Comprehensive Analysis of Food Allergy Research Trends via Large Language Model–Assisted Text Mining (1980–2024)

Food allergy prevalence has risen globally, yet conventional reviews provide limited, long-horizon, quantitative views [1]. We therefore analyzed the 1980–2024 literature using large language model–assisted text mining (LATeM) to map temporal shifts at scale [2]. We analyzed 14,445 English-language articles retrieved from the Web of Science database. The search criteria are detailed in Table S1. A key innovation of LATeM is the use of an LLM to augment topic interpretation and synthesis. In brief, LATeM proceeds as follows: article retrieval and preprocessing; LDA topic modeling with coherence-based model selection; LLM-assisted labeling/interpretation of topics; and expert verification (Figure 1A). Core parameters (e.g., number of topics = 5, passes = 100, α/β = auto) and the software environment are detailed in Supplementary Methods (Appendix S1: S1–S5). Prompt structures and validation steps are summarized in Table S2 and Supplementary Method S7 (Appendix S1). As a sensitivity check, we replicated temporal counts in PubMed (1980–2024; harmonized query) and observed perfect concordance with WoS across 5-year bins (Spearman ρ = 1.00, p < 0.0001); details appear in Figure S1. Temporal analysis revealed a clear chronological evolution of research priorities. Core terms (“food,” “allergy”) remained dominant, but a significant shift was evident: early research (1980s–1990s) focused on fundamental epidemiology, whereas the 2000s saw a marked increase in terms related to severe reactions (“anaphylaxis”) and diagnostics (“skin prick test”). Since 2010, advanced clinical practices (“oral food challenge”), regulatory concepts (“precautionary allergen labeling”), and sophisticated statistical methods (“odds ratio”, “confidence interval”) became prominent. This trajectory reflects an increasing emphasis on quantitative evaluation and evidence-based risk management (Tables S3–S5 and Figure S2). LDA modeling identified five coherent themes across 5-year bins (absolute counts in Figure S3); to contextualize proportional shifts, we annotated total publications above each bar in Figure 1B. Beyond major themes, representative subtopic signals were evident across layers: (1) foods/proteins: “peanut,” “ARA,” and protein/epitope terms feature prominently among high-growth or newly emerged items in later periods (Tables S4 and S5); (2) prevention/treatment: “oral”/“challenge,” “epinephrine,” and “dupilumab” align with shifts toward active management and biologics (Tables S4 and S5); (3) and immune responses: “Th2/Th17,” “cytokines,” and “expression” characterize mechanistic evolutions within immune-focused literature (Table S6). This normalized analysis illustrates a substantial shift in research focus. While foundational research on allergen proteins dominated early periods, the proportion of research dedicated to pediatric anaphylaxis management has surged markedly since the mid-2000s. This pronounced increase likely reflects the rising prevalence of pediatric food allergies and collective efforts to standardize emergency protocols and develop guidelines for schools and caregivers [3]. The recent literature (2020–2024) reflects further specialization. Co-occurrence network analysis (Figure 2) revealed tightly interconnected thematic clusters, including a Pediatric/Quality of Life cluster, a clinical comorbidity cluster (e.g., “asthma,” “immunotherapy”), and a distinct European Regulatory/Policy cluster (e.g., “EFSA,” “regulation”). The emergence of this regulatory cluster highlights the increasing importance of regulatory science as a specialized frontier. As food supply chains have globalized, harmonized, evidence-based frameworks—underpinned by population-level eliciting dose data [4]—like the Voluntary Incidental Trace Allergen Labeling (VITAL) program, have become essential to inform international risk assessment standards and manage cross contamination risks [5]. Furthermore, the LATeM approach was instrumental in identifying the impact of external societal factors by linking emergent terms like “COVID-19” and “food_insecurity” in the underlying abstracts—associations not readily apparent through frequency analysis alone. Taken together, shifts across anaphylaxis/epinephrine, oral food challenge/diagnosis, and EFSA/VITAL indicate a transition beyond descriptive epidemiology toward practice- and policy-integrated scholarship. In conclusion, food allergy research has evolved from foundational epidemiology into a sophisticated, multidisciplinary field encompassing advanced diagnostics, management strategies, and public health policy. The pronounced relative increase in pediatric anaphylaxis management and the emergence of regulatory science discourse are key developments of the past 15 years. These findings align with expert reviews [6]. This is the first rigorous, LLM-assisted bibliometric analysis of this scale in the field. This study demonstrates that LATeM provides a robust, data-driven framework for mapping research priorities, offering valuable insights to guide future scholarship and inform policies promoting food safety. Yoshiyuki Kobayashi led the study conceptualization, methodology, and project administration, performed the formal analysis and validation, and wrote the original draft of the manuscript. Takahiro Inoue conducted the investigation, contributed to writing the original draft, and reviewed and edited the manuscript. Itsuki Kageyama, Yusuke Iwasaki, Rie Ito, Kota Kodama, and Hiroshi Akiyama conducted the investigation and reviewed and edited the manuscript. This work was supported by JSPS KAKENHI Grant Number JP25K18320 and 23K07902, and the Ministry of Health, Labour and Welfare of Japan Grant Number 23KA1007. This work was supported by the Ministry of Health, Labour and Welfare (23KA1007) and Japan Society for the Promotion of Science (JP25K18320 and 23K07902). Kota Kodama has received research donations for their institution from Kumagaya Gumi, Sumitomo Life Insurance, Merge System, Yasuhara Chemical, Tamai Investment Education, and Ernst & Young ShinNihon. All other authors declared no conflicts of interest for this work. The data that support the findings of this study are available in the Supporting Information of this article. Appendix S1: all70169-sup-0001-Supinfo.docx. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.