Introducing Value Traceability

Henning Kuich

Oct 16, 2024

3 mins

We are delighted to announce the first version of value traceability. This new feature marks an important advancement as we continue to build out our code traceability capabilities.

In the context of SDTM, value metadata offers detailed, variable-specific information that enables the precise description of different values within a single variable. For example, this is particularly useful in the LB (Laboratory Test Results) domain, where a single variable may hold multiple types of laboratory results, each with unique characteristics, such as units, ranges, and measurement techniques. This pattern extends throughout both SDTM and ADaM domains, as exemplified in the PARAMCD-AVAL pairs holding efficacy information, as shown in Figure 1 below.

Figure 1: Value-level data structure examples for the SDTM LB domain (a) and ADaM PARAMCD-AVAL pairs (b).

Taking PARAMCD-AVAL pairs as an example, variable traceability applied to AVAL by default displays all derivations for AVAL (Figure 2).

Figure 2: Variable-level filter in Verisian Tracer showing all AVAL derivations.

However, if there are different derivations for PARAMCD-AVAL pairs, understanding their specific implementations remains difficult. To address this, we have developed value traceability, which allows users to filter code and upstream dependencies specific to the desired PARAMCD (see examples in Figures 3-6).

Figure 3: Value-level filter PARAMCD=”AS” in the Verisian Tracer filters for the desired derivation and upstream dependencies (in this example, the variable x only).

Figure 4: Value-level filter PARAMCD=”ER” in the Verisian Tracer filters for desired derivation and upstream dependencies (in this example, the variable y only).

Figure 5: Value-level filter PARAMCD=”OP” in the Verisian Tracer filters for the relevant code and recognizes that no further upstream dependencies are required due to a static value assignment.

Figure 6: Value-level filter PARAMCD=”UH”, which does not exist, and assigns the derivation correctly to the else-branch.

It is important to note that value traceability is a flexible concept. Rather than being limited by value-level conditions predefined in your metadata, the Verisian Tracer now enables you to filter for any condition under which a value assignment for a variable of interest takes place (Figure 7).

Figure 7: Example of a general value-level filter in the Verisian demo study filtering only for LBNRIND value assignments under three conditions.

This release marks a major milestone in the development of code traceability. By covering traceability at dataset-, variable-, and value-level, the Verisian Tracer now provides traceability far beyond any concept defined through CDISC data standards and submission documents like the Define-XMLs. The result is the greatest possible transparency and user friendliness for analysis development, quality control, and regulatory review.

Introducing Value Traceability

Explore Further

Traceability and AI for Better Understanding, Communication, and QC of Trials

The Limitations and Opportunities of Large Language Models

Introducing the Verisian Community