Uncovering new psychoactive substances research trends using large language model-assisted text mining (LATeM)

The emergence of new psychoactive substances (NPS) has become a significant public health concern over the past few decades. This study employed text-mining techniques and large language models (LLMs) to examine NPS research trends from 1990 to 2024. Over 12,000 publications from the Web of Science database were analyzed using the proposed Large Language-assisted Text Mining (LATeM) method to identify key patterns and shifts in research focus. The results indicated an evolution in NPS research from basic pharmacological studies in the 1990s to more diverse approaches encompassing public health, policies, and advanced analytical methods in recent years. The opioid crisis, particularly the emergence of fentanyl, has significantly influenced research priorities since 2020. A notable increase in NPS-related publications was observed, growing from 105 in the 1990s to over 6000 in the 2020s. Furthermore, the research focus shifted from traditional drugs to synthetic cannabinoids and stimulants in the 2000s and 2010s, and subsequently to opioids in the 2020s. Emerging areas of study include forensic applications, environmental impacts, and advanced analytical techniques. This study demonstrates the efficacy of integrating LLMs with text mining to identify emerging trends in NPS research. Our findings underscore the need for continued interdisciplinary collaboration and the development of novel strategies to address the evolving NPS challenges. Future research should prioritize enhancing early detection methods, investigating the long-term health consequences of NPS use, and developing effective prevention and treatment programs.