Group Sequential Design Calculator

A group sequential design lets a trial stop early for efficacy or futility at pre-planned interim analyses while preserving the overall type I error rate. This calculator computes the interim monitoring boundaries, the maximum and expected sample size (or number of events for survival endpoints), and the inflation relative to a fixed design.

How it works

Interim boundaries are derived with the Lan–DeMets alpha-spending framework: you choose how the significance level is spent across looks via a spending function — O'Brien–Fleming (conservative early, near-nominal at the end), Pocock (approximately constant boundaries), or Hwang–Shih–DeCani. The calculator returns per-look Z-boundaries, cumulative alpha spent, and expected sample size under the null and alternative. Boundaries and sample sizes are validated against the gsDesign R package.

When to use it

  • You want the option to stop a superiority (or non-inferiority) trial early for benefit or for futility.
  • Interim analyses are prespecified in the protocol and overseen by an independent data monitoring committee.
  • You are designing with a continuous endpoint (standardized effect) or a time-to-event endpoint (hazard ratio).

Assumptions & limitations

  • Boundaries assume asymptotic normality of the test statistic with independent increments of the information process; strong departures (e.g., markedly time-varying effects) can distort error rates.
  • Futility boundaries here are non-binding — type I error is computed assuming a futility signal may be ignored; the tool does not design binding futility stopping.
  • For survival endpoints the event-to-patient conversion assumes proportional hazards and exponential survival with uniform accrual; non-proportional hazards should be checked separately.
  • Interim looks are supported from 2 to 5; the validity of the design depends on prespecifying the number and timing of looks and the spending function before the trial begins.

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

Loading...