“Our multicancer early-detection blood test detected the strongest signals for the deadliest cancer types, regardless of clinical stage, while signal for indolent cancer types was low,” wrote Kelsey Grossman, the associate director of corporate communications for GRAIL, in an email to Cancer Therapy Advisor. “These findings suggest our blood test may detect the cancers more in need of treatment, rather than contributing to the overdiagnosis of indolent cancers.” GRAIL’s test detected 34% of stage I, 77% of stage II, and 84% of stage III cancers.

Dr Phallen similarly said that future plans for DELFI include training the algorithm to spot the difference between aggressive and nonaggressive cancers. “It’s an extremely important question,” she said. “In the future, we may do some studies, with prostate cancer for example, where we look at the difference in PSA detection versus detection with the DELFI method.” Additional clinical follow-up on patients will also help shed light on whether overdiagnosis might be a problem with the test.

Related Articles

None of the patients reported on in the Nature paper had undergone treatment at the time of testing, but Dr Velculescu said the team is examining how the fragmentation patterns change over the course of treatment. When the treatment was working, he said, the abnormality of the fragmentation patterns declined, whereas when a treatment failed, the abnormal patterns increased. Eventually, it may be possible to train the system to understand the changes in the fragmentation pattern over the lifetime of the tumor.

Continue Reading

To build the best early-detection blood test, a combination of the various approaches could be the way to go. Medical oncologist Filip Janku, MD, PhD, at the University of Texas MD Anderson Cancer Center said, “When ultimately the test for clinical use is developed, I don’t think it will be a test using one method or the other, but I think it will be a combination of several methods.” He pointed out that in the DELFI study, the researchers combined the fragmentation approach with searching for mutations in a subset of patients — and in this way, they improved the test’s sensitivity to 91%.

Dr Diamandis also mentioned that the sensitivity will need to be improved before the test is useful for rare cancers, such as ovarian or pancreatic cancer. “You need to screen 3000 women to find one [case of] ovarian cancer, so the specificity must be higher than what [DELFI developers have] achieved,” he said.

While the fragmentation test shows promise, it’s a proof-of-concept experiment, and more validation needs to be done before it would be ready for the clinic. “It would need to be validated to make sure these algorithms actually work on the large scale,” said Dr Janku.

There also remains the problem of small tumors not shedding enough DNA to do a proper analysis. However, because looking for fragmentation patterns encompasses changes all over the genome, it may be possible to detect cancer patterns even with tiny amounts of circulating DNA. “It’s possible, they say, that even with smaller representation of the genome within the circulation, there may be enough information there to make the call,” said Dr Diamandis. “The smaller the amount of DNA in the circulation, the less confident the call will be.”

While challenges still remain, and testing using fragmentation patterns will ultimately require more validation on larger groups of patients, DELFI’s novel approach to cancer detection with cfDNA could pave a new path. “When I saw just the title, I said, ‘oh, here is another one,’” Dr Diamandis admitted. “But this one has something new and potentially workable … I’m a lot more optimistic on this approach.”

Correction: This article was updated on July 15, 2019, to correct the definition of the acronym for CCGA and to clarify the percentage for which the researchers correctly identified the malignancy’s tissue of origin.


  1. Cristiano S, Leal A, Phallen J, et al. Genome-wide cell-free DNA fragmentation in patients with cancer [published online May 29, 2019]. Nature. 2019;570(7761):385-389. doi: 10.1038/s41586-019-1272-6
  2. GRAIL announces positive new data with multi-cancer early detection blood test from CCGA study [news release]. Menlo Park, CA: GRAIL, Inc. Published May 31, 2019. Accessed June 24, 2019.