PhUSE Working Group: Best Practices for Documenting Metadata
Meeting Date: October 14, 2015
Attendees:
· Sandy (Pfizer)
· Steve (Cytel)
· Kiran (Roche)
· Cathy (Novartis)
· Mina (FDA)
Agenda:
· General/Admin:
o no update from D.J. regarding the stylesheet subgroup
· Review submitted comments on SDRG template:
o There were 4 other sets of comments besides ours:
o https://www.federalregister.gov/articles/2015/07/23/2015-18027/intent-to-review-a-study-data-reviewers-guide-template
o The purpose of reviewing is to gain additional insights/issues to include when we write our white paper
Discussion Notes:
· Amgen comments
o Dictionary/Standards Versions:
§ Information on use of multiple versions (mix of versions) should come from FDA, likely in the Technical Conformance Guide
§ It’s generally not possible to have a single version across all studies within a submission since version updates occur over time, however a single version should be used within a single study, at least for medical coding dictionaries (MedDRA especially). This could possibly differ for CDISC CT, however terms within CDISC CT are not versioned individually and are rarely removed or changed, so could likely just reference whichever latest version was used.
o Agree that it would be helpful to split out SDTM model and IG versions (believe these are no longer paired documents) and to list any TAUGs (Therapeutic Area User Guides) used.
o For the conformance sections, it was recommended to have separate subsections for dataset conformance issues versus define.xml conformance issues. This could be helpful, however there are typically few issues for define.xml itself.
o In terms of CBER requiring a separate document, Mina noted that the majority of biologics now go to CDER and there are not too many CBER applications these days, so there is not a huge impact. We had commented that it would be helpful to have one template that can be used for CDER and CBER submissions.
· AstraZeneca comments
o Agree that protocol design section 2.2 should be required.
o Agree to label change for header in Appendix I.
o Agree that adding Rule ID in the issues table would be helpful. Would need two columns for OpenCDISC rule ID versus FDA rule ID.
o Agree that completion guidelines should state that issue explanations are mandatory.
o Should be OK to combine issues across domains when the same issue occurs in multiple domains. May need to change header in first column of table.
o There are usually a lot of OpenCDISC findings and many of them don’t matter that much, but it can be hard and subjective to identify which are “significant”. Should this be determined by the sponsor? Further clarification from FDA would be helpful, although it would be difficult for them to provide a definitive list of what is considered significant and may differ between reviewers.
§ Examples:
· Many findings for values outside of CDISC CT, but codelists are extensible.
· Duplicates findings due to OpenCDISC not using all of the key variables identified for the specific study.
§ Sponsors will generally try to clean up/address as many OpenCDISC findings as possible. If it’s something that cannot be resolved but has a good explanation, then it is not considered significant, in which case you shouldn’t really have anything “significant”. But if it has an impact to review, it should be noted (and could be considered significant).
§ Significant = anything that could affect interpretation of a study report or analysis?
§ Feeling was that sponsors should be conservative and list everything (Errors and Warnings; Notices not typically included).
§ How should OpenCDISC bugs/defects be documents (not errors in the data, but are errors in OpenCDISC). Would be nice to have some convention or guidance.
· Side topic: Sponsor doesn’t have to submit everything that was collected in the study. Is there any guidance around this? Is it prudent to submit everything?
o Anything used in analysis/reporting should be submitted, however what if certain data was not used for analysis/reporting? Is the sponsor encouraged to submit it anyway?
o Reviewers may want to do their own analysis and may look for the data.
o Anything that is not submitted must be documented accordingly. Should not why not submitted too.
o Best to submit everything or at least have everything prepared in case it is requested.
· Novartis comments
o Besides items already noted, they want to call out use of the TAUGs more.
· Novo Nordisk comments
o Did not have time to get to these; we will review at beginning of next meeting
Next Meeting: week of October 26