Thanks for my colleagure who point out my mistake in calculating the sample size adjusting for the drop out rate. According to the sample size calculation, we estimated that 100 subjects would be needed to detect the statistical differences. Considering 20% drop out rate, what is the # of subjects to be randomized?
It is not 100 + 100*20% = 120
It should be 100 / (1-0.2) = 125
It means that with 125 randomized subjects, if 25 subjects (20% of 125) drop out, we still have 100 subjects left. This is correct way to adjust for the drop out rate.