Improving Research and Discovery

Data analytics has perhaps the biggest potential to enable a new drug discovery and development paradigm, though currently the return on investment at this step of the value chain is hardest to measure. The advent of new genomic technologies has resulted in significant reductions in the cost of genome sequencing. According to the National Human Genome Research Institute, which collects data from the genome sequencing groups it funds, in mid-2015, it cost around $4,000 to create a high-quality draft whole human genome sequence. By late 2015, the price tag was dramatically lower—just $1,500; and if only the protein coding regions were desired—what scientists call the exome—the price tag was lower still, under $1,000 per genome [25]. As sequencing technology at Illumina and other biotechs develops, costs are anticipated to drop further, potentially under $100. As sequencing costs fall, biopharmas are able to rethink how they identify and validate drug targets, leveraging big sets of genetic data to find interesting but rare biological signals that could be the basis for novel medicines.

This approach grows even more powerful if biopharma companies are able to link unstructured phenotypic data, for instance, from patient electronic health records, to genetic data. Combining the genetic data with unstructured historical data via powerful analytics, biopharma companies can uncover not just previously hidden biomarkers, but relate their molecular research to outcomes that previously were not recognized.

Tapping that opportunity is one of the reasons Regeneron formed its Regeneron Genetics Center (RGC) and partnered with Geisinger Health System to sequence the genomes of hundreds of thousands of Geisinger patients [26]. In March 2016, the partners published their first peer-reviewed paper, using sequence data linked to de-identified longitudinal health records to identify a specific genetic mutation that results in a significantly reduced risk of coronary artery disease [27]. The findings corroborate ongoing R & D efforts at Regeneron from animal studies, as well as biochemical and early human genetic experiments, about the novel target's importance in improving clinical outcomes. “This new evidence gives us a lot of confidence about our development strategy and the ability to identify game-changing opportunities,” says Aris Baras, M.D., Head of the RGC [28].

In addition to Regeneron, start-ups such as Berg Health, TwoXAR and Sema4 are on their own, and in partnership with academic groups, bigger biopharmas, and infotechs, using cognitive computing and other big data tools to develop novel drugs for complex disease states [29–31].

Cancer has been one of the areas where these efforts to transform biopharma discovery have been most obvious; that is partly because the same trends driving convergence and precision medicine, topics outlined in Chapters 1 and 4, respectively, reinforce the demand for sophisticated analytics. As outlined in Chapter 1, IBM, via its Watson Health division, is making a clear bid to use cognitive computing to identify and personalize oncology therapies [32].

Going forward, this intersection of genetics and what has historically been health IT will merge and further drive the acceleration of efficient drug R & D. Critically, the failure rate for drugs in late-stage trials is still too high in a world of constrained resources. According to a team of researchers at Sagient Research Systems and the Biotechnology Innovation Organization (BIO), around 40 percent of all drugs fail in phase III [33]. Since the cost of R & D increases sharply from one phase to the next, such late-stage failures represent a very inefficient use of R & D capital. Because of historical pricing flexibility, biopharmas have not had the pressure to focus on R & D efficiency, but that situation is changing rapidly. In the future, biopharma companies will need to better use enabling tools such as artificial intelligence and causal machine learning to improve this portion of the value chain [34].

Improving Clinical Trials

There is no question the current clinical trials paradigm, defined as sequential phase I, II, and III studies punctuated by months of analyses and planning, is outdated. In sharp contrast to the technology and consumer sectors, where big data collected in real time routinely informs product formation, the basic framework for drug development has not changed since the 1960s [35]. Because of the high cost of late-stage drug failures, however, biopharmas are beginning to embrace analytics-driven clinical trial designs to increase the efficiency and success rates of their development programs. These so-called adaptive designs rely on Bayesian algorithms to enable preplanned adjustments to clinical trials to refine hypotheses and reallocate R & D dollars in real time based on clinical data [36].

In addition, as described in Chapter 9, the marriage of clinical data and social media (e.g., Twitter, Facebook, and YouTube, and especially online patient communities) creates a new opportunity to develop clinical descriptions of disease that can inform and accelerate clinical trial enrollment, while –omic data (e.g., genomic, proteomic, microbiomic, etc.) enable the identification of patients likely to respond. Similarly, the incorporation of digital technologies into trial designs allows the detection of early safety and efficacy signals. Indeed, data streams from wearables and sensors allow real-time feedback, creating an opportunity for continuous learning in clinical trials via analytics.

