The pace of progress in blood cancers is accelerating. Over the past decade, the FDA has approved dozens of targeted and immunotherapies, and breakthroughs like CAR T-cell therapy, bispecific antibodies, and MRD testing are transforming how these diseases are treated. What was once acute is increasingly becoming manageable, allowing many patients to live longer, more active lives.
But progress only matters if patients can access it. As treatment options expand, clinical decision-making has become more complex. Clinicians must navigate evolving standards of care while accounting for disease biology, biomarkers, prior therapies, and individual patient considerations. At the same time, many advanced therapies remain concentrated in academic centers and can be difficult to access due to cost and geography - raising real questions about the gap in equity across patient populations.
At Flatiron, we believe that gap is not inevitable. By generating high-quality, real-world evidence, we can help researchers address disparities and give clinicians the insights they need to make smarter, faster decisions across the patient journey.
I recently joined leaders from Bristol Myers Squibb, Roche, and Incyte to discuss how real-world evidence is shaping the future of blood cancer research. You can watch the full webinar here. If you’re short on time, here are the four key insights I took away from the event.
INSIGHT #1: “We need to better understand how therapies are actually used, sequenced, and experienced by patients in the real world”
Dr. Ahmed Sawas, Medical Director, Flatiron Health
This is how my Flatiron colleague Dr. Ahmed Sawas, who moderated our discussion during the event, summed it all up at the end, and I think it goes straight to the heart of the problem — and does a very good job of framing the enormous opportunity for RWE in this space.
There’s a distinction between eligible patients and those who actually receive treatment. Once a therapy moves beyond the controlled environment of a clinical trial, access invariably expands, and the patient population changes. For instance, real-world patients are more often treated in community settings rather than large urban academic centers, they frequently have comorbid conditions, and are far more diverse than trial participants in terms of demographics and socioeconomic status.
Demonstrating safety and efficacy in a trial is only the first step. Real-world success depends on how a therapy is delivered, who receives it, and how it is sequenced with other treatments. Understanding those dynamics requires data that captures not only patient characteristics but also external factors such as geography, practice setting, proximity to specialized centers, insurance coverage, caregiver support, and the patient’s ability to take time off work. These logistical considerations are especially critical for complex modalities like CAR T-cell and bispecific therapies, which are substantially more demanding to administer than traditional treatments.
INSIGHT #2: “While real-world data generally tells us about treatment sequence and outcomes, the ‘why’ behind it all is often more challenging to tease out”
Kim Saverno, PhD, RPh, Senior Director, Medical Affairs, Incyte
Kim Saverno from Incyte made that point early in the webinar, and it stuck with me because it underscores two critical points. First, that there’s a wealth of valuable information in unstructured, narrative data (like clinician notes, biomarker reports, and other free-text sources), and the industry needs to intensify efforts to extract those insights to build a more complete picture of the patient journey. Second, that expectations for RWE have evolved: it’s no longer sufficient to document the patient journey; increasingly, RWE is expected to help explain why that journey took the turns it did.
This is a big change from where RWE stood only a few years ago. We are now using RWE to investigate the whole decision-making process — not just by physicians, but also by patients and caregivers along the entire care continuum. At scale, this fundamentally reshapes how evidence is generated and applied in clinical practice.
INSIGHT #3: “We need longitudinal and connected datasets that follow patients as they flow from care to care within the system”
Fei Fei Liu, Exec. Director, Global HEOR, Bristol Myers Squibb
We talked throughout the webinar about the importance of longitudinal data, and I love that remark from BMS’s Fei Fei Liu because it reminded us of the challenges most patients face along the journey. Many patients begin their care in the community setting, experience a relapse or seek additional options, transition to an academic center for specialized treatment, and then return to their local clinic for follow-up. In between oncology visits, they may also see their primary care physician. The result is often a fragmented journey that can feel like an obstacle course. As much as we wish electronic health records in the US were fully interoperable and seamlessly synchronized across networks and care settings, that’s rarely the case. As a result, RWD needs to integrate disparate data sources in order to reconstruct as complete and accurate a view as possible of each patient’s longitudinal medical journey.
Fortunately, a new generation of AI tools is helping researchers connect the dots. All panelists agreed that specialized large language models (LLMs) are transforming the RWE landscape, enabling teams to capture far more detailed real-world data and make sense of it in record time. In a recent blog, Dr. Sawas described how LLMs are allowing Flatiron to develop bigger, deeper, and faster real-world datasets and shared how those datasets are already being used to analyze treatment pathways for patients with B-cell lymphomas, multiple myeloma, chronic lymphocytic leukemia, and B-cell acute lymphoblastic leukemia. I definitely encourage you to read that blog to learn more about these recent breakthroughs.
INSIGHT #4: “The biggest challenge we face today is not lack of data, but the quality of that data”
Ashwini Shewade, Principal RWD Scientist, Hematology, Roche
Every time we talk about technological progress, especially with AI these days, I’m reminded of Mark Zuckerberg’s famous call for engineers to “move fast and break things.” Obviously we don’t want to do that in our field. Researchers, clinicians, and regulators will only rely on RWD they can trust, and that trust depends on rigorous, fully transparent methodology. We cannot simply ask an AI tool to extract information from a physician’s note and incorporate it into a dataset without robust validation — particularly when generic LLMs may produce different results with each request.
In high-stakes domains like healthcare, speed must be matched by accuracy. Without safeguards in place, even the most sophisticated tools risk undermining confidence rather than advancing evidence generation. At Flatiron, we developed the VALID framework to compare LLM-extracted data to human-abstracted data, using a very rigorous set of metrics, benchmarks, and replication procedures to make sure that what the LLMs produce is at least as good if not better than what an expert in the field would capture.
RWE today is used by blood cancer researchers across the entire drug development lifecycle, from mapping out unmet needs to optimizing I/E criteria, developing synthetic control arms for accelerated applications (here in the U.S. but around the world as well), all the way to supporting postmarketing requirements. Data quality is paramount, and the VALID framework allows us to improve the scale, depth, and speed of our hematology datasets without compromising quality.
Those are my four big takeaways from the webinar, but the full recording is rich with many additional insights, and I invite you to watch it in its entirety when you have an hour to spare. I’m deeply grateful to my fellow panelists for their generosity with their time and for their candid, thoughtful contributions. We still have a great deal of work ahead — as Fei Fei Liu rightly pointed out during the webinar, “success means we leave no one behind, and whoever needs care gets it” — but it was energizing to see that our initiatives at Flatiron are aligned with the expectations of leading real-world data experts. That validation is exactly the momentum we need as we accelerate our data expansion efforts this year.
