Skip to main content
Premium Trial:

Request an Annual Quote

Mutation Count in ctDNA Predicts Patient Response to Immunotherapy

NEW YORK (GenomeWeb) – New data this week has illustrated that a simple method of counting blood-borne DNA mutations in a panel of genes can distinguish patients who have a higher likelihood of responding to cancer immunotherapy.

In the study, published in Clinical Cancer Research on Monday, a research team from the UCSD Moores Cancer Center retrospectively studied liquid biopsy results from a 69 patients treated with these drugs.

Using two different methods — either counting total mutations or only variants of unknown significance — the researchers were able to divide the cohort into two subgroups with significantly different rates of response.

Razelle Kurzrock, the study's senior author and a leader in genomic medicine research at UCSD, said that the impetus for the study was a single case, in which she and her colleagues treated a patient successfully with immunotherapy after noticing an abnormally high number of mutations in her liquid biopsy report from Guardant Health.

The woman was suffering from advanced disease and had no tissue available. "We had one chance to help her and so we took a risk, assuming that the high number of alterations meant that there was a high tumor mutational burden, and gave her a checkpoint inhibitor," Kurzrock said.

"She's gone back to the Mideast now, but she's probably either in complete remission or near-complete remission after being headed for hospice," she added.

Despite this success, the group had relatively low expectations for the study initially, because it didn't seem likely that it would be possible to infer TMB from the limited number of genes assessed in a liquid biopsy assay, Kurzrock said. "But when we started to look at the data it was just obvious to us that there were cutoffs, and when we did the statistics, what looked obvious to us by eyeball turned out to be stat significant," she added.

The results come at a time in which several companies are developing proprietary non-invasive tests that they say will be able to infer whether a cancer patient has a highly mutated tumor and is more likely to respond to immunotherapies.

Testing that targets these more highly-mutated cases — patients with what is now called a high tumor mutational burden — has advanced extremely rapidly, beginning a few years ago with the first data that showed that patients with specific markers for mismatch repair deficiency or microsatellite instability responded much more readily to the new class of immune checkpoint drugs than did other patients.

Other studies found that a mutation counting approach, which distinguishes patients who show a mutation count over a certain threshold in their whole-exome sequencing data, also picks out more likely responders.

Foundation Medicine soon demonstrated that it could use the much smaller number of genes in its FoundationOne tissue sequencing assay to identify the same patients that meet this whole-genome high mutation threshold, making TMB calculation available to ordering physicians as part of its FoundationOne test in late 2016.

Early this spring, Personal Genome Diagnostics said that it was developing a non-invasive NGS assay that gleans patients' TMB from analysis of circulating DNA mutations in a small panel of genes. Guardant Health has also said that it plans to develop a similar test.

Recognizing the advantages of non-invasive testing, Foundation also announced plans last month to create a blood-based version of its tumor mutational burden test. The assay is being developed as a companion diagnostic to Roche/Genentech’s immunotherapy drug Tecentriq (atezolizumab) in first-line treatment of non-small cell lung cancer patients.

The US Food and Drug Administration has not yet approved an immune checkpoint inhibitor specifically for patients with high TMB. But the field recently saw the approval of Merck's PD-1/PD-L1 inhibitor Keytruda (pembrolizumab) for patients with microstatellite instability-high (MSI-H) status or mismatch repair deficient (dMMR) tumors — the first approval for a biomarker-defined population regardless of where in the body their cancer occurred.

In their study published this week, the UCSD team analyzed results from patients who had been tested using Guardant Health's liquid biopsy assay Guardant360, which currently covers 73 genes, and who had also been treated with one of several cancer immunotherapies.

Analyzing test results and outcomes retrospectively for 69 patients with a variety of cancers, the team looked at two different methods for subgrouping patients based on their mutation frequency.

First, they counted the total number of variants. Due to the limited number of genes in Guardant's panel, the team believed they needed as much "genomic real estate" as possible. But they also looked at whether they could use only variants of unknown significance, which is the method that seems to work best in tissue analyses, Kurzrock said, because it removes bias introduced by targeted sequencing panels  that enrich well characterized mutations.

Both methods were able to distinguish a high-responder and low-responder group, although Kurzrock said that the VUS method looked like it correlated a bit better.

Patients that had more than three VUS had a response rate of 45 percent, while among those with three or less only 15 percent responded.

This stratification held fast for other measures like progression-free survival (PFS) and overall survival, Kurzrock said. For example, responders with more than three VUS had an average PFS of 23.2 months versus 11.7 months for those below that threshold. This wasn't statistically significant though — most likely because there were only eight and seven patients in each group, respectively.

As companies like Guardant, Foundation Medicine, and PGDx advance proprietary strategies for blood-based TMB assessment, it is striking that Kurzrock and her colleagues could successfully pick out a patient subset that is significantly more likely to respond to immunotherapy using such a simple counting method.

Kurzrock said, however, that proprietary algorithms are likely to be more accurate than this direct approach, based on her own team's experience using Foundation Medicine's TMB analysis.

"We used to count mutations [from tissue sequencing results] before Foundation Medicine had a formal TMB report," Kurzrock said. "Eight or more we considered high TMB."

"That works … but based on informal analysis we think that whatever they are doing is better than our counting," she added. "We have seen that the correlation with their assay is better."

Based on this, Kurzrock said that the expectation would be that Guardant or other firms would also likely be able to outperform simple counting with a more complicated proprietary method to infer TMB.

The UCSD team's current study was small, so they are now planning to try to extend the results in a larger group of about 400 patients.

The team also plans to study the predictive ability of a liquid biopsy-based TMB assessment prospectively, though this will take longer, Kurzrock said.