The detection of multiple cancer types with an extended set of methylation and protein markers.

Abstract

3040 Background: Cancer is the 2nd leading cause of mortality in the US, with an estimated 1.96M new cases expected in 2023. With guideline recommended screening unavailable for most cancers, multicancer early detection (MCED) tests in development may extend the reach of cancer screening. As part of an ongoing program examining multiple biomarker classes in cancer early detection, we examined the feasibility of combining blood DNA methylation and cancer protein markers for the detection of multiple cancer types. Methods: 236 samples from subjects with cancer (13 tumor organ types) were retrospectively collected and gender/age matched with 146 unscreened non-cancer control samples. Plasma and serum were used for cfDNA methylation analysis of 36 differentially methylated regions and 8 cancer-associated protein markers in a randomized, blinded fashion. Methylated DNA markers (MDMs) were assayed using target enrichment long-probe quantitative-amplified signal assays. Protein markers were assayed with commercial kits. Normalized methylation values/protein concentrations were used to develop a logistic regression model for cancer detection. The model was built with 5 random 0.7/0.3 training/validation splits, keeping case/control, cancer types, gender, and age distributions similar to the original data set. Models for MDMs and protein markers were created, and a combined model including both MDMs and protein markers was also evaluated for performance improvement. Results: A reduced subset of MDM and protein markers was identified to have robust detection of several cancer types. At 99% specificity, overall sensitivity for cancer detection was 68% (95% CI: 62%-74%) for MDMs, 43% (37%-50%) for proteins, and 75% (69%-80%) for MDMs and proteins combined. Conclusions: Building on promising results with mutation and protein biomarker analysis (Science 369:6499, 2020) this pilot study demonstrates the potential of combining DNA methylation and protein biomarkers and contributes to a larger scale effort to develop a multi-biomarker class MCED test (Ann Oncol 2022 Vol. 33:S575). [Table: see text]