CHICAGO (GenomeWeb) – Grail this week shared new data from its CCGA (Circulating Cell-free Genome Atlas) study, an exploratory effort intended to define blood-based molecular signatures that distinguish early cancers from normal control subjects.
The company presented the new data at the American Society for Clinical Oncology annual meeting here, continuing to pave the way for the clinical cancer screening tests it intends to develop.
The broad genomic, transcriptomic, and methylation approaches the company is using for this discovery study don't offer much insight into what the company's eventual diagnostic test will look like. Accordingly, oncologists reflecting on the data at the meeting expressed a mixture of excitement and realism, with ASCO leaders lauding the early results as bringing the field "one step closer" toward a blood test for detecting early cancers.
Especially notable was a CCGA subset analysis focused on lung cancer. Researchers performed a case-control study of 561 non-cancer subjects and 118 patients with lung cancer using three methods the CCGA has adopted — targeted deep sequencing, shallower genome-wide sequencing, and bisulfite sequencing to analyze methylation patterns.
According to the study's authors, detection rates (sensitivity at a 98 percent specificity cutoff) were up to 51 percent for participants with early-stage lung cancers (stages I-IIIA), and 87 percent to 89 percent in those with tumors at stage IIIB and above, depending on which of the three genomic approaches was used.
The lung cancer breakout at ASCO follows Grail's first pan-cancer report from the CCGA, presented earlier this year at the American Association for Cancer Research annual meeting. There, investigators said that they had detected a cancer-like signal in only five of more than 500 cancer-free enrollees, two of whom shortly went on to be diagnosed with a malignancy.
This ability to produce high specificity also held up in the new lung cancer-specific analysis, according to Dana Farber's Geoffrey Oxnard, who discussed the results at a press briefing during the ASCO meeting.
He said that the lung cancer-specific analyses have also included some individuals who were initially considered false positives who have since gone on to develop a clinically confirmed cancer.
Grail has stated that it intends to develop tests that offer early detection of cancers regardless of where they originate in the body, but recent announcements have suggested that the company may be targeting different approaches in different anatomical tumor types.
For example, the firm is commercializing a cell-free DNA sequencing approach developed by Chinese University of Hong Kong researcher Dennis Lo that targets the Epstein-Barr virus to detect nasopharyngeal carcinoma.
For the thoracic oncology community, an important benchmark for a genomic test would be achieving better performance, with fewer false positives, than the imaging techniques doctors use now, namely low-dose CT scans.
Another study released at the meeting reported dismal uptake of CT lung cancer screening, despite randomized trial data that shows it improves outcomes. The false-positive rate and the impact that has on overtreatment appears to be too great for the community to embrace widely. In addition, global use of CT screening is hindered by problems of healthcare and technological access.
Oxnard has long offered a voice of caution in regards to cell-free DNA analyses in cancer, and the rapidly expanding "liquid biopsy" market. But he said that he is sincerely encouraged by what the early CCGA data now suggests. "This was entirely a wacky brainstorm when GRAIL was launched, and now its like 'Wow, we are going to get there,'" he said.
"Two years ago this was a pipe dream… and I didn't believe that this could be done. Today we actually have data that shows it is possible to find early-stage cancer in the blood," he added."It will take some time, but its well on its way."
In other posters presented at ASCO, Grail touched on a variety of other areas it has been investigating in the CCGA, including a report on findings that pertain to clonal hematopoiesis.
As methods have been advanced that seek to detect minute amounts of cancer DNA in the huge background of normal DNA that a blood sample contains, increasing attention has been paid to the phenomenon of clonal hematopoiesis of indeterminate potential (CHIP), whereby mutations accumulate in cancer-associated genes in small clonal populations of white blood cells.
The danger posed by CHIP — both to tissue sequencing and to new blood-based efforts — is that these clonal mutations can potentially be mistaken for a somatic mutation, potentially misdirecting treatment for a patient in the case of tumor tissue sequencing or causing false-positives in the case of blood-based early detection.
First author Charles Swanton from the Francis Crick Institute said that the data coming out the CCGA continues to illustrate the need for future cell-free DNA cancer screening tests account for this phenomenon to avoid false positives.
Among participants with lung cancer in that breakout study, for example, more than 54 percent of somatic mutations detected were derived from white blood cells and not from tumors.
Swanton reported that across more than 1,400 CCGA samples — 576 controls and 836 cancer patients — seven percent of individuals showed a signal of clonal hematopoiesis with a variant allele frequency over 10 percent, while about 40 percent had a signal below 1 percent VAF.
Importantly, the rate of clonal hematopoiesis was similar in the cancer and non-cancer group. And of the CHIP variants identified, over 90 percent were unique to individual patients as opposed to shared.
According to Swanton, this means that there doesn't look to be a bioinformatics or other shortcut that can be taken to resolve the confounding influence of clonal hematopoiesis in blood-based cancer detection.
"Currently we don’t thing there is an alternative to sequencing white blood cells in parallel … because we find very few mutations in common between patients, so there is no a priori way of coping with this," he said.
In a third presentation, the company provided updated data from its larger pan-cancer analyses in the CCGA. Breaking out data from just the bisulfite sequencing aspect of the study's protocol, investigators reported in a poster that detection rates for tumors up to stage III ranged from 56 percent to 80 percent among cancer types that generally cause high mortality, including colorectal, esophageal, head and neck, liver, ovarian, pancreatic, and triple-negative breast cancers, as well as lymphoma and multiple myeloma.
A fourth abstract broke out data specific to breast cancers. According to authors, stage I-III breast cancers had a fairly low detection rate overall — 21 percent if specificity was maintained at the specified 98 percent level.
But interestingly, triple-negative breast cancer had a much stronger signal, with 56 percent detection. The researchers also reported that individuals whose breast cancer was diagnosed through clinical presentation — as a result of symptoms as opposed to thorough screening — had a stronger signal than those diagnosed via mammography: 38 percent and 9 percent respectively.
Oxnard reiterated that its likely that even with optimized methods, developed through machine learning approaches and exhaustive training of various combinatorial assays in the CCGA, there will remain some cancers that just don't yield a blood-based signal.
Hopefully, the continued study will offer opportunities to improve, for example, the 50 percent detection rate that the group saw in earlier stage lung cancers. "We'll try to do it better, based on what we learn here. Maybe we can inch that up," he said.
Also, he argued, the fact that some early cancers aren't presenting a nucleic acid or methylation signal in the blood doesn’t have to be viewed only as a failure. It may also be an opportunity.
"Maybe these cancers are more well-behaved, and maybe we are better at detecting the high-risk cancers," he posited. "We have to work now to understand which we are finding and which we are not, and whether that is biology-based."
As hinted at the breast cancer breakout data, if blood-based testing offers increased sensitivity to more aggressive tumors and lower sensitivity to indolent disease, that could be a good thing, considering that the primary weakness of current screening tools like CT scans and mammograms is the inability to distinguish low-risk findings — resulting in high false-positive rates and associated overtreatment.