Confounder adjustment in vaccine safety studies: comparing three offset terms for case-centered approach

Abstract

The case-centered approach (CCA) can be useful in vaccine safety studies to adjust for time varying confounders, especially seasonality. The method compares the observed odds of vaccination in the time interval prior to the event with the expected odds of vaccination during this interval. The log of the expected odds is included in the analysis model as an offset term and it is critical to correctly estimate vaccine effects. In a recent Zostavax(®) safety study, we used three alternative sources of reasonable expectations for the offset term. We conducted a simulation study to compare the performance of the three offset terms with different types of seasonal confounders and at different degrees of confounding. The offset term calculated by modeling the timing of vaccination based on predictors such as age, sex and site resulted in the least biased estimate of relative risk (RR).