Small-scale randomized controlled trials need more powerful methods of mediational analysis than the Baron–Kenny method Ester Cerin a, * , Lorian M. Taylor b , Eva Leslie a , Neville Owen a a Cancer Prevention Research Centre, School of Population Health, The University of Queensland, Herston Road, Herston, Brisbane, QLD 4006, Australia b Centre for Health Promotion Studies, The University of Alberta, Alberta, Canada Accepted 7 November 2005 Abstract Objective: To devise more-effective physical activity interventions, the mediating mechanisms yielding behavioral change need to be identified. The Baron–Kenny method is most commonly used, but has low statistical power and may not identify mechanisms of behavioral change in small-to-medium size studies. More powerful statistical tests are available. Study Design and Setting: Inactive adults (N 5 52) were randomized to either a print or a print-plus-telephone intervention. Walking and exercise-related social support were assessed at baseline, after the intervention, and 4 weeks later. The Baron–Kenny and three alter- native methods of mediational analysis (Freedman–Schatzkin; MacKinnon et al.; bootstrap method) were used to examine the effects of social support on initial behavior change and maintenance. Results: A significant mediational effect of social support on initial behavior change was indicated by the MacKinnon et al., bootstrap, and, marginally, Freedman–Schatzkin methods, but not by the Baron–Kenny method. No significant mediational effect of social support on maintenance of walking was found. Conclusions: Methodologically rigorous intervention studies to identify mediators of change in physical activity are costly and labor intensive, and may not be feasible with large samples. The use of statistically powerful tests of mediational effects in small-scale studies can inform the development of more effective interventions. Ó 2006 Elsevier Inc. All rights reserved. Keywords: Randomized controlled trials; Mediational effect; Small sample; Exercise-related social support; Walking; Statistical power 1. Introduction To improve the effectiveness of physical activity inter- ventions, there is the need to identify how (analysis of me- diators), for whom (analysis of personal moderators), and under what circumstances (analysis of situational modera- tors) they can lead to increases in physical activity [1]. As well as improving the understanding of key elements in interventions, such research can also help clarify the gen- eral causal mechanisms underlying behavioral change [2]. Research on the mediators of physical activity behavior change has yielded an inconsistent pattern of findings [3,4]. These inconsistencies may in part be due to differences in the statistical power of studies on mediational effects [5]. To find a statistically significant mediational effect (if any exist), intervention studies must (a) produce sizable changes on the hypothesized mediators and outcomes, (b) be based on a sufficiently large sample size, and (c) make use of statistical methods capable of detecting a mediational effect of a specific magnitude for a given sample size. Exercise-related social support from family and friends is a factor associated with being more likely to be physi- cally active [6,7]. To date, however, only a small number of studies have examined the mediational effect of social support on changes in physical activity behavior, using a randomized controlled trial design. This is the recommen- ded methodological framework for testing hypothetical me- diators of behavior change [2]. Of these studies, those with a sample size O250 fully or partially supported the exis- tence of a mediational effect [8–10] and those with !150 participants failed to do so [11,12]. Given that smaller-scale trials typically found small-to-moderate positive associa- tions between the intervention, social support, and physical activity behavior (1%–12% of common variance), it is pos- sible that their failure to identify social support as a signif- icant mediator of changes in physical activity was partly due to a lack of statistical power. * Corresponding author. Tel.: 1852-2817-9846; fax: 1852-2818-8042. E-mail address: ecerin@hku.hk (E. Cerin). 0895-4356/06/$ – see front matter Ó 2006 Elsevier Inc. All rights reserved. doi: 10.1016/j.jclinepi.2005.11.008 Journal of Clinical Epidemiology 59 (2006) 457–464