The First Three EU Joint Clinical Assessments: What the Methods Reveal

Saurabh (Rob) Aggarwal, Sushil Kumar, Ozlem Toploaglu, NOVEL Health Strategies

Executive summary

The EU Health Technology Assessment Regulation has moved from guidance to practice. Three Joint Clinical Assessments are now published, and together they establish how the method operates in the real world. The central finding is straightforward. A JCA does not create differences between products. It exposes the evidence design behind them. Where a randomized trial matched the questions member states asked, the assessment covered every question. Where the evidence was single-arm, most questions went unanswered. For any organization planning a European launch, the implication is direct: the trial, not the submission, determines the outcome, and it must be designed years ahead against the questions the assessment will pose.

The JCA is a clinical instrument, and only a clinical instrument

A JCA delivers a single, pan-European scientific assessment of a medicine's relative clinical effectiveness and safety. Its scope is deliberately bounded. It addresses the clinical domains alone. It sets no price, assigns no added-benefit rating, and issues no recommendation. Those judgments remain with each member state. The assessment establishes what the evidence demonstrates and where it falls short, in a standardized form, so that 27 national authorities can each reach their own decision from one shared clinical analysis.

The framework rests entirely on PICO: Population, Intervention, Comparator, Outcomes. The PICOs define the questions the evidence must answer. The process is anchored to the EMA regulatory timeline, so the clinical assessment proceeds in parallel with approval.

The JCA methods pipeline

Three assessments, three distinct evidence architectures

The three published reports do not share a common structure. Their evidence bases differ, and the methods follow directly from that difference.

The first assessment, tovorafenib (Ojemda) in paediatric low-grade glioma, was conducted by NCPE (Ireland) with IQWiG (Germany) and rested on a single-arm trial. The second and third, lurbinectedin (Zepzelca) and tarlatamab (Imdelltra) in extensive-stage small-cell lung cancer, were led by IQWiG with INFARMED (Portugal) and NNGYK (Hungary) respectively, and both rested on randomized controlled trials. IQWiG has served as assessor or co-assessor on all three, which locates the methodological centre of gravity precisely.

The three JCAs at a glance

Method one: scoping and consolidation determine the burden

The assessment begins with scoping, and scoping is where the first methodological pressure emerges. Each member state defines its PICOs through EU-coordinated surveys, reflecting its own standard of care. Those national inputs are then consolidated into a single assessment scope. Consolidation is the decisive step, and it remains the least clearly defined in practice.

The consequence is that the number of questions can escalate sharply. Comparators differ across countries, and biomarker and histology subpopulations multiply the questions further. The published evidence confirms the variability: across 35 scoping exercises, participation averaged seven countries and eight consolidated PICOs. The tail is long. A single scoping simulation for a first-line immuno-oncology therapy in metastatic non-small-cell lung cancer produced 67 PICOs across 25 countries, driven predominantly by subpopulations.

The first three assessments demonstrate the range concretely. Lurbinectedin required a single PICO. Tarlatamab required seven. Tovorafenib defined eight. A span from one to eight, across three orphan indications, confirms the conclusion: consolidation, more than any other step, determines how demanding an assessment becomes.

PICO scoping and the consolidation problem

Method two: the comparison toolkit, from robust to fragile

With the PICOs fixed, the core methodological task is comparison, estimating the relative effect against each specified comparator. This is where the three reports diverge most, because the method available depends entirely on the evidence supplied.

The strongest instrument is the direct head-to-head randomized trial, where the trial comparator matches the PICO comparator. Lurbinectedin is the model case: one PICO, one trial, one direct estimate, with the trial comparator aligned to the specified comparator.

The next tier applies subpopulations of a single pivotal trial to multiple PICOs. Tarlatamab followed this approach, drawing on subpopulations of the randomized phase 3 DeLLphi-304 study, which compared tarlatamab against chemotherapy with overall survival as the primary endpoint. The method is more demanding, because each subpopulation must sustain a valid comparison.

Where direct evidence is absent, the toolkit turns to indirect comparison, and here the methods grow more complex and less secure. The tarlatamab assessment employed the full range: network meta-analyses, anchored indirect comparisons, and unanchored matching-adjusted indirect comparisons, selected by the data available for each scenario. Anchored comparisons rely on a common comparator and are the more robust indirect option. Unanchored MAICs have no such link and must reweight patient-level data to match on both prognostic factors and effect modifiers.

The weakest configuration, the unanchored MAIC as the sole evidence, defined the first assessment. Tovorafenib's single-arm trial permitted no direct comparison, and the one comparison attempted relied on an unanchored MAIC with a severely reduced effective sample size and no overall-survival data.

