Evaluation of differential contribution of a circulating epithelial cell signal component in a multimodal colorectal neoplasia assay.

Abstract

e15527 Background: Multimodal diagnostic classifiers survey signals from multiple biological compartments and provide iterative, independent and interactive information. Detection of both CRC and AA is critical for noninvasive colorectal screening to improve overall survival and prevent the 2.5-5% annual transition of AA to CRC. FirstSight integrates clinically validated somatic variants from cfDNA and circulating epithelial cells (CECs) adjusting for age and sex. CECs provide information from both intrinsic factors of the adenoma and extrinsic factors such as adenoma microenvironment which facilitates early systemic entry. We sought to assess differential information from CEC signals for the detection of colorectal cancer and/or advanced adenoma (AA). Methods: Blood samples and colonoscopy pathology results were obtained from 438 asymptomatic screening subjects enriched for CRC/AA obtained from 15 US medical centers. The cohort included 18 CRCs and 64 AAs. Somatic variants from cfDNA were identified using NGS and qPCR. CECs were captured by the CellMax biomimetic platform (CMx) using high-avidity EpCAM antibody embedded in the CMx biochip and confirmed with immunostaining (DAPI: nucleus, CK20: epithelial cell and CD45: WBC). CMx platform’s AI/ML analyses of CEC images quantify stain intensities and cellular features. CEC derived signal GM1was evaluated for its predictive capability beyond somatic variants from cfDNA. Results: Genetic or epigenetic variants were not detected in 52% (33/64) of AA cases, limiting the sensitivity of these markers for early disease. However, distinct CEC signals have been identified that aid in the detection of subjects with CRC or AA, or conversely subjects with negative colonoscopies or non-advanced adenomas (nAA). Among them, a novel feature (referred to as GM1) derived from CEC signals was able to differentiate with statistical significance AA from negative/non-neoplastic findings or nAAs in study subjects with negative results in corresponding somatic variants from cfDNA (p < 0.0001). While targeted somatic variants from cfDNA performed well on CRC (17/18), they provided no predictive information for detection of AA for the 353 subjects which were negative for variants, GM1 provided 100.0% (33/33) sensitivity at 35.1% (112/319) specificity, showing its ability to rule out AA with high negative predictive value. Conclusions: Somatic variant detection modes of the First Sight multi-modal assay have high sensitivity for CRC and modest sensitivity for AA. We demonstrate that CEC signal GM1 of FirstSight provides significant independent information for the detection of CRC and AA beyond somatic variants from cfDNA. Additional CEC signals may further improve the sensitivity and specificity for detection of early-stage colorectal neoplasia. The GM1 CEC signal marker will need to be validated further in future studies.