Blood-Based Assay Identifies Hard-to-Detect Gastrointestinal Cancers

A DNA methylation-based cell-free DNA assay demonstrated high specificity and accurate localization of cancers to specific regions of the gastrointestinal tract.

Brian M. Wolpin, MD, MPH

A DNA methylation—based cell-free DNA (cfDNA) assay demonstrated high specificity and accurate localization of cancers to specific regions of the gastrointestinal (GI) tract, according to results of the Circulating Cell-Free Genome Atlas (CCGA) study (NCT02889978), which will be presented during the 2019 Gastrointestinal Cancers Symposium.1

Results showed that the blood-based assay had an 82% (95% CI, 77%-86%) sensitivity rate for cancer detection across stages of disease, at a >99% preset specificity, and a 92% (95% CI, 88%-95%) accuracy rate to predict tissue of origin (TOO) to the GI tract.

“The potential of this test is to diagnose cancer earlier, when it’s more treatable. The ability to do that across cancer types could be quite valuable. Many of the cancer types that this test detects don’t currently have screening tests that allow earlier cancer detection before the cancers cause symptoms,” lead study investigator Brian M. Wolpin, MD, MPH, director of the Gastrointestinal Cancer Center, and director of the Hale Family Center for Pancreatic Cancer Research at Dana-Farber Cancer Institute, stated in a press release.2

Challenges currently exist with early identification of GI cancers, as small tumors are normally not easily detected in routine physical exams. Additionally, screening tests do not exist for cancers of the gallbladder, bile duct, or pancreas and are invasive for colorectal and gastric cancers. Therefore, GI cancers are more commonly diagnosed when they are already at advanced stages and are more difficult to treat.

In the prospective, multicenter, observational, case-control CCGA study with longitudinal follow-up, investigators enrolled approximately 15,000 patients across >20 types of newly diagnosed malignancies, and across all stages, as well as individuals without cancer.

In this second substudy of the CCGA trial, plasma DNA underwent a cross-validated targeted methylation sequencing assay, in order to develop an algorithm that would identify the presence of GI cancer and its location. The data to be presented at the 2019 Gastrointestinal Cancers Symposium will focus on patients with GI cancers as well as those with noncancer controls.

Methylation fragments were combined across targeted genomic regions and were assigned a probability of cancer as well as a predicted TOO. The tumor locations were as follows: esophagus/stomach (upper GI; n = 67), pancreas/gallbladder/extrahepatic bile duct (n = 95), live/intrahepatic bile duct (n = 29), and colon/rectum (n = 121).

Findings showed that the assay had an 82% (95% CI, 77%-86%) overall sensitivity rate, at a >99% preset specificity, for the training set and 81% for the validation set. Overall, the accuracy for defining the GI TOO among the samples, for which TOO was assigned, was 91% for the training set and 89% for the validation set; 6 patients had indeterminate predicted TOO.

When stratified by tumor type, the sensitivity rates were as follows for stage I to III disease: upper GI (75%), pancreas/gallbladder/extrahepatic bile duct (69%), liver/intrahepatic bile duct (79%), colon/rectum (70%). For stage IV tumors, the rates were 100% for upper GI cancer, 96% for pancreas/gallbladder/extrahepatic bile duct, 100% for liver/intrahepatic bile duct, and 93% for colon/rectum.

The predicted TOO accuracy was also identified for these tumor types in stages I to III: upper GI (86%), pancreas/gallbladder/extrahepatic bile duct (94%), liver/intrahepatic bile duct (85%), colon/rectum (96%). When detecting stage IV disease, the predicted TOO was as follows: upper GI (89%), pancreas/gallbladder/extrahepatic bile duct (91%), liver/intrahepatic bile duct (70%), and colon/rectum (100%).

“The data show that evaluating methylation of cell-free DNA within a blood sample, may detect a variety of gastrointestinal cancers with good sensitivity and with a low rate of false positives,” Wolpin stated in the press release. “If further validated with additional testing, this approach has the potential to allow us to diagnose gastrointestinal cancers earlier, when they’re more treatable.”

Two population-based studies are ongoing in an effort to further validate the assay’s screening potential: STRIVE (NCT03085888), which is enrolling females undergoing screening mammograms, and SUMMIT (NCT03934866), which is enrolling individuals without a known cancer diagnosis.


  1. Wolpin BM, Richards DA, Cohn AL, et al. Performance of a blood-based test for the detection of multiple cancer types. J Clin Oncol. 2020;38(suppl 4; abstr 283).
  2. Blood-based test could help identify hard-to-detect gastrointestinal cancers [news release]: Alexandria, VA. American Society of Clinical Oncology. Published January 21, 2020. Accessed January 22, 2020.

<<< View more from 2020 Gastrointestinal Cancers Symposium