Abstract 721: Analytical validation of a multi-cancer early detection test with tissue localization using a cell-free DNA-based targeted methylation assay

Introduction: A blood-based test using cell-free DNA (cfDNA) may address an unmet need for earlier detection of multiple cancers. Here, we report the analytical validation of a targeted methylation-based cfDNA test to detect cancer and tissue of origin (TOO). Methods: Since detection was determined by a machine-learning classifier based on methylation patterns of >105 genomic targets, defining analytical sensitivity by a single common limit of detection (LOD) for each genomic target was not applicable. Thus, we characterized sensitivity with respect to tumor fraction using dilution series of contrived cancer samples containing mixtures of cfDNA from 6 individuals without cancer and 6 participants in the Circulating Cell-free Genome Atlas study (CCGA; NCT02889978) with cancer (breast, colorectal, head and neck, lung, lymphoid neoplasm). Each dilution series was performed at 3 levels near the expected LOD95% (lowest admixture fraction predicted to have 95% detection probability). Analytical specificity was determined by the false positive (FP) rate in 1,204 samples without cancer. Reproducibility and repeatability were characterized by within- and between-run variability in 81 cancer samples and 45 noncancer samples across multiple reagent lots, instruments, and operators. Test performance as a function of cfDNA input (3-100 ng) was assessed by an input titration study with cfDNA from 6 CCGA participants with cancer (colorectal, lung, esophageal, renal, multiple myeloma, lymphoid neoplasm) and 5 individuals not known to have cancer. The effect of 4 potential interferents (hemoglobin, bilirubin, triglycerides, white blood cell genomic DNA) on test performance was also evaluated. Results: In the assessment of sensitivity by tumor fraction, LOD95% ranged from 0.2% (upper bound) to 0.5% across the 5 individual solid tumor cases and was 1.9% for the blood cancer case. Among noncancer samples from CCGA participants, 9 FPs were detected (99.3% specificity; 95% CI, 98.6%-99.7%). Repeatability and reproducibility results were correct in 31/34 (91.2%) and 17/17 (100%) sample pairs with or without cancer in within-run tests and in 77/81 (95.1%) and 45/45 (100%) cancer and noncancer samples in between-run tests. For the input titration study, cancer was detected in 5/6 cancer specimens (156/183 samples) and not detected in any noncancer samples (n = 62) across all input cfDNA levels tested. TOO was correctly localized in all but 2 samples with a cancer signal. None of the tested interferents affected test performance. Conclusions: A targeted methylation-based cfDNA test repeatedly and reproducibly detected cancer with high analytical sensitivity and specificity and localized TOO with high accuracy, which are critical for multi-cancer early detection applications. This validation study supports clinical development of this multi-cancer early detection test. Citation Format: Gregory Alexander, Wendy Lin, Madhuvanthi Ramaiah, Byoungsok Jung, Lijuan Ji, Ekaterina Revenkova, Payal Shah, Christian Croisetiere, Jennifer Berman, Lane Eubank, Gunjan Naik, Jacqueline Brooks, Andrea Mich, Seyedmehdi Shojaee, Neda Ronaghi, Hemanshi Chawla, Xinyi Hou, Qinwen Liu, Christopher Yakym, Patriss Wais Moradi, Meredith Halks-Miller, Nathan Hunkapiller, Sonya Parpart-Li, Alexander Aravanis. Analytical validation of a multi-cancer early detection test with tissue localization using a cell-free DNA-based targeted methylation assay [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 721.