Development of a multi-cancer early detection test using the CCGA cancer methylation atlas

A noninvasive cell-free DNA (cfDNA) blood test that could detect multiple cancers at early stages may decrease cancer mortality. CCGA (NCT02889978) is a prospective, multi-center, observational, case-control study with longitudinal follow-up to support the development of a plasma cfDNA-based multi-cancer early detection assay.

Over 6,000 methylation sequencing data sets (plasma and tumor tissue samples) were generated using samples from participants with newly diagnosed cancer (20 types, all stages) and matched samples from non-cancer participants. The resulting novel atlas of methylation patterns enabled identification of cancer- and tissue type-defining methylation targets that were used to develop a methylation-based cancer detection test. Briefly, the test employed an automated, targeted methylation sequencing assay using a novel high-efficiency methylation chemistry to enrich for the identified methylation targets, and a machine learning classifier that determined cancer status and tissue of origin (TOO). Here, we evaluated test performance in a pre-specified subset of the CCGA cohort (2,301 participants) for 12 cancer types (anorectal, colorectal, esophageal, gastric, head and neck, hormone receptor negative breast, liver, lung, ovarian, and pancreatic cancers, as well as multiple myeloma and lymphoid neoplasms) that, together, account for approximately 63% of all US cancer deaths.

For each cancer type, this multi-cancer methylation test detected cancer at early stages (I-III) at a very high specificity of 99% (or 1% false positive rate); across cancer types, sensitivity ranged from 59 to 86%. Combined cancer detection was 34%, 77%, and 84% at stages I, II, and III, respectively. TOO was provided for 94% of all cancers detected; of these, TOO was correct in 90% of cases.

Detection of multiple deadly cancers was achieved with a single, fixed, low false positive rate and simultaneously provided accurate TOO localization. These data demonstrate the feasibility, and motivate further clinical development of, a multi-cancer approach to cancer detection.