USDM Schedule of Activities: How to Create a Machine-Readable Clinical Trial SoA

Introduction

TransCelerate BioPharma estimates that roughly 90% of clinical trial protocols still exist as unstructured documents, a figure the organization shared in its October 2025 update on the Digital Data Flow initiative [3]. The Schedule of Activities, the table that tells a site coordinator which procedures happen at which visit, sits at the center of that problem. It is usually a Word table with footnotes, merged cells, and conditional logic that a human reads easily and a computer cannot reliably parse without structured interpretation. TransCelerate has also pointed to a four-month average lag between protocol approval and study startup that traces directly back to this gap between how protocols are written and how downstream systems need to consume them [2]. The Unified Study Definitions Model, or USDM, is the CDISC standard built to close that gap, and a dedicated CDISC project is now applying it specifically to the Schedule of Activities [4].

Why This Topic Matters in Clinical Trials

Protocol amendments are expensive, and a number of them involve the kind of change a Schedule of Activities encodes: an added or removed procedure, a shifted visit window, a new assessment tied to an eligibility criterion. A 2024 analysis by Getz and colleagues, published in Therapeutic Innovation & Regulatory Science and built on 950 protocols and 2,188 amendments across 16 sponsors and CROs, found that the share of protocols requiring at least one substantial amendment climbed from 57% to 76% between a 2015 baseline and a 2022 follow-up period [11]. Mean amendments per protocol rose 60% over that span, reaching 3.3, and the time needed to implement a single amendment, from identifying the need to final approval, now averages 260 days [11]. An earlier analysis by the same lead author put a number on what an individual amendment costs: the median direct cost to implement one substantial amendment was $141,000 for a Phase II protocol and $535,000 for a Phase III protocol [12]. When the underlying schedule is a static table instead of structured data, each of those changes has to be manually re-keyed into the EDC build, the CTMS, and the SDTM Trial Design domains, along with whatever other downstream operational documents and systems reference the same schedule. A machine-readable SoA does not eliminate amendments, but it removes the re-transcription step that turns every one of them into avoidable downstream rework.

Current Evidence and Research Landscape

USDM is the reference architecture for TransCelerate's Digital Data Flow initiative, developed jointly with CDISC [1][2]. The project has moved through four main phases since 2021: reference architecture design (July 2021 to July 2022), biomedical concepts and timepoint integration (September 2022 to July 2023), ICH M11 alignment (August 2023 to April 2024), and refinement for complex and device studies (May 2024 to May 2025) [1]. Two workstreams ran alongside that fourth phase rather than as a separate stage: an education track that built CDISC's "Understanding USDM" course through October 2025, and a CORE workstream that converted the conformance rule specifications into machine-executable rules, completed in December 2025 [1][10]. A fifth phase is still being scoped [1]. USDM version 4.0 was released on June 3, 2025, packaging a logical data model, REST API specifications in JSON and YAML, a controlled terminology set, and an implementation guide [1]. CDISC published the first implementation handbook covering trial design domains separately, in mid-2026, building on that v4.0 foundation rather than shipping alongside it [6].

The Schedule of Activities piece is newer and narrower in scope. CDISC's SoA project, launched as a companion effort, exists because prose-based schedules with interpretive footnotes have historically resisted any consistent computable representation [4]. It is explicitly working toward a standardized way to encode visit windows, activity anchors, and conditional branching so that CTMS, EDC, and monitoring systems can all read the same schedule the same way [4].

It is worth separating this from ICH M11, since the two are related but distinct. The EMA adopted the M11 guideline at Step 5 on December 15, 2025 [9]. The FDA finalized its own M11 guidance document, template, and technical specification in May 2026, superseding the revised draft it had circulated in the Federal Register in June 2025 [7][8]. M11 standardizes the protocol document's structure, its section headers, and the technical data fields needed to exchange that content electronically across ICH regions, which is a broader scope than headers alone. What the publicly available M11 materials do not do is define the internal computable structure of the Schedule of Activities itself: how one visit's timing window or one activity's conditional trigger gets represented as data rather than as a sentence. That narrower problem is what CDISC's USDM-based SoA project is building, and the two efforts are designed to converge, since USDM 4.0 already incorporates the ICH M11 CeSHarP conventions as its alignment point [1][7][9].

Operational Impact for Sponsors, CROs, and Sites

Inside USDM, a schedule is not a table so much as a set of linked objects. CDISC's reference implementation and import tooling represent it as a timeline sheet: timepoints run across the columns, activities run down the rows, and a marked cell links a specific activity to a specific timepoint, with support for sequencing, conditions, and branching between them [5]. That spreadsheet view is a working format for building and importing a schedule, not the full normative data model itself, which is the JSON structure described in CDISC's specifications and implementation guidance [5][6]. The formal data model beneath that view adds Encounters for each visit, a ScheduleTimeline that sequences them, and ScheduledActivityInstance and ScheduledDecisionInstance records that tie specific activities, or specific branching decisions, to specific timepoints [5][6]. Activities link out to BiomedicalConcepts, so "fasting glucose at Visit 3" becomes a reusable, standardized definition rather than a line of free text retyped into every protocol that needs it [6]. Epochs, the study phases like screening or treatment, connect through explicit previous and next references rather than implied sequence, which is what lets software reconstruct visit order without a human reading the table first [6].

