As part of our coverage of the 2018 American Society of Clinical Oncology (ASCO) Annual Meeting, held June 1–5 in Chicago, we spoke with lung cancer specialist Dr. Geoff Oxnard, an associate professor of medicine at Boston’s Dana-Farber Cancer Institute and Harvard Medical School. At ASCO, Dr. Oxnard presented data from a study he and colleagues conducted on a blood test that could potentially noninvasively detect early-stage lung cancer.
—Interviewed by Anna Azvolinsky
Cancer Network: First, how is lung cancer currently detected and what is the potential need for an accurate noninvasive lung cancer detection test?
Dr. Oxnard: Currently, lung cancer screening is known to improve lung cancer survival. The randomized NLS [National Lung Screening] trial showed that low-dose CT scan was better than chest x-ray and does make an impact. So, guidelines are increasingly recommending low-dose CT screening for high-risk individuals. The challenge is that it’s not widely used, for a variety of reasons.
Primarily, the logistical infrastructure to make it happen does not exist everywhere. Secondly, there are concerns of false-positives in areas where there is granulomatous disease that will cause benign nodules. For whatever reason, CT scanning makes us recognize that lung cancer screening can be done [and] can make an impact, but there remains a space for an alternative—for approaches that could work with CT screening. So there is a strong rationale for a noninvasive approach, a blood test that could align detection of cancer potentially with fewer logistical challenges, and potentially accurately.
Cancer Network: What was the concept, and the test you and your colleagues are developing that uses subjects’ blood?
Dr. Oxnard: This test uses cell-free DNA, which is currently a standard part of treating advanced lung cancer. We test DNA in the blood all the time to test for key mutations, and we then target those mutations with EGFR inhibitors, ALK inhibitors, ROS inhibitors, et cetera. And because biopsies are inconvenient, we are using cell-free DNA analysis, so-called liquid biopsies, to genotype plasma and look for potentially targetable mutations. This is now standard and FDA-approved ([for example], the liquid biopsy to look for EGFR mutations at initial treatment and [upon treatment] resistance) so we know that this cell-free DNA is robust, scalable, and transportable, and can lead to meaningful information in our clinical care. The question is, can you, instead of looking for specific mutations to target, look for evidence of cancer in general. So this is not cancer genotyping, this is cancer detection, and that is the approach used here. We broadly looked across the genome for any evidence of cancer within the cell-free DNA.
Cancer Network: Can you tell us the important results of this still early, proof-of-principle study? And was there anything in the results that was particularly surprising or novel?
Dr. Oxnard: This was an ATLAS Cancer Genome study screening [Circulating Cell-free Genome Atlas Study], sponsored by GRAIL, Inc. It is a discovery effort, enrolling thousands of patients to identify a cancer signal which is present in the blood of cancer patients but absent in the blood of noncancer patients. It enrolls across the US; the study enrolled patients at more than 140 sites. The cancer and noncancer participants are relatively comparable in terms of age and other clinical variables. So, this is a robust effort to find an approach that can detect cancer with low false-positive rates.
It’s a discovery effort, which means extensive sequencing was done: genome-wide sequencing to look for copy number variation; genome-wide methylation sequencing to look for epigenetic signatures; [and] a targeted sequencing approach across more than 500 genes, to look for specific mutations that could indicate cancer.
One of the most important findings was that you really need to sequence the white blood cells, which are rich with all sorts of mutations, so-called clonal hematopoiesis. These mutations spill into the blood, and you can find these mutations in the cell-free DNA. And you might think there is cancer but [sometimes] it’s not, so you need to control for all of that extra signal in order to achieve high specificity. When you control for the white blood cell mutations, we can develop an approach that can suppress false-positives down to the range of 2%, which is the goal in this assay.
With the understanding that some of these noncancer subjects will have prediagnosis occult cancer, there will be some random rate of cancer in noncancer individuals, so we are not pushing the specificity of 100%. Instead, aiming for a specificity of 98%, we were able to find that these sequencing methods could detect around half of early-stage curable lung cancers and around 90% of advanced-stage lung cancers.
Each of the three different sequencing methods was relatively comparable in terms of detection rate, and the methods could detect cancer in smokers [and] nonsmokers, across stages. Even stage I lung cancers could be detected. And so it does seem to have the legs [indicating] that this could become a diagnostic over time, as a noninvasive cancer detection test.
Cancer Network: What’s next as far as developing and testing this assay further? Do you need to streamline the steps of the assay after you get the patient sample, or is it just [a matter of] validating the assay with additional studies? Also, how unique is this approach for lung cancer? Are there others who are also developing similar assays?
Dr. Oxnard: This approach of doing genome-wide sequencing is very novel. This has not been published before, this broad approach of casting across the whole genome to look for a signal. I don’t know of any other DNA-based detection test like this. There are other tests that have historically tried to use proteomics or other markers. DNA is a really compelling, robust, reliable biomarker in oncology care and that is why this test is really laser-focused on cell-free DNA. This was a discovery effort, and so the next effort is to focus on those areas of the genome that are most important.
The handling of the samples is very simple; the problem is the genomics, figuring out which parts of the genome are most important to sequence and [how] to differentiate signal from noise and cancer from noncancer. So, those are the next steps for optimizing this approach. The eventual diagnostic that will be used is not going to be all three of these sequencing methods. It is going to focus in on the high-yield regions of the genome in order to develop the most impactful and efficient test that can find cancer early.
Once that has been optimized and validated, there are [still] thousands of patients remaining on the TCGA study. We reported on 2,800, but over are 12,000 enrolled. So a lot of patients [are] involved and [there is] a lot of enthusiasm for patients to allow validation of this. And then that final diagnostic will be studied in at-risk patients—so, a real-world screening population. This will really tell us how this test will perform for detecting cancer.
It’s a long road ahead still, but it’s moving very quickly. I did not think we would be here a few years ago. I was quite skeptical, but here we are with good proof-of-principle data, which should convince the skeptics that this is really happening. Patients are enthusiastic, enrollment is very brisk, and so I do expect that we will get the numbers to show that this test has real potential to find cancer early and make a difference. I am really optimistic.