Blood-based DNA methylation marker model for short-term and long-term lung cancer risk prediction
Source: PubMed Central Open Access, NCBI / U.S. National Library of Medicine
Background Screening heavy smokers with low-dose computed tomography (LDCT) has been shown to reduce lung cancer (LC) mortality, however, identifying the specific high-risk population that benefits most, a critical requirement for implementing effective and cost-efficient screening, remains challenging. Methods We developed and validated a blood-based DNA methylation marker model (BBDMM) for all participants, including both ever and never smokers, using the LC risk-informative CpG sites from epigenome-wide association studies (EWAS). The model was developed and internally validated in 2,459 participants from ESTHER, a population-based cohort from Germany. Subsequently, BBDMM was externally validated in exactly same 233 participants drawn from the Norwegian HUNT2 and HUNT3 cohorts with long- and short-term follow-ups and cases identified up to 18 and 6.7 years before LC diagnosis, respectively. Results The BBDMM predicted LC incidence with an area under the curve (AUC) of 0.84 [95% confidence interval (95% CI), 0.80-0.87] in the derivation set. In the independent external validation sets, AUCs of 0.85 (95% CI, 0.80-0.90) and 0.85 (95% CI, 0.80-0.90) were observed in HUNT2 and HUNT3, respectively. Conclusions The BBDMM identified future lung cancer cases with promising potential and the model discrimination was highly stable at different time points. These markers may contribute to the evolution of a blood-based test for predicting LC risk. Graphical abstract Using prediagnosti
Abstract
Background Screening heavy smokers with low-dose computed tomography (LDCT) has been shown to reduce lung cancer (LC) mortality, however, identifying the specific high-risk population that benefits most, a critical requirement for implementing effective and cost-efficient screening, remains challenging. Methods We developed and validated a blood-based DNA methylation marker model (BBDMM) for all participants, including both ever and never smokers, using the LC risk-informative CpG sites from epigenome-wide association studies (EWAS). The model was developed and internally validated in 2,459 participants from ESTHER, a population-based cohort from Germany. Subsequently, BBDMM was externally validated in exactly same 233 participants drawn from the Norwegian HUNT2 and HUNT3 cohorts with long- and short-term follow-ups and cases identified up to 18 and 6.7 years before LC diagnosis, respectively. Results The BBDMM predicted LC incidence with an area under the curve (AUC) of 0.84 [95% confidence interval (95% CI), 0.80-0.87] in the derivation set. In the independent external validation sets, AUCs of 0.85 (95% CI, 0.80-0.90) and 0.85 (95% CI, 0.80-0.90) were observed in HUNT2 and HUNT3, respectively. Conclusions The BBDMM identified future lung cancer cases with promising potential and the model discrimination was highly stable at different time points. These markers may contribute to the evolution of a blood-based test for predicting LC risk. Graphical abstract Using prediagnostic blood samples from participants of large population-based cohorts from Germany and Norway, we identified, evaluated and validated a blood-based DNA methylation marker model with equally good prediction for long-term and short-term lung cancer risk. http://www.w3.org/1999/xlink anchor d33e304 portrait 12916_2026_4973_Figa_HTML.jpg anchor Figa portrait Par5 Abs1