Parexel
Johnson & Johnson
he continuing growth of real-world data (RWD) sources and better-quality real-world evidence (RWE) is changing pharmaceutical development and regulatory decision making for sponsors. For example, in 2023, FDA issued new guidance to industry on the use of RWE to support regulatory decision making, a strong signal of their shift toward more acceptance of these less traditional data sources. As a complement to randomized clinical trials, real-world evidence holds tremendous promise for the future of pharmaceuticals, patient care, equitable access, and how we develop and bring drugs to market.
Wyatt Gotbetter (DIA): Kourtney, what role does RWE play in helping to close the information gaps for diverse populations in healthcare research, and how is RWE helping to advance Johnson & Johnson’s mission and the aims of FDA and other agencies?
Kourtney Davis (KD): Real-world evidence can play a really important role in increasing representation because it complements the evidence we generate in our clinical trials. We do our best, but trials may not always reflect the full range of patients who are indicated for an investigational treatment and may eventually receive treatment post-approval. What we can do with real-world evidence is cast a wider net and that lets us focus on better understanding how a product may work across many different patient populations who are treated in actual clinical practice.
DIA: How are you doing this during clinical development versus the post-approval setting?
KD: Explaining how real-world evidence is generated at Johnson & Johnson will help explain these two domains, pre-approval and post-approval. I lead the pharmaceutical sector epidemiology team, part of the Office of the Chief Medical Officer (OCMO). The OCMO combines safety science, epidemiology, and bioethics expertise to inform decisions we make across the lifecycle, making sure that they’re evidence-based and that patients are at the center of those decisions. One way we do that is using real-world evidence; it allows us to understand safety and effectiveness in context, and also fill in critical gaps among less-studied populations.
Throughout the product development lifecycle, we analyze real-world data to help us further understand the target disease population. Everything in pharmaceutical development is teamwork, so we’re always collaborating in a cross-functional way. Early in development, we work with colleagues in R&D, data science, clinical, medical affairs, regulatory affairs, and biostatistics. We generate characterizations of the disease population so we can better understand the current standard of care and be able to compare our new intervention to that standard. We ask what treatment patterns look like for this target disease population. We also ask how we can identify unmet effectiveness needs and who is not responding to the therapies that are currently available. Plus, we work to understand potential safety or tolerability unmet needs. Maybe some of the current standards of care have toxicities or other safety concerns that a new treatment would want to improve upon. That’s one example of what we do during pre-approval stages.
Another would be understanding patient preferences. We have a benefit-risk team as part of Epidemiology. One of the things they work on around phase 2 with the team responsible for developing a compound is understanding how patients view trade-offs of benefits and risks for an intervention. Having that patient perspective is critical in designing development programs.
DIA: Our opening question was how we’re using RWE to help address and include more diverse populations. Can you share examples or insights around gender, ethnicity, or country of origin that may help that benefit-risk team shape your R&D programs?
KD: When patients are recruited into a patient preference study, we want to make sure the sample of patients is representative of the eventual target patient population that would be recruited into phase 3 trials and that later may use the product in clinical practice.
Part of how we set thresholds or targets for the recruitment of a patient preference study is based on those characterizations we do early on. Part of the descriptive analyses can include questions like: What is the gender distribution for this disease? What is the age distribution? What is the comorbidity pattern of this patient population? What is the race and ethnicity breakdown of this patient population?
DIA: This naturally leads to challenges in implementing these programs and using observational data, in contrast with controlled clinical trials. For 100 years or more as an industry, the idea of the randomized controlled study has survived because we are controlling risk factors, and now we’re inviting them in. What are the challenges in gathering this observational data or trying to draw insights from it?
KD: There are challenges to overcome in how real-world data are collected because most real-world data are repurposed for research. They were either gathered for administrative or billing reasons, so insurance claims are one of the main sources of real-world data; or they’re collected for healthcare delivery, so medical record data can be another source for RWD, as are patient registries. Registries collect data that are specific to the purpose of the registry, so depending on the questions we are trying to answer in our research, they may still not line up perfectly. Some of that real-world data are great for their original purpose, but using those data for research may be challenging. However, most of that data is still extremely worthwhile and we can work around its limitations.
We found that by collaborating with data partners, we can get access to either larger data sets or richer data sets that include more of the fields we need to either characterize the severity and prognosis of the patients or have more valid outcomes in the data set. Partnership is key and, to that end, we have worked increasingly with federated data networks. Federated data networks are a consortium or collection of data partners that work together to answer questions. For example, when we are working on a rare disease treatment, we need multiple data sources to have a large enough pool of data to be able to draw any inferences, and these networks can help us do that. Federated data networks can also create significant efficiencies using a single protocol, standard code, and a common data model.
DIA: Do these federated organizations allow you to have a large enough patient pool to capture that population data?
KD: There are about 300 million people worldwide who live with more than 6,000 different rare diseases. What we’ve been able to do in the federated data networks is bring a collaborative approach to solve the problem of sparse data. A couple of examples: In multiple myeloma, we have helped to sponsor a federated network called HONEUR, which stands for the Haematological Outcomes Network in Europe, that brings together patients with hematologic malignancies to improve outcomes. It now has over 22 data partners from nine countries and brings together data from 60,000 patients that are all mapped to the same common data model. We can answer a lot of different questions when we have data from nine countries and 60,000 patients. You can also look at smaller groups within that pool and answer questions for a targeted therapy. These networks have helped generate critical evidence to understand unmet need and prognosis and inform regulatory approvals and health technology assessment submissions.
