BACKGROUND AND AIMS
Lung cancer is the leading cause of cancer deaths worldwide, with a global estimation of 2.2 million new cases and 1.8 million deaths in 2020.1 Low-dose spiral CT-based screening has shown significant mortality reduction of lung cancer in randomised clinical trials. However, the high false-positive results (around 96%) will lead to high rates of overdiagnosis.2 Liquid biopsy has achieved revolutionary improvements in the early detection of cancers. Here, the authors report the preliminary results of the ASCEND-LUNG trial,3 a prospective case-control study designed to develop early detection models for lung cancer based on multi-omics assays, including cell-free DNA (cfDNA) methylation, mutation, and tumour proteins.
MATERIALS AND METHODS
Blood samples from eligible participants including 230 cancers were prospectively collected from the Department of Thoracic Surgery of Peking University People’s Hospital, Beijing, China, from February 2021 to 15th December 2021. Age-matched non-cancers (n=135) were selected from another study. cfDNA was extracted and sequenced by a customised targeted methylation panel covering approximately 490,000 CpG sites (ELSA-seq, 1000X),4 and an ultradeep target mutation panel containing 168 genes (35000X; matched white blood cells: 10000X). Sixteen tumour proteins were also detected. Early detection models were developed and validated by the support vector machines algorithm, with five-fold cross validation based on multi-omics.
RESULTS
Three early detection models were developed using the methylation, mutation, and protein data of 158 patients with lung cancer, and 135 non-cancer controls, respectively. The specificities of the models were 98.5% (95% confidence interval [CI]: 95.6–100.0%), and 100.0% (95% CI: 97.8–100.0%). The corresponding sensitivities were 72.8% (95% CI: 65.2–79.1%), 18.8% (95% CI: 12.5–26.8%), and 32.1% (95% CI: 25.0–39.7%). A combined model with methylation and protein data could improve the performance of early detection with a specificity of 98.5% (95% CI: 94.8–99.8%) and a sensitivity of 84.6% (95% CI: 78.0–90.0%). Combining all of the data from the three omics yielded comparable results, with a specificity of 98.5% (95.6–100.0%) and a sensitivity of 83.8% (95% CI: 76.6–90.1%). The sensitivities for Stages I–IV were 81.4% (95% CI: 69.1–90.3%), 94.2% (95% CI: 71.3–99.9%), 90.0% (95% CI: 70.8–98.9%), and 85.7% (95% CI: 42.1–99.6%), respectively. The robustness of the multi-omics model was not influenced by the clinical covariates including age, sex, or BMI.
CONCLUSION
In this study, the cfDNA methylation lung cancer detection model showed superior performance compared with the models based on circulating tumour DNA mutation or tumour proteins. The multi-omics detection model including cfDNA methylation and protein could improve sensitivity at a considerably high specificity. However, circulating tumour DNA mutation contributed little to the improvement of the detection performance. This study highlights a potential clinical utility of the multi-omics model with cfDNA methylation and protein markers for detecting lung cancer. The enrollment of validation set is ongoing, and is expected to be completed by March 2023.
