Do We Need to Realign Evidence-Based Versus Precision Medicine?
An interview with geneticist and pathologist Gabriel Bien-Willner, MD, PhD, FCAP, on why we need to change our outlook toward precision medicine in oncology.
To download a PDF of this article,
During this year’s American Society of Clinical Oncology (ASCO) Annual Meeting, I decided to attend a pre-meeting session on a critical timely topic. The half-day session, Genetics and Genomics for the Practicing Clinician, included multiple panelists, each of whom presented real-world cases to form the basis of discussions on both tumor and inherited genetic variations, the molecular testing that is now available to clinicians, and when such tests should be administered.
The entire session was insightful, but what remains with me most was when a gentleman in the audience stood up and posed a question to the panelists concerning precision medicine, during which he made the following provocative statement: “I would argue that evidence-based medicine is incompatible with precision medicine and, as currently practiced, is not effective for cancer care.”
As one whose blog uses the tag line “It’s all about the evidence,” I was immediately intrigued. Following the session, I walked over and asked whether he would be willing to be interviewed for my blog to discuss his thoughts concerning precision-based versus evidence-based medicine in this genomic era—he agreed.
An Introduction
Gabriel Bien-Willner, MD, PhD, FCAP, is board certified in Anatomic Pathology and Molecular Genetic Pathology. Gabe, a classically trained human geneticist and molecular pathologist with deep expertise in next-generation sequencing (NGS), has a long history of providing knowledgeable, critical insight into the molecular basis of disease in cancer patients.
Currently the executive director for medical affairs at Molecular Health, Gabe began his role with the company as medical director, when he was responsible for the daily clinical activities of Molecular Health’s CLIA-certified NGS laboratory following the launch of the company’s cancer genome panel.
Today, the company is focused on developing innovative software solutions for precision medicine for both clinicians and laboratories. Molecular Health has centered its products around the “Dataome” (MH Dataome) technology platform, and its current products are based on analysis of data from a tumor’s genetic composition in addition to numerous types of biomedical data, enabling physicians to create a report including qualified and individualized treatment recommendations for the patient and allowing oncologists to make more fully informed treatment and medication decisions.
As we began our interview, Gabe stressed that he was speaking as an individual who was expressing his personal opinions and views and was not representing those of Molecular Health in any way.
Challenges for Oncologists in the Genomic Era
We began our discussion by reflecting on how unprepared many clinicians feel when it comes to explaining genomic test results to their patients. “We’re in a kind of precarious position,” Gabe explained, “where they are given a very complex genetic test result, and now, the way our medical system is today, they’re expected to be able to explain the results to the patient, but they themselves don’t understand it.” He shared an example from last year’s ASCO Annual Meeting: a breast cancer oncologist was the moderator for a precision-based medicine biomarker session, and she opened the session by saying, “I’m a breast oncologist; I don’t really understand any of this. But, here we go.” Gabe’s reaction? “I just thought ‘Wow! Okay, this is where we are today!’ Everyone knows the value of this, but how could we possibly be using these data correctly? There’s this missing link between the information and making use of that information: the interpretation of that data is left to people who don’t really understand it.”
To address this serious disconnect, Gabe would like to see practice changed, where there is a “learned intermediary.” As he explained, “Just like the doctor is between the treatment and you, I think there needs to be this other, new type of molecular physician who should stand between the genetic data and the oncologist, so that what they receive is something that they are capable of digesting. Computational tools to simplify the workflow and knowledge gaps will be a big part of this process, but I personally feel today that you cannot expect all oncologists to now be experts in genomics. I think that if that is the expectation, we’re setting ourselves up to fail with this endeavor into precision medicine.”
Making a Career Choice
Gabe had always been deeply interested in genetics: “I found it fascinating, the ‘building blocks of our lives’ and, early on, I was particularly fascinated with the prospect of gene therapy—where at the time, in the mid-90s, it was thought that that was going to cure every disease. Of course, it didn’t turn out that way.” As he was considering medical schools, he decided to attend Baylor College of Medicine in Houston, Texas, because he believed that they had one of the best genetics departments in the world. Gabe completed both his medical training and his PhD in genetics and genomics at Baylor.
