Improving the Efficiency and Ensuring the Quality of Data Assembly for Pharmacometric Analysis
Thaddeus Grasela* (1), Jill Fiedler-Kelly (1), Darcy Hitchcock (1)
(1) Cognigen Corporation; Buffalo, NY, USA
Introduction: The creation of an analysis-ready dataset consists of preparing a time-ordered sequence of events for each subject, based on a statement of clear and concise specifications for the analyses. Once data programming begins, the programmer is likely to face a host of issues that arise from deficiencies in the specifications or inconsistencies between the specifications and raw data. These issues typically spawn a series of e-mails and discussions between the project team members. In the process of answering these questions, more specific questions are formulated as the team members clarify issues and resolve uncertainties. The cycle of questioning and discussion is a valuable source of information on how to improve specifications and reduce the time and effort required for data assembly.
Objectives:
- Analyze e-mail communications between team members to identify common issues
- Incorporating findings into formalized programming specifications
Methods: A systematic analysis of communications between programmers and pharmacometricians generated during the course of data assembly for numerous independent projects was performed. Information extraction and discovery techniques were utilized to uncover the most frequently discussed topics, issues and problems. These issues and the attendant queries and answers were used to develop formal programming specification forms.
Results: Inadequate or incomplete information regarding analysis population selection and dosing were common issues. Of these, issues pertaining to dose were most important. For example, proper dosing specification requires instructions on handling first dose versus multiple dose, criteria for ascertaining steady-state, and instructions for managing multiple doses prior to achieving steady-state.
The specification forms were deployed in conjunction with a secure, website (wiki) to capture communications between scientists and programmers in a structured format. System-generated and user-provided metadata serve to facilitate the discussion and resolution of residual issues. Subsequent review will enable future refinement of the specification forms.
Conclusions: The development of these specification forms is anticipated to improve the performance characteristics of data assembly in terms of consistency, reliability, timeliness and quality. Future work will involve quantifying the benefits of these improved programming specification forms.