A single SoA row illustrates the pattern, though the following is illustrative of how the pieces fit together rather than a verbatim USDM instance. A conventional footnoted table might read: "Fasting glucose, Visits 2 through 4, if abnormal at screening." In USDM terms, that becomes an Activity record for fasting glucose linked to its BiomedicalConcept, three ScheduledActivityInstance records tying that activity to the Encounter objects for Visits 2 through 4, and a condition attached to the schedule that is meant to reference the screening result as a data field rather than describe it in prose [5][6]. The clinical meaning does not change, but a condition expressed this way is at least checkable: a validation rule can confirm it points to a field that actually exists elsewhere in the same study design, something no automated system could do against a footnote.

CDISC's own implementation handbook quantifies what that structure changes downstream. Building the SDTM Trial Design domains (arms, elements, visits, inclusion and exclusion criteria, trial summary) from a paper protocol currently takes clinical programmers 8 to 32 hours, followed by two to three weeks of validation [6]. With USDM-structured input, that same automation reduces programmer involvement to roughly one or two hours of final review before validation [6]. The handbook is explicit that this is not a fully unattended process. Complex designs, platform trials, basket studies, and anything with adaptive branching still need a human checking the automated output, and the handbook frames this as "keeping the human in the loop" rather than removing oversight [6]. For a site coordinator or CRA, the practical difference shows up earlier: a computable schedule can generate consistent visit windows and activity checklists automatically, instead of each system's build team independently re-reading the same footnoted table and occasionally reaching different conclusions about it.

Regulatory and Documentation Considerations

The chronology above matters for compliance planning, but the more durable distinction is what each standard actually checks. ICH M11, now final in both the EMA (Step 5, December 2025) and FDA (final guidance, May 2026) versions, governs the document: its section order, its required content, and the technical format for exchanging it electronically [7][8][9]. USDM's conformance layer governs the data underneath that document. CDISC has published CORE conformance rule specifications for USDM, moving from a written specification to versioned, machine-executable rules that were completed in December 2025 [10]. Those rules catch problems such as a previous-encounter reference that does not match the next-reference of the encounter it points to, before the data ever reaches an SDTM build [6][10]. A sponsor building toward both standards at once gets a protocol document that regulators can read consistently and a dataset underneath it that systems can validate automatically, but those remain two separate compliance checks rather than one.

AI and Automation Perspective

The realistic role for automation here is translation and validation, not authorship. USDM's structured format and its CORE rule set give software something concrete to check: whether an epoch sequence is complete, whether every scheduled activity resolves to a defined biomedical concept, whether a timing reference actually points to something that exists [6][10]. That is a tractable problem for automated tooling in a way that reading footnoted prose never was.

Converting an existing, unstructured Schedule of Activities into that format is a harder problem, and it is where the human-in-the-loop framing in CDISC's own handbook matters most [6]. Footnotes carrying conditional logic, such as "Procedure X only if Y was abnormal at screening," do not map cleanly to a fixed schema without someone confirming the mapping is correct. The handbook's estimate that clinical programmer effort drops from days of manual transcription to one or two hours of review only holds once the underlying protocol was authored in, or converted into, USDM format to begin with [6]. For legacy studies and in-flight amendments, the near-term reality is a mix of structured authoring for new protocols and supervised conversion for existing ones, with validation rules doing the checking rather than a fully autonomous parser doing the extraction.

Common Missteps to Avoid

A few confusions come up repeatedly as teams start planning for USDM and M11 together, and they are worth naming directly.

A Starting Checklist for Sponsors and CROs

Teams evaluating where to start with a USDM-based SoA typically work through a version of these steps:

1. 1Confirm which USDM version and CORE rule set a vendor or internal tool targets, since v3.0 and v4.0 rule specifications differ [10].
2. 2Decide whether new protocols will be authored directly in USDM-compatible tooling versus converted from existing templates after the fact, since the two paths carry different review burdens [6].
3. 3Map which downstream systems, EDC, CTMS, SDTM Trial Design builds, will consume the structured schedule, and confirm each can ingest the API format CDISC publishes [1][6].
4. 4Budget explicit human review time for conditional logic and adaptive designs rather than assuming full automation from the first protocol [6].
5. 5Track ICH M11 template alignment separately from USDM conformance, since a regulator-ready document and a validated dataset are two different checks [7][8][9][10].
6. 6Expect implementation maturity to vary by sponsor, CRO, and vendor, since tooling support for USDM v4.0 and the CORE rule set is still uneven across the industry [1][10].

