01 · Documents · what should happen
The rules are trapped in PDFs
Every team re-reads the protocol and re-types its rules into its own tool.
The intelligence layer for clinical trials
Turn every trial into one model AI can act on.
One study ontology for the protocol, its decisions and its data. AI agents work from it. Your team signs off.
Built on established clinical standards
Every element of the model carries a standard meaning.
- CDISCUSDM · CDASH · SDTM
- OpenStudyBuilderStudy definitions
- ICH M11Structured protocol
- TransCelerateDigital Data Flow
- MedDRAAdverse events
- WHODrugMedications
- LOINCLab tests
- SNOMED CTClinical terms
- CDISC ODMData exchange
The problem
Every trial runs on documents, decisions and data. Apart.
People carry the context between systems by email, spreadsheet and re-keying.
02 · Decisions · what people approve
The reasons live in inboxes
Approvals and their rationale sit in email until an auditor asks for them.
03 · Data · what actually happens
Every system keeps its own copy
EDC, CTMS, IRT, labs, ePRO and safety each hold their own version of the rules.
- 6+
- logins per study at nearly 7 in 10 sites
- 60%
- of sites re-key data into sponsor and CRO systems
- 68 days
- to build and release one study database
- $535K
- median cost of one major Phase III amendment
Advarra site surveys (2023, 2024) · Tufts CSDD and Veeva (2017) · Getz et al., Therapeutic Innovation & Regulatory Science (2016). Industry figures, not AccuraTrials results.
The operating layer
The operating system for clinical trials.
Accura OS is built to take in every document, decision and data point a trial produces, from the protocol and its amendments to lab results, eTMF records and site data, and to run them as one governed model. It bends as the trial does: arms added or dropped, amendments applied, nothing rebuilt.
01 · Ingest everything
Every document, decision and data point
The protocol and its amendments, the SAP, consent forms, lab manuals and SOPs, the approvals and committee decisions, and the study data itself. Read once, never retyped.
02 · The model
Model the whole trial once
Documents, decisions and data become typed, linked entities: arms, visits, forms, rules, lab results and approvals, each traced to the clause, record or signature it came from.
03 · Accura OS
A flexible operating layer
Every system and every AI agent runs on the model your team approves. Add an arm or drop one, apply an amendment, change a form: the model moves with the trial, and nothing is rebuilt.
- EDC LIVE
- CTMS
- eTMF
- CENTRAL LABS
- IRT / RTSM
- SAFETY / PV
04 · Connected systems
Every system on one model
AccuraTrials EDC runs on it today, with randomization, ePRO and eConsent built in. Connectors to central labs, eTMF, CTMS and safety systems are in development.
Running the trial
Trials change while they run.The model moves with them.
Arms open, stop and graduate. Amendments land mid-study. Data arrives from sites, labs and safety systems. Because the study is one model, each change is applied to the running study and signed, the data already collected stays where it is, and nothing is rebuilt.
01 · Add or change an arm
A new arm is an amendment, not a new build.
The arm’s visits, forms and checks arrive as a reviewed change to the running study. Every form, check and collected value the change touches is counted before anyone signs, and participants already enrolled keep their study as it was.
Amendment review and apply built into AccuraTrials EDC02 · Stop, graduate or reopen
Arm decisions with a second signature.
Proposed, reviewed by a different qualified person, then executed with an electronic signature. Only new assignments change: participants on the arm keep their treatment, follow-up and data, and arm names stay out of the general audit view to protect the blind.
Governed arm decisions built into AccuraTrials EDC03 · Edit a form mid-study
Every patient’s copy, brought current in place.
Change a form and existing patient forms update where they are. Entered values, open queries and locks stay as they were, each copy is audited, and a signed form changes only if you opt in, which voids its signature with the reason recorded.
Form propagation built into AccuraTrials EDC04 · Reconcile as you go
Differences become decisions, not a cleanup sprint.
Adverse events are reconciled against safety reporting case by case, with each discrepancy resolved and verified by a second investigator. Drug-kit counts are reconciled site by site with signed evidence, and every import is reconciled against its source. Open safety items hold the study’s closeout until they are resolved.
Safety, supply and import reconciliation built into AccuraTrials EDCAI agents
Agents for the workaround every study.
They draft, check and trace from the same model your team reads.
Study build
Configure the study from the protocol: visits, forms, branching and checks, each linked to its clause.
