openstudyval

Open-source R validator for FDA study-data submissions (SDTM, SEND, and ADaM).

This package follows the same workflow as a commercial validator such as Pinnacle 21 Community: point it at a study folder, choose a standard, optionally pass define.xml and controlled terminology, and get a severity-ranked HTML/CSV report.

It is not Pinnacle 21. Rule text, engine, and report layout are original. It implements publicly described expectations from the FDA Study Data Technical Conformance Guide and the eCTD technical-rejection criteria for study data (1734, 1735, 1736, 1738). It does not copy Pinnacle 21 rule IDs, messages, or configuration files, and a clean report is not an FDA or PMDA certification.

Install

From this repository, once you have published it:

# install.packages("remotes")
# remotes::install_github("shekharchalwadigrepo-hub/openstudyval")
install.packages("openstudyval", repos = c('https://shekharchalwadigrepo-hub.r-universe.dev', 'https://cloud.r-project.org'))

After CRAN acceptance:

install.packages("openstudyval")

SAS transport files need the suggested package haven. define.xml content checks need xml2.

install.packages(c("haven", "xml2"))

Validate a study

library(openstudyval)

# Bundled example (issues are intentional)
report <- validate_submission(example_study("sdtm"), standard = "SDTM")
summary(report)

write_validation_report(report, "validation-report.html")

# A real folder of dm.xpt, ae.xpt, ts.xpt, define.xml, ...
report <- validate_submission(
  "m5/datasets/abc123/tabulations/sdtm",
  standard = "SDTM"
)

standard is "SDTM", "SEND", or "ADaM". CSV and RDS are read with base R. .xpt is read with haven.

What is checked

Layer	Examples
Technical rejection	Missing `ts.xpt` / SSTDTC (1734), missing DM or ADSL and define.xml (1736), STUDYID consistency (1738), STF presence note (1735)
XPORT / TCG	Dataset and variable names of 1–8 characters, labels ≤ 40, character values ≤ 200, files over 5 GB
SDTM / SEND	DM required, trial-design presence, DOMAIN matches dataset, USUBJID starts with STUDYID, USUBJID in DM, `--SEQ` uniqueness, ISO 8601 `--DTC`, core codelists, paired test code/name
ADaM	ADSL required, one row per USUBJID, subjects exist on ADSL, PARAM/PARAMCD one-to-one, population flags, TRTSDT ≤ TRTEDT
define.xml	Well-formed XML, ItemGroupDef names vs datasets (requires `xml2`)

MedDRA and WHODrug are licensed and are not shipped. Pass a study codelist:

ct <- data.frame(AESEV = c("MILD", "MODERATE", "SEVERE"))
validate_submission(path, standard = "SDTM", ct = ct)

Add rules without editing the package

Catalog columns: rule_id, publisher, standard, domains, severity, check, argument, message, guidance.

check is one of dataset_required, variable_required, codelist, iso8601, key_unique, one_to_one, nonmissing, regex.

extra <- data.frame(
  rule_id = "CUST_ARM", publisher = "sponsor", standard = "SDTM",
  domains = "DM", severity = "Error", check = "codelist",
  argument = "ARMCD=A|B|SCRNFAIL",
  message = "ARMCD is not in the study arm list.",
  guidance = "Protocol section 6",
  stringsAsFactors = FALSE
)
validate_submission(path, standard = "SDTM", rules = extra)

Domains accept AE|LB, AD*, or *.

Before you submit this package to CRAN

Replace Package Maintainer <you@example.com> in DESCRIPTION with your name and a working email.
Replace YOU/openstudyval in DESCRIPTION, README.md, and this repo URL.
Run R CMD check --as-cran and keep a clean check on at least one platform besides your own. GitHub Actions in .github/workflows/R-CMD-check.yaml does this.
Confirm the license (MIT) is what you want to ship.
Do not claim FDA endorsement or compatibility with a named Pinnacle 21 engine version.

Repository layout

R/                  validator, report writer, rule engine
inst/extdata/rules  core rule catalog
inst/extdata/example_sdtm  intentional-failure demo
inst/extdata/example_adam
tests/testthat
vignettes/validating.Rmd

License

MIT. See LICENSE.