Adaptive and Platform Trials

Why trial design matters

Clinical trials test treatments in people. A classic trial fixes its plan at the start. Newer designs can change in set ways while the trial runs. This page explains adaptive trials, platform trials and master protocols. It is information only, not advice about joining a trial. See also clinical-trial phases explained.

Adaptive designs

Adaptive designs use results accumulating in the trial to modify its course according to pre-specified rules. [1] The FDA has issued guidance for sponsors on the appropriate use of adaptive designs, including principles for designing, conducting and reporting them, and including Bayesian adaptive trials and complex trials that rely on computer simulations. [2] The page I read was current as of November 2019. The key point is "pre-specified": the rules are written before the data are seen. Reviewers note that the term "adaptive design" has sometimes been used misleadingly as an all-encompassing label. [1]

Platform trials and master protocols

An adaptive platform trial can study multiple interventions in a disease or condition over a long period, with interventions entering and leaving the platform under a predefined decision algorithm. [3] In oncology, FDA guidance on master protocols describes trials that simultaneously evaluate more than one investigational drug and/or more than one cancer type within the same overall trial structure. [4] The oncology guidance page was current as of March 2022.

An example: RECOVERY

The RECOVERY trial was a controlled, open-label trial that compared a range of possible treatments in people hospitalized with Covid-19. [5] For one treatment, dexamethasone, 482 of 2,104 people (22.9%) in the dexamethasone group died within 28 days compared with 1,110 of 4,321 (25.7%) on usual care (age-adjusted rate ratio 0.83; 95% confidence interval 0.75 to 0.93). [5] The benefit differed by the level of breathing support: lower death among those on a ventilator or on oxygen, but not among those who needed no respiratory support. [5] This is a large randomized result for the groups studied, and I read only the main report, so this page does not rate it further. It shows how a shared design can answer a specific question well. It does not mean the same treatment helps other conditions.

What these designs can and cannot establish

Trials with an adaptive design are often described as more efficient, informative and ethical, but experts stress that investigators must understand what can and cannot be inferred from the results and how to report them. [1] Being adaptive is not a mark of quality by itself. A platform trial that tests wellness products without a clear outcome, control group or independent oversight would still be weak evidence. Phase labels are also still relevant: Phase 2 studies are not large enough to show benefit on their own. [6] See registered trials vs published results.

How to read a claim about an adaptive trial

• Was the plan registered before the data were seen?
• What exactly was measured: illness outcomes, or a lab marker?
• Who ran and paid for it?
• Did results come from the trial's final analysis or an early look?