TABLE 3.
Operating characteristics for the Bayesian adaptive design case study
| Scenario | Proportion of simulations CPAP declared superior to control | Proportion of simulations NIPPV declared superior to control | Proportion of simulations where at least one intervention is declared superior to control a | Proportion of simulations stopped early for success | Proportion of simulations stopped early for futility | Average total sample size (SD) | Average allocations, number of patients (SD) | ||
|---|---|---|---|---|---|---|---|---|---|
| Control | CPAP | NIPPV | |||||||
| (1) Null | 0.0133 | 0.0144 | 0.0258 | 0.0155 | 0.3925 | 1356 (202) | 515 (92) | 421 (123) | 421 (123) |
| (2) One intervention superior | 0.0108 | 0.8897 | 0.8901 | 0.7470 | 0.0018 | 1126 (273) | 458 (126) | 211 (59) | 458 (127) |
| (3) Both interventions superior | 0.6710 | 0.6745 | 0.9260 | 0.8193 | 0.0004 | 1064 (269) | 391 (106) | 336 (105) | 337 (106) |
| (4) NIPPV > CPAP > Control | 0.1700 | 0.8274 | 0.8604 | 0.7166 | 0.0019 | 1140 (278) | 436 (115) | 285 (102) | 421 (115) |
| (5) Both interventions have small improvement | 0.2559 | 0.2519 | 0.4314 | 0.2821 | 0.0267 | 1370 (221) | 517 (93) | 427 (122) | 427 (123) |
| (6) Harm | 0.0005 | 0.0017 | 0.0021 | 0.0012 | 0.7319 | 1197 (242) | 451 (107) | 373 (117) | 374 (117) |
Note: These results are based on 10,000 simulated trials for each scenario. We assume a mean recruitment rate of 6.5 patients/week and that it took 6 months to reach that rate. It was assumed there would be no dropouts.
Abbreviations: CPAP, continuous positive‐pressure ventilation; NIPPV, non‐invasive intermittent positive‐pressure ventilation.
a
Proportion of simulated trials that declared the trial to be ‘successful’, that is, at least one arm superior to the control at the final analysis (includes trials that stopped early and those that recruited to the maximum sample size). The simulated type I error is italicized.