His initial work in genomics was with his mentor, James R. Lupski, MD, PhD, a member of the National Academy of Sciences, when they were presented with a patient who had been diagnosed with
As Gabe explained, “It turned out that this patient had a translocation, meaning that the chromosomes are broken and then rearranged with other chromosomes about a million base pairs away from the Sox9 gene. So, I hypothesized that there were probably several enhancers or DNA elements that allow the proper activation of this transcription factor, Sox9, whose expression had to be maintained at a very steady level to function properly. And so what was happening was that there were DNA elements upstream and downstream of this translocation breakpoint, and you were losing some of these elements—but not all of them, because the phenotype of this patient was not as severe as with others with mutations in this gene.”
Using novel computational methods and mouse models, Gabe was able to obtain evidence of rearrangements in these DNA elements—ie, enhancers—nearby. “One thing that concerned me was that I spent 4 years proving that there was an enhancer that was actually being activated by Hedgehog signaling, which was driving the expression of Sox9,” Gabe noted. [The hedgehog family of signaling molecules play a critical role in transmitting development signals.] “It took me 4 years to prove that there was 1 enhancer, but we speculated that there must be dozens of such regulators.” He believed that there had to be a way to determine where such enhancers were without the a priori knowledge he had in this case, ie, critical phenotypes and mouse models that gave him this evidence. “There had to be a way to figure out where these elements were. So that was my first foray into genomics,” Gabe explained.
“At the time, there was a very new technology that was just published by a group at the Sanger Institute that I thought could help me find these other elements.” This molecular cytogenetics team, led by Nigel Carter, DPhil, investigated methods to detect changes in the numbers of chromosomes and genes to learn more concerning the causes of particular inherited diseases in humans. Gabe explained that the new technology, called Chip on Chip, relied on using antibodies to capture proteins bound to the DNA elements of interest, which were then frozen. The “captured” DNA would then be hybridized against a microarray chip that included genomic regions of interest, in this case, the regions in Chromosome 17q. “When hybridized, you would see in the analysis where the DNA elements were located, therefore identifying all putative elements in a single experiment without bias.” However, he noted, the technology ultimately did not pan out for multiple reasons: “It wasn’t very scalable or reproducible, it didn’t really work very well at the time, but that was really the beginning of genomics and with DNA for me. And it was really essentially our first approach, not unlike microarray technology, to capture a lot of DNA or RNA information all at once, rather than in a very focused experiment like we’d always done in the past.”
He realized at the end of his PhD that folks who are geneticists tend to go into pediatrics to study the same kind of rare diseases that he was currently studying or perhaps internal medicine, but he found that he did not want to focus on very rare disease. Rather, he wanted to focus on common disease “and make an immediate impact in this space. My goal was to bring genetics and genomics into the clinical space. It always has been, and it still is.” He felt that the best way to accomplish this was in the cancer field through pathology. “There are 2 ways [to enter the field of cancer research]: one is through internal medicine and oncology, and the other is through pathology. Both pathologists and oncologists are experts in cancer, but they are experts in different ways. Oncologists are experts in the treatment of cancer, and pathologists are experts in the diagnosis of cancer, and they’re both experts in the biology of cancer.”
So that is the path that Gabe chose. When he became a resident, he decided to go to Washington University in St. Louis, known as one of the best training programs in this area. And right around that time, next-generation sequencing (NextGen Sequencing or NGS) was developed. Also known as high-throughput sequencing, NGS enables researchers to sequence DNA and RNA much more quickly and inexpensively than Sanger sequencing, representing a paradigm shift that revolutionized the study of molecular biology and genomics. Gabe realized that this NextGen sequencing was a new method that would be transformative. “Even though the chemistry is almost essentially identical to Sanger sequencing,” he explained, “with NGS we’re turning a 2-dimensional process into a 3-dimensional one, where we can perform massively parallel sequencing and capture sufficient data, allowing us to draw conclusions independent of explicit a priori knowledge.” In addition to greatly reducing cost, it has dramatically increased the throughput of genomic sequencing, enabling simultaneous screening of thousands of genetic locations (loci) for disease-causing mutations. “Now we can sequence everything all at once if we want,” Gabe noted, “and figure out what you sequenced later.”
