SME input

[samussiah]

• oncology is a broad therapeutic area - need to focus on sub-areas of oncology
• solid tumor - how quickly and how significantly you can shrink the tumor
• hematological - decrease/increase of parameter
• PRO data - how can it be incorporated alongside traditional clinical data - e.g. quality of life as a differentiator when response rate is only non-inferior
• spider/radar chart - choose a set of continuous/ordinal outcomes and animate change from baseline
• short term and long term dynamics - quality of life
• capture event or progression itself in the bar chart race -
• think about using the solar system animation to visualize state change
• radar chart for PRO? check out Doug's PRO data visualization
• composite PRO data against tumor response data in a Gapminder-esque animated scatter plot over time

[samussiah]

TODO:

High priority

• add view that aligns start of current response across all participants
  • add a control to align bars at start of current response
  • x-range would not change but x-domain would need to be split in two: time to current response on the left, duration of current response on the right... might be unintuitive
  • annotate durable response, e.g. draw a vertical line at 8 weeks indicating durability
• add visit-level pauses at median study day of visit that overlays text describing the current level of response

Low priority

• add a control to set animation speed
  • should not reset animation to beginning
• add a control to set durable response duration
• add a control to sort by duration of current response or by time to current response
• add a control to align bars at end of time to any response
  • similarly x-range would not change but x-domain would need to be split in two: time to any response on the left, duration of current response on the right... might be unintuitive
• add a control to set the upper limit of the x-axis to the current timepoint or the final timepoint
• add an introductory, static explanatory screen - color coding, x-axis, y-axis - like a tutorial
• add interactivity
  • click a bar to view the participant's details
  • click a bar to highlight that participant and their progression

Ideas

• sync speed of animation with rate of change?
  • e.g. speed up between visit windows and slow down around visit windows
  • or have a shorter duration for timepoints without change and a longer duration for timepoints with change
• longer bars are intuitive but how can we emphasize good outcomes over smaller time periods?

Bugs

• response bars on the right side aren't transitioning
• no transition between study and current response views

[samussiah]

Feedback - 2021-04-19

• use OS and PFS to simplify chart - PD/SD/PR/CR is more granular than necessary
• can we simplify to a sort of state diagram to use the solar system animation?
  • how do we capture durability of response?
  • do we care about previous reponses? difficult to capture with the solar system animation
• TODO: add stratification by treamtent group !IMPORTANT
  • how to best compare two groups of different sizes
  • scale the height of the bars to the height of the chart canvas?
  • then do we need to align the response groups across the two panels?

[samussiah]

Feedback - 2021-04-21

• duration of any non-progressive response (which isn't the same as PFS)
• re-shuffling is difficult to follow
  • rethink this feature
  • position participants initially in the category of their first response to avoid excessive reshuffling
  • transition one response category at a time or one direction at a time: responders then non-responders
• transplant: alternative positive response
• transplant and progressive disease aren't necessarily associated with a duration
• CR > PR would be unusual
• incorporate the additional characteristic of something like hematologic improvement
• two dimensions: reduction in bad cells (blasts), improvement in good cells (marrow cells)
• depth of response, hematologic improvement are more important than the timing aspect of the animation in certain situations
• normalizing follow-up between transplants and continued non-progression

[samussiah]

Feedback - 2021-04-23

• how can we compare arms?
  • side-by-side plots
• add progression-free survival
  • if we sort by time to progression we basically have a survival plot - just a matter of sorting the IDs by ascending time to progression - censored participants
• can we highlight participants with specific treatments - NAD, IDT, etc.
• incorporate subgroups in forest plot from paper - filtering or stratifying
• add a subgroup control
• it would be informative to support annotation of events-of-interest - skeletal-related events, hematologic improvement, etc.

[samussiah]

Feedback - 2021-05-03

• Add sort by total duration as a survival curve view.
• Add median duration of response - participants >median sorts upward, <median sorts downward.
  • This approach allows comparison of efficacy between arms.
  • Group participants by treatment group?
• Add a benchmark or destination - durable response?
• Rather than a continuous bar draw one bar segment per day that moves from the right and lands on the left - piling up toward durable response. Or just dropping down from above the bar into place.

Priorities

• Let's not group by response type - PDs are going to happen and take focus away from the positive.
  • Add the overall survival curve view (transplants still at the top?).
  • Add the PFS view.
  • Add a sort and group control.
• Update x-axis unit to months, say, or 28-day cycles.
• Emphasize durability - brighten participants who have achieved durability, overlay a "finish line": x cycles. Make durability configurable.