Comments for Blog post: Crucial Role of Release Control In R For Health Care Organizations

[joseph-rickert]

Please add comments on the R Consortium Blog Post under this issue.

[bms63]

The idea of a Golden suite of packages being available for all sponsors and regulatory agencies sounds like a dream. We all agree that these packages are "gold" and don't need to be assessed internally or externally for validation purpose and so we can focus on others.

[geoskiad]

Question related to this statement: "One suggestion that has gained some traction is that sponsors use the previous minor and latest patched R version for a submission." - How close to a DB lock would one switch the R version used for the CSR analysis? Even in a change cycle of 2 times per year, considering that you do not want to risk timelines due to a last minute change of R version, by the time you reach "submission" you may have quite an "old" R version. Is anyone aware of any potential issues/problems (from Regulatory perspective) with this until now? Or do I misinterpret something here? Note: I have zero personal experience with R based submissions.

[bms63]

Question related to this statement: "One suggestion that has gained some traction is that sponsors use the previous minor and latest patched R version for a submission." - How close to a DB lock would one switch the R version used for the CSR analysis? Even in a change cycle of 2 times per year, considering that you do not want to risk timelines due to a last minute change of R version, by the time you reach "submission" you may have quite an "old" R version. Is anyone aware of any potential issues/problems (from Regulatory perspective) with this until now? Or do I misinterpret something here? Note: I have zero personal experience with R based submissions.

I would not switch R versions between a lock and CSR unless there is some huge risk to the sponsor and regulatory agency. Hopefully, very unlikely to ever happen!

I think engaging with the reviewers early is very important especially if the study is going to happen over many years. The regulatory agencies can usually get access to R versions and prepare accordingly.

[madprogramer]

This blog post resonates a lot with me. Back in the Old World, I know that there is some work being done on CE approval of medical software. Here too, reproducibility of results is a big expectation similar to the FDA approval processes.

I don't think R versioning is as important as it was a few years ago. Namely, renv has matured a good deal. So I would say I am leaning towards the Roche and Eli Lilly way of doing things. renv can also address package versions, on top of the R release version itself and imo should be the standard practice for 'snapshotting'.