ABSTRACT
Up-and-down (UD) dose-finding designs are used extensively in a variety of scientific and engineering fields. The first technical documents describing UD designs date to the 1940s and involve military explosive testing to find the optimal height from which to drop a bomb (Anderson et al. 1946; Dixon and Mood 1948) and hearing-threshold determination (von Békésy 1947). In both of these applications, UD designs are still the method of choice. Other common UD applications include failure-threshold determination in electrical and material engineering (Lagoda and Sonsino 2004) and finding the median effective dose (ED50) in anesthesiology (Pace and Stylianou 2007). As these examples suggest, dosefinding is a general term to describe experiments whose outcomes are binary (“yes/no”), conducted to find the treatment that would trigger a “yes” response at a prespecified frequency. UD procedures comprise a family of sequential dose-finding designs, meaning that treatments are ordered in time and each treatment (except the first one) is determined by previous outcomes rather than being predetermined before the experiment’s start.
