What Is the USDM Study Schema? A Complete Guide to CDISC's Unified Study Definitions Model

Introduction

A phase III protocol amendment took 260 days on average to move from identified need to final sign-off in 2024, and investigative sites operated on mismatched protocol versions for 215 days at a stretch [7]. Those numbers come from Tufts CSDD's most recent benchmarking study, and they explain why a data model with an unglamorous name, the Unified Study Definitions Model, has become one of the more closely watched developments in clinical trial technology. USDM is CDISC's answer to a structural problem: protocols are still written as documents, but nearly every downstream system, from EDC to RTSM to SDTM datasets, needs that same information as structured data. USDM defines what that structured version looks like.

This guide walks through what USDM actually contains, how it relates to CDISC's older standards and to TransCelerate's Digital Data Flow initiative, where it stands with regulators, and what sponsors and CROs need to know before treating it as an operational requirement rather than a technical curiosity.

Why This Topic Matters in Clinical Trials

Clinical protocols are authored once and then re-keyed, by hand, into a dozen downstream systems. A site coordinator reads the schedule of assessments and builds it again in the EDC. A statistician reads the arms and visit windows and rebuilds them in the SDTM Trial Design domains. A vendor reads the randomization scheme and rebuilds it in the IRT. Each re-entry is a chance to introduce a discrepancy between what the protocol says and what the system does, and each one adds days to trial startup.

TransCelerate, the industry consortium that partners with CDISC on this work, has quantified part of that gap directly: it identifies an average four-month lag between protocol approval and study startup, driven in large part by the absence of a common, machine-readable protocol format that systems can consume without manual translation [2]. USDM exists to close that gap by giving every downstream system the same structured source of truth, expressed once, at the point the protocol is designed.

Current Evidence and the Research Record

USDM is not a single document. CDISC describes it as a logical data model built from UML class diagrams and an accompanying data dictionary, paired with a RESTful API specification that supports both JSON and XML, protocol-specific controlled terminology, an implementation guide for upstream and downstream consumers, and a set of machine-executable conformance rules written in YAML [1]. Four full versions have shipped so far. Version 1.0 arrived in August 2022 as an initial reference architecture. Version 2.0 followed in June 2023 with biomedical concepts and time points added. Version 3.0, released in April 2024, aligned the model with the ICH M11 protocol template and extended coverage to more complex study designs, with its conformance rules published that December. Version 4.0, the current release, shipped in June 2025 and rounds out the API specification, implementation guide, and conformance rule set [1].

The reference implementation of that model is TransCelerate's Study Definitions Repository, built as a cloud-based, technology-agnostic, open-source system under an Apache 2.0 license [1]. CDISC also maintains an open-source Python package on GitHub, licensed under MIT, that provides the USDM model classes and an Excel import tool capable of converting a study definition spreadsheet into the equivalent API-compliant JSON [4]. That package carries an honest caveat in its own documentation: it was not originally built for public use, and only informal testing has been performed on it to date [4]. That is worth noting for any team evaluating build-versus-buy decisions around USDM tooling.

Development has proceeded in phases rather than a single release cycle. Phase 1, running from July 2021 to July 2022, established the reference architecture. Phase 2, through July 2023, added biomedical concepts and electronic templates. Phase 3, closing in April 2024, focused on ICH M11 alignment and complex designs. Phase 4, which closed in May 2025, extended the model to observational studies and medical devices. A fifth phase is under discussion with scope not yet finalized [1]. The model is governed by the USDM Governance Group, whose members serve one-year terms and meet monthly, plus additional sessions tied to each development cycle [1].

CDISC has also started publishing implementation handbooks aimed at operational teams rather than standards developers. The first of these, released in June 2026, documents how USDM metadata can automatically populate five SDTM Trial Design domains: Trial Arms, Trial Elements, Trial Visits, Trial Inclusion and Exclusion Criteria, and Trial Summary. The handbook estimates that more than 95% of the information those domains require already exists in the study design captured in USDM, and that automating the mapping can cut a task that currently takes a programmer 8 to 32 hours and two to three weeks of turnaround down to one or two hours of final review [3]. That figure is CDISC's own reported estimate rather than an independently validated benchmark, and it is one sponsors evaluating USDM tooling should ask vendors to reproduce against their own protocols rather than accept at face value.

