|
Activity Number:
|
99
|
|
Type:
|
Invited
|
|
Date/Time:
|
Monday, August 9, 2004 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #300251 |
|
Title:
|
Causal Vaccine Effects on Binary Post-infection Outcomes
|
|
Author(s):
|
Michael G. Hudgens*+ and M. Elizabeth Halloran
|
|
Companies:
|
University of North Carolina, Chapel Hill and Emory University
|
|
Address:
|
Dept. of Biostatistics, School of Public Health, Chapel Hill, NC, 27599,
|
|
Keywords:
|
causal inference ; selection bias ; vaccine
|
|
Abstract:
|
The goal of prophylactic vaccination is to prevent infection or ameliorate disease or decrease infectiousness of individuals who become infected subsequent to vaccination. Thus, evaluation of several vaccine effects of interest condition on being infected. Standard approaches to assessing vaccine effects on the distribution of post-infection outcomes generally do not assess causal effects of vaccination, because comparison groups are selected post-treatment assignment. We consider estimation and testing of causal effects of vaccination on binary post-infection outcomes (denoted VEp) such as secondary transmission and severity of disease. We adopt the approach of Frangakis and Rubin (2002) and define causal effects within principal strata defined on the joint potential infection values under vaccine and placebo. While the causal VEp is generally not identifiable without unverifiable assumptions, using a maximum likelihood based approach we derive bounds on the causal estimands of interest. Sensitivity analysis and testing procedures are also developed. The methodology is demonstrated using data from field studies of a rotavirus vaccine candidate and a pertussis vaccine.
|
