Closes #143 #141 #140 ADRG updates for Resubmission v3

[bms63]

Address concerns for Pilot 5 raised by the FDA in the April and May Submissions working gorup meetings

Clean Global Env

• Contains code snippet in ADRG to clean up Global Environment Remove intermediate datasets from working environment on program calls #141
• Renamed objects for clarity Remove intermediate datasets from working environment on program calls #141

Snippet of code for datasetjson viewing

• Small snippet to easily see datasetjson Include sample code in ADRG #140

Classes and Formats

• Inclusions of explanations form datasetjson r package dev Get clarification on datasetsjson package issues where we mention discussion with package team #143

NAs introduced by coercion

[github-actions]

Dataset QC Check Complete!

Date: 2026-07-06
Run by: Copilot
Git Ref: refs/pull/148/merge
BASE: Pilot 5 Datasets
COMPARE: Pilot 3 Datasets

✅ Dataset: adadas

No issues were found!

✅ Dataset: adae

No issues were found!

✅ Dataset: adlbc

No issues were found!

✅ Dataset: adsl

No issues were found!

✅ Dataset: adtte

No issues were found!

[github-actions]

TLF QC Check Complete!

Date: 2026-07-06
Run by: Copilot
Git Ref: refs/pull/148/merge
BASE: Pilot 5 Datasets
COMPARE: Pilot 3 Datasets

✅ Dataset: tlf-demographic-pilot5.out

old[36:40] vs new[36:40]
  "—————————————————————————————————————————————————————————————————————————————————————————————————"
  ""
  "Program: tlf-demographic.r "
- "2026-06-28 23:13:55.397003"
+ "2024-04-18 20:17:25"
  ""

✅ TLF: tlf-kmplot-pilot5.pdf - Score: 75

[1] "1. Title formatting: First image has 'KMplot for Time to First Dermatologic Event - Safety population' while second image has 'KM plot for Time to First Dermatologic Event-Safety population' (spacing differences around dash)\n\n2. Y-axis label: First image shows 'Survival Percent' while second image shows 'Probability of event'\n\n3. Grid styling: First image has lighter gray gridlines, second image has darker/more prominent gridlines\n\n4. Legend position: First image has legend positioned on the right side of the plot area, second image has legend labeled 'TRT01A' positioned inside the upper right corner of the plot\n\n5. Legend labels: Both contain Placebo, Xanomeline Low Dose, and Xanomeline High Dose but with slightly different formatting\n\n6. Confidence interval shading: The shaded regions around the curves appear slightly different in opacity and style between the two images\n\n7. At-risk table: The numbers differ significantly between the two images at corresponding time points (e.g., at baseline Placebo shows 86 in first image vs 86 in second; Low Dose shows 84 vs 84; but subsequent timepoints show different values)\n\n8. Bottom timestamp: First image shows 'Program ID Internal (2025-09-28 20:11:46.62TZ)' while second shows 'Program: [!!kmplot.r | 202404~18 20:17:28]'\n\n9. Axis ranges and tick marks: Slight differences in the exact positioning and formatting of axis ticks\n\n10. Line thickness: The survival curves appear to have slightly different line weights\n\n11. Overall plot dimensions and aspect ratio appear slightly different"

✅ TLF: tlf-efficacy-pilot5.rtf - Score: 99

[1] "1. Timestamp difference: The first image shows a timestamp of '2024-04-18 20:17:28' at the bottom of the table, while the second image shows '2024-04-18 20:17:18' - a 10-second difference.\n\n2. All other visible elements appear identical: The table title 'ANCOVA of Change from Baseline at Week 20', all column headers (Treatment, Week 20, Change from Baseline), all data values, sample sizes (N), means, standard deviations (SD), confidence intervals (95% CI), and footnotes are exactly the same in both images.\n\n3. The table structure, formatting, borders, and layout are identical.\n\n4. All statistical values including the ANCOVA results (0.074, 0.256 OM, -0.028, -0.135 OM for the two treatment groups and 0.102, 0.392 for differences) are the same.\n\n5. The footnote text explaining the ANCOVA methodology and abbreviations is identical in both images."

✅ TLF: tlf-primary-pilot5.rtf - Score: 99

[1] "After carefully examining both images, I found only one difference:\n\n1. Footnote text difference: In the first image, the footnote at the bottom of the second table reads 'Based on an Analysis of covariance (ANCOVA) model, with treatment and baseline value as a covariate.' In the second image, this same footnote reads 'Based on an Analysis of covariance (ANCOVA) model with treatment and baseline value as a covariate.' The difference is the presence/absence of a comma after 'model' - the first image has a comma while the second image does not.\n\nAll other elements appear identical:\n- Both tables show 'Protocol: CRES26908/Stage 1+2' and 'Population: SAFS01'\n- Table 14.2.1 and Table 14.2.2 titles are identical\n- All numerical values are the same across both images\n- Column headers (Placebo, Low Dose Xanomeline, High Dose Xanomeline) are identical\n- All statistical data points match exactly\n- The layout, formatting, and structure are the same\n- All other footnote text is identical"

[parmsam-pfizer]

Do we need to expand on this, just in case? @bms63

[bms63]

@copilot can you add an explanation to this issue with NAs. I believe it is due to how datasetjson R package is reading in the data.

