Evaluating COVID-19 Vaccine Masking and Unmasking Methods in Two National Pharmacovigilance Databases

The COVID-19 mass vaccination led to a substantial increase in spontaneous reports submitted to pharmacovigilance (PV) databases, potentially introducing masking effects that could conceal safety signals.

To examine the masking effect of COVID-19 vaccines on disproportionality analyses and to compare two unmasking interventions in the Dutch (Lareb database) and Spanish (Farmacovigilancia Española, Datos de Reacciones Adversas, FEDRA) national PV databases: removal of all drug-event combinations (DEC) involving a COVID-19 vaccine versus excluding influential outliers DECs only.

The masking effect was explored retrospectively on the basis of the number of signals of disproportionate reporting (SDR). DECs involving a COVID-19 vaccine were excluded using crude and outlier techniques, and reporting odds ratios were recalculated. Subsets of important medical events (IME) were analysed in both databases.

Both crude and influential outlier removal methods led to reductions in the number of reports, DECs and SDRs. Both in the Lareb database and FEDRA, crude removal excluded 2.1% of DECs, while the outlier method excluded 0.1%. Crude removal had a greater impact on SDRs, reducing them by 9.8% in the Lareb database and 3.9% in FEDRA, compared with 5.7% and 1.1% with the outlier method. In the Lareb database, 1301 SDRs (20 IME-related) were unmasked using crude removal, and 1942 (95 IME-related) with the outlier method. FEDRA showed 1453 and 1226 SDRs unmasked, including 41 and 70 IME-related.

COVID-19 vaccines caused substantial masking in both databases. Both strategies effectively revealed new SDRs, though their impact varied. The choice of approach should be tailored to the database context.