Polishing a Diamond in the Rough: Deconstructing and Reconstructing the Estimate of Population Benefit of Vaccination

• Activity Number: 324
• Type: Contributed
• Date/Time: Tuesday, August 4, 2009 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Oliver M. Bautista*+
• Companies: Merck & Co., Inc.
• Keywords: Vaccine efficacy ; population benefit ; interaction model ; time dependent covariates

Abstract

The population benefit of a prophylactic vaccine such as the quadrivalent HPV vaccine is estimated from an ITT analysis of vaccine efficacy (VE). The ITT population consists of disease-free subjects (who benefit from vaccination) and diseased subjects (who do not benefit from vaccination). A pooled analysis of VE with vaccination group as the only covariate produces a biased estimate of population benefit since it ignores the VE-by-disease status interaction. In a pooled Cox PH model where VE is the percent reduction in vaccine group hazard relative to placebo, the estimated VE in 16 to 26 year-old women against HPV 16-related CIN 2/3 or worse is 38% (95% CI: 23 to 50%). In a model that accounts for the VE-by disease status interaction, the estimated VE is 97% (95% CI: 87 to >99%). Further refinements are achieved using accelerated failure time models with time dependent covariates.