NEW YORK – Crucial questions regarding the clinical utility of Grail's multi-cancer screening test Galleri have yet to be answered by prospective data from intended-use populations, pathologists from the Washington University School of Medicine have noted in a commentary on recently published controversial data on the assay's use in prostate cancer early detection.
At the same time, the commenters offered some clarity on the purpose and scope of the study, countering earlier reactions online that focused on a lack of early-stage sensitivity.
Investigators from Grail, the University of Miami Miller School of Medicine, and the Vancouver Prostate Center at the University of British Columbia faced criticism following their late August publication in JCO Precision Oncology of a prostate cancer sub-analysis from the company's prior case-control CCGA study and ongoing prospective PATHFINDER trial.
In a subsequent editorial led by first author Ann Ayzman and also published in JCO Precision Oncology, the WashU oncologists voiced concerns that most data related to multi-cancer early detection (MCED) testing has largely been taken from case-control studies, including the recent reported data.
Because cancer cases in this type of study are recruited after they've already been diagnosed, there is concern that the reported sensitivity and specificity is an overestimation compared to what would be true in a real-world undiagnosed population.
"Some extrapolated data suggest that in almost all screened cancer types, the positive predictive value [of MCED testing] would be [less than] 10 percent," Ayzman and colleagues wrote, citing, among other caveats, a critique of earlier Grail data by pathologists at Toronto's Mount Sinai Hospital.
Commenters discussing the study on LinkedIn last month expressed shock at how low Grail's test sensitivity was even in a case-control context, especially for early-stage prostate cancer. They highlighted that Galleri is marketed as an early detection test and that its ultimate clinical value rests in shifting cancer diagnoses to earlier, more treatable stages.
The Grail study authors responded that the question of clinical utility is more complicated than that, conceding that while an MCED assay should meaningfully improve the ability to detect cancers at treatable stages, it shouldn't be expected to do this equally across all tumor types. In addition, it is important that the test avoid detecting certain early cancers, including prostate cancers, that don't need to be treated, they said. Grail didn't add comments of its own but highlighted the same points in its initial announcement of the publication.
Ayzman and her coauthors noted along the same lines that the Grail study did provide some clarity on Galleri's strengths. For one, the data cement earlier suggestions from the company that Galleri preferentially detects deadlier cancers and is thus unlikely to lead to overdiagnosis.
"The data suggest that a positive result would be unlikely to detect indolent disease and should prompt ordering clinicians to obtain further diagnostic workup," the editorial authors wrote. "Notably, roughly one-third of detected cases in the CCGA cohort had nonmetastatic disease that was potentially amenable to cure."
In an email, Brandon Mahal, lead investigator on the Grail prostate cancer study and an oncologist at University of Miami's Sylvester Comprehensive Cancer Center, stressed that the goals of his team's analysis were not to provide evidence for clinical sensitivity in an intended-use population.
"A common concern we hear around MCED testing is that it may contribute to overdiagnosis of cancers. That is, we may be detecting cancers that would not grow or spread enough to cause harm over the course of an individual's lifetime," he said.
"Prostate cancer is one cancer type that has a notably high proportion of indolent cancers and is very common in the intended-use population for MCED tests … [so] we were particularly interested in understanding how Galleri test performed in detecting various grades and stages of this disease."
Mahal said that "threading the needle" therapeutically is a particular challenge for prostate cancer. "We want to detect cancers before they metastasize, while at the same time, we do not want to over-detect early-stage cancers that may not ever metastasize."
In the study, Mahal and his coauthors analyzed Galleri's performance across different prostate cancer Gleason grade groups and clinical stages in 420 patients from the CCGA, a three-part case-control study that established the Galleri assay format, also noting performance thus-far in the ongoing PATHFINDER prospective trail.
In the CCGA prostate cancer patients, the test detected only 3 percent of stage I disease, about 5 percent of stage II cancers, 15 percent of stage III, and more than 80 percent of stage IV cases. In PATHFINDER, one cancer has thus far been detected and one missed.
Overall, Mahal said, the results show that individuals with a positive Galleri test and a tumor origin prediction of prostate cancer are likely to have a clinically significant cancer that merits rapid workup and treatment.
He also noted that low sensitivity to early-stage prostate cancer does not mean Galleri would necessarily show low sensitivity for all early-stage cancers. In other words, conclusions about the test's overall clinical value shouldn't rest on this specific sub-analysis.
Because most of the data they analyzed was retrospective, Mahal and his team didn't have the ability to look at clinical outcomes for cancers detected by Galleri versus those the test missed. But as a surrogate, they compared the overall survival of patients in the CCGA cohort with historical records from the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) database.
Patients that Galleri missed had about three times higher overall survival rates than the SEER population, while those it did detect had similar survival rates to the overall historical average.
"Survival comparisons with SEER data may not be definitive, but they can provide critical insights," Mahal wrote. "We found that participants with prostate cancers not detected by the MCED test were three times more likely to survive than would be expected. This result strongly suggests that those cancers not detected by the MCED test are likely indolent cases that would not cause harm."
Moreover, about a third of the Galleri-detected cases in the CCGA group were both high-grade and pre-metastatic, "the types of cases we want to detect," Mahal said, because they are both high risk and potentially curable.
In their commentary, Ayzman and colleagues stressed that regardless of its implications, the current analysis should be held with caution because the number of cases detected in the study was relatively low.
Another important limitation, they wrote, is that 85 percent of patients in the cohort self-reported as White, non-Hispanic. "Prostate cancer is notorious for disproportionately affecting Black men," who face "higher incidence rates, more aggressive disease, and double the mortality risk compared with their White counterparts."
Future efforts would benefit from greater inclusion of the true population that stands to benefit from MCED screening, they argued.