Abstract Body

Background: In the simplest randomized controlled trial (RCT) design, individual patients are randomly allocated to two study arms (Intervention vs Control) to measure the effects of a specified health intervention on a defined outcome. In some cases, however, we wish to examine the effects of multiple interventions in the same trial. Examples of such studies include (1) treatments with multiple dosage regimens or alternative drugs; (2) studies designed to compare the effects of two or more different interventions, alone or in combination; (3) combination interventions where we wish to unpick the effects of different components. We will discuss some of the study designs that can be used to address such questions. A simple design compares two or more intervention conditions with a single control arm. There are sometimes issues of multiple comparisons and, depending on the mode of delivery of the interventions, there may be a need for multiple placebos if a blinded design is required (e.g. double-dummy trials). When the separate or joint effects of interventions are of interest, the factorial design may be an attractive choice, for example the 2×2 factorial in which patients are randomized to four groups given either intervention A alone, intervention B alone, neither or both. Under certain assumptions, the factorial design permits the effect of more than one intervention to be measured in a single trial at a similar cost to a study designed to evaluate a single intervention. The limitations and strengths of this design will be reviewed, and implications for sample size outlined. Such trials may be randomised either at individual or cluster levels. In some cases, the two approaches may be combined, such that clusters are randomly allocated to the presence or absence of one intervention, while individuals or households within clusters are randomly allocated to presence or absence of another intervention (sometimes referred to as a split-plot design). The above design choices will be illustrated with RCTs conducted to measure the effects of interventions against HIV and related infections, including the Zamstar, VOICE and PopART trials. Conclusions: While it is often best to aim for simplicity of design, resources for rigorously-designed RCTs are limited and it is important to ensure that such resources are used efficiently. Sometimes, by careful choice of study design, it may be possible to evaluate several interventions in a single study at little additional cost. Alternative approaches to unpick the effects of combination interventions also need to be considered. In particular mathematical modelling may sometimes be used to supplement RCT evaluations.