Background: Monitoring adherence to antiretroviral therapy (ART) is critical in the management of the HIV/AIDS epidemic, particularly in high-burden settings like South Africa. Group-based trajectory models (GBTMs) offer a promising approach to identify latent classes of adherence behaviour from longitudinal data.

Materials and methods: We applied GBTM to 12 months of electronic monitoring (EM) device data from a 24-month prospective observational study (ADD-ART) of 250 virally-suppressed, people living with HIV in the Western Cape region of South Africa. We then applied a Kaplan-Meier survival analysis on the sample, after stratifying by adherence profile.

Results: The GBTM analysis revealed five distinct adherence trajectories, as illustrated in Figure 1: 1) stable and excellent adherence (19% of participants), 2) stable and acceptable adherence (26%), 3) stable and poor adherence (17%), 4) slowly deteriorating adherence (18%), and 5) rapidly deteriorating adherence (20%). These subgroups exhibited markedly different viral suppression outcomes, with the two deteriorating adherence groups exhibiting a much faster decline in the probability of viral suppression over time.

Conclusions: This analysis approach provides a valuable tool for identifying heterogeneous patterns of longitudinal ART adherence that have important implications for clinical management and viral outcomes. This approach could be extended to other adherence monitoring methods and settings to better target adherence interventions.