Clinical Trial Arenas: The Stadium We Build for One Game
The Two-Year Build
Imagine a sports league that built a brand-new stadium for every game on its schedule, then took it down so the next could go up somewhere else. Games would happen once every three years and tickets would have to cover four billion dollars of construction for three hours of football. No league would do this. It is, with uncomfortable fidelity, how we run clinical trials.
The build starts at the synopsis and runs through the protocol, the regulatory interactions, the CRO selection, the database, site selection and contracts, and the training of everyone who runs the contest. Roughly two years from first document to first patient. Then the contest runs, and the result is not always shared: we usually learn who won, rarely the score, almost never the tape. Nine months later an article appears. Then it is all dismantled, and somewhere else another is built from scratch. Seen this way, the expense of clinical research stops being mysterious: we are not paying for the game, we are paying to build a stadium and then demolish it.
What a Standing Arena Looks Like
There are two general forms. A platform trial is an arena where multiple treatment regimens enter, take their turn and leave. A basket trial is an arena where multiple distinct patient groups, different mutation subtypes, different histologies, are evaluated against a treatment. And huge benefits come when one trial is both a platform and basket trial.
What holds an arena together is the master protocol, whose defining property is modularity: it describes the rules of the game without listing the teams. Arms live in appendices such as sub-studies, in the FDA's language, that plug in and pull out without reopening the governing document. I-SPY 2 has ran this way in neoadjuvant breast cancer since 2010, for more than 10 years, with twenty-seven or twenty-eight arms passing through one arena. Woodcock and LaVange describe the framework in the New England Journal of Medicine in 2017.
What An Arm Gives Up, And What It Gets Back
Suppose an arm wants 100 patients on treatment and 100 on control. Entering a platform where two other arms are already enrolling, it randomizes three-to-one and funds 133 patients: 100 active, 33 control.
In analysis it does not stop at 33. Patients eligible for this arm but randomized to another arm's placebo come back into the comparison, so the arm reads out against 100 or more controls, and the sponsor receives the data package it would have built alone.
The timeline is the more striking number. Building your own arena means roughly two years to first patient. In I-SPY 2 an arm went from decision to enrolling in four months, because the master protocol was approved, only the appendix needed IRB review, the database existed and the sites were open.
Sites, Regulators, And The Human Condition
For a site, the arena replaces a repeated startup problem with a single contract and a standing relationship. For regulators the ground has moved considerably, and the pandemic accelerated it; guidance exists, and the open questions are the ones any trial raises such as non-concurrent controls, historical controls, endpoint selection.
For patients the trade is real and worth stating plainly. You do not choose your cohort: three arms are enrolling and you are randomized among them. You are not blinded across arms which is near-impossible when arms come and go, but you are blinded within your cohort, whose investigator brochure you read and consent to specifically. And in a disease people live with for decades, finishing one arm does not end the relationship: patients can enter the platform again on a different arm.
The Arithmetic At The Level Of A Disease
Five arms running five separate trials, each wanting 100 active and 100 control, enroll 1,000 patients: 500 active, 500 placebo. The same five arms inside one platform at three-to-one enroll 133 each with 500 active and 167 placebos, 667 in total, with every arm's inference as strong as it would have been alone, and probably stronger.
The first number is 333 fewer patients. The second is 333 fewer people on placebo, and that is the one patient organizations raise first.
A third gain gets less attention. When every arm draws on a shared, community-owned control group, the disease itself becomes legible: variation across patients, the behavior of endpoints, the shape of natural history. The HEALEY ALS Platform Trial feeds its data into the PRO-ACT database. Five separate trials would produce five private answers and no shared understanding.
Shots That Would Otherwise Never Be Taken
Limb-girdle muscular dystrophy has 39 identified subtypes, and the rarest are rarely, if ever, put into a trial. A sponsor interested in four of them would need 80 sites for a standalone study, so it never happens. Inside a basket trial the arena is already standing, and the subtypes are already enrolling, so adding four is a small step rather than a program, and nine patients still support an inference when strength is borrowed across groups.
The First Arena Is The Hard One
None of this is free. It is easier to build a stadium for one game: you can customize it, and someone is paying, because this is how they have always done it. Building something modular, for games not yet scheduled and teams not yet named, is a larger lift with a more diffuse beneficiary. That, not the statistics, is the bottleneck.
It is also why initially arenas have been built in rare disease, and why patients have so often built them. The HEALEY ALS Platform Trial carries Sean Healey's name because he was a patient who judged the opportunity worth funding, and once that arena existed each additional arm became a far smaller lift. Patients see most clearly what the current system costs them, and in genetic disease they are looking past their own treatment to their children and their children's children.