A structured approach to industrialize the data sourcing to support model based drug development
Pinault Gregory, Ette Georges, McDevitt Hugh, Steimer Jean-Louis, Buchheit Vincent
Novartis Pharma AG
Objectives: The Novartis Modeling & Simulation (M&S) programmers recently revisited and harmonized their individual experiences of building modeling ready data files from clinical databases and other sources. The authors would like to highlight the different steps undertaken to industrialize the crucial part of any modeling activity which is the data preparation
Methods: As M&S staff increased over the past years [1], the experience and variety of ways to deliver data has also evolved and expanded. Meanwhile, the value of M&S support for key decisions at key stages in the drug development process is now widely recognized and is expected to increase as model based decision making is embedded within the wider organization of Novartis Development. Consequently, any model based analysis is now likely to be reported to health authorities. This implies state of the art preparation and validation [2] of the data. To face the increasing demand in a strict regulatory environment, the authors lead the industrialization of data preparation by revisiting the modeling data file composition, the programming organization, and the data request tracking and management.
Results: The number of clinical parameters of interest and the subjectivity of the modeling approach to investigating the available data, lead to inconsistencies between data requests. The regulators expect validation of data prepared for a modeling analysis to be submitted to them, which impacts the department resources. The request specifications have been fine tuned by identifying classes of variables and defining the key common properties to characterize them. This puts the focus of standardization on the methodology rather than the data output. The carry over of programs amongst individuals and projects has been optimized by simplifying the programming framework and workflow. In alignment with the two previous work streams, a request tracker has been designed. Every data request is saved in a database, which can then be used to build activity reports for management and more.
Conclusions: Industrialization of M&S activities implies faster and better access to validated data. Enhancing the process of bringing the required data together is laying the foundation for further industrialization [3]. The authors believe it is one step forward in blurring the boundaries between data sourcing and exploratory modeling.
References:
[1] Pinault G., et al. Quality, Efficiency and Industrialization Initiatives during the evolution of a dedicated SAS Programming Group. PAGE 18 (2009) Abstr 1513 [www.page-meeting.org/?abstract=1513]PAGE 18 (2009).
[2] Buchheit V., et al. Efficient quality review for modeling input dataset. PAGE 20 (2011) Abstr 2041 [www.page-meeting.org/?abstract=2041].
[3] McDevitt H., et al. A technology roadmap to support model based drug development. PAGE 20 (2011) Abstr 1950 [www.page-meeting.org/?abstract=1950].
