_testing._(48788570773).jpg){kind=link}
- Published
- Updated
- Sources checked
Traceability requires more than matching numbers in filenames. A reviewer should be able to explain which data cut, transformations and analysis rules produced a result in the clinical report, including decisions that changed the meaning or inclusion of individual observations.
Choose a reported result
FDA study-data resources encourage planning standardized data and supporting documentation together. Begin a tracing exercise with an important endpoint or safety table. Make the assumptions behind its title, denominator, analysis population and data cutoff explicit before following the result back through the data pathway.
Draw the linkage chain
Connect the report table to the analysis dataset, then to SDTM and the source entry. Derivation rules, metadata and program versions should explain these links. For external laboratory or device sources, include the transfer file and acceptance record as part of the evidence chain.
Run a tracing exercise
Ask a reviewer who did not produce the data to perform the exercise. An explanation that only the programmer knows is insufficient. Where joins, filters or date rules change the result, the reviewer should understand the decision from documentation and demonstrate it on a representative record.
- Trace a critical table back to a source record.
- Include the external-data transfer version in the chain.
- Reconcile data cuts against outputs in the final delivery.
Close changes through impact assessment
When a correction occurs, identify affected datasets and outputs. Carrying a table generated from an older data cut into a newer report is a common control gap. Show approved files, versions and review status together in the final delivery manifest; file existence alone does not demonstrate suitability.
Frequently asked questions
- Are matching filenames sufficient traceability?
- No. Transformation logic, source and version relationships must also be understandable.
- Does access to code answer every question?
- Code needs the scientific rule, metadata and version context alongside it.