We’re sharing everything these days. Rides. Spare rooms. Pictures of things wrapped in bacon. Our thoughts, 24 hours a day. Next up: Cancer data?
It can’t come soon enough for Charles Sawyers. Sawyers is one of the world’s most decorated cancer researchers, and his long resumé includes crucial work that helped make imatinib (Gleevec) one of the first drugs approved to go after a cancer—in Gleevec’s case, chronic myeloid leukemia—that was driven by a specific genetic mutation.
While much of the biotech world was running around Philadelphia at the giant BIO convention last week, Sawyers was in Salt Lake City helping run a much smaller conference sponsored by the American Association for Cancer Research.
In his wrap-up keynote, he told his cancer research peers that they needed to break down the walls of their institutions and share the ever-greater volume of data coming from research labs, cancer clinics, and medical centers.
Linking hospitals and research centers via data networks isn’t the stuff of screaming headlines, but people like Sawyers say breakthroughs in care with price tags society can afford won’t come regularly without those data connections. So when I heard about what Sawyers and others were saying in Salt Lake last week, I got them on the phone to hear how much of this advocacy was just talk, and how much was leading to action.
After explaining the need for “a new culture where data sharing across institutions happens much more quickly and easily,” Sawyers said he has convinced his colleagues and the leadership at Memorial Sloan-Kettering to get on board. A big plan for extramural sharing is underway, he said, but he couldn’t discuss details except to say there will be a formal announcement in the fall.
One of those groups sharing with MSK could be Intermountain Healthcare, an integrated health provider and insurer based in Salt Lake City. Its director of cancer genomics, Lincoln Nadauld, told me Intermountain was in discussions with “several academic institutions and integrated delivery” groups to share patient outcomes data. “We don’t want to be exclusive,” he said.
A talk Nadauld gave at the conference illustrated why his organization needs to share. He told the audience that an examination of 72 of Intermountain’s cancer patients with advanced disease hinted that those receiving targeted therapies—the vanguard of personalized medicine—saw their cancers halted (called progression free survival, or PFS) for about twice as long as those receiving standards of care like chemotherapy. What’s more, Intermountain calculated that the personalized medicine care, which included notoriously expensive drugs, didn’t cost any more than the standard care.
“We finally have some data that shows across all cancer types this approach does appear to provide some benefit regarding survival,” said Nadauld. “We also thought that more expensive drugs would make the ‘targeted’ cohort more expensive. But the punch line is that we were able to improve PFS without increasing costs. Neither of those things have come out before.”
But there are big caveats. Not only were the sample sizes small—36 patients in each group—the patients were plucked from records retrospectively, which can expose studies even by the most well-meaning researchers to biases. Nadauld also cautioned that the study hasn’t yet been peer-reviewed.
To know whether Intermountain’s findings are a slice of something more profound, at least two things need to happen: Much bigger patient groups have to be compared, and prospective studies that start from scratch and measure into the future, not back into the past, need to be conducted.
And for those to happen, Intermountain needs to share—first, to conduct stronger retrospective studies, but as time goes on and more patients receive targeted therapies, the sharing of outcomes going forward will be important, too.
Sawyers said small sample sizes will hamstring researchers who don’t share. For example, he cited the “long tail” problem of cancer mutations. When tumors are sequenced, a small number of genes tend to show up with frequent mutations. (P53 is one, for example.) But a chart of all mutations will show a scattering of genes with rare mutations—the “long tail” of the chart, that is. “It never hits zero. The more patients you sequence, the longer the tail gets, and you keep discovering very rare mutations,” said Sawyers.
Some of those rare mutations could be “oncogenes”—that is, the ones in the driver’s seat making the tumor grow—but without the pooled data it will be harder to know. It’s not just looking at more patients with the same type of cancer. Powerful analyses across all cancers might reveal that a mutation known in one type of cancer is showing up in patients across the continent with other types of cancer.
And thanks to new types of clinical trials, patients can be grouped and tested based on mutation profile, not on the tissue or organ where the cancer is growing. “If it’s a single case at one institution, no one will change clinical practice based on that case,” said Sawyers. “But if four patients at four institutions responded to a drug, perhaps you can announce it to the world.”
Lillian Siu thinks a lot about cancer trial design. She’s an oncologist at the Princess Margaret Cancer Centre in Toronto, and she told me genotype matching—trying to get cancer patients with unusual tumor mutations into trials with drugs designed to go after those specific mutations—is catching on quickly. These experimental trials—one prominent example is called LUNG-Map and is run by a consortium of government, industry, and advocacy groups; another just launched by the National Cancer Institute goes by the acronym NCI-MATCH—won’t necessarily give a drug company the big push needed to make a case for a drug’s approval, Siu says. They’re meant for now to be early ways to find promising signals that, say, a drug that fights a type of aggressive skin cancer based on a certain mutation might also fight a lung or ovarian cancer where that same mutation shows up.
Although the list of targeted therapies for cancer is growing every year, Siu says it can’t grow fast enough. Drug development moves slowly, even with regulatory bodies like the FDA loosening the reins on drugs for dire medical needs. So Siu advocates for better trial access while we wait for new therapies to enter the clinic.
More data sharing will help institutions group patients with rare mutations into ever-larger subsets. Siu is also looking beyond genomic information, which isn’t the end-all-be-all of biomedical research. As measurements of proteins, microbes, and even the molecular switches that govern gene expression grow more sophisticated, more research centers will have all kinds of ‘omics for every patient. When that comes to pass, it’s hard to say. Right now, only a few centers, like Memorial Sloan Kettering, have what Sawyers calls “a luxury of resources” to capture partial or full genetic profiles of cancer patients.
But the push toward ever-cheaper sequencing continues. And Sawyers sees a potential “virtuous circle” that could ensue: With more data and better analysis, insurance companies would feel more comfortable covering broader genomic testing, which in turn would allow more clinical centers to do more tests, which would add more data to the ever-deepening pool. Right now, he says, we’re in a transition period, with payers not paying for comprehensive genomic workups “because we’re not seeing the benefit.”
It’s all well and good for powerful healthcare groups to promise to work together; it’s another for them to actually connect their systems. When I spoke last week with Matthew Trunnell, the chief information officer of the Broad Institute in Cambridge, MA, who is moving across the country to the Fred Hutchinson Cancer Research Center in Seattle, he told me his new challenge isn’t just to deepen the Hutch’s internal use of complex data, it’s also to connect to other regional health centers, all by tapping into the cloud-computing, data-sharing expertise of Seattle’s powerful tech companies.
Trunnell said life science practitioners are behind other fields in building 21st century big-data systems that begin to create a “mosaic effect”—the emergence of patterns and solutions from ever-larger collections. It’s what Microsoft researcher Jim Gray, who disappeared at sea sailing off the San Francisco coast several years ago, called “the fourth paradigm of science,” Trunnell told me. “It’s letting the data drive the hypothesis. It hasn’t happened yet in the life sciences.”
That’s in part because of privacy concerns—“the big elephant in the room,” says Charles Sawyers—as well to some extent as institutional pride, ego, and proprietorship. Whether the optimism bears out, and more data with more sophisticated analysis leads to better health for all (not just for those who can afford it), remains to be seen. But the data connections are coming. “Instead of access to data from 3000 patients, you will have access to 300,000 patients,” says Lillian Siu. “Big data days are real, they’re no longer just theoretical. You have no choice.”