Abstract:
|
Clinical trial (CT) data are captured and cleansed in an "operational database." Later, data from multiple domains (demography, AEs, etc) are combined into "analysis datasets," which are used for statistical computations and tabulations.
"Analysis Databases" are a new concept. A Study Analysis Database (SAD) is constructed for one CT. A Compound Analysis Database (CAD) combines data from SADs from all CTs of one compound. A SAD or CAD is scientifically engineered and contains CRF data, derived data, metadata, and documentation. Creation of a CAD involves integrating data from multiple studies with varying scientific objectives, often requiring resolution of intricate scientific issues.
A SAD or CAD is structured so that a large majority of all scientific/statistical questions that can reasonably be asked of the data can be answered in just one computational step, without preliminary processing steps (except sorting/subsetting). A SAD or CAD is much more than an operational database, a collection of analysis datasets, or a data warehouse. This paper presents principles and procedures of SAD/CAD construction, and sets the environment for the subsequent papers in the session.
|