The CDISC Standards Family: CDASH, SDTM, ADaM, and Define-XML

What each CDISC standard is, the problem it solves, who maintains it, and how the pieces connect from data collection to a traceable submission

Pharma & Clinical

The orienting map of the CDISC standards. Learn what each foundational standard is and how they nest: CDASH for collection, SDTM for tabulation, ADaM for analysis, Define-XML for the metadata and traceability that wrap them, plus Controlled Terminology (the NCI/EVS-maintained shared vocabulary) and the cSDRG/ADRG reviewer’s guides. Includes the standard-vs-implementation distinction and where the Implementation Guide (IG) fits.

Published

June 30, 2026

Modified

July 7, 2026

TipKey takeaways
  • CDISC (the Clinical Data Interchange Standards Consortium) is the standards organization; its foundational standards are the rules that make every trial’s data follow the same shape — CDASH (collection), SDTM (tabulation), ADaM (analysis), with Controlled Terminology as the shared vocabulary and Define-XML as the metadata that ties it all together.
  • They form a chain: CDASH standardizes collection → SDTM tabulates the collected data into domains → ADaM derives analysis-ready datasets → Define-XML describes the SDTM and ADaM datasets so a reviewer can trace every value back to its source.
  • A standard is not your dataset. Each standard ships a model plus an Implementation Guide (IG) — the rules and templates. Your study’s actual datasets are the implementation; conforming to the IG is what makes them CDISC-compliant.
  • Controlled Terminology (codelists and valid values) is maintained by the CDISC Terminology Team with NCI/EVS and published quarterly — it supplies the standardized coded values SDTM and ADaM require.
  • The reviewer’s guides — cSDRG (for SDTM) and ADRG (for ADaM) — are PHUSE-templated human-readable companions to the datasets, not CDISC standards themselves.

Introduction

A reviewer at the FDA opens your submission and finds a folder of datasets, a define.xml, and a short PDF guide. They navigate all of it — which dataset holds adverse events, what each variable means, how a derived value was computed — without ever calling you. They can do that because every piece follows a structure they have seen a hundred times before.

That shared structure is CDISC. The lesson on the clinical-trial data flow followed a data point through the five-stage flow (CRF → SDTM → ADaM → TLF → submission); the regulatory context explained why regulators make CDISC mandatory; the estimands framework covered how the clinical question is defined. This lesson is the standards themselves — what each one is, the problem it solves, who maintains it, and how they connect. It is the orienting map for the hands-on SDTM and ADaM series that follow.

No code, no setup — this is a concept map. One relationship diagram and two reference tables are all you need to carry into the hands-on series that follow.

A standard is not your dataset

The single most useful distinction up front: a CDISC standard is a set of rules, not a file. Each foundational standard is published as two things working together:

  • A model — the abstract structure (what classes of data exist, what a variable is, the core rules).
  • An Implementation Guide (IG) — the concrete instructions: the specific domains, variables, business rules, and worked examples you actually build to. SDTM has the SDTMIG, ADaM has the ADaMIG, CDASH has the CDASHIG. These foundational standards are catalogued on CDISC’s Foundational Standards page.

There is a one-to-one relationship between a version of a model and a version of its IG, and each IG names the model version it was written against. So when someone says “we’re on SDTMIG v3.4,” they are naming the rulebook their datasets follow.

Your study’s actual dm.xpt, adsl.xpt, and the rest are the implementation — the datasets built to those rules. “CDISC-conformant” means your implementation follows the IG. Keeping the two apart prevents a common beginner confusion: SDTM is not a file you have, it is a standard your file conforms to.

How the standards fit together

The foundational standards are not a flat list — they form a directed chain with a wrapper. Read the diagram as the journey of one trial’s data:

flowchart TD
  CDASH["CDASH<br/>data collection"] --> SDTM["SDTM<br/>tabulation domains"]
  SDTM --> ADaM["ADaM<br/>analysis-ready datasets"]
  CT["Controlled Terminology<br/>shared vocabulary"] -. "coded values" .-> SDTM
  CT -. "coded values" .-> ADaM
  SDTM -. "described by" .-> DX["Define-XML<br/>dataset metadata"]
  ADaM -. "described by" .-> DX
Figure 1

The solid arrows are the data pipeline: CDASH shapes how data is collected, SDTM tabulates it, ADaM derives what gets analyzed. The dashed arrows are the connective tissue: Controlled Terminology feeds standardized values into both SDTM and ADaM, and Define-XML describes both so the whole thing is navigable and traceable. Get this picture and each standard below is just one labeled box.

