Presentation (CTI)

OBJECTIVES: The first Joint Clinical Assessment (JCA) report on tovorafenib, marked the first JCA under Regulation (EU) 2021/2282. As tovorafenib was investigated in a single-arm Phase 2 study (FIREFLY-1) only, no direct randomized evidence was available, requiring an unanchored (Matching-Adjusted) Indirect Comparison (MAIC). Within the JCA framework, the assessment of uncertainty of indirect evidence is central: where evidence derives from indirect sources, HTA bodies must characterize the magnitude and direction of potential bias, raising fundamental questions on acceptability of ITCs under EU HTA.
METHODS: Critical appraisals of the submitted MAIC were extracted from the JCA report (v1.0, 2026), the EU Methodological Guideline for Quantitative Evidence Synthesis (HTA CG, 2024) and IQWiG General Methods (v8.0, 2025). HTA reports from G-BA (six procedures, 2019-2025) and HAS (four procedures, 2021-2025) were analyzed with AI support to document how both agencies handle similar methodological issues.
RESULTS: The JCA assessors identified ten structural criticisms: unanchored design, residual and unmeasured confounding, incomplete effect modifier identification, inconsistent variable selection, low effective sample size, divergent endpoint definitions, absence of multiplicity-controlled testing, unverifiable PFS censoring assumptions, and erroneous sensitivity analyses. Both agencies apply similar criteria: G-BA does not accept unanchored MAICs to quantify added benefit. HAS classifies unanchored MAICs as low level evidence, often resulting in limited ASMR ratings or no designation.
CONCLUSIONS: Criticisms from the JCA report reflect a stable cross-agency consensus: unanchored MAICs without full comparator IPD carry inherently high uncertainty and are insufficient to demonstrate robust benefit. For drugs with single-arm pivotal trials, early IPD access, anchored designs, and pre-specified effect modifier identification is critical under EU HTA. Where these conditions cannot be met, real-world evidence may offer a meaningful alternative, provided methodological rigor and population and endpoint comparability are ensured.