Roche Holding AG (ROG) Earnings Call Transcript & Summary

Welcome from my side. It's Karl speaking. I'm your moderator today. And before I hand over to our host, Alan Hippe, a few remarks from my side. This is the first event on digitization for Roche, but to my best knowledge, it's also the first one for our industry. And as you know, we went virtual with many of our events already a while ago, but when we had the idea to focus on this emerging technology months ago, I have to say there was no COVID-19. But the current crisis around corona even more so emphasize the importance of this theme. We are happy to have with us Alan, our CIO and CFO, the host of the day. He will talk about strategy, IT capabilities and how digitization impacts our entire value chain in Roche. This will be followed by Bryn Roberts. He is approaching the theme from a research point of view. Mark Lee, the Global Head of Personalized Health Care, coming more from the development side. Mike Rivers, approaching the theme from a diagnostic side. He has a very cool video. I only can recommend it to you about digital pathology emphasizing the diagnostic part from that point of view, really, really, really nice. But I mean, here, of course, you need to go into the web in order to see it. That's obvious. And then, we have Steve Guise. He will talk about the interface, which is close to the customers from the company, patients and any other stakeholders. And also, for instance, new recruiters into the company and some hires. So I trust that you had an opportunity to download the presentation from the web. You can ask questions, and I hope you will have questions via the phone or via the web. And if it's via the web, I will read the questions. If it's phone, then we will give you the opportunity directly to dial into the call. And I hope you will enjoy our insights. And with this one, I want to hand over to Alan. Alan, please.

Yes, Karl, thank you so much. Yes, big welcome from my side as well. Really a great opportunity today, really a first-time for Roche and even the industry to talking about digital technology and advanced analytics and what we can make out of it. And I totally agree with you, Karl. I think COVID-19 has given us a platform like perhaps never before where we can really use technology to change things in a very positive way, and especially in healthcare, and I will come to that. But let me first say, I hope that everybody is fine and healthy. It's very important these days. I would like to welcome our presenters as well, and I can assure you, every speaker today is really a distinguished contributor at Roche. And I can even say, well, these are superstars in our company. So you will enjoy that presentation without a doubt. Before I get into the whole topic, we thought we'd give you an update on our COVID-19 activities at Roche when it comes to testing and when it comes to drugs. We would not want to make that a COVID call yet. This is really about digital technologies. But we thought we'd give you an update and hopefully, that will mitigate the questions that will come up referring to that topic. But as said, COVID has set the stage for a couple of changes ahead, which might even be very sustainable. And let me give a couple of ideas, right at the beginning. I think COVID will have an impact on how we deal with partnerships and stakeholder management. So how do we collaborate with external parties? And we see that already, for example, on the commercial side. Lots of HCPs want to connect with us via video or via other technologies. We will have an impact on the people side. Because it's very clear, when you look at the new technologies, the question is, how is the new working place of the future really looking like? And last but not least, I think that's the dominating one. It will transform healthcare. Because what we've seen is, in the last weeks, healthcare systems have shown their vulnerability, their gaps, their failures even. And I think everybody has understood that data and diagnostics make a huge difference. When you look at a disease, when you look really at the spreading of the virus, and I think really data and diagnostics will become more important, which leads to the point that you can shape ecosystems. Let's assume that's around the disease area. I think then really what we can trigger is better outcomes for patients. And if you want to do so, it is certainly great when you have diagnostics, when you have pharma, and when you have clinical insights under one roof. And we have that at Roche. So can we drive better outcomes for patients in the future? Absolutely. But now the quick COVID update on Slide 5. Very, very brief. In the middle of that slide, you see really the Elecsys anti-SARS-CoV-2 assay, which is the antibody assay. And we have launched the antibody test this week. We will produce up to 100 million tests from June on. That means when you put it on an annualized basis, 1.2 billion tests that we can produce, we as Roche alone. So there will be a broad coverage of antibody testing. We have a huge installed base for the e-platforms, roughly 40,000 systems and more in the market. So I think really that gives you the few broad testing will be available. On the left-hand side, you see the molecular testing with the cobas 6000 (sic) [ 6800 ], 8800 platforms. You know we produce roughly 10 million tests for these platforms when it comes to COVID-19 to really detect the virus directly. And the other piece to it, we have also the LightCycler and the MagNA Pure platform, that's another 5 million tests per month, if you like. But here, the installed base is much smaller from cobas 6800, 8800. We have 850 machines in the market, 122 of the 850 in the U.S. and basically nothing in China, which also shows the opportunities ahead. And then on the right-hand side, you see Actemra in severe COVID-19 patients. We are currently recruiting a trial. I think really the end of the recruiting is imminent, we have said beginning of May. And we expect results of that trial in June. The idea is as IL-6 brings the immune system up, and some patients really suffer from that basically the immune system is overreacting. So bring it a little bit down, still being able to kill the virus. But on the other hand, avoiding the ventilator is a huge advantage. So that's the update. And now let me get to the real topic today. And when you go to Slide 6, you see the agenda. I will talk about the industry trends very quickly. Then Roche in Digital, which is basically what we're doing today. And then the Outlook: Building a digital culture across the organization. That's what we want to do in the future and what we're aiming at. So on Slide 7, you see the industry trends in digital. And you see it really broadly and a lot of elements are affected. But one thing is for sure, it's going to change what we can do for patients and what we can do in the entire healthcare ecosystem. And I think there are 2 major digital trends we see in healthcare: One is really consumers are taking their health into their own hands. And I think that's even supported now by the current COVID crisis. Because what you see is, and I said this at the beginning, there is an increasing importance of healthcare data. Now the lesson has understood that healthcare data matters if you want to treat a disease appropriately. Telemedicine will go up. I see a major increase basically everywhere when it comes to telemedicine. And I see an increase in the preparedness to share data. And you see that with the apps used now during the COVID crisis, and that hopefully will apply to other disease areas as well. And another phenomenon that you're going to see is new players have entered healthcare and tech players, so to say. And I don't think that's really, at the very end, competition. I think these are great partners to collaborate with. On Slide 8, you see the increased investment in data. You see on the left-hand side, the cumulative number of clinical decision support companies, which has increased quite a bit. And then on the right-hand side, you see the substantial increase in Big Data-driven investments, and we are one part of this. With that, let's get to the section, Roche in Digital, and directly to Slide 10. And just to set the stage, and I would talk -- like to talk about the ongoing activities in the middle of that page. And you see we have more than 300 key initiatives ongoing. Unfortunately, we can just talk about 17 today. So we've selected 17 use cases out of this 300 -- more than 300 key initiatives. Our digital spend per year is more than CHF 3 billion, which is very significant. I'm aware of that. And the other piece, which is not so known, I think, is that we have already more than 30 digital products in the market. And we will talk about a couple of them. We'll talk about Floodlight, we will talk about some amazing stuff that we're doing. On the left-hand side, you see that we have done also significant M&A investment. One is Foundation Medicine, the world-leading molecular insights business, very much focused on oncology so far, really looking at genomics data. Very exciting. And then we have Flatiron, healthcare biology -- healthcare technology and services company focused on accelerating cancer research and improving patient care. This is basically a company providing real world data, and you will really hear a lot about these 2 terms, and these 2 companies today. On Slide 11, you see really how the Roche IT infrastructure looks like and our capabilities and what we have done during the COVID phase, and I'm really proud of what we have been able to provide during that crisis. And everybody knows who's a little bit familiar with technology, this is nothing which comes from 1 day to another. Over the last couple of years, we have made strategic investments in our infrastructure that enables business continuity during COVID-19, and there is no exception here. When you look at manufacturing, when you look at communication, I think we came up really here with great contributions and we kept Roche up and running. And elements of these investments are shown on the slide, as you can see, such as investment in connectivity, contact center harmonization, the remote collaboration tools, such as the Google Suite that we're using. In fact, this allowed us to move from roughly 40,000 to 110,000 employees working from home, really from 1 day to another literally, and an increase of remote collaboration of roughly 400% with no major business interruption. On Slide 12, you see a lot of data and I don't want to steal Steve Guise's thunder. Steve will talk about that later on, and he is the right guy to talk about it. So he will give you an insight in that data and what it meant for Roche in this period of crisis. With that, I would like to go to the outlook. So what's going to happen in the future, and that's Slide 14. And on 14, I think you see a very remarkable thing because that's the Roche strategy in a very simplified form. In the middle, you see the patient benefit. And our strategy is around 3 pillars: one is Pioneering Products. And you would say, well, that is what you've always done. Yes, we develop medicines, diagnostics and digital products. But you see there is now an edge to it, and that's new because we want to apply appropriate technologies on the basis of a deep understanding of this disease biology, and we are very convinced this is paving the way for the future. The other pillar is Enhancing Access. And what we want to do there is providing tangible value to healthcare systems and enhancing access for our offerings. But to move into this direction, it will be crucial to show health and economic value and this is increasingly based on real world outcome. And we should be able to demonstrate that, so we need the data for it. And last but not least, Shaping the Ecosystems, make it work. And here is the interesting part, as I've said. I think now it's the time where healthcare systems are basically de-frozen and where there is flexibility out there that we can come in and shape the ecosystems with the insights business that we're having, with diagnostics and with pharma. And we want to work together with the stakeholders on solutions to improve the patient outcomes. By doing so, we want to support reducing the complexity and especially the waste in healthcare systems. Improving outcomes, this is what we have on our mind. So what we want to do in Slide 15, and that's what we do already. What digitalization means to Roche. We want to differentiate the core products with digital solutions. We want to leverage digital to enhance stakeholder experience. And certainly, a glimpse of the obvious, we want to optimize the internal value chain through digital. On Slide 16, you have 3 examples, and I want to avoid now to go into every example here deeply. You will hear about augmented products. I think Bryn will talk about this. He will talk about -- you will hear about stakeholder experience. I think that's a little bit on Steve's side. Very exciting and very actual with the recent developments. And then we have the internal value chain. Here, we have an example about blood gas sensors in our diagnostics business. And here, we were able with artificial intelligence to bring the yields up in production by 25%, which is quite a platform. On Slide 17, you see what we're doing in finance. I don't want to stretch that too much, but we're investing into a brand-new ERP system in Roche, which certainly has to be connected to our shared service centers and to everything else that we're doing out there. It will allow us to come up with smart analytics and insight. And at the very end, I think we want to acquire a lot of automation here and want to get to a digital board room. Slide 18 shows you a summary of what we want to do today. We'll have a lot of 17 use cases, along the value chain, which you see on the upper part of the slide. And on the left-hand side, you see what it is all about: meaningful data scale, genomics, real world data, digital biomarkers, digital pathology, decision support and consumer ecosystem. And when you move on even one side, you see really all the use cases that we're going through. And basically, we touch every part of the value chain. One I have touched really because you might miss something on production and supply. And here, I gave you the example with the blood gas sensor production. With that, very excited about the opportunity, how we can shape ecosystem and excited to hand now over to Bryn Roberts, the Global Head of Operations and Informatics in pRED. Bryn, how are you doing?

