Transportability in Health Technology Assessment

The question is no longer “Is the evidence strong enough?” but “Will it work for the population making the reimbursement decision?”

For years, methodological innovation focused on generating robust comparative effectiveness evidence using indirect treatment comparison (ITC) methods such as network meta-analysis (NMA), matching-adjusted indirect comparison (MAIC) and simulated treatment comparison (STC). These approaches enabled comparisons between treatments that had never been evaluated head-to-head.

Can evidence generated in one population (e.g., a clinical trial) be trusted to inform decisions in another (e.g., a country-specific population)?

That question—transportability—was the central theme of a recent University College London (UCL)/Thermo Fisher Scientific health economics and market access symposium bringing together statisticians, health economists, health technology assessment (HTA) experts and industry leaders. Across the presentations and panel discussion, one message emerged consistently: Transportability is rapidly becoming one of the defining methodological challenges in modern HTA.

HTA decisions are always local—even when evidence is global

Randomized clinical trials are increasingly multinational; HTA decisions are not. Every reimbursement decision asks a fundamentally local question: Is this evidence applicable to patients treated within our healthcare system?

Answering this question becomes challenging because economic models routinely combine evidence from multiple sources, each representing different patient populations. As George Bungey, MSc, research scientist, PPD Evidera health economics and market access, Thermo Fisher Scientific, highlighted during the symposium, economic models already rely on substantial assumptions about evidence generalizability—even though those assumptions are rarely stated explicitly. Two distinct concepts deserve greater attention:

• ITC transportability—whether relative treatment effects estimated through ITCs apply to the target economic model population.
• Economic model transportability—whether all (or the most critical) evidence sources used in the model appropriately represent the target population.

The hidden assumptions inside every economic model

Many HTA submissions implicitly define their target population using the pivotal clinical trial, yet reimbursement decisions rarely align with trial populations. Patients in routine clinical practice often differ from those enrolled in pivotal trials, and comparator studies included in ITCs frequently represent different populations. Consequently, treatment effects are routinely transported across populations—sometimes explicitly, but often implicitly. Experts at the symposium argued that these assumptions should be made transparent.

A recent National Institute for Health and Care Excellence (NICE) appraisal of tirzepatide for obesity provides a compelling illustration. The economic model drew evidence from multiple clinical trials and other sources, with a NMA used to estimate the comparative efficacy of tirzepatide and relevant comparators. The NMA assumed that estimated relative treatment effects across included trial populations were transportable to the population represented in the economic model. During the appraisal, additional subgroups of clinical interest were identified. However, subgroup-specific efficacy estimates were not always available, meaning that treatment effects estimated in broader trial populations were assumed to apply to these narrower decision populations. The NICE committee ultimately accepted these assumptions, but the appraisal illustrates an important point: Transportability assumptions are often necessary, and they should also be explicit.

Transportability begins with the indirect comparison

Before considering estimands or statistical adjustment methods, symposium speakers emphasized an important conceptual point: every indirect treatment comparison (ITC) already relies on assumptions about transportability. Jack Sahakian, PhD, vice president, statistical methodology and strategy, PPD Evidera health economics and market access, Thermo Fisher Scientific, illustrated this using a simple pairwise indirect comparison. Traditional ITCs and network meta-analyses assume that relative treatment effects estimated across different studies may be applied to the target population (or any relevant population) for decision making. This assumption becomes more difficult to justify when treatment effects are modified by patient characteristics. Both MAIC and STC methods use individual patient data to adjust treatment effects so that comparisons are made in a common population rather than comparing aggregate trial results. In doing so, they explicitly acknowledge that evidence often needs to be transported from one study population to another before valid comparisons are made.

However, Dr. Sahakian highlighted an additional and often overlooked step. After adjustment, the estimated treatment effect is specific to the comparator trial population. If the objective is to make inference for the pivotal trial population, or any other decision-relevant population, an additional transportability step is required. This final step relies on the shared conditional effect modifier assumption (SEMA), which assumes that treatment effect modifiers influence competing treatments in a comparable way. When this assumption holds, adjusted treatment effects may be transported across populations. When it does not hold, even sophisticated adjustment methods may produce estimates that are inappropriate for the decision population. Further details may be found in this publication.

Transportability begins with the estimand

Historically, discussions around transportability have focused primarily on identifying and adjusting for effect modifiers. The symposium speakers suggested that this perspective is necessary but not sufficient. Before deciding which variables require adjustment, analysts must first define exactly which treatment effect they are attempting to estimate and for which population. As Antonio Remiro-Azócar, PhD, principal scientist, Novo Nordisk and honorary research fellow, University College, London argued, transportability begins with the estimand rather than with the adjustment method itself. Marginal treatment effects are often most relevant for population-level HTA decisions, whereas conditional treatment effects may be more appropriate for subgroup analyses or individual-level economic models.

Importantly, even in the absence of effect modification, marginal treatment effects may differ across populations because they depend on the distribution of prognostic variables. Consequently, two studies with identical conditional treatment effects may still produce different marginal treatment effects, reinforcing the importance of defining both the estimand and the target population.

