Irreproducible results The role of sample size blinding and interim analysis
David Meldrum, M.D.
Volume 102, Issue 3, Pages 656-657
In two reviews previously published in this journal on study design and clinical trials registration, we emphasized the importance of doing a calculation of the sample size required to examine the primary outcome of a trial (1, 2). Then the data must not be analyzed until that number of subjects is reached. If interim analysis or analyses are carried out it is necessary to alter the level of statistical significance required to reject the null hypothesis (3). Unfortunately, the clinicaltrials.gov website does not specifically require recording of a sample size calculation, but rather requires that the investigators record a “target or actual number of patients,” assumed to be equal to the sample size plus enough additional enrolled subjects to make up for expected drop-outs.