Project R-15837

Predictive models for treatment failure and mortality for People Living with HIV in Botswana (Research)

Several observational studies indicates that a lower risk of disease progression and mortality in people leaving with HIV (PLWH) initiating ART at higher CD4 cells count compared to individuals with deferred initiation of ART. However, it is not fully elucidated whether PLWH who received ART early can sustain CD4 count increase and viral load decrease overtime. Therefore, it is crucial to develop statistical models to predict the changes of both CD4 count and viral load over time for PLWH on ART. In Botswana, from 2016 onward a "treat all" policy, in which HIV positive individuals begin ART regardless to their CD$ count. We aimed to characterize the evolution of CD4 counts and viral load in PLWH prior and during the "Treat All" era in Botswana. The study will utilize dataset of PLWH who are enrolled in Botswana national ART program. Both longitudinal CD4 cell counts and viral load measurements and other clinical and demographics data will be extracted from Botswana data warehouse. Several modelling approaches will be used to model that data. A semi-parametric mixed effect model will be used to model the CD4 count and viral load with the other explanatory variables available in the data base. To further model the survival status and CD4 count of PLWH, a joint model will be employed, which aims to develop and identify predictive factors of treatment failure and mortality among PLWH on ART in Botswana. Moreover, a bivariate longitudinal model will be developed for the CD4 change and viral load prior and during "Treat All" era, which is essential to predict treatment failure and/or mortality in the early stages of the treatment.

19 June 2025 - 30 September 2029