admiral-adsl
Derives an ADaM Subject-Level Analysis Dataset (ADSL) using the {admiral} R package and pharmaverse ecosystem. Use when a user needs to create ADSL from SDTM domains, derive standard subject-level variables (treatment dates, disposition, demographics, population flags), or generate QC-ready R code following CDISC ADaM conventions. Requires SDTM input data and an ADaM spec.
RConsortium · 已有 AI 审核记录
安装与使用
运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。
npx skills add RConsortium/pharma-skills --skill admiral-adsl技能说明
name: admiral-adsl description: > Derives an ADaM Subject-Level Analysis Dataset (ADSL) using the {admiral} R package and pharmaverse ecosystem. Use when a user needs to create ADSL from SDTM domains, derive standard subject-level variables (treatment dates, disposition, demographics, population flags), or generate QC-ready R code following CDISC ADaM conventions. Requires SDTM input data and an ADaM spec. license: MIT metadata: author: Navitas Data Sciences version: "0.2" pharmaverse: "true" parent: admiral compatibility: > Requires R with admiral, dplyr, lubridate, and pharmaversesdtm installed. Designed for use in a GxP-compliant environment with access to SDTM datasets and an ADaM ADSL specification.
admiral-adsl
Shared conventions (library setup, pipe style, date rules, flag convention,
# REVIEW:annotations,stopifnot()patterns) are defined in the parent../SKILL.md. The workflow below is ADSL-specific.
Derives a CDISC-conformant ADSL dataset using {admiral}. Outputs executable, QC-ready R code with derivation logic traceable to the ADaM specification.
See admiral-functions reference for function selection guidance. See adsl-conventions reference for CDISC variable conventions.
Inputs
Before generating code, confirm the following are available or explicitly noted as absent:
| Input | Required | Notes |
|---|---|---|
| DM | Yes | Subject spine; one record per USUBJID |
| EX | Yes | Exposure; needed for treatment dates and SAFFL |
| DS | Yes | Disposition; needed for EOSSTT, DCSREAS |
| DV | No | Protocol deviations; needed for PPROTFL |
| MH | No | Medical history flags if protocol requires |
| VS | No | HEIGHTBL, WEIGHTBL, BMIBL if in scope |
| ADaM ADSL spec | Yes | Variable list, derivation rules, grouping cut-points |
| Study context | Yes | Treatment arm names, population flag definitions |
If required domains are absent, stop and request them. If optional domains are absent, omit the corresponding derivations and note this in code comments.
Workflow
Follow these steps in order. Generate code section by section, not as a single block.
Step 1 — Setup and domain loading
library(admiral)
library(dplyr)
library(lubridate)
library(pharmaversesdtm)
# Load SDTM domains
dm <- pharmaversesdtm::dm
ex <- pharmaversesdtm::ex
ds <- pharmaversesdtm::ds
# mh <- pharmaversesdtm::mh # uncomment if in scope
# Confirm one record per USUBJID in DM before proceeding
stopifnot(nrow(dm) == n_distinct(dm$USUBJID))
Step 2 — Subject spine
Start from DM. One record per USUBJID is mandatory at this step and must be preserved throughout. Select only variables needed downstream.
adsl <- dm |>
select(
STUDYID, USUBJID, SUBJID, SITEID,
AGE, AGEU, SEX, RACE, ETHNIC, COUNTRY,
ARM, ARMCD, ACTARM, ACTARMCD,
DMDTC, RFSTDTC, RFENDTC,
DTHFL, DTHDTC
)
Step 3 — Treatment dates (TRTSDTM, TRTSTMF, TRTEDTM, TRTETMF, TRTSDT, TRTEDT)
Derive datetimes first, then extract date-only variables. Always include
time_imputation arguments and always retain imputation flags (TRTSTMF,
TRTETMF) in new_vars — setting flag_imputation = "auto" without capturing
the flag variables provides no traceability benefit.
Remove DOMAIN from EX before merging to avoid variable conflicts.
ex_dtm <- ex |>
select(-DOMAIN) |>
derive_vars_dtm(
dtc = EXSTDTC,
new_vars_prefix = "EXST",
date_imputation = "first",
time_imputation = "first",
flag_imputation = "auto"
) |>
derive_vars_dtm(
dtc = EXENDTC,
new_vars_prefix = "EXEN",
date_imputation = "last",
time_imputation = "last",
flag_imputation = "auto"
)
# TRTSDTM / TRTSTMF: first dose datetime and imputation flag
# REVIEW: The placebo filter (EXTRT == "PLACEBO") must be confirmed against the
# protocol. In some studies EXDOSE > 0 is sufficient; in others EXDOSE = 0
# for placebo and EXTRT must be used. Adjust condition per protocol definition.
adsl <- adsl |>
derive_vars_merged(
dataset_add = ex_dtm,
by_vars = exprs(STUDYID, USUBJID),
new_vars = exprs(TRTSDTM = EXSTDTM, TRTSTMF = EXSTTMF),
order = exprs(EXSTDTM),
mode = "first",
filter_add = (EXDOSE > 0 | EXTRT == "PLACEBO") & !is.na(EXSTDTM)
) |>
# TRTEDTM / TRTETMF: last dose datetime and imputation flag
# REVIEW: If subjects have non-contiguous EX records, TRTEDTM reflects the
# last administration date only. Flag for QC if exposure gaps exist.
derive_vars_merged(
dataset_add = ex_dtm,
by_vars = exprs(STUDYID, USUBJID),
new_vars = exprs(TRTEDTM = EXENDTM, TRTETMF = EXENTMF),
order = exprs(EXENDTM),
mode = "last",
filter_add = (EXDOSE > 0 | EXTRT == "PLACEBO") & !is.na(EXENDTM)
) |>
mutate(
TRTSDT = as.Date(TRTSDTM),
TRTEDT = as.Date(TRTEDTM)
)
Step 4 — Planned and actual treatment (TRT01P, TRT01PN, TRT01A, TRT01AN)
Use derive_vars_merged_lookup() with a treatment lookup tibble — this is the
idiomatic admiral approach for controlled terminology mapping and is preferred
over case_when() or mutate() for treatment arm coding.
TRT01P/TRT01PN: from DM.ARMCD (planned). TRT01A/TRT01AN: from DM.ACTARMCD (actual). These are distinct — derive independently.
# REVIEW: Confirm ARMCD values, treatment labels, and numeric codes against
# the randomisation schedule and ADaM spec before use.
arm_lookup <- tibble::tribble(
~ARMCD, ~TRT01P, ~TRT01PN,
"Pbo", "Placebo", 1L,
"Xan_Lo", "Xanomeline Low Dose", 2L,
"Xan_Hi", "Xanomeline High Dose", 3L
# Screen failure subjects (Scrnfail) are not in the lookup — they receive NA
)
adsl <- adsl |>
derive_vars_merged_lookup(
dataset_add = arm_lookup,
by_vars = exprs(ARMCD),
new_vars = exprs(TRT01P, T
阅读完整来源 ↗