Sample Size and Power Estimates
The primary measure used for sample size estimation
was the IE’s rating of Phase I treatment response. Using
chi-square, power estimates for detecting differences in
treatment response among the four treatment conditions
were computed using the following assumptions:
(1) Ha: P(SRT) = 0.60, P(CBT) = 0.60, P(COMB) = 0.80, and
P(PBO) = 0.30; (2) sample sizes of 136 for each active
treatment condition and 70 for the PBO condition; (3)
no adjustment for multiple comparisons; (4) power set
at 80%; and (5) alpha = 0.05, two-tailed test. Given these
assumptions, power analysis revealed that CAMS was
sufficiently powered to detect a 0.19 difference in Phase
I response rates between PBO and each active treatment
condition and a 0.17 difference in Phase I response rates
between COMB and each active monotherapy condition.