Worked example · AI build workflows in developmentData review
Investigate issues as data arrives. The check fires, the clause is attached, a query is drafted.
Edit checks in production · AI drafting in developmentMedical coding
Suggest MedDRA and WHODrug codes for events and medications. A coder confirms each one.
In development · shaped with medical codersAmendment impact
Trace every change study-wide. See the forms, checks and collected data an amendment touches before it ships.
Illustrative · in developmentAgent run
Every step recorded. Every action approved.
- AuditableInputs, outputs, sources and approvals recorded; the run replays.
- Source groundedEvery step cites the protocol, specification or data behind it.
- Versioned executionAI drafts; approved rules run as versioned, deterministic code.
Reconcile central lab resultsACT-217 · illustrative
- Ingest412 results receivedDone
- MapMatched to visits and LOINC codesDone
- CheckRanges and visit windows3 flagged
- Draft3 queries, each citing its ruleDrafted
- ApproveData manager sign-offAwaiting
Trust
Built for inspection,not for demos.
EDC controls designed to support the rules your study answers to. Study-specific validation and your own procedures still apply.
- 21 CFR Part 11Records · signatures
- ICH E6(R3)Good clinical practice
- HIPAADesigned to support
- GDPRDesigned to support
Infrastructure built to the SOC 2 trust principles, with an information security program aligned to ISO 27001. Neither is a completed audit or certification yet.
Who it’s for
Built for the teamsthat run trials.
01
CROs
Every sponsor’s study on one governed model, build to lock.
- Data managers
- Project managers
- Medical coders
02
Sponsors
One study definition across programs, vendors and amendments.
- Clinical operations
- Data management
- R&D IT
03
Biotechs
A first trial without a long vendor build.
- Clinical leads
- Data managers
- Quality
- 01PilotInside one live study
- 02Full studyThe whole study on one model
- 03PortfolioMore studies, same model
Why now
The U.S. is now askingfor exactly this.
HHS and ARPA-H launched SURPASS on 30 September 2026: a five-year program to take clinical development from 10+ years to under four.
< 4 yearsProgram goal, down from 10+
- TA1Phaseless design engineSimulation plans every stage together.
- TA2Continuous inferenceDecisions on accumulating data, still valid.
- TA3Agentic operationsWhere AccuraTrials fits, as a technology partner.
Independent summary of solicitation ARPA-H-SOL-26-164. AccuraTrials is not affiliated with ARPA-H. Read the SURPASS brief
Questions
What teams ask first.
What is Accura OS?
The operating layer for clinical trials: one governed model built from a trial’s documents, decisions and data, that every system and AI agent runs on. It applies changes to the running trial, such as adding or dropping an arm or applying an amendment, without rebuilding the study. AccuraTrials EDC runs on it today, and connectors to central labs, eTMF, CTMS and safety systems are in development.
What is the AccuraTrials semantic layer?
An ontology of the study. Arms, visits, activities, forms, items, rules, endpoints and decisions are typed entities with explicit relationships, each linked to the protocol clause, specification or approval it came from, so people, systems and AI agents read the same study.
How does AccuraTrials use CDISC USDM and OpenStudyBuilder?
They are the backbone of the model. USDM gives the study definition its structure, OpenStudyBuilder is its open-source foundation, and CDASH, SDTM and controlled terminologies such as MedDRA, WHODrug, LOINC and SNOMED CT give each element a standard meaning.
Do we have to replace our EDC or CTMS?
No. AccuraTrials EDC is in production and can run a study end to end, and the study model is designed to sit across the systems you already use. Connectors to other trial systems are in development.
Can we add or change arms while a trial is running?
Yes. A new or changed arm arrives as a reviewed amendment applied to the running study, with every form, check and collected value it touches counted before anyone signs. Stopping, graduating or reopening an arm is a governed decision: proposed, reviewed by a different qualified person and executed with an electronic signature. Participants already enrolled keep their assignments and their data, and the study is not rebuilt.
How does AccuraTrials handle data reconciliation?
Continuously, inside the study rather than in a cleanup sprint before lock. Adverse events are reconciled against safety reporting case by case, with each discrepancy resolved and verified by a second investigator; drug-kit counts are reconciled site by site with signed evidence; and every import is reconciled against its source. Open safety items hold the study’s closeout until they are resolved.
Give your next study one model to run on.
See the EDC in production and the agent workflows we’re building next.