BACKGROUND. Assessment of real-world vaccine effectiveness (VE) is needed in the general population and in special populations to evaluate dosing intervals, variant protection, and immunity duration. Assessing and addressing exposure and outcome misclassification and confounding in observational studies is critical for results interpretation.
OBJECTIVE. Discuss sources of and methods to address exposure and outcome misclassification and potential confounding when evaluating VE with administrative claims data in observational studies.
DESCRIPTION. The symposium provides four presentations followed by 20-minute Q&A and panel discussion.
(1) Overall review of the symposium and perspectives for regulators (5-10 min): Azadeh Shoaibi, PhD, MHS Introduce the Biologics Effectiveness and Safety (BEST) Initiative within the U.S. Food Drug Administration (FDA), Center for Biologics Evaluation and Research (CBER) surveillance program, which contributes to CBER’s mission to assure the safety and effectiveness of biologic products. Discuss the need for VE evaluation using real-world data (RWD) for public health purposes, challenges associated with using RWD in observational studies, and recognition of an uncertainty level these studies carry.
(2) Overview of causal inference using RWD in observational studies focusing on exposure misclassification and confounding (20 min): Michele Jonsson Funk, PhD. Discuss criteria needed for causal inference from observational data, including addressing confounding and misclassification by exposure, outcome, and covariates. Assumptions needed and methods to address misclassification and confounding will be presented.
(3) COVID-19 vaccine exposure misclassification in administrative claims databases in the U.S. (20 min): John Seeger, PharmD, DrPH Describe a) how COVID-19 vaccination is captured in administrative claims data sources, b) potential sources of COVID-19 vaccine misclassification and their impact on VE results, c) methods to address misclassification by supplementing claims data with immunization registry data, and d) application of quantitative bias analyses to address residual misclassification.
(4) Confounding and selection bias in a COVID-19 VE study (20 min): Bradley Layton, PhD Present a summary of a COVID-19 VE study conducted by FDA and its collaborators. Focus on confounding control at design (through matching on select variables) and analysis (i.e., inverse probability treatment weight, negative controls) stages, how to address selection bias for different time metrics (i.e., calendar time, time since vaccination) when assessing immunity duration and protection against variants.
