CHICAGO (GenomeWeb) – Among the 1,300 or so vendors at the sprawling Healthcare Information and Management Systems (HIMSS) conference in Orlando, Florida, last month was first-time exhibitor Roche Diagnostics.
Roche Diagnostics and its parent, pharmaceutical giant Roche, have naturally been involved in various aspects of precision medicine over recent years as drug development and diagnostics have embraced genomics, machine learning, and other advanced technologies. In the meantime, the Swiss company has been busy not only forging partnerships in medical imaging with the likes of GE Healthcare, but making some high-profile acquisitions and product launches in this realm.
It was five years ago that Roche Diagnostics Information Solutions Chief Medical Officer Okan Ekinci first gave a presentation at the annual HIMSS conference, just as the world's largest health IT event began to notice precision medicine.
A few months later, Flatiron Health and Foundation Medicine struck a partnership to create an integrated data platform to develop oncology treatments. Roche bought a majority stake in Foundation Medicine in 2015 and acquired the rest of that firm three years later. In 2018, Roche purchased Flatiron.
In 2017, Roche acquired Viewics, a privately held laboratory business analytics company. That technology focuses on lab efficiency improvements, utilization of devices, and payor reimbursement for such services.
Around the same time, Roche debuted the Navify suite of clinical workflow and decision support software, starting with the Navify Tumor Board management platform to help oncologists sort through data from electronic health records, laboratory, pathology, and imaging information systems, genomics, and administrative datasets.
Those moves established Roche and its Belmont, California-based Diagnostics Information Solutions division as a player in digital health, bridging the worlds of imaging, sequencing, diagnostics, drug discovery, and treatment.
In a presentation at HIMSS, Roche Diagnostics Information Solutions Chief Medical Officer Okan Ekinci said that hospitals have been clamoring for data-driven clinical decision support technology to help improve both patient outcomes and clinician satisfaction. He also said that diagnostics and pharma could now be "treated equally as digital markers" in assessing a patient's condition.
In a subsequent interview with GenomeWeb, Ekinci discussed Roche's strategy in precision medicine. Systems biology and tumor biology in the context of oncology until recently had not been able to "connect the dots between entities like molecular-level, cellular-level, tissue-level, and organ-level, up until you reach the level of a patient," Ekinci explained. "We're at a pivotal time point, where some of the advances that we see on the medical side allow us right now to start building models around how to use the data in a much better fashion."
Below is an edited transcript of the interview.
Why did you decide that this was the year for Roche Diagnostics to exhibit at HIMSS, in a space that you haven't traditionally been in?
It's really about the unique opportunity that we have now to create value at the intersection of medicine and advanced IT. I would say three things that have changed that make it really attractive for us to be there. One is we really do see now this convergence of medicine, science, and healthcare IT, particularly talking about advanced analytics in that realm and the wealth of knowledge and data that waits to be exploited and used. You see this mismatch from the last 10 years that a lot of people talked about big data but didn't really know how to exploit the value.
Second, while we're talking about this growing complexity to make the clinical decisions in emerging areas like precision medicine, there is particularly a need for support tools that are geared towards emerging standards of care. This gearing towards certain emerging standards of care … is something that makes it attractive for us as well.
Third, I think it's the [Roche] portfolio. That spans from immunodiagnostics, sequencing, companion diagnostics, to circulating tumor DNA, all the way to treatment. I think it is right now exactly the time to be there to exploit the potential of data to facilitate these emerging standards of care.
You said in your presentation at HIMSS that you are bringing the diagnostic and pharma sides of Roche together in the context of precision medicine. Is this just for pharmacogenomics, or is it wider?
It will be even wider. These large units within Roche [have started to collaborate] around the patient care continuum. All of the units have an increasing understanding of the disease continuum by using real-world data and, for instance, how existing drugs work — but not only that, but how different treatment mechanisms work, and more importantly to actually get more data out of the more than 95 percent of the time where a patient is not in the hospital. Learning from that real-world data is very important.
