Biometrics partner for sponsors and CROs

R in FDA submissions in 2026: what the pilots proved and what is still SAS territory

A few years ago, "Can we submit R?" was a debate. Today it is a planning question. FDA does not require any particular statistical software; it asks that the software be documented, including its version. The R Consortium's submissions pilots have shown, step by step, what that looks like in practice.

What the pilots showed

  • Pilot 1: tables and figures produced in R and reviewed by FDA.
  • Pilot 2: an interactive Shiny application delivered to reviewers.
  • Pilot 3: ADaM datasets and outputs built in R end to end.
  • Pilot 4: the same kind of application delivered through WebAssembly and containers; FDA reviewers found the WebAssembly version easier to use.
  • Pilot 5: a package using Dataset-JSON instead of XPT files, resubmitted in January 2026 after minor FDA requests.

Separately, FDA's eCTD technical conformance guide now accepts R file types such as .rds and .rda, which makes it simpler to include R code and data with a submission.

The tools have matured

The pharmaverse packages that most R submission work relies on are now stable and actively maintained: admiral for ADaM derivations, metacore and xportr for metadata and transport files, rtables, gtsummary and cards for tables and analysis results. Version pinning matters: record exactly which package versions produced each output, the same way you would record a SAS version and macro library.

Where SAS still makes sense

Many sponsors have decades of validated SAS macros, standard programs and trained teams. Rewriting them for their own sake rarely pays off. A common pattern we see is SAS for the core SDTM and ADaM pipeline, with R added where it shines: graphics, interactive review, simulation and new analyses.

What to put in place before your first R deliverable

  • A controlled environment: a fixed R version and a package snapshot (for example with renv) for the study.
  • Validation evidence for the packages you rely on, proportionate to the risk of each output.
  • Double programming across languages where it helps: an R production program checked by an independent SAS program catches different kinds of errors.
  • The software, versions and operating system stated in the ADRG.

We programme in both SAS and R and can work in either your environment or ours. If you are deciding where R fits in your next study, we are glad to talk it through.

Have a study that needs statistical support?

Tell us about the study and what you need. A statistician will get back to you.