Geoffrey Oxnard, MD, discusses testing error underlying liquid biopsy discordance, steps clinicians can take to optimize liquid biopsy, and where he sees this evolving in the treatment landscape for patients with cancer.
Geoffrey R. Oxnard, MD
The advent of liquid biopsies in oncology has led to a series of opportunities and challenges, explained Geoffrey R. Oxnard, MD. While the diagnostic approach is noninvasive in terms of cancer genotyping, false-negative and false-positive results continue to be obstacles.
"We know that liquid biopsies have limitations; not all cancers shed tumor DNA, so false-negatives are relatively common," said Oxnard, a medical oncologist at Dana-Farber Cancer Institute and associate professor of medicine at Harvard Medical School.
False-negative results have been associated with low tumor DNA shed, limiting the assay’s ability to detect mutations. False-positive results are associated with the heterogeneity of the tumor, as well as tumor DNA shed from normal cells that cause assays to pick up mutations that are not tumor-derived, according to a paper recently published in JCO Precision Oncology by Oxnard and colleagues at Dana-Farber Cancer Center.1
Oxnard advocates for benchmarking different assays in order to optimize the results of liquid biopsies. In a paper published in in JCO Precision Oncology in March 2019 by Daniel Stetson, MS, et al, the study authors investigated the discordance between liquid and tumor biopsies by comparing 4 next-generation sequencing (NGS) assays for mutations in circulating tumor DNA.2
According to Stetson et al, positive predictive value against tissue ranged between 36% and 80%. However, the assays were more concordant when mutations had an allelic fraction (AF) greater than 1% versus those below 1%.
“One of the main takeaways is that we need to be a little more suspicious about variants that are caught at low levels where assay error can at times get in the way of accuracy,” said Oxnard.
In an interview with OncLive, Oxnard discussed testing error underlying liquid biopsy discordance. He also highlights steps clinicians can take to optimize liquid biopsy and where he sees this evolving in the treatment landscape for patients with cancer.
OncLive: Can you share some background on liquid biopsies and their current utility in cancer?
Oxnard: Cancer genotyping can be logistically tricky because there often is inadequate tumor tissue available for genomic analysis. When a biopsy is inconvenient or impractical, a liquid biopsy can be used in attempt of noninvasively getting the genotype. This can be done in advanced lung cancer where the genotype is unknown, where EGFR testing is an FDA approved approach to get the genotype, and this can be done at resistance where the resistance mechanism is unknown, where plasma genotyping can be used to find a targetable resistance mechanism.
About 20% to 30% of the time a mutation will be missed [with liquid biopsies].. For this reason, the FDA states that these are approved as a test which reflects tumor testing if the blood test is negative because a negative blood test is not definitive.
Can you provide some background on the issue of testing errors with liquid biopsies?
We are coming to better understand why it is that sometimes a liquid biopsy and a tumor biopsy find different results. The most common reason is that there is no tumor DNA in the plasma DNA; that’s a low-shedding tumor that would lead to false-negative results.
Increasingly, we are understanding other reasons underlying this discordance of why a blood test finds something different than a tumor biopsy. For example, another cause of false-negatives is that the mutation that the test is not well built to find the complex variant being detected. For example, amplifications are particularly hard to find in plasma and will commonly be missed by a blood test unless there is high tumor content.
On the flip side, we are also seeing that false-positives are increasingly common. The problem with false-positives is for various reasons. Tumor heterogeneity can explain this, but more often, we see it is related to clonal hematopoiesis, mutations in the white blood cells that are shed into the plasma and are found on the liquid biopsy. They seem like they may be tumor mutation, but in fact they are a benign mutation that is circulating in the blood. Both false-negatives and false-positives can be seen on liquid biopsies, and doctors must be aware of this.
How frequently do false-negatives and false-positives occur?
If you are asking a simple question with a simple assay, you will get a simple answer. For example, if you are just looking for EGFR mutations with a targeted EGFR assay, false-positives are extremely rare, and false-negatives happen about 30% of the time because the sensitivity of these assays in metastatic lung cancer is around 70%.
If you are asking a complicated question with a complicated assay, for example, if you are doing tumor genotyping using an NGS assay, it is in fact quite common to find mutations in the plasma that are not derived from the tumor. How often is that the case? In my anecdotal experience, most liquid biopsies find some low-level mutation that may not be derived from the tumor. What the clinician needs to do is be able to scrutinize any given mutation that is found is likely to be tumor derived or not.
Are there any steps that can be taken by labs to confirm the accuracy of liquid biopsy results?
We would advocate the benchmarking of assays versus each other. This paper is a bake-off paper where multiple different assays were compared to get a better sense of truth. I think laboratories need to do a good job validating their findings against other assays to minimize the chance that assay artifact or error is called a mutation.
Otherwise, there is no real way to screen out the possibility of finding a white cell mutation in the plasma besides paired sequencing of the white blood cells. Cancer detection assays are increasingly sequencing the white blood cells at the same time as the plasma in order to eliminate the possibility of mistakenly calling a white blood cell mutation a tumor mutation.
What is the significance of the results from the study by Dr Stetson and his co-investigators?
By using plasma from cancer patients, Stetson et al compared 3 different NGS assays sequencing plasma DNA to assess their reliability and their concordance. They found that the assays were more often concordant for variants above 1% AF. For mutations below 1%, certain assays found them while some didn’t, and there was concern about both false-positives and false-negatives at this lower region. One of the main takeaways is that we need to be a little more suspicious about variants that are caught at low levels where assay error can at times get in the way of accuracy.
The other important finding was that there are various reasons behind some of the discordance, and one of the most important reasons they found was due to the bioinformatic pipeline. The sequencing assay both involves sequencing and how to interpret the sequencing. One thing they point out is that assays can improve their interpretation or their bioinformatics, how they turn sequencing reads into variant calls. This is a major area where assays can improve their performance by double checking their bioinformatics and optimizing their bioinformatics for greater accuracy.
What is the take home message from this that oncologists should be aware of?
Our main message is that we advocate for better validation of the liquid biopsy assays we are using in the clinic. We as clinicians should demand rigorous benchmarking against comparable assays to optimize accuracy. We realize that false-negatives are going to be seen due to tumor shed, we realize that false-positives may be seen due to clonal hematopoiesis, but assay error, such as false-positives due to artifact or false-negatives due to missed mutations is not satisfactory. We, the clinicians, need to demand greater scrutiny of these assays so they perform optimally for our patients.
How do you see liquid biopsies evolving in the future?
A couple important ongoing areas of investigation include using these tests for cancer detection either as a minimal residual disease assay or as a cancer screening assay. This is an area of active investigation. A separate future application is in cancer monitoring using repeat blood tests to assess the effects of a treatment; both of these approaches are investigational right now, but over the coming years we hope to find and be able to demonstrate assays that are accurate and useful for these purposes.