Informed randomization: A new method of outcome-adaptive randomization
A.H.J. Huizing · Utrecht University Repository (Utrecht University) · 2019
Traditionally, random assignment is performed with an equal probability for each treatment arm. In practice this may be limiting as we would ideally maximize the allocation of respondents to the superior treatment. Outcome-adaptive randomization solves this problem by providing a compromise between random assignment and assignment based on existing expectations of treatment superiority. This paper lays the groundwork for informed randomization; a method of outcome-adaptive randomization that can be flexibly implemented and can handle a wide variety of outcome measures. Informed randomization uses estimates of the unobserved potential outcome of an individual, which includes an estimate of treatment heterogeneity. The estimated potential outcomes are obtained through multiple imputation. As a proof-of-concept, a simulation study with continuous outcomes for two treatment arms is performed. Simulations show that, on average, informed randomization assigns more participants to the superior treatment arm. This comes at the cost of a reduction in statistical power.