Table 5.

Comparing the design with adaptive randomization to a trial with fixed randomization via power, mean sample size, and the proportion of patients randomized to the therapy with the highest response rate. The same early stopping rules are used in both. All values based upon 1,000 simulations per scenario.

Scenario	Adaptive Randomization			Fixed Randomization
	Power	Mean N	% to Best	Power	Mean N	% to Best
Null 0.50 – 0.50 – 0.50	0.031	507	N/A	0.029	499	N/A
One Good 0.50 – 0.50 – 0.65	0.90	483	48	0.88	497	33
Two Good 0.50 – 0.65 – 0.65	0.76	679	84	0.86	687	67
One Middle One Good 0.50 – 0.575 – 0.65	0.68	586	47	0.69	599	33
All Bad 0.25 – 0.25 – 0.25	0.044	524	N/A	0.030	509	N/A
All Really Bad 0.10 – 0.10 – 0.10	0.006	400	N/A	0.028	400	N/A

% to Best = Average proportion of patients randomized to the most effective therapy.

Power = probability of identifying a treatment as best or worst at the 0.975 level.