Real-World Evidence in the EU-JCA: Where RWE Can Close Evidence Gaps and Where Its Limits Lie

Real-world evidence can contribute far more to the EU Joint Clinical Assessment than describing epidemiology and treatment patterns. RWE can support PICO scoping, identify prognostic factors and effect modifiers, enable external control arms, and provide additional information for evidence synthesis. At the same time, broader use of RWE increases the methodological requirements for data quality, confounding control, comparability, and analytical justification. For health technology developers, this creates an important planning question: not whether RWE is relevant to the JCA in principle, but which assessment question RWE is intended to address and whether the underlying data source, study design, and analysis are appropriate for that purpose. Current HTACG guidance only partially captures these different applications. At the same time, the first published JCA reports already show where non-randomized evidence and analyses based on such evidence are likely to face particularly close scrutiny.

Real-world evidence can serve three different functions in the EU JCA

RWE can help prepare the clinical question, support evidence synthesis, or contribute directly to comparative analyses in the JCA. These functions differ substantially in their methodological requirements. Sarri et al. map potential RWE applications across the EU HTA process, from Joint Scientific Consultation (JSC) and JCA through to subsequent national HTA procedures. For the JCA dossier, three main roles can be distinguished:

1. RWE as background information: epidemiology, disease progression, treatment patterns, and characterization of the relevant population.
2. RWE as a basis for and complement to evidence synthesis: identification of prognostic factors and effect modifiers, supplementation of baseline information, or support for more complex evidence networks.
3. RWE as a direct input into comparative analyses: for example, an external control arm for single-arm trials.

This distinction matters. A descriptive analysis of the treatment landscape raises different methodological requirements from an RWD-based estimate of a relative treatment effect. RWE should therefore not be treated as a single, homogeneous evidence category. Its specific intended use is what determines the relevant methodological requirements.

RWE can become relevant to PICO scoping before the JCA

Real-world data can help anticipate potential Member State PICO requirements before the final JCA assessment scope has been established. Treatment patterns, local healthcare data and epidemiological analyses can indicate which populations and comparators may be relevant across different European healthcare settings. This matters because the JCA must consolidate Member State requirements into one or more PICO questions. Differences in healthcare practice across Europe may therefore result in different comparator or population requirements. RWE is consequently relevant before an evidence gap becomes visible in the completed JCA dossier. It can already support the preparation of the clinical questions themselves.

RWE can characterize national comparators and treatment pathways

Claims data, electronic health records, registries and other RWD sources can describe which treatments are actually used in routine care and how treatment pathways differ between countries. Treatment-pattern analyses can, for example, show:

• which treatments are used in a specific line of therapy,
• how frequently individual comparators are used,
• which patient groups receive different treatments,
• how much the standard of care varies across countries.

This can be particularly relevant in rare diseases or indications without a clearly established European standard of care. RWE does not replace the formal scoping process. It can, however, help simulate potential PICO scenarios at an early stage and assess whether the planned clinical evidence is capable of addressing those questions.

RWE can identify prognostic factors and effect modifiers for JCA analyses

RWE can provide an important evidence base for systematically identifying prognostic factors and potential effect modifiers. This becomes particularly relevant when indirect comparisons or other adjusted analyses are required. Observational studies can, for example, investigate which patient characteristics are associated with prognosis or may modify the relative treatment effect. These findings can then inform the statistical analysis plan. HTACG methodology requires relevant prognostic factors and effect modifiers to be identified in a transparent and defensible way. Sarri et al. note that literature reviews, observational studies and clinical expertise can all contribute to this process. The first JCA reports illustrate why this matters in practice. An analysis of the first four published reports identified missing evidence for key variables, unclear selection processes and incomplete consideration of prognostic factors as recurring methodological friction points. RWE can therefore become relevant before the comparative analysis itself: as evidence supporting which variables should be included in an analysis and why.

