Can We Compare Adverse Event's Attributable Risk Using a Self-Controlled Case Series Design for Vaccine Safety? A Guillain-Barré Syndrome Use Case.

INTRODUCTION: Estimating attributable risk (AR) through self-controlled case series (SCCS) analyses alone may limit generalizability because SCCS only incorporate vaccinated patients with the outcome (e.g., Guillain-Barré syndrome [GBS]) who may differ from the recommended vaccinee population. We aimed to demonstrate background event incidence rates' impact on vaccine-specific GBS ARs and to standardize ARs across different vaccine studies by applying a generalizable, population-based GBS background rate to better estimate the expected population-level ARs. METHODS: We identified post-licensure SCCS vaccine studies and GBS background rates using US Medicare data via targeted literature review. GBS control period rates from SCCS vaccine studies were extracted or calculated. Population-level ARs were calculated for each vaccine using two published and two hypothetical background GBS rates and the original SCCS-generated incidence rate ratios (IRRs). RESULTS: Published vaccine-specific GBS IRRs ranged from 2.02 (95% CI 0.93-4.40) for RSVpreF to 4.96 (95% CI 1.43-17.27) for recombinant zoster vaccine (RZV). Study-specific ARs per 100,000 doses ranged from 0.28 (H1N1) to 0.90 (RSVpreF). Vaccines with lower control period GBS rates had a lower attributable risk for a given IRR. After standardization using external background GBS rates, population-level ARs were higher for vaccines with higher IRRs. For example, when using the H1N1 vaccine control period rate as the GBS background rate for AR calculation, RSVpreF had the lowest AR per 100,000 doses (0.21) and RZV the highest (2.26). CONCLUSIONS: AR calculation is dependent on the control period event rate within a given SCCS analysis. ARs derived exclusively from SCCS analyses may lead to incorrect conclusions regarding an adverse event's absolute risk in regulatory communications due to a lack of external validity. Using a representative background rate from the recommended vaccinee population, along with SCCS-derived IRR, offers a more consistent approach to calculate the expected population-level AR and supports vaccine risk-benefit assessments.