RECOMMENDATIONS AND CONCLUSIONS
Selecting appropriate control conditions in behavioral intervention research design and development is a critical decision. Choice of a control condition will depend upon several factors, including phase of intervention development, theory base, available resources, and ethical considerations. At the RCT phase, as should be apparent by this point, no RCT and no control condition are perfect. To help make more informed decisions about control group selection for RCTs, we recommend that the following points be taken into consideration. First, it is important to give as much attention to the selection of the control group as one does to the choice of the intervention. It has been our experience that researchers often think long and hard about which interventions to choose, but often approach control group selection as an afterthought. In this regard, avoid automatically selecting a no-treatment control as a default option, although there may be instances when no-treatment controls are called for, for example, to control for practice effects in cognitive intervention trials. Second, consider the use of multiple controls or comparisons, not always a single control condition. It may be the case that different types of controls are needed for a given study. For example, researchers may need to control for practice effects as well as the amount of social contact, as in the ACTIVE study (Jobe et al., 2001). Third, try to minimize differential dropout or attrition in the control condition by maintaining regular contact with control participants, by developing a strong tracking system, by creating a strong research project identity, and through emphasizing the importance of the control condition prior to randomization. More people may drop out of the control condition than the treatment condition, and often do so nonrandomly, creating problems interpreting the results of randomized clinical intervention trials. Fourth, it is important to pilot test control conditions if they involve some activity, and research staff should be trained in their delivery (Stephenson & Imrie, 1998). In addition, all unplanned contact in both treatment and control conditions should be logged. Finally, along with all of the previously mentioned considerations, researchers must become more aware of the broad array of theoretical, methodological, and ethical issues in control group selection (Gross,
2005). Such efforts should result in a better evidence base upon which to create standard best practice and policy. Having consistent evidence in a series of RCTs is generally considered to establish the intervention as “evidence-based” (i.e., it has sufficient data to support its use and broader scale adoption).