Biopharma companies are only now beginning to identify how to leverage analytical tools such as machine learning in this regard. Many of the first efforts are clustered in neurological disease areas where measuring product efficacy and disease progression have historically depended on subjective survey data. Teva Pharmaceutical Industries, for instance, teamed up with Intel in September 2016 to incorporate data from wearable devices in a phase II trial monitoring disease progression in Huntington's disease to generate objective scores measuring the severity of motor symptoms [37].

Creating the Agile Supply Chain

As cost pressures mount and competition from both start-ups and new entrants increases, larger biopharmas are increasingly focused on optimizing their operational performance, especially the activities tied to manufacturing and supply chain. Advanced analytics has a key role to play here. Beyond improving manufacturing efficiencies, analytics can help companies do a better job of creating real-time forecasts, reducing the need to carry costly excess inventory, as well as creating new drug distribution platforms.

A 2014 study shows that while most biopharmas amass significant amounts of historical manufacturing and supply-chain related data, the collected statistics are siloed across different parts of the organization and the available tools are too limiting to handle biopharma's complex manufacturing processes [38]. As streamlining business processes becomes more important in today's tougher climate, however, biopharma companies will look outside the industry to lessons learned in the retail and consumer technology sectors. Analytics-enabled flexible supply chains like the kind Apple employs to create its iPhones will become the norm. Steps that can optimize the hand-offs between the wholesaler, the retail pharmacy, and the patient encourage care coordination and adherence, which can ultimately promote a drug's value

To that end, GlaxoSmithKline in 2014 teamed up with Cambridge University's Institute for Manufacturing to build a more efficient end-to-end supply chain that brings together equipment manufacturers, regulators, knowledge transfer networks, and healthcare providers [39]. Three years earlier, the biopharmaceutical company also partnered with what, at the time, was a non-obvious player: the Formula One race car manufacturer McLaren. On the racetrack, every second counts, so pit crews use sophisticated data analytics to reduce the time required for tire changes and other repairs. GlaxoSmithKline wanted to adapt such methodology to its own processes, particularly in its consumer health business, where small-batch manufacturing leads to not just lost productivity but lower margins. After working with McLaren for about a year, GlaxoSmithKline reduced downtime in one particular manufacturing plant by 60 percent. Since 2011, the two organizations have worked to scale-up changes and have expanded their efforts to make R & D more efficient via the use of sophisticated sensors in clinical trials [40].

Using Analytics to Rethink Commercial Activities

The shift from fee for service to fee for value and the growth of digital technologies have made what was once one of the most straightforward parts of the biopharma value chain—sales and marketing and life cycle management—one of the most complicated. Indeed, estimates suggest that approximately two-thirds of new drugs fail to meet prelaunch sales expectations in their first year on the market and continue to underperform for the next two years [41].

As noted in Chapters 7 and 8, one reason so many recent drug launches have been suboptimal is the need to have, at launch, data showing real-world utility. Absent such information, increasingly skeptical payers and providers make medicines harder to obtain, either by putting them on expensive formulary tiers or by excluding them altogether [42]. (See the box, “Real-World Data—and Analytics—Play an Increasingly Important Role in Establishing Product Value.”)

Beyond real-world evidence analysis, other commercial activities that will alter as a result of data analytics include strategic payer engagement, a topic discussed in Chapter 8, and the channel and geographic optimization of a biopharma's sales force via multichannel digital tools. Such analytics capabilities better enable drugmakers to interact with the right physicians, key opinion leaders, and health systems based on a therapy's clinical profile. As it becomes increasingly difficult to engage with already busy physicians, the ability to customize the message to the right prescribers is essential. On average, high-prescribing physicians are contacted by drugmakers nearly 3,000 times a year. But many of those interactions are overly general and come as physicians are struggling to deliver higher quality care in shorter appointment times [46].

While there is a major opportunity to make the experience for the provider more tailored and personal, the potential to optimize the patient experience could be even greater. Luca Foschini, Chief Data Scientist at Evidation Health, which develops predictive analytics tools for a variety of healthcare stakeholders, believes the same arc that has enabled customer-centricity in the online advertising and retail spaces will also fundamentally change the patient experience. “Personalization was completely absent 10 years ago and poorly understood five years ago,” he says. “Now it is making a huge difference” [47]. Indeed, as organizations such as Netflix and Amazon have grown more data-savvy, they have embraced predictive analytics to mine a user's viewing or reading habits to make recommendations on future movies or books, further customizing the experience and changing consumer expectations [48].

