Report for the ICMRA workshop on data requirements to support COVID-19 strain updates

Background

COVID-19 vaccine antigen recommendations are made twice per year by WHO TAG-CO-VAC in April-May and December and around April-May by some regulatory authorities for region-specific use.  

Data required for recommendations include: 

  • epidemiological and virological surveillance,  
  • genetic and antigenic characteristics of SARS-CoV-2 variants (impact on antigenicity and immune escape),  
  • immune responses in animals or humans with the authorised antigen composition and with potential updated antigen compositions (to evaluate cross-reactivity against emerging variants),
  • estimates of vaccine effectiveness of currently authorised vaccines against circulating SARS-CoV-2 variants. 

Data and timelines for issuing recommendations were lastly discussed by ICMRA in February 2024 (Report from the ICMRA/WHO workshop on: Global perspectives on COVID-19 vaccines strain update | International Coalition of Medicines Regulatory Authorities (ICMRA)). 

While data in humans are limited due to time constraints, data in animals have been widely generated across vaccine developers both in naïve as well as primed mice. Studies in mice can provide important information to support regulatory decisions for COVID-19 vaccine strain updates. For example, studies in naive animals are valuable for demonstrating antigenic relationships among variants. However, the experience gathered in the past 5 years of vaccine updates has also highlighted some limitations. Variability in immunogenicity results, the influence of priming regimens, and lack of standardization among developers are some of the factors that affect interpretation of such data from mice.  

Conducting effectiveness studies to monitor vaccine performance in a real-life setting is critical, but data remain scarce. Due to the declining uptake of COVID-19 vaccines and the reduced generation of surveillance data, timely collection and interpretation of effectiveness data is challenging, especially for vaccines with a new or updated antigen composition.  

Given the complexity of the regulatory and scientific challenges identified above, regulators continue to highlight the need for global alignment with exchanges between authorities and consideration of the WHO TAG-CO-VAC recommendations. Considerations should also include non-clinical data requirements and interpretation, a research agenda to investigate alternative options, the potential use of artificial intelligence models to predict viral evolution for SARS-CoV-2, and a reflection on vaccine effectiveness studies and alternative data collection strategies. 

An ICMRA workshop was held on 11 May 2026, bringing global regulatory authorities and WHO to address the challenges ahead for COVID-19 vaccine antigen composition decision-making.  

Summary of the workshop

The Chair of the WHO TAG-CO-VAC shared the latest COVID-19 vaccine antigen composition recommendation: monovalent LP.8.1 antigen is the recommended antigen; other antigens such as XFG and NB.1.8.1 could also be used. The recommendation was informed by the current global epidemiological situation of COVID-19 and the circulation of SARS-CoV-2 variants, immune responses to SARS-CoV-2 variants and the performance of currently approved COVID-19 vaccines. The current surveillance data show low SARS-CoV-2 activity globally, with regional differences in variant dominance, largely XFG or NB.1.8.1, and the rise of the genetically and antigenically distinct variant BA.3.2. Pre- and post-vaccination sera from individuals immunized with LP.8.1 demonstrated significant increases in neutralizing activity against JN.1 and its descendant variants, including XFG, NB.1.8.1 and, to a lesser extent, BA.3.2.  

There are persistent challenges in the genomic sequence data, clinical immunogenicity data and vaccine effectiveness studies, all of which remain critical for future decision-making. The WHO TAG-CO-VAC will continue to meet twice per year to review evidence and make recommendations on COVID-19 vaccine composition, with flexibility to reconvene if the epidemiological situation warrants an interim update. 

The European Medicines Agency (EMA) and the USA Food and Drugs Administration (US FDA) discussed the value, limitations, and harmonization needs of animal models —both naive and primed mice— in supporting COVID-19 vaccine antigen composition updates. Naive mice models are widely recognized as valuable for assessing antigenic relationships and intrinsic immunogenicity of new vaccine antigens, providing clear, less biased data, though they do not reflect the complex immune background of human populations and its interplay with the immune responses to vaccines antigens. Primed (i.e. pre-immunized) animal models attempt to mimic human immune history but face challenges due to variability in study design (e.g. source of vaccine, type of platform and regimen for priming, interval to boosting and bleeding). Additionally, differences between animal and human immune systems, and the difficulty of replicating years of human exposure in a short animal study are difficult to reconcile. Regulators clarified that animal data are supportive for vaccine antigen composition recommendations rather than essential for approval of updated vaccines, while also encouraging the development of non-animal approaches and more human data.  

Representatives from IFPMA (including Moderna, Pfizer, GSK, HIPRA, Novavax, and Sanofi) presented their aligned views on animal studies, data requirements, AI-based prediction, and the need for improved coordination with public health agencies and among manufacturers. Manufacturers agreed that primary vaccination studies in naive animals are more informative for immunogenicity assessment, while booster studies in primed animals have diminishing value due to design complexity and limited real-world relevance. There is strong support for integrating non-clinical and clinical evidence, including real-world effectiveness data, though manufacturers noted challenges in timely access to such data and called for more consistent expectations for post-authorization clinical studies. Manufacturers are exploring AI tools for predicting viral evolution and immune escape, recognizing their potential but also the need for further validation and centralized development, and highlighted the importance of robust global surveillance despite declining sequencing. Industry representatives advocated for improved coordination with WHO and regulators, harmonized study designs, and public-private partnerships for conducting vaccine effectiveness studies, drawing lessons from the influenza vaccine process. 

