![]()
An approach using large language model (LLM) technology can increase the ability to identify patients with lung cancers after exposure to the September 11 terrorist attacks on the World Trade Center (WTC), suggests a study in the October
Journal of Occupational and Environmental Medicine.
Led by Julia Lo Cascio, MD, of NYU Langone Health—Long Island, the study identifies many previously unrecognized WTC-related lung cancers, including interactions between patient characteristics and outcomes, type of exposure, and cancer-related gene mutations.
Twenty-five years after the WTC attacks, first responders and exposed citizens continue to be diagnosed with cancers related to dust, fires, and toxins resulting from the disaster. These "9/11 cancers" may be related to gene mutations linked to toxin exposure.
While lung cancer is the third-most WTC-related cancer, the relatively few studies of this issue have focused mainly on defined groups of rescue and recovery workers. The researchers designed a study using advanced LLM technology—a type of artificial intelligence—to identify and gather detailed information on lung cancers in patients with WTC exposure.
Applying LLM analysis to clinical notes of patients with primary lung cancers treated at the NYU Perlmutter Cancer Center between 2011 and 2023, the researchers identified 265 patients with documented WTC exposure. The use of LLM nearly tripled the number of WTC-related lung cancers identified. Many patients were not enrolled in existing WTC registries—including programs providing medical benefits to affected individuals.
About 35% of identified patients were first responders, who tend to have shorter-term but high-volume pollutant exposure; while 16% were residents of southern Manhattan, with longer-term but less-potent exposure. The remaining 48% were commuters, with the lowest likelihood of exposure.
Data analysis revealed a complex series of interactions between WTC exposure, patient characteristics, and cancer outcomes. For example, cancer outcomes were worse for women with past or current smoking, but for men who had never smoked. Regardless of other factors, risk of death was three times higher for men compared to women. Risk was higher for residents of the affected area, suggesting that prolonged exposure led to worse outcomes.
On analysis of genetic data in 149 patients, many had mutations of certain cancer-promoter or cancer-suppressor genes. In particular, mutations of the
EGFR gene were most common among residents of lung cancers: 54%, compared to 36% in commuters and 16% in first responders.
The genetic findings "support the claim that there is no singular '9/11 Gene' that automatically links a patient to exposure," the researchers write. Rather, lung cancer-related gene mutations seem to occur at different proportions in the WTC population.
The study provides new information on a wide range of characteristics and outcomes in lung cancer patients with a history of WTC exposure. Using LLMs to analyze information in the medical record can "identify a broader range of patients that may otherwise have exposure to WTC, but not be included in other registries," Dr Lo Cascio and coauthors write. They hope their research "provides unique insights into a population that has been so far understudied and can be applicable to other areas of lung cancer pathogenesis."
About the Author
Dr. Lo Cascio may be contacted for interviews at julia.locascio(at)nyulangone.org.