Biological, environmental, and socioeconomic factors contribute to lung cancer risk across racial groups. However, current screening guidelines from the US Preventive Services Task Force (USPSTF) are based on age and pack-years of tobacco smoking exposure, which often contributes to disparities in early detection, particularly among Black patients.
Despite having lower cumulative tobacco pack-years compared with white individuals, Black Americans experience higher lung cancer incidence and mortality rates.
Results from a recent study presented Saturday, September 6, at 2025 World Conference on Lung Cancer (WCLC) in Barcelona, confirmed promising findings validating Sybil, a deep-learning artificial intelligence (AI) model, in a diverse lung cancer screening cohort.
According to the data presented, Sybil demonstrated minimal bias and strong performance in predicting 6-year lung cancer risk from baseline low-dose computed tomography (LDCT), with consistent results across all groups.
“This validation study confirms that Sybil performs well in racially and socioeconomically diverse settings, supporting its broader utility for lung cancer screening,” said investigator Mary Pasquinelli, DNP, APRN, FNP-BC, Director of the Lung Screening and Early Detection Program at the University of Illinois Chicago (UI Health).
The study, Validation of Sybil in a Minority Population: Results from the Sybil Implementation Consortium: UIC, MGH, Baptist, Wellstar, will be discussed during the Precision in Lung Cancer Screening session on Monday, September 8. The session will begin at 15:30 CEST in Room 09. The session will also include discussions on the latest in management of incidentally detected pulmonary nodules and of preliminary results from the Taiwan National LDCT Lung Cancer Screening program, among other recent studies.
Precision in Lung Cancer Screening
15:30 CEST, Monday, September 8 • Room 09
Join Mary Pasquinelli, DNP, APRN, FNP-BC, and other experts as they present cutting-edge approaches to lung cancer detection and nodule management using innovative technologies across diverse populations. LEARN MORE
Background
Sybil was developed to predict an individual’s future risk of developing lung cancer within 1 to 6 years using a single LDCT scan. It requires no clinical data or manual annotation and uses an open-source algorithm.
The study aimed to validate the performance of Sybil in a diverse lung screening cohort at the University of Chicago (UIC). Researchers evaluated 2,092 baseline LDCTs between 2014 and 2024 from UI Health’s lung screening program.
Of those enrolled, 68 patients were diagnosed with lung cancer, with follow-up periods ranging from 0 to 10.2 years. Prior Sybil validations in the US were conducted in cohorts that were more than 90% white. However, this new analysis focused on a population where 62% of participants identified as non-Hispanic Black, 16% as non-Hispanic white, 13% as Hispanic, and 4% as Asian.
If a lung cancer screening model has an area under the curve (AUC) of 0.94, there is a 94% chance that the model will correctly rank a randomly chosen patient who develops cancer in the future as higher risk than a randomly chosen patient who does not develop cancer.
The study results showed that Sybil’s AUC performance from years 1 through 6 were reported as follows:
- Year 1: AUC = 0.94
- Year 2: AUC = 0.90
- Year 3: AUC = 0.86
- Year 4: AUC = 0.85
- Year 5: AUC = 0.80
- Year 6: AUC = 0.79
Additionally, the results remained strong when limited to Black participants and after excluding cancers diagnosed within 3 months.
Future Implications
The Sybil Implementation Consortium is led by Lecia V. Sequist, MD, MPH, who is Program Director of the Cancer Early Detection and Diagnostics Clinic and Landry Family Professor of Medicine at Harvard Medical School, Boston, In addition to Dr. Pasquinelli, collaborating investigators include Raymond U. Osarogiagbon, MBBS, FACP, Chief Scientist for Baptist Memorial Health Care and Director of the Multidisciplinary Thoracic Oncology Program and the Thoracic Oncology Research (ThOR) Group at Baptist Cancer Center, Memphis, Tennessee, and William Mayfield, MD, Thoracic Surgery Medical Director for the Lung Cancer Screening Program and Incidental Lung Nodule Program at WellStar Health System, Marietta, Georgia. Dr. Pasquinelli said the consortium will now move forward with prospective clinical trials to integrate the model into real-world lung cancer screening workflows. She said future implementation could include:
- Tailoring lung cancer screening intervals based on Sybil scores
- Personalized population lung screening strategies (similar to colon cancer screening)
- Preventive trials
Lung cancer screening uptake and outcomes vary across racial and socioeconomic groups, with Black patients historically experiencing higher mortality rates and lower screening participation.
“[Sybil] shows promise as a tool for improving early detection and addressing disparities in lung cancer outcomes,” Dr. Pasquinelli said.
Its consistent performance across diverse, real-world settings demonstrates how AI tools like Sybil could transform the current landscape of lung cancer screening and potentially close existing racial and socioeconomic gaps, improving outcomes across a wide range of patient populations, she said.