Thank you. Thanks so much, Alan. A warm welcome from me, and it's a great pleasure to share with you a few examples now, which are going to span transformation of clinical development. And also pointing forward into the prospect of diagnostic, prognostic solutions and potential treatments. So on Slide 22, just starting with the transformation of clinical development. We have a number of tools, digital solutions, deployed across clinical programs in 3 key areas that I've chosen to highlight here. On the left, we have the patient-focused applications around patient experience, engagement. Things like recruitment activities, how we identify patients and reach out to patients who may be appropriate to join our clinical development programs, electronic informed consent and even through to using tools that support medication compliance during the course of programs. In the middle, we think about the investigator, how we support the investigator to optimize the conduct of studies to make their life easier in the way they engage their colleagues and their patients. So investigator meeting app support. And then finally, on the right-hand side, thinking about maybe more technically about novel measurements and endpoints in our clinical programs. And that's where I'm going to spend the bulk of the rest of the discussion. So moving to Slide 23. Just thinking about measurement endpoints and really bringing digital biomarkers into focus here. So this is really the use of digital solutions, typically mobile and sensor technologies, which a patient can carry or have mounted on their person, potentially like gaming platforms as well as we'll touch on. And these are solutions we deploy with the ability to make very sensitive and precise measurements during the course of the study. And obviously, those measurements are objective. They're not relying on rating scales and so forth as has been the tradition. The other advantage is they're continuous and longitudinal, so we can measure 24/7 every day and not just rely on episodes maybe observed in clinical visits. And they're also of low assessment burden for patients. So the experience of a patient using these tools is also enhanced. Because they're using the real world setting day to day, they're also very relevant and/or you could say more ecological measuring the patient in their natural and normal environment and not in an artificial clinical environment. We have digital biomarkers in many indications. A number of those are indicated in the blue here on the right-hand side of the slide. And I'm just going to pick up on a couple of those to exemplify. So in terms of an example of Parkinson's disease, I'm going to look at first the active test. So these are tests where the patient is asked to perform an activity, perhaps on a daily basis. They will hold the phone and perform perhaps a tapping test or drawing a shape, maybe hold the phone to detect tremor or maybe walk with the phone in their pocket so we can look at their walking and turning performance. So that is what we know as active tests. And then on -- moving to Slide 25. The other side of the digital biomarker is really this continuous passive monitoring. So the patient carries the device or the phone with -- during the course of their normal daily activities and from the accelerometry and gyrometry data that we gather from the phone, using deep learning-based models, so AI based, the current neural networks typically, we can actually classify these activities and score them. So we can understand what the patient is doing, if they're climbing stairs, sitting or standing and so forth; and to various degrees, how they're performing those activities. So we can look at things like turning speed if they change direction whilst walking. And this gives us a really good impression of how patients are performing their daily tasks as mentioned. On Slide 26, I just wanted to highlight example of Huntington's disease. This is the full suite of active and passive tests that you see here. And just to highlight, at the bottom in gray there, the process. So once the patient has been taught to use the tools, which we do in teaching fashion in the clinic with the physician. The tests are then performed at home remotely so in a situation that we're in now, with COVID, for example, it would be -- it is of great advantage that we can continue to make assessments of patients without them having to necessarily visit the clinic, so even more important than usual. At this stage, of course, they're transferred securely to us for analysis, and we've talked before about things like privacy by design. So these data are really kept in a very secure ambience and privacy is well respected. On 27, you see this Huntington's app is deployed across all of our clinical development programs for our own molecule. And in addition, a number of natural history studies, which we're partnering with. So it's not just about the specific development of the molecule. It's also about the understanding, insights about the natural history of a disease. And that's something where we've made great progress over the past few years. Understanding actually with this increased precision, sensitivity and objectivity, often what's really going on with some of these symptoms, perhaps at a subclinical level. So moving to Slide 28. Thinking about these digital solutions that we're using in the clinic, in clinical trial settings. They have the opportunity to move downstream to become decision-support tools or augmentation tools actually in the clinical environment. So firstly, around a medicine. So if we have a medicine, such as we do in multiple sclerosis with OCREVUS, can we bring a tool, a digital companion, that will help manage in addition to the drug help the patient and the physician manage their disease more effectively. And that's indeed on Slide 29, what we're doing. So with Floodlight MS, which was originally the biomarker, a platform that was developed alongside OCREVUS in the clinical programs. We're now bringing that forward as a product to launch. So currently, we're working on the first generation. You see the 5 active tests in the suites: pinching tomatoes, drawing shapes, performing a U turn, a 2-minute walk test and then some cognitive testing. And these will be combined with the passive monitoring. So initially, we want to be able to measure function, and we want to generate -- we already have strong clinical evidence from the work we've done, but that will continue to grow over time. Once we have developed sufficient power in this digital platform, we want to measure disease progression in a meaningful way. And eventually, the third-generation we're aiming towards the prediction of disease progression. So looking for markers which will indicate how the patient is likely to progress. And in the first couple of generations of this tool, clearly, what we want to do is ensure that patients are optimally treated for that disease, is brought under control. Jumping to Slide 30, moving away from the digital biomarkers now to other types of tools we developed during the R&D phase, which we now start to bring forward towards the clinic. So predictive algorithms in this first case. So we have a number of algorithms, which we've built, which can predict. In retinal disease, for example, disease progression or perhaps response to certain types of treatment, mechanisms of treatment. In oncology, we also have some very nice predictive algorithm now using a variety of real world data from the Flatiron and Foundation databases that can predict how patients are likely to progress with the disease over the course of time. Both of these tools, we're using actively now are in our own internal programs. And now we start to look at productizing those or creating services that we can bring forward to the real world setting. On Slide 31, we start to think now not only around the medicine, but what about beyond the medicine. So things that really are adjuncts to therapy. So they have some sort of therapeutic benefit in their own right. So digital therapeutics, we would often hear them called. And I just wanted to sort of point forward to these on Slide 32. We have a number of activities ongoing where we're exploring, for example, exergaming as a potential therapeutic modality in conditions like spinal muscular atrophy. You've probably seen we recently announced, actually just in the last few days, announced the partnership with Harman to work on digital tools in logistic spectrum disorders for supporting the daily living activities of people with autism. We know now if we design these tools appropriately, they can be active in both physical and cognitive therapy and skill development. So that's a very active area for us in the future. And we hope to be bringing forward these digital therapeutics in the coming years. And with that, it's my great pleasure to hand over to Mark. Thank you.

