Data-driven Modeling of Long-term CD4 Cell Recovery Trajectories Under Modern Antiretroviral Therapy in People Living With HIV

Background: Although CD4 recovery has been widely studied, its long-term temporal dynamics and phase-specific characteristics in the modern ART era remain incompletely characterized. Methods: This retrospective cohort study included adults with HIV in Nanjing, China, who initiated ART between 2010 and 2019 and maintained viral suppression. The optimal CD4 recovery trajectory model was identified using piecewise linear mixed-effects models with exhaustive grid search. Cumulative probability curves estimated probabilities of CD4 count recovery to ≥500 and ≥350 cells/μL. Results: 2611 individuals contributing 22 970 person-years and 37 959 observations were analyzed. The best-fitting model identified a 4-phase CD4 recovery trajectory with breakpoints at 0.5, 2.5, and 6 years, characterized by the fastest increase during 0-0.5 years (265.4 cells/μL/year), progressive slowing during 0.5-2.5 and 2.5-6 years, and modest growth beyond 6 years (8.6 cells/μL/year). Cumulative probabilities of reaching both thresholds rose steadily but decelerated markedly after 6 years. When stratified by baseline CD4 counts at ART initiation, compared with the 350-499 subgroup, the