AI-Ready FDA Submissions: CDISC, Validation, and Machine-Readable Data
About this session
AI builders working in regulated healthcare need more than accurate models. They need reliable, traceable, and machine-readable data foundations. This session explains how statistical programmers prepare FDA review-ready clinical data packages across NDA, BLA, IND, sNDA, and rolling-review pathways, and how these practices can support responsible AI adoption in regulatory workflows.
Attendees will follow the submission lifecycle from Pre-IND activities through NDA or BLA filing and Complete Response Letter support. The session covers programmer responsibilities within eCTD Modules 2.7 and 5, CDISC implementation using SDTM and ADaM, Define-XML 2.1, and commonly used analysis datasets such as ADSL, ADAE, ADLB, ADTTE, and ADRS or ADTR.
The presentation also examines validation and quality controls involving SDRG, ADRG, Pinnacle 21, Analysis Results Metadata, and dual-programmer review. It connects these established practices with emerging AI-ready submission capabilities, including Dataset-JSON, growing use of R alongside SAS, and machine-readable metadata. Participants will leave with a practical framework for building compliant, validated, and structured clinical data foundations that can support automation and future AI-enabled regulatory processes.
Speaker
Key takeaways
- Learn how CDISC standards, Define-XML, reviewer guides, and validation frameworks create compliant, traceable, and AI-ready FDA submission data. Understand how Dataset-JSON, machine-readable metadata, and R alongside SAS are shaping automation and future AI-enabled regulatory workflows.