Real-World Evidence in Benefit Assessment: Where It Counts

Few terms are understood as differently in market access conversations as real-world evidence. For one side it is the answer to missing comparative data, for the other a methodologically secondary argument. Both are partly right, and the difference does not lie in data quality but in the point in the procedure at which the data is presented.

The hard core: application-accompanying data collection

Germany has created a distinct, binding place for routine-care data. Under Section 35a(3b) SGB V, the G-BA can oblige a company to collect data on use in routine care and evaluate it for a renewed benefit assessment where the evidence available at the time of authorisation is of only limited value for the assessment. The instrument is designed for constellations in which added benefit could not be quantified because of insufficient evidence, in particular for orphan medicines and advanced therapies. The procedure itself is set out in Chapter 5 of the G-BA Code of Procedure, which has been supplemented several times for this purpose. In addition, the authorisation to supply can be restricted under Section 35a(3b) sentence 2 SGB V, so this is not a voluntary companion program.

The first case illustrates the design well. For the gene therapy onasemnogene abeparvovec (Zolgensma), the G-BA obliged a company to run such a collection for the first time in 2020. The basis is a registry study built on the existing SMArtCARE registry, IQWiG developed the concept on behalf of the G-BA, and the final results are to be submitted to the G-BA by 2027 at the latest.

Two consequences that are often missing from planning. First, the collection relies on an existing registry, not on one built after the decision. A company without a suitable registry at authorisation is negotiating about data sources it does not control. Second, the period until reassessment is long. The data quality that will count is created in the years in between, not in the month of analysis.

Outside that point, the randomised trial remains the yardstick

For the regular assessment under Section 35a SGB V, randomised comparative evidence against the appropriate comparator therapy continues to take precedence. IQWiG's General Methods, currently version 8.0 of 19 December 2025, are the reference here; version 7.0 of September 2023 was the one that introduced the concept of application-accompanying data collection into the methods paper in the first place. Non-randomised data is therefore not worthless, but it carries the burden of proof for its own risk of bias. A registry that does not document treatment decisions or does not follow up losses cannot carry that burden.

In practice: demonstrating added benefit solely through routine-care data is the exception, not the rule. Planning for it means planning against the procedure.

At EU level the gap becomes more visible, not smaller

With the joint clinical assessment under the EU HTA Regulation the question sharpens, because the consolidated assessment scope brings together comparators from many member states. For oncology medicines, increasingly authorised on the basis of single-arm trials, only indirect and often unanchored comparisons remain. The European methodological guidelines, grown out of the EUnetHTA 21 work on direct and indirect comparisons and continued by the HTA Coordination Group, explicitly consider single-arm or non-randomised evidence as potentially insufficient for estimating relative effectiveness.

How differently authorities handle this has since been studied: a review of oncology procedures in Germany, France, England and Norway shows that external control arms can be accepted, but inconsistently and under conditions that are known in advance. That is a plannable requirement, not something to negotiate inside the procedure.

For devices and IVDs, real-world evidence is mandatory anyway

Manufacturers of medical devices and IVDs already run most of this data, only under a different name and for a different purpose. PMCF under MDR Annex XIV Part B and PMPF under IVDR Annex XIII Part B require the systematic collection of clinical data after placing on the market. On the reimbursement side, Section 137h SGB V adds the transmission of the state of scientific knowledge, including complete study data, to the G-BA, detailed further by the MeMBV.

The leverage lies in planning these data streams once so that they serve both purposes: the regulatory post-market obligation and the evidence requirement of the assessment. In practice they usually run separately, with different endpoints, and then neither collection answers the other question.

What to do now

• Clarify registry readiness before authorisation. Is there an established registry for the indication that could carry an application-accompanying collection? If not, building one is a multi-year undertaking and belongs in development planning.
• Define endpoints once, use them repeatedly. Patient-relevant endpoints of the assessment, the outcomes of the PICO questions and the metrics of the post-market obligations should come from one list.
• Anticipate risk of bias. Treatment decisions, follow-up, handling of missing values and the comparison group have to be documented before anyone asks to see the analysis.
• Plan data protection in. Registry data is health data. Legal basis, consent architecture and processing agreements co-determine whether an analysis can be used later.

Entourage supports pharma, biotech, MedTech and IVD companies in setting up routine-care data so that it holds up in the assessment procedure: reviewing the registry landscape for the indication, designing an application-accompanying data collection, preparing indirect comparisons and external control arms methodologically, and connecting PMCF and PMPF with the evidence requirements of benefit assessment.