Platform Trial Design Calculator
Designs a multi-arm, multi-stage (MAMS) platform trial in which multiple treatments are compared against a shared control, arms can enter and leave over time, and interim stages allow early stopping. Supports binary, continuous, and survival endpoints with frequentist or Bayesian analysis, and estimates operating characteristics by simulation.
How it works
Treatments share a common control and are evaluated across interim stages, with rules for dropping arms for futility and graduating arms for efficacy; new arms can be added over time. The tool handles how control data are used across arms — concurrent versus non-concurrent controls — and reports per-arm power and the trial-wide error behavior under the chosen configuration.
When to use it
- You are running several treatments against a shared control under one protocol, with arms entering and exiting over time.
- You want interim stages for early efficacy or futility decisions in a multi-arm setting.
- You need per-arm operating characteristics and a defined strategy for control-data handling and multiplicity.
Assumptions & limitations
- To preserve valid randomized comparisons, the primary analysis for an arm should generally use concurrently eligible controls; using non-concurrent control data requires justification, adjustment for temporal trends, and sensitivity analyses, and is best agreed with regulators in advance.
- Because arms are correlated through the shared control and evaluated over time, it is the distribution of the number of false positives — not only a single family-wise rate — that should be characterized by simulation; strong control across different drugs is generally not required by default, while control within a drug (across doses or endpoints) still is.
- Adding or dropping arms and changing the control or background therapy over a long-running trial can affect inference and should be pre-specified.
- Reported operating characteristics are simulation estimates for the modeled scenario, not guarantees for a realized trial.
For the full methodology, derivation, and worked examples, read the complete guide.