Populating ValueLevel

Populating ValueLevel

Valuelevel metadata is typically something YOU have to decide on.
It might be that you do not need valuelevel metadata at all for some domains in your submission, or it may be that you need it for every distinct test code or even beyond that.
Also it makes sense to group valuelevel metadata. For example, systolic and diastolic blood pressure have the same data type, meaning the valuelevel metadata can be shared for them. In other cases, setting valuelevel metadata solely based on the test code might be insufficient. For example, if you have glucose measurements in blood as well as in urine, and for the latter, some is quantitative and some is qualitative. In such cases you will want to make extensive usage of the "WhereClause".
There are some very nice programs on the market that help you make such decisions, making proposals, and having a graphical user interface that makes it easy to create the "WhereClause"s in a graphical way, without needing to understand XML (also XML is very easy to learn).

Did you ever take a CDISC define.xml training? There you learn everything about this.

Populating ValueLevel

What I forgot to say is that when generating valuelevel metadata, the information in the XPT files is far from sufficient when the test result is coded.
For example, if your test result can have the values "SMALL", "MEDIUM", "LARGE" (as in VS "FRMSIZE"), but "LARGE" was never checked in the CRF and thus also not present in the dataset, then you still need to add it to the codelist at the value level. Another example might be that there was an optional item "BMI" (VS) or a specific lab test that was never performed (but mentioned in the protocol), then you still need to add it to the codelist for VSTESTCD or LBTESTCD respectively, including the valuelevel metadata, even if there is no captured data for those tests.

Thanks for the clarification…

Thanks for the clarification. I got the idea that this could be automated from watching the tutorial "Creating Define.xml 2.0 with OpenCDISC" (https://www.youtube.com/watch?v=8PzYO0YlO0I). It seemed like all the ValueLevel and WhereClause information was populated from an Annotated eCRF that can be loaded in the Enterprise edition. Is that the case?

I did not take the CDISC define.xml training yet. I'm currently managing with the information available online. Can you recommend resources that could help in completing the define.xml specs? I am namely looking for explanations about the exact nature of information that should be entered into each column and which column actually require information vs. those that are optional. Although the information that goes into most column is relatively obvious, I had to play around with the information I had entered in the Codelist tab before the define.xml viewed on IE browser displayed the information correctly. I am still confused about the reason why the header terms in the define spec sheet do not correspond to the header terms displayed in the define.xml (ex: Term = Permitted Value (Code) and Decoded Value = Display Value (Decode).    

populating valuelevel

As one of the developers of the define.xml standard, I have seen this kind of problems very regularly. Therefore, I wrote the blog entry "Creating define.xml - best and worst practices". You can find it at: http://cdisc-end-to-end.blogspot.com/2015/03/creating-definexml-best-and-worst.html.
This blog is a real "hit". Since I wrote it, it has been read over 1,500 times.
Also, if you google "define.xml valuelevel metadata", you will find quite a lot of good articles and presentations. If you see one of the names "Lex Jansen", "Kevin Burges", "Sally Cassels" or "Monika Kawohl", then expect high quality. These people are the real experts, and are all members of the define.xml development team.

You can also always contact offline me if you need further information.

VLM in define.xml file is study specific

Hi Adam, 

Value Level Metadata (VLM) in define.xml file is study specific. So far, there are no particular guidance documents about expected granularity of VLM. In general, it should be enough to understand study data by an intended user, e.g., FDA Reviewer.

As a minimum, we recommend to describe a content of SUPP-- datasets in Tabulation data as VLM of QVAL based on QNAM terms.

For Analysis, VLM of AVAL (or AVALC) based on PARAMCD and other variables is recommended.

P21 Enterprise helps to generate VLM for most common cases. aCRF scanning extracts annotations and matches their pages with QNAM values in WhereClause for SUPP-- VLM.  

In Community version, you need to populate VLM manually. We recommend to upload a sample define.xml included into CDISC Define-XML package to generate Excel Specs. It will be a great source of useful examples. Missing functionality in P21C can be compensated by 1. custom programs like SAS macro and 2. LookUp function in MS Excel.

If you need commercial support, please contact our team.

Kind Regards,

Sergiy