Blood-based DNA methylation markers for lung cancer prediction

• Published in: BMJ Oncology
• Main Authors: Onwuka, Justina Ucheojor, Guida, Florence, Langdon, Ryan, Johansson, Mikael, Severi, Gianluca, Milne, Roger L, Dugué, Pierre-Antoine, Southey, Melissa C, Vineis, Paolo, Sandanger, Torkjel, Nøst, Therese Haugdahl, Chadeau-Hyam, Marc, Relton, Caroline, Robbins, Hilary A., Suderman, Matthew, Johansson, Mattias
• Other Authors: Cancer Research UK, US National Cancer Institute
• Format: Article in Journal/Newspaper
• Language: English
• Published: BMJ 2024
• Subjects: Norway; Northern Sweden
• Online Access: http://dx.doi.org/10.1136/bmjonc-2024-000334; https://syndication.highwire.org/content/doi/10.1136/bmjonc-2024-000334

Objective Screening high-risk individuals with low-dose CT reduces mortality from lung cancer, but many lung cancers occur in individuals who are not eligible for screening. Risk biomarkers may be useful to refine risk models and improve screening eligibility criteria. We evaluated if blood-based DNA methylation markers can improve a traditional lung cancer prediction model. Methods and analysis This study used four prospective cohorts with blood samples collected prior to lung cancer diagnosis. The study was restricted to participants with a history of smoking, and one control was individually matched to each lung cancer case using incidence density sampling by cohort, sex, date of blood collection, age and smoking status. To train a DNA methylation-based risk score, we used participants from Melbourne Collaborative Cohort Study-Australia (n=648) and Northern Sweden Health and Disease Study-Sweden (n=380) based on five selected CpG sites. The risk discriminative performance of the methylation score was subsequently validated in participants from European Investigation into Cancer and Nutrition-Italy (n=267) and Norwegian Women and Cancer-Norway (n=185) and compared with that of the questionnaire-based PLCOm2012 lung cancer risk model. Results The area under the receiver operating characteristic curve (AUC) for the PLCOm2012 model in the validation studies was 0.70 (95% CI: 0.65 to 0.75) compared with 0.73 (95% CI: 0.68 to 0.77) for the methylation score model ( P difference =0.07). Incorporating the methylation score with the PLCOm2012 model did not improve the risk discrimination (AUC: 0.73, 95% CI: 0.68 to 0.77, P difference =0.73). Conclusions This study suggests that the methylation-based risk prediction score alone provides similar lung cancer risk-discriminatory performance as the questionnaire-based PLCOm2012 risk model.