NEW YORK – This week's annual European Society for Medical Oncology Congress featured new data from efforts to develop blood-based cancer detection tests, which the oncology community hopes may shift diagnoses to earlier, more treatable stages of disease. But discussion of the results highlighted reasons that doctors are still wary of the ultimate value of these tests.
Johns Hopkins University School of Medicine professor Christopher Douville discussed ongoing work on a multi-analyte assay, which investigators initially trained and independently validated in a retrospective cohort representing 18 tumor types. Exact Sciences has licensed technology associated with the assay, which was previously under the commercial umbrella of a JHU spinout called Thrive Earlier Detection, which Exact acquired four years ago.
The group had presented previous results at ESMO's 2023 congress from their case-control ASCEND-2 study. Exact also shared data from ASCEND-2 earlier this year at the annual meeting of the American Society of Clinical Oncology, demonstrating an overall sensitivity of approximately 51 percent with 98.5 percent specificity and nearly 57 percent sensitivity excluding breast and prostate cancers. Sensitivity for stage I tumors was calculated at 15 percent.
This week, the same investigators shared performance data for specific organ origins, or tumor types, for two of the initial four biomarkers — methylation and proteins.
ASCEND-2 recruited adults aged 50 or older with diagnosed cancer, with suspicion of cancer, or who were apparently healthy controls. Over 150 international sites participated, with more than 600 cancers and 2,000 controls included in both the test training and validation cohorts.
Breaking down performance by tumor type, sensitivities ranged from as low as 12 percent sensitivity in known low-shedding tumor types, like prostate cancer, to 60 to 80 percent detection in liver, lung, and colorectal cancers.
Although the promise of multi-cancer screening tests has been a driver in the diagnostics community for several years now, some experts have urged caution that tuning assays toward high specificity and lower sensitivity — a necessity for broad population screening — might not have the best clinical impact.
At ESMO, Douville shared a new sub-analysis from ASCEND 2 that focused on colorectal cancer and advanced precancerous lesions. "When we re-optimized and assessed performance using cross validation of our MCED-derived markers [but targeting] a lower specificity of 90 percent, we were able to detect 19.4 percent of advanced precancerous lesions and 83.4 percent of CRCs," he said.
Adjusting to a novel biomarker panel and repeating their cross-validation, investigators reported that detection rates for pre-cancers and cancers rose to 31 percent and 88 percent, respectively.
"While this is still preliminary … we are excited to see that the MCED configuration could be specifically optimized for single organ use," Douville said.
In essence, he explained, the data provide a proof of principle that MCED assay platforms can be tuned for more specific single-organ screening at lower specificity.
Discussing the new data, Angela George, an oncologist working with the Institute for Cancer Research, London, raised the issue that although Douville and colleagues showed that they can adjust their test to increase sensitivity in certain cancers, the overall performance for detecting early-stage tumors was still low.
Moreover, the analysis didn't break down stage of diagnosis among the various tumor types detected. "I think the biggest problem that we have when we've looked at some of these tests, when we look at [cancers of the] stomach, ovary, and pancreas, these are cancers which often do not tend to be diagnosed at an early stage, so we're very interested in knowing how the stage of the cancer in these particular organ-specific groups stacked up against the detection rate," she said.
In a paired presentation, Celine Vachon, professor of epidemiology at the Mayo Clinic College of Medicine, presented a new analysis from Grail's STRIVE study, which involved nearly 100,000 individuals across 35 sites undergoing standard mammography screening. Grail's Galleri test uses genome-wide methylation sequencing to identify cancer-associated and organ-specific changes in DNA.
"Hopefully, earlier cancer diagnoses can translate to improved prognosis, but the question is, how early can we detect these preclinical cancer signals," Vachon said.
She cited a recent retrospective study of prospectively collected blood samples, sponsored by the American Cancer Society, which showed that in individuals with a cancer signal detected, the time between that detection and clinical presentation of cancer was an average of 323 days.
Vachon and her colleagues collaborating with Grail on the STRIVE trial were interested in examining the same question in their cohort. Trial participants were enrolled between 2017 and 2018, provided baseline blood samples, and were followed for 30 months.
Among this group, there were approximately 2,400 invasive cancers detected within that time frame. The most prevalent tumor type, as would be expected in a mammography cohort, were breast cancers. But almost 1,400 were not.
The team then looked at a measure dubbed tumor methylation fraction (TMeF), an estimate of the amount of tumor-derived cell-free DNA in the blood, in both cases and controls, deriving a 98 percent cutoff point that she said minimized background noise as a proxy for clinical MCED deployment.
Among all cancers in the group, 17 percent had detectable TMeF at baseline, with sensitivity increasing, predictably, as blood samples ranged closer and closer to clinical diagnosis. For the 12 most deadly cancers in the US, 30 percent showed a detectable TMeF level at their baseline blood draw.
That said, Vachon said, there were also preclinical signals out to 30 months before certain cancers were detected clinically.
Although this heads-up was lower than predicted by the cited retrospective American Cancer Society analysis, Vachon said that differences between the cohorts suggest that Grail's MCED screening test could potentially detect cancers up to 30 months before they would have otherwise been diagnosed. She added, anecdotally, that the STRIVE cohort included a stage II pancreatic cancer diagnosed nearly two and a half years before the subject's tumor presented clinically.
Overall, Vachon concluded, the results suggest that a one-year screen for an MCED test with this type of performance could be a "realistic window" to alter care and outcomes in this population.
That said, the analysis did not break down cancer detection by stage across the whole cohort. As such, it remains unknown what proportion of the 30 percent of deadly cancers with detectable TMeF met that threshold in earlier, more treatable stages versus later stages.
ICR London's George noted that the results were encouraging in terms of the number of tumors detected that tend to present at late stages clinically. But she highlighted the question of whether this type of data can speak to MCED tests' ability to meaningfully impact outcomes by shifting diagnoses to earlier stages.
"What we really need for these tests to actually make a difference to our population as a whole is to be able to shift the stage of diagnosis in a clinically meaningful way," she said.