OBJECTIVES: Health technology assessment (HTA) increasingly leverages real-world data (RWD) to support the evaluation of new technologies. This study assesses how RWD is used for decision-making by different HTA bodies.

METHODS: Technologies submitted to EMA/FDA between 2017-2021 using RWD to support regulatory approval, were identified from grey literature. Ten technologies, covering different disease areas and types of RWD were selected. The EU4+UK HTA reports of the selected technologies were reviewed and the acceptance of RWD was evaluated. To gain further insights, semi-structured interviews with six country-specific HTA experts (with prior experience in submission and assessment of new technologies) were performed.

RESULTS: According to the six HTA experts, RWD is commonly used to characterize disease epidemiology, treatment patterns (e.g., to inform added value assessments) and unmet need. Less commonly (rare diseases and, oncology), RWD may be considered to support efficacy with external control arms (ECAs). For ECA acceptance, demonstration of unfeasibility of direct comparison is necessary, additionally to appropriate study design and statistical analysis. ECA are more likely to be accepted in areas of high unmet need with limited alternatives (e.g., CAR-T cell therapies). Independently of the type of RWD, European HTA bodies consistently favor, representative, country and European RWD over US and rest of the world. The UK, Italy, France and recently Germany, can request RWD for technology re-evaluation. However, in the UK, expert interview, revealed that mature trial data weights more than RWD. Overall, RWD acceptance is context-dependent and heterogenous across countries, with the UK remaining more receptive to consider RWD, Italy relying on RWD to inform managed-entry-agreements, and France endorsing RWD to support technologies re-evaluations.

CONCLUSIONS: In the context of HTA decision, RWD remains valuable to characterize the disease background, however, its use to support efficacy claims remains limited to cases where direct comparison is unfeasible and needs to be methodologically robust.