Code.decode value for CDISC Controlled Terminology: NCI Preferred Term or CDISC Submission Value?

[ASL-rmarshall]

There are currently three types of CDISC controlled terminology referenced for use in USDM spreadsheet:

1. DDF Terminology: terminology specifically created for use in USDM.
2. SDTM Terminology: terminology that was created for use in SDTM whose codelist concepts are also used in USDM.
3. Protocol Terminology: terminology that was created for use in the CDISC Protocol Model, which contains a single codelist that's used in USDM (C174222, Study Arm Type Value Set Terminology).

All CDISC controlled terminology is published in a consistent format, with the following attributes provided for each concept/term in each codelist:

• Code (conceptId): A C-code with a leading character C followed by a sequence of digits used for uniquely identifying each concept in NCI Thesaurus (NCIt), including all CDISC concepts.
• CDISC Submission Value (submissionValue): Contains the approved, semantically correct term for each concept for representation within the context of the CDISC standard for which the terminology is defined (DDF/USDM, SDTM or Protocol).
• CDISC Synonym(s) (synonyms): This identifies the applicable synonyms for the CDISC Submission Value. These terms should not be submitted, but are included for collection and mapping purposes.
• CDISC Definition (definition): This identifies the value of an ALT_DEFINITION property of a CDISC source for a particular concept. In many cases an existing NCI definition has been used. The source for a definition is noted in parentheses (e.g. NCI, CDISC glossary).
• NCI Preferred Term (preferredTerm): This identifies the value of a PREFERRED_NAME property for a concept in NCIt. **NOTE - This column designates the human readable, fully specified preferred term corresponding to the NCI C-code, and is especially helpful for searching NCIt to get the entire concept with links to all instances of the term.

Within the model and supporting documentation for USDM, there is no explicit indication of which of these values should be used when populating the code and decode attributes of the Code class when CDISC terminology is used (i.e., when codeSystem is "http://www.cdisc.org"). Although it safe to assume that the code attribute should contain the "Code" (conceptId) C-code, it is not clear whether the CDISC Submission Value or NCI Preferred Term should be used to populate the decode.

Looking at the example test data that's provided with each release, it seems that the terminology attribute used depends on type of terminology:

• For DDF terminology, the CDISC Submission Value is used. This makes sense because CDISC Submission Value contains the term that have been defined specifically for use within USDM (i.e., the term that is specified in the USDM_CT.xlsx file that is included with the deliverables for each USDM release).
• For SDTM terminology, the NCI Preferred Term is used. This also makes sense because, although the concepts in the codelists are applicable for use in USDM, the CDISC Submission Value has been defined specifically for representation of the concept within the context of the SDTM standard - i.e., as a value in an SDTM dataset. While SDTM representational standards (e.g., text values in upper case) are appropriate for representation of data points in datasets, they are not appropriate for representation of the concept with USDM (e.g., when rendered by reference into narrative context). In the absence specifically defined terms for representation of these concepts within USDM (i.e., these concepts are not defined in DDF terminology), the NCI Preferred Term from the SDTM terminology is used as the next-best alternative.
• For Protocol terminology, the NCI Preferred Term is also used. As the CDISC Protocol standard is similar to (or even a precursor for) USDM and, in most cases, has similar representational requirements - most concepts are created for human-readable representation in a protocol - it might be more appropriate for the CDISC Submission Value to be used.

The IG should be updated to indicate which terminology attribute should be used to populate the decode attribute of the Code class - i.e., to define which values are "correct". Without this specification, it is not possible to define conformance rules (and implement them in CORE) for use of CDISC terminology.