Statistical issues in the use of dynamic allocation methods for balancing baseline covariates

Abstract

Background: The procedure for allocating patients to a treatment arm in comparative clinical trials is frequently chosen with only minor deliberation. This decision may, however, ultimately impact the trial inference, credibility, and even validity of the trial analysis. Cancer researchers are increasingly using dynamic allocation (DA) procedures, which balance treatment arms across baseline prognostic factors for clinical trials in place of historical methods such as simple randomisation or allocation via the random permuted blocks.Methods: This article gives an overview of DA methods, the statistical controversy that surrounds these procedures, and the potential impact on a clinical trial results. Results: Simple examples are provided to illustrate the use of DA methods and the inferential mistakes, notably on the P-value, if incorrect analyses are performed.Interpretation:The decision about which method to use for allocating patients should be given as much consideration as other aspects of a clinical trial. Appropriately choosing between methods can affect the statistical tests required and what inferences are possible, while affecting the trial credibility. Knowledge of the different methods is key to appropriate decision-making. © 2011 Cancer Research UK All rights reserved.