Five months ago, Grail Inc. began marketing the first-ever prescription test intended to detect more than 50 types of cancer — most of which have no recommended screening test — with a mere blood sample.

“We believe we have the potential to transform cancer care by reducing cancer deaths and decreasing health care costs by detecting more cancers earlier,” the five-year-old company declares on its website.

If this is the first you’ve heard of Grail’s potentially revolutionary product, one that may have found a Holy Grail of oncology, don’t feel uninformed.

The basis of Grail’s Galleri test, and similar tests being developed by other companies, is that fragments of DNA, shed into the bloodstream by a tumor, can be isolated and sequenced to find genetic alterations that are signs of cancer.

Huge advances in genomic analysis and data science have bolstered this concept. But as skeptical experts point out, the decades-long quest for a routine, affordable, multi-cancer test to flag a malignancy in a person with no symptoms is still more quixotic than clinched.

Here’s why.

Basic biology

The 1948 discovery that scraps of DNA can get out of cells and float around in the blood had no practical applications until recent decades.

Now, sequencing of circulating tumor DNA is used to predict the aggressiveness of certain cancers, to monitor patients for relapse, and to select treatments targeted to particular cancer-causing mutations.

Sequencing of “cell-free DNA” shed by a fetus into the mother’s blood is the basis of non-invasive prenatal screening for genetic disorders such as Down syndrome.

Grail’s cancer screening test grew out of a study of blood samples from 10 pregnant women who had abnormal prenatal screening results — and subsequently were diagnosed with various malignancies. Grail researchers reexamined the DNA sequencing data from the women’s prenatal tests and discovered the women had genetic abnormalities that did not match the fetal abnormalities — presumably because the women’s tumors were also shedding DNA into their blood.

“In one case, blood was sampled after completion of treatment for colorectal cancer and the abnormal pattern was no longer evident,” the researchers wrote in 2015 in JAMA.

Still, a test that uses circulating tumor DNA for early detection must overcome a basic biological obstacle: Most budding tumors shed so little DNA into the blood, and it breaks down so quickly, that a standard blood sample is not likely to contain an analyzable fragment.

Eleftherios P. Diamandis, head of clinical biochemistry at the University of Toronto and Mount Sinai Hospital in Toronto, has published estimates of how big a tumor would have to be for detection with tumor DNA. He and his associate Clare Fiala extrapolated from other researchers’ studies of the amount of cell-free DNA in prenatal and cancer diagnostics.

A tumor would have to be at least one centimeter in diameter (a third of an inch) and weigh 1 gram (three-hundredths of an ounce) to be detected with a routine 10-milliliter (about three teaspoons) vial of blood, Diamandis and Fiala calculated. Although that size is early stage, it’s big enough to be detected with tried-and-true technology such as mammography or MRI. A tinier clump of malignant cells probably would not shed enough DNA for it to be captured and sequenced.

What about taking a bigger blood sample? Imran S. Haque, a computational biologist and vice president at Recursion Pharmaceuticals, has estimated that 150 to 300 ml of blood — up to half a pint — would be needed. That, he wrote, is “more blood volume than practical for a routine screen.”

Studies have revealed other problems. Some mutated DNA in the blood is harmless and comes from non-cancerous cells. Some healthy people carry cancer-associated mutations without developing cancer.

Strategies have been proposed to overcome these problems. But adding more analytical procedures “is costly and time-consuming, thus limiting the attraction of tumor DNA as a population-based screening test,” Diamandis and colleagues wrote in April in Clinical Chemistry and Laboratory Medicine.

Validation

Grail was founded in 2016 by Illumina, the global giant in DNA sequencing technologies, after Illumina funded the study of pregnant women’s blood samples.

Grail leads the race to create a multi-cancer early detection test, but rivals are also in development, including CancerSEEK, TEC-Seq, PanSeer, and Delfi.

Development of such genomic tests starts with studies that are, by design, rigged. Blood samples from newly diagnosed cancer patients and healthy individuals are analyzed to assess the test’s ability to detect signals of cancer. Most published studies to date involve this essential but exploratory validation.

The results, even if promising, do not prove that the test would work in real-world screening, much less reduce deaths. Definitive clinical trials showing a ”mortality benefit” can take decades.

How does Grail’s Galleri, which is being studied in more than 50 types of cancer, stack up so far?

Of 1,254 cancer-free patients, Galleri correctly found no cancer signal in 99.3% of them, according to a study published in June in Annals of Oncology.