No. Thanks, Bryn, and thank you to the participants for joining us today. Just to build on what Bryn just presented in digital, it's clear that healthcare data and the insights we can derive from such data are the fuel for that strategy that Alan laid out: Pioneering Products, Enhancing Access and Shaping Ecosystems. On Slide 35, we highlight a key theme for us in thinking about healthcare data, this concept of meaningful data at scale, that critical determinants of the range and reliability of insights we can derive from healthcare data depend upon data scale, the number of patients represented, the depth of the data, the amount of data per patient and the quality. And we're going to focus on real world data here. And in particular, our partnership with Flatiron Health. Flatiron is a real world data resource focused in oncology derived from the electronic health records of more than 2 million cancer patients treated across the United States. And in addition to the scale of the data, this is rigorously curated, integrated, structured and unstructured data, the depth and quality of which are demonstrated by this increasing sensitivity we observe as you layer in each of those data types. And this combination of scale, depth and quality enable analyses as you see on the right where we can construct cohorts out of the real world data using the attributes of those patients to mimic the control population from randomized trials. And here's an example from a Tecentriq trial, OAK, showing in the red and blue curves, the remarkably similar survival outcomes for these populations and highlighting the great potential of this type of data across our value chain. And indeed, today, we have more than 100 such projects applying real world data from Flatiron and many other partners in R&D access and regulatory. Now if we go to Slide 36, just some specific examples of our use of real world data in our engagements with many of our partners in the healthcare ecosystem to accelerate the availability and access of our medicines for patients. First example here is Rozlytrek in ROS1+ non-small cell lung cancer. Now owing to the rarity of this population, a randomized trial was deemed not feasible. The availability of Flatiron real world data enabled the study of a cohort of individuals with ROS1 receiving the standard of care, crizotinib. And that then enabled this sort of comparison, which supported the early submission of Rozlytrek in this indication to the EMA and the Japanese Health Authority and ultimately was accepted in support of the approval of Rozlytrek in this indication in Japan. On Slide 37, another example of real world data usage. We're now, in this case, with an HTA in Germany, supporting a reimbursement decision for Perjeta in early-stage breast cancer. And an important question in this determination is the cost of -- and the burden of treatment in the first-line metastatic breast cancer setting. And Flatiron real world data was used to demonstrate the extended duration of treatment with Perjeta in the metastatic population, which allowed us to contextualize the benefit of Perjeta in really reducing the incidence of metastatic disease and the cost and burden incurred, therefore. And this enabled a positive reimbursement decision for Perjeta in this setting in April of 2019. On Slide 38, this is an example of real world data supporting a post-marketing commitment. In this case, Kadcyla in the metastatic breast cancer setting. The question here in the post-marketing commitment was around cardiac risk in Kadcyla treated patients in HER2+ metastatic breast cancer patients who had a depressed cardiac function. And Flatiron real world data, in this case, enabled the study of the largest cohort of Kadcyla patients with this attribute. And the observation across all those individual patients shown on the left, the relatively small number of patients experiencing a cardiac event allowed a conclusion that Kadcyla usage in the setting was not an unacceptable cardiac risk for such patients and resulted in an appropriate label update. On Slide 39. Now let's go beyond the currently available real world data and really highlighting that we will anticipate increasing impact of real world data as additional sources become available and the scale and quality continue to improve. But it's the addition of new dimensions of data that will amplify the impact of real world data across all the applications that we described earlier. And in this case, the Foundation Medicine and Flatiron collaborative effort to build a clinical genomic database, really leveraging the overlap of patients receiving the Foundation medicine tumor profiling test in the Flatiron network, really combining carefully curated clinical data, clinical outcome data, with a leading comprehensive tumor genomic profiling dataset allows for a new set of insights to be derived. And these include studying genomically defined populations and their outcomes in the standard of care, but also understanding the underlying genomics of populations with different clinical outcomes and thereby deriving a better understanding of prognosis, but also resistance and response. And those insights can lead to clinical decision support tools, which can contribute to solutions, but also, on Slide 40, applications in research and early development. And some examples called out here, on the left-hand side for Genentech research in early development. Just 2 examples. One, in a program looking at a targeted therapy, the availability of the clinical genomic database allowed a determination of a pan-tumor approach. Similarly for another targeted program, the understanding of the incidence of brain metastases informed the decision to include a brain-penetrant molecule into the broader development strategy for that program. And on the right-hand side, in pharma research and early development, similarly targeted programs taking advantage of an understanding of outcomes in the standard of care allowed for decisions around including such a target at the portfolio, and then, more broadly speaking, an understanding of the molecular mechanisms for checkpoint inhibitor resistance in the standard of care informing a broad array of cancer immunotherapy projects. And so the availability of such data with applications across the value chain, but importantly, in generating scientific hypothesis that are informing our R&D efforts. On Slide 41, this is a transition now from sort of the clinical genomic data that is now available in the standard of care and thinking about constructing different clinical genomic data through a prospective program. And this is a program we refer to as the prospective clinical genomic platform, which really takes advantage of the collaboration that we have, again, with Flatiron and Foundation Medicine. In this case, prospective definition, allowing us to determine the patient population to focus on, but also to introduce new data types that would not be routinely collected in the standard of care. And then this initial program, which we launched just a few months ago, actually focuses on the serial collection of circulating tumor DNA data. And this, combined with tissue genomics, which really capitalizes on the leading genomics capability of Foundation Medicine, is then paired with the real world data capabilities of Flatiron Health. These are patients in the Flatiron network who then have their clinical outcome data abstracted from their health record, collected in due course of standard patient management. And then this also allows the inclusion in the future of additional data types like digital pathology, which you're going to hear about from Mike, but also clinical imaging as well as biobanking. And the availability of scalable platform really enables clinical research to reach more investigators and more patients and generate more real world data of this type in the real world setting and which we believe will be quite transformative for the clinical research. Now moving on to Slide 42. Now I've already alluded to this concept of increasing depth. And certainly, genomics is part of it. Digital health, as Bryn referred to, will also be part of that increase in depth. But medical imaging is a particularly exciting area where these advanced analytics approaches are allowing us to really think of scalable and reproducible ways of adding this information to our real world data resources. And what's shown here is an analysis by Genentech and Roche scientists, applying these advanced analytics approaches to the DICOM images from these scans, in this case, this is PET-CT, and demonstrating in a validation study that the machine-based model performs remarkably well compared to expert radiologists. In fact, outperforming the radiologist to radiologist concordance in many instances. And this certainly opens the door for more reproducible and quantitative way to assess tumor response and progression in our trials, but also excitingly, this opens the possibility of a scalable and reproducible approach to bringing this data into real world data and enhancing its impact and applications across the portfolio. Now on Slide 43, we're going to talk about a complementary effort that we put in place to enhance the advanced analytics capabilities in the organization. The Roche Advanced Analytics Network is a voluntary self-reinforcing community within Roche of our scientists who are working in advanced analytics. And today, this network includes more than 1,300 people at 40 Roche sites. And through harnessing of the scale and diversity of our scientific community and bringing in the network effects, we've enabled sharing of learnings, peer-to-peer education and mentoring, but also engagement with external experts, strategic relationships with key academic institutions and then also sharing of infrastructure and rapid iteration around tools that are developed internally and externally. And also opening up, on Slide 44, an interesting way of leveraging this community through what we call the RAAN data challenges. And this amounts to like a hackathon or you could refer to this as the crowd sourcing of science. And these are really self-assembling challenges where teams of individuals, scientists around Roche participate in answering a question of key interest to the community. And this example is predicting response to Tecentriq in non-small cell lung cancer. And for these challenges, we provide data of unique scale and quality for training. And then there's a competition for performance of the ultimate model on a test set. And you can see the scale of participation in this challenge with more than 500 individuals representing more than 140 teams. And we believe approaches like this, together with our efforts to continuously advance the scale, depth and quality of healthcare data, are really going to open the pathway to ever more impactful insights that will change drug development but also the delivery of personalized patient care. And so with that, it's my pleasure to turn it over to Mike Rivers.