Gabe subsequently decided to do a postdoc to learn NextGen sequencing in the lab of Robi Mitra, PhD, a new faculty member at Washington University’s Center for Genomic Sciences. While in the lab, Gabe learned computational methods and coding as well as how to use Unix (a multi-user computer operating system). As he emphasized, “To make sense of these data, you really had to create the software yourself, honestly: it really did not exist. If you needed to make some sort of analysis of NextGen sequencing results, you needed your own program to achieve this.”
Gabe was influenced by the progress that was being made at Washington University, “including the launch of the Genomics and Pathology Services, where I believe we were the first academic institution to do a hybrid capture, a large gene panel with NextGen sequencing to capture this kind of information from patients. So, I got to be there for the development of that.”
Ultimately, he joined the faculty at Washington University, but did not stay there for long. Having multiple opportunities, he decided to enter industry and accepted a position as medical director for Molecular Health, which as noted previously, specializes in the development of analytic software and informatics approaches that are necessary to make sense of these complex data sets captured with NextGen sequencing.
A Catch-22 Scenario
Speaking about Molecular Health’s overall mission and goals, Gabe said that “The mission of Molecular Health is to create software applications to allow physicians to make sense of complex clinical data, and right now, we’re focused on cancer and genomic data. But, ultimately, the position of the company is to go well beyond both cancer and genetic data.” Gabe noted that when he was first brought on board, “We were not only focused on the development of software, but we decided to showcase that software by starting a commercial laboratory and medical service, which I had been running. The company has refocused solely on software and software development. The service we were offering was very comprehensive and included a medical review by experts. So you were not just getting a test result, you were receiving an interpretation from a real expert into what these complex results mean.”
I told Gabe that I had been particularly interested in the fact that Molecular Health was offering such specific medical expertise and just how valuable a service this was, since, as he’d stated at the beginning of our conversation, there is simply no way that we can expect all oncologists to become experts in how to communicate genomic and genetic information and how to interpret it. “Yes,” he said, “so it’s a Catch-22 scenario, because there are not enough people like me and other molecular pathologists who are really well-grounded in NextGen sequencing as well as in genetics and genomic principles. You can try to make them accessible to as many people as possible—and I think that, in the future, it will be an entirely new subspecialty of medicine. But today it’s difficult to have those people available to everyone.”
Gabe continued: “But one way that you can make them accessible to everyone is with software that enables you to better understand and interpret the results, that can make people who are not quite experts good enough to understand the information that’s coming out of the system. So, I would say that that’s the direction of the company, and it’s something that I’m helping the company do. I’m not sure whether in the future, even in the long-term future, that’s sufficient, but it’s certainly the biggest dent we can make with this real problem. Yet I do think that there is this future of genomic medicine, that there are going to be people with these skill sets who are more widely available, and I think that they have a critical role to play.”
I responded by emphasizing that I appreciated the development of software to enable the delivery of reports that make clinical sense to the ordering physicians, assisting them in their decision making. So lacking an actual clinical consult with a molecular pathologist, they would still have the report explaining the genetic variants identified, their significance, as well as clinical recommendations, and perhaps clinical trials that would be appropriate for these patients.
He agreed, noting that providing such data interpretation is a large part of what they offer and that his focus with the company is “Also, creating a clinically verified or validated knowledge set—that is, a knowledge database of what variants mean in different disease types, so that there can be an automated interpretation for the high-yield, commonly seen or more well-characterized variants, where experts would tend to agree on the significance. That may be helpful for most patients. There’s always potentially going to be cases where you need a little bit more insight, but it’s a great first step into this field to make it more accessible and understandable to people.”
