Home / Knowledge Base / Article / Domains are Topic-based, Except When They're Based on Structure

For an implementer trying to decide where data belong in SDTM-based datasets, it's pretty clear when data belongs in a trial design dataset, a relationship dataset, one of the new study reference datasets, or one of the special purpose domains.  However, it can be difficult to choose the right general observation class dataset, especially if data are about findings.

The SDTMIG says, "A domain is defined as a collection of logically related observations with a common topic."  For most general observation class domains this means that the content of domain records has a common "topic" represented in the topic variable, --TERM for events, --TRT for interventions, or --TESTCD for findings.

  • For events domains, the common topics are kinds of events (e.g., adverse events, disposition events, healthcare encounters).
  • For interventions domains, the common topics are interventions that play a common role in the study (e.g., exposure to protocol-defined treatments, concomitant medications, or procedure agents).
  • For most findings domains, the common topic is a kind of data, such as
    • Lab data (data from tests on specimens) of a particular type (e.g., microbiology [data about non-host organisms], microscopic [histopathology], pharmacokinetic concentrations [concentrations of drugs and drug metabolites], and general clinical labs).
    • Morphology or physiology data about a particular body system (e.g., cardiovascular, reproductive)
    • Data about tumors or lesions (identification, results)
    • From a particular kind of examination (e.g., ECG, physical examination, vital signs) for some of the earliest SDTMIG domains*. Note: new domains are no longer created based on particular kinds of examinations.

Findings domains, however, include a few domains where the content of the tests (questions) do not have a common topic, but the domain is based on a structure:

  • The Inclusion/Exclusion (IE) domain represents yes/no responses to eligibility criteria.  The content of those eligibility criteria could be almost anything, for example, subject age, lab or other findings, or medical history.
  • The Questionnaires (QS) domain represents data from a questionnaire. The questions may be addressed to a subject or associated person about current or past experiences and their effect on the respondent, or they may be questions answered by a clinician about the subject.  The criteria for considering something a questionnaire are independent of the content of the questions.
  • The Functional Tests (FT) domain represents data about the subject's performance on an activity or task represented as a test.  The criteria for considering something to be a functional test are independent of kind of test or the kind of medical information drawn from the test results.
  • The Disease Response and Clinical Classifications (RS) domain represents formally structured data supporting a published disease response criteria or clinical classification. Again, the criteria for considering something a disease response or clinical classification are independent of the content of the data.

The fact that some findings domains are based on structure rather than content can present difficulties in deciding where particular data should be represented.  When is a set of questions a questionnaire rather than just a set of questions which should be represented in domain(s) based on question content?

The QRS standards page describes the kinds of instruments which CDISC considers questionnaires, functional tests, disease response criteria, and clinical classifications.  The information on QRS standards page as well as the available QRS supplements can be useful in deciding whether particular data should be represented in a structure-based QRS domain or in topic-based findings domain(s).

*The earliest version of the SDTMIG was heavily influenced by the1999 FDA guidance, "Providing Regulatory Submissions in Electronic Format – NDAs", which guidance included the following list of "individual datasets that are generally provided to support each study."

  • Demographics
  • Inclusion criteria
  • Exclusion criteria
  • Concomitant medication
  • Medical history
  • Drug exposure
  • Disposition
  • Efficacy results
  • Human pharmacology and bioavailability/bioequivalence data
  • Microbiology data
  • Adverse Events
  • Lab – chemistry
  • Lab – hematology
  • Lab – urinalysis
  • ECG
  • Vital signs
  • Physical examination