Single-Arm Sample Size Re-estimation Calculator

Re-estimates the sample size at an interim look for a single-arm Phase II trial with a binary endpoint (objective response rate versus a fixed historical benchmark), using either a Bayesian decision rule or a conditional-power promising-zone rule.

How it works

In Bayesian mode the interim efficacy decision uses the posterior probability that the response rate exceeds the historical benchmark, futility uses the predictive probability of eventual success, and final success is judged against a separate final threshold; a Beta–Binomial conjugate model is used. In conditional-power mode, the trial re-estimates sample size in a Mehta–Pocock promising zone. Both modes apply a sample-size cap. The Bayesian final threshold can be recalibrated by simulation to target a specified error rate.

When to use it

  • You are running a single-arm Phase II study with a binary response endpoint and a reliable historical control rate.
  • You want to expand the sample size when interim results are promising, within a cap.
  • You want a Bayesian decision rule (posterior + predictive probability) or a conditional-power rule, with simulated operating characteristics.

Assumptions & limitations

  • The Bayesian thresholds are posterior/predictive probabilities and the conditional-power threshold is not analytically tied to the frequentist type I error; the achieved error rate must be checked by simulation at the null response rate.
  • In this single-arm setting the conditional-power rule is not analytically calibratable and can leave the type I error above the nominal level; the Bayesian predictive rule can be calibrated by adjusting the final threshold via simulation.
  • Conclusions depend on the historical benchmark being stable and comparable — a single-arm design has no concurrent control.
  • Conditional power, predictive probability, and posterior probability answer different questions and should not be interpreted interchangeably.

For the full methodology, derivation, and worked examples, read the complete guide.

Loading...