CHICAGO (GenomeWeb) – Based on the latest sensitivity and specificity data from its case-control analyses in the CCGA (Circulating Cell-free Genome Atlas) study, cancer early detection firm Grail believes that it will be ready to start returning results to clinical trial participants next year, a crucial step in gathering evidence for the clinical validity and utility of its approach.
The new results, shared in a poster on Saturday at the American Society of Clinical Oncology annual meeting here, joined several other presentations by the company this weekend, including additional data comparing outcomes among those whose cancers the Grail test caught and those it missed.
Grail Chief Scientific Officer Alex Aravanis said that the updated sensitivity data was collected for all 20 cancers studied in the CCGA, and for a pre-specified group of 12 of the deadliest cancers, which together account for about 63 percent of cancer mortality.
Significantly improving on results shared in prior conference abstracts, investigators reported on Saturday that an updated version of the company's methylation-based approach was able to detect between 59 and 86 percent of cases in the 12-cancer subset with variation across stages and cancer types.
"Importantly, this was at a very low false positive rate, less than or equal to 1 percent. And the test also could correctly identify the tissue of origin, or site of the cancer with high accuracy," he added.
Specifically, overall detection was 34 percent, 77 percent, and 84 percent at stages I, II, and III respectively. In addition, a tissue of origin result was calculable in 94 percent of cancers detected, correctly identifying the tumor location in 90 percent.
"You really want to find cancers when they're non-metastatic and even better when they're still localized, because outcomes can be dramatically better," Aravanis said. "So [seeing] detection of 77 percent [for Stage II cancer] is quite good and speaks to the potential to have a significant clinical benefit and potential mortality reduction."
"Stage II outcomes are only marginally different than stage I, so that's very encouraging," he added.
Based on these results Grail believes it will be ready to begin returning results to patients so that they can be acted upon next year.
This will take place with a new cohort, as opposed to one of the firm's existing observational studies — which include the fully-enrolled 100,000-person STRIVE study of women over 50 and the more recently announced SUMMIT study which aims to recruit another 50,000 individuals — Aravanis added. He said Grail is in active discussions regarding where, and with what partners, this return of results research will take place, but declined to disclose potential collaborators.
An open question for the Grail data (and for similar reports from competitors) is whether the sensitivity numbers seen in these initial case-control analyses will hold true in prospective analyses of the intended population, in this case individuals without a known cancer.
"It's hard to speculate how the performance will change, but … this is why we think it's important to do that validation," Aravanis said, adding that this is exactly what the company is expecting to read out from its STRIVE and SUMMIT studies.
According to Charlotte Arnold, Grail's vice president of corporate affairs, the company expects to report data from STRIVE next year.
Aravanis added that some of the company's other completed analyses are also contributing to its confidence that it can manifest appropriate sensitivity and specificity as it moves the test forward through validation.
He highlighted a second poster presented at the ASCO meeting, in which a group of investigators led by Dana Farber oncologist Geoffrey Oxnard looked at mortality data from cases in the CCGA.
That team found that if they compared cancers that were detected by Grail to those that were not, there was a much higher mortality for the former: "about an eight times hazard ratio … of dying during the follow up period," Aravanis said.
"This is really important because what it says is that we're preferentially detecting the [deadliest] cancers, which is really what you want to find early and be able to intervene on," he added. "Conversely, what it says is that initial theoretical concern that perhaps what we would detect would be indolent cancers or cancers where there's a lot of overdiagnosis and overtreatment, we're not seeing that."
The data don't provide definitive evidence of a mechanism for why the test may preferentially detect more aggressive cancers, but Aravanis said that a good hypothesis might be that there is a parallel in what biologically allows tumors to shed DNA into the blood and what makes them clinically dangerous.
"If you think about it, cancers that grow fast, that have a lot of cell turnover, that invade, get blood access: that's probably what's biologically required to get material into the blood functionally … so it's probably not a coincidence that, for example, we don't detect as many indolent tumors like early-stage prostate cancer or early-stage hormone receptor-positive breast cancer or thyroid cancer."
Grail was already flanked by a number of competitors in the cancer screening space, including Guardant Health, and Freenome, and ranks are continuing to grow, most recently with the launch last week of Thrive Earlier Detection, which is advancing a method develop at Johns Hopkins.
One aspect that Thrive executives said they see as weighing in their favor is the relative simplicity and cost-effectiveness of their targeted mutation and protein approach, compared to genome-wide methods.
Aravanis didn't provide details about the precise costs of the system that Grail employs for analyzing methylation signals, but he said that the assay used to generate the most recent CCGA data presented at ASCO has a "much lower cost format," than what was employed in previous analyses.
Notably, Grail developed its methylation approach after first comparing several blood-borne DNA signals, including mutations and copy number variation — comparing their performance in the same cohort.
Not only was methylation the strongest differentiator between cancer and non-cancer samples, but the company also analyzed potential complementarity or added value for combining different signals, Aravanis said.
In contrast to what Guardant Health reported earlier this year at the annual meeting of the American Association for Cancer Research for its assay — that it could detect some cancer cases only via mutation, and others only via methylation — for Grail "anything detected by the other approaches was also detected by the methylation approach."
Regarding the competitive landscape in general, Aravanis said that it is exciting to see so many making investments to bring this type of testing to the clinic.
"What we believe will really enable broad adoption is a very low false positive rate … and then also the ability to distinguish the cancer type to aid in the workup," he added.