RWE can describe disease and treatment patterns in the JCA dossier

One of the more direct applications of RWE is the characterization of the disease, population and treatment landscape. Epidemiological and descriptive studies can provide information that goes beyond the controlled setting of a clinical trial. Sarri et al. identify, among other applications:

• epidemiology and the size of the relevant population,
• symptoms and disease progression,
• prognosis,
• treatment patterns,
• population characteristics,
• unmet need.

These data can be particularly useful in rare diseases, where clinical trials may involve small and highly selected populations or where current European treatment guidelines are limited. Real-world treatment patterns can also help inform and justify comparator choices. They can show which treatments are actually used in routine practice and which combinations may, for example, contribute to an individualized treatment comparator. The descriptive use of RWE should, however, be clearly distinguished from the estimation of relative treatment effects. A data source may reliably describe routine care without necessarily being suitable for a causal comparison between two treatments.

RWE used as an external control arm faces particularly high JCA requirements

External control arms can provide a way to construct a comparator population outside a clinical trial, particularly for single-arm trials. Once RWE is used directly to estimate a relative treatment effect, methodological requirements become substantially more demanding. The central challenge is the absence of randomization. Differences between patient populations may influence the observed treatment effect. Relevant confounders, prognostic factors and effect modifiers therefore need to be identified and, where possible, accounted for analytically. A robust comparison requires consideration of questions such as:

• Are the populations sufficiently comparable?
• Are relevant confounders and prognostic factors available?
• Are the same endpoints measured in a comparable way?
• Are index dates and follow-up periods defined consistently?
• How are missing data handled?
• Which variables are included in the adjustment?
• How robust are the results under alternative assumptions?

The HTACG methodology reviewed by Sarri et al. sets a high bar for non-randomized evidence. For unanchored comparative-effectiveness analyses in particular, access to individual patient data is important because adjustment needs to be conducted at patient level. This makes data availability a strategic issue in its own right. A theoretically suitable RWD source provides limited value if relevant variables are missing, individual patient data cannot be accessed in time, or the population is not sufficiently comparable with the trial population.

The first JCA reports show the practical limits of RWE-based comparisons

The first published JCA reports provide practical indications of how closely non-randomized evidence and analyses based on it are scrutinized. Fragmented comparator evidence and the justification of prognostic factors are particularly visible areas of methodological challenge.

Tarlatamab illustrates the challenge of fragmented comparator evidence

The JCA for tarlatamab involved multiple PICO questions and different comparator scenarios. The webinar analysis describes a fragmented comparator landscape that required broader literature searches as well as the use of real-world data and observational registries for comparative analyses. This illustrates a broader issue: the more comparator requirements differ across Member States, the less likely it becomes that all questions can be addressed through direct randomized evidence. RWE can provide additional evidence in such situations. At the same time, methodological uncertainty increases. For non-randomized cohorts, a central requirement is to demonstrate that differences between populations do not materially bias the observed treatment effect. Sensitivity analyses and conservative scenarios therefore become important parts of the assessment.

Prognostic factors become a critical component of RWE-based analyses

The first JCA reports also show that the selection of prognostic factors and effect modifiers needs to be more than a technical modeling decision. It needs to be substantively justified. A statistical adjustment can only account for variables that are known and available in the underlying data. Unmeasured confounding therefore remains a fundamental limitation of non-randomized comparisons. It is not sufficient to include a long list of variables in a model. What matters is a transparent rationale showing:

1. which factors are prognostically relevant,
2. which variables may modify the treatment effect,
3. what evidence supports their selection,
4. which relevant variables are missing from the data,
5. how missing information may affect interpretation.

Well-designed RWE can therefore contribute in two ways: it can form part of a comparative analysis and provide evidence supporting the selection of the variables required for that analysis.

