BACKGROUND: Using real-world data to assess COVID-19 vaccine safety is challenging due to evolving vaccine recommendations, potential incomplete capture of COVID-19 vaccines in data sources, and emerging safety signals.
OBJECTIVE: To describe critical study design considerations in the design of a Pfizer-BioNTech COVID-19 vaccine postauthorization safety study (PASS) in the United States in a rapidly changing COVID-19 environment.
METHODS: This PASS uses primarily claims data from partners in the Sentinel System, with a source population of ~65 million people. Three study designs were considered: cohort design with concurrent unexposed comparators, cohort design with historical unexposed comparators, and self-controlled risk interval (SCRI) design.
The potential for incomplete capture of COVID-19 vaccines in claims (which could lead to misclassification of unexposed status) and the potential for high vaccine uptake in the study population (which could lead to small numbers of unexposed comparators) were critical design considerations. Other key exposure-related considerations were the multidose primary vaccination schedule, added booster doses, prioritizing of high-risk groups, and changing vaccine eligibility criteria over time. Key outcome-related factors were changing healthcare utilization during the pandemic, relatively limited knowledge of the safety profile at study initiation, and emerging safety signals.
RESULTS: The study is planned to use a cohort design with unexposed comparators as it is appropriate for both acute and nonacute events. Concurrent rather than historical comparators were chosen because changes in healthcare utilization during the pandemic could impact outcome ascertainment. Confounding will be addressed by matching exposed to unexposed comparators on time-specific propensity scores.
Before final analysis, the study will assess counts of unvaccinated individuals over time to confirm the feasibility of using concurrent unexposed comparators. Completeness of COVID-19 vaccine data will also be assessed before final analysis by comparing vaccine coverage estimates in data sources to estimates from public health authorities.
Sensitivity analysis will incorporate a SCRI design and a cohort design with historical comparators to address potential misclassification of unexposed status.
CONCLUSIONS: Flexibility is critical in designing this PASS for Pfizer-BioNTech COVID-19 Vaccine using real-world data. Although a cohort design with concurrent unexposed comparators is planned, the primary study design may be converted to SCRI (for acute events) and cohort design with historical comparators (for nonacute events) if vaccine capture is incomplete or if there are insufficient concurrent unexposed comparators.
