Labels and "label mismatch"

Re: Labels and "label mismatch"

Hi Jozef,

The labels requirement comes directly from the IG (section 3.2.2 Conformance in SDTM-IG 3.2), which states:

Conformance with the SDTMIG Domain Models is minimally indicated by:
Using SDTM-specified variable labels for all standard domains (5th bullet)

This requirement is also confirmed by FDA published validation rule FDAC033.

If you have an issue with the rule implementation we would be happy to work with you on correcting it. However, if your issue is with the underlying business rule, then you should take it up with CDISC and the regulatory agencies.

Here is more details on the implementation. The label checks in Validator are implemented as 2 separate rules: SD0063 and SD0063A. The first rule checks that labels for standard variables (Required, Expected, and Permissible) in standard domains match IG exactly including case. So this is the rule that implements the IG requirement. The second rule compares labels for model permissible variables with the generic labels from the Models and was originally implemented as Informational. Its purpose was to make it easier to review the non-standard variable labels as suggested by Anthony Chow (http://www.opencdisc.org/forum/sdtm-313-validation-found-some-errors#comment-916). However, when FDA was publishing their rules they automatically upgraded all Informational messages to Warnings, so SD0063A is currently a Warning. We understand that this is an issue and are working with FDA to correct it for the next release of FDA rules.

Thanks for your feedback,
Max

Label mismatch

I agree with Max that this is the current status.

But shouldn't we be critical to such rules when we see that they decrease data quality, and delay submissions considerably, rather than blindly follow and implement them? 10% of the FDA rules is just damned wrong, another 10% unclear or open for different interpretations. Do we accept all these? I don't.

Two of the goals of SDTM were to improve data quality and allow faster reviews. Many of the rules as they are there now decrease data quality and retard reviews (e.g. reviewers having to go though many pages explaining the "false positives")

Here is another post I recently wrote and is surely of interest: "Airport security, toothpatse and validation rules". It presents a completely new approach based on risk assessment.

Dataset Labels

Can anyone provide a reference in an IG that says the specific case a DATASET label must be?

Pinnacle21 rule warning DD0059 is reported if the dataset label (which is used in define.xml) is uppercase, the expectation seems to be sentence case.

In the SENDIG at least, each domain description table shows the name of the dataset in uppercase as section heading, and sentence case above the table.

Label mismatch

The only thing (as far as I could find out) the SDTM-IG (v.3.2) states in section 3.2.2 (Conformance) is:
"Conformance with the SDTMIG Domain Models is minimally indicated by: ... Using SDTM-specified variable labels for all standard domains". It doesn't state anything about whether that is case sensitive or not.
Side-remark: for the domains, it doesn't even state where each label exactly is, one can only guess from the sections.

IMPORTANT: the above mentioned sentence will be removed from the future SDTM-IG 3.3. My interpretation of that is that we will finally be allowed to (slightly) change the variable label to BETTER explain what the variable means in the context of our own submission.

Also see: http://cdiscguru.blogspot.com/2015/12/sdtm-labels-freedom-or-slavery.html