From: Should RECOVERY have used response adaptive randomisation? Evidence from a simulation study
Operating characteristic where sample size matches RECOVERY | FuR | FeR | Tf | RMCf | |
---|---|---|---|---|---|
Proportion allocated to control (%) | 40.0 | 25.0 | 40.0 | 40.0 | |
Proportion allocated to dexamethasone (%) | 20.0 | 25.0 | 27.0 | 26.6 | |
Proportion allocated to hydroxychloroquine (%) | 20.0 | 25.0 | 14.9 | 15.0 | |
Proportion allocated to lopinavir (%) | 20.0 | 25.0 | 18.1 | 18.4 | |
Mortality (deaths/9600 and percentage) | 2394 (24.9%) | 2388 (24.9%) | 2373 (24.7%) | 2374 (24.7%) | |
Statistical power dexamethasone (%) | 54.6 | 49.8 | 59.9 | 58 | |
Statistical power lopinavir (%) | 9.6 | 9.2 | 6.5 | 7.9 | |
Family-wise error rate (%) | 3.8 | 4.7 | 4.9 | 5.1 | |
Bias (×10−2) | Dexamethasone | 0 | 0 | −4.1 | −5.4 |
Hydroxychloroquine | 10.0 | 6.2 | |||
Lopinavir | − 33.2 | −39.1 | |||
Mean Squared error (×10−4) | Dexamethasone | 1.60 | 1.98 | 1.50 | 1.57 |
Hydroxychloroquine | 1.70 | 1.94 | 2.28 | 2.11 | |
Lopinavir | 1.84 | 1.89 | 1.81 | 1.91 |