Future of ChatGPT in Pharmacovigilance
페이지 정보
본문
ChatGPT is inspiring a broad spectrum of applications across various domains, including pharmacovigilance (PV). With its ability to process vast amounts of textual data and engage in real-time dialogue, ChatGPT offers significant opportunities to strengthen Adverse Drug Reaction (ADR) reporting and improve the accuracy and speed of PV operations. Given the FDA’s estimate that only 1–10% of ADRs are reported to FAERS, with mentions occurring far more frequently on social media, ChatGPT’s value lies in its ability to identify ADRs using Real-World Evidence (RWE) from non-traditional platforms. Its capability is particularly evident in extracting ADR keywords from the unstructured language prevalent on social media.
In a practical test, ChatGPT successfully generated an ADR inventory for the diuretic Lasix that aligned with published research; however, its performance was highly sensitive to phrasing. While it produced an accurate ADR profile using the brand name, it incorrectly associated the drug with other antibiotics when the generic name was used. Furthermore, despite its multilingual availability, ChatGPT currently lacks adequate training data for pharmaceutical information in languages other than English.
Notably, ChatGPT identified hyperglycemia, gastrointestinal issues, and skin rash as the top three ADRs associated with alpelisib, precisely matching the most frequent Grade 3/4 ADRs observed in Phase III clinical trials. Such capabilities in data aggregation and prioritization can be invaluable for the post-marketing surveillance of newly approved or less common medications requiring urgent monitoring. However, while ChatGPT excels at compiling comprehensive lists of relevant ADRs, it falls short in determining causality, and its effectiveness as a Drug-Drug Interaction (DDI) database remains uncertain.
Relying solely on ChatGPT’s responses for clinical decision-making could compromise the scientific rigor of drug safety research and jeopardize patient health. Therefore, ChatGPT-generated outputs must be critically analyzed through the lens of a PV specialist’s advanced knowledge and expertise.
Ref.> Hanyin Wang, Yanyi Jenny Ding & Yuan Luo.Drug Safety. 2023:46;711–713.
https://link.springer.com/article/10.1007/s40264-023-01315-2
