Table 2.
Proportion of trials showing significant improvement for “stopping of recruitment” cases
| Stopping of recruitment | Design | pC = 0.50 | pC = 0.75 | |||||
|---|---|---|---|---|---|---|---|---|
| ∆ = 0.20 | ∆ = 0 | ∆ = − 0.10 | ∆ = 0.20 | ∆ = 0 | ∆ = − 0.10 | |||
| NSTOP = 20 | F1 | 0.416 | 0.021 | 0.002 | 1 | 0.618 | 0.244 | |
| S1 | (a) | 0.414 (0.75) | 0.021 (0.587) | 0.002 (0.245) | 1 (0.003) | 0.615 (0.585) | 0.242 (0.830) | |
| (b) | 0.234 + 0.180 | 0.006 + 0.015 | 0.0004 + 0.0012 | 0.997 + 0.002 | 0.411 + 0.204 | 0.116 + 0.126 | ||
| F2 | 0.054 | 0.006 | 0.002 | 0.048 | 0.005 | 0.001 | ||
| S2 | (a) | 0.118 (1) | 0.042 (1) | 0.060 (1) | 0.054 (1) | 0.019 (1) | 0.044 (1) | |
| (b) | 0 + 0.118 | 0 + 0.042 | 0 + 0.060 | 0 + 0.054 | 0 + 0.019 | 0 + 0.044 | ||
| NSTOP = 50 | F1 | 0.858 | 0.032 | 0.0005 | 1 | 0.972 | 0.619 | |
| S1 | (a) | 0.793 (0.136) | 0.023 (0.020) | 0.0004 (0.0002) | 1 (0) | 0.948 (0.054) | 0.526 (0.181) | |
| (b) | 0.720 + 0.073 | 0.016 + 0.007 | 0.0003 + 0.0001 | 1 + 0 | 0.916 + 0.032 | 0.438 + 0.088 | ||
| F2 | 0.333 | 0.032 | 0.005 | 0.320 | 0.021 | 0.003 | ||
| S2 | (a) | 0.234 (0.987) | 0.016 (0.950) | 0.003 (0.837) | 0.420 (0.998) | 0.020 (0.989) | 0.017 (0.903) | |
| (b) | 0.013 + 0.221 | 0.0001 + 0.016 | 0 + 0.003 | 0.002 + 0.418 | 0 + 0.020 | 0 + 0;017 | ||
| NSTOP = 100 | F1 | 0.988 | 0.028 | 0 | 1 | 1 | 0.875 | |
| S1 | 0.894 (0) | 0.025 (0) | 0.0005 (0) | 1 (0) | 0.981 (0) | 0.661 (0) | ||
| F2 | 0.542 | 0.029 | 0.002 | 0.834 | 0.024 | 0.0008 | ||
| S2 | (a) | 0.425 (0.732) | 0.012 (0.529) | 0.0004 (0.190) | 0.747 (0.765) | 0.012 (0.766) | 0.0004 (0.287) | |
| (b) | 0.255 + 0.170 | 0.005 + 0.007 | 0.0001 + 0.0003 | 0.236 + 0.511 | 0.001 + 0.011 | 0 + 0.0004 | ||
In bold, type-I-errors maintained in [0.024;0.026]
Abbreviations: pC indicates control survival rate; ∆: simulated survival rate difference; F1: fixed single-arm design; S1: group-sequential single-arm design; F2: fixed two-arm design; S2: group-sequential two-arm design; Stopping of recruitment: specific case with a fixed control survival rate and an early stop of the trial due to an outbreak end; For group-sequential trials, the total of trials demonstrating an efficacy of the experimental treatment was defined by the sum of trials with a significant test and trials with an adjusted p-value inferior to 0.025. The p-value was adjusted for underrunning using the method proposed by Whitehead [18]
(a): Proportion of significant tests with proportion of inconclusive group-sequential trials for “changing with time” cases
(b): Proportion of significant tests + proportion of trials with adjusted p-values inferior to 0.025 for group-sequential trials and “stopping of recruitment” cases