ARPA-H-SOL-26-164 · TA3 Agentic Operations
ARPA-H SURPASS and phaseless trials.
SURPASS asks teams to run one persistent, adaptive platform trial that carries each drug from first-in-patient safety to efficacy without stopping between phases. Its Agentic Operations area needs a layer that keeps the master protocol, every arm and every site in step as the trial adapts. AccuraTrials builds that layer: one governed model of the protocol, changed only by reviewed, signed diffs.
The program
What is ARPA-H SURPASS?
- SURPASS
- Simulation-augmented, Real-time Platform Adaptive Seamless Trials: a five-year program from the Advanced Research Projects Agency for Health (ARPA-H) to shorten the clinical development of drugs and biologics with persistent, multi-arm platform trials that share operational infrastructure and control information, combine validated predictions with accumulating human data, and make valid adaptive decisions as evidence emerges.
ARPA-H’s Proactive Health Office released the solicitation, ARPA-H-SOL-26-164, on 30 September 2026 as an Innovative Solutions Opening, and expects to award multiple Other Transaction agreements. The program directly advances HHS’s Operation TrialBlazer. Its stated aim is a future in which clinical development runs as a continuous learning system rather than a sequence of disconnected trials.
The problem it names is familiar to anyone who has run a development program. Taking a drug or biologic through the clinic takes about ten years, costs roughly $1 billion to $2 billion and fails more than 90% of the time, and each phase can require a new protocol, start-up process, control group, site network, contract structure, data system and regulatory submission. Infrastructure is rebuilt instead of maintained, and what one phase learned is not reliably carried into the next.
SURPASS does not fund point tools. Each Core Platform Team builds one platform trial in a single disease area, and ARPA-H judges the integrated clinical and technical performance of the whole platform rather than the completion of any one technical area.
The solicitation, line by line
From ARPA-H-SOL-26-164 as published on 30 September 2026. The solicitation and ARPA-H’s program page govern; check them before you rely on any line here.
- Agency
- ARPA-H, Proactive Health Office. Program manager: Daria Fedyukina, PhD.
- ISO
- Awards
- Multiple Other Transaction agreements, negotiated with the proposers whose solutions are most advantageous to the government.
- OT
- Length
- Five years: Stage I, platform and capability buildout (24 months), then Stage II, capability validation and deployment (36 months).
- 5 YEARS
- Go/no-go
- FDA authorization to proceed for the platform Master Protocol IND and for at least one investigational arm by the end of Stage I.
- STAGE I GATE
- Arms
- At least three drug or biologic arms. At least two should not be approved for any indication, including one with no prior human data, which may enter later.
- ≥ 3 ARMS
- Sponsors
- At least two for-profit drug or biologic sponsors, including one startup or small company. An investigator-initiated arm is encouraged.
- ≥ 2 SPONSORS
- Sites
- At least six by the end of the program, the first enrolling no later than Year 2. One site holds the master protocol IND and the statistical analysis plan, and processes must support IRB approval in under a month.
- ≥ 6 SITES
- Disease
- One disease area per platform, with a pipeline deep enough to sustain the platform for about a decade after funding ends. Neurodegenerative diseases are of particular interest.
- 1 DISEASE
- Platform operator
- A lead operator with the expertise, infrastructure, authority and long-term incentive to run a multi-sponsor platform. A disease-expert clinical institution or comparably durable organization is preferred.
- DURABLE
- Team
- Project managers and experts in biostatistics, regulatory affairs, clinical medicine, AI and predictive modeling, and clinical operations, plus the sponsors. Subcontractors, vendors and partners may supply capabilities.
- ALL 3 TAS
- Arm funding
- ARPA-H will consider funding trial expenses for the first three arms in full and the fourth and fifth in part, and does not anticipate funding arms beyond the fifth.
- ARMS 1–5
- Questions
- SURPASS@ARPA-H.GOV. Answers are posted to the program page.
- BY 2 NOV
The idea
What is a phaseless trial? One platform, first dose to efficacy.
- Phaseless (seamless) trial design
- A design that integrates the phases of clinical development, so each investigational arm moves from first-in-patient safety evaluation through efficacy, and where proposed confirmatory evaluation, inside one persistent platform trial instead of a new trial for each phase. SURPASS treats seamless and phaseless as synonyms.
