NEW YORK – Researchers from cancer detection firm Grail have published new data on a method for determining the amount of cancer in a patient's body from a blood sample without using traditional imaging methods.
The technique, called tumor methylated fraction (TMeF), holds promise for various monitoring use cases, including detection of minute residual disease in patients with early-stage, potentially curable cancers, and monitoring cancer progression or response to treatment in later-stage disease.
Estimates of tumor burden are used across the cancer care spectrum as a prognostic indicator to track treatment response and identify disease recurrence, but current imaging-based methods have notable limitations. Tumors have to be large enough and in the right locations to be imaged accurately, and inferring 3D volume from 2D images doesn't always correlate with actual tumor volume. Image interpretation is also subjective.
Grail first announced its ambitions in blood-based cancer monitoring in 2020. In an email this week, Grail CSO and Senior VP of Research Amoolya Singh said that the company began working on developing a tool to quantify tumor fraction in blood samples while it was still validating its flagship multi-cancer early detection test, Galleri, which is now available clinically.
The firm's TMeF quantification is based on the same concept of assessing methylation patterns in circulating cell-free DNA that Galleri uses to determine whether an otherwise healthy individual has an unknown cancer somewhere in their body.
Other methods are available to estimate the amount of tumor-associated DNA in the blood, mainly by quantifying small variants, and in some cases using tumor tissue sequencing data to help improve sensitivity. Although there are workarounds that allow for tissue-free estimations, and for dealing with confounding factors like clonal hematopoiesis, Grail has maintained that its methylation-based approach could offer a simpler, more accurate alternative.
In a study published late last month in the journal Cancers, Grail investigators described their development of TMeF using an algorithm to identify differentially methylated regions in a reference database of cancer tissue samples compared to cell-free DNA samples from individuals without cancer.
The method then measures the quantity of DNA fragments that match the identified cancer-associated DMR patterns to calculate TMeF.
When the researchers compared TMeF to small variant allele fraction estimates for the same cancer plasma samples, the two were significantly correlated. The group also confirmed that their TMeF estimates were accurate compared to the expected tumor burden in synthetic titrations.
In further tests, the team observed that TMeF increased with cancer stage and tumor size and inversely correlated with a patients' probability of survival. Overall, the authors wrote that TMeF demonstrated an "order of magnitude improvement in the lower limit of accurate quantification over previously published approaches."
They also noted that accuracy is not the only advantage of a methylation-based approach to tumor burden estimation. Based on their data so far, TMeF ctDNA estimates appear to be robust across the clonal evolution of tumors because of the large number of differentially methylated regions that are common among cancer types.
"Due to the broad nature of the signal used to compute TMeF across numerous cancer types, TMeF likely measures ctDNA levels originating from all shedding cancer cells — both from the primary tumor and metastatic sites — regardless of clonal lineage," the Grail team wrote.
The authors said TMeF could be used clinically as imaging is currently, to measure tumor burden at baseline in order to predict patient prognosis. Baseline tumor burden is also being explored as a predictive marker for specific cancer therapies.
Singh said that Grail believes TMeF also has great potential in therapy response monitoring, with emerging data suggesting that early on-treatment response patterns can even predict overall response and outcomes.
"It has [also] been shown that ctDNA abundance can be utilized to predict therapeutic response in various disease areas for different therapeutic agents. This is an active area we are exploring, as well," she said.
Another low hanging fruit is the growing minimal residual disease testing sphere. Guardant Health's commercial MRD test, Guardant Reveal, includes epigenetic signals, along with other molecular targets. Singh said a direct comparison isn't possible, but highlighted Grail's cancer-agnosticism and genome-wide methylation approach as potential distinguishers.
Blood-based tumor fraction methods have also caught the eye of other liquid biopsy firms. In a 2022 study, Foundation Medicine researchers used an aneuploidy-based method added to variant detection to estimate tumor fraction for patients tested with its liquid biopsy genotyping assay FoundationOne Liquid.
With the ability to estimate tumor fraction, Foundation Medicine showed that it could help better discriminate true-negative results from potential false-negatives caused by ctDNA levels present at fractions too low for its test to detect. Identifying these patients could help better determine which patients with negative results would be best served by a follow-up tissue test.
Singh said that Grail's research has shown that methylation-based tumor fraction is also highly correlated with successful cancer detection. The high negative predictive value reported by the Foundation Medicine researchers comports with Grail's analyses.
One thing that is not on the agenda, at least right now, Singh said, is adding tumor burden estimation to Grail's existing Galleri assay, which "aims to provide simple and precise directions to primary care doctors for where cancer may be developing."
Grail is now further validating TMeF in larger, more diverse populations of clinical subjects, which will allow it to measure the effect that things like genetic ancestry, cancer subtypes, and clonal evolution have on its accuracy. However, based on preliminary analyses, the authors wrote, it doesn't look like ancestry-related methylation pattern differences would have a significant effect on test performance.
Singh said that the company also has a number of pharma partnerships in which it is studying TMeF, including MRD studies in early-stage lung cancer assessing both prognostication and recurrence prediction with AstraZeneca.