PROJECT SUMMARY Continued access and adherence to antiretroviral therapy (ART) is crucial for pediatric populations living with HIV to prevent AIDS-related comorbidities. However, there are several unique challenges to treatment adherence in pediatric populations, including caregiver factors, and medication palatability and tolerability. Little work has been done in characterizing patterns and factors related to the timing of ART nonadherence since initiation of therapy in children. Additionally, new ART formulations available for adults have been slower to obtain approval for use in pediatric populations due to difficulty in assessing efficacy in trials. Bringing modern causal inference methods to the analyses of pediatric trials through rigorous estimation of per-protocol effects can help overcome some of these difficulties, and help bring analyses of pediatric trials into the modern age. The objectives of the proposal are to (1) characterize the barriers to pediatric dolutegravir combination therapy adherence, and (2) estimate per-protocol effects of treatment failure by 96 weeks in the pediatric HIV trial, taking into consideration the unique barriers to medication adherence in pediatric populations. Data from the ODYSSEY trial, a phase II/II open-label non-inferiority trial assessing dolutegravir-based ART compared to standard of care in an international sample of children aged 2 to 18, will be used to characterize baseline and time varying factors influencing time to protocol deviation, and then estimate per-protocol effects impacting 96-week treatment failure correcting for such protocol deviations through implementation of modern causal inference methods and machine learning. The completion of the proposed aims will provide evidence on characteristics contributing to timing of protocol deviation and estimates of per-protocol effects of treatment failure by 96 weeks corrected for such protocol deviations in the ODYSSEY trial for first or second line dolutegravir use compared to standard of care in a pediatric population. These results will (1) provide substantial information on factors related to timing of eventual ART nonadherence so that appropriate clinical intervention can be developed to prevent nonadherence related to these factors, and (2) provide an example of modern per-protocol analyses in pediatric trials which will serve as guidelines for future per-protocol analyses in pediatric trials. The training plan outlined in this F30 proposal will provide the applicant with necessary skills in modern causal inference methods and clinical medicine to become a productive and successful physician scientist at the intersection of pediatrics and application of modern causal inference methods for pediatric trials.