Thank you, Mark. Hello, everyone. It's a pleasure to be with you today. On Slide 46, I'll just reorient you to our agenda for today. We've had an excellent tour so far through early research and pharma development and the digital efforts that are underway in those areas. And I want to take a few minutes to turn our attention now to Diagnostics. In the Diagnostics division, we have a number of significant digitalization projects that are broadly focused in the areas of lab operations effectiveness, digital biomarkers and algorithms and clinical decision support solutions. And today, I want to narrow the focus on digital pathology and talk to you a little bit about the digital pathology portfolio that we've been developing at Roche and the investments that we've been making to digitize tissue diagnostics. So on Slide 47, you see kind of our vision of the pathology lab of the future. And I think it's important to note that today, in most pathology labs, the workflow stops at the second circle in this chain. Tissue samples come into the lab via surgical resection or a biopsy. They're fixed and processed and ultimately transformed into very thin slices that go onto a glass slide. Those slides have been stained with reagents, some with antibodies in the case of immunohistochemistry to look for specific biomarkers of interest. And those slides are then presented to a pathologist for manual review on a microscale. That's the workflow today in most pathology labs. And there are some significant limitations to this workflow. First of all, it's very manual and inefficient. Importantly, it requires the pathologists to be present where the slides are created and produced. So the pathologist needs to be in the lab, needs to be in the hospital where the slides are produced. And particularly today, in the environment of COVID-19, we have seen a surge in interest in remote diagnostics and a lot of questions and energy around how can digital pathology enable pathologists to work more effectively remotely and, frankly, more safely. So a key limitation of the current technology. And then also the approach today is quite manual and requires a subjective review from the pathologist. And we've talked a lot today already about artificial intelligence and the potential that it has to bring insights into the diagnostic process. And I think digital pathology is a wonderful example of this. We see an amazing growth in machine learning and deep learning-based image analysis algorithms that have the potential to help pathologists today, but also in this workflow that you see as we move into more and more complex assays and particularly multiplex assays that are looking at multiple biomarkers simultaneously to really enable clinicians to understand the tumor microenvironment, particularly important for immuno-oncology. Digital pathology is absolutely essential to enable the evaluation of these multiplexed samples. So as we look at this workflow here, and we progress beyond the second circle, we see a fully end-to-end digital workflow. The slides are digitized, in whole slide images with a scanner, as you see there in the third circle. And then moving through the workflow, enabling the digital presentation of these images to a pathologist anywhere at any time and providing AI-based support for the analysis. And then the real power of digitizing this very rich information is taking it out of the silo that it's in now and enabling it to be integrated with tools like our NAVIFY tumor board or other clinical decision support downstream applications that really give value to not only the clinician's work, but ultimately to the patients. So let me dive a little deeper into our digital pathology portfolio on Slide 48. Roche has created an end-to-end portfolio. Starts with slide scanning on the left. About 2 years ago, we launched the DP200 slide scanner. This has been a very successful product launch for us and produces an excellent high-quality image as the foundation, not only for the diagnosis, but also for our image analysis activities. I really want to draw your attention though to the centerpiece of our portfolio and the center of this slide, the pathologist workflow software. This is key for us strategically. It's a key focus for us as well because this is the interface that replaces the microscope for the pathologist. It manages the full digital workflow and ultimately is the environment in which the pathologists will trigger these image analysis applications as well. So we have a host of these applications. I've listed some of them on the right side there for breast cancer. We've recently released 1 for PD-L1 (SP263) lung cancer. Roche has created quite a center of excellence in developing these artificial intelligence-based tools, but our vision is really that this pathologist workflow software is a platform, an ecosystem, within the tissue diagnostics lab that is open. We will accept images from our scanner. But also, as we can see on the bottom left, accept DICOM standardized images, and we're very active in driving that standard for pathology. At the same time, we will offer very high value, robust image analysis applications, but we are creating an open environment that allows us to integrate third-party image analysis applications as well, I think, ultimately bring much more value to this ecosystem. This pathologist workflow software is on market today, it's on-premise solution called uPATH. Very excited to tell you that in the next couple of months, we will be releasing this in a cloud format on the NAVIFY platform and rebranding this as NAVIFY Digital Pathology. So this will bring this in alignment with some of the other products that we have, the digital products that we have and allow a higher level of integration. And rather than talk -- try to talk through the workflow of the software, this is when I'd like to show you a brief video capture of the experience that the pathologists will have with this software. So if you're on the webcast, if you're on the phone, I apologize. This is going to be a little tough to follow. But if you're on the webcast, I'd encourage you to look for that video screen and maximize that window for this presentation, and we'll kick off the video right now. This is an image capture of the pathologists logging on to the software. So if you expand that window on the right side of the screen, you should see this coming up now. Hopefully, you are. And I'm just logging in as the pathologist here. And I can do this from any web browser anywhere in the world. So this, again, shows the power of this approach and this technology. The pathologist then goes -- the first screen that is shown is the case view screen for the pathologist. And then the pathologist goes into the place where they would view the slides. And so hopefully, you're seeing that right now. I'm having a little challenge on my screen. So I'm not seeing this clearly myself. But the slides are showed in order, all of the images associated with the case are present. There are 8 slides in this particular case. And the one that I'd like to call your attention to is the one on the bottom middle of the screen. This is the PD-L1 image. And this is the one that has the image analysis application associated with it. You may see that little blue bar at the bottom of that, that says analysis is successful. Just to tell you very quickly what we do with our whole slide image analysis applications is that we have created a clinical workflow that really enables the pathologist to move quickly and efficiently through the process. So as soon as the slide is scanned, we send the image to an analysis server and the whole -- every cell on that image is classified. The coordinates for that cell are sent to -- those cells are sent to a database and the relevant information about those cells, whether they are tumor or nontumor, artifact, immune cell, whatever is important is recorded in that database and waiting for the pathologist to interact with it. And what you're able to see on the video with the workflow is some of the important capabilities within the workflow software of being able to align and synchronize slides and so forth. And then ultimately, when we come to the PD-L1 slide that we want to analyze, we ask the pathologist to annotate around the tumor area. The pathologist provides that annotation. They really can note everything of relevance on the slide, and we're able to return the results very, very quickly because we've done that precompute step ahead of time. So the actual clinical workflow for the pathologist is very seamless in the software, very fast and very seamless. And I'm struggling to see the video here. So I'm hoping that I'm somewhat synchronized with you. But once that analysis is complete, you see a digital overlay that allows the pathologist to see exactly what the algorithm is calling positive and negative tumor. You see detailed statistics on those tumor stain cells, the nontumor stain cell. And all of that information is captured the report and available to be sent on to the lab information system or to other software. So a very powerful ecosystem that really enables, we think, new value -- brings new value to the diagnostic process. So this PD-L1 lung algorithm was launched or released by us just last month. We're very excited to add this to our portfolio, and we have a whole series of additional algorithms as well that are coming. So let's move on to Slide 49, and I'll wrap up the discussion here. Hopefully, you've had just a glimpse of what the user experience for the pathologist is today. We don't have the time to really go into detail, into all the solutions that we're working on. But I think it's important to note that we really see this globally as a large ecosystem of high-value digital products that ultimately will lend value to the clinical decision. So if you look at the left side of this graphic, we just talked about NAVIFY digital pathology. Also on the NAVIFY platform, we have the Mutation Profiler software. This addresses one of the key challenges of next-generation sequencing and dealing with the vast amounts of data that are there, translating that very complex data set into actionable clinical decisions for the clinicians. All of these data can then become inputs into downstream clinical decision support applications like the NAVIFY Tumor Board, which can integrate all of this diverse information, both diagnostic information, demographic information, and really empower clinicians to make good decisions quickly for the patient. And I think the last thing that I would say, just to emphasize, that we have cast this as a future vision. But these products are here today. They are available now. And I really think now is the time for these products, and now is the time to pull these diagnostic pieces out of the silos, integrate this information and really, really generate fantastic results and experience for the clinicians, but ultimately, better and faster results for patients. And with that, I'd like to turn it over to Steve Guise.

