An association of cancer specialists is racing ahead with an ambitious project aimed at improving the quality and efficiency of cancer care by mining patients' electronic health records.
The 30,000-member American Society of Clinical Oncology announced a partnership Wednesday with SAP, the global software giant whose U.S. base is in Newtown Square. They are developing CancerLinQ, a computer network intended to help cancer doctors make treatment decisions for their patients based on the results of comparable patients.
"This has the potential to transform the way we do routine care," said ASCO past president Clifford Hudis, a breast cancer specialist at Memorial Sloan-Kettering Cancer Center.
CancerLinQ is one of several evolving "big-data" initiatives that are uniting powerhouses of oncology and computing. For example, Flatiron Health, the brainchild of two young Wharton School graduates, is backed by Google and some leading cancer centers. IBM's Watson "cognitive" technology - famed for winning on TV's Jeopardy game show - is being trained for oncology with help from Memorial Sloan-Kettering.
The underlying theory is that the growing use of electronic health records opens a rich cache of digital medical information. Stripped of personal identifiers, the data can be analyzed to find the best therapies, reveal harmful effects, and suggest new areas of research.
Since only about 3 percent of patients enroll in the gold-standard type of study, called randomized trials, big data is seen as a faster, cheaper way to gain insights - and keep doctors up to date.
"Big data may greatly expand the capacity to generate new knowledge," two Canadian physicians opined in the Journal of the American Medical Association. And "big data may help with knowledge dissemination."
While it all sounds great, the cheerleaders tend to be long on enthusiasm and vague on details. Even some proponents of big data say it is being hyped.
"It's not going to be the holy grail," said Robert Carlson, chief executive officer of the National Comprehensive Cancer Network. "There's a challenge not only in collecting the data - you also need statistical analytical methods that are standard and validated. It's not clear to me that we have those."
NCCN, a Fort Washington, Pa., alliance of cancer centers that develops clinical practice guidelines, is working with both Flatiron and IBM Watson to integrate that guidance into their analytical platforms.
ASCO, meanwhile, plans to use its own guidelines to let doctors to see how their care stacks up against the best practices and peers' performance.
By the end of this year, the first version of CancerLinQ is supposed to be rolled out at 15 oncology practices around the country that are providing 500,000 patients' records. Among these "vanguard" practices is Cancer Treatment Centers of America, which has a hospital in Philadelphia.
Although ASCO expects to invest "many millions of dollars" in CancerLinQ, it will be provided at "low or no cost" to early adopters, Hudis said.
A prototype of CancerLinQ that focused only on breast cancer was tested in 2013, using records from 170,000 patients around the country. Hudis said the pilot revealed that extended use of a hormonal drug reduced the risk of breast cancer recurrence - the same thing recently shown by two randomized trials.
Yet the goal of that prototype was limited: to prove ASCO could assemble data from differing electronic records systems. "There was not the potential to improve care," Hudis said.
Whether big-data pioneers have the necessary analytical tools, and the buy-in from busy clinicians, remains to be seen. The National Cancer Institute spent seven years developing its $350 million Cancer Biomedical Informatics Grid - caBIG, for short - only to pull the plug in 2011.
A lot of early big-data initiatives in medicine will likely fail, said Bill Hanson, chief medical information officer of the University of Pennsylvania Health System, which is working with Flatiron. "That's just the nature of innovation."
CancerLinQ will be built with SAP's "groundbreaking" database storage and management system, called HANA. It is the basis of a platform now used by Germany's National Center for Tumor Diseases, said David Delaney, chief medical officer of SAP America.
"We have a strong foundation from our previous work," Delaney said.
One of the most promising uses for big-data initiatives like CancerLinQ, experts say, is teasing out information about uncommon occurrences - about cancers, complications, or genetic mutations that an individual doctor rarely sees.
"It's never going to replace clinical judgment," said J. Robert Beck, senior vice president of Fox Chase Cancer Center. "But it may help as we get to a point in the future where we have to go global to test a therapy in patients with different genetic signatures."
"It may be helpful in recognizing unexpected or rare toxicities," echoed Carlson at NCCN.
On the other hand, big data may lead to incorrect conclusions, because patient records can only provide indicators and correlations. Such circumstantial evidence led millions of menopausal women to take hormone replacement therapy, believing it would protect their hearts; a randomized trial ultimately showed hormones actually increase the risk of heart attacks and strokes.
"We need to learn how to use big data," Carlson said, "and how not to."