NEW YORK – Members of the CCGA (Circulating Cell-free Genome Atlas) study consortium today published the most comprehensive results to date showing that the epigenomic cancer screening approach from liquid biopsy developer Grail can distinguish a significant number of cancers, including many early-stage tumors, from normal controls using a simple blood draw.
The results, published in Annals of Oncology, come on the heels of Grail's announcement of its first clinical trial, called Pathfinder, that will be returning results to participating physicians and patients, a goal that CCGA investigator and Dana-Farber Cancer Institute oncologist Geoffrey Oxnard said crystalized as new data in the last year showed that the sensitivity and specificity seen in the training portion of CCGA would hold up when applied to an independent set of validation samples.
The data suggest that Grail's test could be one of the first "to have an impact at the broader population level, and could facilitate an important transition from screening for individual cancers, to screening individuals for all cancer types," Joshua Ofman, Grail's chief medical officer, said in a statement.
Spurred by the results, Grail officially launched a prospective study this January, in which participating institutions are ordering the test and returning results to a planned 6,200 individuals, two-thirds of whom have some known risk factor for developing a cancer.
When the trial was first posted, the clinical partners were California's Sutter Health and Utah's Intermountain Healthcare, and they have since been joined by Dana Farber, which announced its participation alongside the new GGCA publication today, and Oregon Health & Science University, which is now also cited in the trial's National Institutes of Health listing.
Grail's technology, now in development for several years, involves targeted methylation profiling designed to distinguish cancer-associated patterns in DNA fragments shed by tumors into the circulatory system.
While some other companies developing genomic and epigenomic blood tests for cancer screening have targeted specific tumor types like colorectal or lung cancer which in some ways present easier targets for clinical translation, Grail has maintained a goal to usher in a new paradigm of pan-cancer screening.
In a statement highlighting the new CCGA publication, the company stated that half of all current cancer deaths remain unaddressed by current guideline-based screening, because the majority of deadly cancers do not have guideline-recommended screening tests available.
Overall, the ongoing CCGA study has included more than 15,000 participants with or without a diagnosis of cancer, used first to develop the genomic/epigenomic strategy for the test, then to train it to distinguish cancer cases from controls, and finally to validate that discrimination.
The sub-study data reported this week — some of which was also shared at meetings held by the European Society for Medical Oncology and American Society of Clinical Oncology last fall — included about 6,500 individuals, 2,000 of which served as an independent validation cohort.
Tested in this independent validation set, which spanned more than 50 cancer types including breast, colorectal, esophageal, gallbladder, bladder, gastric, ovarian, head and neck, lung, lymphoid leukemia, multiple myeloma, and pancreatic cancer, the GRAIL assay showed 99.3 percent specificity: a false positive rate of less than 1 percent.
The overall detection rate for all cancer types in the cohort was about 44 percent across stages I-III. But for a pre-specified set of 12 of the deadliest cancers, which together account for approximately 63 percent of annual US cancer deaths, sensitivity jumped up to 67.3 percent.
Broken down for each stage, sensitivity was 39 percent for patients with stage I cancer, 69 percent for those with stage II, and 83 percent for those with stage III.
Moreover, for 96 percent of positive results the test could detect a signal of where in the body a cancer should be located. In 93 percent of the cases this prediction was correct.
According to Oxnard, apart from the sensitivity and tumor origin performance itself, the fact that he and his colleagues could see so clearly that performance of the test was consistent from the training into this independent validation set is what made the case for pushing now into a clinical, return-of-results pilot.
"Performance often falls off in validation … especially in small cohorts," he said. "But here we have … the validation supporting the performance — the ability to keep false positives low … and at the same time detect a wide range of compelling cancer types and point to a tissue of origin."
"Those pieces together… that's what it looks like we need for a multi-cancer test on a population scale," he added.
The hope now is that the test's performance on all these measures holds up in Grail's ongoing prospective observational studies, as well as in the new Pathfinder trial.
According to Oxnard, part of the point of assessing the test now in full clinical practice is that unlike some of the tumor-specific screening tools being developed, the implementation of pan-cancer screening will require ground-up development of new clinical practice (and regulatory) pathways.
"What's nice about single-cancer tests is we know what to do next. If the test is for lung cancer, you look in the lungs," Oxnard said.
"The problem is that we are not prepared even how to evaluate the [alternative case]," he added. "We are creating the regulatory framework for these tests as we go, and similarly we are creating the diagnostic workflow."
"How is it going to go? That’s the whole point of a trial like this," he said.
According to Dana Farber, recruitment to Pathfinder is temporarily on hold in light of the current coronavirus pandemic, but Oxnard said that while the institution hasn't yet managed to recruit patients, he knows that other centers had already tested some individuals before things were put on pause. Grail did not immediately respond to questions about how many individuals were recruited or their distribution among the high-risk and normal-risk subgroups.
Oxnard also highlighted earlier work that he and his colleagues shared at last year's ASCO annual meeting, which suggested cancers detected by Grail's assay have a higher mortality rate than those the test misses.
In other words, the test's design may ensure that it preferentially detects the deadliest cancers and misses indolent tumors for which immediate treatment would be unnecessary. The possibility that tests like Grail's might lead to overdetection and overtreatment of tumors that pose no actual threat has been a significant point of concern, which makes data like this especially encouraging, Oxnard said.
"We aren’t trying necessarily to find every neoplasm that exists. A test that does that would have an overdiagnosis problem," he explained. "You want to find the cancer that needs to be found, so that is the upside, maybe, of a sensitivity that is compelling … but not perfect."
Annals of Oncology Editor-in-chief, Fabrice André, also director of research at France's Institut Gustave Roussy, added in a statement announcing the publication that "earlier detection of more than 50 percent of cancers could save millions of lives every year worldwide and could dramatically reduce morbidity induced by aggressive treatments."
Announcing the new CCGA publication, Grail also highlighted a second study appearing in Cancer Epidemiology, Biomarkers and Prevention, which used records from the SEER database to model the impact of early cancer detection on cancer deaths.
Based on the analysis by authors of that study, if one-third of stage IV cancers were diagnosed at stage III, one-third diagnosed at stage II, and one-third diagnosed at stage I, clinicians could expect a 24 percent reduction in cancer deaths.