RShiny Dashboard Approach for Supporting Interactive DMC Review

[pzhang-cims]

Proposal

1. Pain Point
   1. DMC Importance: Per FDA draft guidance on DMC Feb 2024, DMC is a unique committee among all oversight groups that are the only one having access to unblinded accumulating safety and efficacy data. To enable DMC to effectively review ongoing study data and provide informative recommendations, it is important to provide innovative tools for DMC to fulfil its duty.
   2. Before DMC Meeting, DMC Committee usually will have only 1 week review such TFL, which will include thousands of pages. This is similar to a previous argument that in the Posit blog post (see below for reference)
      The prevalent use of .pdf files, while a step up from paper, often results in unwieldy documents with tables sprawling across numerous pages.
   3. DMC Committee will join closed session with independent statistical group, to discuss the unblinded results, usually can be 1 to 2 hours.
   4. During the closed session, DMC reviews the unblinded data in the form of tables/figures/listings (TFLs) and usually raises many at hoc questions and analyses which the static TFLs are not sufficient to address. The current approach is to take additional days for Sponsor and independent statistical group to prepare the results and send it back again.
2. Proposal
   1. Inspired by recent Pilot submissions through Rshiny, DMC review can be similar to FDA review.
   2. Interactive dashboard using R to perform analysis, which can be inclusive to study-specified analysis and endpoint. See {rtables}, {tern}, {teal}
   3. Providing aids to DMC for to effectively and interactively review data (TFLs) for ongoing studies.
   4. Beside interactive review, the dashboard approach could provide audit trails for complying with FDA’s new guidance on DMC practice.
   5. Providing downloadable/portable files for further reference
   6. Although there are some similarities with the NDA submission, the server/ Rshiny dashboard approach to DMC could be much easier, and can improve the effectiveness of DMC practice.

We would like to have a roundtable discussion on this topic regarding feasibility, technical challenges and acceptance.

Expected impact

1. Development Efficiency: R language is natural to develop this. They are all embedded within R environment.
2. Compatibility: Compatible with existed R package, including functional package or statistical package
3. Improved reviewer experience: can review TFLs efficiently and effectively in by interaction. Recall that thousand pages of TFLs previously include same type of tables/figures under different population, same layout for different parameters. If reviewer review the TFLs, they need to search the keyword by ctrl+F. While listing provides a reference level for subject-level information, people usually search for patient’s ID from more than 10K pages. This can be done through filtering from dashboard.
4. Adaptive feature for each study: while usually dashboard/package is developed for general purpose, one might expect to develop study-specific features. This can be done through {teal}. Developer can maintain the server and module instead of maintain static code for deriving TFLs (Teal (r-consortium.org), page 12)
5. Perform adhoc analysis during closed session meeting.

Prior discussions/work

Use of Data Monitoring Committees in Clinical Trials
Teal (r-consortium.org)
pharmaverse blog - Testing Containers and WebAssembly in Submissions to the FDA
How Pharma is Pioneering WebAssembly with webR & Shiny for FDA Clinical Trial Submissions - Posit
Pilot 2 Shiny webR Application (brilliant-elf-2da930.netlify.app)
Just say no to data listings! - PubMed (nih.gov)
WP061.pdf (phuse.s3.eu-central-1.amazonaws.com)

Would you be willing to potentially facilitate this discussion?

Yes definately!