Thanks, Mike. So I've got the pleasure of bringing us home, if you like, into the Q&A. I'd really like to thank Mike, Mark and Bryn for taking us through some really detailed use cases. I'm going to -- if you go to Slide 52, I'm going to talk about our go-to-market model and the digital investments that we're making there to enable a new go-to-start model for the future. And then I'm going to just summarize a few of the capabilities that underpin all of what you've heard about today. And then finally, coming back to some points that Alan made in his opening, I will just describe a little bit the figures you saw behind our COVID-19 response from a technology perspective. On 53, what I wanted to do here was just to summarize a little bit from a pharma perspective where we see digital having its biggest effects over the coming 10 years. And you've heard from Mark and Bryn where we're using data at scale -- meaningful data at scale and digital tools and technologies in terms of digital biomarkers, potential therapeutic solutions. Covering points 1 and 2 on this slide, really industry-leading R&D effectiveness and augmenting our medicines. What we don't have time for today is to delve into point 3. And point 3 talks about how we are building an infrastructure to be able to deliver individualized therapies to patients in the near future. So as we get into personalized cancer vaccine, cellular therapies and the like, we have to build a completely new infrastructure to be able to take samples from patients back to our labs to be able to sequence their tissue, be able to synthesize individual treatments for those patients and shift them back. So I'm sure that in our future events, we'll be able to dive into those topics. But today, I'll spend a bit more time on point 4 here, which is really the go-to-market model. On 54, you'll see a little bit of a From/To perspective. And I would say in our current lockdown coronavirus crisis mode, this has been really brought to the forefront. But what I will say is the trends that you see here are not new trends that have only just emerged in recent weeks and months. This has been a trend over years that we've been building infrastructure for moving from this kind of predominantly in-person static information, kind of field force-driven approach to sales, to moving to often virtual digital channels, much more customized content for different audiences and that will be supported by underlying analytics to make sure that we're delivering the right information to the right audience. So let me step into that on 54 -- 55, rather. So in order to be able to personalize the content that we deliver to different stakeholders, we really need to have globally consistent real-time digital content that contains up-to-date medical information from the latest trials, really understanding which elements of our messages resonate well with different stakeholders. We'll deliver that content to people through their preferred channels. And in some cases, that will still be via face-to-face interactions. But in many cases and in many markets, we see an acceleration towards digital channels. And through this approach, we'll be able to tailor the content that we deliver to people really down to a segment of one. And through the application of technology, we believe not only can we deliver better content to people, but we can also do that at significantly lower costs. On 56, I just want to illustrate how we're using a much more data-driven approach to understand different behaviors in the market, both from our stakeholders that we talk to, also patients and how they use our tools. We're really trying to bring together all of the different elements of the Roche Group between pharma, diagnostics and our insights business to be able to use analytics to explain trends in the past, be able to predict where the market will go in the future and then guide our teams to make the best next step in their approach to the market. And this is fully data-driven. So to realize that ambition, you'll see on 57, we're investing heavily in technology at this point. So we've been running a program called EpiCX, which you see here in the center of the slide, for a couple of years now. And we've been partnering with 2 leading partners in the field, both IQVIA and Veeva, to bring together different digital capabilities to support our medical teams, our sales force, our marketing teams and also those event organizing teams that run conferences around the world. New underlying technologies from IQVIA and Veeva have been adopted because they're industry-leading platforms, and we believe that actually we can differentiate predominantly through our science and through our medicines the technology as an enabler here. But really, what we want to do is make sure that we can bring the message of our science and our medicine to the right audiences. The EpiCX program has already rolled out in a couple of pilot countries, in Australia and in Taiwan. We've also rolled out to a couple of pilot groups within our Genentech U.S. organization. And we will have rolled out this platform across the U.S. by the middle of this year, despite all of the challenges with the lockdown currently. And then over the coming months, we'll deploy in Brazil and then in China, and then follow-up with the rest of the world over the remaining 18 months. So by the time this program is completed, we will have, for the first time, a global ecosystem across pharma connecting together medical, sales, market access across our affiliates, and we will be linking into the diagnostics business as well to get that 1 customer view. So on 58, let me jump now to really just kind of summarizing a little bit the technology that's underpinning all that you've heard today. So from a Roche capabilities perspective, Slide 59 shows that this is a real combination of having the right people, having the right technology underpinnings and clearly, data and information is driving all of this. On the people side, we really like to think of ourselves as the preferred go-to employer for tech talent who really feel that they want to make a difference to the world, make a difference to society. And what we often hear when we reach out into the market is people are looking for having a meaningful contribution. And actually, the purpose of Roche and Genentech of doing now what patients need next is really what attracts them to come and work here. But this is a particularly in-demand talent base, so we've launched a couple of initiatives to really target tech talent in very specific ways. So I invite you to go and spend some time looking at our Code4Life website, which is really targeted at recruiting tech talents by connecting them with the purpose of the company and showing where they can have an impact. And then we're using our Future X Healthcare platform to run kind of challenge events where we put different challenges out to the broader community with academics and students and start-ups and the like to see if they can solve some of healthcare's serious problems with digital approaches. And then this is a great source for talent. On Slide 61, we are -- I just want to walk you through a little bit some of the infrastructure that underpins all of this. So from what you heard from Bryn and then from Mark and then from Mike, you can see we're dealing with enormous volumes of data, whether it's data from sensors being worn by patients in the kind of clinical trial setting that Bryn described, whether it's some of the real world data that we're accessing from our partners in Flatiron and FMI or other sources of real world data that we're pulling in from countries around the world, then we have a huge imaging data sets and huge genomic data sets. What we realize now is that well, 2 years ago, I would say, we saw that the enterprise infrastructure that we built for running the company was not sized for those kinds of volumes of data. So we embarked on building what we call the Roche Science infrastructure, which is a dedicated infrastructure for scientists. It includes high-performance compute and storage, both in house, on-premise, in our data centers. But also bursting out into various cloud providers. So we connect to our Roche Science Cloud to Amazon's AWS platform, Microsoft's Azure platform and also Google's cloud. In the near future, we'll also be connecting to cloud providers in China so that we're able to provide those capabilities to our scientists in China. Through the RSI platform, we're also able to effectively manage the life cycle of huge volumes of data. We produce many petabytes of genomic and imaging data and this seems to be well-managed within the company, but also we need to be able to seamlessly move large volumes of data between our sites, between ourselves and partners. So this infrastructure has been built over the last 2.5 years and now underpins all of the use cases that you heard described by my colleagues. So finally, if we jump to Slide 63, I just want to come back a bit to the current coronavirus situation. And as Alan described in his opening, we were able to seamlessly move a large proportion of our workforce, about 70% of people, overnight to work from home. Now of course, the crisis started a few weeks earlier in China, and we already had to move our Chinese workforce to work from home back in February. And we learned, of course, how well that went during those weeks. In China, we have very specific challenges with access to certain Internet services, which are more restricted in China. So we learned through those few weeks what it would take to do this for the whole company. And then in the middle of March, we basically switched the entire workforce of Roche to work from home overnight. And if you look here on the top right, you can see the remote collaboration. It went through the roof overnight. And we've been relying on Google's G Suite for many years now. We made a switch to the G Suite back in 2013. And this tool and products have really matured in recent years to enable us to really run the company effectively with everybody working from home. We also -- I show on the left the service desk tickets and the on-site requests. You see during the time, we didn't have any real peaks of support needs. Actually, everything worked pretty seamlessly. We had, of course, a drop-off in on-site requests because people were working from home. But our remote service desks that we've set up around the world in our shared service centers didn't experience particularly high volumes. We also consciously moved, over time, many things to the cloud, like the G Suite. But in recent weeks, we moved certain websites and other noncritical tools outside of the Roche VPN protected zone into an access layer, which could be accessed without VPN. We boosted our VPN capability. So I can very confidently say with great pride, we didn't have 1 single incident related to this sudden switch to work from home. I'd just like to leave you on Slide 4 with just a nice story of how our colleagues and our infrastructure helped battle the crisis in Italy. So in Europe, Italy, of course, was the first country to be really affected by COVID-19, and we found the healthcare system to be really overwhelmed. And with all of the infrastructure that we put in place to serve Roche, we were able to deploy that infrastructure to also support the Ministry of Health in Italy. So we used our call center infrastructure and we brought our own employees, about 250 volunteers from across pharma and diagnostics, to actually answer calls on behalf of the Ministry of Health in Italy. And I think in the first few weeks, we picked up about 30% of the call volume there. So this is just 1 area on top of working diagnostics to provide testing and in pharma to run trials with our medicines, but we've also been putting our IT infrastructure and our volunteers to good work in helping fight the crisis in Italy. So finally, let me wrap up with a slide on 65 that you saw with Alan. Our Roche Digital Strategy is summarized very nicely by these 3 buckets. We're using technology to advance our stakeholder experience. We're applying tools to really accelerate the internal value chain, both diagnostics and pharma, and we hope to be able to bring innovation to market at much faster times and at lower cost. And then on the blue side, you have heard from Bryn of all the -- the digital solutions that will surround our medicines and go beyond our medicines into digital therapeutics. So hopefully, between the 4 of us, you've had a nice round picture of the digital strategy at Roche. And now I'd like to pass back to Karl for the Q&A.

