A New Era

Shiny-based Clinical Submissions using WebAssembly

Eric Nantz

Eli Lilly & Company

posit::conf(2024) | August 13, 2024

bit.ly/shinywebr

The results are in!

A clinical trial just achieves their primary database lock. A multitude of analyses and visualizations are generated with the “real” data. Imagine you are reviewing these results. Sure, the primary analysis results meet the benchmark for statistical significance. But this is a complex disease state, requiring you to investigate how key efficacy, lab measures, and safety link together.

Many Results

In the not-so-distant past, the results of a clinical trial were sent to reviewers as paper outputs that could fill a small moving truck

Time to comb through many static outputs. How many outputs are there total? Over 1,000 tables, listings, and graphs.

Can we do better?

Instead of opening / reading these static outputs, what if we had a streamlined interactive application to optimize the review process? The good news is that the technology is here now, thanks to the amazing innovations powered by open-source. Over 10 years ago, Shiny was released to empower R users with the ability to create interactive feature-rich web applications using R code.

More recently, the webR package was revealed at last year’s posit conf to bring the power of web-assembly to the R ecosystem

R Submissions Working Group

• Industry and Regulatory members
• Create submission deliverables with open-source software
• Emulate realistic submission review
• Innovation encouraged!

R/Pharma 2018

I saw firsthand the potential of such a collaboration with the advent of the R/Pharma conference.

Install & Execute Application

• Extract submission bundle
• Install R
• (Optionally install RStudio)
• Install Rtools for package compilation
• Install {renv} R package
• Install application package library
• Run application in R process

Install & Execute Application

• Extract submission bundle
• Launch App in web process
• View App in web browser

Traditional Shiny

Web-Assembly

Engine Swap

• Verify packages compatibility with WebR

  • {golem} {rhino}

• Augment shinylive::export to inject SAS data sets

• User-friendly wrappers

• Test with custom Windows virtual machines

driver.R

source("utils.R")
extract_app_bundle()
run_app_webassembly()

With Great Power…

• Dynamically updates p-values based on filtered observation
• Potential for confusion and misinterpretation
• Compromise: Allow filtering within visualization module only

The Future Ahead

• Submission to FDA in progress
• Investigate reproducibility features
• Streamline execution process
• Complete container version (Docker)

Acknowledgements

• Sam Parmar (Pfizer)

• Ben Straub (GSK)

• Robert Devine (J&J)

• Saghir Bashir (Argenx)

• Ning Leng (Roche)

• Joe Rickert

• Pedro Silva (Jumping Rivers)

• André Veríssimo

• Vedha Viyash

• Pawel Rzymkiewicz (OlympCode)

• Paul Schuette

• Hye Soo Cho

Thank You!

R-Podcast

R Weekly Highlights

Shiny Developer Series

@rpodcast

eric-nantz

rpodcast