The majority of biopharma organizations are investing in real-world evidence (RWE) technologies, and are beginning to use RWE across every phase of the product lifecycle—from discovery through post-launch and outcomes-based agreements with payers. This is helping organizations move products through the pipeline more quickly, and develop market access and launch strategies.
With such a wide range of applications for real-world data (RWD), biopharma organizations are tasked with choosing the organizational models that work best for their teams so they can capture and communicate the value of RWE.
Two organizations which have adopted enterprise-wide approaches to RWE generation, Sanofi and Janssen R&D, joined a webinar Aetion hosted with Informa to discuss how they’ve structured their teams for success, and how they’ve used RWE to accelerate drug development for COVID-19 and beyond.
Read on to learn more from the conversation between Najat Khan, Ph.D., Chief Data Science Officer of Janssen R&D, and Bob LoCasale, Ph.D., Head of RWE at Sanofi, including their advice for biopharma organizations looking to develop RWE capabilities.
Responses have been edited for clarity and length.
Q: How is RWE structured at your organization? What decisions have contributed to your approach?
NAJAT KHAN (NK): At Janssen R&D, we’ve intentionally integrated data science across all of our therapeutic areas and functions, and it’s embedded in our strategies and priorities to help advance our pipeline.
We have a centralized structure for incorporating data science, which encompasses everything from artificial intelligence and machine learning to RWE and epidemiology methodologies. We also have data engineers who ensure that the data platforms are used effectively so we can run analyses at scale, as well as data science product portfolio leads who blend medical and scientific understanding with program design expertise to help craft proficient data science programs.
BOB LOCASALE (BL): Our RWE team is centrally responsible for designing and delivering external-facing RWE, which leverages mostly pre-existing data. The team is made up of expert methodologists and epidemiologists, outcomes researchers, biostatisticians, and data and literature analysts.
We’re in the process of transitioning from a Center of Excellence model to a “hub and spoke” model. My team remains the center of RWE expertise, but we’re building individualized “spokes” of RWE functions within global business units. Some of these even have “spoke-within-spoke” models for more specific functions, such as R&D.
This shift is in direct response to a rising demand for RWE across the product lifecycle. Ultimately, we aim to bring RWE functions even closer to global product and brand teams, and even to local affiliates, so we can provide real-time, rapid insights while maintaining a center, which will continue to focus on publishing and generating regulatory-grade RWE.
Q: Can you share an example of how RWE enables your R&D organization?
BL: One example, which we executed a few years ago and continue to grow today, was a new indication-seeking exercise. We leveraged our clinical expertise within the R&D function, as well as other internal knowledge, to identify and build an immunology cohort within a real-world, electronic medical record (EMR)-based dataset. We then ran unsupervised machine learning to identify phenotypic clusters of patients. This helped create a list of new indications, which we ranked according to proximity to anchor indications.
In the end, we were able to identify about 90 percent of the anchor indications, and a variety of other new indications that were yet unexplored. This was a much more rapid process of identifying new indications than we had done previously and provided newer information that we had in the past.
NK: RWD has been central to our work on the COVID-19 vaccine. We used data first to ensure we had diverse clinical trial populations, which is especially important due to the high incidence of the disease, and to help ensure we ran the trials effectively. Now we’re looking into using external control arms for additional studies.
When the COVID-19 pandemic started, there were a lot of unknowns: who is at highest risk, what endpoint do you pick for studies, how do you predict the next hotspot? We needed rapid answers, so we developed a machine learning model to run on top of the RWD—which included data on global, state, and county-level infection rates, mobility data, and other consumer data—to predict future hotspots and social compliance, and to understand where pandemic mitigation policies would be enforced more strongly. This model helped us make predictions four months in advance, at a county level.
Now, our predictions are over 90 percent accurate. This is a first-of-its-kind model that has helped us not only accelerate development timelines by six to eight weeks—which is especially important during a pandemic—but also establish a rich dataset, which we can use to run sub-analyses.
Q: Can you speak to the critical role data plays in generating RWE?
NK: It’s important not only to have access to the data, but to have access to the right types of data and be proactive about using it. The investments that we made in an end-to-end platform, and in linking data across our internal and external datasets, has helped accelerate how we use the data to build the right models.