The comparison-methods hierarchy

The certainty gap the method has not resolved

A material methodological gap runs beneath the entire framework, and it bears most heavily on the indirect comparisons the JCA increasingly requires. Certainty of evidence is intended to be graded, but the established instruments were built for trials, not for reweighted indirect comparisons. There is no risk-of-bias tool specific to MAICs, and without one, GRADE cannot be applied; GRADE offers no method for rating certainty from a MAIC. In one review of oncology MAICs, only 2.6% met all reporting criteria.

This is the central tension in the method. The framework demands comparative evidence for every PICO, which drives developers toward indirect approaches when trials cannot cover every comparator. The discipline has no agreed means of grading the confidence those approaches warrant. The EU practical guideline on direct and indirect comparisons imposes structure, requiring rigorous similarity assessment and justification of any proxy for a missing effect modifier, but the underlying grading gap persists.

Evidence design determined coverage

Considered together, the three assessments establish the governing conclusion: the JCA exposes evidence design; it does not compensate for it.

Tovorafenib brought single-arm data, and seven of eight PICOs could not be assessed. This is not necessarily a planning failure. In an ultra-rare paediatric cancer, a randomized comparator arm may be neither feasible nor ethical. The method, however, is unforgiving of that constraint. It poses the comparative question regardless, and records the gap where the answer cannot be supplied, transferring the burden to national assessment.

Lurbinectedin and tarlatamab brought randomized evidence, and both covered every PICO. The second and third assessments provide comparative evidence across all member-state PICOs, in contrast to the first, which produced usable evidence for one of eight. The determinant is not the disease or the sponsor. It is the trial design and comparator strategy underpinning each dossier.

Coverage tracks evidence design

Implications for evidence strategy

Three conclusions follow for European evidence planning.

First, the primary determinant of a clean JCA is a randomized trial whose comparator matches the comparators member states will specify. Lurbinectedin demonstrates the return on that alignment.

Second, where a single trial cannot cover every PICO, the fallback is a sequence of indirect comparisons, and each step from anchored to unanchored adds fragility and invites greater uncertainty. Anticipating the likely PICOs and testing them through Joint Scientific Consultation is therefore a trial-design decision, taken years in advance, not a submission activity.

Third, for single-arm and orphan programmes, the realistic position is that the method will expose the evidence gap irrespective of the strength of the underlying signal. The evidence-generation plan must account for that from the outset.

Outlook

Three assessments constitute a narrow base, and only tovorafenib has been examined in depth in the public literature. The two small-cell lung cancer reports confirm that the framework can address a full set of PICOs when the evidence supports it, and they also reveal the escalating methodological complexity required to do so. The definitive test will arrive as the framework moves from orphan indications into common diseases, where comparators are more entrenched, subpopulations proliferate, and PICO counts approach the levels so far observed only in simulation. The methods established in these first three reports are the standard against which every subsequent assessment will be measured.

Sources: European Commission JCA reports on tovorafenib (Ojemda), lurbinectedin (Zepzelca) and tarlatamab (Imdelltra), 2026; EU practical guideline on direct and indirect comparisons; JMAHP analysis of JCA PICO scoping, 2026; Office of Health Economics, Lumanity, Fortrea and Becaris Publishing analyses, 2026; DeLLphi-304 (JCO/ASCO 2025) and IMforte trial evidence.

Report Lead Author Bios

SAURABH (ROB) AGGARWAL, PhD
Principal & Co-Founder, NOVEL Health Strategies

17+ years at NOVEL | 300+ US & global engagements | 160+ publications | Johns Hopkins| IIT

• Leads NOVEL’s work in healthcare strategy, evidence generation and life sciences innovation

• Prior experience at Sanford C. Bernstein, IMS/IQVIA and Parexel

• Author of 160+ publications, including the recurring Nature Biotechnology series What’s Fueling the Biotech Engine

• Work cited in 400+ publications and featured in The New York Times, Nature Medicine, Scientific American and other leading outlets

• Recipient of 2 AMCP Platinum Medals and recognition from AACR and the Sidney Kimmel Foundation

• PhD, Johns Hopkins University

OZLEM TOPALOGLU, PhD, MPH
Senior Director, NOVEL Health Strategies

10+ years at NOVEL | 5+ years at FDA | CMS | Cochrane | Johns Hopkins

• Leads work across evidence generation, epidemiology, HEOR and health policy

• Spent 5+ years at FDA, overseeing post-approval studies, epidemiology and literature reviews

• Prior experience at CMS and Parexel

• Cochrane co-author and contributor to the Tufts Cost-Effectiveness Analysis Registry

• Reviewed 100+ cost-effectiveness studies

• Recipient of 8 FDA awards

• MPH, Johns Hopkins Bloomberg School of Public Health | PhD, Georg-August University