(A-358) Using Tree-Based Scan Statistics to Monitor Routine Vaccines for Possible Associations with Guillain-Barré Syndrome in the Vaccine Safety Datalink

Background: Tree-based scan statistics are used to study the safety of specific medical products across a broad range of diagnoses organized in a hierarchical structure, where an exposure of interest is chosen and patients’ post-exposure data are scanned for clustering of these diagnoses. Here, we applied a case-crossover design, choosing instead an outcome of interest and scanning hierarchically organized vaccine exposures among patients who experienced the outcome to identify possible associations.

Objectives: To use tree-based scan statistics and a hierarchical structure (tree) of vaccines to determine which vaccines or groups of vaccines, might be associated with Guillain-Barré syndrome (GBS).

Methods: We used a case-crossover design that is outcome-focused and self-controlled, automatically adjusting for time-invariant confounding. The study population was adults aged 18+ years in the U.S. Vaccine Safety Datalink in 2017-2023. Incident GBS cases were identified with ICD-10-CM code G61.0 in an emergency department or inpatient setting with no G61.0 or G65.0 observed in any setting in the prior 400 days. We assessed all U.S. vaccines in the 84 days prior to the incident GBS code, using a vaccine tree that contained 99 unique CVX codes organized into 7 major vaccine categories and up to 5 levels. We used an unconditional Bernoulli tree-based scan statistic with a risk window of Days 1-21 (primary) and Days 1-42 (secondary) after vaccination and a control window of Days 43-84. We required a multiplicity-adjusted p-value ≤0.05 to signal an association.

Results: There were 2,584 incident GBS cases, of which 40% were ≥65 years old and 24.7% had at least one vaccination code in the preceding 0-84 days. The primary analysis detected statistical signals for influenza vaccines (odds ratio (OR): 1.88) and some subcategories; Ad26.COV2.S COVID vaccine (OR: 12); AS01-adjuvanted vaccines, including recombinant zoster (OR: 2.58); pneumococcal vaccines, including conjugate (PCV) (OR: 3); and tetanus-diphtheria-acellular pertussis (Tdap) vaccine (OR: 2.52). The secondary analysis similarly detected statistical signals for these vaccines, excluding PCV and Tdap, with lower odds ratios.

Conclusions: Using a new hierarchical organization of vaccines and tree-based scan statistics, we identified statistical signals of potential associations between 5 vaccines or vaccine groups and GBS. These results are qualitatively consistent with previously reported work. This method holds promise for studying other outcomes suspected of being associated with vaccination if case-finding algorithms are relatively specific. The possibility of detecting associations among groups of vaccines with similar delivery platforms (e.g., adjuvanted) is an advantage of tree-based data-mining.