There is also so much variability in the data, especially at a global level. So in order to run diverse trials that are reflective of a real-world, global population for our COVID vaccine trials, we needed to establish a strong understanding of ex-U.S. data. Then, in order to use the data, we had to do a lot of data cleaning and back testing with local authorities to be able to pull data from across the world. It was quite the effort, but the data component is crucial and foundational to running a study.
BL: The data is ultra-critical. We have made fantastic advancements in analytics; now we need to have fantastic advancements in data—beyond EMRs and claims, which we’ve used for decades. We’ve made progress in data curation, which is great, but we need to continue to link and evolve the data so we can remedy the data missingness issue.
Additionally, while we have the ability to collect secondary data—data that exists for purposes other than research—it often lacks behavioral data. You can get some of this in the unstructured space, but a lot of it needs to be collected directly from patients. We need pre-existing, prospective models to marry all of this data together and solve for missingness, then we can have confidence that we’re capturing all of the potential confounders that could have a role in the interpretation.
In terms of data access, we certainly can’t just license all data. Therefore, we need to be able to partner with organizations and create different models to access the data. With that comes the importance of the tools to be able to interrogate data. Partnering with Aetion provides the robust tools for generating high quality insights and running studies in a more rapid way, but also at a high level of quality.
Ultimately, you need a marriage of data sources, along with the tools that allow you to interrogate that data properly, to reach a better study output, and, in the end, build a better product.
Q: How can RWE support research on rare diseases?
BL: For rare diseases, where there is often unmet medical need and limited treatment options, existing and routinely collected health care data aren’t sufficient to power analyses. We need to think about primary data collection for very specific research projects, so we can reserve these data as a registry to answer other research questions.
Another area of opportunity with RWD for rare disease is to help identify patients earlier in their disease, before they’re diagnosed. We’ve been using advanced analytics within Sanofi to see if we can identify patterns in care through larger datasets which would potentially signal a patient eventually having a rare disease earlier in their pathway. We could then build this into a clinical decision support system tool to signal to the provider that the patient could have this rare disease.
NK: We know that in rare diseases, patients go an average of five or more years of being misdiagnosed. For example, in diseases like pulmonary arterial hypertension (PAH), which has non-specific symptoms like breathlessness and fatigue, PAH is initially overlooked whilst more common respiratory conditions, such as asthma and chronic obstructive pulmonary disease (COPD), are investigated as possible causes; it’s not until the heart is investigated that PAH or another type of pulmonary hypertension will be suspected.
In addition, while early detection is certainly helpful, the data is quite fragmented. This raises the question of how you can both identify patients that have a rare disease and gain access to the necessary data. In the U.S., it’s quite challenging to create a robust, longitudinal history for patients, then parse out the signatures that highlight disease progression. This is a bit easier in some countries in Europe, which have cradle-to-grave, fully connected datasets.
Q: What advice would you give other biopharma organizations looking to incorporate RWE?
NK: My advice would be to look at your program and identify three to five top priorities to tackle, then leverage the internal and external ecosystems to do so. It’s also important to understand the problems you’re solving for. Once you’ve figured that out, doing the right feasibility and choosing the right datasets and algorithms becomes much more manageable. You also need to understand the limitations you face, and take a balanced approach.
BL: You need to target the part of the business process that you think RWD can change, because it ultimately needs to impact the business operating income to gain necessary leadership support for the investment in an RWE program. So you need to measure the impact well, that way you can show the value of RWE versus the alternative.
Finally, RWE is not a quick and cheap solution to answer questions. The things you want to impact and change with RWE are important, and ultimately improve patient lives, so it’s important to make the investment in high quality RWE tools and programs.
Q: What do you see as the future of RWE?
NK: We are at an inflection point for RWE. The external data ecosystem is improving and evolving, but no dataset is perfect. In general, there are more and more proof points of real problems that we’ve solved with RWE to enhance the probability of a program’s success.
BL: RWE, for me, is about patient-generated data, and uncovering the heterogeneity within patient populations to generate insights on groups beyond homogenous clinical trial populations. The power of RWE is clear for me in that space, and we can achieve it by enhancing our data input.