As biopharmas develop their analytics capabilities, one can envision them taking similar strides to customize the patient experience either directly, where regulations permit, or through collaboration with providers. They might use wearable data to provide tailored interactions to patients based on their health status, for instance, to promote medication adherence or healthy behaviors such as better sleep hygiene. Data analytics could also be a valuable tool for connecting more directly with patients on aspects of disease education. Mining metadata from social media (e.g., Twitter, Facebook, and YouTube), biopharmas could begin to develop clinical descriptions of disease that are consistent with how patients themselves discuss their symptoms. This terminology can then be used to develop more meaningful education-based and patient-based portals.

Biopharmas interested in these new approaches must tread carefully, however. Guidance on how biopharmas can interact with patients safely and legally are still evolving, and there is clear evidence that patients want brand-agnostic information, not materials linked to a specific therapy. Indeed, success will depend partly on establishing a new social contract that makes patients more willing to share their personal data with non-provider stakeholders in the health space [49].

Future Uses

As drug companies fully integrate external and internal data to answer stakeholders' needs and gather product evidence about value, they have an opportunity to create new types of data-rich partnerships that are less transactional (selling pills) and more relational (engaging around health outcomes). (See the box, “The Dynamic Data Market.”) These include new risk-sharing models, described in Chapters 7 and 8, as well as beyond-the-pill service models where the intellectual property is not a therapeutic but an algorithm.

Consider that biopharmas can partner with payers to combine the latter's claims and authorization data with its proprietary clinical trial data and develop algorithms to forecast the monetary impact associated with using a novel but relatively untested product versus the potential benefits. The biopharma could devise a pricing contract that preserves pricing flexibility but limits the potential monetary risk to the payer if utilization exceeds the forecast. This kind of analytics-based outcomes payment can improve the biopharma's productivity, accelerating the collection of valuable outcomes data while simultaneously generating revenue. An added but important bonus: by sharing data in this manner, the biopharma strengthens its relationship with a critical customer.

Companies are already edging into this arena. For example, Vifor Fresenius Medical Care Renal Pharma, a joint venture between Vifor Pharma and Fresenius Medical Care, is developing an outcomes-based service to predict, and ultimately prevent, debilitating anemia in patients with chronic kidney disease. The service hinges on structured data collected from laboratory reports. As enabling technologies such as cloud computing and machine learning evolve, however, future algorithms will also take advantage of environmental and personal data as well [50].

Data Integration

The first two challenges, data integration and data quality, are largely technological hurdles that companies have started, albeit slowly, to address. Because of the fragmented nature of health data generation—via wearables, medical records, lab tests, and payer claims—one of the first steps is gaining access to the relevant data. As biopharmas invest more heavily in real-world evidence collection post-launch, they will generate some of these data themselves. Other types of data, for instance, anonymized electronic health record information or claims data, will come via partnerships with providers or payers. As noted, biopharmas are already active in this arena. Still other data, for example, air quality forecasts or consumer search patterns, can be purchased from third-party sources, including infotech players.

In addition to obtaining access to the needed data, biopharmas must create a common platform, whether via a third-party cloud or an internally built warehouse, to promote those all-important collisions of data (Figure 10-7). That common location encourages users to make connections between different types of data that were historically stored in discrete so-called “data lakes” inside or outside the organization. By integrating those different flows of data, the objective is to turn individual lakes into a larger data ocean. At even the most sophisticated biopharmas, however, the current level of in-house analytics capabilities makes this prospect more an aspirational goal than a reality.

To encourage data sharing, especially across large biopharma organizations, management teams will need to develop systems that consistently structure data in ways that preserve contextual relationships, even when the information has been de-identified due to privacy regulations. This is particularly important for unstructured data, as well as non-health data that has an impact on patient wellness (e.g., environmental data related to air quality for patients with asthma or chronic obstructive pulmonary disease). As such, strict data governance protocols, including using master data management techniques that resolve inconsistencies in how data are defined, are important to establish early on. Indeed, researchers at the MIT Sloan Center for Information Systems Research recommend that bigger companies consider designating a “data dictator” to establish common definitions and data management protocols for the safe and appropriate handling of data [54].

Data Quality

For data to be useable and generate new insights, it has to be of high quality. This is not exactly a new problem for biopharma data teams, but it has been made worse by the volume of data being generated and the tendency for it to be stored in silos. The problem is, if the “small data” are not trustworthy, having a bigger pool does not necessarily help. It just increases the risk that biopharmas draw erroneous conclusions based on their analyses. Moreover, if biopharmas are to avoid the garbage in, garbage out phenomenon, they must actively avoid what researchers at Northeastern University and Harvard University term “big data hubris,” in which “big data are a substitute for, rather than a supplement to, traditional collection and analysis” [55]. (See the box, “Google Flu Trends' Unanticipated Big Data Lesson.”)

