SAN FRANCISCO — A single blood-based test proved accurate in detecting multiple gastrointestinal (GI) cancers, and in most cases could pinpoint a tumor’s location, a prospective case-control study showed.
With a specificity of greater than 99%, sensitivity of a targeted methylation-based cell-free DNA assay reached 82% on a training set of 312 patients with established GI cancers and 81% on a validation set of 135 GI cancer patients, Brian Wolpin, MD, MPH, of the Dana-Farber Cancer Institute in Boston, reported here.
“A multi-cancer early detection test evaluating cell-free DNA methylation may be a useful test to detect GI cancers, and may guide further evaluation and work-up,” Wolpin said during his presentation at the Gastrointestinal Cancers Symposium.
“If you don’t have a high enough specificity, you get a lot of false positives and that leads to a lot of follow-up testing and a lot of costs to the healthcare system,” Wolpin told MedPage Today.
Among more than 2,000 cancer-free controls, the classifier achieved a specificity of 99.8% in the training set (three false positives out of 1,521 individuals) and 99.3% in the validation set (four of 610 individuals).
“The ability of this sort of approach to have a specificity greater than 99% is sort of a prerequisite in many ways to get started,” he said. “That’s been quite an achievement to get to.”
Accuracy of the assay improved by disease stage. For patients with stage I cancers, sensitivity was just under 50% but stage IV cases approached 100%. For stage I-III disease combined, sensitivity was 73% in the training set and 71% in the validation set.
The assay was also highly predictive of a tumor’s location, and correctly predicted the tumor organ in 91% of patients in the training set and 89% of those in the validation set.
Session moderator George Fisher, MD, PhD, of Stanford University School of Medicine in California, asked whether the test was “sufficiently specific” to detect the location of cancers in patients who present with liver metastases or intraperitoneal disease but with an unknown primary tumor.
“With the targeted methylation approach, there is a high potential to be able to define the organ of origin,” said Wolpin, adding that at the moment the efforts have been focused on the “early-detection space” and on a population scale.
“But I think, as you’re suggesting, there are other ways that this type of approach could be used in the future that could be beneficial to patients,” he said.
Wolpin was presenting results from the gastrointestinal cancer cohort of the Circulating Cell-free Genome Atlas (CCGA) study, a multicenter observational trial that has enrolled over 15,000 individuals across 142 sites, including 2,185 newly diagnosed patients with over 20 different cancer types.
For the GI portion of the CCGA study, the test’s accuracy was tested in 447 patients with GI cancers — 174 with colorectal tumors, 25 with stomach tumors, 71 with esophageal tumors, 123 with pancreas tumors, 14 with gallbladder tumors, and 40 with tumors of the liver or biliary duct.
“Blood tests that can identify cancer in asymptomatic individuals, particularly GI cancers that can be difficult to detect in early stages, could change cancer diagnostics by making it easier to accurately screen for and identify these cancers earlier,” Muhammad Shaalan Beg, MD, of UT Southwestern Medical Center in Dallas, said in an American Society of Clinical Oncology statement. “The preliminary results seen in this study, however, will need to be validated by screening large populations of asymptomatic individuals.”
Indeed, a third phase of the CCGA study will seek to further validate the results. The current study made no estimate for the test’s positive or negative predictive value in a population-based screening setting.
Wolpin highlighted that the vast majority of patients with pancreas, biliary, or gallbladder cancers present with metastatic disease.
“If you could find 50% to 70% of those patients when they had stage I or II disease, that’s a big improvement over what we do now,” said Wolpin. “There is clearly still more work that needs to be done.”
Last Updated January 24, 2020
The study was funded by GRAIL.
Wolpin reported financial relationships with Celgene, G1 Therapeutics, BioLineRx, Genentech, GRAIL, and Eli Lilly. Numerous co-authors are employed by GRAIL.
Beg disclosed honoraria or consulting/advisory board relationships with Array BioPharma, Boston Biomedical, Bristol-Myers Squibb, and Ipsen, as well as research funding from various companies.