Abstract 295P
Background
Current lung cancer (LC) screening methods, like low-dose CT (LDCT), face challenges with high false positives and low compliance. Circulating tumor DNA (ctDNA) tests show promise but are costly and lack sensitivity for early-stage tumors. This study presents a novel LC detection method leveraging multiple LC-specific features in cell-free DNA (cfDNA) through a cost-effective shallow genome-wide sequencing approach. We prospectively evaluated this assay in LC screening-eligible individuals and symptomatic patients referred from primary care settings.
Methods
A multimodal ctDNA assay with shallow sequencing coverage (0.5x)was developed, integrating fragmentomic, nucleosome, end-motif, and copy number alteration analyses from a single blood sample. A machine-learning model was trained using a retrospective case-control cohort of 157 LC patients and 239 healthy controls. The performance of assay was prospectively validated in an external cohort of 67 high-risk and LC screening-eligible individuals.
Results
The model showed robust performance with an AUC of 0.97, achieving 90% sensitivity and 92% specificity in retrospective cohorts. Sensitivity for early-stage tumors (stage I/II) was 75%, surpassing hotspot mutation-based and multi-cancer assay (SPOT-MAS). In the prospective cohort, 67 participants with symptoms suggestive of LC underwent primary imaging. Among them, 39 showed no suspected lesions, all correctly predicted as negative by the multi-feature model. The remaining 28 participants, identified with suspected lesions, underwent contrast-enhanced CT, which yielded 4 negative and 24 positive imaging results. Of these, 7 cases were confirmed as LC, while 17 were diagnosed with benign lesions. The assay detected cancer signals in 5 of the 7 LC cases, achieving 71.4% sensitivity, 100% specificity, and a 100% positive predictive value (5/5), significantly reducing falsepositive rates compared to LDCT.
Conclusions
This multimodal, shallow-depth cfDNA assay offers a noninvasive, cost-effective, and accurate approach for LC detection, addressing key limitations of current screening methods like LDCT, such as high false positives and low compliance. It shows promise as a complementary tool for early LC detection.
Legal entity responsible for the study
Medical Genetics Institute.
Funding
Gene Solutions.
Disclosure
All authors have declared no conflicts of interest.