VUS Results: Challenges for Clinical Management
Variants of unknown significance (VUS) are results where DNA alterations are detected, but there is not currently sufficient data to classify whether it is neutral or deleterious. For example,
During the ASCO pre-meeting, Gabe had said that it was extremely irresponsible not to share potential knowledge about the existence of these variants or what some of these variants can mean—a perspective that truly resonated with me. He had given the example of an epidermal growth factor receptor (EGFR) point mutation of unknown significance. In the United States, approximately 15% of patients with non-small cell lung cancer have mutations in EGFR. Certain EGFR mutations have been identified that may predict a positive response to particular agents, known as tyrosine kinase inhibitors, that target EGFR. Gabe noted that if an EGFR VUS were identified that was likely damaging and in a known regulatory domain of the protein, you could make a reasonable clinical judgment about the mutation’s potential significance. I asked whether he could expand on this critical perspective.
“The standard right now is that we’re looking at NextGen sequencing and precision medicine through the prism of non-complex clinical laboratory testing,” Gabe noted. “We want our laboratory testing to be precise, right? With most of this testing, there’s 1 value that we care about where we want to know with a high degree of reliability and precision that we have the right answer.”
“For example,” Gabe continued, “if you’re having a blood test for hemoglobin, you want to know whatever that number result is. As a physician, you know how to interpret that, and you want to know from a laboratory perspective that that is the accurate reading. We know that there is a normal range of distribution of signal, below which and above which is abnormal and within that range is considered a normal range. That’s a typical laboratory test. The problem is that with NextGen sequencing, it’s not one test.”
“Rather,” Gabe said, “you are testing for every nucleotide position, for everything you’re sequencing, which could potentially be mutated in a number of different ways. So, in reality, the number of variables in that test, depending on the size of the panel that you’re testing and the kind of testing that you’re doing, can seem to approach the infinite. And when you’re looking at all that data, it’s not a test result like other chemistry tests, it’s not a yes or no or normal or abnormal, it’s a very complex relationship of multiple variables all at once—and up to thousands at a time if you’re doing an exome sequencing capture. So you’re really practicing medicine by interpreting the complexity of the data to summarize what that data means.”
In other words, as he explained, “It may be similar to a primary care physician’s interaction with a new patient. A patient encounter can be broken down to a series of variables, and each variable seen as a ‘test’: the way they look, their chief complaint category and description, every component of the physical exam, the lab test that was ordered, everything you do with that patient. But the reality is that the doctor does not see it that way. The doctor looks at the patient as a whole, he thinks about what’s best for the patient in light of all the evidence presented before him—and that’s really how I see the interpretation of these complex data. You also have to consider the patient, their history, and their family history sometimes while reviewing the sequencing data. You need to consider all the genes that are sequenced, what the variants are, what the disease is—and the fact that the same variables in different diseases do, in fact, give you different answers. So looking at this exercise as a lab test is overly simplifying a complex process. Practicing medicine by interpreting results is how you’re going to get the most out of this and the most out of precision medicine.”
The sophistication of the analysis also raises the specter of confronting the challenge of VUS. Gabe continued, “From a traditional laboratory test perspective, if you see a variant that you’ve never seen before, you don’t know what it means because there’s no evidence that it means anything, so it’s reasonable to ignore it. That’s not the right approach. Instead, the correct approach is ‘What is this patient’s disease? Let’s look at this with suspicion. What is the patient’s age? What is the patient’s sex?’ A BRCA variant for breast cancer in a 36-year-old means something very different than it could in an 87-year-old with prostate cancer. And then you can go into more details.”
Gabe then posed the following questions:
- What is this gene’s function? Is it a tumor suppressor? Is it an oncogene?
- Even if it’s never been seen before, is this variant likely damaging protein function? If so, why?
- What domain of the protein does this mutation occupy? Is it in an inactivating domain? Is it in an activating domain?
He then emphasized the following: “You have to consider all of these. And even if the gene itself has not been reported to have multiple mutations, if it is a regulator of another gene that is known to have critical importance, how is that going to affect this patient?”
Internal server error