Anatomy of a Platform Trial

One Trial, Changing Treatment Arms

A conventional picture of a clinical trial has a fixed beginning, a fixed set of treatment groups, and a fixed end to recruitment. A platform trial has a different organizing idea: a trial-wide framework can continue while the treatments being evaluated change. Some arms may complete recruitment while other arms remain active, and additional arms may be considered under the same overarching protocol.

The important distinction is between the life of the trial as an organized study and the recruitment period of a particular treatment arm. These periods need not start or finish together. Making the distinction explicit helps describe what the trial is, what its protocol governs, and how its timeline should be reported. It does not, by itself, establish that any treatment comparison is unbiased or that results from different periods can be combined.

“Continuing” describes the organizing framework, not uninterrupted recruitment or an obligation to run indefinitely. A platform trial may have pauses, changes to its protocol, periods with different numbers of active arms, and a planned or eventual trial-wide closeout. Its defining feature is not simply that a study has several treatments. It is that an overarching protocol and infrastructure support evaluation of changing treatment arms over time.

What the Master Protocol and Infrastructure Do

The master protocol provides the common structure within which arms are evaluated. Depending on the trial, it can specify trial-wide procedures for matters such as governance, eligibility framework, outcome measurement, safety oversight, data management, and the statistical plan. Arm-specific details can supplement that structure—for example, a particular treatment regimen or a relevant comparison—without making every arm a wholly separate trial.

The infrastructure is the practical organization that carries the framework forward. It may include coordinating and data centers, participating sites, shared operating procedures, oversight committees, and systems for maintaining trial records. Infrastructure can persist while the arm configuration changes. Conversely, using the same sites or data system does not alone prove that several studies form one platform trial: the relationship to the master protocol and the continuing trial-wide organization matters.

The master protocol is not necessarily one unchanging document. Amendments can update the framework, and an arm can have additional protocol material. What matters for a structural description is that the trial-wide framework governs how the arms belong to the organized trial. The protocol documents should make that relationship understandable; a label such as “platform” is not a substitute for examining it.

Representing the Two Levels of Time

To describe the timelines precisely, first state what counts as the operational life of the overall trial. For illustration, let \(T_0\) be the date trial-wide operations begin and \(T_1\) the date trial-wide closeout is complete. Those boundaries should be defined for the description at hand: protocol approval, first site activation, first participant recruitment, and final closeout are different possible milestones, not interchangeable dates.

For arm \(j\), let \(a_j\) be the beginning of that arm’s recruitment period and \(b_j\) the end of that period. Write the overall operational interval as \(I_T=[T_0,T_1)\), and the arm-specific recruitment interval as \(I_j=[a_j,b_j)\). The half-open interval convention includes its start and excludes its end, which makes adjacent periods straightforward to describe. In the usual case where recruitment occurs within the operating trial, \(I_j\subseteq I_T\).

Here \(D_T\) is the duration of the defined trial-wide operational period, and \(d_j\) is the recruitment duration for arm \(j\). Because the arm interval lies within the trial interval, \(d_j\leq D_T\). But this containment does not imply that the arm starts when the trial starts or that its recruitment ends when the trial closes. Participants already recruited into an arm may also continue in follow-up after recruitment to that arm has ended.

The relationship can be justified directly from the interval boundaries. Containment gives \(T_0\leq a_j\) and \(b_j\leq T_1\). Therefore,

The sum of recruitment durations across arms is not generally the overall trial duration. Overlapping arm periods count the same calendar time more than once in the sum; gaps between recruitment periods count in the trial duration if the trial remains operational. Trial-wide preparation before first recruitment and follow-up or closeout after the last recruitment can also contribute to its life. Thus neither the sum of arm durations nor the span of one arm should be reported as the trial’s duration without checking the stated boundaries.

Reading a Timeline Without Confusing Its Units

A useful timeline distinguishes at least two things: the overall trial-wide period and each arm’s recruitment interval. If follow-up continues after recruitment ends, showing it separately can prevent another confusion: an arm’s recruitment endpoint is not necessarily the endpoint of participant follow-up, and neither automatically defines trial-wide closeout.

The following hypothetical example shows why adding arm durations does not produce the trial duration. All dates are chosen for illustration.

If a timeline gives only the trial’s opening and closing dates, it may conceal when individual arms recruited. If it gives only arm dates, it may omit the trial-wide work before, between, or after those periods. The goal is not to impose a particular timeline format, but to make the unit and boundaries of each reported duration explicit.

How the Structure Compares With Alternatives

A fixed multi-arm trial can evaluate several treatments under one protocol while retaining a planned set of arms over its design. A platform trial also uses a trial-wide framework, but is organized to accommodate changing arms over time. In a collection of separate trials, each study has its own protocol and trial organization, even if the studies happen to share sites, investigators, or a data system.

These descriptions concern structure, not the quality of a treatment comparison. Calling a study a platform does not show that its arms recruited at the same time, shared a control group, or support a particular pooled analysis. Those questions require details about the design and data. The structural description should not be mistaken for a statistical conclusion.

Common Errors and Useful Checks

A clear description can be checked by asking whether the trial-wide and arm-specific timelines have been named separately, whether the start and end milestones are stated, and whether recruitment is being confused with follow-up or closeout. The checks below address structural description, not the validity of a treatment-effect estimate.

• Equating an arm with the whole trial. An arm can recruit for only part of the overall trial period. Report its recruitment interval as arm-specific.
• Adding arm durations and calling the result the trial duration. Overlapping periods are counted multiple times in that sum. State that the sum is arm-months, or calculate the trial-wide duration from trial-wide boundaries.
• Assuming that shared infrastructure alone establishes one platform trial. Shared sites or data systems are not enough; describe how the arms are governed by the master protocol and continuing trial organization.
• Treating “continuing” as “unchanged” or “uninterrupted.” A platform may be amended, pause recruitment, or change its arms while retaining its organizing framework.
• Using “last recruitment” as a synonym for “trial closeout.” Recruitment may end before follow-up, data work, or other trial-wide operations are complete. Specify the milestone used for the overall endpoint.

A practical diagnostic is to draw a single line for the operational life of the trial and a separate line for every arm’s recruitment. If an arm appears to begin before the declared trial start, or its recruitment ends after the declared trial closeout, then either the stated boundaries or the timeline needs clarification. Such a discrepancy is a prompt to check definitions and records, not by itself proof that the trial is or is not a platform.