Search the first three EU Joint Clinical Assessment (JCA) reports for real-world evidence, and you won't find much. All three lean almost entirely on comparative effectiveness drawn from randomized controlled trials. That's a striking data point for anyone who has spent the past two years hearing that real-world evidence would be central to how Europe assesses new oncology medicines.
I recently joined Karen Facey, Senior HTA Advisor at the Universities of Oxford, Utrecht, and Edinburgh and RWE4Decisions, and Hillary Keenan, Senior Director, Epidemiology, Global Evidence and Outcomes at Takeda, for a webinar on building an HTA-ready evidence strategy for European oncology. Watch the recording on-demand.
If you only have a few minutes, these are the insights that stuck with me.
Insight #1: Real-world data is playing a different role than we expected.
“There is no real-world data. That's the bottom line.” That was Karen's read on the first three JCA reports, and it's a useful corrective to some of the assumptions our industry has been making. But as Karen pointed out, the JCA itself is a summary of the evidence, not the reimbursement decision. That decision still happens at the member-state level, where countries doing economic evaluation have long incorporated real-world data in some form.
Hillary made a related point, one our own audience poll that day seemed to confirm, that the readiness gap isn't really about whether RWE matters, but whether teams have fit-for-purpose data lined up before the compressed submission window opens. What we're watching right now is the very early stage of a process that's still figuring out where RWE belongs, not evidence that it doesn't belong at all.
I think about this the same way I think about the gap between trial populations and the patients I treat as a clinician: what a regulatory document says on paper and what happens once real patients and real health systems get involved are two different things. National uptake, shaped by local real-world data, is where I expect this story to actually play out.
Insight #2: Heterogeneity is reshaping how PICOs get built.
One reason RWE may be light in these early reports is structural. Care varies so widely across the EU's 27 member states that assessors are dissecting broad submissions into much narrower subpopulations, each requiring its own comparator. Karen described early JCAs where a single indication was split into seven or eight distinct population-intervention-comparator-outcome combinations, while a submission with one clean, well-aligned PICO moved through far more smoothly.
This is exactly where connected, harmonized real-world data earns its place. Care pathways, terminology, even how labs and outcomes get recorded, differ by country and by care setting, whether that's community oncology, an academic center, or a larger hospital system. Getting ahead of that fragmentation means standardizing data models and definitions well before a dossier is due. It's one of the reasons global interoperability matters so much to me: it's not an abstract technical goal, it's what lets us answer a real HTA question with confidence across very different healthcare systems.
Insight #3: “As early as possible” is a planning requirement.
Every one of us on the panel converged on some version of this: evidence planning has to start earlier than most teams currently start it. I think about this in terms of mapping PICO feasibility against available data during trial design itself, not waiting until the dossier stage to find out where the gaps are. Hillary pushed the point further upstream, into comparator arm selection, arguing that teams need to understand real-world standard of care early enough to know whether trial data alone answers the HTA question or whether it needs real-world evidence to supplement it.
Karen added something critical: the lead time required simply to access and assess a new real-world data source. That ramp-up, plus an honest gap assessment of what evidence you have versus what you'll be asked for, has to be built into the timeline from day one. She'd like to see this go even further with class-of-therapy conversations (e.g., gene therapies as a category) happening well before any individual product reaches assessment.
Insight #4: AI is earning trust for the easy jobs, while the hard jobs are still being negotiated.
Our discussion on AI evaluated both sides of the AI narrative. Karen described real movement in HTA bodies' comfort with AI over the past year, mostly around what she called "augmented intelligence": efficiency gains, meeting notes, the more mundane operational lift. The open question is AI's role in evidence generation itself, and the EU's HTA Coordination Group recently published principles built around three ideas: accountability, transparency, and reporting. Hillary emphasized that it isn’t just about keeping a human “in the loop,” it’s about keeping a human “at the helm.”
That framing lines up closely with how we already think about validating real-world evidence, and it's the same logic behind the quality frameworks we've built for AI-enabled data curation at Flatiron. We developed our VALID framework specifically to compare AI-extracted data against human-abstracted data, using rigorous benchmarks and replication analyses, because the standard was never whether a human is somewhere "in the loop." It's whether someone is accountable when that output eventually informs whether a treatment gets reimbursed.
Insight #5: A credible comparator strategy starts with the real-world standard of care, not just the regulatory sign-off.
Hillary’s approach is to survey the real-world data early enough to know what your comparator arm actually looks like in practice, use it to sanity-check the guideline-based assumptions in your evidence plan, and pair it with early scientific advice from HTA bodies to confirm you’re on the right track before you’re locked into a design. Karen added that even when a comparator gets sign-off from regulators, payers will push back if it doesn't reflect local care, so assessing real-world data from different member states matters in addition to what your local teams are telling you. I'd add that this only works if the underlying data can hold up under scrutiny: transparent methodology, recent data, and consistent data models across geographies, so that pooled or cross-country analyses reflect a harmonized process rather than data that happened to be curated in different ways.
None of what we discussed suggests real-world evidence is losing relevance in European HTA. If anything, it suggests the opposite: these first JCAs are surfacing exactly where the gaps are, and national-level decisions, still shaped heavily by local real-world data, are where I expect those gaps to close. That's consistent with what I've seen across Flatiron's global oncology network, where connected, harmonized data across geographies is often what makes a comparator strategy defensible once it reaches the country level.
Those are my five takeaways, but there's a lot more in the full conversation, and I'd encourage you to watch it when you have time. I'm grateful to Karen and Hillary for such a thoughtful discussion, and to everyone who joined us live with such sharp questions. We're still early in this process, three JCAs in, but that's exactly why it's worth engaging now, while the model is still being built rather than fixed in stone.
.png)