Abstract 3369: Utility of ctDNA-based targeted methylation MRD assay for hematological malignancies

Introduction: Minimal or measurable residual disease (MRD) is an emerging independent predictor of progression-free and overall survival in several hematologic diseases. Despite many different methods used in clinical trials, few assays have clinical utility as a surrogate endpoint in a limited number of hematologic malignancies. MRD assays for lymphomas are mostly performed in research settings due to lack of harmonized methods, heterogeneity of diseases, and need for a primary tumor sample. We explored the feasibility of a pan hematologic malignancy classifier (“pan heme classifier”) based on GRAIL’s methylation platform as a potential tumor-agnostic plasma-based MRD assay for hematologic malignancies. Methods: First, 428 plasma samples from various hematologic malignancies including diffuse large B-cell lymphoma (DLBCL), follicular lymphoma (FL), mantle cell lymphoma (MCL), acute myeloid leukemia (AML), chronic lymphocytic leukemia (CLL), and multiple myeloma (MM) were blindly tested retrospectively using the pan heme classifier. The majority of samples (375/428; 88%) were from relapsed/refractory (R/R) disease. Cancer signal detection and signal origin prediction were explored at various detection thresholds. Next, sensitivity of the pan heme classifier was evaluated using 74 (22 and 14 unique patients in CLL and MCL, respectively) post-treatment samples with orthogonal methods (flow cytometry in CLL and clonoSEQ in MCL). Finally, dilution experiments were carried out in 4 DLBCL and 4 CLL patient samples. Results: Of 428 samples, 408 (95%) passed quality control. At a prespecified detection threshold, we observed high (91%) cancer detection rate with high cancer signal origin accuracies: 100% in CLL, >98% in MM, >95% in non-Hodgkin lymphoma (DLBCL, FL, MCL), and 87% in AML samples. As expected, targeting lower detection thresholds resulted in increased cancer-positive calls. Cancer was reproducibly detected in 48 of 54 (89%) cases where paired samples were taken prior to treatment (screening and cycle 1, day 1), demonstrating high biologic reproducibility. Analysis of posttreatment CLL and MCL samples with cell-based orthogonal MRD assays suggests the current assay limit of detection (LOD) is ~10-3-10-4 MVAF (a methylation-based tumor fraction estimate). These data suggest that cancer signal score is strongly correlated to the observed MVAF. Furthermore, serially diluted DLBCL and CLL patient plasma samples were spiked into healthy volunteer plasma sam