Grail researchers have highlighted the importance of this result, because in real-world screening, a false alarm, or “false positive,” is a big deal, triggering anxiety and diagnostic workups that may include surgical biopsies. In a 2020 paper, the researchers calculated that for every 100,000 people screened in real-world conditions, Galleri would return 1,406 positive results, of which 691 would be false positives — less than 1% of those screened.

» READ MORE: False-alarm mammograms can discourage cancer screening, research finds

But William C. Taylor, a Harvard University physician and population medicine researcher, called that a “misleading” miscalculation. If 691 of 1,406 positive results are wrong, the false positive rate is 49% — almost half.

While flagging a cancer that doesn’t exist is a big deal, missing one that does exist could be devastating.

Of the 2,823 newly diagnosed cancer patients in the validation study, Galleri correctly identified the vast majority who had advanced, often incurable cancers — 81% of Stage 3 and 93% of Stage 4.

Galleri also pinpointed the organ of origin for 93% of the cancers it identified by analyzing “methylation” — patterns of hydrocarbons added to the DNA fragments.

The test’s accuracy fell, however, at the earlier stages that are the key to Grail’s laudable aspirations. Overall, the test missed 82% of Stage 1 and almost 60% of Stage 2 cancers — although this rate varied based on cancer type. For example, the test identified more than 70% of early-stage pancreatic tumors, a particularly aggressive malignancy that is usually diagnosed too late.

Overdiagnosis

A counterintuitive problem with some cancer screening tests has been identified: Early detection is not always the best thing because some tumors never become harmful. Prostate cancer screening with the PSA blood test and, to a lesser extent, breast cancer screening with mammography, have come under fire for “overdiagnosis” — finding tiny, innocuous tumors. Such tumors usually wind up being treated to err on the safe side, but treatment is not innocuous. Surgery or radiation for prostate cancer, for example, may leave the patient with sexual or urinary problems, or both.

In August, Grail researchers published a new analysis of their validation data that concludes Galleri “will not lead to overdiagnosis.”

To reach that conclusion, they correlated survival rates with Galleri detection rates. It turned out that the test detected 89% of the cancers of patients who died within three years, but only 44% of the cancers of survivors.

The researchers interpreted this to mean their test has “prognostic significance” because “cancers not detected … had better prognosis than cancers detected.”

Since this pattern could also be due to Galleri’s failure to detect most early-stage cancers, the researchers looked at some other indicators. The patients with cancers that Galleri missed, especially those with late-stage cancers, had “notably better survival” than would be expected, based on federal cancer surveillance data.

Joshua Ofman, Grail’s chief medical officer, said this suggests Galleri won’t detect cancers that don’t need treatment.

“We are likely not detecting the cancers that are indolent,” he said.

Diamandis rejected that idea.

“These authors spin their [poor] findings to their advantage by saying that if they do not detect the tumor, it has better prognosis,” he said. “Grail is missing most clinically important tumors and along with them, as expected, it is also missing the clinically unimportant ones.”

Home-brewed tests

In May 2019, Grail announced that its test had been granted “breakthrough device” designation by the U.S. Food and Drug Administration. Manufacturers voluntarily seek the designation to help speed up development, review and approval.

Grail now has several large clinical studies underway to evaluate how the test performs in real-world situations.

“We are absolutely committed to going through the process to get FDA approval,” Ofman said. “That will take some time. There’s a lot of work to be done.”

How can Grail already be marketing the unapproved test — with a list price of $950, not covered by insurance — directly to doctors?

Like many biotech companies, Grail is using 1988 regulatory pathway that critics call a loophole. Any lab that is certified by the Centers for Medicare and Medicaid Services to perform high-complexity tests can market a new one without review, as long as the test is designed, manufactured and used within that single lab.

Such tests used to be called “home brew” because they were as simple as making beer at home. Now, laboratory-developed tests are often more like high-tech alchemy.

Beginning two decades ago, the FDA announced its intention to regulate lab developed-tests that posed risks to public health. (FDA pressure put the kibosh on one example, an ovarian cancer screening test called OvaSure, but not before some screened women removed healthy ovaries.)

However, opposition from both industry and academia has mostly stymied the FDA’s proposals for an overhaul.

“My prediction,” Diamandis said, “is that when this [Galleri] test is used for real screening in asymptomatic people, the results will be a disaster because most early cancers will be missed, and a lot of people without cancer will be sent for unnecessary biopsies.”

Grail’s Ofman said, “We feel very comfortable that we are putting a well-validated test into the market.”