SDTM, ADaM and SEND Data Viewer
Drug companies send FDA their trial data as SAS files in standard layouts called SDTM, ADaM and SEND, with a define.xml that explains each column. Open them in your browser and see what a reviewer would flag.
Open the .xpt datasets and define.xml from a study data package. Browse every record, read the define, and see where the package breaks FDA's conformance rules.
- Open: Folder or files
- Processing: in your browser
- Checks: 18 rules
Open SDTM, ADaM or SEND datasets
Drop a folder of .xpt files with its define.xml, or choose them.
Nothing is uploaded: files are read in this browser tab.
Browse every dataset
Records with their variable labels, filtered by any value, without SAS or a viewer license.
Read define.xml
Datasets, variables, origins and codelists, laid out like the stylesheet but searchable.
Check what FDA checks first
Name and label limits, lengths, identifiers, --SEQ, and whether the data match define.xml.
Stays on your device
Study data never leave this tab. Close it or clear the package to release them.
About this data
What it reads
FDA asks for study data as SAS transport files, version 5: one dataset per .xpt file, with a define.xml that describes every dataset, variable and codelist. SDTM holds the collected clinical data, ADaM the analysis datasets built from it, and SEND the nonclinical (animal) study data.
The transport format dates from IBM mainframes: numbers are stored as IBM floating point, not the format your computer uses, and SAS dates as days since 1960. The viewer converts both. Define-XML 2.0 and 2.1 are read for datasets, variables, origins and codelists.
The checks
Each check names its rule and links to it. The FDA rules come from the Study Data Technical Conformance Guide (version 6.3, October 2026); the identifier and --SEQ rules from the CDISC SDTM and ADaM implementation guides; the define.xml comparisons from Define-XML.
- One version 5 dataset per file, named like the file. FDA reads SAS transport version 5 only. Each .xpt should hold one dataset whose name matches the file name (ae in ae.xpt).
- Files over 5 GB must be split. FDA asks for datasets larger than 5 GB to be split into files no larger than 5 GB, placed in a 'split' folder alongside the full dataset.
- Dataset names up to 8 letters and digits. Version 5 transport files allow 8-character names that start with a letter. Longer names are truncated or rejected.
- Variable names up to 8 letters and digits. Variable names longer than 8 characters, or with characters other than letters, digits and underscores, do not survive version 5 transport.
- Labels up to 40 characters. Dataset and variable labels longer than 40 characters are cut off in version 5 transport files.
- Every dataset and variable has a label. Reviewers see the label, not the code. FDA asks that the dataset label clearly describe its contents.
- Text values up to 200 characters. Version 5 transport stores at most 200 bytes per character variable. Longer text goes in supplemental qualifiers (SUPP--) or split variables.
- Text in plain ASCII. Values outside printable ASCII can display wrongly or be truncated in FDA's tools. Accented letters and smart quotes are the usual cause.
- Required identifiers present. Subject-level SDTM and SEND datasets carry STUDYID, DOMAIN and USUBJID (SEND may use POOLID). ADaM datasets carry STUDYID and USUBJID. Supplemental and relationship datasets use RDOMAIN.
- DOMAIN matches the dataset. The DOMAIN value should be the dataset's two-letter code (split datasets such as LBCH keep DOMAIN = LB).
- --SEQ unique within each subject. The sequence number identifies one record per subject in a domain. Duplicates break links from SUPP-- and RELREC and are one of the reasons FDA cites for unusable data.
- Datasets match define.xml. Every submitted dataset should be described in define.xml, and every dataset it describes should be in the package.
- Variables match define.xml. A variable in the data but not in define.xml has no documented origin or meaning; one in define.xml but not in the data is a broken promise to the reviewer.
- Types agree with define.xml. A text type in define.xml should be a character variable in the data, and integer or float should be numeric.
- Lengths agree with define.xml. For character variables, the length in define.xml should be the length stored in the transport file.
- Labels agree with define.xml. The label in the transport file and the description in define.xml should be the same text.
- Values are in the define.xml codelist. When define.xml attaches a codelist to a variable, every value in the data should be one of its terms. External dictionaries such as MedDRA are not checked.
- Text lengths no longer than the data. FDA asks that each text column's length be the longest value of that variable across the study's datasets (within the dataset for SUPP--). Padding inflates file size: USUBJID with 18-character values should have length 18, not 200.
Frequently asked questions
- Are my datasets uploaded?
- No. The page reads the files you choose inside your browser tab. Nothing is sent to Arca or anyone else, and closing the tab or pressing Clear releases them.
- Which files can it open?
- SAS transport files (.xpt) in version 5, the format FDA requires, and version 8, plus define.xml in Define-XML 2.0 and 2.1 (and the label and origin fields of 1.0). Choose a folder to open a whole package; a folder with both tabulation and analysis data opens as separate packages.
- Does this replace Pinnacle 21 or FDA's validator?
- No. It runs the checks reviewers hit first and explains each one, so you can triage a package from a CRO before running the full validator. Controlled terminology versions, value-level metadata and cross-domain rules are not checked.
- Why are dates shown as calendar dates?
- SAS stores dates as days since 1 January 1960. When a numeric variable carries a date or datetime format, such as DATE9. or E8601DT., the viewer shows the date that SAS would print.