Abstract 910P
Background
Liquid biopsy offers an optimal approach for the early detection of lung cancer that is the leading cause of cancer mortality worldwide. Here, we report the results of AssesSment of early-deteCtion basEd oN liquiD biopsy in LUNG cancer (ASCEND-LUNG, NCT04817046), a prospective study designed to develop early detection models for lung cancer based on multi-omics assays including cell-free DNA (cfDNA) methylation, mutation, and tumor proteins.
Methods
Blood samples from lung cancer patients were prospectively collected from Peking University People's Hospital since February 2021 (data cut-off December 15, 2021). Age matched non-cancer controls were selected. A methylation panel of ∼490,000 CpG sites sequenced by the ELSA-seq, a mutation panel of ultradeep sequenced 168 genes, and tumor protein assays were performed. Early detection models were developed by 5-fold cross validation.
Results
In total, the multi-omics data of 158 patients with lung cancer and 135 non-cancer controls were included in the model development. The specificities for methylation, mutation, and protein models were 98.5% (95% CI, 95.6%‒100%), 100% (97.8%‒100%) and 100% (97.8%‒100%), respectively. The sensitivities of them were 72.8% (65.2%‒79.1%), 18.8% (12.5%‒26.8), and 32.1% (25.0%‒39.7%), respectively. Combing all three models could improve the performance of early detection with a sensitivity of 83.8% (76.6%‒90.1%) and a specificity of 98.5% (95.6%‒100%). Table: 910P
Performance of the early detection models for lung cancer
| Early detection models for lung cancer | ||||
| Performance | Multi-omics | Methylation | Mutation | Protein |
| Specificity (95%CI) | 98.5% (95.6‒100%) | 98.5% (95.6‒100%) | 100% (97.8‒100%) | 100% (97.8‒100%) |
| Sensitivity (95%CI) | ||||
| Total | 83.8% (76.6‒90.1%) | 72.8% (65.2‒79.1%) | 18.8% (12.5‒26.8%) | 32.1% (25.0‒39.7%) |
| Stage I | 79.7% (67.8‒89.8%) | 65.0% (53.8‒75.0%) | 1.7% (0‒6.8%) | 25.3% (16.5‒35.4%) |
| Stage II | 77.8% (55.6‒94.4%) | 70.4% (51.9‒85.2%) | 26.3% (10.5‒47.4%) | 34.6% (15.4‒53.8%) |
| Stage III | 96.0% (88.0‒100%) | 87.5% (75.0‒97.5%) | 40.0% (20.0‒60.0%) | 35.0% (20.0‒50.0%) |
| Stage IV | 88.9% (66.7‒100%) | 81.8% (54.5‒100%) | 55.6% (22.2‒88.9%) | 63.6% (36.4‒90.9%) |
Conclusions
In this study, the methylation-based lung cancer early detection model showed superior performance compared with the models based on mutation or protein. The multi-omics combined model can further improve the sensitivity at a considerably high specificity. Our study highlights a potential clinical utility of the multi-omics-based early detection model for lung cancer. The enrollment of validation set is ongoing and expected to be completed by March 2023.
Clinical trial identification
NCT04817046.
Editorial acknowledgement
Legal entity responsible for the study
Kezhong Chen.
Funding
Research Unit of Intelligence Diagnosis and Treatment in Early Non-small Cell Lung Cancer, Chinese Academy of Medical Sciences（2021RU002), National Natural Science Foundation of China (No.82072566 and No.81602001) and Peking University People's Hospital Research and Development Funds (RS2019-01).
Disclosure
All authors have declared no conflicts of interest.