Usefulness and limitations of ChatGPT in getting information on teratogenic drugs exposed in pregnancy.

Abstract

There is a growing interest in using artificial intelligence (AI) technology to obtain information on the risk and safety of drugs in pregnancy. Inadvertent drug exposure in pregnant women is inevitable in acute or chronic diseases, especially in unplanned pregnancies. According to the Korean mother safe counselling center database (2010-2023), many Korean pregnant women and their families asked about the teratogenicity of their exposed drugs. The most frequently used drug was topiramate (n=2,018), followed by isotretinoin (n=1,972), dexamethasone (n=1,279), and doxycycline (n=1,119). Unexpectedly, thalidomide, a notorious teratogen that causes phocomelia, was included. It has been withdrawn from general prescriptions since 1961, except for the treatment of multiple myeloma. Her partner used it for the treatment of multiple myeloma. In this study, we evaluated the usefulness of AI Chat generative pretrained transformer (Chat GPT) by comparing information between AI and a literature review on isotretinoin, a well-known teratogen to which Korean pregnant women are frequently exposed. ChatGPT provides general information on teratogenicity for pregnant women and medical providers rather than on their exposure. Thus, AI can induce unnecessary termination of pregnancy due to misinformation and misperception in cases of notorious teratogens such as isotretinoin. Therefore, counseling on the teratogenicity of medication exposure in pregnancy must be performed with ChatGPT, as well as a literature review. Further studies are required to obtain more individualized information using AI in the field of teratology.