Traditional group sequential design | Numbers (n = 43) | Percentages (76.8%) | |
---|---|---|---|
How was interim analyses planned | Pre-planned | 39 | 90.7% |
Pre-planned by DSMB | 4 | 9.3% | |
Frequency of interim analyses planned | N = 1 | 9 | 20.9% |
N = 2 | 16 | 37.2% | |
N > 2 | 18 | 41.9% | |
How was the interim analysis set | Time | 15 | 34.9% |
Patients | 5 | 11.6% | |
Events | 20 | 46.5% | |
NA | 3 | 7% | |
Pre-planned decision rule | The O’Brien-Fleming group sequential boundaries | 3 | 7% |
Haybittle and Peto type of stopping rule | 7 | 16.3% | |
An alpha spending function (Lan and DeMets): the O’Brien- Fleming-type | 10 | 23.3% | |
Triangular sequential design (A Wang-Tsiatis group sequential design) | 2 | 4.7% | |
Asymmetrical group sequential (a Peto-type boundary) | 1 | 2.3% | |
Others | 9 | 20.9% | |
NA | 11 | 25.6% | |
Was the trial stopped early | Yes | 29 | 67.4% |
No | 14 | 32.6% | |
Reason for stopping early | Futility | 5 | 11.6% |
Efficacy | 14 | 32.6% | |
Safety | 9 | 20.9% | |
Others | 1 | 2.3% | |
Adaptive designs | Numbers (n = 13) | Percentages (23.2%) | |
How was interim analyses planned | Pre-planned | 13 | 100% |
Frequency of interim analyses planned | N = 1 | 9 | 69.2% |
N = 2 | 2 | 15.4% | |
N > 2 | 2 | 15.4% | |
How was the interim analysis set | Time | 0 | 0 |
Patients | 10 | 76.9% | |
Events | 2 | 15.4% | |
NA | 1 | 7.7% | |
Was the trial stopped early | Yes | 7 | 53.8% |
No | 6 | 46.2% | |
Reason for stopping early | Futility | 4 | 30.8% |
Efficacy | 2 | 15.4% | |
Safety | - | - | |
Others (slow enrollment) | 1 | 7.7% | |
Type | Methods | ||
Re-estimation of sample size | Bayesian adaptive approach | 2 | 15.7% |
Simulation | 2 | 15.7% | |
Others | 2 | 15.7% | |
NA | 2 | 15.7% | |
Response-adaptive randomization | Bayesian approach | 1 | 7.7% |
Dose-Response | Bayesian method | 1 | 7.7% |
Enrichment | Bayesian method | 1 | 7.7% |
Two-stage adaptive | Simulation used to assess the type I error. A group sequential approach with efficacy boundary based on a gamma (− 10) α spending function. | 1 | 7.7% |
Seamless phase IIb/III | Simulation used to assess the type I error. A truncated Levin-Robbins sequential elimination procedure for selecting a right dose. | 1 | 7.7% |