RConsortium/submissions-pilot2

Additional table: Patients Remaining in Study

Closed this issue · 1 comments

Per feedback from FDA: I do have a suggestion for an additional table of Subjects Remaining in Study, as illustrated below (thank you Hye Soo). Not sure why the End of Treatment numbers go back up (LOCF ?), but it does suggest a pronounced differential drop out effect.

image

Documenting the data processing logic I used to create the summary table:

  • The example summary shared by FDA reviewers was derived from the labs data set (adlbc). I decided to use the adsl data set instead, as it offers a simple way of obtaining the patient disposition status as well as the visit number they left the treatment period (VISNUMEN).
  • I created a simple (long-format) data frame that expands appropriate visit records for each patient's participation in the trial, up to the recorded visit number in VISNUMEN, which allows me to summarize the counts of patients who were present in each visit. Example output below:
USUBJID TRT01P VISNUMEN VISITNUM VALUE VISIT
01-701-1015 Placebo 12 3 1 Baseline
01-701-1015 Placebo 12 4 1 Week 2
01-701-1015 Placebo 12 5 1 Week 4
01-701-1015 Placebo 12 7 1 Week 6
01-701-1015 Placebo 12 8 1 Week 8
01-701-1015 Placebo 12 9 1 Week 12
01-701-1015 Placebo 12 10 1 Week 16
01-701-1015 Placebo 12 11 1 Week 20
01-701-1015 Placebo 12 12 1 Week 24
01-701-1023 Placebo 5 3 1 Baseline
01-701-1023 Placebo 5 4 1 Week 2
01-701-1023 Placebo 5 5 1 Week 4
  • The long data frame was then fed into a custom {Tplyr} table. The key step was to ensure the number of patients was based off of the "original" adsl data set and not this custom long-form version. Thankfully this was super easy to implement. I verified that the number of patients who completed week 24 matches Table 14-1.02 (Summary of End of Study Data) within the CDISC Pilot 1 CSR.