Evaluating how a COVID-19 seasonal vaccination program performed might help to plan future campaigns. This study aims to estimate the relative effectiveness (rVE) against severe COVID-19 of a seasonal booster dose over calendar time and by time since administration. We conducted a retrospective cohort analysis among 13,083,855 persons aged ≥60 years who were eligible to receive a seasonal booster at the start of the 2022-2023 vaccination campaign in Italy. We estimated rVE against severe COVID-19 (hospitalization or death) of a seasonal booster dose of bivalent (original/Omicron BA.4-5) mRNA vaccines by two-month calendar interval and at different times post-administration. We used multivariable Cox regression models, including vaccination as time-dependent exposure, to estimate adjusted hazard ratios (HR) and rVEs as [(1-HR)X100]. The rVE of a seasonal booster decreased from 64.9% (95% CI: 59.8-69.4) in October-November 2022 to 22.0% (95% CI: 15.4-28.0) in April-May 2023, when the majority of vaccinated persons (67%) had received the booster at least 4-6 months earlier. During the epidemic phase with prevalent circulation of the Omicron BA.5 subvariant, rVE of a seasonal booster received ≤90 days earlier was 83.0% (95% CI: 79.1-86.1), compared to 37.4% (95% CI: 25.5-47.5) during prevalent circulation of the Omicron XBB subvariant. During the XBB epidemic phase, rVE was estimated at 15.8% (95% CI: 9.1-20.1) 181-369 days post-administration of the booster dose. In all the analyses we observed similar trends of rVE between persons aged 60-79 and those ≥80 years, although estimates were somewhat lower for the oldest group. A seasonal booster dose received during the vaccination campaign provided additional protection against severe COVID-19 up to April-May 2023, after which the incidence of severe COVID-19 was much reduced. The results also suggest that the Omicron XBB subvariant might have partly escaped the immunity provided by the seasonal booster targeting the original and Omicron BA.4-5 strains of SARS-CoV-2.