NEW YORK – Several early adopters of Grail's blood-based multi-cancer screening test, Galleri, shared their experiences implementing the assay clinically and studying it observationally at the annual meeting of the American Society of Clinical Oncology over the weekend.
The new data comes at an unexpectedly conspicuous moment for the company after it was revealed last Friday across major media outlets that one of Grail's vendors, PWN Health, had mistakenly sent 400 positive test results to patients who were not supposed to receive them.
Although this was a reporting error, rather than a failure of the test itself, it comes at a tenuous moment for Grail, which is currently the subject of a financial regulatory battle between the US Federal Trade Commission and Illumina, which acquired its former spinout in August 2021.
Aside from that complication, the company and its hopeful competitors also face significant skepticism from stakeholders who are concerned that Galleri and other MCED tests won't offer clinical impacts that merit their costs to the healthcare system.
Real-world and observational outcomes don't answer these questions the way randomized data can, but the showing by early adopters at the ASCO meeting did convey the excitement and enthusiasm these pioneers have for a new, potentially revolutionary screening tool. Grail's test involves sequencing cell-free DNA extracted from a blood sample and analyzing it for cancer-associated methylation patterns.
In an oral presentation, University of Oxford oncologist Brian Nicholson reported initial results from a study called SYMPLIFY, a prospective assessment of the clinical performance of Grail's test in individuals who haven't yet been diagnosed with cancer but who have signs and symptoms that may indicate a tumor.
The study is the first large-scale prospective evaluation of an MCED test in a symptomatic population. Overall, Nicholson and his colleagues studied 5,461 participants recruited over five months from five clinical pathways at 44 hospital sites in the UK. These included patients referred for urgent imaging, endoscopy, and other diagnostic modalities to investigate suspicious symptoms. The team tested patient blood samples and then compared Grail's predictions to the standard-of-care diagnosis.
Grail's assay detected a cancer signal in 323 cases, 244 of whom also had a cancer diagnosis via standard-of-care examinations and 79 of whom did not. Investigators calculated a positive predictive value for Galleri of about 76 percent and a negative predictive value of nearly 98 percent. They also reported an overall sensitivity of 66 percent and a specificity of 98 percent. According to Nicholson and his colleagues, sensitivity increased with increasing age and cancer stage, from 24 percent sensitivity for stage I disease to 95 percent in stage IV. In addition, the test's prediction of cancer location or origin was accurate 85 percent of the time.
Discussing the Oxford study, Hartford HealthCare oncologist Peter Paul Yu said that the performance drop in early-stage disease calls into question whether Galleri could demonstrate adequate clinical utility in this symptomatic setting.
Nicholson argued that there's no question that a positive test should lead to further investigation, which can be meaningfully informed by Galleri's tumor location prediction. In contrast, a negative test didn't seem to push down the risk of cancer far enough to rule out further investigation, unless limited to the upper GI pathway, where negative predictive value was much higher.
Study investigators believe their results can be used to further develop the Galleri classifier to enhance its negative predictive value in symptomatic patients. If that can be achieved, Nicholson said he sees a potential use case in primary care to select patients for referral and direct that referral to the correct pathway — something he argued should be pursued in prospective interventional trials.
In a second session, Eliezer Van Allen, chief of the division of population sciences at Dana-Farber Cancer Institute, discussed a set of posters presented at the meeting, including a follow-up to Grail's PATHFINDER study and an analysis of early real-world clinical experience using Galleri's tumor origin prediction to guide targeted diagnostic next steps.
Van Allen said the "provocative" title given to him for his discussion — "Multi-Cancer Early Detection: Ready for Clinical Implementation?" — highlights the excitement and the uncertainty of the current moment regarding multi-cancer screening tests.
"I hope I don't need to convince a room in the prevention section of ASCO that early detection in cancer prevention is something that we should all be aspiring to achieve across the board," he said.
The PATHFINDER follow-up study compared cancer detection results using two versions of the Grail assay and assessed follow-up diagnostic evaluations — whether they matched the test's origin prediction and whether they were successful.
Among 39 patients with a cancer signal detected by both test versions, resulting diagnostic evaluations were consistent with cancer signal origin in 30 patients, or 77 percent of cases, the authors from Memorial Sloan Kettering Cancer Center reported. Diagnostic resolution was achieved in 32 of the 39 patients, with the other seven requiring additional work-up. This additional work-up led to diagnostic resolution in all seven cases, with three patients diagnosed with cancer and the other four found to be clear of cancer as of the end of the study period.
Van Allen said a particularly interesting finding from the study was on the utility of whole-body imaging. In about half the cases, this work-up didn't actually contribute to diagnostic resolution. "This was for a variety of reasons but is interesting and … something worth keeping an eye on in this particular space," he said.
He also highlighted that the mean number of imaging tests and diagnostic procedures for participants with false positive and true positive tests was similar, something that gives a "glimpse of [future health, economic, and other impacts] if this rolls out in the larger population setting."
Van Allen also highlighted a poster authored by clinicians at Adventist Health, which summarized experiences from an initial set of about 53,000 Galleri tests offered clinically.
Notably, "the real-world cancer signal detection rate was actually 0.95 percent," Van Allen said, which is "pretty close" to what investigators saw in the original PATHFINDER study, with distributions also largely concordant, though patients 54 years old and under were excluded from the PATHFINDER trial but did make up a small group in the Adventist study.
Among those with a positive test in the Adventist cohort, 32 percent were detected at stages I and II, and 61 percent at stages I, II, or III.
"In terms of cancer signal origin — the ability to actually map back from testing to the anatomical site — they achieved a 91 percent rate of success … which is also consistent with some of the other studies that are happening in the non-real-world setting," Van Allen added.
In addition, more than 60 percent of the positive results represented cancers that aren't currently screened for on a population level. Although follow-up is ongoing, early data from a limited subset of patients indicates successful diagnosis of invasive cancer across multiple tumor types, including stage I and II cancers, the authors wrote.