Abstract 112: Analytical performance of a cfDNA-based targeted methylation multi-cancer early detection test for population-scale screening

Abstract

Abstract Introduction: cfDNA-based tests have the potential to facilitate earlier detection of cancer and provide greater opportunity for curative intervention. Previously, we demonstrated that a cfDNA-based targeted methylation multi-cancer early detection (MCED) test can detect multiple cancers across all stages with high specificity, and predict the signal origin (SO; ie, tissue of origin) with high accuracy; we also validated the repeatability and reproducibility of its performance. Here, we validate a recent version of this MCED test planned for use as a screening tool. Methods: In the Circulating Cell-free Genome Atlas study (NCT02889978), cfDNA was analyzed using targeted methylation sequencing and machine learning to classify samples as cancer or non-cancer, and predict SO. Dilutions of cfDNA samples from participants with early-stage lung, head and neck, sarcoma, upper gastrointestinal, colorectal, or ovarian cancer into cfDNA from individuals without cancer were analyzed to establish the limit of detection (LOD95%) defined by the lowest variant allele frequency (VAF) at which accurate classification of cancer signals was attained in at least 95% of replicates. Analytical specificity was assessed by the true negative rate in 66 non-cancer cfDNA samples. Classification performance was evaluated as a function of cfDNA input (0.5-100 ng) using cancer samples (liver/bile duct, lung, head and neck, ovarian, breast, bladder/urothelial, uterine). Repeatability within-run and reproducibility between-runs were characterized by pairwise comparisons among 53 cancer samples (anorectal, breast, head and neck, uterine) and among 62 non-cancer samples processed across multiple reagent lots, instruments, and operators. The effects of high levels of potential interferents hemoglobin, bilirubin, triglycerides, and genomic DNA on classification when spiked into 88 plasma samples along with cfDNA from individuals with cancer (liver/bile duct or lung) or when spiked into 87 non-cancer samples were evaluated. Results: LOD95% was 0.11% VAF. No false positives were detected in non-cancer samples (100% analytical specificity). Across all input cfDNA levels tested, in 50/50 (100%) cancer samples, a cancer signal was detected with a correct SO prediction. All sample pairs in within-run (n=110) and between-run (n=696) analyses were concordant with respect to cancer/non-cancer classification and SO prediction (100% repeatability and reproducibility). None of the tested interferents affected cancer signal detection (100% correct) or SO prediction (100% accuracy). Conclusions: These results suggest that the analytical performance of a cfDNA-based targeted methylation MCED screening test is robust for clinical implementation. This MCED test can be a complementary tool to the existing repertoire of cancer screening options currently available for clinical use. Citation Format: Gregory E. Alexander, Byoungsok Jung, Lijuan Ji, Ekaterina Revenkova, Payal Shah, Jacqueline Brooks, Jeremy Carter, Zhao Dong, Lane Eubank, Maryam Hosseini, Xinyi Hou, Hannah Kiarie, Neda Ronaghi, Fabian E. Ortega, Madhuvanthi Ramaiah, Kate Rhodes, Rita Shaknovich, Seyedmehdi Shojaee, Sonya Parpart-Li, Nathan Hunkapiller. Analytical performance of a cfDNA-based targeted methylation multi-cancer early detection test for population-scale screening [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 112.