FDA's R Pilot Review Letters: The Honest Gap Between Proof-of-Concept and Reviewer-Ready

The R Consortium’s FDA Submissions Working Group has now accumulated something genuinely useful: formal FDA review letters for both Pilot 4 (WebAssembly/Docker) and near-final feedback on Pilot 5 (Dataset-JSON) — the first checkable public record of where FDA’s reviewer environment actually sits.

For Pilot 4, FDA confirmed technical feasibility of both approaches and then spent the rest of the letter explaining why neither is ready for routine use. WebAssembly still requires renv for dependency restoration and a local web server to render the app — the reproducibility burden is repackaged, not eliminated. Docker fared worse: FDA’s letter explicitly flags WSL dependency, licensing constraints on agency machines, and concludes that containers do not yet meet the robustness standard for regulatory submissions. The counterintuitive finding worth circulating internally: FDA reviewers preferred WebAssembly over containers precisely because it only needed a browser. The working group expected Docker to be the more familiar path. It was not.

For Pilot 5, the operational blocker is external: Dataset-JSON v1.1 is not yet supported in Pinnacle 21 validation software, forcing a fallback to .xpt for validation. FDA also flagged intermediate dataset handling, triggering a January 2026 resubmission. Until Certara adds P21 support, the Dataset-JSON migration path is practically blocked for most sponsors — that is the single number to track.

Against those constraints, two quiet wins: J&J’s hybrid R/SAS submission (SAS for ADaM, R for TFLs) received FDA approval within timeline and is now a replicable operational blueprint; and the August 2025 FDA Study Data Technical Conformance Guide now explicitly permits R-native file extensions in eCTD packages — making the .txt conversion workaround obsolete. Teams still running that workaround should flag it now.

Protocol read: The R-in-submission story is closer to “real but specific paths only” than the WG’s optimism sometimes suggests — hybrid R/SAS is the replicable win, while container-only submissions and Dataset-JSON migrations remain gated on infrastructure outside the sponsor’s control.

What to do now:

• Treat J&J’s hybrid R/SAS (SAS for ADaM, R for TFLs) as the operational blueprint for R-inclusive submissions in 2026.
• Hold off on Dataset-JSON migration plans until Certara adds Pinnacle 21 support; track that one product update rather than the broader CDISC roadmap.
• Stop using .txt extension workarounds for R artefacts; the August 2025 Technical Conformance Guide permits R-native extensions in eCTD.