Skip to main content
Premium Trial:

Request an Annual Quote

Pancreatic Cancer Study Finds NGS-Based Liquid, Tissue Biopsies Provide Discordant Results


NEW YORK (GenomeWeb) – A team led by researchers from molecular profiling firm Perthera has published a study comparing results from liquid and tissue biopsies in pancreatic cancer patients.

The study, published this month in Oncotarget, found significant discordance between the two approaches, suggesting that, at present, liquid biopsies are not suitable replacements for tissue biopsies, said George Mason University's Emanuel Petricoin, an author on the paper and co-founder of Perthera.

The researchers looked at 34 pancreatic cancer patients from a number of community-based and academic oncology practices, analyzing them with either of two blood-based next-generation sequencing assays: Guardant Health's Guardant360 test (26 patients), which sequences a 68-gene panel in cell-free DNA on an Illumina HiSeq 2500, and Cynvenio's ClearID (eight patients), which sequences a 50-gene panel on a Thermo Fisher Scientific IonTorrent PGM. Of these 34 patients, 23 also had tumor tissue sequenced using Foundation Medicine's FoundationOne assay, a 321-gene panel sequenced on the Illumina HiSeq 2500.

The liquid biopsy testing detected mutations in 19 patients, with mutations identified in a median of two genes per patient. Tissue biopsies identified mutations in a median of 13 genes per patient, a result that is likely due in large part to the larger number of genes sequenced by the tumor tissue assay.

The researchers also compared results for four commonly mutated genes in pancreatic cancer in the 23 patients who received both liquid and tissue biopsies. They found that nine (39 percent) had matching KRAS status, with the blood-based assays failing to detect 14 KRAS variants detected by the tissue-based assay; and six (26 percent) had matching TP53 status, with the blood-based assays missing 15 variants detected by the tissue assay. The blood-based tests detected no mutations in CDKN2A or SMAD4, whereas the tissue assay detected mutations in 11 and six patients, respectively.

The discordance in the KRAS results were of particular note given that this gene is thought to be mutated in around 90 percent of pancreatic cancers, Petricoin said.

"KRAS is such a great molecular rock to tie [the analysis] onto," he said. "You either see it or you don't see it. And if you don't see it, then it really tells you there's some analytical sensitivity issue, because the tumor has it."

"KRAS should be found and it was not, in a majority of the patients," he added. "That should give one pause."

Justin Odegaard, a senior medical director and lab director at Guardant Health, suggested, however, that the patients analyzed in the Oncotarget study were not appropriate candidates for the Guardant360 test.

As the authors noted, nearly all the subjects "had received some form or systemic chemotherapy" before their liquid biopsies and such treatment "likely has a strong impact on the levels of [cell-free] DNA."

Of the 34 patients sampled, eight had either no or minimal disease burden, and 17 (a group that encompassed the eight with minimal or no disease burden) were either responding to their current therapy or were stable.

The Guardant360 assay is intended for selecting patient therapies and, as such, is meant to be used in patients who are either beginning treatment or are considering switching treatment after progressing on their current treatment, Odegaard said.

"The test is a therapy selector," he said. "You want to use it when the patient is progressing or when you're looking to change therapies. Obviously, you're not looking to change therapies if somebody's doing well, if the patient is responding."

"The test wasn't really being used properly in most of these cases," Odegaard added. "If you use the test at the wrong time, on the wrong patients, you can get information that can be misleading."

Looking only at the 17 patients with extensive tumor burdens and who were progressing on their current therapy, the liquid biopsy assays detected KRAS mutations at a higher rate than in the overall group — seven out of 17 (41 percent) compared to 10 out of 34 (29 percent)—though still at not as high a rate as Petricoin suggested they should be found. Tissue biopsies were available for nine of the 17 patients with extensive tumor burdens and who were progressing on their current treatment. The tissue-based NGS assay detected KRAS mutations in seven of those nine (78 percent).

Despite the discordance observed in the study, Petricoin noted that liquid biopsies are still a valuable source of information on patient tumors, particularly in the case of conditions like pancreatic cancer, where tissue biopsies are often not possible.

"I don't want it to sound like I'm saying that patients who can't get their tumor biopsied shouldn't consider liquid biopsy," he said. "We just have to be circumspect in knowing a tumor biopsy should remain the gold standard. If at all possible, you should get that tumor biopsied."

Odegaard suggested, as well, that some discordance between liquid and tissue biopsies is to be expected, noting that previous research has demonstrated that the two types of assays "capture unique datasets."

"It has been recorded over and over again that these are distinct thing," he said. "Plasma is a little more superficial look at the entire tumor, whereas [tumor tissue] provides a deeper look at one specific portion of the tumor. Tumor heterogeneity basically means those are not going to be the same."

Previous studies have found significantly higher concordance between liquid and tissue biopsies than did the Oncotarget paper. For instance, in one 165-patient study cited by the authors, liquid biopsy detected tumor tissue variants with 85 percent sensitivity. In another 17-patient study, liquid biopsy achieved 90 percent sensitivity.

On the other hand, a recent study in JAMA Oncology comparing the FoundationOne tissue test and the Guardant360 assay found significant discrepancies between the two. In an analysis of samples from nine cancer patients, only 22 percent of variants were reported by both companies.

The two assays identified a total of 45 mutations across the nine patients, but only 10 of those mutations were identified by both tests, and in two of the nine patients, there was no agreement between the mutations identified by the two tests. These differences translated to different treatment suggestions, as well. The tests identified a total 36 candidate drugs based on the patient mutations, but only nine drugs were recommended by both, and in the case of five patients, there was no overlap between the drugs recommended by the two tests.

Perthera was not involved in this study, but the outcomes are relevant to the company's goal of using molecular profiling to guide cancer treatment. Launched in 2013 by Petricoin and venture capitalist Dendy Young, the company aims to provide cancer patients, particularly in the community setting, the sort of molecular profiling and expert interpretation typically available at major cancer centers. It calls this its Precision Cancer Analysis (PCA) service.

The company does not perform the assays itself but helps physicians access genomic, proteomic, and phosphoproteomic profiling services from outside providers, serving as an intermediary by handling the many steps involved in obtaining and sending out patient tumor samples to molecular profiling firms and returning assay results.

The Oncotarget study was part of Perthera's ongoing process of evaluating molecular testing methods to determine which it should include as part of its service. Since its launch, the company has focused primarily on pancreatic cancer through a relationship with the Pancreatic Cancer Action Network. More recently, it has partnered with patient advocacy group Lung Cancer Alliance, and plans to begin doing more work in this disease.


That will include similar evaluations of lung cancer tests, including liquid biopsies, Petricoin said. "We're going to be working with the Lung Cancer Alliance to look at different technologies, different companies' approaches, in the lung cancer field."

"We sit at a good place because with these advocacy groups, they can work with us to look at what really is kind of the state-of-the-art in the field," he said. "We can look at different competing, technologies, and because we're not a lab, we don't have any dog in the fight, so we can look across platforms to see what's working, what's not working. We definitely want to continue to do that."

Petricoin does have a relationship with at least one firm Perthera uses for its analyses. The company gets phosphoproteomic data from Theranostics Health, of which Petricoin is also a co-founder.