Question about variable lengths - SDTM versus ADaM

j Joan

on September 17, 2013

Hi,

I am wondering. What is the rationale in 1.4.1 that has lengths tested strictly for SDTM to avoid excess, and does not do the same check for ADaM datasets? I'm not sure why excess is okay for one and is impermissible in the other. Note I am not requesting or advocating for more length checks for ADaM.

Thanks!

Joan

Forums: General Discussion

s Sergiy

on October 1, 2013

Hi Joan, Historically ADaM

Hi Joan,

Historically ADaM validation is limited to a set of rules in "ADaM v1.2 Validation Checks" document. Those rules were developed by CDISC team based on text of the ADaM Implementation Guide.

I agree with you that some additional requirements for analysis data may be introduced for particular usage. E.g., FDA submission.

Kind Regards,

Sergiy

d Dirk

on April 10, 2014

On the topic ADaM vs. SDTM lengths

Hi guys,

Good work being done BTW!

Some thing to consider for these length checks in ADaM if/when implemented: there is the copy rule, i.e. Those variables in ADaM that come from SDTM should be exact copies (var. name, label, content, and attributes).

From this, I would understand that the ADaM dataset alone does not have the necessary info. to determine the length since an ADaM dataset does not necessarily have all records that are in its SDTM predecessor, hence setting the length of a var to max(length in ADaM-set) may be wrong...

Note: 99% sure about this: length is an attribute and the doc. For ADaM v2.0 explicitely mentions attributes (section 4.1 says: "If a variable exists in an SDTM domain that can be used for an analysis without any change, then this variable should be included in the analysis dataset “as is”, with all SDTM attributes retained. ").

As an extension question on this: How are SDTM copies in ADaM to be treated wrt their define.xml metadata? Are these to be considered "attributes" as per the above quote? I am thinking primarily about variable role: some variable synonym in SDTM may conceptually become different in ADaM so that its role is different. Is the role variable metadata piece actually relevant for ADaM? The ADaM IG seems to not mention it and the model has it only for a couple of variables... For now, I have been using the copy-rule and where a variable is not SDTM I set these to "Analysis" which is in a very general way always true, but at the same time feels wrong...

All thoughts very welcome, :-Dirk