Abstract:
|
A number of methods in current clinical trial practice allow mid-trial decisions to be considered based on interim trial results. Potential actions include stopping the trial, whether for proven efficacy or for futility, or revising the trial sample size. Some decision thresholds are commonly described on a scale of conditional power, that is, the chance that the trial will succeed based on certain assumptions on the effect size governing the rest of the trial. CP calculations sometimes utilize the original study design hypothesis, while sometimes effect sizes consistent with current data trends are assumed. While CP may have some interpretive value, there is at times a tendency to over-interpret these as success probabilities, which can be problematic because interim data may not accurately reflect true treatment effects. This may lead to risks of setting criteria that seem intuitively reasonable but do not optimally achieve their objectives. We discuss and illustrate how criteria are better viewed as an aspect of trial design, and set based on thorough understanding of the operating characteristics they induce, rather than on intuitive interpretations as chances of trial success.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.