NEW YORK (GenomeWeb) — As a number of companies forge ahead with blood-based cell-free tumor DNA or circulating tumor cell "liquid biopsy" tests for cancer diagnostics, Foundation Medicine is taking a more cautious approach.
"We fundamentally believe that cell-free DNA tests, in the vast majority of clinical settings, should only be used when it's absolutely impossible to get hold of an FFPE biopsy, or that biopsy has been exhausted with the pathology work-up," Foundation Medicine CSO Phil Stephens told GenomeWeb recently.
"The biopsy has been the gold standard for diagnosis for decades," he said, and most clinical trials of targeted therapies for FDA approval have involved biopsies. "Until one does a systematic correlation between what's going on in the cell-free DNA and what's going on in the biopsy, we don't believe it's appropriate to make informed treatment decisions" based on cell-free DNA tests, Stephens added.
One area the company believes ctDNA assays may have an edge over biopsy tests is in patients who relapse after targeted therapy. "In that unique clinical setting, there may be some advantages with cell-free DNA," he said, adding that further studies are needed.
Before Foundation Medicine will launch its own ctDNA test, "there are some key questions that need to be addressed," he said, for example which tumor types shed sufficient DNA into the blood, and how well genomic alterations in the tumor and in cell-free DNA correlate.
Companies such as Guardant Health, Personal Genome Diagnostics, and Inivata have been developing and or already offer next-gen sequencing-based cell-free DNA cancer profiling assays and presented their latest studies at the American Association for Cancer Research meeting in Philadelphia last month.
Earlier this year, Foundation Medicine said it plans to launch a cell-free ctDNA assay for its pharmaceutical partners by the end of this year and a commercial version of the test in early 2016.
The company is currently embarking on a study of 1,000 to 2,000 patients with different types of solid tumors, in which it will compare its tissue-based targeted sequencing test, FoundationOne, with its new cell-free DNA assay to see how well the results correlate and how they could be used to inform clinical treatment decisions, Stephens said. In addition, it is considering studies in relapse patients after targeted therapy.
By reaching out to physicians that have been using FoundationOne, the company has already signed up almost 1,000 patients for retrospective and prospective analyses. The entire study, to be completed early next year, will be biased towards patients most likely to benefit from molecular profiling, such as patients who relapsed after standard non-targeted treatment. But it will also include patients, up to a fifth, with earlier-stage and lower-grade disease, allowing the researchers to study how different types of tumors at different stages shed DNA into the bloodstream.
The cell-free DNA test, which includes 65 genes with "the most druggable genomic alterations," is already up and running in the company's laboratory and has undergone initial analytical validation, Stephens said.
At the AACR meeting last month, Foundation Medicine presented a poster with preliminary results. The assay performed well using cell lines mixed in different ratios, detecting most genomic alterations accurately. In 75 patients with different stage disease, including breast, lung, and colon cancer, there was broad correlation between tumor stage and the amount of ctDNA.
Early data also suggest that results from cell-free DNA tests are "somewhat correlated with what's going on in the FFPE biopsy, but there is not 100 percent correlation," Stephens said.
In general, several challenges remain with cell-free ctDNA assays, he said. One is that not all tumor types shed the same amount of DNA into the plasma — up to a quarter of patients may not have any measureable ctDNA. Therefore, a negative result from a cell-free DNA assay does not necessarily mean the patient's tumor has no therapeutically relevant genomic alterations.
For example, in a set of 20 sarcoma patient samples, the FoundationOne test found an average of three alterations per patient, whereas a commercial cell-free DNA assay from another company found no genomic alterations in 18 of these patients. While this does not mean a ctDNA assay is inappropriate for all sarcomas, he said, it indicates that certain tumor types release less DNA into the circulation than others. "It's likely to be a subset of sarcoma patients and we need to nail down exactly what it is," he said, and one of the goals of the ongoing study is to reveal which tumor types are unlikely to release sufficient ctDNA.
Another challenge is that tissue-based and cell-free DNA-based tests may find different therapeutically relevant mutations, and without further studies, it is difficult to decide which results are the most relevant for clinical decision-making.
According to Stephens, there are two main reasons for discordant results between the two tests: genomic heterogeneity in the tumor may not be seen in the biopsy but reflected in the liquid biopsy, and cell-free DNA tests can yield false-positive results due to low-level contamination. "The propensity for false positives because of the low minor allele frequencies of the circulating tumor DNA in many patients is really a tremendous challenge," he said.
Foundation Medicine has seen several cases where inconsistent results were likely caused by false-positive ctDNA test results. In one patient, for example, the FoundationOne test found a RET fusion, but a cell-free DNA assay from another company missed that fusion and found an EGFR mutation instead, leaving the patient's doctor to wonder which therapy would be most appropriate. Foundation Medicine conducted its own cell-free DNA test, which confirmed the RET fusion and showed that the patient "absolutely did not have an EGFR mutation," he said.
Is this really a CTC?
Besides a cell-free ctDNA sequencing test, Foundation Medicine has been working on an NGS-based assay for detecting genomic alterations in circulating tumor cells, or CTCs, and presented a poster on this work at the AACR meeting last month.
The company has optimized its assay so it works in just a few dozen cells with almost the same accuracy as in FFPE biopsies, Stephens said. In a collaboration with researchers at the University of Wisconsin, Foundation Medicine scientists showed they can genomically profile CTCs from lung cancer patients with "accuracy sufficient to characterize actionable targets for clinical decision-making," according to the poster abstract.
In principle, CTCs offer a number of advantages over cell-free DNA. For example, because their DNA is not mixed with normal DNA, tumor alterations can be detected more easily. Also, because the DNA is not fragmented, it is easier to sequence entire genes and detect copy number variants. Cells can be assayed in other ways as well, for example by RNA-seq or proteomics. However, one needs to be sure that the cells isolated for analysis are in fact circulating tumor cells, Stephens said.
Around 50 companies in the US provide methods for CTC isolation, according to Stephens, many of them based on cellular phenotypes, such as the expression of certain cell surface markers.
But in a study involving nine patients with pancreatic cancer, Foundation Medicine scientists were unable to find a KRAS mutation in CTCs isolated using technology from an unnamed commercial collaborator, even though they could find the mutation in eight of the nine patients using biopsy material or cell-free DNA. The cells isolated fit the phenotype of what "the vast majority" of CTC companies would define as CTCs, he said, yet they were likely not cancer cells.
The reason they chose pancreatic cancer for the study, he said, is that this cancer type shows little heterogeneity for driver mutations, so CTCs would likely contain these mutations. The company is conducting similar studies with other providers of CTC isolation technology in order to demonstrate that their cells are indeed CTCs.
"We just strongly felt we needed to put these data out there because there is a dogma in the CTC field that you can phenotypically characterize them, and this was the first extremely compelling evidence that that notion is incorrect," he said.