Blood Simple? Grail’s Cancer Screening Quest Faces Tough Questions

Xconomy National — 

[Corrected, 2/4/16, 2:05pm. See below.] Three weeks ago the world’s biggest genomic sequencing company, Illumina (NASDAQ: ILMN), unveiled a spinout called Grail to make blood tests that could spot all kinds of cancer in seemingly healthy people, perhaps as soon as 2019.

If you know anything about the history of screening healthy people for cancer, you’re right to be skeptical.

The people behind Grail can sympathize. “I was the in-house skeptic,” says Rick Klausner, Illumina’s chief medical officer since 2013. “I told Illumina, ‘Don’t go near it.'”

Klausner, who has also run the National Institutes of Health’s National Cancer Institute and was the top global health official at the Bill and Melinda Gates Foundation, has changed his tune.

But from conversations with Klausner and many others in the field of cancer research, treatment, and prevention, it’s clear that the scope and audacity of Grail’s plans will require measurements and analysis that stretch the understanding of biology, as well as strategies that run counter to much of the current thinking in public health.

“In the past we’ve made promises to the public that didn’t go well,” says Scott Lippman, director of the University of California San Diego’s Moores Cancer Center, speaking about the cancer medical field. (He is not affiliated with Grail.) “Given the molecular complexity of cancer and its premalignant origins, this is a big promise to say we’ll take a drop of blood, tell you if you have cancer, where it is, and how we might cure it.”

Grail launched last month with more than $100 million committed from Illumina, which is majority owner, Arch Venture Partners, Sutter Hill Ventures, Bill Gates, and Bezos Expeditions, the umbrella organization for founder Jeff Bezos’s extracurricular interests and investments.

Illumina warned investors its earnings per share would dip by 15 cents this year because of its Grail investment. It’s only a 3 or 4 percent drop, based on Illumina’s own estimates, but it’s tangible.

Grail has yet to name a CEO. An announcement could come soon. So it was Klausner, a Grail board member, who spent time on the phone with me last week to explain how Grail intends to develop a blood test that combs through the entire genomes of healthy people for the faintest traces of cancer—even tumors that are too small to cause symptoms or be detected in other ways.

Grail must also prove to doctors, patients, health regulators and insurers that its high-tech detective work actually results in better outcomes for patients. The need to make people’s lives better cannot be overstated. “The major challenge in early cancer detection is that there is always harm to some degree, no matter what the screening test,” says Lippman. “So the amount of benefit you have to show must be even higher.”

For decades, many clinicians, medical societies, and advocacy groups like Susan G. Komen have urged healthy people to undergo regular testing for various cancers. Screening saves lives, goes the mantra. Problem is, most tests have made it into wide use without first undergoing the long-term, randomized trials that many researchers believe are the definitive way to show that a test actually saves lives rather than, for example, uncovering tumors that were never destined to spread or otherwise harm the patient.

What’s more, screening can hasten death or injury through unnecessary biopsies, surgeries, and drug regimens for people wrongly diagnosed or whose cancer might never have progressed.

Take, for example, the prostate-specific antigen, or PSA, screen that was once a staple of men’s annual checkups. At best, one in 1,000 men avoids death from prostate cancer after taking the test, according to U.S. Preventive Services Task Force, a volunteer group of medical evidence and prevention experts overseen by the U.S. Agency for Healthcare Research and Quality (AHRQ).

But of those 1,000 men, 30 to 40 will develop erectile dysfunction or incontinence, two will have a heart attack or other serious cardiovascular event, and one will develop a serious blood clot—all from the cancer treatment. These numbers don’t even count complications from invasive biopsy. Given this unfavorable balance of benefits and harms, the Task Force now recommends against routine use of the test.

“The point of screening is to reduce suffering and death, not just to find more stuff,” says Jennifer Croswell, medical officer at the AHRQ.

So here comes the liquid biopsy: a run-of-the-mill blood draw to catch rare bits of DNA and similar molecules shed by a patient’s cancer cells and analyze the mutations in them. Liquid biopsies are starting to gain acceptance as a way to help doctors care for patients already diagnosed with cancer—to help guide treatment based on the tumor’s genetic profile, for example, or to monitor a patient’s response to a drug regimen or surgery.

But is there enough cancer DNA floating around in the blood to red-flag microscopic tumors in otherwise healthy people? Even if Grail—or its competitors in the cancer screening race—could produce a test that was 100 percent reliable at detecting cancer (a statistical impossibility, but let’s just say), could it distinguish between a slow-growing, non-threatening tumor—a turtle, as Croswell puts it—from a fast-growing, dangerous cancer—a rabbit? And could the test makers eventually prove their tests are reducing suffering and death, not just finding more “stuff”?

