Parallel group trials allocate each participant to a single intervention for comparison with one or more alternative interventions. In contrast, cross-over trials allocate each participant to a sequence of interventions. A simple randomized cross-over design is an ‘AB/BA’ design in which participants are randomized initially to intervention A or intervention B, and then ‘cross over’ to intervention B or intervention A, respectively. It can be seen that data from the first period of a cross-over trial represent a parallel group trial, a feature referred to in Section 16.4.5. In keeping with the rest of the Handbook, we will use E and C to refer to interventions, rather than A and B.
Cross-over designs offer a number of possible advantages over parallel group trials. Among these are (i) that each participant acts as his or her own control, eliminating among-participant variation; (ii) that, consequently, fewer participants are required to obtain the same power; and (iii) that every participant receives every intervention, which allows the determination of the best intervention or preference for an individual participant. A readable introduction to cross-over trials is given by Senn (Senn 2002). More detailed discussion of meta-analyses involving cross-over trials is provided by Elbourne et al. (Elbourne 2002), and some empirical evidence on their inclusion in systematic reviews by Lathyris et al. (Lathyris 2007).