8286314 A Large language model for identifying industry-funded research on the carcinogenicity of benzene and cobalt metal

<h3>Objective</h3> Industry-funded research is a threat to the validity of scientific inference on carcinogenic hazards. Scientists require tools to better identify industry-funded studies and minimize the influence of industry bias in evidence synthesis reviews. We applied a novel large language model (LLM)-based tool named InfluenceMapper to demonstrate and evaluate its performance in identifying industry sponsored research on the carcinogenicity of two occupational exposures of commercial interest by major industries, benzene and cobalt. <h3>Material and Methods</h3> We identified all epidemiological, animal bioassay, and mechanistic studies included in systematic reviews on the carcinogenicity of benzene and cobalt conducted by the WHO/IARC Monographs Programme. InfluenceMapper extracted sponsoring entities disclosed in study publications and classified 40 possible relationship types between all entities and the study and each author. A human classified entities as industry or industry-funded. Positive predictive values described the extent of false positive relationships. <h3>Results</h3> Analyses included 1,520 studies for both agents. We identified 184 disclosed industry or industry-funded entities from InfluenceMapper output that were involved in 453 distinct study-entity and author-entity relationships. For each agent, between 4-8% of studies were funded by industry and 1-4% of studies had at least one author that disclosed receiving industry funding. Industry trade associations funded 19 studies published in 14 journals over a 37-year span. After funding, the most prevalent disclosed relationships with industry were receiving data, holding employment, paid consulting, and providing expert testimony. Industry sponsored research consisted predominantly of mechanistic studies. Positive predictive values were excellent (&gt;98%) for study-entity relationships but declined for relationships with individual authors. <h3>Conclusion</h3> LLM-based tools can significantly expedite and bolster the detection of industry sponsored research in cancer prevention. Possible use cases include facilitating the assessment of bias from industry studies in evidence synthesis reviews and alerting scientists to the influence of industry on scientific inference about carcinogenic hazards.