The standards, one by one

CDASH — the collection standard

CDASH (Clinical Data Acquisition Standards Harmonization) standardizes how data is collected on the case report form. It defines basic standards for the organization, naming, and description of the variables you capture, so collection looks the same across studies and sponsors.

It ships as the CDASH Model, the CDASHIG, and a metadata table. Crucially, the CDASH Model is structured to align with SDTM: it gives naming conventions and metadata that map collected fields cleanly to their SDTMIG variables. That alignment is the point — collecting to CDASH gives clear traceability of the collected data into SDTM, with no guesswork at the mapping step. CRFs are designed to CDASH by the data-management team.

SDTM — the tabulation standard

SDTM (Study Data Tabulation Model) reshapes the collected data into a fixed, regulator-expected structure. It organizes data into domains — standardized tables identified by a two-letter code (DM demographics, AE adverse events, VS vital signs, LB labs, EX exposure, and many more). The SDTMIG supplies the specific domain models, assumptions, business rules, and examples you build to.

SDTM is tabulation: a faithful, standardized record of what was observed, mostly one row per observation. It is deliberately not shaped for statistics — its job is to make any study’s data findable in the same place every time. It is one of the two dataset packages a sponsor submits, and is typically produced by clinical or statistical programmers.

ADaM — the analysis standard

ADaM (Analysis Data Model) takes the standardized SDTM data and turns it into analysis-ready datasets. CDISC defines ADaM as the standard that supports “efficient generation, replication, and review of clinical trial statistical analyses” and “traceability between analysis results, analysis data, and data represented in the Study Data Tabulation Model (SDTM).” ADaM is built with SDTM as its source and Define-XML for its metadata — it sits squarely inside the CDISC end-to-end flow.

Where SDTM tabulates, ADaM derives: treatment variables, population flags, baseline, change-from-baseline, visit windows. It defines a small set of dataset structures — ADSL (one row per subject, the subject-level backbone), BDS (Basic Data Structure, for measurements over time), and OCCDS (Occurrence Data Structure, for events such as adverse events). The ADaMIG spells out the rules. Traceability is ADaM’s defining property — every analysis value must walk back to its SDTM source — and it is the reason ADaM exists as a separate stage rather than analyzing SDTM directly. You run statistics on ADaM, not SDTM.

Controlled Terminology — the shared vocabulary

A standardized structure is only half the battle; the values must be standardized too. CDISC Controlled Terminology is the set of codelists and valid values used with data items in CDISC-defined datasets — the agreed vocabulary for things like sex, units, severity, and lab test codes. It is what makes “this AE was severe” mean the same coded thing in every study.

It is maintained by the CDISC Terminology Team in collaboration with the National Cancer Institute’s Enterprise Vocabulary Services (NCI/EVS), supports all the foundational standards (SDTM, CDASH, ADaM, and SEND — the Standard for Exchange of Nonclinical Data), and is published quarterly. Because it supplies the values FDA and PMDA (Japan’s Pharmaceuticals and Medical Devices Agency) require in conformant datasets, using the current applicable terminology version is part of conformance — not an afterthought.

Define-XML — the metadata that ties it together

Define-XML is a data-exchange standard that transmits metadata describing the structure of your datasets — variables, codelists, value-level metadata, and how each derived value was computed. It provides this metadata for the SDTM (and SEND) tabulation datasets and the ADaM analysis datasets.

Define-XML is the machine-readable backbone of traceability: it is how a reviewer navigates from a result on a table back to the dataset, the variable, and the derivation behind it — without opening your code. It is required by both the FDA and the Japanese regulatory authorities for the datasets in an electronic submission, which is why “the define.xml” is mentioned in the same breath as the datasets it describes. (The same connect-the-metadata idea, pushed to full automation across the study lifecycle, is the goal of the CDISC 360i initiative — standards-driven, metadata-driven automation from design through submission.)

The reviewer’s guides: cSDRG and ADRG

A define.xml is machine-readable; reviewers also get a human-readable companion that narrates the context a metadata file cannot. Two of them travel with the data:

  • The cSDRG (Clinical Study Data Reviewer’s Guide) — for the SDTM package.
  • The ADRG (Analysis Data Reviewer’s Guide) — for the ADaM package.

