Information-Theoretic Criteria for Optimizing Designs of Individually Randomized Stepped-Wedge Clinical Trials

Abstract

Clinical trials are essential for advancing medical knowledge and improving health care, with Randomized Clinical Trials (RCTs) considered the gold standard for minimizing bias and generating reliable evidence on treatment efficacy and safety. Stepped-wedge individual RCTs, which randomize participants into sequences transitioning from control to intervention at staggered time points, are increasingly adopted. To improve their design, we propose an information-theoretic framework based on D– and A–optimality criteria for participant allocation to sequences. Our approach leverages semidefinite programming for automated computation and is applicable across a range of settings, varying in: (i) number of sequences, (ii) attrition rates, (iii) optimality criteria, (iv) error correlation structures, and (v) multi-objective designs using the ϵ-constraint method.