Unblinded Sample Size Re-estimation Calculator

Re-estimates the sample size at an interim analysis using the unblinded interim treatment effect, within a promising-zone framework, and combines the two stages with an inverse-normal combination test so that the final decision controls the type I error regardless of how the sample size was changed.

How it works

At the interim, the observed conditional power places the trial in a futility, unfavorable, promising, or favorable zone (Mehta–Pocock). In the promising zone the sample size is increased (up to a cap) based on the observed effect. The final test uses the inverse-normal (weighted-Z) combination test of Lehmacher & Wassmer / Müller & Schäfer, with pre-specified stage weights, so the two stages are combined validly. Continuous, binary, and time-to-event endpoints are supported.

When to use it

  • You want to adapt the sample size based on the observed interim effect, not just a nuisance parameter.
  • You need the final analysis to control the type I error despite a data-dependent sample-size change.
  • You are using a single interim analysis with a pre-specified information fraction.

Assumptions & limitations

  • Type I error is controlled at the nominal level only when the combination test is used with pre-specified, fixed stage weights; disabling the combination test (naive re-estimation) can inflate the type I error, and the analysis is then exploratory.
  • The combination-test guarantee relies on non-overlapping stage data and on the interim information fraction matching what was pre-planned.
  • Re-estimation acts only within the promising zone and is subject to the maximum-inflation cap.
  • The simulation tab reports the achieved type I error and power as estimates for the modeled scenario; there can be a small power cost relative to a fixed design when no increase is triggered.

For the full methodology, derivation, and worked examples, see the guide for your design:

Loading...