Table 2 Operating characteristics for the proposed Bayesian sequential designs for the OSCAR triala
From: Using Bayesian adaptive designs to improve phase III trials: a respiratory care example
Design | Scenario: control vs HFOV primary outcome rate | Average duration (weeks) | Average sample size (SD) | Proportion stopped early for success | Overall Proportion Successful b | Proportion stopped early for futility |
|---|---|---|---|---|---|---|
Design 1: Fixed design | No difference: 45% vs 45% | 196 | 1006 (0) | NA | 0.0283 | NA |
Target difference: 45% vs 36% | 196 | 1006 (0) | NA | 0.8219 | NA | |
Small difference: 45% vs 40% | 196 | 1006 (0) | NA | 0.3503 | NA | |
Large difference: 45% vs 30% | 196 | 1006 (0) | NA | 0.9979 | NA | |
Treatment harmful: 45% vs 50% | 196 | 1006 (0) | NA | 0.0003 | NA | |
Design 2: Interim analysis at 250, 500 and 750 patients | No difference: 45% vs 45% | 103 | 519 (236) | 0.0123 | 0.0268 | 0.8956 |
Target difference: 45% vs 36% | 145 | 730 (227) | 0.5319 | 0.7793 | 0.1434 | |
Small difference: 45% vs 40% | 141 | 719 (260) | 0.1607 | 0.3302 | 0.4828 | |
Large difference: 45% vs 30% | 114 | 560 (127) | 0.9592 | 0.9932 | 0.0058 | |
Treatment harmful: 45% vs 50% | 75 | 367 (155) | 0.0004 | 0.0004 | 0.9949 | |
Design 3: Interim analysis at 335 and 670 patients | No difference: 45% vs 45% | 130 | 664 (207) | 0.0070 | 0.0264 | 0.8123 |
Target difference: 45% vs 36% | 163 | 828 (179) | 0.4314 | 0.8079 | 0.0829 | |
Small difference: 45% vs 40% | 161 | 825 (207) | 0.1151 | 0.3431 | 0.3593 | |
Large difference: 45% vs 30% | 139 | 696 (90) | 0.9214 | 0.9971 | 0.0019 | |
Treatment harmful: 45% vs 50% | 109 | 555 (170) | 0.0000 | 0.0005 | 0.9841 | |
Design 4: Interim analysis at 335, 500 and 670 patients | No difference: 45% vs 45% | 112 | 564 (202) | 0.0087 | 0.0229 | 0.8578 |
Target difference: 45% vs 36% | 147 | 741 (230) | 0.4723 | 0.7873 | 0.1227 | |
Small difference: 45% vs 40% | 146 | 742 (242) | 0.1362 | 0.3334 | 0.4371 | |
Large difference: 45% vs 30% | 114 | 557 (131) | 0.9334 | 0.9958 | 0.0033 | |
Treatment harmful: 45% vs 50% | 91 | 454 (108) | 0.0002 | 0.0003 | 0.9895 | |
Design 5: Interim analysis at 503 and 755 patients | No difference: 45% vs 45% | 138 | 712 (158) | 0.0099 | 0.0249 | 0.8637 |
Target difference: 45% vs 36% | 160 | 812 (176) | 0.5237 | 0.813 | 0.0922 | |
Small difference: 45% vs 40% | 163 | 834 (174) | 0.1501 | 0.3426 | 0.4062 | |
Large difference: 45% vs 30% | 134 | 667 (139) | 0.9596 | 0.9968 | 0.0015 | |
Treatment harmful: 45% vs 50% | 126 | 645 (129) | 0.0002 | 0.0005 | 0.9915 | |
Design 6: Interim analysis at 503, 755 and 880 patients | No difference: 45% vs 45% | 136 | 702 (143) | 0.0130 | 0.0270 | 0.9381 |
Target difference: 45% vs 36% | 158 | 792 (159) | 0.6320 | 0.7990 | 0.1415 | |
Small difference: 45% vs 40% | 158 | 810 (156) | 0.2035 | 0.3376 | 0.5420 | |
Large difference: 45% vs 30% | 134 | 664 (132) | 0.9847 | 0.9966 | 0.0027 | |
Treatment harmful: 45% vs 50% | 127 | 644 (127) | 0.0002 | 0.0005 | 0.9972 |