Clinical Trials

Biography

Biography: Wenle Zhao

Abstract

The superiority of maximum tolerated imbalance (MTI) randomization procedures over the traditional permuted block design has been well documented in literature. However, the permuted block design remains the most used method in clinical trial practices. This trend may reflect a lack of awareness of the disadvantages of permuted block design and a lack of easy implementation methods of MTI procedures. This presentation evaluates  the statistical properties and implementation methods for five MTI procedures (Big Stick Design, Biased Coin Design with Imbalance Tolerance, Ehrenfest Urn Design, Block Urn Design, Asymptotic Maximal Procedure) and permuted block design based on the conditional allocation probability. Analytic results for the probability of deterministic assignments, the probability of complete random assignments, the probability of correct guess, as well as the standard deviation of treatment imbalance for these randomization designs. Recommendations for the selection of proper randomization design are provided for trials with and without stratification based on the consideration of treatment allocation randomness. Furthermore, a unified framework for central real-time subject randomization is proposed for the implementation of different randomization algorithms based on their conditional allocation probability functions. This approach removes the requirement of pre-generated randomization sequences, eliminates the risk of treatment allocation concealment failures, and providers potentials for the use of hierarchy randomization using MTI procedures. Compared to stratified permuted block design, this new strategy offers more effective controls for both imbalances in multiple baseline covariates and the overall treatment imbalance. This randomization design has been implemented in more than 10 multicenter clinical trials in the NIH Stroke Trials Network