Data Privacy

In many markets, privacy regulations require healthcare organizations to remove patient identifying information before the information is shared. That can make it more difficult to conduct the predictive and prescriptive analytics most likely to offer the biggest return on investment. For instance, a biopharma company may want to demonstrate the value of a new cardiovascular medicine that might compete with older, cheaper generics. It could use predictive and prescriptive analytics to identify the 20 percent of the patient population that is at greatest risk of a costly adverse event and begin to conduct clinical and observational trials to showcase the value. But to build a complete picture, the biopharma will likely want to integrate a range of different data, for instance, information transmitted from digitally enabled scales or activity trackers. Without established processes for data governance and integration, it could be difficult to combine the data in meaningful ways that give a complete and actionable picture.

There are also emerging compliance risks for biopharma companies. As data are combined from different organizations, there is still the risk of re-identification, particularly if health data are combined with governmental information such as US voter registration details. To limit the risks, data management techniques that include the aggregation of smaller sample sizes are helpful [57].

But biopharmas will also have to review their own cybersecurity plans to make sure adequate safeguards exist for the highly sensitive patient data stored within their organizations or in the cloud. Data generated from IoT-enabled devices such as smart asthma or glucose monitors could be particularly vulnerable to hacks by cybercriminals who might use details to commit fraud [58]. Indeed, even if no hack actually occurs, the threat of a potential cybercrime can damage a company's reputation and its product revenues. That is what happened to Hospira in 2015, when the US Food and Drug Administration issued a warning about software vulnerabilities associated with the manufacturer's Symbiq drug pumps. Although Hospira initially worked on software updates to prevent remote hacking, it ultimately decided to remove the product from the market [59].

Though many biopharmas clearly understand the potential business risks cyber threats pose, a 2015 survey by EY found significant gaps in capabilities. Interestingly, 65 percent of life sciences organizations surveyed revealed a significant cyber incident that was not otherwise identified by their security team; a similar percentage, 61 percent, also noted they do not have a dedicated cybersecurity role to focus on threats tied to emerging technologies. Closing this capability gap will be important as biopharmas rely more heavily on big data [60].

Pace of Change

In addition to the challenges described above, biopharmas developing data analytics capabilities face a fourth major challenge: the pace of change. It may be surprising, but just five years ago, electronic health records were not widely implemented across health systems, and wearable usage was limited to early adopters. Since then, there have been dramatic improvements in enabling technologies that ease data access and integration, and specific analytics tools required to make sense of the information are now available. The question for biopharma companies in this dynamic environment is, how do you invest in data analytics in ways that are flexible and scalable?

This is not a trivial question. Typically, these kinds of information technology investments take months, if not years, to implement, along with large amounts of capital. Given the velocity with which technology standards and platforms change, and the fact that biopharma business models themselves are evolving because of macroeconomic forces, there is a real risk that, by the time a biopharma creates an internal analytics engine, it will no longer be fit for purpose. That is a potential problem when revenue growth at many of the biggest biopharmas in the industry is already slowing.

Organizational and Cultural Barriers

It is relatively easy to create a narrative that draws a direct line between improved data analytics and better market performance. Unfortunately, the data proving the value of data largely do not yet exist; one exception, perhaps, is the use of data analytics to optimize marketing to key providers. Given the potential significant upfront costs associated with creating analytics expertise, this lack of clarity about return on investment creates a further disincentive to accelerate investment.

In addition, there are other cultural and organizational barriers that might limit willingness to invest in analytics. Big data analytics is not a core capability for most biopharma organizations. Traditionally, data experts sat in a backroom buying and manipulating databases. Data was not seen as a strategic imperative that will accelerate work across all aspects of the biopharma value chain. This siloed thinking makes it more difficult to integrate different kinds of data across the biopharma organization and, thus, reap full value for the investment.

Biopharma firms also need new kinds of expertise, particularly data scientists able to extract meaning from various types of structured and unstructured data. Not only do such individuals understand the limitations of different types of data, but they are capable of quickly building predictive models to reflect real-time changes, for example, in product sales or clinical trial recruitment.

Individuals with these capabilities are a rare commodity: the strategic consulting firm McKinsey predicts a global excess demand of 1.5 million data scientists [61]. And since many data scientists come from mathematics or engineering backgrounds, these executives will also need healthcare-specific training [62].

