Using a balancing procedure in multicenter clinical trials. Simulation of patient allocation based on a trial of ventilation tubes for otitis media with effusion in infants.

Abstract

A basic issue in randomized controlled trials (RCTs) is whether we can safely assume comparability between groups at baseline with respect to all potentially important prognostic factors. In other words, did randomization work sufficiently well? In small trials balanced allocation procedures are employed, whereas in large-scale trials simple randomization will do. The question is: When should balancing be considered? We performed a simulation study in which we varied the number of categories in the prognostic factors and the number of patients. Simulation showed that, in all instances, a balancing procedure almost always led to perfect or almost perfect balance, while the imbalance with simple randomization was larger. To study the effect of balanced and random allocation on subgroup analyses in our OME trial, we compared the quotient of the width of the confidence intervals (CI). The widest CI in random allocation over the 13 hospitals was on average 13% wider than in balanced allocation. Investigators should always consider balanced allocation, especially in categories with a low number of patients and when subgroup analysis over many categories is requested.