Long-term HIV dynamic models incorporating drug adherence and resistance to treatment for prediction of virological responses

Long-term therapy with antiretroviral (ARV) agents in HIV-infected patients often results in failure to suppress the viral load. Imperfect adherence and drug susceptibility to prescribed antiviral drugs are important factors explaining the resurgence of virus. A better understanding of the factors responsible for the virological failure is critical for the development of new treatment strategies. In this paper, we develop a mechanism-based reparameterized differential equation model for characterizing long-term viral dynamics with ARV therapy. In this model we directly incorporate drug susceptibility and drug adherence (measured by medication event monitoring system (MEMS) and questionnaires) into a function of treatment efficacy. A Bayesian nonlinear mixed-effects modeling approach is investigated for estimating dynamic parameters by fitting the model to viral load data from an AIDS clinical trial. The effects of drug adherence interaction with drug resistance-based models are compared using (i) the sum of the squared residual (SSR) from individual subjects and (ii) the deviance information criterion (DIC), a Bayesian version of the classical deviance for model assessment, designed from complex hierarchical model settings. The results indicate that the drug adherence combined with confounding factor, drug resistance in viral dynamic modeling significantly predict virologic responses. Our study suggests that long-term reparameterized dynamic models are powerful and effective in establishing a relationship of antiviral responses with drug adherence and susceptibility.