Table III.
Robustness simulation study results to compare Bayesian Doubly Optimal Group Sequential (BDOGS) with HSD optimal procedure when the assumption of proportional hazards is not met in a clinical trial with five analyses. The true median TTF is 12 months for A.
| True hazard |
Method | False pos. |
Power | Sample size, δ =0 (δ=3)
|
|||||
|---|---|---|---|---|---|---|---|---|---|
| Mean | 2.5 per cent | 25 per cent | 50 per cent | 75 per cent | 97.5 per cent | ||||
| Proportional hazards assumption met | |||||||||
| Exp | BDOGS | 0.05 | 0.80 | 639 (650) | 505 (409) | 638 (675) | 675 (680) | 680 (680) | 680 (680) |
| HSD | 0.05 | 0.80 | 611 (627) | 383 (398) | 537 (566) | 656 (680) | 680 (680) | 680 (680) | |
| Robustness study—proportional hazards assumption not met | |||||||||
| WI | BDOGS | 0.03 | 0.90 | 375 (397) | 346 (365) | 363 (381) | 371 (390) | 381 (401) | 483 (515) |
| HSD | 0.05 | 0.99 | 584 (473) | 363 (369) | 498 (390) | 610 (427) | 680 (522) | 680 (680) | |
| PI | BDOGS | 0.01 | 0.84 | 357 (388) | 334 (350) | 349 (365) | 357 (377) | 364 (385) | 380 (503) |
| HSD | 0.05 | 0.99 | 572 (389) | 346 (352) | 481 (368) | 592 (377) | 680 (388) | 680 (506) | |
| VS | BDOGS | 0.05 | 0.98 | 384 (380) | 335 (351) | 351 (368) | 362 (377) | 382 (387) | 488 (492) |
| HSD | 0.04 | 0.99 | 578 (429) | 350 (354) | 486 (373) | 594 (388) | 680 (490) | 680 (624) | |
| LN2 | BDOGS | 0.05 | 0.36 | 455 (501) | 381 (395) | 402 (418) | 415 (436) | 445 (591) | 680 (680) |
| HSD | 0.05 | 0.36 | 629 (646) | 398 (413) | 579 (680) | 680 (680) | 680 (680) | 680 (680) | |
| WD | BDOGS | 0.04 | 0.25 | 406 (460) | 373 (388) | 391 (408) | 401 (421) | 411 (535) | 542 (680) |
| HSD | 0.05 | 0.20 | 629 (648) | 394 (409) | 572 (680) | 680 (680) | 680 (680) | 680 (680) | |
| PD | BDOGS | 0.05 | 0.27 | 473 (436) | 398 (385) | 420 (404) | 433 (415) | 543 (431) | 680 (590) |
| HSD | 0.05 | 0.36 | 637 (645) | 417 (403) | 593 (680) | 680 (680) | 680 (680) | 680 (680) | |