Skip to main content
. 2023 Aug 25;24:556. doi: 10.1186/s13063-023-07586-5

Table 2.

Choice of non-inferiority margin and significance level for the analysis of D3 based on observed confirmed viral rebound risk using the Smooth Away From Expected (SAFE) frontier and the “modify significance level if change non-inferiority margin” analysis method. The power, type 1 error and probability of changing the margin are shown for each true value of control event risk

Control event risk (P0)a 1% 2% 3% 4% 5% 6% 7% 8% 9% 10% 11% 12% 13% 14% 15%
NI margin 5.0% 5.8% 6.5% 7.3% 8.0% 8.9% 9.5% 9.9% 10.0% 10.0% 10.0% 10.0% 10.0% 10.0% 10.0%
Significance level 0.50% 0.50% 0.50% 0.50% 0.50% 0.50% 0.50% 0.50% 2.50% 2.50% 2.50% 2.50% 2.50% 2.50% 2.50%
Power 91.8% 83.8% 79.0% 75.9% 74.2% 74.1% 74.9% 75.9% 77.2% 78.6% 79.3% 78.8% 77.3% 75.2% 73.0%
Type 1 error 3.06% 2.86% 2.69% 2.72% 2.77% 2.65% 2.62% 2.52% 2.52% 2.58% 2.58% 2.58% 2.57% 2.56% 2.56%
P (change margin)b 100% 100% 100% 99.9% 98.8% 95.0% 86.0% 71.4% 53.6% 36.2% 22.0% 12.1% 6.1% 2.8% 1.2%

The column in bold corresponds to the sample size calculation assumption made at the design stage

aThe choice of non-inferiority margin and significance level will depend on the observed confirmed viral rebound risk. The power, type 1 error and probability of changing the margin depend on the true control event risk

bThe probability of changing the margin is the probability that, for a given true control event risk, the observed control event risk will be lower than 9%, hence leading to using a non-inferiority margin in the analysis different from the originally planned 10%