Regulation — pharma-specific

ICH E8(R1) + E9(R1) — study design + statistical principles for AI-in-trials

E8(R1) frames the general considerations for clinical studies; E9(R1) extends statistical principles including the estimands framework. When AI touches eligibility or analysis, the intercurrent-event language + quality-by-design language of both guidelines apply directly.

What the guidelines actually say

"Quality by design in clinical research means that quality should be built into the scientific and operational design and conduct of clinical studies … Quality factors that are critical to the reliability of information and the protection of study participants should be identified prospectively." ICH E8(R1) General Considerations for Clinical Studies (2021)

"An intercurrent event is an event occurring after treatment initiation that affects either the interpretation or the existence of the measurements associated with the clinical question of interest." ICH E9(R1) Addendum on Estimands and Sensitivity Analysis (2019, updates 2026)

What this means in plain English

Three practical implications when AI participates in eligibility or analysis:

Quality factors prospective. AI-driven eligibility is a quality factor that must be identified up front — not audited only after enrollment closes.
Estimand clarity. If the target population differs from the enrolled population because AI screening skewed representation, the estimand definition needs an explicit intercurrent-event handling.
Sensitivity analyses. E9(R1) expects sponsors to test estimand robustness. AI-driven representation skew is a first-class sensitivity-analysis input.

What triggers the exposure in the sample

P3 exclude-lane rate rose from 22% baseline to 75% recent. That skew is an intercurrent event affecting the target-vs-enrolled-population map. Under E9(R1), it is exactly the kind of finding a statistical analysis plan should address with a sensitivity analysis + estimand-attribution note — and under E8(R1), it should have been flagged prospectively as a quality risk on the trial-eligibility AI.

What the $499 Snapshot shows against these rules

Dated evidence of AI-driven representation skew — the intercurrent-event flag record
Per-patient-group differential quantified — direct input to sensitivity-analysis re-planning
Quality-by-design gap surfaced early enough for prospective SAP updates
Independent-verifier signature — distinct from the sponsor's biostats vendor

See the lane-shift chart that quantifies the skew →

How does this help me?

An EMA CHMP or FDA CDER statistician who sees an unaddressed AI-driven representation skew in the SAP will push back at scientific advice or pre-NDA. Having independent evidence early lets the biostats team fold the finding into the estimand strategy rather than defend after the fact.

Read: ICH E8(R1)+E9(R1) -- what it saves in SAP + estimand review →

$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.