1: In lapply(d[dbl_cols], as.double) : NAs introduced by coercion
2: In lapply(d[dbl_cols], as.double) : NAs introduced by coercion
3: In lapply(d[dbl_cols], as.double) : NAs introduced by coercion
4: In lapply(d[dbl_cols], as.double) : NAs introduced by coercion

Here is location for repsonse from developers of datasetjson - atorus-research/datasetjson#76. Can you also link the issue like how it was done in other sections please

[Copilot]

Added the explanation and linked the upstream datasetjson issue in the ADRG section. The new text notes that read_dataset_json() warns when as.double() converts string placeholders like "NA" to missing numeric values, which matches the package team’s explanation in issue #76. Addressed in 6f8d656.

[parmsam-pfizer]

Only one comment. The rest of the changes look good to me!

[parmsam-pfizer]

@copilot Update label = scales::label_number(accuracy = 0.01) to labels = scales::label_number(accuracy = 0.01) in tlf-kmplot.r

[bms63]

@parmsam-pfizer i remembered we have to rebuild the combined pdf manually :(. I'll try and do that tmr

[robertdevine]

@bms63 @parmsam-pfizer @RConsortium/pilot5 - fresh run of p5 combined report. Reproducible in Windows, (w10,w11) Linux, macOS - bare metal (sponsor-issued laptop) and vm.
cmb-report-tlf-pilot5.pdf

*[For anyone interested, testng different formats - what happens with Table 14-3.01 depending upon which pdf rendering engine the table (in some cases) spans two pages which if not explicitly included in the image graphics does not render onto the pdf report. That said, a verbose log of the p5 primary program output clearly shows the Table 14-03.1 is captured in its entirety and matches exactly the previous pilot submmission reports.]

[parmsam]

Pages 2 and 4 seem to be different when comparing to the previous pilot5-cmb-report-manual.pdf

• On page 2, LS Means differences and confidence intervals for the pairwise comparisons are missing, as well
  as the entire Xan High vs. Xan Low sub-group comparison
• On page 4, the plot is less wide than before.

[robertdevine]

@parmsam @bms63 @RConsortium/pilot5 Today's run of pilot5 combined report attached. Appears visually exact to pilot1 and pilot3 reports.
cmb-report-tlf-pilot5.pdf

The page 4 item we can format as necessary. If memory serves from the general meeting an agreed setting would be established. Possibly fine already given the substantial similarity to prior submission reports. Ty

[parmsam]

Thanks, @robertdevine! Agreed that page 4 matches with P1 and P3 but I touched base with Ben and we're pretty sure FDA reviewers wanted us to reformat that page 4 plot so the width was increased, so it is more easily readable. Sorry, we'll need to regenerate that page to also be wider width. Everything else looks great!

[parmsam]

Fixed pdf included in this message:
pilot5-cmb-report-manual.pdf

Had to comment out line of code from qmd file. See #152 for PR from fork into this branch which has the fixed PDF and qmd.

[robertdevine]

Thanks @parmsam-pfizer @RConsortium/pilot5 - yes, did manage to expand the width of the table on page 4. closer to p1 and p3 Also the page 5 km-plot is wider in the pilot5 report than in pilot 1 and pilot 3 and the ordinate label position does not match the earlier pilots. Reasonable to expect slight differences in quarto outputs and the rtf route used in the earlier pilots. Should be good to go on it - these are all slight formatting changes at the render. Ty
cmb-report-tlf-pilot5.pdf

[bms63]

The programs need to be re-run and combined into the pdf with updated timestamps. 3/4 have timestamps from January

Lintr is failing as well - I think the commented out code can just be removed if it isn't being used to run the programs. I was using it to run one of the programs - I think??

[parmsam-pfizer]

One step forward, two steps back, haha. Sorry, I should time to fix this fully tomorrow evening when I have access to my Windows machine. Easy fixes. Removed the line already to get the lintr to pass. I'll ensure the file timestamps get updated. We should be able to reach out to Beverly on Monday.

[parmsam-pfizer]

Update label = scales::label_number(accuracy = 0.01) to labels = scales::label_number(accuracy = 0.01) in tlf-kmplot.r

This is actually incorrect b/c scale_ggsurvfit applies some modifications. Could also be a version issue with the ones we used compared to the latest one.

[parmsam-pfizer]

Made a bunch of fixes and reran the combined report to get it working. Tagging @bms63.

Updated combined pdf: https://github.com/RConsortium/submissions-pilot5-datasetjson/blob/1d8fc8ea09f7582dd746370b00ec5b0c71f4ba41/pilot5-submission/pilot5-programs/pilot5-cmb-report-manual.pdf