You will be relying on wearable data sensors and observations of daily living that you can't get in a clinical setting?
Just thinking about future drug development, if you can learn, for instance, from the real-world data what characteristics may actually lead to a better response rate … you can design clinical studies where real-world data could replace the reference cohort.
With the care decentralization that we have going on, especially in the US, where patients are being treated outside the hospital more often and then also become more frequently survivors of cancer so they need more and more monitoring, we actually need to have much more longitudinal coverage of all the patients' data. Only then I think we can also make incremental improvements in care. So, when we talk about clinical decision support, we need to make sure in the future that we cover as much as possible the disease continuum. We still can influence the patient's outcome if we help them, for instance, to come back for a certain diagnostic test after a certain while.
Precision medicine means so many things to so many people and it's getting redefined pretty much every day. What is the Roche vision for precision medicine?
On the highest level, of course, it's about delivering the right therapy to the right patient at the right time. But that particular sentence meant 10 years ago a totally different thing than it does today and that it will mean five years from now.
Ten or more years in the past, we were selling blockbusters for larger patient populations. Then came this era where we were dealing with companion diagnostics, so we tried to identify subgroups of patients. This was already some sort of personalization of care or precision medicine without being named that. Now there is this third wave [of personalized healthcare], with really targeted therapies and immunotherapy and the increasing use of molecular and tissue diagnostics and profiling where you then can do the decision on the therapy.
For us, it's about increasing personalization of care, and it is a natural continuum of the scientific insights generated. It's not an arbitrary choice to go for further individualization. Common cancers are becoming more like conglomerates of rare cancers today. We see this as a trend, and it's a trend that is unstoppable. With increasing characterization of patients using more diagnostic markers, we start getting to smaller and smaller cohorts of patients. I'm not going to use the term "N-of-1 medicine" here because that's a controversial one, but we do see that the cohorts are getting smaller. We have much smaller patient populations that we're dealing with when it comes to making a therapeutic decision, especially with the genomic alterations that we're talking about.
Our mission is to keep up with the implications that science has at that point and … deliver the value of that further differentiation to every single patient. Science is speeding up a lot, and keeping up in clinical practice becomes really dependent on what kind of tools you use to ... turn insights into clinical practice.
At some point in time, we may come to the conclusion that we have so much data in different data domains that we can actually start modeling disease evolution, for instance. We can start understanding even without doing any further trials, for instance, that a combination of different trials and the knowledge that we have allows us to infer something about an individual patient. If I can characterize the patient to an extent in silico that it resembles the [current status of the] patient, then I can start simulations and see the outcome without doing any further trial. I think at that point, data in silico becomes a "digital twin."
What do you mean by "digital twin"?
A digital twin is a kind of dynamic, virtual, in silico representation of something. In this case, it's the patient, but originally this term comes from industry and material science. When you think about airplanes and the blades of the [engine] turbines, with material science knowledge, you can now model what happens at what point with a certain geometry of a blade. If you turn that turbine to X speed, statistically, something happens in the blade that results in a blade failure and then turbine failure, and then you have a catastrophic impact. This kind of modeling is the foundation of the digital twin thinking. You take sensors and all of the data into your digital model and you continue learning.
Applying it to healthcare, let's take genomic alterations databases. Currently, our knowledge bases are fragmented. Everybody has bits and pieces of those databases. But think about bringing those things continuously together so we have a bigger and bigger and better and fuller picture of [mutations] and outcomes and therapy responses. If we can start getting a fuller picture and then a more longitudinal picture and more variety of data that we pull into the context of this, we can build a model that will help us to do better predictions about the outcome of a certain patient.
In a tumor board when all of the patient data are on the table, this in silico model of a patient could then be utilized to add one piece of information. For instance, if I use drug A, what is the calculated likelihood of [a desired] outcome [for this particular patient]?
Does this mean that N-of-1 medicine is coming?