ML-NMR represents an important methodological step forward

David Phillippo, PhD, research fellow in evidence synthesis, Bristol Medical School (PHS), presented multilevel network meta-regression (ML-NMR) as a natural evolution of population-adjusted indirect comparison methods. Unlike traditional MAIC or STC approaches, which are generally limited to pairwise comparisons and a single target population, ML-NMR analyses complete evidence networks and can estimate both marginal and conditional treatment effects for multiple decision-relevant populations within a single analytical framework.

As Dr. Phillippo noted, larger evidence networks and multiple target populations are becoming increasingly common, particularly with implementation of the European Joint Clinical Assessment (JCA). Rather than conducting multiple independent analyses for different jurisdictions, a single ML-NMR analysis may estimate treatment effects for the comparator trial population, the pivotal trial population and other decision-relevant populations, allowing direct comparison across populations while making transportability assumptions more transparent.

Methodological advances also have implications beyond evidence synthesis. As Dr. Phillippo highlighted, more sophisticated evidence synthesis methods also require more sophisticated economic models. For example, population-adjusted analyses may produce time-varying marginal treatment effects, challenging traditional modelling approaches based on constant hazard ratios. This raises broader questions about whether economic modelling methods should evolve alongside advances in evidence synthesis. Rather than viewing improved evidence synthesis as an isolated statistical advance, the symposium concluded that statistical methodology and economic modelling must evolve together.

Transparency may become just as important as methodology

Perhaps the most practical recommendation emerging from the symposium was not a new statistical technique; it was greater transparency. The symposium speakers concluded that future HTA submissions should explicitly document target populations, evidence sources, transportability assumptions and the rationale for extrapolating evidence across populations. Making these assumptions explicitly would strengthen the transparency and credibility of future HTA submissions.

Looking ahead

Transportability is unlikely to remain a niche methodological topic. Multinational clinical development, increasing reliance on indirect treatment comparisons, expanding use of real-world evidence and implementation of the JCA will all increase the need to demonstrate that evidence is applicable to local decision populations. Ultimately, the challenge for HTA is no longer simply producing robust comparative evidence—it is producing evidence that is credible, relevant and decision-ready. In that sense, transportability is rapidly becoming one of the defining principles of evidence credibility in modern HTA.

Panel discussion: from methodological innovation to practical implementation

The panelists focused on the practical implementation of transportability within HTA. Several practical themes emerged:

Payer perspective, JCA and HTAs outside of NICE

Martin Parkinson, BSc, executive director, EU HTA regulation lead, PPD Evidera health economics & market access, Thermo Fisher Scientific, framed the discussion from the perspective of healthcare payers, arguing that the primary responsibility of HTA agencies is to maximize health gain for their own populations using finite healthcare resources. From that perspective, transportability is not an optional methodological refinement but a necessary component of evidence assessment.

The discussion also highlighted the implications of the JCA, which aims to establish a common clinical assessment across Europe. However, reimbursement decisions remain national responsibilities, with each Member State defining its own comparators, treatment pathways and target populations. The panel noted that although JCA will increase the need for transportability analyses, wider adoption will require greater methodological guidance and capability within HTA agencies.

The role of data, synthetic data and RWE

A recurring theme was that many transportability methods exist because analysts are attempting to compensate for limitations in the available evidence rather than because those methods represent the ideal analytical solution. Greater availability of patient-level information, more complete reporting of baseline characteristics, and richer descriptions of covariate distributions could substantially reduce uncertainty in transportability assessments.

Panelists also discussed the potential role of synthetic datasets, which may provide a practical compromise between protecting patient privacy and enabling more informative evidence synthesis. Looking further ahead, the discussion emphasized that real-world evidence should not be viewed solely as a source of data at market entry. Instead, observational data collected during routine clinical practice could help evaluate whether treatment effects observed in trials remain applicable after reimbursement, allowing transportability to become an ongoing process rather than a one-time assessment.

Complexity is increasing; so should transparency

The panel speakers also challenged the perception that increasing methodological complexity should be viewed as a barrier to implementation. Healthcare reimbursement decisions involve substantial opportunity costs, and the relevant question is whether transportability methods produce evidence that is more credible and relevant for the populations in which decisions are made.

Overall, the panel reinforced the central message: transportability is no longer simply an issue for indirect treatment comparisons, but a unifying concept linking evidence generation, evidence synthesis, economic modelling and healthcare decision making.

This article summarizes presentations and panel discussions from the UCL–Thermo Fisher Scientific Symposium on Transportability in HTA (July 6, 2026), featuring contributions from:

• George Bungey, MSc, research scientist, PPD Evidera health economics and market access, Thermo Fisher Scientific
• Jack Sahakian, PhD, vice president, statistical methodology & strategy, PPD Evidera health economics and market access, Thermo Fisher Scientific
• David Phillippo, PhD, research fellow in evidence synthesis, Bristol Medical School (PHS)
• Antonio Remiro-Azócar, PhD, principal scientist, Novo Nordisk and honorary research fellow, University College, London
• Gianluca Baio, PhD, professor of statistics and health economics, University College London
• Martin Parkinson, BSc, executive director, EU HTA regulation lead, PPD Evidera health economics & market access, Thermo Fisher Scientific