In the sequential model a drug passes through Phase 1 (safety, tolerability, PK/PD, dose finding), Phase 2 (dose and exposure-response, preliminary efficacy) and Phase 3 (confirmatory efficacy and safety), and every stop between them is a restart. A phaseless platform keeps the objectives of each phase and drops the restarts: one master protocol, one site network, shared controls, and prespecified rules for adding, adapting, stopping or graduating arms.
At a minimum, a SURPASS design must carry every arm from first-in-patient safety evaluation through preliminary efficacy. For an ultra-rare disease it should span Phases 1, 2 and 3 and be capable of supporting registration.
Phaseless does not mean unstructured. A design may skip formal phase gates, but the objectives of each relevant phase still have to be met, and in the solicitation’s words, “continuous” or “real-time” inference does not mean unrestricted or undocumented examination of trial data. Every analysis and adaptation runs through authorized, reproducible and auditable procedures defined in the master protocol, the arm-specific appendices and the statistical analysis plan. That is a records problem as much as a statistics problem.
| Aspect | Sequential phases | Phaseless platform |
|---|---|---|
| Protocol | A new protocol for each phase and each drug | One master protocol, with arm-specific appendices built to be modified as arms join |
| Start-up | Sites, contracts and data systems set up again for each trial | A persistent site network and shared operational infrastructure, maintained rather than rebuilt |
| Controls | Control participants recruited separately for each intervention | Control information shared across concurrent arms, augmented with in silico methods only where evidence supports it |
| Decisions | Go/no-go at the end of each phase | Prespecified interim analyses, with a final target of monthly or more often |
| Arms | One intervention per trial | Arms enter, stop or graduate under prespecified decision rules |
| Learning | Lost between phases and between drugs | Simulation assumptions updated as blinded data accumulate, and the design re-simulated with real-world parameters |
Drawn from sections 2.1 to 2.2.4 of the SURPASS solicitation.
Three technical areas
Design, inference, operations. Judged as one platform.
Every Core Platform Team proposes and performs work in all three areas. No single organization is expected to hold every capability; the prime performer answers for how they integrate.
-
TA1
Phaseless Design Engine
A simulation-based design environment that plans first-in-patient, efficacy and, where proposed, confirmatory stages together. It integrates in silico models, such as digital twins, synthetic controls, disease-progression and PK/PD models, each with a defined role, its data inputs and the decision it informs.
Final success criteria
- A non-programmer trialist builds a simulator in ≤ 1 month
- Valid re-simulation after a design change in ≤ 5 hours
- ≥ 2 major decisions per arm informed by an in silico model
- Control allocation cut ≥ 80% by Year 5 with in silico methods
-
TA2
Continuous Inference
A continuous-inference and adaptation-validity engine: adaptive-platform, anytime-valid, group sequential or Bayesian methods that permit decisions on accumulating data without compromising validity, run through an independent statistical analysis committee with DSMB reporting and a blinding plan.
Final success criteria
- Prespecified interim analyses monthly or more often
- Sample size ≥ 50% below reference designs
- Correct go/no-go decisions 2.5× faster
- Time to the first effective drug ≥ 50% shorter
-
TA3Where AccuraTrials works
Agentic Operations
AI-enabled operational systems and an integrated clinical operations framework that let the platform and its arms start, enroll, execute, adapt and generate decision-ready data at speed, while keeping participant safety, data integrity, compliance, quality and inspection readiness intact.
Final success criteria
- Platform IND submission to first subject dosed in < 3 months
- New arm, FDA “may proceed” to first dose in < 1 month
- Data collection ≤ 2 days; cleaning and datasets ≤ 1 week
- Human hours at ≤ ¼ of historical FTE hours
TA3 · Agentic Operations
The seam between a decision and every site that acts on it.
When TA2 authorizes an adaptation, TA3 has to implement it across the master protocol, every affected arm and every site, then hand TA1 and TA2 data with its provenance intact. This is how that seam runs on AccuraTrials.
01
TA2 decides, under its own rules
An adaptation such as enriched eligibility, a stopped arm or a re-estimated sample size is decided by the platform’s prespecified rules, its independent statistical analysis committee and the decision rights in its governance framework. AccuraTrials does not make that decision. It implements the authorized change.
02
Compiled like the original
The change to the master protocol or an arm appendix is compiled into the same study model as the original and compared with the running platform. The result is a diff: which criteria, forms, edit checks and visit windows change, and how much collected data sits behind each one.
03
Reviewed and signed by people
Nothing deploys on the model’s authority. The people your governance framework names review the diff and approve it with an electronic signature that requires re-authentication and a recorded reason.
