A Monte Carlo comparison of three up-and-down designs for dose ranging.

One of the first objectives in the clinical study of a new drug is to determine the range of effective dosages to be studied further. That is, to determine the threshold and plateau doses of the dose response. To this end, three up-and-down designs are described for use when a dichotomous (binary) response of primary interest can be defined. Each design involves entering the subjects chronologically, with each subject's dose being chosen based on the response of the preceeding subject: up one dose level after a response, down one after no response. The three designs were compared using Monte Carlo simulated results from a typical sample size of 30 observations each and dose-response relationship with a range of 5 increasing doses. The design involving one observation on each of 30 subjects was found to be most efficient, although the other two involving three observations on each of 10 subjects performed adequately.