Each explains study-specific decisions, conformance findings, and anything a reviewer needs to read the data correctly. They are PHUSE-templated community deliverables, not CDISC standards themselves — a distinction worth keeping straight. The cSDRG template has been reviewed by the FDA and published in the Federal Register; the FDA prefers its use but does not strictly require the template, and an ADRG is recommended as part of a standards-compliant analysis submission.

The CDISC standards at a glance

One reference table for the whole family — what it is, where it acts, and who maintains it:

Standard What it is Stage Maintained / produced by
CDASH Rules for collecting data on the CRF, aligned to SDTM Collection CDISC
SDTM Model for tabulating collected data into fixed domains Tabulation CDISC
Controlled Terminology Codelists & valid values (the shared vocabulary) SDTM & ADaM CDISC Terminology Team + NCI/EVS
ADaM Model for analysis-ready, derived, traceable datasets Analysis CDISC
Define-XML Machine-readable metadata describing the datasets Wraps SDTM & ADaM CDISC (data-exchange standard)
cSDRG / ADRG Human-readable reviewer’s guides for SDTM / ADaM Submission PHUSE templates (sponsor-authored)

The first four standards are accessible in machine-readable form through the CDISC Library, the authoritative metadata source the pharmaverse tools read from.

The chain in one view

The relationships above are easy to state as a small table. Here is the chain as input → output links:

Step Standard Stage Takes in Produces
1 CDASH Collection CRF / study design standardized data collection
2 SDTM Tabulation collected CRF data tabulation domains (DM, AE, VS, …)
3 ADaM Analysis SDTM domains analysis-ready datasets (ADSL, ADAE, …)
4 Define-XML Metadata SDTM + ADaM datasets machine-readable dataset metadata

Read down the Takes in and Produces columns and the pipeline is explicit: each standard’s output is the next one’s input, and Define-XML takes the SDTM and ADaM datasets as its input to produce the metadata that describes them. Controlled Terminology is not a step in this chain — it is the vocabulary that feeds the values inside the SDTM and ADaM outputs, which is why it sits on dashed arrows in the diagram rather than in this ordered table.

Where this fits

These foundations lessons give you the frame; the rest of the pillar builds on it. Each foundational standard named here becomes a hands-on series — SDTM programming (building DM, AE, VS, EX domains and applying Controlled Terminology), ADaM with admiral (ADSL, then BDS and OCCDS datasets), and the submission-packaging work that produces Define-XML and the reviewer’s guides. The next lesson in this series closes the foundations with the industry’s SAS-to-R shift — why an open-source R pipeline is now a first-class way to implement every standard you just met. Where a hands-on series exists, the pillar hub links to it.

🟢 With an AI agent

Ask Prova “what is the difference between SDTM and ADaM, and where does Define-XML fit?” — it answers grounded in this pillar’s lessons, with runnable R you can try on example pharmaverse data. The runtime is the judge. Ask Prova →

Common issues

These are the conceptual slips that trip up people new to the CDISC standards.

Treating SDTM or ADaM as a file rather than a standard. “SDTM” is a model plus its IG — the rules. Your dm.xpt is an implementation of those rules. When you say a dataset “is SDTM,” you mean it conforms to the SDTMIG. The same goes for ADaM. Mixing up the rulebook and the file is the root of most early confusion.

Confusing CDASH with SDTM. Both are domain-shaped and look similar, but they act at different stages: CDASH standardizes collection (the CRF), SDTM standardizes tabulation (the submission dataset). CDASH is built to map cleanly into SDTM — that alignment is its purpose — but they are not the same standard, and you collect to one and submit the other.

Thinking the reviewer’s guides are CDISC standards. The cSDRG and ADRG are PHUSE-templated documents that accompany the datasets; they are not part of the CDISC foundational standards. They narrate study-specific context a define.xml cannot. Treat them as required submission companions, not as a CDISC standard you “conform to.”

Frequently asked questions

CDISC is the Clinical Data Interchange Standards Consortium. Its foundational standards are CDASH (data collection), SDTM (tabulation), and ADaM (analysis), supported by Controlled Terminology (the shared vocabulary) and Define-XML (the dataset metadata). Together they make every trial’s data follow the same structure so regulators can review submissions consistently.

SDTM standardizes collected data into fixed domains for tabulation — a faithful record of what was observed, mostly one row per observation. ADaM derives analysis-ready datasets from SDTM — with derived variables (flags, baseline, change-from-baseline) and full traceability back to SDTM. You build SDTM to organize the data and ADaM to analyze it; you run statistics on ADaM, not SDTM.

