OBJECTIVES: This study aims to describe how health economic analyses were assessed in France to appraise COVID-19 treatments in a context of parametric and structural uncertainty.

METHODS: Descriptive Review of economic opinions from the French HTA body (HAS), published from January 2020 to June 1^st 2023 on COVID-19 treatments. A focus on uncertainty management is made, considering epidemiological context and ability of modelling to quantify and explore uncertainty to guide decision making.

RESULTS: Five economic opinions published by HAS on COVID-19 treatments were analyzed concerning 3 treatments (including prophylaxis) to prevent severe forms of COVID-19.

Economic modeling relied on 1 Markov model, and 3 decision tree+Markov models, all modeling ambulatory and hospital care for COVID-19, then long-COVID and death. No one modeled inter-individual transmissions nor reinfections.

HAS’ economic committee wasn’t able to conclude on efficiency of the products due to global major uncertainty. Global major uncertainty objected were based on the following:

• • Systemic uncertainty due to epidemiological context: number and severity of epidemic waves, clinical data temporality and transposability, evolution of level of care through time and loss of efficacy of treatments.
  • Modelling and methodological uncertainty due to assumptions made, considering lack of data and knowledge in the COVID-19 context.

However, the committee did not invalidate the methodology for any of the manufacturers' analyses, neither for cost-utility analyses (N=4) nor budget impact analyses (N=4), and for 2 opinions, analyses were qualified as “well documented”.

CONCLUSIONS: While modelling was used by Ministry of Health in the early COVID-19 pandemic to guide public health measure in France, authorities now show stronger risk-aversion towards economic models’ uncertainty. HAS’ clinical committee was able to conclude on COVID-19 treatments’ efficacy, and other HTA agencies such as NICE concludes on efficiency whereas HAS’ economic committee did not. However real-world data on COVID-19 are available to tackle parametric and contextual uncertainty.