n Nancy
on

 

Is it possible to specify links to multiple supplemental files in a single variable source/derivation, or do we need to provide value-level metadata pointing to each file separately?  We're creating Define-XML via P21C using the Excel specs template, and have tried associating all of the files with the same comment, and then referencing that comment from the variable metadata, but only the last file defined in the list is displayed in the resulting Define-XML file.  What are we doing wrong? 

Forums: Define.xml

Sergiy
on August 18, 2022

Hi Nancy, 

You need to introduce Value Level Metadata. So, you can provide a reference to different Origins and define related conditions for each VLM item.

Kind Regards,
Sergiy 

n Nancy
on August 19, 2022

Thanks, Sergiy!  We'll try doing that.

Thanks,
   Nancy

Lex
on August 19, 2022

The ideal approach when a variable has multiple origins is to use Value Level Metadata. However, Define-XML 2.1 supports the use of multiple def:Origins when there are multiple sources for a single Variable but no way to construct a ValueList that clearly identifies the cases where each def:Origin applies.

it was not clear in this example whether you actually had multiple origins, or a variable with one origin, but just multiple documents attached to that origin. Multiple documents would just mean multiple def:DocumentRef child elements within the def:Origin element. The latter should not be solved with Value Level Metadata.

Best, Lex

 

j Jozef
on August 20, 2022

In addition to Lex's excellent comment, I think you should also think about whether you really need multiple documents for explaining something (like a derivation) about a single variable. I think this may be confusing for a reviewer, needing to jump between several documents.
So you may think about combining the documents and then reorganizing the single one, so that all information about a single variable comes together and  is not distributed over different documents.
Anyhow, also use def:PDFPageRef to point exactly where in the document the reviewer can find the information. def:PDFPageRef can either point to a page (or page range) in the document, or to a section, so-called "named destination". The latter then of course needs to have been created in the PDF document.
I have seen so many cases that when the user clicks on the link for the document in the HTML rendering of the define.xml, that the PDF opens on the first page instead of on the page or section where the information is. Thus, testing all links in the rendering is of utmost importance - you really don't want to frustrate your reviewer.

Regarding creating all this, there are some excellent GUI-based tools that allow to do all of this by drag-and-drop and wizards.

s Siru
on December 20, 2023

Hi Sergiy,

I also have a question about multiple origin varible. In our project, an independent assessement would required for tumor imaging. And we have both investigator results (origin: CRF) and BIRC results (origin: eDT). My question is will all variables have both origin need value level meatadata ? 

For example, in TU domain, we have TULNKID, TULOC, TUDTC, VISIT, TUORRES, TUSTRESC, TUTESTCD, TUTEST, TUEVAL, all these variables have value collected in CRF and from eDT . Then all these variables need value level metadata ? 

Thanks.

Sissi

j Jozef
on December 21, 2023

I have some doubts whether TUTESTCD, TULNKID, TUSTRESC, maybe also VISIT, are "collected".
--TESTCD/--TEST are usually "assigned", --LNKID also is, or is "derived", --STRESC is often copied from --ORRES, or derived from it (e.g. unit transformation). --DTC usually will have the same origin as --ORRES.
Only assign valuelevel metadata to variables that are really collected!
I don't of course know your exact case, but my experience is that in 80% of the cases, it is sufficient to only assign valuelevel metadata to --ORRES.

s Siru
on December 22, 2023

Hi Jozef,

Thank you. The IRC trasnfered data is in SDTM format, and we use the --TESTCD, --TEST,--LNKID......they transferred directly. So for these varialbe, your suggestion is we still can use "assigned" other than "eDT" ?

 

Thank you.

Sissi

j Jozef
on December 22, 2023

Hi Sissi,
Do you mean that you received the SDTM files without a define.xml?
If so, that would be very bad practice: one should not try to generate a define.xml from SDTM files, one should have the define.xml as the "specification" of the SDTM mapping.
Or did I misunderstand you?
 

s Siru
on December 28, 2023

Hi Jozef,

I would say the files we recived are not SDTM actually, they are just in SDTM format and contain the SDTM variables. We have a data specification which specifies variable name, label, variable type, format or controlled terms, description, possible values. And we need to combine the SDTM-like data and EDC data into a real SDTM. I'm not sure I've made myself clear.

For example, in the specification the varlable TRLNKID, it described such as "Target Lesions : T1-T2; Non-Target Lesions : NT1-NTn; New Lesions : N1-Nn;Split Lesion: T01.1, T01.2…". So I'm not sure these vairbles' origin is assigned or eDT.

Thank you very much.

Sissi

j Jozef
on December 28, 2023

If you can export from the EDC system in CDISC-ODM format, everything becomes much more easy ...
There are several tools on the market that easily, and in an extremely user-friendly way, allow you to set up the mappings and generate the SDTM datasets.
If you need to start from Excel "specifications" ... OMG ...

s Siru
on December 29, 2023

Hi Jozef,

I think you misunderstood me. What I mean is for TU/TR/RS, we have two parts of data, one is from EDC or we can say investigator, another is from indenpendent assessor. The indenpendent assessor's part of data was transferred in SDTM format contains --TESTCD/--TEST/--LNKID/--ORRES and so on. And for EDC data, we will populate these variables too. Then we'll set two parts of data together to have a final SDTM, so most of the vraibles have 2 origins, one is eDT and one is collected/assigned.

I hope I finally explained it. Thank you very much.

Sissi

j Jozef
on December 30, 2023

Without seeing your data and your possibilities to work with them, it is hard to give you any advice.
For example, there is nothing such as an "SDTM format". SDTM is a categorization model, for which the transform format can be XPT, XML, JSON, CSV ...
Also, can the EDC system export in CDISC ODM-XML format?
Also, unclear is whether you want to keep the datasets from the two sources separate ("split domains") or that you want to merge them. Do you have the tools to merge datasets?

Anyway, as Lex already wrote, and I keep confirming, using "ValueLists" in the define.xml allows you describe different origins.

If you need more advice, you can always contact me by e-mail, my address is easy to find out.

Want a demo?

Let’s Talk.

We're eager to share and ready to listen.

Cookie Policy

Pinnacle 21 uses cookies to make our site easier for you to use. By continuing to use this website, you agree to our use of cookies. For more info visit our Privacy Policy.