A quick way to keep these related standards straight:

Independent industry commentary has been more measured than CDISC's own materials. TransCelerate's own framing of the initiative points at the same automation potential, describing it as removing "non-value-added activities" from trial startup [2], and one sponsored industry commentary, published by Clinical Trials Arena in July 2025, credits USDM more specifically with the potential for automated EDC configuration, more consistent RTSM dispensing schedules, and fewer manual transcription errors [8]. That same commentary also argues the automation only holds if the protocol is genuinely finalized before configuration starts, and that delaying setup until then could extend lead times rather than shorten them, though this is one commentator's assessment rather than a measured outcome [8]. A separate technical review of M11 implementation makes a related, narrower point from the systems side: it argues that a protocol can conform fully to the M11 structure while the receiving EDC or IRT still lacks the connector, field mapping, or version-handling logic needed to actually use the data, meaning structural conformance and functional automation are not the same thing [13]. Protocol amendments compound the problem regardless of which framing proves right. As the Tufts data shows, amendments now touch 76% of phase I through IV trials and can force rework across every connected system [7]. The same amendment study found that the mean number of amendments per protocol rose 60%, from 2.1 to 3.3, over the period studied, and that 77% of those amendments were judged unavoidable rather than a result of poor planning [7]. A model built around a stable, structured protocol has to account for a protocol environment that is amending more often, not less.

Operational Impact for Sponsors, CROs, and Sites

For a data manager, USDM's practical value shows up first in the Trial Design domains. Instead of a programmer reading a protocol PDF and manually building Trial Arms, Trial Elements, and Trial Visits datasets, those structures are derived from the same USDM instance used to configure the EDC, which removes one of the more common sources of protocol-to-database mismatch [3].

For clinical operations teams, the bigger effect is on startup sequencing. When the schedule of activities exists as structured data rather than a table embedded in a Word document, it can populate site training materials, monitoring plans, and IRT visit windows from a single source rather than three separately maintained copies. TransCelerate frames this as the shift from document-centric to data-centric trial conduct, with the schedule of activities named as one of the more tractable places to start [2].

For sites, the amendment burden is where USDM's promise is most tested. A 215-day average window in which different sites run on different protocol versions is a coordination problem today, largely managed through spreadsheets and manual tracking [7]. A structured, versioned protocol definition does not eliminate that window, but it does make it possible to trace exactly which systems still reflect the prior version and which have been updated, something a document-based workflow struggles to do reliably.

Vendor readiness is the piece most sponsors underestimate. One published technical analysis counted 575 individual data-element entries in the ICH M11 technical specification, 288 of them required, 168 conditional, and 118 optional, each tied to controlled terminology for electronic exchange, and noted that the specification deliberately does not prescribe which file format, JSON, XML, or FHIR, a system must use to move that data [12]. USDM fills that gap for CDISC-aligned tools, but only where both the sending and receiving systems actually implement it. Before assuming an EDC or IRT vendor can consume a USDM-structured protocol, sponsors are better served asking three direct questions during vendor due diligence than assuming the answer: can the EDC build a casebook skeleton straight from a USDM study definition through its API, can the IRT or RTSM derive randomization strata and dispensing schedules from that same structured source, and which USDM version does the system validate against [12]. Public evidence of how consistently EDC and IRT vendors can answer those questions today is thin, which is itself a reason to ask them directly in the current procurement cycle rather than assume the capability exists. The answer worth trusting is a working demonstration against one of the sponsor's own protocols, ideally including a mid-study amendment, rather than a capability claimed in an RFP response [12].

