Résumé
A key question in evolutionary epidemiology is to determine differences in the conditions that may allow some mutant strains to spread in a population where a resident strain is already circulating. Evolutionary invasion analyses assume that the immunity is long-lasting for previously infected individuals making it difficult to study traits such as immune escape. We relax this last assumption and allow the environment faced by the mutant to fluctuate outside of any epidemiological equilibrium. We introduce an original two-strains non-Markovian model that accounts for realistic immunity waning and cross-immunity, inspired by the case of SARS-CoV-2 variants. We show that mutants with increased contagiousness or with some immune escape abilities are more likely to invade the population. We also show that the timing of the introduction of mutant strain in the population is key because it is associated with the population's immunisation status. Our results underline the importance of immune waning and non-equilibrium dynamics on infectious disease evolution.