RWE can provide indirect inputs into JCA evidence synthesis

RWE does not need to serve as the comparator itself to influence a JCA analysis. Real-world data can also provide additional information used to synthesize randomized evidence. Sarri et al. describe several potential applications. RWE can, for example:

• inform the natural history of a disease,
• supplement baseline characteristics,
• contribute to reweighting analyses of RCT populations,
• provide prior information for Bayesian evidence synthesis,
• help connect otherwise disconnected evidence networks,
• support the assessment of surrogate outcomes.

These applications are methodologically distinct and cannot be evaluated generically. They nevertheless show that the role of RWE in the JCA extends well beyond the traditional external control arm. One example is the transferability of an RCT population. If the trial population differs from the population relevant to routine care, RWD can provide information about the distribution of relevant patient characteristics in practice. Under appropriate methodological conditions, these data can inform analyses assessing the generalizability of trial findings. RWE can therefore provide a bridge between highly controlled trial evidence and the populations for which HTA decisions ultimately need to be made.

Current EU HTAR guidance only partially captures the potential of RWE

According to Sarri et al., current HTACG methodology provides only limited specific guidance on broader RWE use in the JCA. The most developed guidance concerns non-randomized evidence used directly in comparative analyses. The methodological framework is less developed for other potential uses. The authors discuss, among other areas, the potential for further methodological development around:

• external control arms,
• target trial emulation,
• transportability analyses,
• advanced methods for integrating RWE into evidence synthesis,
• synthetic data.

These points represent the authors' assessment and recommendations. They are not additional HTACG requirements that have already been adopted. The gap between available RWD infrastructure and detailed JCA methodology is nevertheless notable. DARWIN EU and the European Health Data Space are creating substantial European infrastructure for the use of health data. At the same time, the methodological integration of these data into the JCA is not yet described in detail for all potential applications.

High-quality RWE for the JCA requires four conditions

The suitability of RWE for a JCA depends on more than the size of a database. Data quality, transparency, comparability and analytical robustness determine whether real-world data can generate credible real-world evidence.

1. Data quality determines the value of RWE

Claims data, EHRs and registries are often not collected primarily for the specific research question being investigated. This can create challenges around completeness, coding, measurement and clinical validity. Before conducting an analysis, it is therefore necessary to assess whether the required populations, exposures, outcomes, confounders and prognostic factors can be captured with sufficient quality.

2. Transparency and reproducibility support credibility

RWE analyses need to be documented so that assessors can understand the data source, study design, variable definitions and analytical decisions. The publication refers, among others, to standards such as STaRT-RWE, RECORD and RECORD-PE. Transparency is particularly important in a European assessment because the same analysis may be interpreted by assessors from different HTA systems.

3. RWD from different countries are not automatically comparable

Healthcare systems, coding practices, data collection and data access differ across Europe. A large German claims database therefore does not automatically represent care in France, Spain or other Member States. Conversely, multinational data sources may require substantial methodological harmonization before populations or outcomes can be analyzed jointly. The geographic origin of RWD is therefore not merely an operational issue. It directly affects the external validity of the resulting evidence.

4. Comparative RWE analyses require a defensible analytical rationale

The more directly RWE is used to estimate treatment effects, the more important confounding, exchangeability and sensitivity analyses become. Methods such as propensity score matching or weighting can address observed differences. They do not, however, automatically eliminate the limitations of a non-randomized data source or study design. Analytical sophistication cannot substitute for weaknesses in the underlying data or design.

RWE for the EU JCA needs to be planned early

RWE generation for a JCA can require several years of preparation. Sarri et al. recommend planning RWE early, preferably two to three years in advance. They also place repeated PICO simulation and additional evidence planning across product development, with more intensive preparation beginning at least around 1.5 years before the JCA submission. There are clear operational reasons for this. Access to RWD can take time. Data sources need to be assessed, protocols developed, variables operationalized and analyses conducted. Multinational projects also involve different data access arrangements, governance structures and privacy requirements. At the same time, only a limited period remains after the final JCA assessment scope has been established to adapt the evidence package. The first JCA procedures reinforce this challenge. Therefore work-at-risk evidence generation is based on an early PICO simulation as one possible strategy: evidence is prepared for plausible PICO scenarios before the final assessment scope becomes available. This approach is particularly relevant for RWE. A new RWD analysis can rarely be designed, approved, conducted and fully documented within a few weeks.

