As Real-World Evidence Gains Ground in Oncology, Where Does Flatiron Health Fit In?

In oncology, new therapies arrive at a dizzying pace—targeted drugs, antibody-drug conjugates, next-generation immunotherapies. Yet for clinicians and drug developers, every advance seems to spawn as many questions as answers. Randomized Controlled Trials (RCTs) show what works under ideal conditions. They rarely reveal how those same treatments perform across diverse patients, shifting standards of care, and the messy realities of everyday practice.

That widening gap has thrust real-world evidence (RWE) into the spotlight. Once dismissed as secondary, real-world evidence is now indispensable for regulators, for trial designers, and for physicians deciding what to prescribe tomorrow.

At the center of this transformation is Dr. Emily Castellanos, Head of Research Oncology at Flatiron Health. In a conversation with GeneOnline, the oncologist who sits at the intersection of clinical care and data science describes how she and her team turn millions of de-identified patient records into evidence that actually moves the needle. Here, she explains why real-world evidence has gone from “nice-to-have” to “must-have,” and how Flatiron is helping shape the next trend of cancer research and care.

Flatiron Health Elevates Real-World Data into Decision-Grade Clinical Evidence

Flatiron’s platform is no longer defined simply by the scale of its oncology database. Its differentiating value lies in how real-world data is systematically transformed into decision-grade clinical evidence.

“The platform is foundational,” Castellanos noted, “but what sits on top of it is our clinical and scientific expertise, combined with deep collaboration with partners and industry stakeholders, to ensure the evidence we generate is fit for real-world decision-making.”

For biopharma companies, this distinction has become increasingly important. The challenge is no longer access to data, but the ability to reduce uncertainty at each stage of development. Increasingly, real-world evidence is being embedded across the entire product lifecycle, from early clinical development and trial design to post-approval studies and label expansion. It provides continuity in an increasingly complex and fragmented treatment landscape

RWE Answers Questions RCTs Cannot

Despite its rising profile, real-world evidence is not trying to replace RCTs. It exists to answer questions that RCTs simply cannot. “Real-world evidence and randomized controlled trials serve fundamentally different purposes,” Castellanos said. “Our goal is not to replace RCTs.”

RCTs remain the gold standard for establishing efficacy under tightly controlled conditions. But their very strength, including strict eligibility criteria, standardized protocols and limited patient diversity, also limits what they can tell us about how medicines perform once they reach the clinic. Real-world data fills that gap, showing how therapies are actually used, how patients with comorbidities respond, and how treatment strategies evolve over time.

“I think real-world evidence can play an important role in helping physicians answer questions that may not be answered in a clinical trial,” she added. These are questions around comparative effectiveness, optimal sequencing, and duration of therapy that often emerge only after approval. 

For example, the P-VERIFY study leveraged Flatiron data to evaluate real-world progression-free survival of three CDK4/6 inhibitors plus an aromatase inhibitor in HR+/HER2- metastatic breast cancer, addressing comparative effectiveness questions that trials alone could not fully resolve.

Flatiron’s track record speaks for itself: its data have supported more than 40 regulatory decisions by the FDA and EMA, offering regulators a clearer picture of how drugs behave in broader and more representative populations.

Trust in RWE Depends on Transparency and Fit-for-Purpose Use

As real-world evidence gains influence, skepticism persists. According to Castellanos, the biggest threat to trust lies not in the data itself, but in how it is used.

“Trust in real-world evidence gets broken when the data is not used in a fit-for-purpose way,” she emphasized.

Not every dataset is suitable for every question. Misalignment between the data source, the analytic method, and the intended use can lead to misleading conclusions. Flatiron counters this with methodological transparency and rigorous validation. The company has published detailed frameworks for defining real-world response and progression, along with the VALID framework (Figure1) for assessing the quality of AI-extracted data.

Equally important is its collaborative approach. Rather than simply handing over datasets, Flatiron works side-by-side with sponsors to refine study questions, align on methodology, and ensure regulatory acceptability. The result is evidence that is both robust and genuinely useful, showcasing the proper usage.

Multimodal Data Opens New Doors and New Challenges

Linking clinical records with genomics, imaging, and claims data is one of the most exciting frontiers in real-world evidence. It is also one of the most technically demanding.

“The challenge isn’t just linking the datasets,” Castellanos explained. “It’s ensuring statistical power, clinical relevance, and fitness for use.” Every additional data layer shrinks the eligible patient population. She offers a simple analogy: “Think about a deck of cards. Start with 52. Filter for hearts, you have 13. Filter for face cards, you have 3. With every filter, the population shrinks.”

Flatiron’s answer is starting from a scale as large as possible. With more than five million patient records and growing, the company pairs this vast dataset with sophisticated AI that can extract structured information from unstructured clinical notes at efficient speed and accuracy. Yet scale without quality is meaningless. The company continuously validates AI outputs against human-curated benchmarks, runs replication studies, and insists on full data provenance so every finding can be traced back to its source.

From Trial Design to Label Updates

The proof is already visible in the clinic and the boardroom.

When Amgen developed Lumakras (sotorasib) for KRAS G12C-mutated non-small cell lung cancer, a relatively rare patient population, regulators asked for a natural history study to better understand how these patients were being treated, and what their clinical outcomes were. Flatiron’s linked clinical-genomic database delivered data on more than 7,000 patients, evidence the FDA ultimately cited in its approval decision.

In another case, RWE from Flatiron was used to understand the safety of Kadcyla (ado-trastuzumab emtansine) in breast cancer patients with pre-existing cardiac conditions. This data enabled fulfillment of the sponsor’s post marketing commitment entirely with real-world evidence, cutting roughly five years off the anticipated study timeline and supporting two label updates.

Similarly, Flatiron real-world data also supported FDA approval of a new every-two-week dosing regimen for Erbitux (cetuximab), a schedule which enabled reduction in the number of patient visits while maintaining patient safety and clinical effectiveness.

RWE is also being integrated into how trials themselves are designed. Sponsors now use real-world populations to test inclusion and exclusion criteria, improve enrollment diversity, and build external control arms or digital twins, tools that can accelerate development while maintaining scientific integrity.

Toward a True Learning Health System

As the oncology community prepares for the upcoming ASCO Annual Meeting, Castellanos sees real-world evidence evolving from descriptive analysis to true decision-enabling evidence which can sit alongside randomized trial results in treatment guidelines and daily clinical conversations.

“Trials tell us what works under controlled conditions,” she explained. “Real-world evidence helps us understand how that actually translates into practice across broader patients, different settings, and over time.”

In the end, the goal is simple but profound. It is to close the loop between research, regulation, and care so that every patient’s experience improves outcomes for the next. “Patients are at the center of our mission,” Castellanos highlighted. “This data comes from real-world care, and it has to go back to improve that care for it to be meaningful.” As oncology innovation continues its breakneck pace, that continuous feedback loop may prove to be the most powerful tool of all.

Author

Steven Chung

紅杏枝頭春意鬧固然是眾人所追求的風景，但我願意說，青蘿拂行衣更能湧生感嘆。 願在海量晦澀的生醫資訊中，為您點亮一座心燈塔。鐘煌翔