The End of the Phase Trial: Evidence as a Continuous Stream

The thesis

Phase I, II and III are not scientific truths — they are batch processing inherited from a world of paper forms and quarterly board meetings. We freeze a protocol, enrol for years, lock the database, and only then learn. It is the slowest possible way to acquire knowledge.

The redesign

Run one continuous evidence platform per disease. Patients enter a perpetual, master-protocol trial; arms are added and dropped by pre-registered Bayesian rules; dose, population and endpoints adapt as evidence accrues. The "phases" dissolve into a single flowing posterior probability of benefit and harm that regulators can watch in real time.

This raises the bar, not lowers it

• Fewer patients on losing arms. Response-adaptive randomisation moves people toward what is working.
• Failure is fast and cheap. Bad molecules die in weeks, freeing capital for better ones.
• Continuous safety. Signals surface the moment they emerge, not at database lock.

The hard part is governance

This only works if the decision rules are frozen even though the data are not — pre-registered, machine-checked, and auditable. That is a solvable engineering and policy problem, and it is the real work.

Call to the industry

Stop optimising the trial. Redesign the unit of evidence. The winners of the next decade will treat evidence like a live data stream — versioned, monitored and always-on — not like a report you write once and defend forever.

A first-principles provocation from the Anxya Health Futures desk. Directional and informational — not medical, legal, financial or regulatory advice. The point is to move the debate, then do the hard validation work.