A panel of regulators and industry representatives debated the future role of animal models, harmonization strategies for primed animal models, and the integration of human data and AI tools. Panellists agreed on the continued relevance of naive animal models for antigenic and immunogenicity characterization of individual variants, while recognizing that in isolation this data might exaggerate the advantage of an updated composition. The use of naïve infection animal models, coupled with naïve vaccinated models and antigenic mapping, was considered potentially helpful based on preliminary studies. However, the complexities involved in establishing infection models, including the potential lack of BSL3 facilities requiring the use of CROs, may prolong timelines, which would not be supportive of the rapid decisions that are required. Hence, it remains important to generate and assess different types of data, including data in primed animals despite the complexities.  

The characteristics of the BA.3.2 variant highlight the need to assess data in primed animals, since analysis of data from such models demonstrates the influence of past exposure to BA.3.2 cross-neutralisation. The value for harmonized priming regimens that attempt to reflect human exposure is particularly relevant, since the bulk of the adult human population was first exposed to either an ancestral or early omicron lineage strain. Using only recent vaccine variants for priming tends to reduce the breath of immunity, skewing the results in a way that it is not relevant for extrapolation to humans due to immune imprinting induced by the earlier strains. Thus, various streamlined priming models could be considered to avoid over-immunisation, by using either an agreed mix of both old and new variants, or some combination of infection followed by vaccination (or vice-versa), which should be investigated.  

There was broad support for collecting post-vaccination human sera annually from diverse age groups and regions to inform strain selection, with suggestions for collaborative networks to facilitate data sharing and analysis. The need for sera from trials and of the effectiveness data to retrospectively validate the predictive value of animal models was highlighted. Notably, industry expressed their willingness to further develop and validate AI-based prediction models for viral evolution and immune escape. This approach has retrospectively demonstrated a strong correlation between clinical data and reduced cross-neutralization. Thus, the use of computational tools might be considered in addition to updated booster regimen in mice, to support strain recommendation decisions in the future. The panellists highlighted the need to maintain strong genomic surveillance and vaccine effectiveness studies, despite declining resources, and discussed the potential for wastewater analysis and other supplemental data sources. The WHO Coronavirus Network (CoViNet), which aims to bring together surveillance programmes and reference laboratories to support enhanced epidemiological monitoring and laboratory (genotypic and phenotypic) assessment of SARS-CoV-2 and other coronaviruses of public health importance, could potentially further support these activities.  

Workshop participants agreed on the following points:

  • Harmonization of Animal Model Study Design: both studies in naïve as well as primed animals should continue to support regulatory decisions regarding strain recommendations. The use of primed models remains of value to indicate the direction for strain selection, but a greater effort to harmonisation of study design is needed. In particular, streamlined priming with a minimum set of vaccinations to keep the breadth as close as possible to what is seen in humans is essential. Studies should be conducted with different settings and strategies to evaluate the various options, contrasting the results in animals with the cross-neutralisation data obtained with sera from clinical trials, combined with results of post-approval effectiveness studies and infection studies of animals (see below). Vaccine availability for priming, priming regimens and intervals, heterologous versus homologous approaches, as well as the assays involved should also be considered.  
  • Exploration of Infection-Based Animal Models: Define a research agenda to explore infection-based animal models (e.g. hamster infection models) as a possible supplement to current vaccination animal models for antigenic characterization.  
  • Collection and Sharing of Human Sera for Cross-Neutralization Studies: Establish a collaborative clinical framework involving industry and other entities or trial networks (e.g. through a public private partnership) for annual collection and sharing of post-vaccination human sera across different geographies, and age groups to support cross-neutralization testing of novel variants by individual manufacturers. It is considered critical that this dataset be available by February or March each year for strain update decisions. Sera panels should be generated from standardised clinical trials conducted in young, adult and elderly groups, each including at least 25-30 individuals. Datasets from such standardised trials could be combined from each region, which should include COVInet, the WHO coronavirus laboratory network.
  • Evaluation and Validation of AI-Based Prediction Models: Continue development and retrospective validation of AI-based prediction models for SARS-CoV-2 evolution and immune escape, including validation against robust historical datasets in the clinical space.
  • Maintenance of Genomic Surveillance and Vaccine Effectiveness Data: ICMRA advocates for sustained investment in genomic surveillance (including wastewater analysis as useful supplemental dataset in the broader sense) and vaccine effectiveness data generation to inform timely and robust strain update recommendations.
  • Need for collaboration: Industry is encouraged to collaborate on the development of such streamlined animal models and the other points raised above, to bring forward the results to regulators in a collaborative and active way.