Getting a drug approved is slow, and a lot of the wait is the data. A trial runs for years, then the results are packaged up and sent to the regulator, who reads through the lot. On 28 April, the US Food and Drug Administration said it wants to change that, starting with two pilot trials that feed their results to the agency as they happen.
What real-time actually means here
Instead of a single submission at the finish line, the two proof-of-concept trials will report their endpoints and data signals to the FDA continuously. AI and cloud tools will flag patterns and clean data as it arrives, not months later. The agency took public comments on the plan until 29 May, so the framework is still being shaped.
How much time it could save
The FDA's chief AI officer, Jeremy Walsh, said the approach could cut 20 to 40 per cent off the overall length of a trial. That is the agency's own estimate, and the pilots will test whether it holds. Even the low end would be meaningful. Months are not an abstraction to a patient waiting on a treatment for a serious illness.
Where speed could cost something
Watching trial data in real time, with software helping interpret it, brings an obvious risk: that speed erodes the caution that makes approvals trustworthy. The FDA is framing this as efficiency, not a lower bar. The pilots are small for a reason. The test will be whether the standard of evidence stays put as the clock speeds up.
Why it reaches beyond the US
The FDA sets the tempo that other regulators, including those across Asia, tend to follow. If real-time reporting works, expect the model to travel. For anyone who has watched a family member wait out a slow approval, a shorter path is the kind of behind-the-scenes change that eventually shows up in a clinic.
The pilots have names, and the selection timetable is set
The two proof-of-concept trials are run by AstraZeneca and Amgen, which the FDA named in its 28 April announcement as partners in studies that report signals to the agency as they occur.
The initiative was announced on 28 April and the request for information took comments until 29 May. The FDA said it intended to publish final selection criteria in July and complete pilot selections in August, a quick timetable for a regulator.
One design detail sharpens what is actually being tested. Walsh has described the pilot as starting with signal data, and as an exercise in reassessing which data the agency needs in order to make a decision at all. That is a narrower and more interesting question than shortening a timeline. A trial that reports continuously produces far more information than a submission at the end; the pilot is partly about finding out how much of it the FDA should be looking at.
The 20 to 40 per cent estimate is still an estimate
The figure was the agency's own when this was written, and it remains the agency's own. Two trials that began reporting this year will not produce a verdict on trial duration for years, because the thing being measured is how long a trial takes.
Two stories since show what happens when claims outrun the record
Two health-technology stories over the summer suggest the pressure on the evidentiary standard is real, and that it arrives through the announcement rather than through the standard itself.
In July, Johnson & Johnson said the FDA had authorised its OTTAVA surgical robot. On close reading, almost nothing in the announcement could be checked: both performance claims cited unpublished internal documentation, and the FDA database that would confirm the authorisation held no record of it at the time. That is normal at four days out, which is exactly why the announcement arrived before the record did.
In August, Google announced blood-pressure and insulin-resistance trend features for the Pixel Watch 5, described as rolling out soon. Google says the features are not FDA cleared and are intended for general wellness.
Neither is an indictment of real-time reporting. Both illustrate the same mechanism the pilot has to guard against: the speed at which a health claim can be published has already outrun the speed at which anyone outside the company can verify it. Continuously feeding data to a regulator does not make that data public. A faster but equally opaque process would just concentrate more power in fewer, less visible steps.
What travels, and what does not
The Pixel Watch features show the tempo mechanism running in the other direction. A consumer device can offer health trend estimates with no FDA clearance behind them, because a general-wellness trend, unlike a diagnostic reading, does not require it.
Regional regulators import the FDA's tempo more readily than its scrutiny. If real-time reporting shortens approvals and the model travels, the evidentiary standard has to travel with it, and that half is considerably harder to export.
Correction, 18 October 2026: An earlier version gave the date of the FDA announcement as 27 May; it was 28 April, and that announcement already named AstraZeneca and Amgen as the pilot partners. The July and August selection milestones are now described as the FDA's stated plans, an unsourced reference to a same-day cancer approval has been removed, and the description of the Pixel Watch 5 features has been corrected.
View author profile → · Editorial policy
