This presentation introduces the concept of adaptive design, a method of research in which one or more elements is deliberately changed mid-study.
Heuser, A., Huynh, M., and Zhou, C. (2016, February). Introduction to Adaptive Designs [Conference workshop]. American Statistical Association Conference on Statistical Practices, San Diego.
Unlike conventional studies or clinical trials, where the design is pre-specified, adaptive designs are a class of study designs with adaptive or flexible study characteristics. Things that can vary include the sample size, variable dose or intervention, or flexible number of study arms. This flexibility is guided by examination of the accumulated data at an interim point in the study, and initiating changes in the design that may make the studies more efficient, more likely to demonstrate an effect of the intervention if one exists, or more informative (e.g., by providing broader dose-response information.) This course will introduce adaptive designs, explore their weaknesses and strengths, illustrate the techniques with examples, and allow attendees to have hands-on experience with a toolkit designed to get started in using adaptive designs.
This half-day course will focus on the adaptive design methods that have been used in applied research, and the adaptive designs used in clinical trials will be only a subset of those methods. Given the short timeframe of the course, we not be able to highlight all possible use. Clinical trials are an initial focus because of our experience at NIH, but we will discuss any applications depending on the composition of the course participants. If the participants are primarily non-clinical researchers, we will use non-clinical examples to teach the fundamental methods. We will have a fairly complete set of references for examples of how adaptive designs can be used. Throughout the course we will also highlight and reference applications of adaptive designs in social sciences and program evaluation research, survey research, and of course, clinical research.