Controlled adherence with long-acting injectable cabotegravir plus rilpivirine: determinants and clinical impact of injection-schedule adherence in the Spanish RELATIVITY cohort

BACKGROUND: Long-acting cabotegravir plus rilpivirine (CAB+RPV LA) transforms adherence from daily self-administered oral therapy to scheduled healthcare system-supported injections. The real-world determinants and consequences of injection schedule adherence remain incompletely defined. METHODS: We conducted a nationwide observational cohort study within the RELATIVITY program, combining retrospective data collection for participants already receiving CAB+RPV LA with prospective follow-up of new initiations that included virologically suppressed adults who initiated CAB+RPV LA in routine care in Spain. Injection-schedule adherence was calculated using a proportion-of-days-covered framework adapted to long-acting injectable ART, based on the proportion of evaluable injections administered within a predefined ± 7-day target window, and categorized as complete (100% of evaluable injections within the target window), high (90-99.9%), or suboptimal (<90%; including early and/or late injections outside the target window). Univariable associations with noncomplete injection-schedule adherence were assessed, whereas adjusted determinants of late injection outcomes were evaluated using multivariable models. The time to the first late injection and the time to definitive discontinuation were analyzed using Kaplan-Meier methods and Cox regression. A separate multivariable logistic regression model evaluated the occurrence of two or more late injections. An exploratory cross-validated prediction model was developed to estimate the risk of noncomplete injection-schedule adherence. RESULTS: Among 3349 participants, 3280 had evaluable injection-schedule data. Complete injection-schedule adherence was observed in 2309/3280 (70.4%), high adherence in 948/3280 (28.9%), and suboptimal adherence in 23/3280 (0.7%). At least one late injection occurred in 1254 participants (38.2%), whereas two or more late injections occurred in 436 (13.3%). Adherence remained high across age, sex, nationality, comorbidity burden, substance use, previous virological failure, oral lead-in, and baseline CD4+/CD8+ strata. In univariable analyses, baseline characteristics showed limited associations with noncomplete adherence. In adjusted models of late-injection outcomes, center volume, prior ART duration, foreign nationality, BMI, and comorbidity burden showed small-to-moderate associations. Virological failure remained infrequent across adherence categories. The exploratory prediction model showed limited discrimination (cross-validated AUC, 0.53), indicating that baseline variables alone were insufficient for clinically useful prediction. CONCLUSION: In this large real-world cohort, CAB+RPV LA achieved very high injection-schedule adherence and rare suboptimal adherence. Baseline characteristics had limited ability to predict schedule deviations, supporting CAB+RPV LA as a controlled-adherence strategy, the success of which depends on reliable delivery systems as much as patient selection.