Background <p>Long-term persistence of spike-specific memory B cells (MBC) upon mRNA SARS-CoV-2 vaccination remains poorly explored in people living with HIV (PLWH). We previously observed that the primary two-dose immunization elicited a B cell response quantitatively similar but phenotypically different compared to healthy participants (HC), with higher prevalence of CD27^-IgD^- double-negative cells and lower proportion of CD27⁺ Ig-switched MBC.</p> Methods <p>The PatoVac_COV longitudinal prospective single-centre study included 74 PLWH vaccinated with mRNA SARS-CoV-2 vaccines. Building upon our previous findings, blood samples were collected before and after the third and fourth doses, and two years after the first dose. PBMC were analysed for the RBD-specific MBC response by multidimensional flow cytometry, while spike-specific IgG were tested in plasma samples by ELISA and by surrogate virus neutralization assay. A supervised decision tree model was used to identify clinical and demographic predictors of RBD-specific MBC persistence.</p> Results <p>The third vaccine dose robustly recalled RBD-specific MBC, driving their differentiation into Ig-switched resting MBC, a subset undetectable before boosting and dominant also in HC. The fourth vaccine dose minimally impacted on the MBC response. IgG⁺ resting MBC, recognizing both <i>wild type</i> and Omicron BA.2 strains, were still present two years after vaccination. Clinical and demographic factors associated to the persistence of IgG⁺ resting RBD-specific MBC were time since initiation of current ART, BMI, CD8⁺ cells count, granulocytes count and blood creatinine levels.</p> Conclusions <p>These findings highlight the impact of boosting, particularly the third dose, in enhancing and shaping the RBD-specific MBC in PLWH. Clinical and demographic variables associated with long-term MBC persistence may serve as potential biomarkers for predicting durable vaccine responses in PLWH.</p>