Discussion Club: FDA/PhUSE CSS Working Group 4 Update Budapest 2012

From PHUSE Wiki
Jump to: navigation, search

FDA/PhUSE CSS Working Group 4 Update at the annual PhUSE Conference 2012


The FDA/PhUSE CSS Working Group 4 collaborates on the topic "Standards Implementation Issues with the CDISC Data Models". The group has narrowed their focus to working towards the delivery of a Data Guide. A small sub-team has put a lot of effort into drafting some initial thoughts around standard data guide and the discussion club has a brainstorming discussion on this topic.

Question 1: Do you create a ‘data guide’ and what does it contain?

Almost everyone created some type of data guide but might have called it something else. In general, people listed the following things they might include in the guide.

  • Explain how the analysis data was created
  • Explain errors and data issues within the data
  • Orientation of the database very much like a table of contents
  • Any data rules that were used for generating data sets that cannot easily be summarized in variable level metadata
  • Information about SAS code that is included in the submission
  • How Different approaches were captured – legacy vs.SDTM
  • Description of any data dependencies – raw-SDTM- ADaM and within a model (e.g. ADSL to other structures)

Everyone agreed that there needs to be clear communication between what should be in the data guide versus define.xml. The FDA representative said they recently had a submission where a lot of information was pulled out of the define.xml into a data guide and it was a disaster. But the draft guidance says there are a Submission Reviewer Guide and a Data Reviewers Guide. What is the difference?

Question 2: How much information is replicated in the data guide?

There was a lot of discussion about the sections in the current draft which refer to the study diagram and inclusion/exclusion criteria. This led to an overall discussion about how much information to duplicate. Everyone was in agreement to not duplicate information from the protocol and SAP but to reference that information where needed.

Question 3: At what level should the data guide be created?

Most people thought the data guide should contain information across the submission as well as study specific information. There was not consensus if this should be one large document or different documents broken out. Most people did think there should be some summary that links the pieces. Technically, some indicated that hyperlinking could be an issue and not sure what eCTD rules would allow. Also, there were some different opinions as to whether the Data Guide or define.xml should be the starting point. Most people thought define.xml should be the starting point with references to one or more data guides.

Some people also suggested structured study metadata that would be included at a submission level. Not sure PRM contains what we really need for the submission.

Another good comment was whether the data guide is a guiding document or a supportive document.

Question 4: What should be included in the data validation section?

The data validation section should:

  • Clearly communicate warnings/errors in the data
  • Not be just a dump of a compliance report
  • Should include links to the published checks
  • Needs to include sponsor defined checks that go above and beyond industry checks
  • Include some more clear text around checks that maybe are not intuitive out of the box

People seemed to circle around having a structured table which summarized the warnings rolled up and then maybe some narratives where serious issues/errors occurred.

General Comments

  • This seems to be SDTM centric. Are we missing anything by not at least thinking about ADaM
  • How to handle and document different version of SDTM – not getting clear feedback from FDA
  • Separate discussion about ADaM.
    • Are people including analysis results metadata? In general no, but those that are doing it are only including key analysis
    • Slight changes in analysis concepts from study to study
    • Really don’t know as most of people have not documented full ADaM metadata yet
  • Isn’t define.xml enough? Most said no but some are just submitting the define.

Last revision by Cdecker,10/19/2012