Search This Blog

Saturday, June 2, 2018

#ASCO18: Grail’s cancer blood test shows ‘proof of principle,’ but challenges remain


A long way to go, but getting there — that’s the verdict on the highly anticipated data Grail released Saturday about its liquid biopsy for cancer.
The Illumina spinoff is almost as well known for its executive departures and ability to raise buckets of money as for its out-of-the-park goal: detecting tumors super-early, when even cancers with a horrible prognosis might be treatable, by analyzing DNA that has escaped its cells and is floating in the blood.
There are still years of work before Grail has a cancer blood test ready for clinical use. But the four studies it presented at the annual meeting of the American Society of Clinical Oncology demonstrate “a proof of principle,” said geneticist Fergus Couch of the Mayo Clinic, a co-author of a study using Grail’s system to detect breast cancer. “These methods are better for some cancers than others, and there are some cancers we didn’t pick up. But it bodes well that we can detect cancers and not make many mistakes in terms of telling someone she has cancer when she doesn’t.”
The basic idea of circulating cell-free DNA is that tumor cells constantly burst and die, releasing their DNA into the blood. If that DNA contains tumor-specific mutations or other abnormalities, it might reveal the presence of cancer. The challenge is to differentiate cancer-marking DNA from DNA that might look suspicious but doesn’t come from malignant cells.
The early data from Grail’s Circulating Cancer Genome Atlas study suggest that the problem of such “false positives” is tractable. CCGA is enrolling 15,000 participants (it has just over 12,000 so far), 70 percent with cancer and 30 percent without. After a blood sample is drawn at one of the 142 participating medical institutions, Grail analyzes it three ways: detecting mutations in 507 cancer-associated genes; with DNA sequencing to detect gene duplications or losses, which can indicate cancer; and by an assay for abnormal DNA activation, called methylation patterns.
“This is the most exhaustive sequencing you can imagine,” said Dr. Geoffrey Oxnard of Dana-Farber Cancer Institute, who led the lung cancer part of CCGA. “The scope is groundbreaking, because we don’t want to miss any cancer.”

For its ASCO data, Grail analyzed a subset of CCGA including 878 participants with newly diagnosed untreated cancer (20 types) and 749 healthy participants.
Of these, 127 people had lung cancer. After setting the bar for what counts as a cancer signal high enough to keep false positives below 2 percent, the three assays picked up lung cancer about equally well. The methylation test detected 41 percent of early-stage (I-IIIA) lung cancers and 89 percent of late-stage (IIIB-IV) ones. Genome sequencing picked up 38 percent and 87 percent, while the 507-gene assay found 51 percent of early-stage cancers and 89 percent of late-stage ones.
That cell-free DNA picks up advanced lung cancers better than early ones is no surprise, since advanced tumors are larger, among other things. But it’s also a problem. The whole point of cancer blood tests is to detect cancers super-early.
Oxnard called detecting 50 percent of early lung cancers “awesome,” since it might give such patients a much better chance at survival than if their lung cancer is found later.
Outside experts aren’t so sure. “I’m looking for something higher,” said Dr. Debra Patt of Texas Oncology. Low-dose CT finds just under 90 percent of early lung cancers, she pointed out, making it the number to beat. “Ultimately [Grail’s test] is likely to be a better mousetrap, but it still has to evolve.”
Dr. Otis Brawley, chef medical officer of the American Cancer Society, said he’s optimistic that Grail can get its sensitivity numbers up to the 75 percent to 80 percent range, “but it’s going to take time.”
Thoracic oncologist Dr. Naiyer Rizvi of NewYork Presbyterian Columbia University Medical Center, who is not involved in the Grail study, wonders if “early” detection might still be too late: “If a tumor is already shedding DNA into the blood, it might be acting like advanced disease” in terms of the threat it poses. He said the true measure of a test like Grail’s is whether people identified “early” have a higher cure rate than those diagnosed only after they have symptoms. “I view this as very preliminary data,” Rizvi said.
On the positive side, the rate of false positives in Grail’s test seems low. The company had previously announced that the assays detected cancer-like signals (incorrectly, it seemed) in only five of 580 healthy volunteers — but two of those were soon diagnosed with cancer (ovarian and endometrial). That suggests Grail’s tests could have a false positive rate below 1 or 2 percent.
Part of Grail’s secret sauce is subtracting out “this-might-be-cancer” signals from white blood cells. Through aging and stress, those cells accumulate mutations at high rates but in fact never become, say, leukemia. Some 54 percent of the mutations detected in the lung cancer part of CCGA were from white blood cells, not tumors.
“You have to get rid of those to get down to your true tumor-associated mutations,” said Mayo’s Couch. “I’m not aware of anyone else [developing cancer blood tests] who’s doing this.”
The breast cancer part of CCGA, which included 358 women with invasive breast cancer, ran the same three assays. The methylation assay detected 56 percent of triple-negative breast cancers, 34 percent of HER2-positive ones, and 11 percent of hormone-receptor-positive/HER2-negative tumors, all in stages I to III.
“That’s a little lower than we would like,” said Couch. “We’d like to be above 80 percent. But this is just a first pass to show how the three methods do.”
CCGA is also reporting data from 1,627 participants on detecting 20 tumor types at any stage. Again setting the bar high enough to keep false positives below 2 percent, the assays detected 63 percent of colorectal cancers, 58 percent of esophageal, 56 percent of head and neck, 80 percent of ovarian, and 60 percent of pancreatic, in each case for stages I to III. But the test flopped, detecting less than 10 percent, in cancers of the thyroid, uterus, and kidney.
“In some other cancers we do have a lower signal,” said Dr. Anne-Renee Hartman, Grail’s vice president of clinical development. But CCGA is intended to identify the most promising of the assays and optimize them, she said. “We do not intend to bring all three assays into the final product.”
An even bigger challenge than detecting cancer early is distinguishing tumors that are so indolent they will never threaten health. Ideally, a blood test would not flag those cancers, since a diagnosis would send patients for more testing and, in many cases, risky treatment for no purpose. Research to determine that lies well in the future.

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.