THE FIRST JCA REPORT: WHEN THE FRAMEWORK SETS AN IMPOSSIBLE TASK

June 19, 2026

Last week in Istanbul at the HTAi Annual Meeting, the first published EU JCA report was presented. The manufacturer welcomed it publicly. Parts of the rare disease community cited it as proof the framework is broken for rare paediatric oncology.

Having read the dossier and heard the reactions on the ground, here is my five cents: Ipsen's evidence generation was largely done right. The trap was the framework.

What the Evidence Landscape Offered

Tovorafenib received EMA conditional MA in April 2026 for relapsed/refractory paediatric LGG with BRAF alteration after ≥1 prior systemic therapy. FIREFLY-1 was a single-arm Phase 2 trial — the expected design for an indication with no standardised second-line treatment.

Before submission, Ipsen did the right things. The SLR confirmed no comparative data existed for ~75% of the trial population (BRAF fusions/rearrangements). An RWE feasibility study was conducted — one US database was explored, deemed unsuitable (insufficient R/R patients with BRAF alterations, inadequate follow-up), and the conclusion was correctly documented: no fit-for-purpose external control existed. And even if US RWE had been feasible, IQWiG's documented position is that US real-world data does not transfer to the EU healthcare context.

The honest conclusion from all three exercises was that comparative evidence did not exist. That is the evidence environment, not a submission failure.

The Structural Trap

The HTACG assigned 8 PICOs. Six had no comparative data anywhere — not because Ipsen failed to generate it, but because the disease landscape has none. Conditional MA was the right decision for patients. But it is a JCA liability: EMA says the evidence base is incomplete and approves anyway; JCA asks for comparative evidence to determine added therapeutic value. These point in opposite directions. For six PICOs the framework documented the gap and moved on.

For PICO 5 — the BRAF V600E subpopulation (~22 patients in FIREFLY-1) — a comparator study existed: Bouffet 2023, R/R dabrafenib+trametinib in BRAF V600 LGG, n=36. Ipsen attempted an unanchored MAIC. That was the wrong call.

The mathematics make this clear. Arm 1 had 13 V600E patients. After weighting: ESS = 5.82. Fewer than six effective patients. Had both arms been included (22 patients), Scenario 1 shows ESS = 11.54 — still far below the minimum for meaningful inference on PFS in oncology (50–100 events). There was also a baked-in endpoint incompatibility: Bouffet 2023 used modified RANO-HGG criteria (Minor Responses excluded); FIREFLY-1 used full RANO-LGG (Minor Responses included). This is a field-level problem from pre-harmonisation trial design — not Ipsen's fault, but it made any response rate comparison structurally invalid.

The correct response to PICO 5 was the same response given to the other six: declare infeasible, document the reasons — subgroup too small for MAIC inference, positivity assumption unverifiable, endpoint frameworks incompatible — and present FIREFLY-1 data transparently as single-arm evidence.

What Went Wrong

Instead, the comparison was attempted. And the execution gave assessors additional reasons to document conduct failures. Arm 2 efficacy data was withheld (halving the matching population and driving ESS from 11.54 to 5.82). Investigator-assessed PFS from Bouffet 2023 was omitted, flagged as selective reporting. The 3-year data cut (May 2025) was not database-locked when the JCA opened (June 2025) — a planning failure, not deliberate withholding, but the result was data assessors could not validate. The formal clarification meeting produced no minutes and no clear answers.

What the Framework Must Learn

The Regulation does not yet have implemented rare disease proportionality provisions. Eight PICOs applied to a heterogeneous ultra-rare paediatric indication where the mathematics of comparative assessment could not be satisfied is the framework operating at full specification against an evidence environment that cannot support it.

The fix is a feasibility gate before PICO assignment: check whether the mathematical preconditions for indirect comparison exist — sufficient populations, comparable endpoints, verifiable positivity — before asking the developer to produce evidence for them. If they fail, redesign the PICO. This case should be the precedent that drives that change.

THE CORE TAKEAWAY

The first JCA report is not a verdict on a developer. It is a collision between a framework built for comparative evidence and an indication where that evidence cannot exist — and neither side had a clear protocol for how to handle it.

Ipsen's error was not the evidence generation. It was attempting a comparison that the mathematics could never support, and executing it in a way that created an additional conduct record alongside the statistical failure.

The lesson for developers in the JCA pipeline is not to generate more comparative data. It is to know when declaring a comparison infeasible is the stronger position — and to make that argument proactively, with the structured rationale ready before assessors find it themselves.

KNOW WHEN NOT TO ATTEMPT THE COMPARISON. THAT IS THE ARCHITECTURE.

Navigating this unyielding framework requires a specialized architectural bridge. At Sur Access, we tailor-make your JCA strategy and run predictive PICO simulations—ensuring your clinical evidence is insulated against biostatistical audits, whether you are launching a conventional oncology blockbuster or an ultra-rare orphan technology.

Partner with an expert who has been in the room.