ABSTRACT
Sometimes neither the cluster randomized trial (Chapter 4) nor the multisite trial (Chapter 6) is a satisfactory choice for the design of a trial. Such a dilemma arises when randomization at subject level risks control group contamination (Craven, Marsh, Debus, & Jayasinghe, 2001; Keogh-Brown et al., 2007; Plewis & Hurry, 1998; Slymen, Elder, Litrownik, Ayala, & Campbell, 2003; Rhoads, 2011), while randomization at cluster level risks selection bias as well as slow and/or incomplete recruitment.
