Table 7.

Statistical power for testing the treatment effect under the null and alternative hypotheses for all patients and for erlotinib naïve patients by marker status in the frequentist’s equal randomization design. A p-value cut-off of 0.108 for the Fisher’s exact test was chosen to control the type I error rates for comparing TX 2, 3, 4 to TX 1 under the null hypothesis to 10%.

Scenario	TX	Margin	Naïve Margin	Naïve MK(−,−)	Naïve MK(+,−)	Naïve MK(−,+)	Overall	All TX
Null	2	0.007	0.054	0.044	0.018	0.003	0.097	0.200
	3	0.006	0.058	0.053	0.018	0.001	0.099
	4	0.004	0.036	0.050	0.004	0.001	0.071
Alt.	2	0.579	0.644	0.487	0.076	0.031	0.741	0.923
	3	0.593	0.651	0.495	0.081	0.035	0.756
	4	0.639	0.650	0.495	0.021	0.135	0.758