DIA: Before RWE was called RWE, we would often work with patient advocacy groups and look at registries in the rare and orphan space. You mentioned HTA, which suggests the burden of disease and the experience of the caregiver and the family. Are you also using registries and working with patient advocacy groups for this purpose?
KD: We are still working with registries in multiple therapeutic areas. Johnson & Johnson has several therapies in pulmonary hypertension, a progressive pathophysiological disorder that could be associated with cardiovascular and respiratory diseases with significant unmet need in those patients. The PHederation network is a public-private partnership led by Johnson & Johnson that has six different data sources including multiple registries: about 12,000 patients around the world that have pulmonary hypertension. That core data is registry data.
DIA: We can only have data from those individuals who are accessing the healthcare systems where they live. Do you still see blind spots in the available data?
KD: We do have challenges in select data sets, especially the ones not created for research purposes. Race or ethnicity is often missing from these data sets, or not captured. If it is included, it may not be completed for 30% to 50% of persons in a data set. Sexual orientation and gender identity are almost always missing from routinely collected health data sets as well. These variables could be missing for legitimate reasons, such as patient privacy, and we want to respect that. Often, we are not the owner of the data; we are just using it for research purposes with the appropriate consent, and we need to protect patient privacy. That said, we also want to be able to build future data sets that allow us to answer questions that matter to patients. It is a balance between keeping data secure so that it is trustworthy information, but also being able to answer questions that are important for improving health outcomes.
DIA: We mentioned that FDA and other regulatory bodies are looking to improve inclusion and health equity: Do we have the data sources and other means to fulfill their mandates? What’s being asked versus what can we deliver with RWE today?
KD: Regulators have become more explicit in their requests for diversity data, opening the door for significant improvements in the availability of more representative data to inform regulatory and clinical decisions, and we are making a lot of progress in addressing information gaps in populations that are underrepresented in clinical trials. If you look back at where we were with representation of women in clinical trials 20 or 30 years ago, we have come a long way. For example, if you look at a representative real-world data set like SEER, the US National Cancer Registry, we are likely to have a pretty good idea of the racial, ethnic, age, and gender or sex breakdown for a cancer population that would be treated in the real world.
Where we don’t have a resource like SEER, or we don’t have a nationally representative registry for a disease, there is more work to do. You may need to combine different data sets to try to get at what is representative of the target population eligible for a future treatment. We are doing the best we can and constantly challenging ourselves to do better. Again, working with data partners and patient advocacy groups, as well as working in public-private partnership-type consortia, is one of the most important ways to improve data representativeness, quality, and transparency.
DIA: It’s impossible to have any conversation in life sciences in 2024 without discussing AI. How is AI helping to address some of the challenges and further your aims in this space, especially as data is more plentiful?
KD: It’s very early to know exactly how AI is going to transform real-world evidence generation, drug development, and healthcare, but we can say confidently that we know that the transformation is coming. There are lots of opportunities for AI to facilitate identification and summarization of epidemiology information that will drive efficiency. Activities that are typically done manually and repetitively will be sped up by AI tools. AI is going to help us process unstructured data in medical charts and help us to conduct and complete data validation studies much quicker, leading us to more robust evidence generation. As researchers, AI will allow us to focus more of our energy utilizing valid information analysis and interpretation activities as opposed to manual extraction tasks that can be time- and resource-intensive.
AI can also aggregate and analyze huge amounts of data and help us identify patterns and trends in those data. For instance, we started our conversation talking about unmet medical need: we will be able to use AI to set up algorithms to generate hypotheses very quickly, but the human is still going to need to review results for sense checking and integrating with other evidence. Predictive analytics is another place where AI is going to be very powerful and helpful to us. Internally, it can help us target our interventions as we design our programs, but it will also help in clinical practice. Real-world evidence will be at the practitioner’s fingertips, to be able to show a patient with a set of specific characteristics this benefit-risk profile versus that benefit-risk profile. I’m very excited about what it can do to improve decision-making for regulators, for healthcare providers, and ultimately for patients.
DIA: There’s also a nice complementarity, perhaps a virtuous cycle, of using the technology and the RWE to allow us to identify patients where they are, maybe based upon clinical presentation, to bring them into the right modes of care, perhaps clinical trials, and thereby enrich the RWE available to the field. Where do you see the RWE field in five to 10 years, and how will that help advance health equity and inclusion?
KD: We already mentioned the data are changing. The more data providers understand what good data looks like, the more likely we are able to access data sets that better reflect the patient populations who may eventually receive the treatments. The FDA guidance for clinical trial diversity plans was an important step, asking companies to set diversity targets that are evidence-based. As the data have become better, we have been able to build more robust diversity plans, set more accurate recruitment targets for clinical trials, and then plan accordingly. In other words, design our studies and select sites in the right locations with the right recruitment strategy so we can achieve those targets. The kind of data we’re going to be able to generate post-approval should also continue to fill in health equity gaps. Not only do we want more representative trials, but we want to be able to immediately follow up after authorization and generate evidence that demonstrates the benefit-risk profile in diverse populations treated in clinical practice.
Pragmatic clinical trials are another promising area, one that has been a dream of many health authorities and epidemiologists around the world. We would love to see real-world data help us not just plan studies but actually be part of the study data collection. By embedding studies in real-world data systems with randomization at the point of care, patients enrolled in the pragmatic trials can be monitored longer term for effectiveness and safety outcomes with lower burden to patients, enabling the creation of very rich data sets. These real-world data sets can answer many questions for multiple stakeholders and help close several of the evidence silos we’ve discussed today. In the end, to overcome health equity gaps, it’s important to lead with both science and care. Innovation that is evidence-based and patient-centered will be key as we look at what’s possible over the next 5-10 years.