Open Source in Filings - R Validation in 2026 (2026 EU Summit)

Unresolved reproducibility challenges and possible support models for regulatory reviewers of R submissions

2026
EU Summit
Validation
Published

October 5, 2026

Are Concerns About Open Source in Filings Solved?

Discussion summary from the inaugural R/Pharma EU Summit at Novartis in Basel on October 5, 2026, alongside BioTechX.

See also: 2026 US Discussion

Short Answer: Not Fully

The main unresolved problem is how regulatory reviewers can reliably reproduce submitted results. This is especially difficult when a reviewer does not use R or lacks technical support, the required environment, or package setup. Even when a reviewer uses R, reproducibility is not guaranteed.

Main Concerns

  1. Reproducibility: reviewers unfamiliar with R face significant barriers to rerunning analyses.
  2. Validation and verification: people use these terms differently, particularly in submissions mixing programming languages. Sponsors need to state acceptable discrepancies and explain expected differences explicitly.
  3. Reviewer readiness: education and technical readiness vary between agencies. Code alone is insufficient; reviewers may need setup support, instructions for running analyses, and context for interpreting them.
  4. Dependencies and version changes: a large package stack behind a clinical study report (CSR) increases installation demands. Reproducing results years later may be difficult as packages and systems change. SAS was perceived as more backward-compatible, potentially making this a concern for open source.
  5. Human support: there is no clear support model when reproduction fails. Help may require agency IT, specialist technical teams, and a shared support structure outside the agency.

Possible Solutions Discussed

These were proposals, not established solutions or regulatory commitments:

  • Set expectations early: tell reviewers up front that a submission uses open-source tools.
  • Improve submission documentation: use a clear Analysis Data Reviewer’s Guide (ADRG) template. An R Consortium submissions working group pilot was described as working on this. Video walkthroughs could demonstrate setup and explain what to review in the ADRG and submission.
  • Make statisticians the first contact: they are well placed to explain both technical differences and expected analytical discrepancies when reviewers cannot reproduce R results.
  • Define acceptable discrepancies in advance.
  • Provide standardised package environments: industry could jointly create a curated, regulator-friendly, installable environment rather than only a shared repository. Collaboration with regulator IT was suggested; the specific examples, possibly FDA and China’s CDE, were uncertain in the summary.
  • Explore infrastructure options: WebAssembly and webR were mentioned.
  • Fund technical support inside agencies: industry could collectively fund IT staff to set up environments and answer questions. Participants perceived a support gap, while noting conflicts of interest would make support from an individual pharma company problematic.
  • Use AI skills: these might help set up environments or explain discrepancies when regulators re-check R results in another language.

Takeaway

The problem extends beyond validating packages. Reviewers need reproducible environments, clear expectations and documentation, and a workable human support model.