Table 2

Summary of simulation results

Scenario	Number of patients, mean (SD)					Reached maximum accrual (proportion)§	Proportion of trials found one or more arms efficacious at completion*	Proportion of trials found one or more arms efficacious and stopped early†	Proportion of trials in which arm was found efficacious‡
Scenario	Total sample size	Control	Group 2	Group 3	Group 4	Reached maximum accrual (proportion)§			Group 2	Group 3	Group 4
1	456 (76)	140 (28)	105 (40)	106 (41)	105 (40)	0.63	0.05	0.01	0.021	0.021	0.021
2	437 (93)	150 (37)	42 (20)	122 (47)	122 (47)	0.53	0.03	0.01	0.000	0.025	0.024
3	466 (58)	140 (21)	96 (34)	96 (34)	135 (22)	0.63	0.52	0.37	0.024	0.026	0.887
4	431 (80)	143 (29)	42 (20)	107 (39)	140 (29)	0.42	0.35	0.58	0.000	0.039	0.929
5	457 (667)	129 (20)	89 (32)	120 (21)	120 (21)	0.58	0.52	0.42	0.039	0.846	0.849
6	374 (95)	117 (33)	41 (13)	108 (33)	108 (33)	0.21	0.17	0.79	0.000	0.883	0.887
7	400 (109)	108 (29)	97 (28)	97 (28)	97 (28)	0.39	0.34	0.61	0.806	0.808	0.808
8	441 (84)	128 (24)	98 (37)	98 (37)	117 (22)	0.53	0.53	0.47	0.025	0.024	1.000
9	395 (107)	129 (35)	33 (14)	113 (43)	120 (31)	0.35	0.35	0.65	0.000	0.042	1.000

The table shows the summary of 10 000 trial simulations. The total sample size for a frequentist trial with one control and three active arms, with power of 0.8 and two-sided comparison-type I error of 0.05, would be 500. The total sample size for three two-arm frequentist trials run in series, with power of 0.8 and two-sided comparison-type I error of 0.05, would be 774.
*Proportion of trials that concluded one or more arms were efficacious at the last look.
†Proportion of trials that finished early and concluded one or more arms were efficacious.
‡Success group-proportion of trials in which the arm was found efficacious.
§Proportion of trials that reached the maximum sample size