I think it's about the transition towards allowing in our clinical decision support environment modeling and simulation as an entity that helps become more and more precise in predicting potential outcomes. Then I think we will start feeling more comfortable about the hyperfragmentation that's going on. Otherwise, I think we will just come to a trade-off point where we cannot create more insights with the traditional science we have today. It's about whether we start trusting modeling and simulation in healthcare.
Clinical decision support is a wide field with differing definitions. What is Navify's place in the market?
If you look at the decision support tools out in the market, oftentimes these are rigid tools that have been built for a certain purpose and they are not really dynamic in the sense that they're not further evolving over time.
We said precision oncology, for instance, as an initial use case of this. We said tumor board is the environment where we want to be in because the pain point is so high. Navify is a CDS platform in the first place. It's not just a tumor board solution, [but what] we call a workflow product. We designed Navify as a platform for workflow products that can [support] applications that can be either developed in-house or sourced from outside. That, of course, requires bigger investment and larger long-term thinking, but I think with that we are much better geared than many of our competitors who are rather focused on designing a product around one single use case.
We partnered with an MD Anderson spinoff called MolecularMatch, a company that looked into how to curate clinical trial data and publication data so that they quickly better match with patient characteristics. What we do is create an output from [Navify] Tumor Board that is in near-real time sent to MolecularMatch. They do the search. They do have the database that is most up to date, and then they immediately send us the feedback.
There is this convergence of imaging and molecular diagnostics, and some of that is because, particularly in rare diseases and cancers, clinicians want and need both kinds of test results, right?
There this trend that we have started seeing over the course of the last five years, which is that imaging is actually undergoing a significant transformation towards quantification. I'm not just talking about these standard quantification things in terms of getting parametric maps from images, but it's more like using AI increasingly to "de-mask" and identify more features in the images than a human eye would not be able to identify.
Publications have shown actually that we can both identify disease progression earlier or even identify patient responses to certain therapies earlier. That has fueled our ideas that we want to treat imaging parameters increasingly similarly like diagnostic parameters — in vitro and in vivo coming together. The first step of that is to bring the report into the workflow context, because this is the more urgent need of our customers.
If you look at tumor boards that are conducted in the US today, you have a lot of discussions that are imaging-centric, so it's not that there is just a pathology-centric approach. In hematology-oncology, people would start around the pathology report to discuss the options. But in other areas where surgical treatment is the key decision-making point, there it's more imaging-centric. Based on the very simple fact that the workflow is around that, we said we need to integrate much more strongly on the imaging side, and hence we have a partnership with GE Healthcare to facilitate that.
How are you integrating with vendors of electronic health records and laboratory information management systems?
Our biggest value proposition when it comes to getting significant amounts of data in a structured fashion into the Tumor Board environment is probably the integration that we do with all of the EHR vendors in the markets where we are active in. We do not focus on one single EHR vendor, but rather have Accenture as our global IT partner, who has experience with all of EHRs in the markets where we are active. The integration of [Navify clinical decision support] to the EHR is done by Accenture worldwide.
Why did Roche decide to get into bioinformatics?
We can envision a future where you have a couple of thousand sites that constantly generate structured data in oncology. You could assume that there is an intrinsic value in that aggregated data. Our mid-term vision would be to exploit the value of that data that we then have generated together. Of course, by default, Roche has no access to the data whatsoever. However, in the future, if we find large collaboration partners who will be interested to see what you can do with larger amounts of data aggregated in an anonymous fashion, then the next layer of value exploitation can begin, for instance, looking into things like comparing your own individual patient to a larger cohort of patients.
When we think about the hyperfragmentation, how will you be able to find in N dimensions that you want to compare patient to patient, a patient even within your geography, meaning let's say within the US, you will need to tap into much larger cohorts of patients. One of the parts of the vision is also to see how we can facilitate data standardization across regions.
We also envision … that we can help our customers with data science expertise. There is [a shortage of] data science expertise at the hospital level, and even academic medical centers lack real good data science expertise.