Ask Mallard
Study design

Platform trial

4 min read

Use an ongoing trial structure to test interventions as arms enter or leave under rules written before results are known.

NHMRC EVIDENCE LEVEL · 2009Level II · InterventionProvisional design classification. Study quality and certainty of findings require separate appraisal.
Why this level?Participants or clusters are randomly allocated to concurrent comparison groups. NHMRC 2009 source

The version taught here uses random allocation and concurrent controls for treatment comparisons, matching NHMRC intervention level II. Platform is an infrastructure label; other allocation choices require their own classification.

Fit

When to use it, and when not to

Use it when

  • You expect several interventions to need testing over time.
  • You can maintain common governance, outcomes and a planned route for changing arms.

Not when

  • You need one fixed comparison with no changing arms. Use a parallel randomized trial. Read: Randomized controlled trial
  • All arms are fixed from the start. A fixed multi-arm trial may suffice.

Before you start

What you need

Decide these first
  1. The arms and their entry and stopping rulesSpecify eligibility, timing and who can change the protocol.
  2. The control patients for each comparisonDefine which patients were eligible while each arm was open.
  3. The primary outcome and contrastoutcomeTypesummaryMeasureDefine the treatment effect for each comparison.
  4. Design operating characteristicsPlan simulation of error rates, power and sample sizes across plausible time trends and recruitment patterns.
It rests on these
  1. The featured design randomizes treatment within eligible open arms.How to check: Record assignment probabilities and keep allocation concealed.
  2. Control data suit the comparison period.How to check: Prefer concurrent controls. Any use of earlier controls needs explicit assumptions and sensitivity analyses.
  3. Adaptations follow planned rules.How to check: Assess the whole design by simulation, including multiple claims and shared controls.

Plain language

How it works

A platform keeps a common trial structure running as intervention arms change. New arms enter and unsuccessful arms can leave under planned rules. Eligible patients are randomized among the open options in the version taught here.

What to preserve

Each treatment claim needs its own defined control set and outcome. Shared controls can improve efficiency but do not erase time trends. Assess changes in usual care and the joint effect of adaptation and multiple testing before launch. Use CONSORT principles for each randomized comparison. Report changes in arms and control eligibility.

Platform trial in outline
Arms may change over time. Each claim needs a defined control set and planned adaptation rules.Platform trialOpen eligible treatment armsRandomize among current optionsCompare with eligible concurrent controlsAdd or stop arms by planned rulesArms may change over time. Each claim needs a defined control set and planned adaptation rules.
Conceptual diagram. It shows the structure of the idea, not data from any study.

Worked example

A worked example

Invented for illustration
Setting
An ongoing trial adds a new treatment arm.
Question
Does the new treatment reduce the primary event?
Approach
Compare 200 assigned to the new arm with 200 eligible concurrent controls. Earlier controls are excluded from this illustration.
Result
Events occur in 40 of 200 on new care and 50 of 200 controls. The risk difference is -5 per 100 and the risk ratio is 0.8.
What it means
This is one concurrent randomized contrast. It does not settle every other arm or validate the platform's stopping rules.

In practice

Recent studies PubMed search · not vetted

Searching PubMed…

Watch out

Common mistakes

The one to remember

A shared protocol does not make all comparisons exchangeable. Calendar time, eligibility and changes in usual care matter, especially when controls are not concurrent.

  • Using every earlier control as though care had stayed constant.Instead: Define time overlap and assess changing eligibility, prognosis and usual care.
  • Sizing one pair of arms and calling the platform sized.Instead: Simulate the planned entry, stopping and testing rules. Report comparison-specific power and overall error behavior.

Inside Mallard

How Mallard uses this

Plans do not link to this design yet. Mallard has no dedicated sizing form for platform trial.

No Mallard code example points to this page.
Atlas node design.platform_trial, the id a plan cue records.

Related Learn pages

Related designs and methods

Evidence

Sources & further reading

  1. Dexamethasone in Hospitalized Patients with Covid-19 (opens in a new tab) · N Engl J Med · doi 10.1056/NEJMoa2021436
  2. Interleukin-6 Receptor Antagonists in Critically Ill Patients with Covid-19 (opens in a new tab) · N Engl J Med · doi 10.1056/NEJMoa2100433
  3. CONSORT 2010 statement: updated guidelines for reporting parallel group randomised trials (opens in a new tab) · BMJ · doi 10.1136/bmj.c332
Prepared with AI assistance; source titles, authors, years and DOIs checked against PubMed on 2026-11-06. No independent professional approval is claimed. RECOVERY reports a randomized dexamethasone comparison within a continuing treatment program. REMAP-CAP reports a randomized adaptive platform with prespecified decision criteria.
Last source review: 2026-11-06
AI-generated educational content
Mallard uses reputable methodological and reporting sources, but this material may contain errors or omit important context. Verify important methodological decisions with the cited sources and, when appropriate, a qualified research methodologist or statistician.