Detecting many kinds of cancers early, with a simple blood test for DNA shed by tumor cells, is a seductive prospect. Such tests, often called liquid biopsies, are now stepping out of the lab. This week a pioneering real-world study confirms some of their promise—and highlights a potential drawback that may soon confront physicians, regulators, and the public: a small but not insignificant number of alarming but false detections.
In the first trial of such a test in a large, general population, instead of in groups with known tumors or at high risk of cancer, researchers screened nearly 10,000 women with CancerSEEK, an assay for circulating tumor DNA developed at Johns Hopkins University. It detected 26 previously unknown tumors, they report online in Science and at a virtual session of an American Association for Cancer Research meeting today.
Observers say the results are promising, but leave key questions unanswered. “I’m very impressed with this paper. They did all the right things. But it does not mean the assay should be put on the market and sold,” says breast cancer researcher Daniel Hayes of the University of Michigan, Ann Arbor. It’s not yet clear that the early warnings saved the women’s lives. And 101 of the women who had a positive test and received followup imaging turned out not to have cancer; 22 of them underwent fruitless invasive tests such as an endoscopy.
The earlier cancer is detected, the better—but not always, as some tumors may never grow enough to need any treatment. And false alarms can lead to needless anxiety and unnecessary procedures. Such concerns fuel debates over the frequency and worth of existing screening methods such as mammograms and colonoscopies, and the concerns are amplified for cancer blood tests because they promise to make it easy to screen millions of healthy people.
To explore those issues with an early version of CancerSEEK, the Geisinger Health System in Pennsylvania recruited women between the ages of 65 and 75 who had no known cancer history. CancerSEEK pulls circulating DNA from a blood sample and looks for mutations in 16 genes known to drive the growth of various cancers. It combines this information with established protein biomarkers for some cancers. If a woman’s test was positive, an expert panel decided whether she should get a second blood test to confirm the initial finding and rule out molecular changes in her blood that might have fooled the first test. If the positive test appeared real, the woman was invited to undergo full-body imaging.
One-year follow-up results of the DETECT-A study of nearly 10,000 women, ages 65 to 75.
|Initial blood test||9911|
|Cancers detected with blood test||26|
|Cancers detected with standard screening||24|
|Cancers detected with neither approach||46|
|Total cancers detected||96|
|Imaging based on false blood test||101|
|Invasive diagnostics after false blood test||22|
During participants’ first year in the study, dubbed DETECT-A, the Hopkins test picked up 10 types of cancer in 26 women, roughly double the number found with conventional screening. Seven of the cancers, such as ovarian, have no approved screening test and are often deadly because they are rarely found early. Seventeen of the 26 women had early-stage cancers that had not spread far, beyond nearby muscle and lymph nodes. Most of the 26 had surgery or are undergoing treatment and 12 are in remission, suggesting they did benefit from the test.
The DETECT-A study “makes great strides in its approach,” says ovarian cancer researcher David Huntsman of the University of British Columbia, Vancouver. But more data attesting to an overall survival benefit from testing are needed, he and others stress, before the blood test should be approved for widespread use.
The 101 false positives is a modest number given the scale of the study, Huntsman and Hayes say. But the test also had false negatives. Conventional screening over the DETECT-A study’s first year of follow-up found breast, lung, and colon cancers in 24 women whose CancerSEEK test was negative. Tumors in 46 more women became apparent through symptoms or other reasons. The Hopkins team views its test, which it has continued to improve, as “additive and complementary” to standard screening, says Nickolas Papadopoulos, who led the study with Kenneth Kinzler and Bert Vogelstein.
Some public health experts have worried that if a blood test were available, people might shun these other, proven screens. But participating in the study didn’t deter the women from continuing mammograms. And in a follow-up survey completed by nearly 6900 women, only 0.3% said they regretted participating in the study, including only one of 120 with a false positive or negative result. The DETECT-A study “suggests that multicancer blood-based screening is safe and feasible,” says oncologist Minetta Liu of the Mayo Clinic.
Liu is part of an academic group running studies of another blood test, developed by the biotech Grail, which looks at patterns of methyl groups on cell-free DNA to detect cancers. In a March paper in the Annals of Oncology and at the virtual cancer meeting, Grail shared the latest results from a study that enrolled 15,000 people, some with newly diagnosed cancer and others with no evidence of the disease. Its test for 50 cancers had a detection rate of 44% for early- to midstage cancer and also determined the site of the known cancer in 93% of cases, which could reduce the need for full-body imaging. In December 2019, the company launched PATHFINDER, a prospective study of 6200 people—all older than 50 and 70% at elevated risk for cancer. “This is an important next step toward the commercialization of our test,” says Grail spokesperson Kelsey Grossman.
As for when CancerSEEK might reach the clinic, a startup company called Thrive owns the rights to further develop the test. It raised $110 million last year and is planning a large followup trial designed to earn regulatory approval. Questions persist about how officials and expert groups that set screening guidelines, such as the U.S. Preventive Services Taskforce, will weigh early cancer detection tests. Will decade-plus studies that show survival benefits be required, for example, as they were for other screening methods such as mammograms and lung scans for smokers? Or can proxies, such as data showing the tests find more early cancers than existing screening methods, satisfy the various groups? “That is what everybody in the field is grappling with now,” Papadopoulos says.