What is USDM?

USDM stands for Unified Study Definitions Model. It is the CDISC standard that describes a clinical study design (its arms, visits, activities, objectives, eligibility criteria and interventions) as structured data that software can read, instead of prose in a protocol document that people have to read and re-enter.

The problem it solves

A clinical trial protocol is a document. It is long, it is written for clinicians and regulators, and the facts in it are spread across paragraphs, tables and footnotes. Every system that takes part in running the study needs some of those facts: the data collection system needs the visits and assessments, the trial management system needs the schedule, the registry needs the design and the outcome measures, the statistical team needs the endpoints.

None of those systems can read a document, so people do it for them. The same schedule of activities is read and typed into several systems by several teams, each interpreting it slightly differently, and all of it is repeated when the protocol is amended. USDM replaces the retyping with a single description of the study that every system can read.

Who maintains it

CDISC, the organisation behind SDTM, CDASH and ADaM, develops and publishes USDM. It does so in collaboration with TransCelerate BioPharma, a non-profit consortium of pharmaceutical companies, as the reference architecture for TransCelerate’s Digital Data Flow (DDF) initiative. TransCelerate also maintains an open-source reference implementation, the Study Definitions Repository.

The standard is developed in the open. The model, the API specification, the controlled terminology and the implementation guide are all in the cdisc-org/DDF-RA repository on GitHub, and anyone can read them.

The five parts of the standard

USDM is not one file. A release is made of five deliverables, and it helps to know which one answers which kind of question.

  1. The logical model

    A UML class diagram of 86 classes and the relationships between them. This is the standard itself; the other four parts exist to make it usable.

  2. The API specification

    An OpenAPI definition for sending and receiving a whole study as JSON. It fixes the file format most people mean when they say a USDM file.

  3. Controlled terminology

    The codelists behind coded attributes such as arm type, timing type and objective level, published through NCI EVS like the rest of CDISC terminology.

  4. The implementation guide

    The explanation: how each part of the model is meant to be used, with worked examples and mappings to other standards.

  5. Conformance rules

    A set of machine-checkable rules that say what a correct USDM file looks like beyond what the JSON schema can express.

What USDM covers

Version 4.0 defines 86 classes carrying 833 attributes. They fall into ten areas, each linked here to its part of the object model reference.

Study, versions and amendments
The root of every USDM file: the study, each version of its plan, its titles and identifiers, and the amendments that moved it from one version to the next.
Organizations, roles and sites
Who is involved and in what capacity: sponsors, registries, CROs, sites, named people and who is masked.
Study design
The shape of the trial: arms, epochs, the cells where they meet, the elements inside each cell, and the indication under study.
Schedule and timing
The schedule of activities as data: visits, activities, the timeline that orders them, and the timings and windows between them.
Objectives, endpoints and estimands
What the study sets out to learn and how it will be measured, including the ICH E9(R1) estimand attributes.
Population and eligibility
Who can take part: the planned population, cohorts within it, and each inclusion and exclusion criterion.
Interventions and products
What participants receive: the intervention, how it is administered, and the product, ingredients, strengths and devices behind it.
Biomedical concepts
The data layer: what is actually collected when an activity is performed, down to the property and its allowed responses.
Documents and narrative
The protocol as a document: its versions, its sections of prose, and the templated text and abbreviations that prose uses.
Codes, quantities and extensions
The building blocks every other class uses: coded values, numbers with units, ranges, and the mechanism for carrying data the standard does not define.

What a USDM file looks like

When people say a USDM file they mean JSON in the shape the API specification defines. The outermost object is a wrapper: it holds one study and says which version of the model the file follows. Everything else is inside the study.

A USDM file, with the contents collapsed
{
  "usdmVersion": "4.0.0",
  "systemName": "Concordare Learn USDM",
  "systemVersion": "1.0",
  "study": {
    "id": "Study_1",
    "name": "EX101-201",
    "label": "EX-101 in Stage 1 Hypertension",
    "description": "An illustrative study written for the Concordare Learn USDM guide. The sponsor, product and identifiers are fictional.",
    "versions": [ … 1 StudyVersion ],
    "documentedBy": [ … 1 StudyDefinitionDocument ],
    "instanceType": "Study"
  }
}

Three conventions run through every file:

  • Every object has an id that is unique within the file and an instanceType that names its class.
  • An object that belongs to one parent is nested inside it. An object used in several places is written once and pointed to by id, in an attribute whose name ends in Id or Ids.
  • Coded values are full objects, not bare strings: a code, the code system, its version and the human-readable decode.

The examples page walks through a complete small study one piece at a time.

How it got to version 4.0

VersionReleasedWhat it added
USDM 1.0August 2022The base model for a structured study design. Simple designs only, and no implementation guide.
USDM 2.0June 2023Study timing and timelines for complex visit structures, biomedical concepts to support EDC automation, and enough protocol content to populate a protocol template. The first version with an implementation guide.
USDM 3.0April 2024Coverage of the draft ICH M11 protocol template, more of the SDTM trial design datasets, elements for trial registries, and cohorts for complex designs.
USDM 4.0June 2025Separate interventional and observational study designs, device studies, richer amendments, estimands, abbreviations, roles and interventions, aligned to the March 2025 review draft of ICH M11. Published with conformance rule specifications.

Version 4.0 was released on 3 June 2025 and is the current published version. It was aligned to the review draft of ICH M11 that was available at the time. ICH adopted the final M11 guideline in November 2025, and the USDM 4.0 implementation guide says a 4.1 release is planned to align with it.

What it is used for

Passing a design between tools

A protocol authoring tool writes the study as USDM and a downstream system reads it, with no one retyping the schedule of activities in between. This is the use the standard was created for.

Building data collection

Biomedical concepts attached to activities say what data each one collects. An EDC system can use them to pick matching forms from its library, or to build new ones.

Generating SDTM trial design datasets

Arms, epochs, elements, cells, visits and criteria in USDM line up with the SDTM trial design domains (TA, TE, TV, TI and TS), so those datasets can be generated instead of written by hand.

Filling trial registries

Much of a ClinicalTrials.gov record can be filled from a USDM study: identification, sponsor, design, arms and interventions, outcome measures and eligibility. Part of an EU CTIS application can be filled the same way.

Producing the protocol document

Because USDM holds the protocol's text as well as its structure, a study in USDM can be rendered into a protocol template such as ICH M11 or the TransCelerate Common Protocol Template.

What USDM is not

  • Not a regulatory requirement. No regulator currently requires a submission in USDM. ICH M11, which defines the harmonised protocol and its exchange format, is the regulatory guideline; USDM is the industry data model aligned to it.
  • Not the study data. USDM describes the plan. Data collected from participants is described by other standards, such as CDASH and SDTM.
  • Not an authoring method. The standard says what a study definition contains, not how a sponsor arrives at it. Some write the design in a tool that produces USDM; others convert an existing protocol.
  • Not a site’s operating plan. USDM holds the study as the sponsor designed it. How a particular site will carry it out is outside its scope, which is the subject of USDM at the research site.

Keep going