Abstract 909P
Background
Liquid biopsy is promising for cancer early detection to reduce mortality, particularly for those cancers without effective screening paradigms. The Pan-canceR early detectiOn by Multi-omIcS biomarkErs study (PROMISE, NCT04972201) is a prospective multicenter case-control study to assess the performance of the multi-cancer detection blood test (MCDBT) in the early detection of nine cancers in the lung, colorectum, liver, esophagus, pancreas, head and neck, ovary, and biliary tract.
Methods
Blood samples were prospectively collected from cancer patients and non-cancer individuals. A targeted cell-free DNA (cfDNA) methylation panel of ∼490,000 CpG sites, a 168-gene mutation panel, and a 16-protein assay were applied. Participants stratified by age and clinical status were split into training (n = 981) and test sets (n = 492). The MCDBT models were developed on the training set and then validated on the test set.
Results
In the test set, specificities of the methylation, mutation, and protein based MCDBT models were 99.2% (95% CI: 97.0‒99.9%), 99.1% (96.8‒99.9%), and 99.6% (97.7‒100.0%), respectively (Table). The sensitivities for the three MCDBT models were 72.4% (66.5‒77.8%), 51.7% (44.1‒59.2%) and 47.8% (40.8‒54.9%), respectively. The multi-omics MCDBT model combining the methylation, mutation, and protein markers yielded a higher sensitivity of 79.0% (73.5‒83.8%) at a specificity of 97.9% (95.1‒99.3%). The accuracies of the top-one and top-two predicted origins were 75.3% (68.4‒81.3%) and 90.9% (85.8‒94.6%), respectively. Table: 909P
Performance of the MCDBT models
| Performance | MCDBT models | |||
| Multi-omics | Methylation | Mutation | Protein | |
| Specificity (95% CI) | 97.9% (95.1‒99.3%) | 99.2% (97.0‒99.9%) | 99.1% (96.8‒99.9%) | 99.6% (97.7‒100%) |
| Sensitivity (95% CI) | ||||
| Total | 79.0% (73.5‒83.8%) | 72.4% (66.5‒77.8%) | 51.7% (44.1‒59.2%) | 47.8% (40.8‒54.9%) |
| Stage I | 61.9% (48.8‒73.9%) | 50.8% (37.9‒63.6%) | 33.3% (19.6‒49.5%) | 32.1% (19.9‒46.3%) |
| Stage II | 70.2% (56.6‒81.6%) | 63.2% (49.3‒75.6%) | 38.9% (23.1‒56.5%) | 38.6% (24.4‒54.5%) |
| Stage III | 86.7% (76.8‒93.4%) | 80.0% (69.2‒88.4%) | 53.7% (39.6‒67.4%) | 50.8% (38.1‒63.4%) |
| Stage IV | 95.2% (86.5‒99.0%) | 93.5% (84.3‒98.2%) | 75.0% (60.4‒86.4%) | 72.1% (56.3‒84.7%) |
Conclusions
The methylation-based MCDBT model exhibited a superior sensitivity at a high specificity in the early detection of nine cancers. The multi-omics model improved sensitivity with a minimal impact on specificity. A further expanded study (NCT04817306) is currently recruiting.
Clinical trial identification
NCT04972201.
Editorial acknowledgement
Legal entity responsible for the study
Qiang Gao.
Funding
Guangzhou Burning Rock Dx Co., Ltd.
Disclosure
C. Wang, X. Yang, S. Fang, Y. Zhang, G. Wang, F. Liu, X. Wen, J. Zhao, G. Zhou, B. Li, S. Cai, Z. Zhang: Financial Interests, Institutional, Full or part-time Employment: Burning Rock Biotech. All other authors have declared no conflicts of interest.