Statistical Calculators › Group Sequential, Adaptive, and Interim Analysis › Sample Size Re-estimation (Blinded)
← All Calculators

Group Sequential, Adaptive, and Interim Analysis

Sample Size Re-estimation (Blinded)

Re-estimate the required sample size for a two-sample comparison of means using a blinded internal-pilot variance estimate. Treatment assignment remains concealed: the interim variance is estimated from the pooled observations and the prespecified treatment effect is retained.

Trial Design & Interim Data

Equal allocation, common variance, two-sided two-sample t-test framework. Enter the blinded pooled variance observed at the interim analysis.

Re-estimated Sample Size

The final recommendation is constrained to the entered minimum and maximum total sample sizes and cannot be smaller than the interim sample.
Enter the interim variance and trial assumptions, then click Re-estimate Sample Size.

Methodology

This calculator implements the blinded internal-pilot approach for a two-sample comparison of means with equal allocation and a common variance. identifies this procedure as its MTT24 table, “Blinded Internal Pilot Sample Size Re-estimation for Two Sample t-test for Inequality.” own materials describe blinded sample-size re-estimation as an adaptive procedure in which nuisance parameters such as variance can be reassessed without unblinding treatment assignment.

1. Blinded interim variance

At the interim analysis, treatment labels are ignored and the variance is estimated from the pooled observations. Let N1 be the total number of observations available at the interim analysis and let SOS2 denote the resulting one-sample pooled variance.

SOS2 = 1 / (N1 − 1) × Σ(Yi − Ȳ)2

Because pooling two treatment groups also captures part of the treatment difference, the simple blinded variance estimator can be adjusted using the prespecified treatment difference δ.

2. Treatment-effect-adjusted variance

For equal allocation, the adjusted one-sample variance estimator used here is:

Sadj2 = [ (N1 − 1)SOS2 − N1δ2/4 ] / (N1 − 2)

If the adjustment is disabled, the calculator uses SOS2 directly. The adjusted form follows the blinded variance-reestimation approach associated with Gould and Shih and the subsequent treatment of simple blinded sample-size adjustment by Kieser and Friede.

3. Re-estimated sample size

For a two-sided comparison of two equally sized groups, the large-sample sample-size equation is applied using the updated variance:

nper group = 2 Sadj2 (z1−α/2 + z1−β)2 / δ2

Nraw = 2 nper group

The calculated total is rounded upward to an even integer so that the two treatment groups remain equally sized. The final value is then constrained to be at least the interim sample size and at least the specified minimum, but no greater than the specified maximum.

Nfinal = max{Ninterim, Nmin, min(Nmax, ceileven(Nraw))}

Important: Blinded sample-size re-estimation addresses uncertainty in nuisance parameters such as variance; it does not use an interim estimate of the treatment effect to make an efficacy decision. The internal-pilot literature distinguishes this from unblinded promising-zone or conditional-power sample-size re-estimation.

Worked validation example

A two-group study begins with 43 participants per group (86 total). At the blinded interim analysis, 22 participants per group (44 total) have provided the primary endpoint. Suppose the prespecified effect is δ = 1.00, the blinded pooled interim variance is SOS2 = 3.00, two-sided α = 0.05, and target power is 90%. Set the allowable total sample-size range to 44 through 172.

Sadj2 = [43(3.00) − 44(1.00)2/4] / 42 = 2.80952381

nper group = 2(2.80952381)(1.959964 + 1.281552)2 = 59.04171

Nraw = 118.08342 → Nfinal = 120

Thus the implementation should return 120 total participants (60 per group) for this validation case.

References

Kieser, M. & Friede, T. (2003). “Simple procedures for blinded sample size adjustment that do not affect the type I error rate.” Statistics in Medicine, 22, 3571–3581. DOI: 10.1002/sim.1585.

Gould, A. L. & Shih, W. J. (1992). Methods for sample size adjustment in clinical trials using blinded interim data. The blinded one-sample variance approach is part of the methodological foundation for internal-pilot sample-size reassessment.

Statsols /. Two Sample Mean Blinded SSR Template and procedure listing for MTT24, “Blinded Internal Pilot Sample Size Re-estimation for Two Sample t-test for Inequality.”

Kairalla, J. A., Coffey, C. S., & Muller, K. E. (2008). “GLUMIP 2.0: SAS/IML Software for Planning Internal Pilots.” Journal of Statistical Software, 28(7).