| Metadata excerpt: ADSL variables used in this report | ||||
| define.xml-style excerpt, hand-curated | ||||
| Dataset | Variable | Label | Type | Codelist |
|---|---|---|---|---|
| ADSL | USUBJID | Unique Subject Identifier | text | |
| ADSL | TRT01P | Planned Treatment for Period 01 | text | TRT01P |
| ADSL | TRT01A | Actual Treatment for Period 01 | text | TRT01P |
| ADSL | AGE | Age | integer | |
| ADSL | AGEGR1 | Pooled Age Group 1 | text | AGEGR1 |
| ADSL | SEX | Sex | text | SEX |
| ADSL | RACE | Race | text | RACE |
| ADSL | SAFFL | Safety Population Flag | text | NY |
| ADSL | ITTFL | Intent-To-Treat Population Flag | text | NY |
| ADSL | PPROTFL | Per-Protocol Population Flag | text | NY |
| ADSL | COMPFL | Completers Population Flag | text | NY |
| ADSL | DCSREAS | Reason for Discontinuation from Study | text | |
| ADSL | TRTDURD | Total Treatment Duration (Days) | integer | |
Illustrative excerpt only, covering the ADSL variables referenced elsewhere in this report; not a validated define.xml. NY is the CDISC controlled terminology codelist for No/Yes flags. |
||||
Appendix B — Reproducibility
B.1 Source data
The analysis datasets are the CDISC pilot study datasets (study identifier CDISCPILOT01) distributed in the pharmaverseadam package. R/00-adam.R reads those datasets, restricts them to the randomised treatment groups, derives the treatment-emergent flags used throughout the report, and derives the time-to-event dataset with admiral::derive_param_tte().
B.2 Metadata excerpt (define.xml style)
A define.xml documents every submitted dataset and variable: its name, label, data type and, where applicable, the controlled terminology codelist it draws from. Tools such as metacore build a validated metadata object from that specification; the excerpt below hand-curates a subset of ADSL variables used in this report in the same shape, without generating or validating a full define.xml.
B.3 Analysis results traceability
Every summary table in this report is computed by gtsummary, which retains the Analysis Results Dataset (ARD) behind each statistic, following the CDISC Analysis Results Data standard. Each table’s ARD can be recovered from its table object, and cards::bind_ard() combines the ARDs from several tables into a single dataset, which is the mechanism used for quality control of the numbers reported here.
disposition_ard <- gtsummary::gather_ard(tlf_disposition(adsl))[[1L]]
demographics_ard <- gtsummary::gather_ard(tlf_demographics(adsl))[[1L]]
ae_overview_ard <- gtsummary::gather_ard(tlf_ae_overview(adae, adsl))[[1L]]
ard <- cards::bind_ard(disposition_ard, demographics_ard, ae_overview_ard, .quiet = TRUE)
head(as.data.frame(ard)[, c("variable", "stat_name", "stat")], 8) variable stat_name stat
1 Reason for discontinuation n 8
2 Reason for discontinuation N 28
3 Reason for discontinuation p 0.2857143
4 Reason for discontinuation n 9
5 Reason for discontinuation N 28
6 Reason for discontinuation p 0.3214286
7 Reason for discontinuation n 2
8 Reason for discontinuation N 28The combined ARD traces every statistic back to the table (Table 14.1.2, Table 14.1.3 or Table 14.3.1.1) and variable that produced it:
as.data.frame(dplyr::count(ard, variable, name = "n_rows")) variable n_rows
1 ..ard_total_n.. 1
2 AGE 32
3 AGEGR1 36
4 Any TEAE leading to withdrawal 26
5 Any TEAE with fatal outcome 26
6 Any dermatologic TEAE 26
7 Any drug-related TEAE 26
8 Any serious TEAE 26
9 Any severe TEAE 26
10 Any treatment-emergent adverse event (TEAE) 26
11 Completed the study 26
12 Died 26
13 Discontinued the study 26
14 ETHNIC 35
15 RACE 45
16 Randomised 26
17 Reason for discontinuation 90
18 SEX 36
19 TRT01A 12
20 TRT01P 12
21 TRTDURD 32
22 Treated (safety population) 26B.4 Session information
| Package versions used for this report | ||
| Package | Version | Source |
|---|---|---|
| dplyr | 1.2.1 | RSPM |
| ggplot2 | 4.0.3 | RSPM |
| gtsummary | 2.5.1 | RSPM |
Key packages are cited in the reference list: admiral (Mancini et al. 2026), gtsummary (Sjoberg et al. 2021), survival (Therneau 2026), ggsurvfit (Sjoberg et al. 2026), gt (Iannone et al. 2026).
B.5 Table shells
Table shells fix column layout, statistics and placeholder text at SAP sign-off, before database lock. The final report differs from the shell only in the numbers, not in structure. The mockup in Table B.1 is the pre-lock shell for Table 14.2.1; the populated table in Table B.2 is the same table after database lock, reproduced here for direct comparison.
| Treatment group | Subjects | Subjects with an event | Median time to event (days) | Hazard ratio (95% CI) | p-value |
|---|---|---|---|---|---|
| Placebo | XX | XX (XX.X%) | XXX (XXX, XXX) | Reference | |
| Xanomeline low dose | XX | XX (XX.X%) | XXX (XXX, XXX) | X.XX (X.XX, X.XX) | X.XXX |
| Xanomeline high dose | XX | XX (XX.X%) | XXX (XXX, XXX) | X.XX (X.XX, X.XX) | X.XXX |
XX marks cells populated only after database lock. Column layout and statistics are fixed at SAP sign-off. |
|||||
| Treatment group | Subjects | Subjects with an event | Median time to event (days) | Hazard ratio (95% CI) | p-value |
|---|---|---|---|---|---|
| Placebo | 86 | 20 (23.3%) | Not reached | Reference | |
| Xanomeline Low Dose | 84 | 39 (46.4%) | 80 (55, NA) | 2.98 (1.73, 5.13) | <0.001 |
| Xanomeline High Dose | 84 | 39 (46.4%) | 89 (50, NA) | 3.34 (1.94, 5.75) | <0.001 |
| Medians and confidence intervals are Kaplan-Meier estimates. Hazard ratios come from a Cox proportional hazards model with planned treatment as the only covariate and placebo as the reference group. A median is reported as not reached where fewer than half the subjects in the group had an event. | |||||
quarto_version <- system2("quarto", "--version", stdout = TRUE, stderr = TRUE)
typst_version <- system2("quarto", c("typst", "--version"), stdout = TRUE, stderr = TRUE)
cat(
"R: ", R.version.string,
"\nQuarto:", quarto_version,
"\nTypst: ", sub("^typst ", "", typst_version)
)R: R version 4.6.1 (2026-06-24)
Quarto: 1.11.1
Typst: 0.15.1 (9dfd3a08)The exact package versions are pinned in renv.lock. Running renv::restore() followed by quarto render reproduces this report from a clean checkout.