Drug-associated bipolar disorder: A disproportionality analysis of the FAERS database

Drug-induced bipolar disorder (BD) is easily neglected clinically. We mined the U.S. Food and Drug Administration (FDA) Adverse Event Reporting System (FAERS) database to screen high BD-risk drugs and assess BD risk warnings in drug labels.

Data were extracted from the FAERS database spanning the first quarter of 2004 to the third quarter of 2025. Preferred terms (PTs) related to BD, including 'bipolar disorder', were identified using the Medical Dictionary for Regulatory Activities (MedDRA v27.1). Generic drug names were standardized via the DrugBank database. Sex-stratified and BD subtype-stratified disproportionality analyses were performed.

A total of 23,607,454 adverse event reports were analyzed, with 17,620 cases linked to BD. Varenicline (537 cases) was associated with the highest number of BD reports. Based on disproportionality analysis, the top five medications with the highest reporting odds ratios (RORs), proportional reporting ratio (PRR), empirical bayes geometric mean (EBGM) and information component (IC) were pizotifen, tafenoquine, flupenthixol, naltrexone, and pemoline. Among the top 50 medications, 44 (88%) lacked any mention of BD risk in their prescribing information. Time-to-onset analysis revealed that 51% of cases (n = 993) occurred within 0-30 days of drug exposure. Signals differed markedly by sex: varenicline showed higher male risk, while esomeprazole and oxycodone carried mild female-specific risks. Naltrexone strongly associated with BD-I and oxcarbazepine with BD-II, with elevated subtype-specific RORs limited by small subtype samples.

Most high BD-signal drugs lack label warnings. Drug-induced BD presents distinct sex and subtype-specific risks, requiring gender- and subtype-personalized clinical monitoring.