|
Abstract:
|
To stop unsuccessful clinical study as early as possible for drug development, biopharmaceutical statisticians are expected to compare the efficacy endpoint in an early phase study to a target objective response rate (ORR). The sample sizes in these trials are often small. Assuming the number of responders follows a Binomial distribution, Bayesian approach is applied to calculate the posterior probability of exceeding the target ORR given an experimental observation. A predefined go/no-go threshold is then selected according to the posterior probability. This decision threshold can be varied with different model assumption. However, the type of uncertainty due to the various distance between the threshold and all possible results has not been addressed. A normalized Average Distance between all possible outcomes and a Decision threshold (ADD) is proposed to quantify the uncertainty. A shiny app tool will be provided to implement the proposed method.
|
