IARC 60th Anniversary - 19-21 May 2026
Session : Lung Cancer Screening, Early Detection, and Prevention: Addressing the Leading Cause of Cancer Deaths
A multi-ancestry polygenic risk score improves lung cancer risk stratification across diverse populations
NATARAJAN P. 17,21,22, ALDRICH M. 16, PARRA E. 11, KACHURI L. 4, ZHANG H. 19, HUNG R. 2,10, ADLER N. 1,2, CHEN T. 3, FU M. 4, DAI J. 14, TRUONG B. 17, BETTI M. 16, HELLWEGE J. 16, KOYAMA S. 17, BYUN J. 5, AMOS C. 5, KARTSONAKI C. 20, WALTERS R. 20, LAN Q. 6, CHRISTIANI D. 18, JOHANSSON M. 7, MCKAY J. 7, LANDI M. 6, LIU G. 8,12,13, LE MARCHAND L. 9, INTERNATIONAL LUNG CANCER CONSORTIUM T. 7
1 Department of Anthropology, University of Toronto, Toronto, Canada; 2 Lunenfeld-Tanenbaum Research Institute, Sinai Health, Toronto, Canada; 3 Department of Biostatistics, Harvard T.H. Chan School of Public Health, Boston, United States; 4 Stanford University, Stanford, United States; 5 University of New Mexico, Albuquerque, United States; 6 National Cancer Institute, Bethesda, United States; 7 Genomic Epidemiology Branch, International Agency for Research on Cancer, World Health Organization, Lyon, France; 8 Division of Medical Oncology and Hematology, Princess Margaret Cancer Centre, Toronto, Canada; 9 Epidemiology Program, University of Hawaii Cancer Centre, Honolulu, United States; 10 Epidemiology Division, Dalla Lana School of Public Health, University of Toronto, Toronto, Canada; 11 Department of Anthropology, University of Toronto Mississauga, Mississauga, Canada; 12 Departments of Medical Biophysics, Pharmacology and Toxicity, and IMS, University of Toronto, Toronto, Canada; 13 Division of Biostatistics, Dalla Lana School of Public Health, University of Toronto, Toronto, Canada; 14 Department of Epidemiology, Nanjing Medical Univeristy, Nanjing, China; 15 Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, United States; 16 Division of Genetic Medicine, Vanderbilt University Medical Center, Nashville, United States; 17 Broad Institute of MIT and Harvard, Cambridge, United States; 18 Harvard T.H. Chan School of Public Health, Boston, United States; 19 Division of Cancer Epidemiology and Genetics, National Cancer Institute, Bethesda, United States; 20 Nuffield Department of Population Health, University of Oxford, Oxford, United Kingdom; 21 Cardiology Division, Massachusetts General Hospital, Harvard Medical School, Boston, United States; 22 Center for Genomic Medicine, Department of Medicine, Massachusetts General Hospital, Boston, United States
Background: Currently implemented lung cancer risk prediction models consider a limited set of risk factors and insufficiently stratify patients for screening, leaving a large proportion of lung cancer patients, particularly those from non-European ancestries, ineligible for low-dose CT scans. Polygenic risk scores (PRS) have demonstrated strong potential for improving lung cancer risk stratification but are predominantly derived from participants of European-ancestry, limiting their applicability to other populations.
Objectives: We aimed to develop a novel multi-ancestry PRS for lung cancer and evaluate its risk prediction and stratification performance in diverse cohorts.
Methods: We trained four multi-ancestry PRS models (PRS-CSx, JointPRS, CT-SLEB, PROSPER) using genome-wide association study (GWAS) summary statistics from European (43,421 cases 402,179 controls), East Asian (12,769 cases and 262,269 controls), and African American (4,617 cases and 69,020 controls) populations. We benchmarked and validated each PRS in the All of Us cohort, a multi-ancestry cohort consisting of 1,603 lung cancer cases and 352,217 controls. Based on odds ratio per standard deviation (OR per SD) and area-under the curve (AUC) evaluation metrics, we selected and further validated the best-performing PRS in the Nanjing Lung Cancer Cohort (NJLCC, East Asian participants), Vanderbilt University Medical Centre (VUMC, African American participants), and Mass General Brigham Biobank (MGBB, primarily European participants) to assess PRS portability. Lastly, by incorporating the PRS into an absolute risk model that accounts for smoking status, we evaluated how PRS risk strata affect the age at which All of Us participants meet a 6-year 1.51% screening threshold.
Results: PRS-CSx demonstrated the best performance, with an overall OR of 1.44 (95% CI: [1.38, 1.52]) and covariate-adjusted AUC (conditional on sex, age, and the top 16 principal components) of 0.60 (95% CI: [0.59, 0.61]). PRS-CSx also had comparable performance within ancestry groups (European OR = 1.49, 95% CI: [1.41, 1.58]; African OR = 1.36, 95% CI: [1.21, 1.53]; East Asian OR = 1.39, 95% CI: [0.98, 1.98]; Admixed American OR = 1.31, 95% CI: [1.07, 1.61]; Other OR = 1.29, 95% CI: [1.04, 1.60]). Similar results were observed in the additional datasets; MGBB European OR = 1.26, 95% CI [1.21, 1.32], NJLCC OR = 1.43, 95% CI [1.34, 1.53], VUMC OR = 1.18, 95% CI [1.08, 1.29]. Furthermore, in the All of Us cohort, ever-smokers in the highest quintile of the PRS distribution reached the 1.51% 6-year risk threshold of developing lung cancer 3-5 years earlier than ever-smokers with average genetic risk (40-60% PRS percentile) in four of five ancestry groups.
Conclusions: Integrating a multi-ancestry PRS into a lung cancer risk prediction model improves patient stratification, addressing disparities in screening eligibility across diverse populations. This approach has the potential to improve early detection and reduce lung cancer mortality equitably across ancestry groups.