How Kitsa Fits Into This Problem

KScribe, Kitsa's protocol and regulatory document generation tool, sits downstream of this same shift. As sponsors move toward structured, USDM-conformant study definitions, document generation tools that can consume that structured data directly, rather than re-extracting a Schedule of Activities from a PDF each time a related document is drafted, are positioned to keep a protocol's schedule consistent with the ICF, the IB, and the CSR that reference it. That is a design direction the industry is moving toward, not a claim about the current capability of any specific tool.

Key Takeaways

• USDM is CDISC and TransCelerate's reference model for representing clinical trial protocols as structured data instead of prose, with version 4.0 released June 3, 2025 [1].
• A dedicated CDISC project is applying USDM specifically to the Schedule of Activities, since footnoted, prose-based schedules have historically resisted consistent machine parsing [4].
• The share of protocols requiring at least one substantial amendment rose from 57% to 76% between 2015 and 2022, and a single amendment's median direct cost runs $141,000 (Phase II) to $535,000 (Phase III), the operational cost a computable SoA is built to reduce [11][12].
• USDM represents schedules through linked Encounters, ScheduleTimelines, and ScheduledActivityInstances rather than a flat table, enabling automated SDTM Trial Design domain generation [5][6].
• CDISC's own handbook reports that automation can cut clinical programmer effort on trial design domains from 8 to 32 hours down to roughly 1 to 2 hours of final review [6].
• ICH M11 (finalized by the EMA in December 2025 and by the FDA in May 2026) standardizes protocol document structure and exchange; USDM's SoA project separately standardizes the computable schedule data underneath it [7][8][9].
• Converting legacy, unstructured schedules into USDM format still requires human review, particularly for conditional footnote logic and complex adaptive designs [6].

FAQ

References

1. [1]CDISC. "Digital Data Flow (DDF) for Clinical Trial Protocols." CDISC.org. https://www.cdisc.org/ddf
2. [2]TransCelerate BioPharma. "Digital Data Flow Initiative." TransCeleratebiopharmainc.com. https://www.transceleratebiopharmainc.com/initiatives/digital-data-flow/
3. [3]TransCelerate BioPharma. "TransCelerate Advances Industrywide Shift To Digital Protocols." PR Newswire, October 8, 2025. https://www.prnewswire.com/news-releases/transcelerate-advances-industrywide-shift-to-digital-protocols-302577913.html
4. [4]CDISC. "Advancing Schedule of Activities with USDM: CDISC SoA Project Launch." Webinar. https://www.cdisc.org/events/webinar/advancing-schedule-activities-usdm-cdisc-soa-project-launch
5. [5]CDISC. "usdm" GitHub repository, README and Excel Timeline (Schedule of Activities) documentation. https://github.com/cdisc-org/usdm/blob/main/README.md
6. [6]CDISC. "USDM Handbook 1: Leveraging USDM Metadata to Automatically Construct the Trial Design Domains." Version 1.0. https://www.cdisc.org/sites/default/files/2026-06/USDM-HB1%20v1.0_FINAL.pdf
7. [7]U.S. Food and Drug Administration. "M11 Clinical Electronic Structured Harmonised Protocol." Guidance Document, M11 Template, and Technical Specification Document, Final Guidance, May 2026. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/m11-clinical-electronic-structured-harmonised-protocol
8. [8]U.S. Food and Drug Administration / Federal Register. "M11 Technical Specification: Clinical Electronic Structured Harmonised Protocol; Draft Technical Specification and Template; Availability." 90 FR 24146, June 6, 2025. https://www.federalregister.gov/documents/2025/06/06/2025-10359/m11-technical-specification-clinical-electronic-structured-harmonised-protocol-international-council
9. [9]European Medicines Agency. "ICH M11 Guideline: Clinical Study Protocol Template and Technical Specifications." Step 5, December 15, 2025. https://www.ema.europa.eu/en/ich-m11-guideline-clinical-study-protocol-template-technical-specifications-scientific-guideline
10. [10]CDISC. "USDM v3.0 and v4.0 Errata." CDISC.org. https://www.cdisc.org/ddf/usdm-errata-3-4
11. [11]Getz, K.A., Smith, Z., Botto, E., Murphy, E., and Dauchy, A. "New Benchmarks on Protocol Amendment Practices, Trends and their Impact on Clinical Trial Performance." Therapeutic Innovation & Regulatory Science, Vol. 58, March 4, 2024. https://link.springer.com/article/10.1007/s43441-024-00622-9
12. [12]Getz, K.A., Stergiopoulos, S., Short, M., Surgeon, L., Krauss, R., Pretorius, S., Desmond, J., and Dunn, D. "The Impact of Protocol Amendments on Clinical Trial Performance and Cost." Therapeutic Innovation & Regulatory Science, Vol. 50, No. 4, 2016, pp. 436-441. https://journals.sagepub.com/doi/abs/10.1177/2168479016632271