Blinded Sample Size Re-estimation Calculator

Re-estimates the required sample size at an interim look using only the blinded (pooled) data — the nuisance parameter such as the outcome variance or overall event rate — while keeping the assumed treatment effect fixed. Because treatment assignments are never revealed, the design does not use interim treatment-effect information.

How it works

Using the Kieser–Friede approach, the tool estimates the nuisance parameter from the pooled interim data (a lumped variance for continuous endpoints, the pooled rate for binary, the blinded event probability for survival) and re-derives the sample size needed to keep the planned power for the pre-specified effect. The re-estimated size is capped by a maximum-inflation factor. The tool provides closed-form analytical results, with an optional simulation to assess achieved type I error and power under the modeled scenario.

When to use it

  • Your original sample size depended on an uncertain nuisance parameter (variance, control rate, or event rate) that you want to update at an interim.
  • You want to preserve blinding and avoid using interim treatment-effect estimates.
  • You need to keep the design's power robust to a misspecified nuisance parameter.

Assumptions & limitations

  • Type I error preservation for blinded re-estimation holds because only the blinded nuisance parameter is used — it depends on maintaining blinding and not incorporating any treatment-effect information at the interim.
  • The blinded (pooled) variance is slightly positively biased when a true effect is present, which can make the re-estimated sample size conservative.
  • Only the nuisance parameter is updated; a misspecified treatment effect is not corrected by blinded re-estimation.
  • The simulation option assesses achieved behavior for the modeled scenario rather than establishing it in general.

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

Loading...