The Next Adaptive Trial Bottleneck Is Controlled Change at the Site

SURPASS is proposing a continuous trial operating model

ARPA-H launched SURPASS on September 30, 2026, alongside related projects addressing site capacity, national data infrastructure, and patient navigation. SURPASS itself is organized around three technical areas: a simulation-supported design engine, a continuous inference engine, and an operations layer intended to automate startup, arm onboarding, data cleaning, and dataset construction. The solicitation remains open, with solution summaries due November 30, 2026. ([arpa-h.gov](https://arpa-h.gov/news-and-events/hhs-launches-surpass-and-new-efforts-accelerate-faster-smarter-clinical-trials))

This is more than a faster version of the familiar multicenter trial. A perpetual platform may evaluate multiple interventions under a shared protocol, use a common control, and add or remove treatment arms as evidence accumulates. The infrastructure remains active even as the specific scientific questions change.

That continuity can reduce repeated startup work at the program level. It can also move operational complexity downstream. If a platform changes faster than sites can validate, train, document, and execute those changes, theoretical efficiency becomes protocol deviation risk.

Feasibility has to be assessed twice

Conventional feasibility often asks whether a site has the investigator, population, equipment, and staff required for a defined protocol. A platform trial needs two related assessments.

The first is platform feasibility. Can the site support the master protocol, long-term governance model, core assessments, data expectations, randomization process, safety reporting, and ongoing training burden? The second is arm-level feasibility. Can the site safely deliver a particular intervention with its specific eligibility criteria, pharmacy requirements, laboratory handling, procedures, and follow-up schedule?

These questions should not be collapsed into one optimistic enrollment estimate. A site may be an excellent platform participant but lack the facilities for one arm. It may support an oral product but not an infusion requiring observation, specialized preparation, or after-hours coverage. Partial participation should be designed into the operating model rather than treated as a site failure.

A 2026 JAMA Network Open survey of experienced platform-trial teams found that shared infrastructure offered real advantages, but respondents also reported greater planning needs, complex randomization and drug-supply requirements, insufficient site resources, and difficulty adding new comparisons. ([pmc.ncbi.nlm.nih.gov](https://pmc.ncbi.nlm.nih.gov/articles/PMC13047460/))

• Define the site’s core platform capabilities separately from arm-specific capabilities.
• Document which interventions, procedures, visit types, and participant populations the site can support.
• Reassess competing studies, staffing, equipment, and referral access before each new arm opens.
• Do not assume prior platform participation automatically establishes readiness for the next intervention.

Every adaptation needs a deployment pathway

A sponsor decision to add or close an arm is not the same as site implementation. Between the two are regulatory review, contract and budget changes, system configuration, drug shipment, pharmacy preparation, staff training, delegation updates, consent deployment, participant communication, and confirmation that obsolete materials are no longer in use.

The platform therefore needs a formal change-deployment process with named owners and readiness gates. A site should not be considered ready because an amendment was emailed or a training module was assigned. Readiness means the applicable approvals are in place, affected personnel can execute the revised workflow, systems reflect the correct version, required supplies are available, and the site can demonstrate which participants are affected.

Version control becomes especially important when sites participate in different combinations of arms. A coordinator must be able to determine the correct consent form, eligibility pathway, randomization instructions, laboratory kit, pharmacy manual, and visit schedule for the participant in front of them. That determination cannot depend on memory or a crowded email inbox.

ICH E6(R3) requires trial processes to be operationally feasible, proportionate to risk, and supported by clear roles, fit-for-purpose systems, proactive risk management, and traceable records. Those expectations apply just as strongly when the protocol is designed to change. ([database.ich.org](https://database.ich.org/sites/default/files/ICH_E6%28R3%29_Step4_FinalGuideline_2025_0106_ErrorCorrections_2025_1024.pdf))

• Maintain a site-specific applicability matrix for the master protocol and each active arm.
• Use documented go-live criteria rather than a single activation date.
• Include pharmacy, laboratory, recruitment, regulatory, data, and participant-facing teams in change assessment.
• Test revised randomization and data workflows before live participant use.
• Archive superseded tools promptly while preserving required records and traceability.