Grail says yes to all of the above, and company officials say the first pan-cancer tests for broad swaths of the population could be available in 2019 for $1,000, perhaps less. Here’s their plan.

Phase one has a narrower focus, to test only people who have been screened with other methods—for example, a mammogram for possible breast cancer, a CT scan for lung cancer, or PSA for prostate cancer—that have found something suspicious. Instead of heading right to an invasive biopsy, they could do a confirmatory test (also known as a reflex test) with Grail’s product. Reflex tests could become available in the next year or two, says Klausner.

Near-term clinical studies to prove their worth should take months, not years, he says, because healthy people are getting screened for cancer all the time with mammography, PSA tests, and the like. If immediately after a positive screen they take a Grail blood test and also submit to a tissue biopsy, Grail can compare results. If the tissue and liquid biopsies match up over hundreds, even thousands of patients, it might be evidence enough to justify diverting people away from tissue biopsies after a negative liquid biopsy.

Grail is already talking to “a variety of partners” to set up those studies, says Klausner.

Phase two—far more ambitious—is proving Grail’s blood test should be used to screen for a cancer in a wide healthy population, not just as a follow-up for those who’ve already had a positive mammogram, CT scan, or PSA test. To do so, Grail wants to partner with a large national government—perhaps more than one—that has single payer healthcare, Klausner says.

The Grail test would become part of the national health program, just as an experiment, at first. That process would then lead to huge reams of data—the whole genomes of hundreds upon hundreds of thousands of people—that Grail hopes would prove the test’s utility.

The massive trial would hinge upon a concept called “stage shift.” The medical field measures a cancer’s size and reach in stages. Stage one is the smallest and most contained, and stage four typically means it has spread throughout the body and is nearly always fatal. Grail wants to detect people’s cancer at the lowest stage possible—shift the detection from higher to lower stages—because, Klausner says, a lot of data suggest doing so means better outcomes.

The broad outline from Klausner goes like this: Test as many people as possible to get baseline numbers on how many people have stage one cancer, how many have stage two, and so on. Those who test negative get follow-up tests at regular intervals. Over time, if the screen works as hoped, the cancers found during each follow-up round will exhibit a stage shift: an increase in stage one and two cancers and a decrease in stage three and four cancers, which implies that by catching (and treating) the lower-stage tumors the program is preventing some high-stage tumors from developing.

There are big logistical and statistical caveats with this approach.

When testing for a stage shift, finding more cancers earlier is only relevant if the number of late stage cancers does, in fact, go down, says Barry Kramer, director of the Division of Cancer Prevention at the National Cancer Institute. If there is little or no decrease in late stage disease, he says, there is probably no reason to use the screen. In other words, the shift must be absolute, not relative, to be promising. The difference in practice is subtle and tricky to tease out, says Kramer.

Klausner understands the danger—“we’re not naive about these problems”—and says Grail will indeed test for an absolute, not relative stage shift.

[This paragraph previously said Grail “will not” do a randomized trial. It has been changed to reflect more uncertainty. We apologize for the error.] But Grail isn’t likely to do what many, including Kramer, say is the best way to know if a screening method is doing more good than harm: randomized controlled trials measuring a test’s effect on patients’ mortality. “Speaking very generally, the accepted measure of screening’s effectiveness is preventing or delaying deaths,” says Donald Berry, a biostatistics professor at University of Texas M.D. Anderson Cancer Center and a member of Grail’s advisory board.

Berry declined to comment on Grail’s plans specifically. But Klausner said he doubts Grail will design randomization into its trial, adding that randomization isn’t necessarily the best way forward in a massive population study. Details of the trial design will emerge over the next year, he says, after input from Grail advisors like Berry and from “the broader medical community. It will be an exhaustive and interactive process.”

And while measuring the effect of the screening test on mortality is still a long-term goal for the Grail study, Klausner has his eye on a nearer horizon: stage shift data that are so dramatic that the medical community is convinced to adopt the test by 2019.

It’s a shockingly ambitious timeframe, possible to even contemplate because Grail has the backing of Illumina, its majority owner and provider of discounted technology, and the financial backing of some of the world’s richest people. They’ll need the horsepower and cash to crunch the entire genomes of hundreds of thousands of people with “ultra deep”—read “extremely redundant and extremely expensive”—sequencing.

Not everyone is convinced the technology—including what Klausner says is a “secret sauce” of analytical improvements—can find the right bits of DNA sloughed off by tiny tumors into a patient’s bloodstream. “You’re saying a few million cancer cells [in a microscopic tumor] will have shed exactly the four or five mutations you need to look at,” says Chirag Patil, a neurosurgeon at Cedars-Sinai in Los Angeles who also researches precision cancer treatments.