04
Deployed once, for every site
Every site runs on the same approved version of the model, so a change is made once instead of being re-keyed site by site and system by system. Your team deploys it the same day it is approved, after whatever validation your process requires.
05
On the record, and downstream
The change is written to the audit trail with its effective date, prior value and reason. Data that flows on to TA1 models and TA2 analyses carries its source, its model version and its full edit history.
TA3 metrics
Nine numbers TA3 is judged on. And which ones we move.
The interim and final success criteria from the solicitation, with what AccuraTrials contributes to each, including the ones that belong to someone else on the team.
| Metric | Interim | Final | AccuraTrials’ part |
|---|---|---|---|
| Platform IND submission to first subject dosed | < 6 months | < 3 months | In part The study build is drafted from the master protocol for review instead of specified and programmed by hand: about 48 hours to a reviewable setup in our first deployment. Contracting, IRB approval and site activation sit with the team. |
| FDA “may proceed” to first subject dosed, each new arm | < 3 months | < 1 month | In part A new arm’s appendix is compiled into the running platform as a reviewed change, so onboarding its forms, checks and visits is a review, not a rebuild. |
| Contract execution, each new arm | < 1 month | < 1 week | Partner Contracting belongs to the platform operator and the sponsors. |
| IRB approval, each new arm | ≤ 3 weeks | < 1 week | Partner Submission and review sit with the team’s central or commercial IRB. The reviewed diff documents exactly what an amendment changes. |
| Actual vs predicted enrollment, across arms | Within ±50% | Within ±25% | Partner Prediction is a TA1 model. Enrollment data leaves AccuraTrials EDC with its provenance. |
| Data collection time | ≤ 1 week | ≤ 2 days | Ours Sites enter data into AccuraTrials EDC with checks running as they type. Paper source documents can pre-fill forms by OCR for field-by-field review, ePRO arrives through a patient portal, and EHR integration is FHIR-ready. |
| Data cleaning and dataset construction | ≤ 3 weeks | ≤ 1 week | Ours Errors are stopped at entry, queries open on the exact field, coding uses MedDRA and WHODrug, and data exports as CDISC ODM XML with its audit trail. |
| Human hours vs historical FTE hours | ≤ ½ | ≤ ¼ | In part The protocol is compiled once instead of re-keyed into each system, and amendments arrive as reviewed diffs instead of re-specification. |
| Sponsor cost per arm vs three CRO quotes | Equal to the lowest | Below the lowest | In part Follows from the rows above. Pricing is usage-based and quoted on a call. |
Metrics from section 2.2.5 of the solicitation. Human hours are measured at first site activation, first subject dosed, the first and third arms fully recruited, and program end. The 48-hour figure is our internally reported first-deployment observation to a reviewable build, not an independent benchmark; go-live depends on the sponsor’s own review and validation.
How AccuraTrials fits
TA3, capability by capability. Including the ones that are not ours.
The solicitation lists twenty-three TA3 capabilities. These are the ones AccuraTrials does today, which product does them, and what a team still needs from a partner.
| TA3 capability | What AccuraTrials does | Product |
|---|---|---|
| Master protocol and arm documents | Compiles the master protocol and each arm appendix into one machine-readable model on CDISC USDM and ICH M11. Every rule links to its page and clause, and an ambiguous clause becomes a labelled gap for a person, never a guess. | Cortex |
| New-arm onboarding | Drafts a new arm’s visits, forms and edit checks from its appendix. The build imports into quarantine: reviewable section by section, and unable to enroll anyone until a named person activates it. | Studio |
| Implementing TA2 adaptations | An authorized adaptation enters as a protocol change, is compiled like the original, and is shown as a reviewed diff with the collected data behind each affected artifact, then deployed after sign-off. | Cortex · Studio |
| Data collection and reconciliation | Data capture built to 21 CFR Part 11, with OCR pre-fill from source documents, ePRO and eConsent, source data verification and FHIR-ready EHR integration. | EDC |
| Data cleaning and analysis-ready datasets | Edit checks drafted from the protocol and enforced at entry, queries on the exact field, MedDRA and WHODrug coding, data locks, CDASH-aligned fields and CDISC ODM XML export with the audit trail. | EDC |
| Clinical supply | Randomization and RTSM are built in, with blinded allocation, kit inventory, shipments and dispensing at sites, and allocation events in the same audit trail. | EDC |
| Protocol deviations | Evaluated on submit and by a time-based sweep, categorized per ICH E6(R3), graded minor, major or critical, and delivered with the clause and the data that tripped them. | EDC |
| Inspection-ready documentation | A hash-chained, tamper-evident audit trail, e-signatures bound to the record version, and every AI suggestion logged with what a person accepted, changed or rejected. | EDC |
| Traceable data for TA1 and TA2 | Every value traces to its source and every rule to its clause, the same protocol version always compiles to the same model, and exports carry the audit trail. | All four |
| Operations against TA1 predictions | Accura OS checks connected trial systems against the same model. It runs on AccuraTrials EDC today, with connectors to four more trial systems in prototype. | OS · in build |
| Everything else | Contracting and site activation, IRB and regulatory submissions, regulatory-path tools, site management, pre-screening, recruitment and retention, safety reporting, vendor oversight, the Trial Master File and resource forecasting come from the platform operator, a CRO or specialist vendors. CTMS and safety systems are integrated over REST, not replaced. | Partner |
Accura Studio is shipping. AccuraTrials EDC is live in a paid, regulated Phase I trial. Accura OS is in build. TA3 systems must operate inside the platform’s clinical, statistical, regulatory and governance structure, and so does ours.