Analytics as a Strategic Imperative

Inexpensive computing power, the cloud, and increasingly robust algorithms powered by artificial intelligence are the primary technological drivers behind the current data analytics opportunity. However, these technology advances should be viewed as necessary, but not sufficient, for success. The overarching goal is to institutionalize a set of behaviors that promote using analytics as a core competency to generate better business insights. Such sweeping behavior change will only come about when biopharma's senior leadership views analytics as a strategic imperative.

That means having a top-level executive committed to analytics initiatives, who is capable of advocating for resources and elevating the subject to the board as appropriate. This person could be a chief information officer (CIO) or a chief data officer (CDO). The title is less important than the ability to have someone responsible at a business-wide level for articulating how analytics will support the overall business strategy. “We see many organizations that have spun up initiatives and are spending a lot of money, (but) do not necessarily have a clear point of view on how value will be delivered,” notes Chris Mazzei, chief analytics officer for EY [64]. Indeed, a CIO or CDO can help connect the dots between interesting ongoing pilots so the companies can reduce duplication and translate successful initiatives to other groups in the organization.

Changing Behaviors

One of the key jobs for management is to create the incentives that promote the appropriate sharing of data and analytics tools across the organization. This is both a people management issue as well as a technological issue. For larger companies, one way to accelerate the process is to prioritize the creation of a center of excellence focused on developing certain crucial analytics processes. This senior management team would be responsible for doing the following: developing protocols for data governance and aggregation; centralizing methods, tools, and models so they can be shared easily across the organization; creating shared metrics for success; and promoting new innovations. By focusing on these kinds of activities, senior management can drive behavior change, promoting an environment where the results of prior analyses are factored into new analytics projects. Importantly, biopharma companies that use this approach spend less time searching for the right data and tools and more time performing analyses to drive important business decisions.

For this approach to succeed, however, the activities of the center of excellence must be closely integrated with the actual business. Having a team of data scientists produce multiple product launch scenarios that are then handed off to a biopharma brand unit to assess and implement is less likely to succeed than if the sales and marketing people interact with the analysts as a single team to develop a precision launch plan. Only by working together can the analysts and the business organization build credibility with each other and establish clear alignment between the end analyses and the business unit's priority—in this example, a successful launch. Indeed, absent a strong working relationship between business leaders and data scientists, there is a real risk that biopharma decision makers revert to business as usual, making decisions that are not driven by analytics and evidence but rooted in their preconceived notions and historical biases.

Partner for Flexibility

Bigger biopharmas face the option of building their own in-house capabilities either organically or via acquisitions, while smaller, earlier-stage biotechs, because of capital constraints, will almost certainly need to partner to access the data, the platform-enabling analytics tools, and the human expertise. However, just because bigger companies have the wherewithal to “go it alone” does not mean they should. Building a robust analytics engine requires significant up-front investments in time and money to create the appropriate infrastructure and capabilities. Because the technology is changing so fast, there is a real risk of spending hundreds of millions of dollars to create a platform that is unusable or out of date given the changing healthcare market.

To maintain flexibility, companies should consider whether an “as a service” model can be applied to various aspects of data analytics (Figure 10-8). Many biopharmas already outsource certain elements of their business, partnering with third parties around manufacturing or aspects of research like assay development. Why not analytics as well?

Companies can partner with various third parties to get access to the capabilities that are most valuable based on their maturity and life cycle. A pre-commercial biotech in late-stage clinical trials could begin working with a commercial analytics specialist to develop analytics to optimize a pending product launch; a commercial biotech with just a handful of products might partner with an analytics provider to develop more customized pricing agreements with key health systems. Even larger biopharmas might be interested in working with bigger analytics players to integrate patient-centric data into their clinical trial design given changing regulations about when and how such data should be used. Simply put, by taking an analytics-as-a-service approach, biopharmas can scale-up—or scale-down—their use of analytics assets based on their changing needs. They can also limit their risks by contracting with a third party that can help generate the analytical insights.

Many of the biggest biopharmas will examine the analytics as a service option and decide that data-driven insights are such a core part of their business that they need to “own” these capabilities. Indeed, companies that are serious about moving beyond the pill may find that an essential component of their intellectual property is an algorithm designed to maximize product sales following launch. In this instance, companies will want to retain full control of the algorithm, and that means keeping the capabilities in house.

Even in this situation, given the variety of data that play a role in health, biopharma companies will need to establish partnerships with third parties. At a minimum, this means reaching outside the organization to access different types of relevant data to combine it with internally available data. But it also may mean sharing data with partners or suppliers to create new ways of doing business, for example, working with a payer to create an outcomes-based pricing model for a new but expensive drug. As the need to pursue such partnerships grows, companies will want to structure these agreements so that they retain control of the processes resulting in their data-driven insights.