An Optimizing Up-and-Down Design
Assume that the probability of success is unimodal as a function of dose. Take the response to be binary and the possible treatment space to be a lattice. The Optimizing Up-and-Down Design allocates treatments to pairs of subjects in a way that causes the treatment distribution to cluster around the treatment with maximum success probability. This procedure is constructed to use accruing information to limit the number of patients that are exposed to doses with high probabilities of failure. The Optimizing Up-and-Down Design is motivated by Kiefer and Wolfowitz’s stochastic approximation procedure. In this paper, we compare its performance to stochastic approximation. As an estimator of the best dose, simulation studies demonstrate that the mode of the empirical treatment distribution using the Optimizing Up-and-Down Design converges faster than does the usual estimator using stochastic approximation.
Keywordsconstant gain stochastic algorithm Markov chains phase I/II clinical trial random walk designs stationary treatment distribution stochastic approximation
Unable to display preview. Download preview PDF.
- Hardwick, J. P. and Stout, Q. (1998). Personal communication. Google Scholar
- Kpamegan, E. E. and Flournoy, N. (2000). Up-and-down designs for selecting the dose with maximum success probability. Submitted.Google Scholar
- Wetherwill, G. B. (1963). Sequential estimation of quantal response curves. Journal of the Royal Statistical Society B 25, 1–48.Google Scholar
- Hardwick, J., Meyers, M. and Stout, Q.F. (2000). Directed walks for dose response problems with competing failure modes. (Submitted).Google Scholar