Background: ctDNA has the potential to identify patients (pts) with early stage cancer; however, current assays are challenged by limited sensitivity (~50%), reliance on a single analyte (e.g. somatic mutation detection), and/or the need for tumor tissue or genomic DNA sequencing to interpret ctDNA results. Recent studies have demonstrated that ctDNA can be detected using other biomarkers including DNA methylation. We developed a technology in which both somatic mutations and epigenomic alterations can be analyzed in a single assay.

Methods: Using a large database of cell-free DNA (cfDNA) profiles generated from advanced cancer patients, we designed a targeted sequencing assay that detects somatic variants, methylation alterations, and other epigenomic variations at transcription factor binding sites associated with CRC. Total cfDNA was extracted, partitioned based on methylation level, and analyzed. Data were then filtered using a variant classifier to differentiate tumor- from non-tumor-derived alterations without a priori knowledge of tissue or germline sequencing results. A machine learning model was trained on 111 cfDNA samples from 38 late stage and 10 early stage CRC pts and 63 age-matched cancer-free controls. For the independent test set, plasma samples (4-5mL) were collected from 72 pts with stage I-IV CRC prior to and 4 weeks after (N = 50, total of 122 samples) surgical resection. 35 age-matched cancer-free controls were similarly analyzed in the test set.

Results: Of the 72 pts, 62.5% were male, and median age at CRC diagnosis was 61.5 years (range 36-85). Stage distribution was 52.8% stage I/II, 40.3% stage III, and 6.9% stage IV. In the 50 pts with post-surgical samples, clinical follow-up was available for 49 (median post-surgery follow-up: 314 days; range 15-472). Utilizing this assay, pre-surgery ctDNA detection rate was 94% (68/72); 97% in stage I/II, 90% in stage III, and 100% in stage IV. Epigenomic analysis significantly enhanced ctDNA detection relative to somatic mutational analysis alone (94% vs. 56%; p<0.0001). Specificity in age-matched cancer-free controls was 94%.

Discussion: Utilizing a plasma-only sequencing assay incorporating somatic genomic variant detection, epigenomic analysis, and a bioinformatic classifier to filter non-tumor derived variants, ctDNA detection rate in early stage CRC (I-III) is 94% (63/67; 95% confidence interval 86%;98%) with 94% specificity, far outperforming the detection rate of somatic sequence variant detection alone. Clinical follow-up is ongoing to evaluate post-surgery ctDNA detection rate and disease recurrence. These results have significant implications for the clinical utility of ctDNA in early stage cancer management.

Citation Format: Seung-Tae Kim, Victoria M. Raymond, Joon Oh Park, Elena Zotenko, Young Suk Park, Matthew Schultz, Won Ki Kang, Oscar Westesson, Hee-Cheol Kim, Yupeng He, Justin I. Odegaard, Stefanie A. Mortimer, William J. Greenleaf, Ariel Jaimovich, Jeeyun Lee, AmirAli Talasaz. Combined genomic and epigenomic assessment of cell-free circulating tumor DNA (ctDNA) improves assay sensitivity in early-stage colorectal cancer (CRC) [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 916.