Life Sciences R&D & Medical

Q&A: How biopharma can find common ground on real-world evidence

By Anna Sato, Qin Ye, Matt Docherty, and Nathaniel Leung

Jan. 13, 2023 | Q&A | 7-minute read

Q&A: How biopharma can find common ground on real-world evidence

Real-world data and evidence (RWD/RWE) has the potential to significantly enhance and shape the future of clinical development in biopharma, an idea we explored earlier this year in an article describing a collaboration between AstraZeneca and ZS. To understand what needs to change for this to happen, Qin Ye, a principal in ZS’s RWE practice, sat down with Mats Sundgren, Ph.D., an executive health data strategy advisor for industry and academia. Sundgren has driven RWE progress for over a decade in leading in roles at both AstraZeneca and IgniteData, while Ye has a background in medicine and health informatics. Their conversation, moderated by ZS Consultant Anna Sato, focused on imagining the capabilities of RWD and RWE and discussed how biopharma teams can improve their RWE functions.


Anna Sato: What needs to happen for RWE to fulfill its potential to impact the clinical development life cycle?


Qin Ye: There is a common language gap and a multidisciplinary challenge to overcome. The typical clinical trial team, from study physicians to biostatisticians, are accustomed to controlled settings and very structured data, with a consistent process on data collection and data lineage tracking. When you have that kind of expertise, naturally, there are concerns with RWD, including lack of structure, varying levels of completeness and potential biases. These concerns can be a barrier for clinical teams considering RWD, especially if they intend to engage in regulatory submissions. I think there’s still a bit of a gap in understanding and expectations.


On the other hand, the RWD analytics team may be very proficient with RWD and feel comfortable using advanced analytics to analyze RWD to generate meaningful insights. However, they may not necessarily have an understanding of the key decisions and processes for clinical trials, such as the expectations of key stakeholders and regulators.


There’s a gap between what you’re comfortable with versus what’s needed, and I see that as one of the key challenges. Even though there is a common goal, people from other disciplines can have totally different assumptions.


Mats Sundgren: I’m thinking about four factors. The first is RWD needs to be high quality and that’s going to take time. Natural unstructured data will have an implication someday, but its capabilities need to mature further. For now, I think the king is going to be structured data.


Number two, I would like to see more collaboration among vendors, industry and healthcare around driving these services and using RWD. There should be transparency about methods and services and I’d like to see regulators roll that forward. I think accountable care organizations and hospitals where RWD is created are particularly important stakeholders here.


The third thing is education, and that goes for stakeholders within the pharmaceutical industry. There needs to be education about the potential here. I also think that’s a need for healthcare organizations (HCOs) as well.


The fourth factor is probably accreditation. This is a factor I have brought up in my conversations around electronic health record to electronic data capture (EHR2EDC) initiatives, particularly with IgniteData, the EHR2EDC consortium and the EHR for Clinical Research program. With accreditation of RWE use, we could drive the implementation and adherence of RWE, providing a driver’s license of sorts.


QY: Just to add to that, I think right now electronic medical record data collection is primarily for patient care, billing and capture of service details. It wasn’t designed to support regulatory decision-making. I think creating synergy between those two worlds and then introducing accreditations would create incentives for how systems, physicians and clinicians can improve the quality of data collection in standard patient care.


AS: What are some of the prominent AI and machine learning (ML) applications you envision for providing external evidence in clinical development?


MS: If you go beyond the three basic services of feasibility, recruitment and execution, I think using a sophisticated algorithm on the back end of hospitals could help identify patients for trials. For instance, this would be useful for nonalcoholic steatohepatitis (NASH) patients because they are notoriously difficult to identify without a painful screening program invoking invasive biomarkers.


Irritable bowel disease could be similar because it’s still based on questionnaires and outcomes research-related qualitative measures. With much more sophisticated RWD available now, I think it’s absolutely possible to utilize algorithms for patient identification in this disease area as well.


In ophthalmology, applying AI and ML to imaging data could make it possible to identify diseases of the eye much faster. Also, in areas like Alzheimer’s disease and central nervous system conditions, it may be possible to use data-intensive medical scans and images to build algorithms detecting deviations. This not only enhances the ability to do higher-quality trials, but also allows providers to better help their patients. Again, it’s a very nice synergy.


QY: I also think leveraging deep learning to enable patient phenotyping based on digital biomarkers and large combined data sets would be very interesting. This could be used to both predict and retrospectively evaluate how a patient actually responds to treatment. Using this to shape trial design could help accelerate the pace of precision medicine.


I think this could be a very important capability to enable truly enriched trial design, as many of today’s designs are largely based on approximations and estimations from physicians. Taking from the NASH example, on top of disease prediction, screening is also a huge challenge since most patients are under-diagnosed. When designing a trial, investigators need to understand disease progression and separate and identify patient subgroups reliably. Here, deep learning shows potential to identify trends around both phenotypes and genotypes. From there, being able to apply them to the right treatment for those subtypes will be very important and very exciting.


AS: Do you have any tips for others trying to bolster RWD and RWE initiatives in their organizations, in terms of setting up a collaborative working model for evidence generation and maximizing output?


MS: It probably goes back to my third factor about collaboration. I think we need initiatives that bring the industry, vendors and hospitals together. And hopefully regulators would say this should be supported by transparency and standards.


But back to your question. I think, yes, bring me the evidence. Once we have tiny but hard evidence of success, that helps to change the mindset of the industry and hospitals. This can be done within one company, a set of HCOs and of course with a vendor partner. It would be better if you could share, go across different therapeutic areas and perhaps different kinds of constellations, and arrive at the conclusion that bringing this accumulated evidence really works. So, why is this not happening? It’s because everyone is so busy and it’s an expensive business. I think Janssen Pharmaceuticals is one of the brave ones—they are very collaborative as well as competitive.


AS: Whats at risk if companies dont invest sufficiently in RWE for clinical development?


QY: I think a couple things are at risk. The success of pharmaceutical or health tech companies depends on their ability to accelerate clinical development and their ability to create a sustainable process to reliably and scientifically answer their stakeholders’ questions. When a company does not have adequate RWE capabilities, they will likely have a diminished data-driven ability to design and execute trials, engage with the FDA and other stakeholders and create integrated evidence plans to deliver the evidence package systematically.


This creates two risks. The first is a biopharma company cannot realize the full potential of their product, which impacts their commercial success. The other risk is without adequate RWE capabilities, the organization may bring in a lot of confusing or incomplete data that would be difficult for any stakeholder to interpret. Then, they may not be able to achieve their ultimate objective of improving patient outcomes.


AS: Any other closing thoughts?


MS: I have an additional one, Anna. I completely agree with Qin, and I also think there’s something more hindering innovation. Not only can the pharma industry deliver medicines faster through RWE support, but it also can enable new research that wouldn’t otherwise be able to go forward.


For example, the vision for using EHR2EDC during my time at AstraZeneca and now IgniteData was to reduce site burden. Sites can now recruit patients more quickly and operate more cost efficiently. Another thing I’d add is a little bit more subtle and philosophical. If innovations such as EHR2EDC were to operate at full swing, it could actually allow us to perform research that otherwise wouldn’t happen or develop medicines that otherwise would never make it to the patient. I think that’s quite big.


To enable this, I will go back to my previous point on collaboration. I think John Nash put it best when he said, “The best for the group comes when everyone in the group does what’s best for himself and the group.”

Add insights to your inbox

We’ll send you content you’ll want to read – and put to use.

About the author(s)