driven eligibility
The Guidance sets FDA's expectation that sponsors enroll clinically representative populations. When AI screens eligibility, "representative" becomes a measurement question about the classifier — not just the recruitment plan. A silent exclude-lane differential on the elderly-multimorbid patient group is the shape FDA drafted this Guidance to detect.
What the guidance actually says
"FDA encourages sponsors to enroll participants who reflect the diversity of the population that will use the medical product, if approved … A Diversity Plan should include measurable enrollment goals, operational plans to meet those goals, and a discussion of how the sponsor will monitor progress."
FDA Guidance for Industry: Diversity Plans to Improve Enrollment of Participants from Underrepresented Racial and Ethnic Populations in Clinical Trials (April 2022 draft, codified expectations in FDORA 2022)
"Sponsors should describe … the plan to reach enrollment goals, including specific measures for recruitment and retention of participants from underrepresented racial and ethnic populations."
FDA Diversity Plans Guidance — monitoring + measurement expectation
What this means in plain English
Two obligations when AI participates in eligibility screening:
- Measurable representative-population goals. The Diversity Plan appendix names specific enrollment targets. Those targets are auditable.
- Monitoring. "How will the sponsor monitor progress?" AI-driven eligibility monitoring falls to the sponsor — not the clinical-AI vendor whose classifier is being monitored.
An AI classifier that systematically over-routes an underrepresented subgroup to the ineligible_exclude lane is the operational failure mode this Guidance is calibrated against. Retraining timing, recruitment mix, and per-subgroup exclusion rate are all in scope.
What triggers the exposure in the sample
P3 (age 65+, 3+ comorbidities, minority representation) exclude-lane rate rose from 22% baseline to 75% recent. A 53 percentage-point group differential in the wrong direction for a Diversity-Plan target subgroup is precisely the shape a FDA Type C meeting reviewer or EMA scientific-advice rapporteur would flag. The Diversity Plan appendix commitment — "sponsor monitors AI-driven eligibility decisions for representative-population parity across pre-specified subgroups" — is now harder to answer without dated independent evidence.
What the $499 Snapshot shows against this rule
- Per-patient-group exclude-lane rate baseline vs recent — the direct parity measurement
- Timestamped drift-event log — when the differential first emerged
- Independent-verifier signature — not the sponsor grading its own recruitment
- 3 fix-first items scoped to the trial-eligibility AI — the "operational plan to meet goals" restart
See the lane-shift chart that produces the finding →
$499 Snapshot. 3 business days.
Independent-verifier determination scoped to your sponsor's AI surface + 3 fix-first items + signed declaration.
Buy $499
Snapshot credit applies to Baseline ($2,500) or Enterprise Attestation ($35-55K) upgrade within 30 days.