FDA requirementes for datasets

[DmytroYermak]

Hi all,

Could you please help with a link on FDA requirements/regulations: where clearly stated that datasets for submissions have to be in SAS format.

Thank you!

[RW9]

Currently the standard is to provide (unfortunately as its a dreadful file format) XPT files along with define.xml, reviewers guide and such like.  It should be compliant with the CDISC model, so most people use Pinnacle21 to validate structure and code lists.

They may move to dataset.xml for data, however I haven't seen any movement for this.  As for links and such like this has nothing to do with SAS per se, look at the CDISC website:

https://www.cdisc.org/

[DmytroYermak]

Hi,

Does it mean that XPT format can be received from any other formats like .xlsx or .accdb/.mdb or .r ? If so, what is the reason why sas is being used?

Should program codes be submitted along with XPT files?

[RW9]

These are all CDISC questions, not SAS ones.  

XPT is used as it is "open source" - which means you can find the file specification online which is freely described (its is also ancient, and a real pain to work with).

No, xlsx, accdb etc. are not acceptable as far as I am aware - there is clear guidance one what is to go in an FDA package.

SAS is the preferred statistical software as it is a validated system.  R is not currently although its making headway and some companies are attempting to validate it.

Yes, all programs are being asked for now along with data - not sure if it is a requirement as such yet, but its certainly on its way to being.

[DmytroYermak]

No, xlsx, accdb etc. are not acceptable as far as I am aware - there is clear guidance one what is to go in an FDA package.

I wanted to ask if the formats above can be converted into XPT. If just .sas7bdat can be converted into XPT - this is it - the requirement to use just SAS.

And could you please say, if know, what the 'validation' can mean for R. It is 'open source' and nobody will care/take responsibilities for the correctness of the program, will not it? Or CRAN should do it? But the CRAN is 'unofficial', isn't it?

[RW9]

"I wanted to ask if the formats above can be converted into XPT. If just .sas7bdat can be converted into XPT - this is it - the requirement to use just SAS." - Yes, for instance you can read Excel files into SAS then create XPTs from the datasets.  Is there a magic button which converts file X to file Y, no, not generally.  Nor would you do this any way, data sent to FDA has to be in a certain structure, as given in the CDISC models.  Read the guidance over at CDISC on how to send data packages to FDA, it is all clearly stated and defined there and outside the scope of a SAS forum. 

"Validation" - SAS is validated software, which means that SAS certifiy the software will do what it says it will do according to the documentation.  R is open source, meaning anyone can contribute to the code base, either directly or via packages.  Noone provides assurances that the code will do what it says, or not change etc.  This is the major difference between the two.  I don't believe CRAN provides any assurances that the code works as it should.