Why big data must be used in the fight against cancer
Cancer is one of the leading causes of death, with some 1.7 million Americans expected to be diagnosed with some form of the disease this year alone. Yet in the vast majority of cases, the details of their disease and treatment—from the tumor’s composition to which drugs were tried and what the results of those treatments were—remain stuck in medical records that offer no help to others facing similar circumstances.
Instead, most of what’s known comes from the 3 percent of patients who’ve taken part in clinical trials, leaving enormous gaps in our understanding of the multiplicity of diseases that are grouped together as cancer and why certain patients respond or don’t to various treatment regimens.
In many ways, it’s puzzling that the power of big data is not being unleashed to the extent it might to help cure cancer. Therefore, there’s a real opportunity for big data providers, the healthcare industry, health policy makers and the new administration in Washington to step up.
We are exposed every day to advances in healthcare technologies that are creating the opportunity for truly personalized medicine, but greater advances are possible. One of the most significant involves the creation of massive databases that would guide treatment plans based on the evidence of what works and what doesn’t in specific situations.
Already, both public- and private-sector efforts are underway to build such databases of health information from cancer patients that could rapidly enhance our disease understanding and our ability to personalize therapy, potentially extending lifespans and quality of life for millions of people.
Such large-scale data exchange was a key element of the Obama administration’s cancer “moonshot” program overseen by Vice President Biden. The incoming Trump administration has yet to take a stand on it. The current Congress can help by voting to continue funding. The new president can make an enormous difference in the lives of cancer patients by pledging his support.
Why does this matter? Take a step back for a minute to understand the problem of cancer treatment. Tumors are marked by their heterogeneity and instability. A treatment regimen that shrinks one person’s tumor may have no impact on another person’s, and even within the same individual a treatment may appear to work while missing a cluster of cells with a different genetic makeup. Furthermore, tumors are continually changing because of the evolutionary pressure imposed by cancer treatments, which often causes resistance.
Personalized medicine is changing cancer treatment by enabling doctors to target therapies to tumors with specific molecular aberrations. However, our reliance on clinical trial data, which are often limited to specific late-stage cancers, has curtailed the impact of personalized medicine to date. As an example, a multiple myeloma (blood cancer) patient who might benefit from a therapy used more traditionally in melanoma (skin cancer) may never receive it because their tumor was never tested for the molecular aberration relevant to the melanoma therapy.
To get the most out of these new treatments, there needs to be broader testing up front—so that tumors’ molecular fingerprints are identified at diagnosis—and over time—with changes to their molecular composition monitored throughout treatment.
Unfortunately, today, the reverse happens: Broader testing is often done only in later stages of the disease, at which point tumors are more heterogeneous, resistant and generally harder to treat.
This isn’t just a question of adding more tests, but of using information to improve treatment. By combining the information gleaned from testing with other clinical and outcomes data, big data analytics can be used to learn more about the variety of cancers. Each new piece of information may seem insignificant on its own, but when added to the database, becomes meaningful. The more data, the more likelihood of uncovering previously hidden biological associations, identifying treatment options that work at the outset and reducing the use of drugs that won’t work (and often come with horrific side effects).
The initial elements of the technological infrastructure are now being put into place. The American Society of Clinical Oncology launched CancerLinQ, a health platform that is expected to use data to better manage cancer treatment. As of last June, ASCO’s CancerLinQ had partnered with 58 oncology practices and built its dataset to include 750,000 patient records. Its goal is to learn from the 97 percent of patients who don’t go through clinical trials, and to uncover patterns about tumor characteristics and treatment options that can improve care for cancer patients.
Private efforts are also underway. These include forward-thinking providers such as US Oncology and Intermountain Healthcare, which is also a contributor to CancerLinQ, innovative testing laboratories such as Foundation Medicine, as well as health information companies such as the Google-backed startup Flatiron Health.
Many obstacles remain. One of the biggest is simply getting the data, and massaging it to a standardized format so that it can be parsed and analyzed. Another is the array of barriers to information sharing. Here, cancer patients themselves can play a part by signing on consent forms to share data for research purposes if asked. Under health privacy rules, data can be shared if it is “de-identified,” that is washed clean of certain personally identifying characteristics. Research that relies on health data that may be personally identifiable requires consent.
Finally, there is the role that government funding and centralized support can play in accelerating information sharing. Lives are at stake, which is why we urge Congress and President Trump to continue to drive forward with the big data elements of the cancer moonshot program.
The lesson of cancer treatment and research is that, while cancers get lumped together, no tumor and no patient are ever really average. Only by creating a gigantic data resource to uncover the molecular underpinnings of specific tumors will cancer treatment truly advance beyond today’s haphazard care.