The Importance of Routine Viral Load Monitoring: The Effect of Delayed Regimen Modification Following Virologic Failure of First-Line Antiretroviral Treatment among HIV-Infected Adults in Sub-Saharan Africa
Background: Resource limitations currently preclude use of routine plasma HIV RNA testing in much of Africa. Alternative approaches for detecting treatment failure are known to have poor sensitivity and specificity; however, the impact that lack of viral load testing will have on patient mortality remains unclear.
Specific Aims: We will investigate the importance of routine plasma HIV RNA level monitoring by estimating the effect of delayed regimen modification following virological failure on first-line therapy. We hypothesize that delayed regimen modification will be associated with increased mortality among subjects whose CD4+ T cell counts do not meet WHO criteria for immunologic failure. Such a finding would support the claim that routine viral load testing will improve patient outcomes. We will further investigate the prognostic significance of cumulative exposure to HIV RNA level (viral load Area Under the Curve) during on first-line therapy, with and without controlling for concurrent CD4. This aim is motivated by the hypothesis that exposure to replicating virus results in increased patient mortality via pathways not mediated by changes in CD4+ T cell count.
Design: Our research questions will be addressed using five prospective cohorts in Sub-Saharan Africa where routine viral load testing is available, but in which there is nonetheless substantial variability in the delay that occurs between loss of virological suppression and regimen modification, Study subjects will be patients who initiate first-line NNRTI-based therapy and subsequently develop virological failure.
Statistical Analysis: Time-dependent confounding by indication complicates the analysis of longitudinal antiretroviral treatment patterns. We will address this challenge using marginal structural-based methods, including history-adjusted marginal structural models and direct effect models. Estimation will use inverse probability weights and targeted maximum likelihood.
Duration of Study: We project a submitted manuscript for each of the two specific aims by the end of 24 months.