Stated plainly
What we are, and what we will not claim.
A TA3 partner, not the platform
We would join a Core Platform Team as its operations layer. The disease expertise, the statistics and the sites come from the team.
- Independent
- AccuraTrials is not affiliated with or endorsed by ARPA-H or HHS and has no SURPASS award. This page summarizes the public solicitation.
- NO AFFILIATION
- Not the operator
- SURPASS prefers a disease-expert clinical institution or comparably durable organization to run the platform. We support that operator; we do not replace it.
- TA3 PARTNER
- Not TA1 or TA2
- We do not build simulators, digital twins or adaptive statistical methods. We implement the decisions they produce and give them traceable data.
- SCOPE
- No platform trial yet
- AccuraTrials EDC is live in a paid, regulated Phase I trial. It has not yet run a multi-arm platform trial under a master protocol.
- STATUS
- Agentic, not unsupervised
- The AI drafts and proposes; a qualified person reviews, approves and e-signs before anything reaches a live study. The solicitation asks the same of TA3: automation does not displace the authority of sponsors, investigators, sites, committees, IRBs or regulators.
- HUMAN SIGN-OFF
- Figures as reported
- The 48-hour setup and 100% source-link figures are internal observations from our first deployment, not independent benchmarks.
- REPORTED
Key dates
The SURPASS timeline. All times Eastern.
Dates and deadlines
The pitch dates are estimates that ARPA-H will confirm in its Solution Summary feedback letters. Registration and any changes are posted on the program page.
- 15 October 2026
- Informational webinar, online, 12 to 1 pm. Registration closes 12 October.
- WEBINAR
- 28 October 2026
- Registration closes for Proposers’ Day.
- REGISTER
- 2 November 2026
- Questions due to SURPASS@ARPA-H.GOV. Later questions may not be answered.
- QUESTIONS
- 6 and 9 November 2026
- Proposers’ Day. Day 1 is hybrid, in Washington, DC and online, 9 am to 5 pm; day 2 is online, 9 am to noon.
- PROPOSERS’ DAY
- 30 November 2026
- Solution Summary due by 5 pm through solutions.arpa-h.gov. Required to go further: four pages, not counting the cover page, the rough order of magnitude and references.
- REQUIRED
- 22 January 2027
- Solution Pitch Package due by 5 pm: a deck of no more than 21 slides, an Excel Gantt chart, a preliminary master protocol, a two-page in silico and AI model description, a scope of work and a cost workbook.
- ESTIMATED
- 1 to 5 February 2027
- Virtual pitches: a 15-minute presentation and up to 45 minutes of questions.
- ESTIMATED
FAQ
SURPASS and phaseless trials, answered plainly.
Updated October 2026
What is ARPA-H SURPASS?
SURPASS (Simulation-augmented, Real-time Platform Adaptive Seamless Trials) is a five-year program from the Advanced Research Projects Agency for Health, released on 30 September 2026 as solicitation ARPA-H-SOL-26-164. It funds Core Platform Teams to build persistent, adaptive, multi-arm platform trials that share infrastructure and control information, use validated predictive models and make valid adaptive decisions as data accumulate, so drugs and biologics can be evaluated faster, at lower cost and with fewer participants.
What is a phaseless clinical trial?
