Assessing Mediation in HIV Intervention Studies

• Activity Number: 459
• Type: Topic Contributed
• Date/Time: Wednesday, August 9, 2006 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Felicia Hardnett and Sherri Pals*+ and Craig Borkowf and Ann O'Leary and Jeffrey Parsons and Cynthia Gomez
• Companies: Centers for Disease Control and Prevention and Centers for Disease Control and Prevention and Centers for Disease Control and Prevention and City University of New York-Hunter College and University of California, San Francisco
• Keywords: mediation ; asymmetric confidence limits ; HIV intervention studies ; intervention trials ; intervening variable effect ; indirect effect

Abstract

Recent literature suggests the widely used approach for assessing mediation, the causal st eps method, can be severely underpowered in many research settings. In the present study, we compare results obtained by 95% asymmetric confidence limit (ACL) construction with results obtained by the causal steps method. ACL construction uncovered significant mediating factors not identified using the causal steps method. We also identified marginally significant suppressors. Constructing ACLs is preferred for this type of analysis because of its statistical power and because it does not require the intervention to have a significant effect on the outcome. This method is also preferred because the distribution of the mediation effect is asymmetric. Also, ACLs present the size of the mediating effect, rather than just a binary decision of significance.