Leveraging R and Shiny for Point and Click ADaM Analysis Ian Fleming and Fred Hofstetter NJ CDISC User Group January 2015
Agenda Lifecycle of a TFL The Promise of Standards Overview of the Tool ADaM Viewer Demonstration Q&A
Lifecycle of a TFL
How does Pharma get here?
How do we get to a TFL? Protocol/Analysis Plan Data Collection Protocol – Design and Endpoints Analysis Plan – Details about the analysis Data Collection Case Report Forms Collection of Data at sites in data capture systems SDTM/ADaM SDTM: Collection of raw data in data sets ADaM: Creation of analysis data in data sets TFL Programmed Tables, figures, and listings Included in study report or Integrated Summary
Extensive Process Transitions from one form to another require significant effort Significant amount of single use programs Use of “Validated Systems” Typically SAS Macro based infrastructure Company specific infrastructure
The Promise of Standards CDISC formed in 1997 “to develop and support global, platform-independent data standards that enable information system interoperability to improve medical research and related areas of healthcare.”
Jetpacks
Fundamental Question Why hasn’t standards adoption brought the levels of efficiency that we were expecting Tools? The standards? The Industry? How do we explore the cause?
Rapid Prototyping Originated in manufacturing Facilitates real world testing of solutions Development occurs through iteration De-facto standard methodology for web development
Rapid Prototyping Design Prototype Test Collect Feedback Improve
The ADaM Viewer
Motivations Proof of Concept for Rapid Prototyping methodology The ability to build standard tools off of ADaM data The feasibility of R and Shiny for this type of work
Brainstorming Requirements Ability to read in ADaM submission transport files Ability to produce minimal set of standard summaries Point and click interface – no end user programming required No install needed FREE!
CDISC Tools Lots of tools for some standards CDASH (EDC systems, standard CRFs, etc.) ODM (in/out from different data collection systems) SDTM (validation, data visualization tools)
ADaM
Technology Options SAS? Java? R? Need license(s) No quick/easy point and click without other tools Extensive knowledge of SAS stack needed Java? Lot of coding Steep skill set R?
R Early History – 1990 Ross Ihaka and Robert Gentleman Department of Statistics at the University of Auckland Open source statistical analysis software based on S programming language Package based Functional specific extensions
R: Early History https://www.stat.auckland.ac.nz/~ihaka/downloads/Massey.pdf If you want to know more…
Shiny Web application framework for R Package installed in R Interactive data analysis with real time code execution based on user input Web technology without having to know web technology Minimal Infrastructure requirements
Fully functional prototype 3 weeks later… Prototyping complete Fully functional prototype Ability to read in ADaM submission transport files Ability to produce standard types of summaries Point and click interface No install needed
Demonstration
Results Rapid Prototyping Standard tool for ADaM Analysis 3 weeks from concept to full prototype 2 resources working in their spare time Standard tool for ADaM Analysis Consistently create results across any ADaM data R and Shiny Very easy to create and deploy
Additional Benefits Ability for non-technical people to look at analyses Removing roadblocks to data Ad-hoc confirmation of current analyses Easily extendable Easily accessible Low/No cost
Summary Rapid prototyping is a valuable tool Next step: incorporate into our development process and interactions with users R provides tools and packages for quick and powerful application development Next step: how can we leverage this on a larger scale? Able to produce easy point and click analysis for ADaM Next step: Options for a universally available solution?
Questions