Skip to content
Omnikalc

Group sequential boundaries calculator

Choose the design, the number of planned looks and the current look to get the z boundary and the nominal p-value for an interim analysis at an overall two-sided 5 % level.

Updated on
Your inputs

Value out of range

Value out of range

Results

Boundary at the current look |z|

3.471

Nominal two-sided p-value at this look
0.00052
Information fraction
33.3 %
Boundary at the first look
3.471
Boundary at the final look
2.004
The final look is slightly stricter than a single test at 1.96
Nominal p-value at the final look
0.04507
Overall false positive risk if each look were tested at 5 %
10.7 %
Testing at 5 % at every look inflates the false positive rate

The group sequential boundaries calculator gives the z boundary and the nominal p-value for an interim analysis with an O’Brien-Fleming or Pocock design at an overall two-sided 5 % level.

How the calculation works

  • Choose the design and the number of planned looks.
  • Enter the current look.
  • The calculator gives the boundary and nominal p-value at this look, the information fraction and the first and final boundaries.

Example

A trial plans 3 looks with an O’Brien-Fleming design, whose constant is 2.004. At the first look, the boundary is 2.004 × √(3/1) = 3.471, with a nominal p-value of 0.00052. At the second look it is 2.004 × √(3/2) = 2.454, and at the final look 2.004, a nominal p-value of 0.04507.

Why it matters

The calculator helps clinical trial statisticians, members of data monitoring committees, and students who learn why interim analyses need adjusted thresholds.

Good to know

The constants are published values for equally spaced looks and a two-sided overall 5 % level; other levels or spacing need software. The boundaries apply to a normal test statistic z computed from all the data accumulated up to the look. Stopping for benefit should rely on more than one boundary crossing: safety data, consistency and the size of the effect also count. A trial that stops early tends to overestimate the effect, so the estimate should be adjusted. The risk shown for naive testing at each look comes from the classic calculations of Armitage and colleagues. Decide the plan before looking at any data and write it into the protocol, along with who sees the interim results and who makes the decision to stop.

The statistical power calculator links effect, sample and power, the SPRT calculator gives a sequential test for defects, the p-value calculator converts z to p and the Bonferroni correction calculator handles multiple tests.

Frequently asked questions

What are group sequential boundaries?

When a trial is analyzed several times before the end, testing at 5 % each time would raise the overall false positive rate well above 5 %. Group sequential designs set stricter thresholds at each look so that the overall error stays at 5 %.

O’Brien-Fleming or Pocock?

O’Brien-Fleming boundaries are very strict early and relax toward the end, so the final analysis is close to a usual test. Pocock boundaries are constant, which makes early stopping easier but the final threshold stricter. O’Brien-Fleming is the most used.

Are the looks equally spaced?

This calculator assumes equal increments of information between looks. With unequal spacing, an alpha-spending function is used instead. Plan the number and timing of looks before the trial begins.

Was this calculator helpful?

Thanks for your feedback!

Results are estimates provided for information purposes. They do not replace professional advice.