President Obama’s Precision Medicine Initiative (PMI), announced in January and budgeted this year for $215 million, is underway. It will include a voluntary research cohort of 1 million people and is already funding cancer initiatives, including a national “cancer knowledge network.” Eventually a broader knowledge network incorporating molecular data, medical histories with social and physical environment information, and health outcomes will take shape. The goal is a “transformational leap” enabling more accurate diagnoses, more individually targeted disease prevention and treatment strategies, and better health outcomes (Sci Tran Med 2015;7:300ps17).

Yet precision medicine’s transformational aspirations are controversial. “There will be some hits and some niche successes, no question about it,” said Michael Joyner, MD, a professor of anesthesiology and researcher at the Mayo Clinic in Rochester, Minnesota. “But is this going to be broadly transformative? I doubt it.” This critique is based in part on the failure of genomics to fundamentally change clinical outcomes in cancer and heart disease. Precision medicine proponents counter that the limited successes of genomics to date can be vastly multiplied by exploiting the knowledge network, of which genetic information is just one component.

Cancer Successes and Failures

Cancer, the PMI’s initial focus, highlights both the strengths and the pitfalls of the precision medicine approach. Particular genetic defects in tumors are now routinely targeted with very specific drugs. For example, more than two dozen inhibitors of receptor tyrosine kinases, which propagate cell growth signaling pathways, have been approved by the Food and Drug Administration (FDA) during the last 15 years, in a wide variety of tumor types. “Obviously there have been some successes,” said Antonio Fojo, MD, PhD, an oncologist and professor of medicine at Columbia University Medical Center in New York City. “But in the majority of the clinical trials, the benefit has been very small, measured as two months of longer survival.”

In any case, Fojo noted, this isn’t today’s precision medicine concept. It’s data-driven therapy, much like what has been done for decades targeting the estrogen receptor with tamoxifen in women with breast cancer. Precision medicine aspires to improve health outcomes quickly by broadly sharing accumulated individual data, even in the absence of traditional clinical trials. This is already happening in cancer treatment, in which community oncologists now routinely order expanded gene panels or whole-exome sequencing of patient tumors and prescribe targeted therapies off-label upon finding target gene mutations. “We hope that precision medicine in some form works,” added Fojo. “But the current paradigm that looks for a mutation and then gives an FDA-approved agent in an off-label indication, for that we don’t have clinical trial data.”

Some existing data, in fact, suggests that it won’t work. A multicenter, randomized trial in France compared molecularly selected drugs and standard cytotoxic chemotherapy and found no difference in tumor growth (Lancet Oncol 2015;16:1324–34). An earlier pilot study showed a clinical benefit, but used patients as their own controls (J Clin Oncol 2010;28:4877–83). Several “basket trials” that target mutations across tumor types are now underway. One is the National Cancer Institute’s Molecular Analysis for Therapy Choice (MATCH) trial, which aims to match more than 1,000 patients nationally to more than 20 targeted agents. “The NCI-MATCH is an excellent trial,” said Fojo, who nevertheless predicts failure.

Fogo’s bearish outlook on NCI-MATCH lies in the fact that most adult tumors are hard to target effectively. Multiple genes drive tumor growth, defeating single-agent therapies; tumors invariably develop drug resistance; and intratumoral heterogeneity—the existence of multiple clones within a given tumor—allows for escape. And a gene mutation driving a melanoma, for example, doesn’t mean the same mutation in colon cancer should be targeted, as a recent failed trial showed (J Clin Oncol 2015;33:4032–8).

A Systems Approach

Precision medicine proponents acknowledge the obstacles and the complexity, but point to evidence that a very limited number of signaling pathways are driving any given tumor, making it vulnerable to combination treatment. While most targeted combination therapies to date have only added toxicity, Fojo indicated, he sees promise in work at Columbia and elsewhere that uses a systems biology approach, taking into account tumors’ cellular and molecular context and not just gene mutations. “That’s a hypothesis that needs to be tested,” he said.

Early detection of cancer is another goal of precision medicine. “Being able to sequence at orders of magnitude deeper than what is currently being done is going to reveal tumors and pathways much earlier,” said Keith Yamamoto, PhD, vice chancellor of science policy and strategy and vice dean for research at the University of California, San Francisco (UCSF). Yamamoto cites work at startup company Grail to screen asymptomatic patients for circulating tumor DNA in blood for early diagnosis.

Precision medicine incorporates far more than genetic information. “Clearly, genetics is not enough,” said Yamamoto. “There are an untold number of other parameters that are affecting the disease, including complex environmental and experiential factors that we need to keep building into our decision-making framework.”

The Challenge of Heart Disease

Yamamoto considers this multi-tiered orthogonal approach the key to tackling genetically complex common diseases. Genomics has made far less impact with heart disease than with cancer, although there have been isolated successes. For example, many potentially lethal heart arrhythmias known as channelopathies are now detectable by gene sequencing. Three-quarters of all cases of one channelopathy type, long QT syndrome, are caused by mutations in just three genes, and sequencing is now standard when these arrhythmias are suspected.

But this knowledge, which dates back 20 years, has not led to major differences in how these patients are treated, much less a cure. And many people harboring these mutations never develop arrhythmias, making the presence of the mutations on a genetic screening test hard to interpret. In heart disease in general, as more information comes to light, “it’s getting more confusing, not less confusing,” Joyner observed. “Areas where you had clean stories, like the channelopathies, have now become ambiguous.”

Precision medicine would cope with this complexity by layering different kinds of biological, environmental, and behavioral data for individuals and then computationally identifying disease links from the massive data output. However, Joyner does not expect this to work. His skepticism is based in part on a statistical concept, the multiple comparisons problem, which holds that some random associations will appear statistically significant when many variables are considered simultaneously—a massive signal-to-noise problem. He also cites the receiver-operator curve, the likelihood of encountering false positives and false negatives when looking at imperfectly predictive markers. He notes that the known genetic risk factors for complex diseases affect an individual’s risk only slightly, making them useless for discriminating disease from non-disease, even when combined.

There are statistical solutions to the multiple comparisons problem, although the scale and variety of the data are daunting. “There’s no doubt that the challenge of being able to pull together meaningful comparisons of very diverse types of data, and massive amounts of it, is a huge problem,” said Yamamoto. “But it’s a computational problem that the computer scientists and companies that deal in information seem to think can be met.” And, according to National Institutes of Health (NIH) director Francis Collins, gene variants that individually poorly predict risk for common disease nevertheless should collectively point to biological pathways that are causative. “By finding so many of these variants, you can build pathways, molecular networks, that start to make sense about what the disease is all about,” Collins said.

These ideas will be thoroughly tested, not just by Obama’s PMI and NIH, but also by elite institutions like UCSF and Columbia that have made precision medicine a centerpiece of their research agendas. This worries Joyner, he said, in part because it could lead to medical overtreatment, and also because there will be less support for conventional medical research. “There are finite resources, and are you going to make the most progress possible in alleviating human suffering with this approach?” he asked. “Or are there more creative ways to do things? And what are the limits of big science?”

Ken Garber is a freelance writer in Ann Arbor, Michigan. Email: kengarber@prodigy.net.