Yes, thanks a lot. Very impressive. And as a user, I can really confirm that we have a very, very stable and convenient IT environment. So I also wanted to thank all of you as a kind of a customer. The first question is coming via the web from Simon Baker from Redburn. Alan, this is going to you. We have heard that the cost of IT infrastructure like servers, cloud, capacity has increased during the pandemic due to demand. How has this affected the costs or timelines of your ongoing IT investment?

Can you hear me, Karl? Yes, you can hear me now. Very reasonable question. But the good thing is that we have monitored these costs because certainly, I think we had to transfer people to home and we had to give them an equipment which enables them to work from home. But what I can say is, and there were some additional costs but it was very, very limited. And the great thing has been that we did all the investments pre- the crisis. I think it was not like that we had to do a lot during the crisis. We had a little bit here and there, and that was very limited. And I would say it's not even a double-digit million amount. It's -- I would say it's a smaller 1 million digit amount that we have invested in the last weeks. We had really the staff and the infrastructure available.

Yes. Thanks a lot. Maybe we can take the operator a question from the call via the phone.

The first question from the phone comes from Michael Leuchten with UBS.

So 3 questions, please. Firstly, I was just wondering if you could talk to the interaction with the regulatory authorities when it comes to your new biomarkers and endpoints, as you outlined in your slides. Is that an open channel? Is that a parallel communication that can happen with the regulatory agencies? Or are they not quite living in the digitalized world yet that allows that conversation? My second question is about missing data when we think about digitalization and artificial intelligence in early and maybe mid-stage R&D. So when I think about clinical trials and how to report, it always strikes me that there's an underreporting on trials that have failed. So when machine learning is taken into consideration, is that a problem? Do we miss too much data at this point in time that we'll run into problems when it comes to AI? I would love to hear your thoughts. And then thirdly, Alan, in your opening remarks, you made very interesting comments about how much more efficient your manufacturing now is. I was wondering if you could elaborate on that. I think the number you gave was 20-something percent. And I was just wondering if you could flesh it out because it seems a remarkable achievement.

Maybe, Mark, you can start with the questions here on the interaction with the health authorities, endpoints and the data acceptance.

Yes. I mean, I think in terms of the digital biomarkers, and Bryn may want to chime in as well. I mean I think there's been -- there's certainly a path down the software as a medical device path, where I think the standards for the underlying software, the level of clinical evidence and validation, analytical performance, I mean, these are areas that I think are well understood but evolving, nonetheless. Engagement there, I think, is good. And -- but I mean, with the advent of new technologies, including the application of advanced analytics, I think it is an evolving space. But this is an opportunity, as Alan says, to really help engage in the evolution of the ecosystem. With respect to the advanced analytics and missing data, I think this really speaks to the importance, the critical importance of great data science. An understanding of the missingness of data, the gaps in the data, the sources of bias, I mean, they absolutely can impact the reliability of the conclusions you draw even when you are applying these new methods. In fact, if anything, I would say, it even heightens the need to really understand the underlying science. And so this is an area that we have prioritized. And while you're simultaneously addressing the missingness of data, the quality of data, you need to develop the methods to understand what you can and cannot really conclude out of any given data source.

Maybe I can just build on that, Mark. This is Bryn. So in terms of endpoints in our studies, currently, the digital biomarkers are secondary and exploratory endpoints. But in all cases, as Mark said, we're in discussions with the authorities. So they will be included -- those data will be included in the discussions with authorities and then potential submission packages to follow. It will be a little bit longer before they're finally primary endpoints. And I think that we just have to build more evidence of relevance and concordance. And we have a lot of that evidence already, so they're very powerful, very positive responses from the authorities I have to say. And then on the missingness of data, I think it's, just to build on Mark's point, it's a problem of science -- has been a problem of science rather, underreporting of negative results in the literatures. Something we've had to deal with for many years. And actually, in many ways, the missing patterns in data also tell us an awful lot. So there's actually a whole area of data science around the missingness elements of data, which we do exploit ourselves.

Yes should I take this second question, Karl?

Yes, please. Yes.

Before I do so, let me add to the missing data and AI. And I think that's a great question because I remember -- and Mark, you remember, Bryn, you all remember that. Really, we were there at the beginning, when we worked on the term meaningful data at scale, where we came up then with the point comprehensive data set really around patients. I remember that vividly. And when we had all the questions, what quality of data is necessary, how can we cure data? And I cannot agree more with the point to say it needs really great data science to progress here. This is not simple. Just having a big pot full of data is not giving you any insight. And especially in our field, that's why we narrow it normally down to clinical insights because that implies already that there has to be a certain data quality that you can tap into. And that then gives you really data which are really meaningful at the very end for the patient and for these clinical trials. The manufacturing piece, it was just 1 example I have to say, and it was a rather small one. As said, I think it was really about the blood gas business, which is a small blood gas business and 1 sensor that we produce here. But in all honestly, I can say that was not really a very profitable exercise that we underwent with the center because we had a huge yield at the time. And I think it's basically a quality check that we're doing. So there is a lot of imaging. We do imaging during the time, the center is produced. And then really after, I don't know, 5 steps of production, I think we take an image. And then this image is really analyzed with artificial intelligence. And then we can already say after 5 steps, whether this really this sensor is fulfilling the quality standards or not. In the past, we produced that standard to the end and then tested it. And certainly, then the damage was done and completely done for the sensor. So it's just one thing where imaging helps with quality checks and then combined with artificial intelligence.

Yes. Thanks a lot. Just for your background information, for the presenters and for those people who are listening online, there are about over 600 people online, over 600 people. So that is really a high interest, so thanks for your interest in this call. I wanted to continue with one question from Michael Leacock over the line here. He asks, how is the reaction of pathologists to NAVIFY? Do they trust what they see and do they worry about what they don't see? For example, are all slides presented are only PD-L1 positive slides? So the question is on the capabilities of these algorithms and the openness from the market to kick on board these new technologies.

