This presentation reflects the views of the author and should not be construed to represent FDA's views or policies.

About Myself

• A statistician/statistical analyst in CDER's Office of Biostatistics for over a year
• Previously in the Office of Pharmaceutical Quality as an ORISE Fellow working on post-marketing stat analysis
• Developing R Shiny Apps for clinical reviews, such as
  • Patient-reported outcomes (PRO) visualizations
  • Forest plots for subgroups
  • Multiplicity procedures
• Providing statistical support to reviews and data standards
• Organizing an R Shiny Users Group within CDER

Agenda

1. What is Shiny?
2. How does Shiny work?
3. How is Shiny used at the FDA?
4. How can I start using Shiny?
5. Take Home Points

Shiny Intro

• Only requires programming experience in R to develop an app
• Used in various disciplines ranging from biostatistics, statistics education, finance, etc.
• Free of charge to develop but cost of a server is expected for vast distribution
• New features are continuously being developed by the RStudio team

Shiny’s Advantages

• Data analyses and visualizations can be programmed into a user-interface format starting from base R code
• Combined with RMarkdown, routine reports can be generated for higher productivity
• In the regulatory setting, standardized tools can be created for efficiency in conducting reviews
• No HTML, CSS, or JavaScript experience required but could provide enhancements
• Access to Shiny apps can be easily done through a Shiny server

Components of a Shiny App

• ui.R
  • Designates the layout of user inputs and outputs
  • Assigns the variables, in the forms of slider bars, drop-down menus, etc., to receive reactive values from users
  • Sends it to server.R
• server.R
  • Performs computations and tasks
  • Receives user inputs to execute code
  • Returns ouputs to ui.R

Launching a Shiny App (locally)

• runApp tracing the directory

runApp("C:\\Jimmy\\MyFolder")

• Hit Run App in RStudio
  
  • An advantage of using RStudio is that you can easily choose how the app is launched

R Shiny Users Group

• Initial planning meeting took place on May 4, 2017
• Branched out from the Data Review Committee (DRC) in the Office of Biostatistics
• Each session would focus on a specific topic/theme, such as
  • Shiny demos from other offices
  • Shiny challenges
  • Shiny updates

Shiny Apps at the FDA

Examples of currently in use and in development apps

• Forest plots for subgroup analysis
• Psychometrics analysis with MPlus
• Patient-reported outcomes (PRO) visualizations
• Gantt chart for review timeline
• Data quality assessment
• Text mining on medical terms

R Shiny MediaWiki Homepage

Purpose:

• Communicates information to staff across CDER
• Provides a brief introduction to Shiny and its capabilities
• Hosts information for each Shiny app and instructions on how to connect to them
• (Future) Hosts R Shiny users group information such as goals, members, projects, status, accomplishments, etc.

R Shiny MediaWiki Homepage

R Shiny MediaWiki Homepage

Shiny Server

• A Shiny server allows users to directly launch an app using its URL in an internet browser
• No need for having R and/or RStudio installed to run apps
• Pilot study will start in CDER's Office of Biostatistics with several apps on the server
• Beta testers will be recruited to “break” the apps
• Test out the security and compliance of a server

Example: R Shiny Forest Plots App

• Produces forest plots for subgroup analysis
• Two data structures are supported: imputed and non-imputed
• Users have the flexibility to choose the parameters to focus on
• Additional adjustments can be made by users
• Output can be directly exported from the app

Example: R Shiny Forest Plots App

Example: R Shiny Forest Plots App

Example: R Shiny Forest Plots App

Example: R Shiny Multiplicity App

• Applies multiplicity procedures in scenarios with multiple endpoints, doses, populations, etc.
• A variety of procedures can be used ranging from the classic Bonferroni to the more advanced Superchain (Kordzakhia and Dmitrienko, 2012)
• Adjusted p-values, decision rules, and confidence limits can be obtained
• Users can output results in either HTML or Word format using knitr

Example: R Shiny Multiplicity App

Example: R Shiny Multiplicity App

Example: R Shiny Multiplicity App

Example: R Shiny Multiplicity App

Starting with Shiny

• https://shiny.rstudio.com/tutorial/
  • RStudio has an excellent tutorial with 7 lessons to start learning about Shiny
  • More advanced articles are also available for those who have an in-depth experience
• https://shiny.rstudio.com/tutorial/js-lesson1/
  • A JavaScript widget tutorial is available to create JS widgets in Shiny
• http://www.statistics.calpoly.edu/shiny
  • A collection of Shiny apps developed by Cal Poly's Shiny group

Take Home Points

• Shiny is a free library in R that is used to develop interactive web applications
• There is a great flexibility for various things that could be created using Shiny
• Shiny apps can be easily accessible through a browser on a computer, tablet, or phone
• Offices in FDA recently have begun developing Shiny apps that are “fit for purpose”
• Shiny apps can serve as standardized tools to increase efficiency in drug reviews
• An R Shiny users group is being proposed (PhUSE, ASA Statistical Computing Section, and others?)