The US Department of Health and Human Services (HHS) has launched SURPASS, an Advanced Research Projects Agency for Health (ARPA-H) programme that seeks to replace the traditional phased model of drug development with continuous, adaptive platform trials.
The Simulation-augmented, Real-time Platform Adaptive Seamless Trials (SURPASS) programme was announced at the end of September alongside three complementary projects.
ARPA-H says clinical development of drugs and biologics currently takes more than a decade, costs up to $2bn on average and fails more than 90% of the time.
It aims to shorten development to under four years.
What is SURPASS?
SURPASS combines predictive computational models, shared trial infrastructure, common control groups and real-time analysis.
It has the following three technical areas:
- A phaseless design engine that uses digital twins and other predictive models to simulate clinical and operational outcomes before launch, so trials can run with fewer patients
- A continuous inference engine that enables always-valid, real-time analysis and rapid adaptations, reducing the need for large conventional control groups
- An agentic operations layer that automates trial start-up, supports onboarding of new treatment arms and speeds data collection, cleaning and dataset construction.
The three companion projects address infrastructure.
STACK will use AI to activate new clinical sites, COMMONS will build a national, privacy-by-design consent and data architecture and CINCH will help patients contribute real-world data and find suitable trials.
HHS Secretary Robert F. Kennedy, Jr said the department is “taking on the delays, duplication and unnecessary costs that slow clinical trials down.”
The effort sits within Operation TrialBlazer, launched in June to encourage early-phase studies to stay in the US rather than move to faster locations such as China and Australia.
Execution is the test
Dr Richard Graham, Chair and co-founder of clinical trial technology company TruTechnologies, said the operational layer is what makes SURPASS notable.
“I’ve said repeatedly this year that if the US wants to become more competitive in clinical development, speeding up regulatory pathways alone won’t get us there,” he explained.
"We also have to modernise how clinical trials are executed once they are underway. A more continuous model depends on execution keeping pace and SURPASS explicitly builds that into its requirements.”
What this means for manufacturing and supply
The SURPASS solicitation sets out no new manufacturing, scale-up or comparability requirements and Dr Graham is cautious about predicting the response from sponsors and CDMOs.
“What it does make clear is that a seamless model only works if drug supply can keep pace and clinical supply management is explicitly part of the operational framework,” he said.
Traditional phase transitions give sponsors natural pauses to re-forecast demand, lock in formulations and reserve manufacturing capacity.
"Exactly how that affects scale-up, comparability or batch planning will depend heavily on the product and programme.”
The one-week data target
ARPA-H has set a final success target of one week for data cleaning and construction of analysis-ready datasets, but Dr Graham said that target exposes a visibility problem.
ARPA-H’s one-week target is difficult to reconcile with a trial model where sponsors may spend most of that week simply waiting to see what happens at the site.
He cited a 2017 Tufts CSDD study that found an average of 8.1 days from patient visit to EDC entry.
In a neurology study with a large pharma sponsor, he added, site activity that typically took 5-11 days to appear in EDC became visible to the sponsor in about three minutes after TruTechnologies’ platform was implemented.
Graham added that Tufts CSDD found protocol deviations per pivotal Phase III trial rose 56% between 2020 and 2025, while Phase II and III trials still take about 60 days on average to reach database lock.
The biggest barrier is that data quality is still too often treated as a downstream cleanup exercise instead of something built into trial execution from the start.
“The updated ICH E6(R3) guideline points in the same direction,” he said.
"To get anywhere near a one-week target, sponsors have to build quality into execution by helping sites get protocol-required work right as it happens and identifying issues while there’s still time to correct them.”
What happens next
ARPA-H will hold an informational webinar on 15 October and a Proposers’ Day on 6 November.
For manufacturers, the immediate question is whether supply chains that have been built around the traditional discrete phase gates can support a continuous model.