To consistently find those multiple mutations, “you’d have to have a breakthrough technology that in real time is sampling your blood, not taking a tiny snapshot” with a small blood draw, says Patil. (He thinks morning urine, which has spent all night collecting DNA fragments, will eventually reveal better results.)

At Johns Hopkins School of Medicine in Baltimore, a research team reported in a 2014 paper that the smallest tumors shed enough DNA into the blood to be detected about half the time. Klausner says pushing the number higher than that isn’t the main goal. “The challenge is not whether you can you find every cancer,” he says. “The most challenging thing is detecting the difference between the cancer we care about and the overdiagnosis of cancer,” which is shorthand for mistaking for dangerous cancer either tumors that aren’t destined to be dangerous or things that look like cancer but aren’t.

Healthcare providers are becoming more sensitive to cancer overdiagnosis and the resulting overtreatment. In prostate cancer, for example, clinicians are already building programs to divert patients with likely nonthreatening tumors away from treatment and into a monitoring program. Even though Behfar Ehdaie, a urologic surgeon at New York’s Memorial Sloan Kettering Cancer Center makes his living removing men’s prostates, he eventually wants to see 60 to 70 percent of his patients moved instead into the hospital’s “active surveillance” program.

Which cancers to treat and which ones to just keep an eye on is just one of a bevy of medical questions raised by the prospect of a screening test like Grail’s.

What about the tests that don’t pick up anything? Does that mean someone is cancer free?

“As clinicians we need to avoid false reassurance,” says Joshua Schiffman, a pediatric oncologist at the University of Utah. He specializes in treating families with high cancer risk, including Li-Fraumeni Syndrome, in which a genetic mutation means carriers have a nearly 100 percent chance of getting cancer in their lives.

Or what if a patient’s blood screen picks up a cancer DNA signal, but a follow-up image test shows nothing?

When asked a similar question last month on a conference call, Illumina CEO and Grail chairman Jay Flatley said perhaps the patient simply gets monitored, or retested three or six months later “to measure the rate of change of the overall mutation in the blood.”

“It could be that rate of change determines when you need to begin treating it, or if you ever need to begin treating it, because in some cases these mutations exist and the body is handling it and the cancer will never become lethal,” said Flatley.

It’s unclear if Grail knows more about that rate of change concept than it’s letting on. Beyond one data fragment revealed last month—Flatley said Illumina has an assay that can pick up from a single tube of blood 88 percent of cancerous mutations typically found in lung cancer tissue samples, better than previous attempts—no one is saying much about what’s under the hood. “People who haven’t seen the data will be skeptical, and those who have seen it, joined the Grail scientific advisory board,” says Bob Nelsen of Arch Venture Partners, a Grail investor. (I reached out to all five non-Illumina SAB members; those who responded declined to comment on the Grail data.)

Klausner did describe, however, the data that turned him from skeptic to believer. They came out of a prenatal test from Verinata Health, a company Illumina bought in 2013.

The test was picking up advanced and previously undetected cancer in pregnant women who were checking the health of their babies. It was a small number, says Klausner, but “consistent” with what statistics would predict in that particular group of women. The light bulb went on, he says, when there were no false positives among the 150,000 women tested. Blood tests had the potential to make the right call—at least with advanced cancers.

Despite the new competition, some liquid biopsy companies I spoke with were excited to see how Grail makes out. “I’m a fan of the effort,” says Andre Marziali, CEO of Boreal Genomics in Vancouver, BC, which provides underlying technology for liquid biopsies but doesn’t intend to market tests on its own. He says Grail will collect an enormous amount of clinical data, more than any other entity could do. Whether its test becomes a good business or not, “Grail will accelerate the arrival of ctDNA screening,” says Marziali, using the shorthand for “circulating tumor DNA.”

Victor Velculescu, one of the Johns Hopkins researchers at the forefront of liquid biopsy research—and a co-interim CEO of Hopkins spinout Personal Genome Diagnostics—says the arrival is just a matter of time. “Within five years one can expect some sort of clinical test to be available for healthy populations,” says Velculescu.

New technologies often acquire an aura of inevitability. It’s important that checks and balances are put in place to ensure liquid biopsies that screen healthy people actually work. Because, as we learned from the prostate test, better health through screening is not inevitable, as Memorial Sloan Kettering urologist Behfar Ehdaie notes: “Probably one million men in the PSA era have gotten surgery that did no good, at best, or perhaps did them harm.”