A phaseless, or seamless, trial integrates the phases of clinical development into one persistent trial. Instead of a new protocol, site network and control group for Phase 1, Phase 2 and Phase 3, each investigational arm moves from first-in-patient safety evaluation to efficacy, and where proposed to confirmatory evaluation, under one master protocol with prespecified rules for adapting, stopping or graduating arms. SURPASS uses seamless and phaseless as synonyms.
What are the three SURPASS technical areas?
TA1, the Phaseless Design Engine, is a simulation-based design environment that integrates in silico models such as digital twins. TA2, Continuous Inference, is an engine for valid adaptive decisions on accumulating data, run through an independent statistical analysis committee. TA3, Agentic Operations, is the AI-enabled operational system that lets the platform and its arms start, enroll, execute, adapt and produce decision-ready data. Every Core Platform Team must cover all three.
What does SURPASS TA3 Agentic Operations cover?
The solicitation lists capabilities from pre-trial activities, master protocol and arm-specific document management and new-arm onboarding, through contracting, site activation, IRB and regulatory workflows, recruitment, data collection, data cleaning and analysis-ready datasets, to deviation management, safety reporting, clinical supply, vendor oversight, the Trial Master File, operational implementation of TA2 adaptations and inspection-ready documentation. TA3 must also give TA1 and TA2 timely, traceable and reproducible data.
How will SURPASS TA3 be measured?
The final success criteria include under three months from platform IND submission to first subject dosed, under one month from FDA “may proceed” to first subject dosed for each new arm, contract execution and IRB approval for each new arm in under one week, enrollment within 25% of prediction, data collection within two days, data cleaning and dataset construction within one week, human hours at a quarter of historical FTE hours, and sponsor cost per arm below the lowest of three CRO quotes.
What is the SURPASS Stage I go/no-go gate?
FDA authorization to proceed for the platform’s Master Protocol IND, and for at least one investigational arm, by the end of Stage I is a required go/no-go criterion for continuing into Stage II. Stage I, platform and capability buildout, lasts 24 months; Stage II, capability validation and deployment, lasts 36 months.
Who can lead a SURPASS Core Platform Team?
The prime can be any organization able to sustain the platform trial after the program ends. Each team must name a lead platform operator with the expertise, infrastructure, authority and long-term incentive to run a multi-sponsor platform, and a disease-expert clinical institution or comparably durable organization is preferred. Teams should include at least two for-profit drug or biologic sponsors, one of them a startup or small company. FFRDCs and government entities, including federal employees, may not take part as prime or sub-performers.
When are SURPASS Solution Summaries due?
By 5 pm Eastern on 30 November 2026, through the ARPA-H Solutions site. A Solution Summary is required to go further. It is limited to four pages, excluding the cover page, the rough order of magnitude and references, and must cover the disease and arms, the TA1 to TA3 innovations, regulatory strategy, sustainability after the program and why the solution beats its competitors. Solution Pitch Packages are due an estimated 22 January 2027.
How does AccuraTrials fit into SURPASS?
As a TA3 technology partner to a Core Platform Team. AccuraTrials compiles the master protocol and its arm appendices into one governed model, builds and amends the study from it as reviewed diffs, captures data in AccuraTrials EDC, built to 21 CFR Part 11, and keeps every rule, change and data point traceable to its clause and its approver. Contracting, IRB, recruitment, safety reporting and the Trial Master File come from partners, and TA1 and TA2 from the team’s modelers and statisticians.
Does agentic AI replace sponsors, investigators or IRBs in SURPASS?
No. The solicitation states that AI-enabled automation does not alter or displace the responsibilities and decision-making authority of sponsors, investigators, clinical sites, statistical committees, the DSMB, IRBs or regulators. AccuraTrials is built the same way: the AI drafts and proposes, and a qualified person reviews, approves and e-signs before anything reaches a live study.
Is AccuraTrials affiliated with ARPA-H?
No. AccuraTrials is an independent company with no SURPASS award, and this page is not endorsed by ARPA-H or HHS. It summarizes the public solicitation; ARPA-H’s SURPASS program page and the solicitation on SAM.gov are the authoritative sources.
How do we talk to AccuraTrials about a SURPASS team?
Email us at jgui@accuratrials.com. We want to hear from lead platform operators, drug and biologic sponsors, clinical sites, and the biostatistics and in silico modeling groups building TA1 and TA2.
Building a SURPASS team? Bring the master protocol.
Thirty minutes with our team: we model one protocol live on the real platform and walk through how an adaptation reaches every site.
Or write directly: jgui@accuratrials.com