Regulatory and Documentation Considerations

USDM's regulatory relevance runs through ICH M11, the harmonized protocol guideline, template, and technical specification for structured, machine-readable protocols. M11's scope is specific. The guideline itself defines its applicability to "interventional clinical trials of medicinal products" across all phases and therapeutic areas, with medicinal product defined broadly enough to include drug-device combination products when they are developed as a drug; protocol content requirements otherwise remain governed by ICH E6 and E8(R1) [11]. Observational studies and trials of standalone medical devices, meaning devices not developed as part of a drug product, sit outside that scope, even though USDM's own Phase 4 development separately extended the data model to cover those study types [1]. The ICH Assembly adopted the M11 package at Step 4, the point of full multi-region technical agreement, on November 19, 2025 [11]. The European Medicines Agency then brought the guideline into its own framework at Step 5, first publishing those documents on December 15, 2025, with the guideline taking effect in the EU on June 11, 2026 [5]. The FDA finalized its own implementing guidance on May 22, 2026, closing out draft versions that had been in circulation since December 2022 and June 2025 [6][10]. That final guidance still carries the disclaimer FDA attaches to essentially all of its guidance documents: it describes the agency's "current thinking," creates no legally enforceable obligations, and permits a sponsor to use an alternative approach that satisfies the underlying statute or regulation [10]. In practice, that means M11 is now the FDA's settled, current position on structured protocol format rather than a draft awaiting comment, even though FDA guidance remains procedurally non-binding by design regardless of whether it is labeled draft or final.

CDISC positions USDM as complementary to, not a replacement for, its longer-standing foundational standards. The model is designed to populate SDTM Trial Design domains automatically rather than to change what SDTM requires [3], and SDTM itself remains one of the data formats the FDA lists as currently supported for study data submitted to CDER and CBER, alongside SEND, ADaM, and Define-XML [9]. USDM sits upstream of that submission step: it structures the protocol design so the Trial Design domains can be generated rather than transcribed, but it does not change what those domains must contain once ICH M11 or FDA-specific format requirements apply to a given study.

AI and Automation Perspective

The realistic role for AI here is narrower than a lot of vendor language suggests. USDM gives automation something concrete to work against: a structured schema with defined entities and a conformance rule set that can be checked programmatically rather than reviewed by eye [1]. That is what makes tasks like Trial Design domain population plausible candidates for automation in the first place. It is not, on its own, a reason to expect protocol authoring itself to become fully automated.

Two limitations matter in practice. First, USDM's automation gains depend on the underlying protocol data being complete and correctly structured before configuration begins; a partially specified or informally amended protocol produces a partially reliable automated output, not a partially manual one [8][12]. Second, the model's conformance rules can confirm that a USDM instance is internally consistent, but they cannot confirm that the clinical judgment behind a schedule of activities or an eligibility criterion is sound. Human review of the underlying protocol content stays necessary regardless of how much of the downstream configuration gets automated. Vendors and sponsors piloting USDM-based automation should validate outputs against a known protocol before extending the approach to production studies, which is the same caution CDISC's own implementation handbook builds into its phased rollout guidance [3].

Piloting well comes down to a short list of concrete checks rather than a general readiness assessment: which USDM version a system validates against, whether it consumes the model through the published API rather than a custom export, whether it runs USDM's own machine-executable conformance rules before accepting a study definition, how it handles a mid-study amendment against an already-configured build, and whether it preserves an audit trail linking each downstream artifact back to the USDM instance it was generated from [1][4]. A pilot that skips any of these tends to surface the gap only after the protocol has already amended once.

How Kitsa Fits Into This Problem

The gap USDM addresses, protocol content that exists in one format but is needed as structured data across many systems, is the same gap KScribe was built to work inside of. When protocol content such as arms, visit schedules, and eligibility criteria is captured once as structured data rather than drafted independently in each document, that same source can carry through into the other regulatory documents a study needs, including the informed consent form, the investigator's brochure, the DSUR, and the CSR. KScribe applies that principle to document generation the way USDM applies it to system-to-system data exchange: the protocol's underlying data is the shared source, rather than each document being a separate drafting exercise. As USDM and ICH M11 adoption move from pilot to standard practice, having that protocol content structured correctly from the point of authoring, rather than retrofitted into structure afterward, becomes more valuable, not less.