Yes. This is Mike. I'll take that. Yes, I think that's a great question. And I think what we see is a growing acceptance by pathologists. And really, I think, a welcoming for these technologies. Yes, it's a change. And I think the key for me is that we need to bring value to the workflow. I think some of the challenges in the past with some kind of legacy digital pathology solutions is they just really didn't provide an experience and ultimately value to the pathologists' workflow. I think that's what we were trying to highlight in the demonstration today is really taking advantage of the digital environment to do things that you can't do manually, like compare the H&E slide, whether looking at the morphology through the IHC slide, be able to look at those side-by-side, synchronize them and so forth. And so I think there's a growing comfort level, a growing acceptance. I will say, too, that in our algorithms, today, we are not presenting a black box to the pathologist. We still provide all the slides to the pathologists. The pathologist still ultimately has all of the control. I think we're taking steps down this path. And that's why we produced a nice digital overlay and give them a chance to really compare what the algorithm is calling versus what they might manually call. And ultimately, the decision is still in their hands. So I think our approach is to make it very easy for them to accept this and to move this and to embrace this technology to bring value to their workflow, efficiency to their work and essentially make it an easy decision for them to make. And we're seeing more and more interest and really a pull from the market, I would say, for this.

Thank you. Maybe we can take the next question via the phone. Operator, please?

Next question from the phone comes from the line of Elizabeth Walton with Credit Suisse.

I have 2, both sort of broad questions. So firstly, do you view these technologies you discussed today as an enabler for everybody or a differentiator for Roche? And if it's differentiating, where is Roche doing things differently or where are you significantly ahead of your peers? And then secondly, many of the technologies you've discussed today look like they can help optimize and reduce your R&D spend, how could your technologies reduce your SG&A spend, which we typically see as high in this industry? Or should we really be thinking about these technologies as enabling mostly R&D savings?

So the question is on how we differentiate what we do versus other companies. I have to say I really have difficulties to ask now on the team because we are basically differentiating ourselves from many others. But why Steve, wouldn't you like to take that question?

So maybe I'll start with the second question because I think that was really highlighted in my part of the presentation, which is, yes, we spent a lot of time talking about the impact of digital on R&D. This is really enabling our science. But when it comes to the other parts of the value chain, I mean, the go-to-market model is very much now being enabled through digital approaches. So as we switch to focus more on pulling information via digital channels, this really lower our costs in actually the sales and marketing side. But it will also, with our ERP investments that Alan highlighted in the opening, we are looking to streamline all of our back-end processes with our new SAP rollout and leveraging further the shared service center infrastructure that we've built in Kuala Lumpur, in Budapest and in San Jose, Costa Rica. We'll be moving a lot of back-end processes into lower-cost locations. So I think the combination of technologies will help across the whole value chain.

Thank you. I mean, for the first question, maybe, Alan, you can give some comments from your side. Because you have the best overview of all of us. Maybe Bryn from his perspective from a more research-driven part also could comment on.

I think really, and Bryn can come in here, I think really what I perceive as being outstanding is really our clinical insights business. I think we have really, meanwhile, such an amount of quality data on hand. You see that on the real world data side with Flatiron, but also our own clinical trials and how we can combine that based on great data science that we have in-house. And Mark, please come in here as well. I think here, that's the difference. It's not like that we just have data. On one hand, we have enormous and meaningful data at hand. That's one element. But the other element is that we also know how to deal with that data and how to extract really insights and outcome based on the data. And I think here, we are absolutely leading. But like that we -- one competitor comes to my mind who really can come up with the same magnitude and with the same setup that we have here. Perhaps I'll leave it there. I'm then happy also to comment on SG&A. But Mark and Bryn, both, what's your view on the contribution here where Roche is ahead?

So I think the underlying technologies, things like artificial intelligence and the use of mobile devices, this is available to all companies, of course. And everybody is exploring, to some degree, these. I think where we have a unique value, we are creating unique values in the combination of our data, our specific molecules and indications of interest. So the example of digital biomarkers, I think there's no one really further ahead than us. There's certainly players that are there. And actually, we're collaborating with other pharma companies, also those who we compete with as well. And the key here is that we do want to actually push forward the science for society. It's not just about us as a company. And actually, we all benefit if we set the standards for things like digital biomarkers. If there's a single standard, and it really is the best that we can produce, together with collaborators, then everybody benefits: patients, the healthcare systems as well as the pharma industry.

Yes. Maybe just to build on that. I mean, I think eventually, what we would envision is that care delivery of the future is really an integration of diagnostics, therapeutics, together with data-driven decision support. And ultimately, these -- the technologies for data-driven decision support will need to be community standards. There shouldn't be proprietary tools for every single provider. They really will need to become adopted as standards. But I think the differentiator is to have this incorporation all the way back in R&D is to envision how these pieces fit together from the start. And so, as Bryn says, I mean, to incorporate digital biomarkers, the use of all these data assets and advanced analytics as part of the development of the products. And then ultimately, in the way we engage the ecosystem is something which I think will be an important differentiator.

Yes. Thank you. What -- it's a more fundamental question from Mark Purcell from Morgan Stanley. He was wondering about the benefit of open sources, digital versus closed system? And second question was on the reimbursement of those tier technologies. How can you actually make money to put it in my words with those systems in the future? Maybe. Yes, Mark, you can give it a try, open versus closed. And Bryn, this is also maybe one for you from the research point of view and then the reimbursement situation for that.

Well I mean, I'll start with the reimbursement. I think that's -- those models will still need to be worked out. I mean I think, increasingly, we're seeing an interest in value-based models where we expect these digital tools to deliver more personalized care, more personalized patient management and coordinately more value. And I think that's going to be an important angle. In terms of the open stores, I think this will depend on the specific application. But I do believe as I said earlier, that these platforms will need ultimately to become community standards. And I think how we can approach that could include these kind of open source approaches.

Yes. Let me comment on the point about -- well, how can we make money? I think that's how you phrased it, Karl, and the reimbursement. Well, I think we have business models running. I think we have Flatiron, we have FMI. Foundation medicine is providing based on genomics data and then careful evaluation of that data also from a comprehensive view. And then comes up with the guidance for the patient, for the doctor in case of the appropriate treatment. Dialysis patient, when it comes to oncology. And that's really -- that's a business model which is running. Foundation Medicine has sales and is growing nicely. And then we have Flatiron, where basically, we give access to data. Also to outside companies, I think, it cannot be that we have real world data and nobody else can benefit. But certainly, I think, basically, we sell that data to the outside world because we have quite some efforts in really bringing that data to a quality when it becomes really clinically meaningful. So I think basically here, Flatiron is working together with every meaningful healthcare company around the globe. And that's another business model that we have. And we shouldn't forget that FMI and Flatiron have contributed more than CHF 500 million in sales in 2019 already and have nice growth rates.

I think just a comment on growth. Yes. Sorry about that. Just to comment real quickly on open versus closed systems. I think what we have found, at least in digital pathology in the diagnostics space, is this technology is moving so quickly that I don't think a closed system can be successful. So we are really committed to an open ecosystem. Yes, we produce an end-to-end solution for many reasons. I think that's valuable. But we're also very committed to connecting to value-added third party applications, algorithms, images, et cetera, to make sure that the whole ecosystem is valuable for the end user.

Okay. Maybe we can take a question from the telephone line. Operator?

Next question from the phone comes from Charles Rhyee with Cowen.

This is Cal Sternick on for Charles. I think this question might be more for Mark. You discussed using the EHR data from Flatiron as part of your real world data strategy. As you think longer-term and the use of real world data expands, do you have an interest or maybe a need for broader EHR data sets from EHR vendors like Cerner's HealtheIntent platform, or maybe some of those being developed by Epic and Allscripts?

Yes. The short answer is yes. I think as we look to data that represents patients in other therapeutic areas, also in other geographies. And then on top of that is an interest in understanding individuals at earlier, earlier stages of disease, even extending back into the presymptomatic or sort of prevention setting. Data from broader and broader populations will be important. And so this, I think, goes back to the concept of scale, depth and quality. This is not just about oncology. And so I think you should -- I mean, this is something that we're very keen to pursue.

Yes. Thank you. Maybe we can take another question from the web -- from the phone. Operator, please.

The next question from the phone comes from the line of Keyur Parekh with Goldman Sachs.

