Signal mining and analysis of adverse events associated with Isotretinoin: A 20-Year real-world pharmacovigilance study based on FAERS and EudraVigilance databases

Abstract

Objective To systematically characterize the post-marketing safety signals of isotretinoin using real-world data from the U.S. FDA Adverse Event Reporting System (FAERS), with independent external validation in the European EudraVigilance (EV) database. Methods Adverse event (AE) reports in FAERS from 2004Q1 to 2024Q3 were analyzed. Signal detection was conducted using four complementary disproportionality algorithms: reporting odds ratio (ROR), proportional reporting ratio (PRR), Bayesian confidence propagation neural network (BCPNN), and the multi-item gamma Poisson shrinker (MGPS). Signals concurrently detected by all four methods were defined as robust. Key findings were subsequently examined in EV as an external reference. Results Among 50,519 patients contributing 142,160 isotretinoin-associated AE reports, 469 statistically robust signals were identified, spanning 25 system organ classes (SOCs). Signals were most concentrated in psychiatric disorders (75, 15.99%), gastrointestinal disorders (58, 12.37%), and congenital, familial, and genetic disorders (50, 10.66%). The strongest association was observed for inflammatory bowel disease (IBD; ROR = 579.14); however, temporal clustering and reporter-type profiling suggested substantial stimulated reporting, potentially driven by litigation, warranting cautious interpretation. Several high-ranking signals were not described in current product labeling, including nasal vestibulitis, hypertrophic anal papilla, and SAPHO syndrome. Pregnancy-related signals were prominent, with unintended pregnancy showing a strong signal (ROR = 91.39). External validation in EV demonstrated high concordance, supporting the robustness and reproducibility of the findings. Conclusions Isotretinoin is associated with a broad spectrum of pharmacovigilance signals, with disproportionate representation of psychiatric, gastrointestinal, and pregnancy-related events. While multiple previously unlabelled signals emerged, their clinical relevance remains to be established. These findings underscore the need for strengthened clinical monitoring, rigorous pregnancy prevention strategies, and careful interpretation of spontaneous reporting data.