The Analysis Data Model extends the CDISC framework beyond raw data tabulation to address the analysis-ready datasets used for statistical programming and regulatory reporting. While SDTM captures data as collected, ADaM datasets are derived datasets optimized for analysis, containing derived variables, analysis flags, and structures that directly support the statistical analyses specified in the Statistical Analysis Plan. This separation of tabulation and analysis data maintains traceability while providing efficient datasets for programming tables, listings, and figures.

ADaM defines several fundamental dataset types and structures. The Subject-Level Analysis Dataset contains one record per subject with variables summarizing subject characteristics, treatment assignment, and disposition. Basic Data Structure datasets contain one or more records per subject per analysis parameter per time point, supporting analyses of endpoints with repeated measurements. Time-to-Event datasets contain one record per subject per analysis parameter for survival analyses. Occurrence Data Structure datasets support analysis of adverse events and other categorical outcomes. Each structure type has specified required variables and conventions that ensure consistency.

The relationship between SDTM and ADaM is foundational to regulatory submission integrity. ADaM datasets must be traceable to their SDTM sources, meaning that any derived variable or analysis population flag can be verified against the underlying tabulation data. This traceability is documented through ADaM analysis dataset specifications and define files that describe the derivation of each variable. Regulatory reviewers rely on this traceability to verify that analysis results accurately reflect the collected data, making clear documentation of SDTM-to-ADaM mapping essential for submission success.