ABSTRACT Background. Pharmacovigilance plays an important role in drug safety monitoring, particularly in post- marketing surveillance after clinical trials, especially when it comes to discovering Adverse Drug Events in the real-world environment. Methods. The approach taken in this study was to create a multi-source AI pharmacovigilance framework using openFDA/FAERS, DailyMed, ClinicalTrials.gov, and PubMed. Semaglutide and sertraline were chosen as a test case because they are from different therapeutic classes and thus offer a good test of the capacity of the framework to recover interpretable safety profiles in different adverse- event domains. Results. Cross-source overlap was summarized using agreement and divergence scores. Semaglutide showed a gastrointestinal profile and sertraline a broader neuropsychiatric profile, with clinically meaningful core safety signals despite low overall cross-source agreement and high divergence (semaglutide: agreement score 0.046, divergence score 0.954, 38 high-confidence terms; sertraline: agreement score 0.082, divergence score 0.918, 32 high-confidence terms). Conclusions. After the normalization and filtering, most terms extracted remained source-specific, as the structures were different for spontaneous reporting, regulatory labelling, posting of results during trials, and published literature reporting. The framework was able to identify clinically plausible core safety signals for both drugs. Keywords: pharmacovigilance, FDA Adverse Event Reporting System (FAERS), semaglutide, sertraline, adverse drug events, multi-source data integration, drug safety
Kasula et al. (Fri,) studied this question.