The Limits of Personalized Medicine: You Don't Need 800 Cable Channels (Or Tumor Genetic Mutations)

H. Jack West, MD


December 21, 2018

One tangible milestone of our ongoing progress in personalizing cancer care is that every few months, the US Food and Drug Administration (FDA) approves of an impressively effective novel therapy for a new molecular target. The latest is larotrectinib, particularly notable for its "primary tumor agnostic" approval for any cancer harboring an NTRK fusion.[1]

Combined with the potential benefit of testing for microsatellite instability and the growing array of primary tumor-specific mutations that directly guide clinical management in many cancers, we now have an increasingly clear rationale for broad next-generation sequencing (NGS) that is consistent with the FDA approval of this strategy for patients with any advanced-stage solid tumor.[2]

However, there are data—though limited—identifying a discordance of results across different companies and assays.[3] There is also a lack of prospective randomized trial data demonstrating a clinical benefit from broad NGS testing compared with more limited molecular testing that includes only markers for which approved therapies are available.[4]

There is a potential downside to the 'more is better' approach.

Nevertheless, the world has moved on. NGS is now routinely incorporated into oncology practice. Once we are testing for more markers than you can count on one hand, the argument goes, it's more cost- and tissue-efficient to run an NGS panel. It's the same argument for taking the premium cable package with 800 channels, including 790 you don't really want, when it's cheaper than paying for HBO, Showtime, and the NFL package separately.

But there is a potential downside to the "more is better" approach. Those extra channels make watching television more confusing, and it now takes 20 minutes of futile channel surfing just to come to the same conclusion: There's still nothing on that's worth watching. Or you end up getting sucked in to a show about quilting, broadcast in a language you don't speak, rather than doing something more productive with your time.

Along with the gains achieved by broad molecular testing, it is time to acknowledge that this practice is not without its challenges, and that the benefits are far more likely in some settings than in others. Newly reported survey data on NGS testing[5] revealed that its use among US oncologists is quite variable; 25% refer their patients to another location or provider for NGS testing, and 21% reported that results rarely or never changed their management. Testing was significantly more common among oncologists under the age of 50, those with a faculty appointment, those with genomics training, and those with access to a molecular tumor board. Beyond the question of pursuing NGS testing, 51% of respondents reported that NGS test reports—often dozens of pages long and replete with management recommendations of varying quality—are often or sometimes difficult to interpret.

I suspect that there is a bit more 'black art' in molecular oncology than we are inclined to admit.

Having waded through NGS reports myself for the past several years, I would argue that there is good reason to believe that they create a murky mess rather than a clear picture. As often as not, the reports feature a laundry list of either clinically irrelevant mutations or seemingly inscrutable variants of markers that we may recognize but are unable to translate into a sensible clinical plan. Too often I have seen patients for second opinions or had calls or emails from colleagues about patients who have had well-established, evidence-based treatment recommendations bypassed in favor of dubious recommendations for targeted therapy, based on preclinical data in a mouse model that was cited in an NGS report—in other words, recommendations that should be qualified as having an evidence level just below grade C.

The European Society for Medical Oncology (ESMO) has initiated a project that recognizes the variability in the strength of evidence for various markers and their putatively paired targeted therapies. This information has been formalized into the ESMO Scale for Clinical Actionability of molecular Targets (ESCAT) scale,[6] which articulates specific tiers of applicability to molecular marker findings. These recommendations range from those ready for implementation in routine clinical decisions, to those with preclinical evidence, and even to those with no evidence for actionability.

Along with guidelines such as those from the National Comprehensive Cancer Network and less formalized recommendations from sources such as UpToDate, these sources will likely be increasingly used to help oncologists identify which targets should receive biomarker-driven therapies, as well as the biomarkers that should be considered rabbit holes of dubious therapeutic choices. Currently, many oncologists feel decidedly unconfident about their ability to translate NGS reports into practice, and some are probably misguided in their optimism for its use.

I hope that, in the near future, a collection of molecular tumor board experts can test the concordance of their treatment recommendations based on interpreting a collection of representative NGS reports featuring a typical list of ambiguous markers along with the more obviously interpretable ones. To my knowledge, we have never seen the results of such a needed experiment. I suspect it would be humbling to learn that there is a bit more "black art" in molecular oncology than we are inclined to admit. It would also be invaluable to define the incremental clinical benefit conferred by expert-guided recommendations.

It would certainly be humbling to see poor concordance and/or no significant benefit from this insight. In the more likely event that we see a marked benefit, we will then need to address how to distribute this insight beyond limited centers in the coming years, particularly when the majority of patients in the United States are treated in a community-based setting. As the survey above highlighted, some community-based oncologists are addressing this gulf by referring patients to specialty centers and physicians, but if this expertise becomes a resource we cannot scale, it will potentially create a class of "haves" and "have nots" for genomic support. This isn't stem cell transplant therapy which is applicable for a small minority of patients; "molecular oncology" may well become integrated into the standard practice of oncology.

The future is here—it's just not evenly distributed.

We need tools to help oncologists help themselves, including resources for the growing number of patients who seek their own reports and actively shape their own care. The American Society of Clinical Oncology (ASCO) and some other organizations and institutions are addressing this problem by developing molecular tumor boards where community oncologists may submit cases and learn from the commentary. These will likely be valuable efforts. But if personalized cancer care is to move past a tipping point from limited availability to routine care, the very fact that tumor molecular profiles are seen as a "personalized" recommendation may require that expert-level decision support be available on demand for exponentially larger patient populations. One novel approach to address this issue is to incorporate artificial intelligence in the interpretation of genomic alterations,[7] which I believe is an ideal application of technology to medicine. In the coming years, we should hope and expect to see far more complex algorithms provide the depth and breadth of insight beyond what the currently limited number of qualified experts can possibly provide.

Science fiction writer William Gibson is attributed with the observation, "The future is here—it's just not evenly distributed." We are still defining the benefits of NGS, but it is time for us to acknowledge that for molecular oncology to deliver on its promise, we need to demonstrate that it confers the benefits we expect from it, and that its potential benefits are not limited to a small minority of beneficiaries with the knowledge required to navigate that world.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.