Indirect Treatment Comparisons in France: Why HAS Exposes the Limits of Multi-Market Evidence Strategies
For over a decade, market access teams have observed a persistent pattern: indirect treatment comparisons (ITCs) that satisfy the National Institute for Health and Care Excellence and align with the EU Joint Clinical Assessment often fail to demonstrate added benefit before France’s Haute Autorité de Santé (HAS). Recent cross-HTA research demonstrates that France is not being difficult it is consistently applying a coherent framework grounded in its distinct regulatory and clinical context. This report clarifies why HAS’s approach represents methodological rigour rather than institutional conservatism and what evidence teams must do differently to succeed in the French market.
France’s Coherent but Different Framework
Recent multi-country analysis reveals that HTA acceptance of ITCs varies significantly across European jurisdictions, but this divergence is not primarily statistical[1]. Rather, it reflects alignment with each system’s specific decision context. France does not reject ITCs due to methodological flaws; it rejects them when they fail to address the mandated question: Is this comparison valid within French clinical practice? This distinction is critical. While NICE and EU JCA optimise for decision coherence and cross-country harmonisation, respectively, HAS optimises for clinical reality within the French healthcare system. Consequently, France’s rejection of an ITC often reflects not statistical invalidity, but misalignment with the French population and treatment pathway[2].
The key insight is that validity must be defined relative to the target population, not merely the dataset. An indirect comparison may be statistically robust and globally coherent while remaining decision-irrelevant in France if the populations reflected in trial networks do not align with French prescribing behaviour or the patients actually treated within French clinical practice. This population-first approach represents a shift from abstract statistical validity toward pragmatic clinical applicability.
Target Populations: The Critical Failure Point
Recent methodological advances in matching-adjusted indirect comparisons (MAICs) clarify why population alignment matters. A MAIC calibrated to a multinational trial population may produce a credible estimate in aggregate, but if that population does not reflect French healthcare practice, the result becomes strategically useless[3]. This explains a common misinterpretation: “The ITC is statistically correct, but HAS does not accept it.” The precise reading is different—the ITC answered the wrong population question.
Market access teams typically define target populations from trial inclusion and exclusion criteria. HAS requires a prior step: defining the population that HAS is actually evaluating. This distinction is not semantic. French guidelines, treatment algorithms, and reimbursement eligibility criteria define a specific patient cohort. An ITC that does not anchor to this population, regardless of statistical rigour, fails the foundational validity requirement. Building France into the evidence network at the protocol stage, rather than conducting post hoc analysis, is essential. French comparators must be incorporated early. French treatment pathways must be reflected in the indirect comparison chain. When these elements are absent, even well-executed ITCs cannot compensate.
Transitivity Reframed: From Statistical Condition to Clinical Property
Classical network meta-analysis (NMA) treats transitivity as a statistical condition: similarity of populations, consistency of effect modifiers, and exchangeability across trials. HAS extends this significantly. In the French framework, transitivity is a property of the care pathway itself[4]. To satisfy HAS’s requirements, an ITC must demonstrate that treatment pathways are comparable within France, effect modifiers reflect local prescribing behaviour, and the indirect comparison chain mirrors real therapeutic sequencing. This means France asks not, “Is this network statistically valid?” but rather, “Would this comparison remain valid if it were conducted within French clinical practice?”
This reframing has profound implications. An ITC that satisfies statistical transitivity may fail clinical transitivity. For example, a network connecting Drug A to Drug B through Drug C in multinational trials might not be clinically transitive in France if French physicians rarely use Drug C in the relevant treatment position. The sequence matters. The populations matter. The real-world pathway matters more than the statistical elegance of the network.
Scientific Rationale: Risk of Bias in Network Evidence
Recent research on risk of bias in NMAs reinforces the legitimacy of HAS’s caution[5]. Even well-conducted networks are vulnerable to structural bias in comparator selection, heterogeneity masking, model-dependent effect inflation, and interpretive overreach. HTA decisions are not abstract statistical exercises; they are real-world allocation decisions under uncertainty with direct economic and clinical consequences. Given these vulnerabilities, HAS’s cautious approach reflects methodological realism rather than conservatism. The agency implicitly acknowledges that the uncertainty inherent in indirect comparisons, sensitivity to population definition, and structural bias risks create an upper ceiling for confidence in indirect evidence. Consequently, ASMR (Added Scientific Value in Medicines) grading reflects not arbitrary constraints but epistemological judgment: ASMR I–II requires direct comparative evidence; ASMR III may be possible with highly robust, well-aligned ITCs; ASMR IV–V defaults when alignment is weak[6].
Economic and Strategic Implications
The ASMR scale directly influences pricing outcomes through the Comité économique des produits de santé. More critically, France serves as a reference-pricing anchor for Spain, Italy, Belgium, and other markets. An ITC design failure in France is not a local issue; it is a multi-market revenue event. Products that fail to achieve meaningful ASMR ratings in France face pricing pressure across Southern Europe. This creates a strategic imperative: evidence design must align with French decision requirements from day one, not as an afterthought.
Practical Implications for Evidence Teams
The scientific and strategic requirements are now clear. First, define the target population by beginning with the population HAS is evaluating rather than the population available in trial datasets. Second, build France into the evidence network—incorporate French comparators and care pathways at protocol stage. Third, treat transitivity as local validation by explicitly demonstrating clinical comparability within the French system. Fourth, prefer anchored comparisons, which reduce assumption burden and increase interpretability. Finally, align evidence strategy with pricing strategy early, since ASMR is the mechanism through which evidence becomes economic value.
France’s approach does not fragment the European evidence strategy; it refines it. HAS exposes when an ITC is statistically elegant, globally coherent, and locally misaligned. In doing so, it signals a broader shift in HTA: the constraint is no longer data generation but ensuring that evidence precisely answers the decision each system is actually making. HAS is not an exception; it is an early signal of where HTA is moving. Understanding this framework transforms the challenge from “How do we convince France to accept our ITC?” to “How do we design evidence that answers France’s actual question?” The distinction is fundamental and demands strategic recalibration of evidence planning across multi-market programs.
References
[1] Macabeo, B., et al. (2024). Cross-country acceptance of indirect treatment comparisons in oncology: A multi-HTA analysis. Health Policy, 148, 104891.
[2] National Institute for Health and Care Excellence. (2023). Guidelines for health technology assessment. NICE.
[3] NICE Decision Support Unit. (2022). Technical Support Document 18: Population adjusted indirect comparisons. London School of Hygiene & Tropical Medicine.
[4] Dias, S., Welton, N. J., Caldwell, D. M., & Ades, A. E. (2010). Checking consistency in mixed treatment comparison meta-analysis. Statistics in Medicine, 29(7-8), 932-944.
[5] Sterne, J. A. C., Savović, J., Page, M. J., et al. (2019). RoB 2: A revised tool for assessing risk of bias in randomised trials. BMJ, 366, l4898.
[6] Haute Autorité de Santé. (2023). Guide for determining the added therapeutic value of medicinal products. HAS.