A standard is the rulebook — a model plus its Implementation Guide (IG) (the SDTMIG, the ADaMIG, the CDASHIG). An implementation is your study’s actual datasets, built to those rules. “CDISC-conformant” means the implementation follows the IG. There is a one-to-one relationship between a model version and its IG version.

Controlled Terminology is the set of codelists and valid values used with data items in CDISC datasets — the agreed coded vocabulary (sex, units, severity, lab test codes, and more). It is maintained by the CDISC Terminology Team in collaboration with NCI/EVS (the National Cancer Institute’s Enterprise Vocabulary Services) and published quarterly, and supplies the values FDA and PMDA require in conformant datasets.

Define-XML is a machine-readable metadata standard that describes the structure of your SDTM and ADaM datasets — variables, codelists, value-level metadata, and how derived values were computed. It is the backbone of traceability: a reviewer uses it to navigate from a result back to its source without opening your code. It is required by the FDA and PMDA for every study in each electronic submission.

They are the reviewer’s guides that travel with the datasets: the cSDRG (Clinical Study Data Reviewer’s Guide) for the SDTM package and the ADRG (Analysis Data Reviewer’s Guide) for the ADaM package. They are human-readable companions to the define.xml, narrating study-specific decisions and conformance findings. They use PHUSE templates and are not CDISC standards themselves.

Test your understanding

For each description below, name the CDISC standard (or document) it refers to:

  1. “The rules for how a variable is named and collected on the case report form, aligned so the data maps cleanly into the tabulation datasets.”
  2. “The model that organizes observed data into two-letter domains like DM, AE, and VS — a faithful, mostly one-row-per-observation record of what happened.”
  3. “The codelists and valid values, maintained with NCI/EVS and published quarterly, that supply the standardized coded values the datasets must use.”
  4. “The machine-readable metadata file, required by the FDA and PMDA, that lets a reviewer trace a result back to its source dataset and derivation.”

Walk the chain: collect → tabulate → analyze → describe. Then remember the one standard that is a vocabulary feeding values into the datasets rather than a step in the pipeline.

  1. CDASH — the data-collection standard, structured to map cleanly into SDTM.
  2. SDTM — the tabulation model, organizing data into domains (DM, AE, VS, …).
  3. Controlled Terminology — the shared vocabulary, maintained by the CDISC Terminology Team with NCI/EVS and published quarterly.
  4. Define-XML — the machine-readable dataset metadata, required for FDA and PMDA submissions. (ADaM, the analysis standard, would be the answer to “the derived, analysis-ready datasets that trace back to SDTM.”)

A. “SDTM” is the name of a specific file (sdtm.xpt) you submit to the FDA. B. A CDISC standard is a model plus an Implementation Guide; your study’s datasets are the implementation that conforms to it. C. The cSDRG and ADRG are CDISC foundational standards alongside SDTM and ADaM.

B. A standard is the rulebook (a model + its IG, e.g. the SDTMIG); your actual datasets are the implementation built to it. A is wrong — there is no single sdtm.xpt; SDTM is the standard your domain datasets (dm.xpt, ae.xpt, …) conform to. C is wrong — the cSDRG and ADRG are PHUSE-templated reviewer’s guides, not CDISC standards.

References

Primary, authoritative sources for the standards described above:

Conclusion

The CDISC standards are a chain with a wrapper. CDASH standardizes how data is collected, SDTM tabulates it into fixed domains, ADaM derives the analysis-ready datasets, and Controlled Terminology supplies the shared coded values that flow through both — while Define-XML describes the SDTM and ADaM datasets so a reviewer can trace any value back to its source, with the cSDRG and ADRG narrating the rest. Keep two ideas in mind and the whole family stays clear: each standard’s output is the next one’s input, and a standard is the rulebook your study’s datasets implement. With this map in hand, the hands-on SDTM and ADaM series ahead are just the work of building real datasets to rules you already understand.

Reuse

Citation

BibTeX citation:
@online{2026,
  author = {},
  title = {The {CDISC} {Standards} {Family:} {CDASH,} {SDTM,} {ADaM,}
    and {Define-XML}},
  date = {2026-06-30},
  url = {https://www.datanovia.com/learn/pharma-clinical/01-regulatory-cdisc-foundations/04-cdisc-standards},
  langid = {en}
}
For attribution, please cite this work as:
“The CDISC Standards Family: CDASH, SDTM, ADaM, and Define-XML.” 2026. June 30. https://www.datanovia.com/learn/pharma-clinical/01-regulatory-cdisc-foundations/04-cdisc-standards.