Key Takeaways

• USDM is CDISC's logical data model, API specification, terminology set, and conformance rule set for representing clinical study design as structured, machine-readable data rather than narrative text [1].
• Four versions have shipped between August 2022 and June 2025, with development organized into phases run jointly with TransCelerate's Digital Data Flow initiative [1].
• TransCelerate's reference implementation, the Study Definitions Repository, is open source under Apache 2.0; CDISC's own Python tooling on GitHub is MIT-licensed and explicitly marked as not yet production-hardened [1][4].
• CDISC's first implementation handbook reports, based on its own analysis, that USDM can cut SDTM Trial Design domain programming from 8 to 32 hours down to 1 to 2 hours of review, since over 95% of the needed information already exists in the study design; this is CDISC's own figure, not an independently validated benchmark [3].
• The ICH Assembly adopted ICH M11 at Step 4 on November 19, 2025, the EU brought it to Step 5 on December 15, 2025 (effective June 11, 2026), and the FDA finalized its own implementing guidance on May 22, 2026, ending its 2022 draft status; FDA guidance stays procedurally non-binding whether draft or final [5][6][10][11].
• Protocol amendments now affect 76% of phase I through IV trials, up from 57% in the earlier benchmark, which is the operational reality any USDM implementation has to be designed around [7].
• One published technical analysis counted 575 data-element entries in the ICH M11 technical specification, but the specification leaves the exchange file format open; sponsors should confirm directly with EDC and IRT vendors whether their systems can ingest USDM-structured data, since public evidence of consistent vendor support is thin [12].
• USDM structures the protocol upstream of submission; it does not change the SDTM, ADaM, SEND, or Define-XML formats the FDA lists as currently supported for study data submission [9].

FAQ

References

1. [1]CDISC. "Digital Data Flow." CDISC. https://www.cdisc.org/ddf
2. [2]TransCelerate BioPharma Inc. "Digital Data Flow Initiative." TransCelerate BioPharma. https://www.transceleratebiopharmainc.com/initiatives/digital-data-flow/
3. [3]CDISC. "USDM Handbook 1: Leveraging USDM Metadata to Automatically Construct the SDTM Trial Design Model Domains," v1.0. CDISC, June 2026. https://www.cdisc.org/sites/default/files/2026-06/USDM-HB1%20v1.0_FINAL.pdf
4. [4]CDISC. "usdm" GitHub repository. CDISC-ORG. https://github.com/cdisc-org/usdm
5. [5]European Medicines Agency. "ICH M11 Guideline, Clinical Study Protocol Template and Technical Specifications." EMA, Step 5 documents first published December 15, 2025; guideline effective in the EU June 11, 2026. https://www.ema.europa.eu/en/ich-m11-guideline-clinical-study-protocol-template-technical-specifications-scientific-guideline
6. [6]U.S. Food and Drug Administration. "M11 Clinical Electronic Structured Harmonised Protocol (CeSHarP)" (final guidance). FDA, May 2026. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/m11-clinical-electronic-structured-harmonised-protocol
7. [7]Getz KA, et al. "New Benchmarks on Protocol Amendment Practices, Trends and their Impact on Clinical Trial Performance." Therapeutic Innovation & Regulatory Science, Vol. 58, No. 3, May 2024. https://link.springer.com/article/10.1007/s43441-024-00622-9
8. [8]Thompson N. "The Promise and Challenges of the Unified Study Definition Model (USDM)." Clinical Trials Arena, July 25, 2025. https://www.clinicaltrialsarena.com/sponsored/the-promise-and-challenges-of-the-unified-study-definition-model-usdm/
9. [9]U.S. Food and Drug Administration. "Study Data for Submission to CDER and CBER." FDA, Study Data Standards Resources. https://www.fda.gov/industry/study-data-standards-resources/study-data-submission-cder-and-cber
10. [10]U.S. Food and Drug Administration. "M11 Clinical Electronic Structured Harmonised Protocol (CeSHarP); International Council for Harmonisation; Guidance for Industry; Availability." Federal Register, 91 FR 30310, May 22, 2026. https://www.federalregister.gov/documents/2026/05/22/2026-10295/m11-clinical-electronic-structured-harmonised-protocol-cesharp-international-council-for
11. [11]International Council for Harmonisation. "ICH M11 Guideline on Clinical Electronic Structured Harmonised Protocol" (Step 4 final document). ICH, November 19, 2025. https://database.ich.org/sites/default/files/ICH_Step4_M11_Final_Guideline_2025_1119.pdf
12. [12]Sakara Digital. "ICH M11 in Force: What a Machine-Readable Protocol Changes." Sakara Digital, 2026. https://sakaradigital.com/blog/ich-m11-machine-readable-protocol-clinical-systems/
13. [13]BGA Software Consultancy. "ICH M11 and USDM v4: The Digital Protocol Reshapes Trials." BGA Software, 2026. https://bgasoft.com/blog/ich-m11-usdm-digital-protocol/