RWE needs to be reassessed for national HTA after the JCA

RWE generated or used for the JCA is not automatically suitable for every national HTA process. National HTA systems retain their own methodological requirements and assess evidence within their own healthcare context. Sarri et al. therefore describe a potential need to update, localize, supplement or re-analyze RWE after the JCA. Reasons can include:

• a national comparator differs from those represented in the JCA,
• the national target population differs from the European PICO population,
• local treatment patterns are relevant,
• additional subgroups are assessed nationally,
• national HTA bodies apply different requirements to non-randomized evidence,
• local baseline risks, costs or healthcare data are needed for economic evaluation.

Reuse of RWE is therefore not simply a question of technical availability. The critical question is whether the evidence answers the specific national assessment question.

RWE in the German delta dossier needs to address the German assessment question

In the German delta dossier, RWE may be particularly relevant where information is needed on the German target population, treatment patterns or additional evidence questions. The methodological requirements of the German early benefit assessment remain applicable. RWE can, for example, support epidemiological estimates, characterize German treatment patterns, describe the local healthcare context or investigate prognostic factors. Particular caution is required for comparative-effectiveness analyses. The fact that an RWE-based comparison formed part of the JCA does not automatically make that analysis suitable for the German benefit assessment. The first practical experience illustrates this point. The webinar analysis of the tovorafenib procedure notes that MAIC analyses used in the JCA were mentioned in the German procedure but were not presented as suitable evidence for the AMNOG assessment because they relied on aggregate study data. This illustrates the core delta dossier challenge for RWE: European evidence and national requirements are closely connected, but they are not identical. The key question is therefore not: Which RWE is already available from the JCA? The key question is: Which German assessment question should the RWE answer, and do the data source, study design and analysis meet the relevant methodological requirements?

What role RWE can play across the EU HTA process

· Phase · Potential RWE contribution · Key methodological question ·

· JSC · Evidence planning, data sources and early PICO simulation · Which RWE questions should be discussed with assessors early? ·

· PICO scoping · Treatment patterns, comparators and populations · Do the data represent relevant European healthcare contexts? ·

· JCA background · Epidemiology, disease characterization and treatment patterns · Is the data source representative and current? ·

· Evidence synthesis · Prognostic factors, effect modifiers and supplementary evidence · Are variable selection and assumptions sufficiently justified? ·

· Comparative effectiveness · External control arms and non-randomized comparisons · Are exchangeability and confounding adequately addressed? ·

· National HTA · Local treatment patterns, populations and additional analyses · Is European RWE transferable to the national question? ·

· German delta dossier · Epidemiology, local care and supplementary analyses · Does the evidence meet German benefit assessment requirements? ·

The different applications illustrate why a general statement about the “acceptance of RWE in the JCA” is of limited value. Requirements depend heavily on the role RWE is intended to play in the assessment.

What manufacturers should prepare for RWE in the EU JCA

RWE should not be treated as a late-stage solution for unexpected evidence gaps. Data access, study design and analytical requirements make early integration into the evidence strategy essential. Four points are particularly important:

1. Define the RWE question before selecting the data source. A large database is only useful if it captures the required population, comparators, outcomes and confounders with sufficient quality.
2. Link PICO simulation and RWE planning early. Plausible comparator and population scenarios can reveal where additional evidence generation may be required.
3. Identify prognostic factors and effect modifiers systematically. Their selection should be based on transparent evidence and should not emerge only after the analytical model has been chosen.
4. Assess JCA and national requirements together during design and analysis. This makes it easier to identify which RWE is likely to be reusable at European level and where national additions may be required.

The first JCA reports already underline an important point: advanced statistical methods can address some evidence gaps, but they cannot fully compensate for weaknesses in the underlying data or study design.