Real-time inference depends on real-time operational discipline

Continuous analysis is only useful when the underlying data are timely, complete, and interpretable. In a traditional study, delayed entry may produce queries and monitoring findings. In an adaptive study, the same delay may affect an interim analysis, the timing of an arm decision, or the information available to a data monitoring committee.

Sponsors should identify which data are time-critical and translate that requirement into site workflows. Not every field needs immediate entry. Randomization variables, primary endpoint data, key safety information, treatment exposure, discontinuations, and events that influence adaptation decisions may require tighter turnaround than lower-priority data.

This is an appropriate use of quality by design. The goal is not to demand same-day completion of every form. It is to protect the data and processes that are critical to participant safety and reliable decisions. FDA’s adaptive-design guidance similarly emphasizes that adaptive trials require careful advance planning, conduct, simulation, and reporting. ([fda.gov](https://www.fda.gov/regulatory-information/search-fda-guidance-documents/adaptive-design-clinical-trials-drugs-and-biologics-guidance-industry))

Operational agreements should specify when the clock starts, who owns each handoff, what happens when source information is unavailable, and how overdue critical data are escalated. Sites then need staffing and coverage that match those expectations, including during vacations, weekends, investigator absences, and periods of high enrollment.

• Classify data by decision impact rather than giving every field the same deadline.
• Track turnaround from participant visit to source completion, data entry, review, and query resolution.
• Create backup coverage for critical assessments and safety reporting.
• Notify sites before major analyses so outstanding high-impact data can be addressed without revealing protected aggregate information.

Consent and pharmacy workflows cannot be afterthoughts

Adaptive platforms can be difficult to explain because the set of available interventions may change over time. Participants need a clear account of what is available to them, how assignment works, which information may change, and what future changes could require new information or reconsent. The explanation should remain understandable even when the underlying statistical design is complex.

Sites need advance rules for participants already screening, randomized, receiving treatment, or in follow-up when an arm closes or materially changes. The sponsor, IRB, investigator, and site team must be aligned before participant-facing communication begins.

Pharmacy readiness is equally consequential. Each added arm may introduce new storage conditions, accountability records, preparation instructions, blinding controls, expiry management, and emergency procedures. Published experience with adaptive platform trials has identified investigational pharmacy as a central operational function for bringing new agents into an active platform. ([doi.org](https://doi.org/10.1136/bmjopen-2021-060664?utm_source=openai))

A platform should maintain a reusable onboarding structure, but each product still needs an arm-specific risk review. Shared infrastructure does not make distinct investigational products interchangeable.

• Develop plain-language explanations of the platform before enrollment starts.
• Map consent requirements to each possible participant status during an adaptation.
• Include the investigational pharmacy in feasibility and amendment planning from the beginning.
• Require documented pharmacy release before randomization to a newly opened arm.

The site contract must pay for continuity, not just visits

Traditional budgets are often built around startup fees, participant visits, procedures, and closeout. That structure can underfund the work that keeps a platform functional between adaptations: recurring training, feasibility reassessment, document replacement, pharmacy setup, system testing, staff meetings, and participant communication.

Sponsors should establish standing fees or defined unit costs for adding and closing arms, implementing substantial amendments, retraining personnel, revising consent, and performing nonparticipant-specific transition work. Sites should not have to renegotiate from zero every time the platform does what it was designed to do.

The larger lesson from SURPASS is that continuous trials require continuous operational capacity. Automation may reduce duplicated work, but it will not eliminate local accountability. The sites most prepared for this model will not simply be the sites with the largest databases or highest historical enrollment. They will be the sites that can absorb change quickly, prove readiness, and keep participant protection and data reliability stable while the trial evolves.