Information-adaptive clinical trials with selective recruitment and binary outcomes

Information-adaptive designs estimate how much statistical information an individual is expected to provide and allocates them to treatment arms in a manner that maximises information gain. Information-adaptive designs with selective recruitment preferentially recruit individuals that are estimated to be statistically informative onto a clinical trial. Individuals that are expected to contribute less information have a lower probability of recruitment. The informativeness of an individual depends on their covariate values and how information is defined is a critical element of information-adaptive designs. In this paper we define and evaluate four different methods for quantifying statistical information. Using both experimental data and numerical simulations we show that selective recruitment designs can offer a substantial increase in statistical power compared to randomised designs. In some cases we observe that statistical power was increased by simply excluding weakly informative patients that would otherwise have been recruited in a randomised trial. In trials without selective recruitment we find that allocating individuals to treatment arms according to information-adaptive protocols leads to an increase in statistical power. Consequently, selective recruitment designs can potentially achieve successful trials using fewer recruits thereby offering economic and ethical advantages.