One for you, Alan. One more broader question. How do you think about kind of your return on investment for every dollar that you kind of spend on AI and IT. Kind of obviously, there are differing kind of requests on your investments. So just kind of help us understand your investment decision-making process on kind of spending more on R&D versus spending more on kind of digitalization, et cetera? And then kind of a second bigger picture question for the broader team. What do you guys think will be the kind of start of the S-Curve kind of on digitalization in health care? So far, kind of it feels like health care is on a few kind of sectors where AI and digitalization hasn't really had a big impact on the day-to-day. Clearly, the COVID dynamics are changing things. So is this kind of the moment when we get on the S-Curve or do you think there is a technological breakthrough that needs to happen over the next few years that would kind of fast forward that process?

Okay, that's a great question. I can say right away. And I think you're very familiar with pharma and even this [ dia ]. And you know how that goes when it comes to return on investment. And I can assure you one thing here, we have some cases where we can really calculate returns very, very well. And we have some cases where it's tougher to do it. All I can say at this stage is we'd be -- without getting too specific, is that we want to really basically take costs out in the cost of sales. We want to take out costs in G&A, in M&D. And basically, we want to bump up our R&D investments. And that's really where we go. I think we are a company based on innovation. We want to do the right thing for patients, and we want to make a difference for patients. There, we want to invest into R&D. That's really what our thoughts circle around every day. I think without technology and without the approaches that I've described today and that we as a team have described today, I think it wouldn't be possible to do that. It's very clear. I think I gave that manufacturing example. I talked about the P/E environment in finance. You can apply that to procurement. You can apply that as said to manufacturing. I think we have so many examples here which really enable us to bring basically costs down in certain fields outside of R&D and even within R&D. I think you've heard Bryn. You have heard Mark. We have additional opportunities to be more productive and work on the productivity. And that's also -- that's what's requested. And you can even say that's a request from society. Because, well, that -- the industry where we can debate how productive the industry is, but there is a request for better outcomes for patients and I think really with that approach, at least in my opinion, that's the only way we can deliver that. I think to answer your question, I think we have to go more into the specifics, well, this more than CHF 3 billion per year. It's a big number. And it's distributed very well in the company. As said, in more than 300 initiatives that we're having. But you think -- I think all in, the performance that you have seen from Roche, I think already in '19, I think it shows also on the manufacturing side, for example, that we really get the right return from the investment we're making.

Okay. And then on the S-curve utilization in health care? Let's say, what is the next level of return? Maybe Mark, you can give it a try.

I'm not sure I'm the -- I'm going to pass this back to Alan, if that's okay.

I do always the same.

Honestly, I have to say I didn't -- acoustically, I couldn't hear the question very well. Sorry, Karl, can you repeat the question, please?

We have a volunteer. Steve is volunteering.

Let's go. Let me see. Please, go ahead.

I would say all of the technologies to enable the digitalization of healthcare have been there for a while. And I think the things that Mark and Bryn described around how digital tools are used to help R&D, but also in this delivery of care, I think the use of digital tools to monitor diseases in our clinical trials is one step. But with the COVID-19 crisis, we have seen a health care system that was very reluctant to embrace things like telemedicine to suddenly switch overnight where all interactions with health care professionals that can be done virtually are being done virtually. If you see such an acceleration in the last few months, I can only see that, that's going to be a kind of kick start to this S-curve. Because the reality is the technology is there. There are companies with venture capital there, chomping at the bits, waiting to invest in this space. And I do believe that this will be a significant movement in the market.

Thank you. So we may have time for one last question before we have to conclude. I know that some of you do have onward obligations. So maybe we take a last one from the phone please, operator.

The next question from the phone comes from Peter Welford with Jefferies.

I've got 2, but hopefully they're both pretty brief. The first one is just with regards to the point of view of medical device. I'm curious there what your feedback has been so far from regulators with regards to, I guess, the use of AI in some of these devices in the sense that clearly, when they approve it, that the device or the algorithm morphs over time, for want of a better word. Do you think the regulators are yet up to or willing to accept devices with that sort of capability? Or is this going to require a continuous regulatory assessment or refiling perhaps? Secondly, then I want to come back to the age-old question, I guess, of pharma diagnostics within Roche. And I guess my question is, do you think by having pharma and diagnostics both within the company, that provides you a benefit with regards to digital? And I guess, spinning it the other way as well, do you think with the advent of digital, if we go forward 10, 20 years, given some of the things you've outlined, actually, to some extent, there's less benefit having them in-house given everyone can perhaps access a lot of information more easily? And therefore, the benefits of having it available to you internally is somewhat mitigated?

Maybe we can start the first one with Bryn then a last final comment from Alan because pharma diagnostics is definitely on top of your expertise at the moment, so for this overarching team here. So maybe Bryn first.

Thank you. Really fascinating question. So I think there's 2 scenarios. One is that we launched a medical device product, which has a static AI. So the model is not learning actively, and we version that in the usual way. But I think the one you're mentioning is a constant learning system. And there, we have a -- we're active in the program with the FDA on pre-certification. So rather than necessarily having the product certified, it's really the -- it's the process that certifies the quality control measures, the approach to how we control and contain that learning model. So we don't yet have the full answer. I think there's a lot of discussion out there still about the ethics and the -- how much we allow these things to learn over time. But it's a very active discussion we're having with the regulators.

Thank you. Alan?

Yes, let me answer the question about having pharma and dia in Roche and what's the benefit here and what does that mean for digital. And I would like to start with an example here, because I'm sure the question in autumn will be, you have symptoms, you have fever, you feel bad. And you ask yourself, okay, do I have COVID? Do I have the flu? Or do I have normal cold? I'm sure this question will arise. And certainly, it would be great if you had a test available. And potentially, that's the case. I'm relatively predictive here, but I think that's the case. And then I think that let's assume you have a flu, for example. The outcome is you have a flu. Then you can take a drug against the flu. And the point is it's not just the drug against the flu itself. It's really a drug which helps to shut the virus down. That means to avoid that the virus is spreading, let's say, after 24 hours. And that's Xofluza. Xofluza is the drug which can provide that. I think everybody has now understood what you have to do to avoid a pandemic. And I think really Xofluza can help to do that as it shuts down the virus from spreading after 24 hours compared to Tamiflu, where that takes 72 hours. But to get there and to take Xofluza, it takes testing. It really takes testing to know whether you have the flu because when you have something else, a cold or have COVID, I think Xofluza is not helping a lot. So I think that demonstrates where we really, really want to go, where we want to have really diagnostics on one hand, which leads you to the right treatment. You can extend that very easily. I think Mark has talked about great examples today, where it is so important in oncology. I think in oncology, without having sequencing applied, without comprehensive patient data, it is basically impossible to get to the right treatment for the patient. And I think really that's the field which is vacant, which is very meaningful when you look really at the importance that treatments have in that field. And that will also apply to other fields. Yes. You've seen it in multiple sclerosis. You've seen it today in Parkinson's, where you really look how patients really potentially deteriorate over time and where we will also raise the question, what's the right treatment for them here. So I think that you have pharma and dia under 1 roof, that's a great progress. But I think, in my experience in the last years, the missing piece is to have really clinical insights data available, which even helps here, which is surrounding a single diagnostic test, gives you the access to comprehensive data, gives you the access to benchmarking data. And that, at the very end, really drives better patient outcomes. And better, how should I say, targeted treatments for patients at the very end. And this is what we mean by shaping these ecosystems. So I think it is an enormous benefit that we have a diagnostics business, together with our pharma business. But as said, the missing piece has been that we have also this data insights business that we have on hand. And is that driving digital? I think the session today answers that easily. I think without these technologies, you couldn't create the insights that we can create.

That's an excellent summary, Alan. Allow me to thank some of the team members now for the really great contribution they did and for the help on making that call work in here. [ Jared, Mary Frances, Sofia, Barry, Kevin, Ono, Jan, Nadia and Beatrice and Isa ] for the organization of the call, logistics and, of course, the speakers. Thanks a lot for making helping us here in Investor Relations, making this whole event really a success. And I thank you, in the end. We had over 700 people online, which is really, really a large crowd. And thanks for your interest in Roche. Alan, you also wanted to conclude something as host for this call?

Yes, Karl, I want to thank you. I want to thank you, the team and the presenters for an outstanding session. I want to thank all the listeners for being interested in Roche. Thanks for your interest and your commitment.

Thank you all. Bye-bye.