When the first Joint Clinical Assessment for tovorafenib was published, I described it as an “empty grid”. Eight PICOs had been defined across the European assessment scope. Only one comparative analysis ultimately made it into the JCA: an unanchored MAIC of tovorafenib versus dabrafenib + trametinib in patients with a BRAF V600E mutation.

Germany has now published its initial benefit assessment of Ojemda®. And the evidence grid has moved again.

The assessment question changes

Tovorafenib is an orphan drug. Under the German AMNOG framework, additional benefit is considered established by marketing authorisation while the statutory revenue threshold is not exceeded. The G-BA therefore assesses the extent of additional benefit within the orphan-drug pathway.

Europe had constructed eight comparative PICOs. The current German pathway starts from a different legal question. That alone changes the role of the European evidence architecture. And it illustrates something we will probably see repeatedly under EU HTA: clinical evidence may be assessed jointly, while its downstream decision context remains national.

The evidence package moves as well

The JCA mainly relied on the two-year FIREFLY-1 data cut. For Germany, the company submitted a more mature three-year cut from June 2025. The G-BA used it because of the longer follow-up.

There is also a remarkable timing issue: when the German procedure started on 15 May 2026, the final JCA report had not yet been formally published. It therefore could not inform the initial German benefit assessment. The G-BA worked from the European JCA dossier submitted by the company, the German dossier, regulatory documents and the FIREFLY-1 study material.

So one of the first operational lessons of EU HTA is already becoming visible: European and national HTA may be sequential in legislation, while important parts of the assessment still run in parallel.

Then the only European comparative analysis disappears

This is probably the most striking part of the case: the unanchored MAIC versus dabrafenib + trametinib was the only relative comparison retained in the JCA. And even there, the European assessors were deeply cautious. They highlighted residual confounding, population differences, substantial statistical uncertainty and effective sample sizes ranging from only 5.81 to 14.26.

The German assessment takes another step: the G-BA does not use the MAIC for the benefit assessment. It states that MAIC analyses based on aggregated comparator data are generally not considered appropriate in this context, notes the very small FIREFLY-1 subpopulation available for the analysis, and points out that ORR and PFS do not address patient-relevant effectiveness outcomes for this assessment. So:

Eight European PICOs became one comparative analysis. That one analysis becomes zero in the German benefit assessment. That is quite a journey for the evidence.

And there is an intriguing institutional twist

Germany’s IQWiG was co-assessor of the European JCA, alongside Ireland’s NCPE. The German orphan-drug assessment, however, is performed by the G-BA itself. Both institutions live inside the same national HTA system. And Germany has traditionally taken a very strict view of indirect comparisons and confounding. The G-BA now applies a distinctly German evidentiary filter, including a strict methodological position on MAIC and patient relevance.

The institutional picture is therefore fascinating: at European level, IQWiG participates in an assessment that retains the MAIC, documents its results and dissects its uncertainty. Back in Germany, the G-BA excludes that same analysis from the national benefit assessment. This is best understood as a difference in mandate: the JCA assesses and describes the available relative clinical evidence, the German process asks whether that evidence is usable within a specific national benefit-assessment framework.

Same country. Different institution. Different mandate. Different fate of the evidence. Geography alone clearly does not explain downstream divergence.

The filtering continues at endpoint level

Tumour response does not survive the German assessment because its patient relevance remains unclear in the submitted operationalisation. PFS is also excluded. The assessment cannot establish how often progression represents clinically meaningful deterioration rather than radiological change alone. Visual acuity is considered patient-relevant in principle – an important distinction – yet uncertainties around testing procedures, standardisation and the analysed population prevent its use in the assessment. Patient-reported symptom and quality-of-life outcomes fall away because of insufficient response rates.

The summary table of the German assessment consequently becomes remarkably narrow: mortality and descriptive safety data from a single-arm study. The evidence has been filtered again. This time through national concepts of patient relevance, validity and methodological acceptability.

And the story may change again over the product lifecycle

Ipsen has already consulted the G-BA on the scenario in which Tovorafenib would enter a regular comparator-based assessment; for example after crossing the German orphan-drug revenue threshold. For that future situation, the G-BA defined an individualised therapy framework. Within the BRAF V600E subgroup, dabrafenib + trametinib is clearly a highly relevant established treatment option. That makes the European MAIC particularly interesting. Its signals are uncomfortable: PFS numerically favoured dabrafenib + trametinib, with an HR of 4.88 and an RMST difference of -6.64 months. Severe adverse events also numerically favoured the comparator, although all of these estimates carry the profound uncertainties highlighted by the JCA.

Today, that comparison has little role in the German orphan assessment. Later in the commercial life of the asset, the comparative question could become much more consequential.

That is why evidence architecture cannot stop at launch.

The downstream lesson

The first JCA is starting to show us where fragmentation may reappear: in legal pathways, institutional mandates, methodological acceptance, definitions of patient relevance, data maturity, procedural timing – and across the commercial lifecycle of the asset.

For companies, the familiar linear model – JCA first, national adaptation afterwards – already looks inadequate. My conclusion is more fundamental: evidence strategy has to be designed around decision systems, not submissions.

A clinical programme that satisfies the European assessment question may still lose key comparators, endpoints or analyses as it enters national HTA. And evidence that looks secondary at launch may become decisive later in the product lifecycle.

That changes how companies should prepare: the relevant question is no longer simply whether the JCA evidence package is robust enough. It is whether the evidence architecture can withstand several downstream tests at once – European PICOs, national definitions of patient relevance, country-specific comparator frameworks and future lifecycle events.

For tovorafenib, one example is particularly revealing: the comparative evidence against dabrafenib + trametinib has almost no role in the current German orphan assessment. Yet exactly that comparison could become strategically important if the product later enters a regular comparator-based AMNOG assessment.

Evidence that is dormant today can become decisive tomorrow. The architecture has to anticipate that future state from day one. Tovorafenib is an unusual first case – orphan drug, conditional approval, single-arm pivotal evidence and overlapping European and German timelines. That is precisely what makes it revealing.

Edge cases are often where architecture becomes visible.

And this case is not finished

Today’s German assessment is an important milestone, not the final word. The publication on 17 August opens the written consultation. Statements from the manufacturer and other eligible organisations and experts can be submitted until 7 September 2026. The AMNOG process then includes an oral hearing before the G-BA moves towards its final resolution, currently scheduled for early November.

The next phase will therefore test which parts of the initial assessment actually survive scrutiny.

I will be watching three things in particular: whether Ipsen can successfully challenge the methodological objections, whether clinical experts can strengthen the case for endpoints excluded from the initial assessment, and how the now-published JCA report enters the discussion on the way to the final decision.

I would resist predicting the eventual benefit category today. But I am prepared to make one broader prediction: EU HTA will reduce duplication at the level of clinical assessment. It will not eliminate strategic divergence downstream. Tovorafenib is giving us an early view of where that divergence can re-emerge – and how quickly the meaning of evidence can change as it moves between institutions.

The European assessment showed us the grid. The initial German assessment showed us how that grid can be reshaped.

In November, we will see which parts finally survive.

Stay tuned.